Download as pdf or txt
Download as pdf or txt
You are on page 1of 234

Perceptron, Inc.

Vector Field Service Manual

Part Number 009-0506


Revision B
March, 2008
PERCEPTRON INFORMATION NOTICE
The information presented herein provides general operating instructions, engineering data and
maintenance procedures for Perceptron equipment. Perceptron assumes no obligation or liability for the
information presented herein or the use of the same, all such information being utilized at the recipient's
own risk.
In no event shall Perceptron be liable for any direct, consequential, incidental or other damages of any kind
resulting from the use of this information, including, but not limited to, loss of profits or revenues, loss of
use of, or damage to, the equipment, or other property damage or injuries to persons.
The availability, presentation and conveyance of such information by Perceptron to non-Perceptron
personnel does not constitute an explicit or implicit endorsement of, promotion of, or liability for such
information. Except that which may be expressly provided for in service manuals or other written
instructions, the recipient is cautioned against unauthorized attempts at maintenance of Perceptron
equipment and related activities and that such actions could result in serious damage to such equipment, the
malfunction of such equipment, injury to persons or damage to property.
In the event that unauthorized maintenance, installation, service, disassembly, adjustments and/or repairs to
such equipment are conducted by anyone other than Perceptron personnel without prior written
authorization from Perceptron, the warranty provided by Perceptron with respect to the equipment shall be
voided, and product performance may be adversely affected.
Perceptron makes no representations or guarantees that improvements, upgrades or changes will not occur.
In the interest of technical progress, the information in this manual is subject to change without notice. The
examples and diagrams in this document are included solely for illustrative purposes and may vary from
those which actually appear on your system.
TriCam, TriLite, VeriStar, DDM, OCF, LASAR, Wheelworks, ScanWorks, IPNet, and Perceptron are
registered trademarks of Perceptron, Inc., and DataCam, Micro-Flash, Visual Fixturing, Sensing the Future,
Optiflex, The Focus on Process, LinearStar, and InterSystems are trademarks of Perceptron, Inc.. The
hardware, software, and processes described herein may be protected under one or more of the following
patents: US 4541721; US 4645348; US 4647208; US 4666303; US 4745469; US 4841460; US
4862598; US 4899218; US 5006721; US 5078496, US75/728,106, EP 0543900; CA 1294427; CA
1304932; EP 280941; EP 306731; DE P3875984.5; DE P3877655.0; Case # 4226-200037 and Case #
4226-200038; and 6,180,939 B1.
Copyright © 2003-2008 by PERCEPTRON, INC.
ALL RIGHTS RESERVED
Printed in the United States of America. No part of this publication may be reproduced, distributed,
stored in a retrieval system, or transmitted in any form or by any means by media now known or
hereafter discovered, including, and without limitation, print, electronic, mechanical, photocopying,
electrocopying, recording or otherwise, as well as the right to display and transmit this work publicly
on-line without prior permission from Perceptron, Inc. This documentation contains proprietary and
confidential technical information. This document and the information in it are for the sole use of the
customer and facility to which it has been assigned. Transfer of this material to unauthorized
personnel or business entities is in violation of Perceptron’s rights under the United States Copyright
Act.
Revision History 009-0506
Date Revision Description
December, 2007 A Initial Release
March, 2008 B Hard Fixture design changed in software.
Added words on the Sphere algorithm. More graphics are
needed to complete.
The new FlexiCam auxiliary lighting is now discussed.
Added Chapter 4. Concerning how to solve ambient light
issues.
Contents
1. Introduction 1
1.1. Purpose of this Manual 1
1.2. Related Documentation 1
1.3. Mathematical Notations Used in this Manual 3

2. Feature Lighting 4
2.1. Understanding the Measurement 4
2.2. The Lighting Toolkit 5
2.3. Special Considerations for Drawn Hole/Slot Lighting 6
2.3.1. Lighting Strategies 9
2.4. Light Intensity With Respect to Distance 26
2.5. FlexiCam (FMS) and MiniCam Lighting 27
2.6. Minimizing the Light Source Distance to the Part 29
2.7. SpotLite Size 30

3. Fundamental Measurement Principles in Perceptron 32


3.1. TriCam Sensor Principles 32
3.1.1. Contour Sensor 32
3.1.2. Surface Sensor 32
3.2. Coordinate Frames used in Vector 33
3.2.1. The Coordinate Frames Overview 33
3.2.2. Coordinate Frames for a Typical Contour Sensor 34
3.2.3. Coordinate Frames for a Typical Surface Sensor 35
3.2.4. Coordinate Frames for a FlexiCam or MiniCam Sensor35
3.2.5. Overview of the Tools that Relate the Coordinate Frames 36
3.3. The Geometric Principles of Perceptron Measurements 37
3.3.1. Introduction to Algorithms and 3-D Measurements 37
3.3.2. Basic Geometry of Perceptron Measurement 40
3.3.3. The Reporting Options 50
3.3.4. Interviewing the Customer 57

4. How to Manage Ambient Light 61


4.1. Minimizing the Effects of Ambient Light in the Sensor 62
4.2. Changing the Overhead Lights 63
4.3. Blocking the light path 64
4.4. Using Sensor Aim to Control Ambient Light 65
4.5. Using Auto-Exposure to Help with Ambient Light 69

5. Setting Up Algorithms 70

009-0506, Rev. B Contents • iii


5.1. Configuring the Reporting for Y-left/Y-right 70
5.2. Defining the Features 72
5.2.1. Feature Vectors 72
5.2.2. Defining the Range Feature 73
5.2.3. Defining the Corner Feature 74
5.2.4. Defining the Edge Feature 76
5.2.5. Defining Extra Feature Vectors for Hole, Slot and Cylinder Features
79
5.2.6. Defining the Cylinder Feature 79
5.2.7. Defining the Stud Feature 82
5.2.8. Defining the Patch Feature 83
5.2.9. Defining the Hole Feature 83
5.2.10. Defining the Slot Feature 85
5.2.11. Defining the Gap-and-Flush Feature 86
5.2.12. Defining the Sphere Feature 87
5.3. Algorithm Setup Basics 87
5.3.1. Exposure Settings for Structured Lighting 87
5.3.2. Multipath Issues 88
5.3.3. Window Settings 90
5.3.4. Laser Filters for Structured Light Images 91
5.3.5. Bend and Line Extents 94
5.3.6. Threshold Settings 95
5.3.7. Perspective Distortion of Triangulation 95
5.4. Algorithm Setup Basics for Hole and Slot images 96
5.4.1. Exposure settings for Feature Images 96
5.4.2. Final Verification of Hole and Slot Measurement 96
5.5. Common Concepts for All Algorithms 97
5.5.1. Re-Teaching an Inspection Point with Certain Changes97
5.5.2. Calibrating to Metal 98
5.5.3. Test Measure Results 98
5.6. The Y-Z Plotting tool in WinSen 98
5.7. The Advanced Analysis 99
5.8. Range Algorithm 100
5.8.1. Search for Laser Line from the Top or Bottom Setting101
5.8.2. Range Algorithm Y-Z Plot results 101
5.9. The 1-D Min Range Algorithm 102
5.10. The 2-D Min Range Algorithm 109
5.11. Corner Algorithm 111
5.11.1. Teaching the Corner 111
5.12. Edge Algorithms 113
5.12.1. Hemmed Edge 117
5.12.2. Vertex Edge 118
5.12.3. Simple Edge 119
5.12.4. MinRange YZ 119
5.12.5. Edge Y-Z Plot results 120
5.13. Floating Point (FP) Edge Algorithm 124
5.13.1. The Report Options for the Floating Point Edge Algorithm 127
5.13.2. Teaching the Floating Point Edge 127
5.13.3. Rules of Programming Floating Point Edge 129
5.13.4. Floating Point Edge Y-Z Plot results 130
5.14. Floating Point (FP) Corner Algorithm 130

iv • Contents 009-0506, Rev. B


5.14.1. Using the Report Options for the FP Corner Algorithm132
5.14.2. Teaching the Floating Point Corner 133
5.14.3. Rules of Programming Floating Point Corner 136
5.14.4. Floating Point Corner Y- Z Plot Results 136
5.15. Cylinder Algorithm 136
5.15.1. Parameters for Cylinder 136
5.15.2. Setup for Standard Cylinder Measurements 137
5.15.3. Threaded Stud Measurements using Cylinder 142
5.16. Hole Algorithm 147
5.16.1. Overview of the Function of the Hole Algorithm 147
5.16.2. Parameters for the Feature Image 148
5.16.3. Teaching the Hole Algorithm 149
5.16.4. Interpreting the Hole Teach Display 154
5.16.5. Interpreting the Hole Show-Algorithm Display 157
5.17. Slot Algorithm 157
5.17.1. Overview of the Function of the Slot Algorithm 158
5.18. Complex Slot Algorithm 162
5.19. Gap and Flush Algorithm 168
5.20. Hole Presence Algorithm 168
5.20.1. Hole Presence Result Computations 173
5.21. Slot Presence Algorithm 174
5.22. Slug Hole Algorithm 176
5.23. Slug Slot Algorithm 178
5.24. Sphere Algorithm 178
5.25. Inspection Point Test Measure 178
5.26. Relationships 180
5.26.1. When to use Relationships 180
5.26.2. Creating Relationships 181
5.26.3. Relationship Limits and Alarms 183
5.26.4. Editing Equations 183

Appendix A. Error Codes 185


A.1 Error Code Listings 185

Appendix B. Mechanical Considerations 189


B.1 Introduction 189
B.2 Automotive Industry Experience 190
B.3 Mechanical Factors 190
B.3.1 Abbe effects 190
B.3.2 Vibration 191
B.3.3 Mathematical Modeling of Vibration 192
B.3.4 Temperature 192
B.4 Common Problems 193
B.4.1 Design Issues 193
B.4.2 Specific Design Recommendations 199
B.4.3 Designing for Special Features 200
B.5 Installation Issues 200
B.6 Mechanical Considerations for FMS 200

Appendix C. Exposure Index Table 203

009-0506, Rev. B Contents • v


Appendix D. Tools to Upgrade an IPNet to Vector 206
D.1 Overview 206
D.1.1 What the conversion tools do 206
D.1.2 What this tool does not do 206
D.1.3 Overview of the IPNet-to-Vector configuration transfer process 207
D.2 Calibration Options 207
D.3 The Step-by-Step Conversion Procedure 208

Glossary of Terms 210

Index 225

vi • Contents 009-0506, Rev. B


1. Introduction

1.1 Purpose of this Manual


This manual is an internal document intended for use by Perceptron Project
Engineers only – not for distribution to customers.
It covers all those aspects of system set-up and commissioning which are
important for minimizing sensor and algorithm errors and optimizing the
measurement results. Following these guidelines should greatly reduce the time
required to get the information the customer needs. With shortened setup times,
the time required to do CMM comparisons is also reduced.

1.2 Related Documentation


Some other manuals the Perceptron Project Engineer should have available in
his/her laptop computer when setting up and commissioning a Perceptron
AutoGauge system include:
• 009-0461 AutoGuide Getting Started Manual
• 009-0450-0203: Analysis and Reports Manual for AutoGauge
• 009-0476 and 009-0392: Visual Fixturing™ for Vector and IPNet
AutoGauge. The 009-0476 discusses configuration and user interface
for Vector. The 009-0392 manual discusses Visual Fixturing strategies,
theory and shim testing and shim test results analysis, but one should
ignore all the IPNet user interface discussion.
• 009-0580: Advanced Analysis Diagnostics manual. This is the same as
the on-line help for this topic. It covers the algorithm, measurement
geometry and sensor aiming strategies.
• 009-0429: PRCPAnalysis Manual, which has a discussion of offset
problems and correction strategies, which are not fixed in the manual
offset input of Vector.
• “Vector tools.xls” provides the tools to highlight and fix any offset
problems.
• 009-0185: Correlation/Comparison Process Between MCMM and
OCMM. This manual does not describe the steps required for the user
interface. For a discussion on the user interface, see 009-0555.

009-0506, Rev. B 009-0506009-0506 Introduction • 1


• 009-0447: C3 Card Manual
• 009-0541: Vector System Calibration Manual (Soon to be released. Use
009-0402 Appendix C until it is. Only the introduction and sections C.1
and C2 apply to current Vector software.)
• 009-0555: Comparison Utility for the Vector Platform.
• 009-0559: Vector Log Retriever Manual
The following manuals are important for FMS:
• 009-0376 FlexiCam User’s Guide
• 009-0401 FMS User’s Manual (IPNet)
• 009-0474 Installation Guide for IPNet AutoGauge FMS and Hybrid
Systems
• 009-0563 AutoGauge FMS System Operation Manual (Vector platform)

2 • Introduction 009-0506, Rev. B 009-0506009-0506


1.3 Mathematical Notations Used in this Manual
There are two equivalent mathematical notations that are used in this manual.
They can be either horizontal or vertical as shown by the following examples
showing the equation for any point on a line. Using P for a point on the line,
and V for the IJK vector, P has separate X, Y and Z values, and similarly V
has separate I, J and K values. Using the equation to describe a line in 3D as
the example, the following are different forms for the same equation. For
any point [X Y Z] on this line, the equation may look like any one of the
following:
[X Y Z] = [PX PY PZ] + k * [VI VJ VK]

⎡X ⎤ ⎡ PX ⎤ ⎡VI ⎤ ⎡ PX + k *V X ⎤
r
⎢ Y ⎥ = P + kV = ⎢ P ⎥ + k * ⎢V ⎥ = ⎢ P + k *V ⎥
⎢ ⎥ ⎢ Y⎥ ⎢ J⎥ ⎢ Y Y ⎥

⎢⎣ Z ⎥⎦ ⎢⎣ PZ ⎥⎦ ⎢⎣VK ⎥⎦ ⎢⎣ PZ + k *VZ ⎥⎦
where the lower case k is any arbitrary scalar value.
All IJK vectors are assumed to be of unit length. This can be verified by the
following calculation:

(Length of V ) = V = V I2 + V J2 + VK2 = 1.00000

009-0506, Rev. B 009-0506009-0506 Introduction • 3


2. Feature Lighting
“A candle in the right location is better
than a stage light in the wrong location.”
- A quote attributed to Bob Dewar, retired
Executive Engineer for Perceptron.

Proper application of lighting is absolutely essential to making the Hole,


Slot or Complex Slot algorithms work reliably and accurately. There are
three suggested lighting strategies: stereo lighting, specular lighting, and
backlighting. A brief description of each follows.

Cautions:
Any other type of lighting than the suggested stereo, specular and
backlighting may cause invalid measurements. Only the easiest of setups
with very lambertian surfaces will tolerate other types of lighting.
Simply achieving repeatability does not guarantee accuracy in all
algorithms. This is even more true of algorithms using flood lighting and
gray-scale image processing.
Make sure that the sensor is positioned for best focus before starting the
lighting process. Lighting an out-of-focus feature will be made more
difficult.
If the cutout option is eliminating more than 180 degrees of a hole, the
resulting accuracy is likely to be very poor. The lighting should be
adjusted to provide at least 180 degrees of edges.

2.1 Understanding the Measurement


Before starting the selection of lighting tools and lighting strategies, one
should get an understanding of the surface to be measured first. The
following questions should have answers before the first decision:
ƒ Is this a drawn/extruded hole or slot feature?
ƒ If front lighting is the selected option, are scratches or other surface
defects expected on the production parts?
ƒ Is backlighting possible? If so, will this provide sufficiently better
results than front lighting to justify the extra work?

4 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


ƒ Is the current part in station representative of the product? If not,
then the next two questions cannot be answered using this part.
ƒ Is the surface more specular or lambertian? (See the glossary for the
definitions of these terms.)
ƒ How flat is the surface near the feature? Will this remain constant
during production?
ƒ Are any ambient light changes expected in production?
The desired outcome is to have very sharp contrast around the edges of the
feature. If, for example, a lighting strategy for a diffuse surface is chosen,
and the surface is specular, the desired contrast is lost. Choosing a specular
lighting strategy is more difficult to set up, and requires more attention, but
will be much more reliable for most of the defects that are commonly found.
Many of the above questions do not have a clear yes or no answer, but it is
important to give the best answers possible before starting.
Finally, if the measurement is participating in a Visual Fixturing result, then
the lighting becomes even more important as the accuracy will have an
influence on all other fixtured results.
Different manufacturing processes will create different surface conditions.
For example, most Chrysler-made parts in the U.S. are more lambertian than
others, and therefore are more easily lit.

2.2 The Lighting Toolkit


There are two types of lighting tools available for fixed sensors, the TriLite
and the SpotLite (Perceptron’s proprietary LED flood lights), which are not
always equivalent. The physical appearance is the same, the fact that they
project the light to the subject area is the same, but there are important
differences to consider before choosing the light type and the way each is
used. The following table is to help with this understanding:

SpotLite SpotLite with


TriLite (515-0108-01 or Diffuser
Light Property (515-0095) 515-0108–02) (515-0108-03)
Round spot with
Projected light sharp edges. See
Scattered Scattered
pattern SpotLite Size table in
this chapter.
More power is Less power required, Less light energy is
consumed, as because light is directed towards
Efficiency
compared to a directed to only the subject area, possibly
SpotLite. spot. requiring more lights

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 5


SpotLite SpotLite with
TriLite (515-0108-01 or Diffuser
Light Property (515-0095) 515-0108–02) (515-0108-03)
5x5mm (this means
that this light cannot
Apparent light
50x50mm be used with specular 50x50mm
source size
surfaces without a
diffuser)
TriLite’s scattered
SpotLite can project
Light projected light may be too dim
at distances over 2 Not recommended.
from long distance at exceptionally long
meters much better.
standoff distances.
Sharp cutoff can be
Close-proximity Scattered light can used to minimize Scattered light can
interference interfere with other interference with interfere with other
between sensors measurements other sensor’s sensors
lighting
LEDs must be out of 5x5mm size means
Best choice, where
Light source used as focus, as in-focus that this cannot be
50x50mm size will
backlight light is uneven. A used for direct
be adequate.
diffuser can help. backlighting.
Light properties for Best choice, but make Long exposure times
backlighting with Long exposure times sure that the spot size can be required in
diffuser or reflector can be required in is adequate for the some configurations,
(Figure 2-15 and some configurations. feature and the part additional lights may
Figure 2-16) variation be required
For backlighting Sharp cutoff means
hinge holes and Scattered light makes that light can be Scattered light makes
suspension holes as this type of directed only at the this type of
shown in Figure backlighting difficult. backing metal for backlighting difficult.
2-20 superior contrast.
The standard SpotLites have two different configurations, each projecting a
different size spot. These spot sizes vary according to the standoff of the
light source and the light configuration. See section 2.7 for more detail.
The SpotLite diffuser kit is designed to replace both the Fresnel lens and the
black apertures that are standard on the two different styles of SpotLite.

2.3 Special Considerations for Drawn Hole/Slot Lighting


Drawn/Extruded holes are commonly used in automotive holes. They are
designed to easily accept a fixturing pin in the process. As such, it is
common for these locations to be measured for Visual Fixturing, or
providing significant information about the assembly process.

6 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


The use of these holes creates a special set of problems for measurement
using vision, specifically, what all other tools use for locating this type of
feature cannot be seen with a set of lighting tools and camera. This is not to
say that it cannot be measured, however, care must be taken to setup a sensor
and lighting to make this possible. Given the importance of making this
measurement accurate, and the impact that a less-than-accurate result would
have, investing the time to get the proper lighting is critical.
First, it is important to understand the anatomy of a drawn hole, as shown in
Figure 2-1. The anatomy of the drawn slot is quite similar.

Figure 2-1. Basic Anatomy of a Drawn Hole (shown as a section cut)

The idea of the drawn hole is that a locating pin is driven into the hole, will
be reasonably centered in the hole. A locating pin that is slightly off center
will use the shape of the draw to reposition the part so that the pin is driven
in, making contact to the flange. The pin will expand the flange slightly, but
not so much that it matches the hole diameter. The bottom of the flange must
be narrower than the expected pin size. The bottom of the flange may not be
exactly centered, only close.

Note: The bottom of the flange is typically not a reliable measurement.


Flange depths are typically not controlled and the narrowest part of the
flange may not be properly centered. Measuring this structure with front
lighting or backlighting typically does not yield valid results.

Mechanical pins contact the flange, and CMMs therefore measure this
feature on the flange for the best possible result. However, a single camera
pointed into this hole cannot see the flange, no matter what lighting is
provided. Accurate measurements that can correlate rely on making the light
go as close to the bottom of this hole as possible, in a manner that is
symmetric around the hole. Therefore, this is the first objective of lighting a
drawn feature.
The second objective is just as important. Since the algorithm processes the
image as deep into this hole as possible, the region in the center of the hole
must be completely black.
Figure 2-2 shows improper lighting of a drawn feature, where two lights
were mounted directly to the sensor with two 835-0071 clamps. The amount
of light seen in the center of the hole is the problem, and for more specular
surfaces, this lighting will fail to properly light the edges for proper contrast.

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 7


Figure 2-2. An improperly lit drawn hole

Figure 2-2 can be made repeatable, and while it is possible to make this
accurate, one must take care to be sure that the process variation does not
create problems for the cutout locations. Contrast changes around the edges
will create accuracy problems. Monitoring the accuracy of the image left-to-
right is recommended. If there are any problems with the left-to-right
accuracy, this will result in poor performance in the Y-axis direction of the
sensor.
Figure 2-3 has substantially improved lighting, where the lights have been
moved further away from the receive optics. This was done by mounting one
light on the front of the sensor as far away from the optics as possible, and
using a ball-tube mount to support the other. However, there are still some
concerns. The size of the light source in the image is small, as illustrated by
the very bright spots on the top and bottom of the hole edges.

Figure 2-3. An improved lighting on the same target with stereo lights

The edges have proper contrast in this image, but there are issues that need to
be considered carefully. If parts that are either more specular overall, or
parts with shiny scratches around these edges, then it is possible the results
will be inaccurate. As these holes may receive locating pins that can scratch
the surface around the all-important edges randomly, this possibility must be
considered. Given enough correct points, the outlier elimination will
properly remove these points.
Figure 2-4 shows the best lighting of all. The light position is the same as
that of Figure 2-3, only diffusers are installed to the lights to increase the size
of the light source. The results are that the top and bottom bright spots
shown in Figure 2-3 are no longer concentrated, however, more important is
the light around the edges are more even. For this hole and this specific
condition, there is no significant difference; it is the tolerance to common
production defects.

8 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


Figure 2-4. Best lighting using stereo lights with diffuser

Installing a diffuser is recommended, but the additional exposure time must


be considered. Also, consider that the light is now scattered, and might
interfere with other sensors.

Lighting with the FlexiCam or MiniCam will frequently show a center in


the image that is not black. As it is the goal of the algorithm to measure as
deep as possible in the hole, any light in the middle will compromise the
results. Consider moving the sensor/light combination in closer as
discussed in section 2.5. If this does not work, consider using a fixed
sensor to measure the hole.

2.3.1 Lighting Strategies


2.3.1.1 Stereo Lighting
Stereo Lighting is the first choice, and the recommended strategy for drawn
or threaded holes, and for parts that have an irregularly surface. This lighting
strategy can be used for surfaces that have sufficient lambertian properties.
Two TriLites or SpotLites must be dedicated to the sensor, and in rare cases,
three or four. The basic stereo lighting consists of one light source on the left
side, the other on the right, as displayed in Figure 2-5. In some cases, the
TriLite floodlights must be mounted with separate hardware away from the
sensor. Standoffs more than 400mm frequently require separate mounting to
support the lights. For drawn holes, and holes with surfaces just behind the
hole, separate supports to spread the lighting is also frequently required.

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 9


Figure 2-5. Stereo Lighting

Figure 2-6 shows one of the recommended options for mounting stereo
lighting. The considerations for using this style of lighting include:
1- This mounting consumes 120-150mm of standoff clearance.
2- There is a minimal amount of additional hardware required to obtain the
light separation.
3- For TriLites and SpotLites with diffusers, the lights being closer to the
part are an advantage. (See section 2.6)
4- The distance to the part should not be significantly different to maintain
the balance of light intensity between the two light sources.

10 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


Figure 2-6. One Option for Stereo Lighting (Large SpotLites shown)

Stereo Lighting of Drawn Holes and Slots:


As discussed in section 2.3, the objective of the lighting must be to get the
deepest part of the hole well lit, while making the center of the hole very
dark, providing maximum contrast. This frequently means that the light
source must be large, and as close to the feature as possible. If a small light
source is used on a specular drawn hole, the algorithm will have difficulty
with the extremes of light intensity around the deeper areas where the
measurement point needs to be. Repeatability is not a good indicator of
accuracy in this situation.

Figure 2-7. Stereo Lighting of a Drawn Hole

Figure 2-7 shows the idealized objective of stereo lighting of a drawn hole.
The lights are set so that they are at substantial angles, therefore providing

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 11


light deep into the hole. At the same time, the light that goes into the hole is
at such an angle that it cannot create backlit areas in the center, which is the
other objective stated in section 2.3. Note that the flange is not lit in a
manner that the sensor can see; however, this measurement will still be a
measurement that can be correlated and is accurate.

Stereo Lighting of Slots:


For slots using stereo lighting, especially drawn slots, it is important to locate
the lights on opposite sides of the narrow direction of the slot as shown in
Figure 2-9, in contrast to the less desirable lighting in Figure 2-8. The
lighting in Figure 2-9 will light up the straight segments of the slot, which are
the most important. This also has the added benefit of minimizing the light
that could come back through the hole. This is especially important for the
center of a drawn slot, where it must be as dark as possible, just as it is
important to have a dark center for a drawn hole. When the slot has other
material just behind the opening, any extra light return from inside the hole
will degrade the measurement quality.
The most important feature of the slot is usually the straight edges. Very few
slots use the rounded ends for measurement or fixturing. Where the slots do
not use the rounded ends, the lighting efforts need to put the appropriate
priority on these important straight segments.

For most slot measurements with either drawn metal or metal behind the slot,
it is generally recommended that the Y-axis of the sensor be aimed such that
it is parallel to the long direction of the slot. This allows one light to be
mounted to the sensor, and therefore use only one ball-tube mount to hold the
other light. The clear exception to this suggested orientation is when the
light interferes with the part transfer, in which case, using two ball-tube
mounting is the recommended configuration.

12 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


Figure 2-8. Less than optimal lighting of a slot

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 13


Figure 2-9. Preferred lighting of a slot

Stereo Lighting of Threaded Holes:


For threaded holes, there is one constant, which is that threads are rarely
different from 60 degrees. This means that the top of the threads are 30°
angles. Therefore, with a sensor aimed straight into the threaded hole, the
ideal lighting is to locate the light sources at 60 degrees away. The threaded
surface is usually machined, meaning that it is very specular.
While it is not required to locate the lights to a 60-degree standard, it is
important to understand that for threaded holes, the lights need to be
substantially angled to work reliably.

2.3.1.2 Specular Lighting


Specular Lighting is the second choice in lighting a part. This type of
lighting uses one, or more SpotLites or TriLites with the sensor, set at an
angle opposite the sensor, as shown in Figure 2-10. If a SpotLite is used, a
diffuser is mandatory.
Specular lighting is using the reflecting quality of the surface. If the surface
is expected to be consistently dull (lambertian), then there is no need for
using this lighting strategy. The reason that this is a second choice of

14 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


lighting is that it accommodates tremendous variation in the surface
conditions. Scratches or other surface changes will not affect the
measurement quality. Scratches near a fixturing hole are common, where the
pin misses the intended target and the part slides over the pin to seat. This
creates a dark region around the scratch where the light is positioned for a
diffuse return.
The size of the light source is critical, and a diffuser or additional light
sources may be required, particularly when using long standoff sensors with
highly specular surfaces, or irregular surfaces. Moving the light source as
close to the hole as possible can be a useful technique for making the
apparent size larger. Before starting the installation of this strategy, the size
of the light source should be reviewed carefully. Note that SpotLites with the
Fresnel lens have a very small light source size. Therefore, moving these in
closer has little effect other than increasing the intensity of light.

Figure 2-10. Specular Lighting

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 15


Figure 2-11. Specular lighting using a mirror

Figure 2-11 shows the extreme case of specular lighting, using a mirror.
Note how the size and location of the light source is now very important. If
the light source is the same distance as the sensor, the effective area that it
can light is about ½ of the size of the source. With TriLites or a SpotLites
with a diffuser installed, the 50x50mm source becomes a 25x25 reflection.
Moving the light source to half this distance of the standoff of the sensor now
makes the reflected light source size appear 2/3 the size, or 33x33mm.

16 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


Figure 2-11 also makes it obvious that the location of the light source is
critical for using specular lighting. For a flat surface, the mirror positioned in
this figure is an easy means of adjusting and validating the light sources
position. Some of the considerations for this technique include:
ƒ If the surface is not flat, or has large variation, this must be
accommodated by even a larger area, or moving the mirror
appropriately to test the coverage.
ƒ If the single light source is not large enough, multiple lights or a
diffuser must be considered.
ƒ If the exposure is too high with the mirror, temporarily reduce
the exposure. If there is not enough dynamic range to examine
this image, then there is a lot of diffuse content in the original
surface. This suggests that a diffuse lighting strategy might be a
better choice. Dropping exposures by more than 3-4 stops is a
good indication that the surface is likely to be diffuse enough. A
neutral-density filter may be needed to reduce the amount of
light.

Note: This is not usually recommended for drawn holes or slots, threaded
holes, or holes with very irregular surfaces.

Figure 2-12. Example of a proper specular lighting

The above figure shows a good example of specular lighting on a machined


surface. Note that the image shows sharp focus, and the surface has even
lighting. Very important is the 180 degrees of usable arc, as indicated by the
red dots. Anything less than 180 degrees of measured points around the
edges means that the lighting quality still needs more work. The geometry of
this part with the equal-but-opposite lighting made the lower half of the
image unusable. This part would have done better with an aim that was
straight on, as in the stereo lighting, but using three lights instead of two.
Circumstances made the specular lighting approach the only choice.

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 17


The use of the mirror test shown in Figure 2-11 was mandatory to achieve
accurate results.

2.3.1.3 Using Diffuse Lighting


Other types of front lighting (not the stereo or specular lighting described
above) only work with surfaces that are very lambertian (diffuse, dull, not
shiny, flat, etc.). If any part of the surface around the hole is shiny, then this
will look dark to the sensor, and can cause problems if these dark patches are
too numerous for the algorithm to work properly. Shiny areas may result
from scratches, certain types of galvanizing, or irregular surfaces. Although
most holes can use diffuse lighting, be aware of the problems, and choose
another lighting technique if there are concerns about the image quality.
A problem with this type of front lighting is how the edges of the hole may
be lit. A punched hole, for example, may be lit in such a manner that the
visible edge will create a football (American football, not soccer football!)
shape in the image. It is best to eliminate as much as possible and use a
cutout to eliminate all of the offending area.

Figure 2-13. An image of a hole viewed from 50° viewing angle

Figure 2-13 is a common image for FlexiCam sensors, where the light and
the camera are in a single unit. The edge points found on the far edge of
Figure 2-13 could be found at either the top or the bottom, depending on
surface condition. Top or bottom points are frequently random for different
parts. Therefore, a cutout must be configured to eliminate the incorrect
points on the far edge.
Using the “Angle between Sensor-Case Z and Surface Normal”, the
difference of the correct and incorrect points could be as much as the sine of
this angle multiplied by the metal thickness. At the 50° shown, this would be
0.77 times the metal thickness. For metal that is only 0.7mm thick, this
would be sin(50°) × 0.7 ≅ 0.5mm , which is more error than most users can
accept. In this instance, a 180° cutout would be recommended to eliminate
this inaccuracy. A 180° cutout leaves a few points at each end that could be
wrong, but these points are insignificant or will automatically exclude these
points with the “Outlier Elimination” option.

18 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


The more of the far edge that is dark, the smaller the cutout needs to be.
Consider this if it is possible.

Figure 2-13 is a good candidate for backlighting, and this is discussed in


the next section.

Off-axis measurements of holes or slots are sometimes the result of


measuring the side of the metal to be used, such as a clevis. The resulting
inaccuracy may exceed the variation of the thickness of the metal.
Measuring the outside of metal should be considered carefully when the
images will be better with more normal sensor aiming and more lighting
control..

Hinge holes, whether located on the body-in-white or the closure panel,


require this type of lighting. Many hinge holes have irregular surfaces that
are quite specular. For any surfaces that are specular and irregular, the size
of the light source is critical. Consider using a SpotLite with a
reflector/diffuser or a TriLite.
The lighting on slots uses the same rules as holes, with the exception that the
emphasis must be placed on the straight sides since this is the most important
part of the feature. Figure 2-14 shows this, and the edges that must not be
used. Just as the hole, this is a typical image of a slot for an off axis
FlexiCam sensor.

Figure 2-14. A front lit slot with an off axis sensor and lighting

As with the holes, try to get more on axis, moving the light to darken the far
edges or consider the backlit options. The image in Figure 2-14 must use the
Complex Slot algorithm with the far curved and straight edges marked as
“Not Used”.

2.3.1.4 Backlighting
Backlighting involves shining the TriLite or SpotLite at a diffuse surface
(Figure 2-15), which will reflect light onto the back of the hole, relative to

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 19


the sensor. Painting a surface gloss white is suggested. The glossy surface is
recommended, as it is much easier to clean.
Deep-machined holes and slots can be backlit successfully, where the light
bounces off the sides of the hole. Using this approach, the lighting of holes
as deep as 100mm have been very successful. This requires a diffuser for a
SpotLite, given the small apparent size of the light source when viewed
directly. Similar to the deep-machined hole, a pierced hole with light
reflecting off the edges will work the best. If the walls of a hole are not lit,
then the image is no longer round, and cutouts must be used for the round
portion. Slot measurements need similar logic applied.

Figure 2-15. Backlighting a hole with a diffuse reflector

An alternative is to place the light source and a large diffuser under the hole
shining towards the sensor, as in Figure 2-16. Spartech Polycast 2447 or
2067 are some of the recommended materials for the diffuser. The cautions
for this include:
ƒ This is an acrylic plastic product, which means that it must be
cleaned carefully. Polycarbonate plastics are available, and are more
durable, but still should be cleaned carefully.
ƒ Mounting this into position requires some design effort.
ƒ The density of the diffuser must be carefully considered. Too dense
yields too little signal, and not dense enough creates an uneven
distribution of backlight.
Glass may be used instead, but is more difficult to mount safely.

20 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


When using backlighting, a sensor that is off axis from the hole may have too
much or too little light from the edges of a hole. Moving the light or the
sensor is necessary. Ideally, an equal amount of light should come from the
edge as well as the center. Failing this, cutouts may be used, however, be
careful to preserve more than half the circle for measurement. For the slot,
the light must be set for the important straight segments, and the curved
edges are usually not important.
Using backlighting for drawn holes, drawn slots and threaded holes rarely
work properly. The problem is the depth of these holes. What is seen is the
very bottom of the hole, not the top, and sometimes reflected light from the
sides of the holes. Backlighting for these types of holes is not recommended.

Figure 2-16. Alternative to backlighting a hole using a diffuser

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 21


When using backlighting, a sensor that is off axis from the hole may have too
much or too little light from the edges of a hole. Moving the light or the
sensor is necessary. Ideally, an equal amount of light should come from the
edge as well as the center. Failing this, cutouts may be used, however, be
careful to preserve more than half the circle for measurement.
Taking the same example that is shown in Figure 2-13, simple backlighting
this hole would give us the image in Figure 2-17.

Figure 2-17. The image of a backlit hole viewed from 50° viewing angle

The edges on the far side of the hole in this image should never be used.
They should always be cut using the hole, and never use the complex slot to
measure any hole, including this example.
A significant improvement over the image in Figure 2-17 would be to light
the far side edge as shown in Figure 2-18. It is possible in some situations to
adjust the position of a single light to give us this kind of image. Adding a
second light to light the edge is another idea to get this significantly
improved image.

Figure 2-18. The same backlit hole example with the far side edge lit

22 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


Figure 2-19. A backlit slot

Using backlighting for the slot is the same as the hole. If it is not possible to
light any of the far side edges, one should not use these edges. In Figure
2-19, only the red and green areas are valid. For many slots, the distortion of
the metal is very likely to make the far side edges part to part variation too
high. Using the edges on the far side has risk due to this variation. As with
the hole, positioning the light or adding a second light to view the far side
edge is helpful, especially the straight segment on the far side of the slot.
For an extruded (drawn) hole, other types of lighting will case significant
problems as a direct result of the part-to-part variation of the appearance of
the edge. This will yield better results than what is shown in Figure 2-17 and
should be done whenever practical. Using two cutouts is a typical
requirement for this image.
For the slot, the light must be set for the important straight segments, and the
curved edges are usually not the important feature.
Similar to the equal-but-opposite lighting, the size of the light source can be
validated with the part removed from the station. The apparent light source
without the part should cover the entire hole, with the expected variation.
Some hinge holes and suspension can be backlit with a SpotLite, particularly
the hole closer to the sensor. The SpotLite is recommended for this
application because of the fact that it has a very sharp cut-off. This means
that light can be aimed to light the background, without lighting the upper
surface.

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 23


Figure 2-20. Using the sharp cut-off of the SpotLite for backlighting a hole

2.3.1.5 Equal-but-Opposite Lighting for Drawn Holes


Equal-but-opposite is configured exactly like specular lighting, but the
reasons for this configuration are substantially different with drawn holes or
drawn slots. Specular lighting is primarily for flat, specular surfaces, such as
machined surfaces or even some types of galvanized. This discussion will
focus primarily on the drawn hole or drawn slot. The concept is to locate the
light source at nearly an equal but opposite angles so that what is lit and what
is seen is well balanced, as shown in Figure 2-21 below.

24 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


Figure 2-21. Equal-but-opposite lighting for a drawn hole

The angles in the Figure 2-21 are exaggerated, only to highlight the concept.
Angles much more normal to the surface are recommended. The angles must
not be so close to normal that the hole center is not perfectly black.
As the sensor cannot see the gray area because it is not lit, and the green area
because it is not in the sensor’s line of site, this means that measurement can
be correlated strongly to the desired flange measurement.
Either stereo lighting or the equal-but-opposite lighting is recommended for
drawn holes or slots. Other lighting strategies will compromise the results,
and it is only a question of degree. It is important to get as close to one of
these two concepts as possible.

Figure 2-22. Improperly lit drawn hole

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 25


Figure 2-22 shows a hole that is improperly lit, which can only be measured
with a cutout larger than 180 degrees. The bottom part of this image is the
bottom of the flange, and the top part, while a better location to do the
measurement, has some shadowing and will create poor accuracy.

Figure 2-23. A drawn hole with equal-but-opposite lighting

Figure 2-23 shows a better image of the same hole as Figure 2-16. Using
equal-but-opposite lighting, more than 180 degrees of the edge on the top is
now usable. Note the contrast at the left and right edges of the hole.
Teaching this still requires a cutout, but as shown in Figure 2-24, the result is
adequate.

Figure 2-24. A taught drawn hole using equal-but-opposite lighting.

The above images are shown with an extremely large draw radius,
specifically to illustrate the complexities. Typical draw radii are not nearly
this large, and therefore somewhat easier to handle than these. For smaller
draw radii, it is quite reasonable to select the sharp-edge hole option.

2.4 Light Intensity With Respect to Distance


For a basic introduction to the concepts, assume that the sensor standoff and
the light standoff are the same. If we then vary the distance of these
standoffs, the amount of light that seen by the sensor is the issue reviewed.
If we use a diffuse source such as the TriLite or a SpotLite with a diffuser,
the amount of light from this diffuse source that actually makes it to the part
surface will be inversely proportional to the square of the distance. If we are
using the diffuse (lambertian) return from the part surface, the amount of
light returned from the surface is also an inverse square. If the sensor and
light combination is moved together, this turns into being inversely
proportional to the 4th power of the distance. See section 2.5.

26 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


A SpotLite with no diffuser has a Fresnel lens that directs the light, and
therefore the amount of light reaching the surface is approximately inversely
proportional to the square of the spot size. (Note that this is a very rough
approximation.) See section 2.7 for computing the SpotLite projected size
with respect to the standoff of the light.
If we think about a hole that has material behind the hole that we wish to
measure, moving the light closer to the part significantly increases the
contrast of the desired hole and decreases the light return from the surface
behind the hole.
The real world does not follow these concepts exactly, but is close enough to
be quite important. These principles must guide the lighting decisions, and
will make the difference between success and failure. This is particularly
important to understand when using the FlexiCam and MiniCam sensors.

2.5 FlexiCam (FMS) and MiniCam Lighting


The following is an overview of the rules for using FlexiCam and MiniCam
sensors, particularly for FMS applications:
1) Use the auxiliary lighting that is now available for the FlexiCam.
When this lighting is mounted on the robot end effector, the size of the
sensor and lighting combination does increase. This may make it
difficult to put the sensor into tight locations.
2) Specify the S100 sensors over the S200 sensors wherever the
application permits or use the S200 with the auxiliary lighting. Using
the S200 without the optional auxiliary lighting frequently fails with
holes and slots with metal behind, smaller features or drawn holes and
slots.
3) For the FlexiCam sensor, consider compromising focus for the
improved contrast achieved by pushing the sensor/light combination
closer to the part. This is particularly important for those holes and
slots with objects close behind the hole such as backing metal.
4) The sensor should be aimed as normal as possible to all holes. If it is
not possible, then a 180° cutout or larger is very likely to be required.
5) Whenever possible, orient the longer direction of light 90 degrees to
the long direction of the slot
FMS systems typically use only the lighting found on the FlexiCam or
MiniCam sensor. This “one-size-fits-all” concept will work in most cases,
given the very close proximity of the light source and the sensor (100m) to
the part surface. For the few systems with the 200mm standoff sensor, this
extra standoff may create more lighting challenges.
For a few situations where this lighting scheme is not adequate, consider
adding fixed lighting. Typically, these lights may be left on. The only
reason to turn them off is interference with other measurements in very close
proximity to this light.
It is quite rare that adding another light to the robot end-effector will be
possible. The additional light has the obvious disadvantages of increasing
the size and load. The increase in size is the most severe problem in most

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 27


applications, frequently making some measurements in tight areas
impossible, and increasing the probability of wrecks.
Do not attempt to use a TriLite or SpotLite with a FlexiCam or MiniCam
sensor. The wavelength of the FlexiCam sensor is 630 nm (red), where these
two lights are specifically designed for the 675 nm (a deeper red) wavelength
of the TriCam sensors. The band-pass filter is designed to eliminate all other
light other than the specific light that is appropriate for that sensor. This
band-pass filter is sufficient to defeat these 675 nm light sources. A light
source requires sufficient light near the 630 nm wavelength to be effective.
For the same reason, a FlexiCam SpotLite will not work with a TriCam.
A broad-spectrum light such as an incandescent or fluorescent lamps with
sufficient intensity can work.
An important consideration with FMS lighting is the fact that the light source
and the receive optic is part of the same unit. If we again assume the light
returns is a diffuse return, the amount of light seen by the sensor will also be
approximately an inverse 4th power function, being a square function for the
amount of light making it to the surface, and a square function for the amount
of light returned. This is the reason for the recommendation to use the
100mm standoff sensors instead of the 200mm standoff.
Holes and slots with backing metal may work better with the sensor pushed
closer to the part, which will enhance the contrast at some loss of focus. It is
worth some time to experiment with the contrast and focus compromise with
FMS to realize the best working distance. For an S100, pushing it closer to
80mm instead of the 100mm standoff can make a very significant difference
in the image contrast with a small loss of focus. Given the extra standoff of
the S200, locating the sensor around 175 mm for a similar effect. However,
the S200 may need the auxiliary lighting mounted if this 175mm distance
still does not yield a workable image from the built-in lighting.
Another consideration is the shape of the light source of FlexiCam and
MiniCam sensors, which is a 50x40 mm rectangle. Aligning the 50 mm
perpendicular to the slot will also improve the contrast of the slot,
particularly on the critically important straight edges.
FlexiCam lighting of Punched Holes or Slots:
Punched holes have thickness. Whenever the sensor/light combination is
moved away from normal, a 180° cutout is then the only possible way to
make the measurement. If the off-axis condition is moved further from
normal, a cutout greater than 180° may be needed, and the resulting
measurement uncertainty increases.
Similarly with slots, you can only use two of the four sides. Using both
straight sides where one of the sides shows the metal thickness may be
possible, but be aware that as the punch wears, the appearance of this
thickness can change. It is possible to see a slow drift until the a new, sharp
punch is installed, creating an apparent mean change that is not real.
FlexiCam lighting of Drawn (Extruded) Holes or Slots:
As with punched holes, whenever the sensor/light combination is moved
away from normal, a 180° cutout for the far side is then the only possible

28 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


way to make the measurement. The far side measurement goes deep into the
hole, where the near side is quite shallow. If the off-axis condition is moved
further from normal, the sharp edge at the bottom of the hole may be seen,
which is not useable. Cutouts larger than 180° are never recommended for
drawn holes. Changes in the appearance of the metal or the size of the draw
radius will increase the measurement uncertainty with off-axis aiming.
Similar problems exist for drawn slots. Aims that are not close enough to
normal may require the Complex Slot algorithm option. With the sensor
rotated around the short axis of the slot means that the curved end that is
furthest away from the sensor cannot be used. This is better than being
rotated around the other direction given that only one side may be usable.
The straight sides of the slot are the most important part of the feature to
measure accurately.
If the aim is such that one of the two straight sides is not useable, this can be
made to work using only the straight side closest to the sensor. This is just
like the hole, where changes in the appearance of the metal or the size of the
draw radius will increase the uncertainty of the measurement.
FlexiCam lighting of Threaded Holes:
The FlexiCam lighting frequently is not adequate for lighting threaded holes
as the light is too close to the receiving optics. What this means is that there
are two likely results:
- There is no contrast between the hole edges and the center. This
is common for holes with backing metal close to the hole.
- All of the threads are lit, top to bottom.
If the threaded hole is deep and all of the threads are lit, a large variation of
the position of the hole will create invalid results. For holes that are not so
deep, the sensor must be positioned exactly at the center.
The best that can be done with the existing FlexiCam lighting with threads is
to use the close-in lighting technique described in Section 2.4. Take care to
validate the measurements. If the lighting cannot work for your threaded
hole, or any other feature, advising the customer that the measurement cannot
work is better than just waiting for a failed CMM comparison result.

2.6 Minimizing the Light Source Distance to the Part


For some specular surfaces, the apparent size of the light source is quite
important. This is the reason that professional photographers use light
sources that are quite large. Many features that Perceptron sensors are
required to measure are quite specular, and require large light sources.
Remembering that what is important to hole measurements is the contrast
around the edges, if there are specular regions that can appear near these
regions, these edges will appear quite bright if there is a light source at the
mirror angle, but if no such light source is there, it will be quite dark.
With the standard SpotLite not using the diffuser kit, the location of the light
source does not matter. The apparent light source size using one of the
Fresnel lenses is too small, and therefore only in situations where the spot

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 29


size is important does the distance to the part have any impact on the results.
Therefore, the remainder of this discussion will concentrate on the SpotLite
with the diffuser, and the TriLite.
Listing the reasons why these light sources should be kept as close to the part
as possible, there are:
1) The apparent light source size increases, which is important if there
is any specular return from the surface.
2) The contrast between the desired surface and undesired surface
typically improves significantly, as in a backed hole example.
3) The contrast between the diffuse and specular returns from the
surface is reduced.
4) For TriLites and SpotLites with diffusers, the amount of light making
it to the part surface reduces the required exposure times.
5) The amount of light making it to the surface helps overcome any
ambient light interference.
If proper lighting cannot be achieved with moving the light sources closer to
the part, the alternative is to place more light sources around the feature.
This obviously adds cost and adds hardware, but is frequently the only
alternative.

2.7 SpotLite Size


The size of the light must minimally cover the hole or slot plus the process
variation, making sure that the appropriate area is covered. This is also
important to understand for balancing stereo lighting, as the amount of
energy from each light source should be balanced. The light intensity will be
an inverse of the square of the diameter of the spot. Computing the size of
the projected spot may be done with the following formulae for the two
different SpotLites:
Small spot size = 0.057 * Standoff + 42.5
Large spot size = 0.25 * Standoff + 43.0
These formulae create the following table:

Standoff Standard Wide-angle


Distance SpotLite Size SpotLite Size
in mm (515-0108-01) (515-0108-02)
100 48 68
200 54 93
300 60 118
400 65 143
500 71 168
600 77 193
700 82 218
800 88 243
900 94 268

30 • Feature Lighting 009-0506, Rev. B 009-0506009-0506


Standoff Standard Wide-angle
Distance SpotLite Size SpotLite Size
in mm (515-0108-01) (515-0108-02)
1000 99 293
1100 105 318
1200 111 343
1300 117 368
1400 122 393
1500 128 418
1600 134 443
1700 139 468
1800 145 493
1900 151 518
2000 156 543
A SpotLite size computer is also built into the PRCPAnalysis tool, and may
be used without any IPNet configuration xml.
The amount of light output from a SpotLite is always constant. Therefore, a
spot size that has twice the diameter also has ¼ the intensity. This is because
the intensity of light is inversely proportional to the area of the spot, and
therefore inversely proportional to the square of the diameter of the spot.
⎛ Diameter 2 ⎞
⎜⎜ A = Πr 2 = Π * ⎟⎟
⎝ 4 ⎠
Consider the ratio of intensity of a large SpotLite to a small SpotLite at a
standoff of 1000mm. This ratio would be:
2
⎛ 1 ⎞
⎜ ⎟
⎝ 293 ⎠
2
= 0.11
⎛ 1 ⎞
⎜ ⎟
⎝ 99 ⎠
The large spot is expected to produce be approximately one ninth the amount
of light at the surface as compared to the small spot. Therefore, the exposure
time must be about 9 times as long to gather the same amount of light. Also
consider the effect of ambient light, which is 9 times as significant.

009-0506, Rev. B 009-0506009-0506 Feature Lighting • 31


3. Fundamental Measurement
Principles in Perceptron

3.1 TriCam Sensor Principles

3.1.1 Contour Sensor

Figure 3-1. The Structure of the Contour Sensor

Figure 3-1 shows a side view of the contour sensor, with the in-focus region drawn. The
architecture of the sensor allows the full laser plane to be in focus, and at the same time, the
optics can be designed for greater light collection, therefore reducing exposure times
required. Designed specifically for the 2-D sectioning and 1-D range measurements, its
design is optimized for this purpose. This is in contrast to a surface sensor, where there are
compromises in the mechanical and optical design for the purpose of 3-D measurements.
One of the common problems with structured lighting is multipath. The contour sensor
design helps suppress the multipath, which is explained in more detail in section 5.3.2.

3.1.2 Surface Sensor


Digital TriCam Surface sensors are capable of measuring all feature types, but are
specifically configured to measure holes and slots using two images. A surface sensor can
use just the structured laser light for sectioning features exactly the same way that a

32 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


Contour sensor can, but with certain restrictions. No other lighting or images are required
for these algorithms.
For holes and slots, the SpotLite or TriLite (Perceptron's proprietary LED flood lights)
floods the surface with light, allowing the TriCam to locate the feature. This image is the
first image, and locates the feature in the X- and Y-axes of the sensor case coordinates.
Next, the TriCam sensor projects a laser line of light and triangulates on it to compute the
range Z value from the second image. Using both images, the X, Y, Z center location of
the feature is then computed of a hole, slot, or rectangle.

Figure 3-2. The Surface Sensor Overview

Figure 3-2 shows the overview of the surface sensor. Note that the in-focus region is a
smaller region of the field of view. There are no specifications for what this region is, only
because it is dependant on the feature. For example, a large hole with sharp-edges can
tolerate being more out of focus than a small hole or a drawn hole. Typically, the in-focus
distances allow a ratio of ⅓ close, and ⅔ far for the same degree of focus. This ⅓- ⅔ ratio
is the well-known rule of thumb in photography.
The depth of focus is a design compromise, where the larger the depth of field is a tradeoff
against how much light can be accepted. Photographers know that a fast lens also means a
shallow depth of focus, and a slow lens will mean higher exposures, but more depth. This
is an important understanding to know where and how this sensor may be used.
The above design tradeoffs result in the surface sensor being a sensor that is more for 3-D
measurement and not a universal solution for all situations. Although it can measure all
types of features, it cannot perform 1-D and 2-D measurements as effectively as the
contour sensor. In addition, the orientation of the laser plane and optics allow for better
part clearance for the types of features that it typically measures.
The FlexiCam and MiniCam sensors are also surface sensors. This means that they will
also have a depth of focus limitation.

3.2 Coordinate Frames used in Vector

3.2.1 The Coordinate Frames Overview


The following are some the definition of the coordinate frames used in AutoGauge. This is
not a complete list, and does not include any robot coordinate frames.

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 33


Part coordinates: This is the coordinate system in which all body coordinate reporting is
presented. This is also the coordinate frame that the CAD XYZ points and feature data are
defined. Other names for this coordinate frame include body coordinates, world
coordinates, user coordinates and CAD coordinates. All deviations are reported in this
coordinate frame. Many North American manufacturers refer to this as body coordinates.
Visual Fixturing Coordinates: This coordinate frame is only slightly different from the
nominal Part coordinates as it represents how far out of position the actual part is from the
nominal location. This is computed for every part measured as the fixture for each part is
expected to be different. The Visual Fixturing transform defines the Relationship between
the nominal and the actual part position.
Sensor Coordinates: This coordinate frame is the coordinate frame the sensor uses for its
measurement. The origin of this coordinate frame is in the laser plane and the X-axis is
defined normal to the laser plane. The origin is nominally the center of the field of view,
but is different due to sensor build variation. Therefore all points in the x = 0 plane in the
laser plane. The positive Z-axis points away from the front face of the sensor. For contour
sensors, the positive X-axis points towards the receive optic, and for most surface sensors,
the positive X-axis also points towards the receive optic. The Y-axis is the right-hand rule
definition using both the X- and Z-axis definition.
Sensor-Case Coordinates: This may be referred to as simply “case coordinates” in this
text. The Sensor-Case Coordinates are defined such that the orientation of each axis is
either parallel or perpendicular to the surfaces on the outside structure of the sensor. The
orientation of this coordinate frame may be seen on the label of the each sensor. The origin
of this frame is approximately at the field-of-view center. The positive Z-axis always
points away from the sensor and is normal to the front of the sensor.

The Relationship between the sensor coordinates and the sensor-case coordinates is
defined when the sensor is calibrated at the time it is manufactured. The Sensor-to-Case
transform is the result of this calibration.

3.2.2 Coordinate Frames for a Typical Contour Sensor


The following figure shows the typical coordinate frames for the contour sensor. Note that
the center of the field of view, the origin of the sensor case coordinates and the origin of the
sensor coordinates are slightly different. Nominally, they are the same.

Figure 3-3. Typical Contour Sensor Coordinate Frames

34 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


The Y-axis in the above figure points out of the page for both coordinate frames so that the
coordinates are right-handed.

3.2.3 Coordinate Frames for a Typical Surface Sensor


The following figure shows the typical coordinate frames for a surface sensor. Just as the
contour sensor, the center of the field of view, the origin of the sensor case coordinates and
the sensor coordinates are slightly different due to variations in the sensor build.

Figure 3-4. Typical Surface Sensor Coordinate Frames

In the above drawing, the Y-axis of the sensor coordinates is out of the page, and the Y-
axis of the case coordinates is into the page. The FlexiCam and MiniCam sensors have
different sensor coordinates, where both the X- and Y-axes point in the opposite directions
from the above drawing.

3.2.4 Coordinate Frames for a FlexiCam or MiniCam Sensor


Both the FlexiCam and MiniCam sensors are surface sensors, and have similar depth of
focus limitations as any other surface sensor. The sensor coordinate frame is upside down
when compared to any other surface sensor.

Figure 3-5. The coordinate frame for FlexiCam and MiniCam sensors

In this sensor, all axes point roughly in the same direction, which is quite different from
other surface sensors.

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 35


3.2.5 Overview of the Tools that Relate the Coordinate Frames
Part Angles: This defines the Relationship between the gravity reference and the part
coordinates. These references can also be viewed as being relative to the horizon. The part
angles are defined to establish this Relationship. If this is not properly defined, the sensor
angles computed with ConeStar will not be correct. Any sensor calibrated using either
manual angles or SensorCal will fail without proper part angle definition. The Vector
System Calibration Manual (009-0541) describes how to establish the proper part angles.
Relating sensors to part coordinates:
Each sensor’s position and orientation must be known relative to the body coordinates.
Vector supports several means of performing this calibration, all using properly defined
part angles. See the Vector System Calibration Manual (009-0541) for more details.
Sensor-to-Case transform: Perceptron calibration defines this transform for each sensor
during the calibration process. This result is stored in the Digital TriCam controller.

3.2.5.1 The Options for Determining the Sensor Position


The following are calibration techniques to relate a sensor to the body coordinate frame:
ƒ ConeStar: These two calibration techniques relate the absolute position of the
sensor to the body coordinate frame in six degrees of freedom, locating X, Y, Z
translation and Rx, Ry and Rz rotations. Sensor angles are computed and
displayed if the correct part angles are entered.
These techniques are described in the Vector System Calibration Manual (009-
0541).
ƒ SensorCal: This calibration technique uses a laser tracker or theodolite system to
discover the sensor orientation with respect to the body coordinate system. It is
capable of measuring Rx, Ry and Rz accurately, but cannot locate the translation of
the sensor in X, Y and Z coordinates. See “PRCPAnalysis Manual”, part number
009-0429, or “Using TriCam™ Case Measurement to Define Sensor Orientation”,
part number 009-0388.
ƒ Manual Sensor Angles: These measurements define how a given sensor’s case
coordinates relate to the gravity coordinate system. This type of calibration is
generally not recommended over ConeStar or SensorCal because it is only accurate
to around 1-2 degrees, using the approved tools.

Caution: Entering manual angles for a sensor that has been position calibrated
with ConeStar will overwrite the calibration determined by the position calibration. As
the original ConeStar information is kept, this calibration can be recovered.

See the Vector System Calibration Manual (009-0541) for a detailed description of how to
use the different sensor position calibration methods.

36 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


3.3 The Geometric Principles of Perceptron Measurements

3.3.1 Introduction to Algorithms and 3-D Measurements


Different measurements have different geometry. The geometry can be categorized as 1-D,
2-D or 3-D. As a 1-D result or feature can have all X, Y, and Z components, it may require
all 3 axes be enabled for reporting for proper understanding. Any loss of a significant axis
that should be reporting creates the truncated results that are the source of much confusion.
It is extremely important to understand which geometry applies to which algorithm to
understand the offsets and reported results. For a 1-D algorithm, an offset should lie in the
line of measurement, equivalent to a single CMM hit. The results of a 2-D sectioning
algorithm should always lie in a plane of measurement. Edges and corners are 2-D
algorithms that require geometric construction for the final result.
All results and offsets should be contained in the expected geometry. If, for example, a 1-
D offset is outside of the expected measurement plane as the result of a CMM comparison,
one must ask how that happened. Reasons include:
- One of the two machines is measuring the wrong feature.
- The results of one or both machines have been truncated (see the glossary).
- The geometric construction used on the two machines is not consistent with each
other.
When describing a CMM comparison/correlation problem to a customer where an offset
lies outside the expected geometry, the geometry must be considered in the root-cause
analysis. This understanding includes the geometric constructions used on both machines.

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 37


The following table identifies the geometry of each of the Vector algorithms:

Feature Feature IPNet/Vector


Geometry Types Algorithms name changes
Range Range “1-D MinRange” was the
1-D IPNet “MinRange Flat”
Patch 1-D Min Range
Corner (S-Corner) “Cylinder” was the IPNet
“Arc Center”
All Edges, Vertex, Hemmed,
Corner
Simple and Floating Point
2-D Edge
Cylinder, including stud “2-D Min Range” was the
(Sectioning) Cylinder IPNet “Min Range”
2-D Min Range
Gap and Flush
Gap and Flush
MinRange YZ
Hole and Slug Hole
Hole
Slot, Slug Slot and Complex
3-D Slot
Slot
Stud

Non- Hole Hole Presence


measuring Slot Slot Presence

Color Color Master Exposure


Figure 3-6 Overview of feature types and the algorithm

The above table will be important for understanding the discussion of offset problems and
correction options that follow.
For a simple example of an algorithm with 1-D geometry, a Range feature result that is
NTF (normal-to-feature) is expected to result in the value in the described [I J K] line.
Therefore, a proper result should be of the follow form:
∆X = k * I ⎡∆X ⎤ ⎡I ⎤
∆Y = k * J , or ⎢ ∆Y ⎥ = k * ⎢ J ⎥
⎢ ⎥ ⎢ ⎥
∆Z = k * K ⎢⎣ ∆Z ⎥⎦ ⎢⎣ K ⎥⎦
where [I J K] is the defined feature (normal-to-surface) vector, k is some value that makes
the [∆X ∆Y ∆Z] the appropriate reported deviation. If a reported deviation [∆X ∆Y ∆Z]
does not follow this form (i.e. does not lie on the [I J K] line), this is not normally
expected. Some of the reasons for not lying on the expected measurement line include
truncated measurements. Configuration problems can also result in measurements that do
not fall on the expected measurement line. See the glossary for the definitions of truncated
deviations as this is one common configuration problem.

38 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


A 2-D sectioning measurement is one that lies in a plane, which means that the features can
be reduced to a line. For example, a corner measurement result locates two planes, whose
intersection is a line.
Measuring 1-D and 2-D must be performed inside of the sensor’s laser plane, which rarely
comes close to the requested measurement. The Vector software will convert these
measurements into something that the users can understand. The following sections
describe how each of these measurements are performed, and provide an overview of the
geometric constructions that do these conversions.
Measuring a 3-D feature such as a hole or slot requires more than a single laser line. The
gray-scale image of this feature is combined with the laser image.

3.3.1.1 Hard Fixturing points


A hard fixturing point is a concept that defines a mechanical fixture point. For example, a
body-in-white could be located in the up/down direction with pads, but free to move about
in the in/out, fore/aft or rotation about the z-axis. In this instance, one could combine the
sensor measurements with the hard-fixture points that define the up/down location of the
part. Such points could be 1-D, 2-D or 3-D. Most common are the 1-D locators which are
rest pads. For this body-in-white example, there would be a minimum of 1-D locators that
could be defined.
The following table is a review of the typical hard-tool fixturing references. It is not
complete, and only a list of the more common part-locating tools for general reference.

Feature Positioning tool Number directions


locating the feature controlled and type
Surface Rest pad, with opposing clamp 1 NTM
Edge Surface, where part is pushed to 1 NTM or NTF*
“V” groove (Used mostly for rough
Edge 2 NTF
locating and not fixturing)
Edge or
Two rest-pad surfaces 2 NTM
Corner
Hole Pin with flat mating surface 3**
Hole Pin without flat mating surface 2 body coordinate***
Slot Pin with flat mating surface 2 body coordinate***
Slot Diamond pin with flat mating surface 3**
*
Edges are sometimes pushed into flat surfaces that are neither aligned to a NTM nor body
coordinate direction.
**
3-D features that locate in 3 directions do the same thing, regardless of reporting
configuration.
***
NTM holes and slots exist, but are rare, found in a few checking fixtures. Most holes
are aligned to the body coordinate frame.

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 39


3.3.1.2 Anchor points
Anchor points are used as a tool for Visual Fixturing analysis, limits and alarms. See the
Visual Fixturing on the Vector Platform manual (part number 009-0476).

3.3.2 Basic Geometry of Perceptron Measurement


3.3.2.1 The Angles of 1-D Measurements
The one-dimensional algorithms (Range, and 1-D Min Range, measurements only reporting
a single plane) all are measuring against a line. In the CMM, the results of a 1-D
measurement are reduced to a single hit with the CMM probe. Given this geometry, there
are two questions that should be answered, which are:
1) How well is the sensor aimed to the CAD measurement line? This is equivalent to
what is the best possible actual measurement line using the current sensor aim.
2) How close is the actual measurement line to the CAD measurement line?
Items that influence the interpretation of the geometry evaluation of the measurement
include:
1) How flat is the surface around the measurement point?
2) How much expected change in surface orientation is expected during production?
3) How well does the IJK represent the actual surface?
4) If the image is of a corner, have we taught to the correct side of the corner?
The 3-D views of the following were created from using the file Measurement-to-Feature
Angle - somewhat bad 20071018.3D, which should be on your computer. If this file is not
on your computer, run the current version of “Get Manuals *******.bat” found in
\\storage\Automotive\PManagement\manuals when connected to Perceptron’s VPN. The
SpinFire reader tool that is available at www.actify.com/v2/products/SFReader/index.htm
will display this on most computers.
The first concept is the sensor-to-feature angle. This is the angle between the sensor’s laser
plane and the IJK specified for the feature. The ideal aim of 0° would be such that the
normal-to-surface IJK matches the sensor’s laser plane. The larger angle can make the
measurement results less accurate, but angles up to about 45 degrees are acceptable for
most features. Smaller sensor-to-feature angles can be required if the feature is not flat
around the nominal point. Larger angles can be acceptable on very flat surfaces that match
the nominal IJK defined for the feature.
Figure 3-7 and Figure 3-8 show two different perspectives of an example configuration of a
sensor-to-feature of approximately 15 degrees. The blue arrow is the feature IJK and is
normal to the blue plane. The blue plane represents the nominal feature. The red is the
plane of laser light. The gray arc shows the sensor-to-feature angle between the laser plane
and the nominal IJK vector.

40 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


Figure 3-7. Perspective of CAD surface and sensor laser plane

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 41


Figure 3-8. Cross-section view normal to both laser and CAD surface plane

Perceptron software compensates for any angular differences, given the assumptions
described in the measurement-to-feature discussion that will be later. Therefore it is not
necessary to drive these angles to 0°.
This primary purpose of this calculation of the sensor-to-feature angle is done to provide a
cross-check of sensor aim to the defined nominal surface. One must also consider the
flatness of the surface and how well the IJK matches the actual surfaces to understand what
the larger angles mean. A large deviation with a curved surface can be a problem for
accuracy. Therefore, the image quality is the most important consideration and must have
the priority.
Consider now the process of teaching the Range algorithm. This process establishes the
actual line of measurement. When one clicks the point on the line during teaching, uses the
actual surface is used to construct the line of measurement. The following graphics show
examples of a taught Range. The measurement line in the image is shown as the blue line.
All measurement results will be along this blue line.

42 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


Figure 3-9. Teaching the Range algorithm

The blue line in Figure 3-9 constructs a line of measurement that exists in the laser plane.
The blue line does not look normal in the image, only because of the perspective view of
the triangulation sensor. It is constructed to be perpendicular to the actual surface at the
point that is selected and taught. The algorithm constructs all measurement results at the
intersection of this blue line and the imaged white line. Once taught, the measurement line
will never change unless the algorithm is taught to another surface.
Note that the XYZ position (translation) of the entities is not part of these concepts. All of
the plane, line and IJK features are positioned in these figures are slightly skew to a
common point for the figures.
Now looking at the measurements from a view normal to the laser plane, we have Figure
3-10. In other words, Figure 3-10 is a section cut. Such a section cut will show the
nominal plane and the actual plane as lines. For this example, the actual part is not quite
parallel to the nominal part.

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 43


Figure 3-10. Normal-to-laser view of 1-D measurements, which shows why a measurement-to-feature angle will be larger than the
sensor-to-feature angle.

Now, using the same example shown in Figure 3-7 and Figure 3-8, here is what this
construction of the measurement line might look like the following using a real-world
perspective. The red line represents the measurement line established by the teaching
process. It exists in the laser plane and normal to the intersection of the plane and the
actual part, which might look like Figure 3-11 below:

Figure 3-11. A view of a taught Range algorithm in the sensor space

44 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


Figure 3-11 and Figure 3-12 are the same as the figures in the above description for the
sensor-to-feature angle with different perspectives and, and different features added or
removed. Figure 3-12 adds the nominal information omitted in the Figure 3-11. Since
there is a noticeable difference between the nominal and the actual surface, the
measurement-to-feature angle is larger than the sensor-to-feature angle. For these figures,
the sensor-to-feature angle is 15 degrees and the measurement-to-feature is 25 degrees.
The 10 degree increase between these angles would be alarmed as a problem.
Comparing the above figure to the one below, the features added are:
- The red taught measurement line inside the laser plane.
- The sensor-to-feature angle shown as the gray arc.
- The measurement-to-feature shown as the red arc.

Figure 3-12. The resulting measurement-to-feature angle as compared to the sensor-to-feature angle

A noticeably higher measurement-to-feature angle such as these figures show can be the
result of:
- An incorrectly defined nominal IJK. This may be the result of a mistake in the
CAD creation of the IJK or a data-entry error. Also look for features with very
similar looking names. Misreading these names is a common source of incorrect
IJK definitions.
- Incorrect negating the J of the IJK for left/right mirroring of the part..

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 45


- The sensor is aimed at a surface with a different orientation than the blue nominal.
The sensor may be aimed at a different location than what the XYZ and IJK
defines. If the feature is curved or bent, the original taught point could be
improperly selected.
- If one side of the corner is to be measured with a range, the error could be the
result of teaching the wrong side of the corner. Simply moving the taught point to
the correct side of the corner and teaching the correct side will yield a much
smaller measurement-to-feature angle.
- The defined sensor orientation is incorrect. Given a wrong sensor orientation, one
can teach the algorithm to the correct surface, but the measurement line is not
correctly oriented. For manually measured sensor angles, an error in measuring
these angles can cause such a problem. Simply moving a sensor without updating
the sensor-to-part information can result in a large measurement-to-feature angle.
For FMS, there are several calibration steps involved, including the tool-finder
calibration, robot-to-part calibration and the reading of the robot pose for this point.
Therefore, it is quite important to review these two angles with an understanding of the
importance of these values. These angles are displayed on the Advance Analysis tab.

3.3.2.2 The XYZ Translations of 1-D Measurements


For inspection points where the absolute sensor position is known, it is now possible to
compute the measurement distances. Similar to the angle results, there are two results
shown, the sensor-to-feature and the measurement-to-feature distances. Starting with a
cross-section view normal to the laser plane and the plane of the part, we have the
following:

Figure 3-13. Sensor-to-feature distance shown in section view, normal to both CAD and sensor laser plane.

Switching to a perspective drawing normal to the surface, a hypothetical taught


measurement line has been added in the figure below that is somewhat less than optimal.

46 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


Figure 3-14. Less than optimal measurement-to-feature distance

The above figure shows that for the current sensor aiming and teaching, the measurement-
to-feature distance is somewhat greater than the sensor-to-feature distance. During the
teaching process, the operator is asked to select the point in the image where the
measurement is to be taken, which is used to define the taught measurement line. The
Vector then constructs the best normal on the selected point in that image, which sets the
line in which all measurements will be taken.
If the measurement-to-feature distance is considerably larger, obviously, this requires
immediate investigation. One common reason for this is simply entering the sign for the
in/out measurement for the wrong side.
On a truly flat surface that does not alter in orientation, this difference may not be
important and does not need to be optimized. On curved surfaces, surfaces that change in
orientation from one part to the next, or surfaces where the IJK does not represent the
orientation of the plane, even a small distance between the measurement and the feature
could be quite important.
Correcting this could be as simple as moving the taught line to a better location in the
image, assuming that the sensor-to-feature distance is acceptable. If the sensor-to-feature
distance is too large for the measurement, moving the sensor is required to get closer to the
desired measurement.

3.3.2.3 Geometry of 2-D (Sectioning) Measurements


The two-dimensional algorithm results are expressed in a plane, measuring a feature that
can be reduced to a single line intersecting this plane. Using the Corner algorithm, the
following describes the geometry:

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 47


Figure 3-15. Line Construction from Vector input

Taking the cross product of the IJKs for both the left side and the right side yields an IJK
normal to both vectors. The Nominal [XYZ] point and this cross-product IJK result now
describe the line of this feature, which is the intersection of both planes. The Edge
algorithm uses the same construction, however algorithms that don’t have plane
constructions such as the Cylinder only uses a single XYZ and IJK. The Cylinder XYZ
and IJK describe the centerline of the cylinder to be measured, which is already a line
without any construction.
The following figure illustrates the sensor-to-feature angle for a corner feature. The red
vector is the normal to the sensor’s laser plane, the black shows the IJK as described. The
sensor-to-feature angle expresses the difference between these two IJK vectors.

Figure 3-16. Illustration of the sensor-to-feature angle for a corner

For example, if we have a sensor-to-feature angle that is 45 degrees, that means that for
every millimeter of process variation, the measurement point moves along the feature by as
much as a millimeter. If the measurement is for a straight feature with an accurate IJK, this

48 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


may not be a problem. If either the feature is curved, or the IJKs are not accurate, the
reported measurements will not be accurate. The shim test results, CMM comparisons, or
Visual Fixturing might fail as a result.
Where the absolute sensor position has been calibrated, the distance between the laser plane
and the nominal point is shown. The intersection of the defined feature and the laser plane
to the CAD XYZ is shown.
Just as in the 1-D case, we have similar issues when the sensor-to-feature angle is higher
than expected:
• The sensor orientation is not correct. Manually measured and entered sensor angles are
one common source of this error.
• The IJK is not correct for this surface.
For the sensor-to-feature distances, we must consider the following issues:
ƒ How straight is the feature around the measurement point?
ƒ How much expected change in feature orientation is expected during production?
ƒ How well do the IJKs represent the actual feature?
Figure 3-17 illustrates the sensor-to-feature distance for a corner. The cross product of the
two surface-normal IJKs is shown, and this would be similar for an Edge. Cylinders and
Min Range algorithms use just the entered IJK.

Figure 3-17. Illustration of the sensor-to-feature distance for a corner

This distance is not the shortest distance between the point and the plane, but the distance
along the line of the feature.

3.3.2.4 Geometry of 3-D Hole and Slot Measurements


The Hole and Slot algorithms use the IJK surface normal to accommodate the need to
understand the surface orientation. Perceptron sensors use two images for these features,
the gray-scale image and the structured-light image. Neither of these images measures the
orientation of the surface in front of the sensor. The IJK surface normal is used to construct
the final measurement in the absence of a direct measurement of orientation. Therefore, if

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 49


the IJK does not match the surface presented to the sensor, the measurement results will be
incorrect.
The second use of the IJK surface normal is to help with another geometric issue related to
the algorithms. For a sensor that is not oriented normal to the plane of the hole, the circular
hole will show an image of an ellipse in the image. The shape and the orientation of the
ellipse in the image must be understood for the Perceptron algorithms to correctly measure.
Perspective distortions also make squares appear as a parallelogram. The square might
appear as a rectangle or a rhombus from different perspectives. Slots with straight or
curved sides also have requirements for the algorithms to correct for these perspective
distortions.

3.3.3 The Reporting Options


From a perspective of measurement and reporting, 3-D measurements are the simplest. For
body coordinate reporting, the concepts of deviations and offsets are quite clear, unlike
those measuring planes and lines. For 1-D and 2-D measurements, the sensor aiming may
not coincide with the desired measurement results. Therefore, the Vector software will
take the as-measured results for these 1-D and 2-D algorithms, and convert those results
into something that the user expects, not just what the sensor happens to be measuring as
shown in the previous section. With only the geometry used for the actual measurements
as shown in the previous sections, the results cannot be understood by the intended user of
the data. These results must be converted using the CAD data and the reporting
configurations selected during the teach process.
The following sections break down these concepts according to measurement types and the
appropriate geometric constructions. Specifically, the breakdown is 1-D, 2-D and 3-D.
Selecting the wrong reporting mode for measurement could have no impact with some
features and on other features could make it impossible to obtain good CMM comparison
results. Generally, the 1-D and 2-D measurements are the more sensitive to reporting mode
configurations than the 3-D.

3.3.3.1 Reporting modes for 1-D measurements


Referring to Figure 3-18, we see that the measurement is not something that can be
understood in a coordinate frame without some conversion. Say for example we want a
normal-to-feature result, but the sensor is not aimed in such a manner that it can present
this result without some conversion. Just converting the raw sensor deviations to body
coordinates is not enough, since these results are not along the normal-to-feature line.
As there are an infinite number of points in the plane of the actual part, reporting this
deviation can be done from the nominal point to any point on the actual part plane, and be a
valid answer. There are only a few ways to show this nominal-to-actual deviation that can
be understood. Vector software allows the following selections:
• Normal to Feature (NTF)
• X-, Y- or Z-axis approach vector
• NTM
Only NTM is not a reporting mode that relates to the user’s coordinate frame, all others
only report in this coordinate frame. For the approach vector, a CMM will commonly take

50 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


the probe and approach the surface using the IJK normal direction, and the results are
therefore only in one axis.
The NTF body-coordinate reporting is the most common, and is illustrated in the following
figure. This is equivalent to a CMM that uses the normal IJK to make the measurement hit,
and therefore all results are expected along the NTF line.
The sensor measurement is shown as vector A, which is converted into the NTF vector,
also shown. This creates a result that is expressed as an [X Y Z] deviation that is normal
to both the nominal part plane. This deviation is the difference of the actual part and the
nominal part, along the [I J K] surface normal.

Figure 3-18. The Geometry of 1-D NTF

The approach vector mode is also analogous to a CMM measurement. If the CMM
positions the Y- and Z-axis, moving only the X-axis to make the hit, all measurements are
in this X-axis line. This is equivalent to the Vector’s X-axis approach vector mode of
reporting.
The concept of NTM is simply the length of the NTF deviation with an appropriate sign for
the direction. NTM has been traditionally signed positive “as the metal grows”, so in this
case, the NTM result would be − X 2 + Y 2 + Z 2 . Vector signs the NTM result

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 51


according to the direction of the [I J K] surface normal vector, which should point towards
the sensor. Most CAD results will be in this direction, and most CMM vectors will be the
opposite. Simply negating all three components of the [I J K] vector may be required to
get the proper signing of NTM reporting. Using either the positive or the negative direction
of the surface normal vector will not change any other aspect of the Vector reporting,
calibration or compensation.
The following figure shows how the 1-D approach vector results are to be interpreted. The
X approach vector is exactly equivalent to a CMM moving its probe in only the X-axis,
without moving the Y- or Z-axis. Therefore, the expected deviation will always be [X 0
0].

Figure 3-19. The Geometry of 1-D Approach Vector Results

Note that the Vector selection of the approach vector is only labeled as X-axis, Y-axis or Z-
axis.

52 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


Not all approach vector options will be valid. For example, if we have a feature that has an
IJK vector of [0.707 0 -0.707], only the X- and Z-axis approach vectors are valid
selections, however the Y-axis approach vector cannot be used.
Note that for simple features where the IJK has a 1 in any component, these choices are
redundant. If the IJK is a [1 0 0] vector, then the NTF and the X-axis approach vector
results will be the same. The NTM results will be similar to these X-axis results in this
example, with only a possible negative result.

3.3.3.2 Reporting modes for 2-D measurements


2-D measurements reduce into a line, therefore the measurement must be performed in a
plane that sections through this line.
Just as the 1-D has an infinite number of valid measurements, the 2-D measurements also
have an infinite number of correct answers. And just as the 1-D is reducible to a few
answers that Vector offers, the 2-D only has a few options that it allows. These are:
• Normal to Feature (NTF)
• X-, Y-, or Z-axis cutting plane
• NTM
The NTF concept is shown in the following figure, using a corner feature as the example.
The desired results for NTF are in a plane normal to the feature line, but the sensor is
positioned such that it is not in this orientation. Vector adjusts the results such that all XYZ
deviations exist in this plane normal to the feature line.

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 53


Figure 3-20. NTF reporting for 2-D feature.

The NTM results for 2-D features do not relate to the user’s coordinate frame, just as NTM
results for 1-D features do. For a corner example, there will be an NTM result for the left
side plane of the corner, and one for the right side plane. It can be thought of as two 1-D
measurement results. However, a CMM can only emulate the corner results with three hits.
With two single hits, the measurement result could be so wrong as to miss one plane or the
other entirely.

54 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


Figure 3-21. NTM reporting for a Corner measurement

Although NTM results do not relate to a coordinate frame, the XYZ and IJK’s must be
properly setup to report NTM.
Unlike a corner or an edge, an NTM result for a feature without a plane reference does not
make sense. Cylinder and Min Range are two examples of such measurements.
The cutting plane reporting modes are all body coordinate concepts, and relate to how the
feature’s line is sectioned. Typically, NTF is used, which is the equivalent to a CMM
constructing the measurement result operating in the plane normal to the feature line. If the
CMM measures the entire feature and geometric constructions at a fixed Z height, this is
equivalent to the cutting plane concept. All measured deviations would therefore have a
Z=0 results. For Vector, this concept is illustrated in the following figure.

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 55


Figure 3-22. The Z-axis (Up/Down) Cutting plane on an A-pillar measurement

3.3.3.3 Reporting modes for 3-D measurements


As mentioned in the beginning of this chapter, 3-D measurements have no need for
compensation, as the results are locating a point. Therefore, there are no approach vectors
or cutting plane concepts. Theoretically, one could describe an NTM result for holes and
slots whose IJK vector does not align with an axis. The need for an NTM reporting for 3-D
measurements in practice has been extremely rare, and best done with Relationship tools.
The concept for NTM exists for measuring holes and slots, but is rarely used or
recommended. One possible use is to confirm that Visual Fixturing is working properly on
a hole that is not axis-aligned.
The important understandings of 3-D measurements are the geometric constructions used
that yield this single point result. For example, CMMs frequently measure the arcs at each
end for the construction of the slot, even though this can be quite incorrect when the slot
has any distortion. Similarly, using a 3- or 4-point touch on a distorted hole can create
incorrect results.

56 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


3.3.4 Interviewing the Customer
The customer must understand the differences between NTM and XYZ coordinate
reporting. The older IPNet system design required this decision before any teaching, this
selection is still important. It is still quite important that the proper choice is made. The
questions must focus on the 1-D measurements first. In the rare situation that there are 2-D
measurements, these will be second. The 3-D measurements are never reported in NTM.
There are two problems with making incorrect choices. The first is that the reports can be
cluttered with unused numbers. For example, if the NTM reports for certain inspection
points are never used, then there is no point to presenting these values in any report.
The second problem occurs if there’s a competing measurement tool, the chances of
comparing the wrong reported values increases.

3.3.4.1 Understanding the Competing Measurement Tools


It is always a good idea to know about the competing measurement tools, and as stated,
even more important if the measurements are to be compared to one of these tools. You
must examine this device carefully to be very certain what to expect. Here are some
generalizations that are typical:
- Remember that NTM is a metal-grows or shrinks concept. It is also a product-
oriented concept. One would expect NTM reporting to be only found on Class-A
surfaces and features that influence what the customer can see. Subassemblies that
would not use NTM reporting include underbody, any suspension component and
frame systems. Since these subassemblies use only body-coordinate reporting, the
CMM is almost always the competing measurement device.
- CMMs can be programmed to report NTM, but typically report only body
coordinate deviations. The most common and easy NTM report is with the 1-D flat
features, but can also include an edge measured in 1-D. CMMs can report 3-D
features (holes and slots) and cylinder features, but require some very difficult-to-
write code to do these features.
- Using a CMM for gap and flush measurements is rare, because of one or more of
the following reasons:
o The CMM must be programmed to do the required geometric constructions to
produce the NTM results
o The CMM will have a lot of trouble finding the gap with a probe tip
o The CMM typically breaks a lot of probes measuring gaps, especially during
the manual operations that are typically used to find the gap
This should be explained to the customer before he commits a CMM to measuring
gap and flush. Most programs measuring gap and flush have been aborted because
the above points were not completely understood.
- Hand-held devices rarely have any connection to a coordinate frame, and therefore
only show 1-D, or a pair of 1-D NTM measurements. The typical gap-and-flush
(also know as margin-and-flush) measurement tools are examples of hand-held
tools and do not have any connection to a global coordinate frame. The most
common tools for measuring gap-and-flush are step gauges and dial indicators or a
DātaMyte device. There is no connection to a coordinate frame with these tools,
and can only produce NTM results. One of a few examples of a hand-held device

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 57


that can report in a coordinate frame is Leica T-Probe and T-Scan, where the laser
tracker is the link between this tool and a coordinate frame.
- Hard-tooled gauges typically only show 1-D NTM deviations for surface, corner
and edge measurements. Names like ring gauges, Blue Bucks, R1 and F1 gauges
are common in North America. All these names are used for the hard-tooled
gauges that only produce NTM deviations. Dial indicators, feeler gauges, step
gauges and LVDTs are also 1-D NTM tools. See the “Correlation/Comparison
Process Between MCMM and OCMM” manual (part number 009-0185) for a
discussion on how some of these tools are used.
- All 3-D features (holes and slots) should be configured as body coordinate
reporting. The need to report these 3-D features as NTM is extremely rare,
difficult to understand and have no standards. The situation where a hole or slot is
not oriented to match the body-coordinate frame is also quite rare and the only
situation where NTM would be different than the body-coordinate deviations. For
these rare situations where NTM is required, Relationship equations are strongly
recommended for these reasons.
- One common configuration choice for gap-and-flush measurement systems is to
report all 3-D features as XYZ coordinates, and the remaining 1-D and 2-D
features as NTM. This is the simple rule-of-thumb to apply to these systems to
avoid confusion.

Never use the terms NTF or “normal to feature” during the customer interview. These
are not industry standard words, and may confuse CMM experts and other metrology
people. Body-coordinate, part-coordinate or XYZ coordinate deviations are the standard
words that will be understood.

3.3.4.2 Some Information that can help


Many times the questions asked get incorrect answers. Some of this is because the wrong
people are asked, and sometimes it is because the people do not understand enough to
properly answer the questions. If a CMM is the competing measurement tool, you should
pose the questions to the person programming the CMM. Not all CMM programmers even
understand what NTM reporting is, so be prepared to fully explain the concepts.
If there is any reason to doubt the answers, the following information can be collected to
help make an informed decision include:
- The CMM programmer must understand the equations and the programming
required to create the NTM results. If the CMM programmer does not understand
the concepts, equations or how to program the CMM to do this, then one can
assume that the results will be body-coordinate deviations only.
- If a CMM is the competing measurement device, and already programmed to
measure the part, ask if a measurement report is available. Such a report can
provide some very important clues. The NTM 1-D results (i.e. range
measurements) are typically shown as a single value that is not labeled with a
body-coordinate axis name such as “N” or “T”. A common format of such a report
has three rows or columns for X, Y and Z deviations. Many CMM reports show a
4th position for the NTM results. For holes or slots, the 4th column/row may be for
diameter, not NTM and may be labeled as “D” or “diam”. Be very certain of what
is reported in this position and what results and feature types put values into each

58 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


component of this report. Check for the numbers being reasonable in the report.
For example, 1-D NTM deviations should be scaled up by factors in the range of
±1.0 to ±1.7 of the largest body-coordinate deviation. They will never be scaled
down.
- Examine the measurement tool carefully with the understanding of the previous
points. If a non-contact tool is involved, you will need to know something about
the algorithm(s) used. For hand-held devices, look for any connection to a global
coordinate frame. There are a very few hand-held devices that have such a
connection through a laser tracker, a large field-of-view photogrammetry system,
etc.
- If there are any edge or corner measurements reporting NTM, they are typically
reported as two 1-D results. If they are 2-D geometric constructions reporting in a
global coordinate system, then the details of the algorithm need to be understood.
- Be sure to review the signing conventions for NTM reporting. Section 3.3.3
describes the typical conventions used for signing (which direction is positive, and
which is negative.

3.3.4.3 Creating and Installing NTM Offsets


Looking ahead to when the offsets will be generated, you and the customer must agree on a
way to process the offsets. The rules that you need to understand for NTM offsets include:
- Comparing NTM measurements obviously create NTM offsets. Similarly,
comparing XYZ measurements create XYZ offsets. Never enter one form of
offsets into another.
- For 1-D offsets, the most significant component of the X-, Y- or Z-offset in body
coordinates will be a scaled version of the NTM offsets, given proper
measurements. This scale factor can be negative or positive. Only features with
and IJK that have IJKs of [+1 0 0], [0 +1 0] or [0 0 +1] do not have a scale
factor.
- One way to fix the signs on some NTM configurations would be to make negate all
IJK values such that the most significant component is signed such that it matches
the coordinate reporting directions. Vector signs the NTM results by the IJK
direction. There are many details to understand before doing this. Contact
Perceptron for these details.
- It is much easier to convert body-coordinate deviations and offsets to NTM than
the other way around. All non-truncated body-coordinate deviations and offsets
can be converted to NTM with simple dot-product computations. Another way to
convert a non-truncated 1-D offsets or deviations is by using the computation
± ∆X 2 + ∆Y 2 + ∆Z 2 .

- While there are easy ways to convert 1-D NTM offsets to body coordinates, the 2-
D NTM offsets are not so easily converted.
- Remember that Relationships have no offset capability. Therefore, if you use
Relationships to generate NTM reporting, there is no way to easily automatically
put the NTM offset into the IPNet.
Please be aware of the out-of-line and out-of-plane concepts for 1-D and 2-D algorithms.
Since correct measurements cannot generate offsets that are out of line or out of plane, such

009-0506, Rev. B 009-0506009-0506 Fundamental Measurement Principles in Perceptron • 59


a condition is a strong indication of a serious problem. Manually entered offset data and
incorrect conversions/calculations are common reasons for NTM checkpoints to have
invalid/incorrect offsets.

60 • Fundamental Measurement Principles in Perceptron 009-0506, Rev. B 009-0506009-0506


4. How to Manage Ambient Light
Ambient lighting can disrupt either the repeatability or accuracy of a measurement.
Ambient light is not always destructive, but at its worst, it will destroy the signal-to-noise
of the image. The following discussion is intended to provide some alternatives to the
ambient light problems.
It is very rare that ambient light directly hits the feature, and fixing this problem is obvious.
More likely the light that bounces around and is seen by the sensor is the problem. Tracing
ambient light problems is quite a bit like the game of pocket billiards, but the bounces are
three dimensions. If specular (shiny) surfaces are involved the bounces are at equal-but-
opposite angels. If lambertian (dull) surfaces are involved with a very bright light, then the
light gets scattered in many directions instead of following the equal-but-opposite rule. In
one extreme example illustrated in Figure 4-1 where both effects were involved, sunlight
from a skylight hit a light-colored wall and bounced off of the shiny metal. This would
shut down a demo system for half an hour every morning that there were no clouds to
scatter the sunlight hitting the wall. Had this been a real production system, one or more of
the suggestions discussed in this chapter would have been required to be sure that the
system would continue working properly. Tracing the light rays similar to this example in
production systems is required to design a fix for such problems.

009-0506, Rev. B 009-0506009-0506 How to Manage Ambient Light • 61


Figure 4-1. Ambient light with multiple bounces

Painted parts create a bigger problem than bare metal. Black requires a very long exposure
time and is extremely specular. This is the color to choose for exploring any ambient light
problems.

4.1 Minimizing the Effects of Ambient Light in the Sensor


Perceptron sensors work with a very small band of light. For the TriCam, this is
approximately 675nM and for FlexiCam and MiniCam sensors, this is 630nM in the red
spectrum. Each sensor has a very narrow band-pass filter to eliminate the other wave
lengths of light. Ambient light that is violet, blue, yellow or even orange cannot be seen by
the sensor. Therefore it takes a strong ambient light that has the red that will pass through
this filter.
The eye can barely see the difference between the 630 and 675nM wave lengths. Both look
like a deep red, but the sensor that is intended for one wave length of light cannot see the
other. The band-pass filter will cut out so much light that it can barely be seen in the
image, typically below noise thresholds.
Ambient lighting can be from one of the following sources:
- The plant lighting
- A window or overhead skylight
- Indirectly from a wall
- Lighting from another sensor, SpotLite or TriLite
Ambient lighting can be seen from:
- Very strong light directly hitting the part that is diffuse.

62 • How to Manage Ambient Light 009-0506, Rev. B 009-0506009-0506


- Light going past an edge or through a hole or slot.
- Reflected light from the shiny metal around the feature (specular light).
The best and obvious solution is to eliminate as many ambient light problems as possible.
The next sections are dedicated to the techniques that can control of the effects of ambient
lighting problems.
The size of the interfering light source should also be considered. A bulb is a point source,
and a window, skylight or wall can create a very large light source. The typical tube-type
fluorescent light fixture is composed of multiple tubes with a white reflector, which in
combination can be as large as a window. The larger light sources can require a very large
light block, or make it difficult to reposition a sensor to avoid the reflection.
Ambient light will be out of focus in the image. This means that even a point source of
light will be defocused and spread out. Out of focus sometimes means that the light
intensity will be at lower levels for the smaller light sources, which is generally good news
for the algorithm. However, this is not always good enough for all point sources.
Defocused imaging of ambient light does not always help with the large light sources or
diffuse surfaces such as a white wall or a very large light fixture. Maximizing the signal
and minimizing the noise is the goal and just good practice.

With windows or skylights as ambient light sources, do not forget to consider the
different possible weather conditions. Different positions of the sun, snow or bright hazy
days have caused failure modes. Direct or indirect light from light-colored walls have
been problem sources. While not possible to eliminate all situations, taking some time to
consider the possibilities can avoid future problems could easily eliminate a return trip to
find and fix a light problem.

Do not miss any opportunity to examine machine vision systems that have sensors that do
not have the narrow band-pass filters. Sensors that try to measure 2-D or 3-D using
structured laser light are quite sensitive to any ambient light without a filter. The sensors
that do not have a narrow band-pass filter will only work in a very dark environment.

4.2 Changing the Overhead Lights


The lighting in the manufacturing environment is typically fluorescent, mercury vapor or
high-pressure sodium vapor. All of these light sources can have the 630 or 675nM wave
lengths of light, and sometimes cause problems. However, be aware that with mercury
vapor lighting, one can find specific light bulbs that do not have the 630 or 675nM light.
Simply changing out the offending light bulbs with these types of bulbs may solve the
problem.
For example, the GE HR400DX33 is a 400W mercury vapor bulb that is known to be dark
around the 675nM red light used in TriCam sensors. These lights may be dark for the
630nM red in FlexiCam and MiniCam sensors, but that needs to be confirmed.
Simply turning off the lights is an obvious solution. However, if there is a need for the
lights in this area, they will be turned on, and may not be turned off for a long time.

009-0506, Rev. B 009-0506009-0506 How to Manage Ambient Light • 63


4.3 Blocking the light path
Blocking the light at any point in its path is the objective with this technique. For a single-
point light source, this can be as simple as the following diagram:

Figure 4-2. Blocking the red light path with green plastic as a filter

Figure 4-2 shows blocking the reflected path from a corner feature with a green filter. The
important point is the green filter is capable of eliminating the majority of red light.
Shields that block light from windows or skylights require much larger filters given that the
light source is much larger. Also consider the variation that comes with the weather.
Sensors have been known to fail with light haze on a sunny day. Some sensors have only
failed when there is snow on a sunny day.
Spartech Polycast (www.spartech.com/polycast/colored_sheet.html) is a source for colored
acrylic sheets that can be used as filters. Polycast 2414 is a green color and capable of
eliminating nearly 85-90% of the red light. Although Polycast 2111 is a lighter green and
eliminates only about 40% of the red, this may be enough. For extreme situations, Polycast
2092 is a really deep, dark green color, but eliminates almost all of the red light. Similar
specification for weld curtains is not currently available.
A cloth or metal shield can also be used to totally block the light. Gray filters are also an
option; however, a 50% gray filter will pass only 50% of all colors, including the undesired
red. Although not as capable as the green colors in removing the red light, some consider a
station surrounded in gray more attractive than a station covered in a green color.
Yellow filters have been used, and sometimes remove enough light to make enough of a
difference. Never use red filters for removing ambient light problems. Red filters are
specifically designed to pass red light, and are not appropriate for this problem.
A simple test for a filter is to put the filter in front of a sensor while observing the image.
A feature image is the best for this purpose, but either feature or structured-light images
may be used. Darker images will result from the more effective filters. Testing the
materials before spending any installation time is a quick way to evaluate what to expect.

64 • How to Manage Ambient Light 009-0506, Rev. B 009-0506009-0506


4.4 Using Sensor Aim to Control Ambient Light
There are several different possibilities based on the feature and the location of the light
source. This breaks down according to the feature being measured. The assumption for all
of the following is that the ambient light cannot be blocked as described in section 4.3.
Hole or slot
For light behind the hole or slot, the obvious idea is to move the sensor so that it does not
see the offending ambient light through the hole. There are some factors to consider before
moving the sensor to such a position, including:
- For drawn holes and slots, an off-axis position may make the measurement
results worse than if the sensor is on-axis.
- The ambient light or the only available positions of the SpotLite or TriLite may
light up the sides of the hole or slot inappropriately.
- There may be other things behind the hole that are now seen.
For light that is interfering with the surface surrounding the hole or slot, and back-lighting
is used, one should be able to block the light as this light typically comes from behind the
sensor. This situation is quite rare.
For ambient light that is interfering with the structure laser light, see the discussion on
reflection from surfaces.
Ambient reflection from flat surfaces
This problem can be from any of the following features:
- A Range
- Either side of a corner
- The plane of an edge measurement
- The structured light image of a hole or slot
If the sensor is nearly normal to the surface, reflected ambient light should not be a
problem that cannot be controlled by blocking the light. The light coming from behind the
sensor is rather easy to block. Therefore this discussion will assume that the problem is
that the sensor is at a fairly steep angle to the reflecting surface.
There are several different repositioning of the sensor that could be effective. Starting with
rotating the sensor about the sensor case X-axis is shown as a possible solution in Figure
4-3.

009-0506, Rev. B 009-0506009-0506 How to Manage Ambient Light • 65


Figure 4-3. Rotating the sensor about the sensor’s X-axis to solve a problem

By simply rotating the sensor about the x-axis to the position of the red sensor, the
reflection will be suppressed.
Using this same example and viewing it from two different vantage points, we have another
solution shown in Figure 4-4 using a contour sensor:

66 • How to Manage Ambient Light 009-0506, Rev. B 009-0506009-0506


Figure 4-4. Different views of the reflection problem with a contour sensor

Figure 4-4 shows two different views on the same corner measurement and light source.
The bottom view of Figure 4-4 shows two different sensor locations. The one in the sensor
location shown in pink has the undesired reflection, where the same sensor turned to the red
position has no such reflection. Both locations show the back of the sensor in the bottom
view, and the sensor windows are pointed towards the corner shown in gray. The sensor
position in red is where the sensor has been rotated by 180° about the sensor-case Z-axis.
With this, the light source is no longer a problem. The FlexiCam sensor can be relocated to
accomplish the same effect as shown here with a contour sensor.
Ambient with edge features

009-0506, Rev. B 009-0506009-0506 How to Manage Ambient Light • 67


There are two situations that can cause problems with measuring edge features. These are
light from behind the edges and reflections on the edge. The reflected light problem is
shown in Figure 4-5 below.

Figure 4-5. An edge with a reflected light problem

A different problem exists with light behind the edge from the view of the sensor, shown in
Figure 4-6 below

Figure 4-6. Ambient light as seen behind the edge

In both the reflected light and light behind the feature situations, the problem can be solved
by simply turning the sensor around its Z-axis so that the optic is looking downward
instead of up.

68 • How to Manage Ambient Light 009-0506, Rev. B 009-0506009-0506


Figure 4-7. A sensor position that solves both problems

The simple rule for Notice that Figure 4-5 and Figure 4-6 have different problems that can be
minimizing the solved with the aiming shown in Figure 4-7. The simple rule is that if you
ambient light can point the optics down instead of up, the ambient light problem is much
problems for edges. less likely for reflections or looking past the edge. This is particularly
important if there are skylights in the environment where the system is
located.

4.5 Using Auto-Exposure to Help with Ambient Light


Auto-exposure is designed to establish an exposure level for the best image signal.
Ambient light is a noise problem that can either disrupt or destroy the signal and therefore
make the algorithms less accurate. Auto-exposure is not really designed for the purpose of
handling the noise; however, there are a few rare instances where adjusting the exposure
level can help.
Therefore, this should only be tried as all other methods have been tried and failed, but do
not expect this to be a solution.

009-0506, Rev. B 009-0506009-0506 How to Manage Ambient Light • 69


5. Setting Up Algorithms
The following sections describe to the user how to teach and calibrate
AutoGauge/AutoGuide algorithms from the various screens. The ultimate goal is to be able
to configure each inspection point so that it can be properly taught and create the desired
measurement.

Each feature must be defined, which then allows the individual inspection points to be
defined around these features. Each Inspection point must be "taught" to use a particular
algorithm. This data from the feature will be used as parameters for a given measurement
algorithm.

There are some constant items in the algorithms, because all Perceptron dimensional
measurement sensors use light-sectioning triangulation. As the tools and solutions are
common to these images, this section provides an overview of these common items.

Understanding these geometric concepts is important to be able to configure the system


properly. Refer to chapter 3.3 for a detailed description.

5.1 Configuring the Reporting for Y-left/Y-right


This function is now referred to the “Cross-Car” or “Crosscar” function. The Cross-Car
function creates the “fat-car/skinny-car” reporting concept. For most manufacturers, the
y-axis is reporting the in/out dimensions. (The one exception to this is some very old
Nissan plants.) On the left side of the part, the y-axis deviations must be reported in a left-
handed coordinate frame to show the positive-is-outboard, negative-is-inboard convention.
For the typical y-axis cross-car convention, one should do the following steps:
- Make sure that the cell is in “Program Mode”.
- Use a right click to open the properties on the part.
- The “Part Space Properties” page may now be configured to look like the
following figure:

70 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


5-1. The Part-Space Properties page

- Now that this is configured to report cross-car, each inspection point properties
page has the additional “Cross-car” characteristic.
- Navigating to “Tools” → “Rename Characteristics” allows you to change the
Cross-car name. The Characteristics page might look like:

5-2. The Characteristic Names page

This changes the name of Cross-car to “Y”. The Y is now named to “Right-Y” to
clearly show that this selection creates right-handed deviation reporting.
- The inspection points that were formerly reporting Y should now be reporting the
Cross-car option. This should be validated before creating any reports.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 71


5.2 Defining the Features
This section describes the concepts common to all algorithms that perform measurement
functions. Each feature must be created, and require the XYZ points, IJK vectors and any
other nominal data required. These XYZ points and IJK vectors will be used by the system
for the algorithms and correct reporting. If any of these entries are incorrect, this can make
the reported results unusable.

Note: The drawings shown in section 5.2 are theoretical ideal aims. For some
algorithms, the theoretical ideals do not produce good images. For all situations, the
sensor must be to acquire the best measurement possible, which usually means the best
image possible. The tools exist in the software to help you achieve this goal, and the
aiming to achieve the best measurements many be very different than what this section
shows.

5.2.1 Feature Vectors


The features as defined with the nominal data may be used to define the orientation of a
surface or show the direction of a line. For example, the XYZ and IJK define the centerline
of a cylinder. In mathematical terms, a line or a plane can be described as a point and a
vector, so all of these features are defined with an XYZ and IJK. These direction vectors
and body-coordinate XYZ locations are part of the Physical Feature configuration.
The Vector system uses these feature definitions for any or all of the following:
• Visual Fixturing computations and compensations,
• ConeStar calibration and information,
• NTM and body coordinate reporting options,
• Hole and slot algorithm setup and processing,
• Identification of conflicts of the data defining these features,
• Conflicts between the feature definition, sensor aiming and the configured
algorithm.
While the definition of an IJK vector is important, the plus/minus direction is typically not
important. Only in the NTM reporting is the sign of the IJK considered. This plus/minus
direction may be used to yield the proper signing of the reported deviations. If the NTM
reporting requires changing the direction of the IJK, you must negate the I, J and K values.
Negating only one or two components is not legal or correct. Note that negating any
component of a unit vector yields a unit vector, therefore the unit-length test built into
Vector does not help identify such signing problems.
These IJK vectors should always have a unit length (1.0). Most devices that create these
IJKs will create unit length IJKs. If they are not unit length, the IJK may have been
incorrectly entered into the Vector or there may be some error in transferring the numbers.
An IJK vector of unit length may still have an error, most commonly a sign error in one
component of the IJK. The vector length does not change with this kind of mistake.
Some algorithms cannot be taught without the feature data, such as the hole or slot
algorithms. This information is used to interpret the feature (flood-lit) image, and required
to make the tools used for teaching function. This is required to compute the expected
perspective distortion, a critical component for successful measurement results.

72 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


In all situations, proper IJKs are quite important for the functioning of the system. As such,
it is preferred that these vectors are generated by the CAD data, however CMM data can be
useable.

Caution! Feature data extracted from CMM programming are frequently not normal
to some features. While it is true that the majority is normal to the surfaces, the
operations on edges and other small features are frequently incorrect. It may be
intentional or accidental. This creates an incorrectly defined axis move. These incorrect
IJK values should never be used to define features. The quality of these CMM results
should also be reviewed carefully and may not compare well to Perceptron results..

5.2.2 Defining the Range Feature


The Range feature is a 1-D measurement on a relatively flat surface whose results will be
expressed in the 3-D domain for the given part.

Figure 5-3. Range Feature

Figure 5-4.Another example of a Range Feature (1-D Min Range Algorithm)

Given the requirements for a single XYZ point and a single IJK direction vector, the Range
feature parameters are defined in the properties page looks like Figure 5-5 below.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 73


Figure 5-5. Range Feature Parameters

5.2.2.1 Algorithms that measure the Range Features


- Range
- 1-D Min Range
- MinRange YZ

5.2.3 Defining the Corner Feature


The Nominal Point describes the point in the user’s coordinate frame where the nominal
location of the measurement exists. It is entered as an X, Y, Z value and describes the
theoretical intersection point, or vertex, of the two planes (intersection point of the two
lines in the section cut) forming the corner.
Nearly all CMM programs and many CAD results do not give you the theoretical intersect
point. If you only have the two surface points, use the “Corner Point example” worksheet
in the “Vector tools.xls” tool.

74 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


The two Plane-Normal vectors describe the orientation of the planes. These are entered as
i, j, k vectors normal to the planes. The NTM reports will use these direction vectors to
report Corner Plane Normal 1 and Corner Plane Normal 2 respectively. Otherwise, the
vectors can be interchanged without any effects on system calibration or reporting.
The Feature Direction Vector is computed to be normal to both vectors. This cannot be
modified, and is for information purposes.

Figure 5-6. Corner Algorithm

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 75


Figure 5-7.Corner Feature Parameters

5.2.3.1 Algorithms that measure the Corner Features


- S-Corner
- Floating Point Corner

5.2.3.2 Converting Two XYZs to a Theoretical Intersect


For Floating Point Corner, CMMs will have programmed two XYZs, and not the single
XYZ of the theoretical intersect. Some CAD processes also develop two points instead of
one theoretical intersect. Converting these XYZ pair to a theoretical intersect may be
accomplished using “Vector tools.xls”. See the “Corner Point example” worksheet, where
the “Planes Intersect” function is used to develop the theoretical intersect point.

5.2.4 Defining the Edge Feature


The Nominal Point describes the point in body coordinates where the nominal location of
the measurement is referenced. It is entered as an X, Y, Z value and describes the
intersection point or vertex of the line on the plane and edge of the surface.

76 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Nearly all CMM programs and many CAD results do not give you the theoretical intersect
point. If you only have the surface point and the edge point, use the “Corner Point
example” worksheet in the “Vector tools.xls” tool.
The Surface Normal Vector describes the orientation of the plane or surface of the part. It
is entered as an I, J, K vector normal to this plane.
The Edge Normal Vector describes the orientation of the edge or end of the part. It is
entered as an I, J, K vector normal to this edge.
The angle between the Surface Normal and the Edge Normal Vectors are expected to be 90
degrees. If they are not, there may be a problem in the definition.
The Feature Direction Vector is computed to be normal to both the Surface Normal and
Edge Normal Vectors. This cannot be modified, and is for information purposes.

Figure 5-8. Edge Algorithm

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 77


Figure 5-9. Edge Parameters

Note the angle between the Surface-Normal and Edge-Normal vectors is displayed. If this
is not 90 degrees, this is a possible indication of a data-entry or a data-source problem.
Check both vectors carefully if this result is not approximately 90 degrees. For example,
CMM approach vectors sometimes do not represent the feature correctly, and will show an
angle that is not 90 degrees. Therefore the CMM approach vectors must not be used when
one or both IJKs are not normal to the metal.
It is reasonable for a corner to be measured as an edge where one side is not useable for the
corner algorithm. In this situation, the surface and edge vectors will not be 90 degrees.
This angle should match the feature, and visual check is an important validation of the
correct IJK values.

78 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


5.2.4.1 Algorithms that measure the Edge Features
- Vertex Edge
- Hemmed Edge
- Simple Edge
- Floating Point Edge
- MinRange YZ (for glass edges)

5.2.4.2 Converting Two XYZs to a Theoretical Intersect


See section 5.2.3.2. The reasons that this may happen for edges is the same as corners, and
the solution is also the same.

5.2.5 Defining Extra Feature Vectors for Hole, Slot and Cylinder Features
The Surface Normal Vector is always required for all holes and slots. Similarly, the
Feature Direction vector is required for cylinders. One should always ask for these vectors
from the CAD. The Feature Direction for Hole and Slot algorithms and the Primary
Feature Normal vector that is needed by these algorithms are not always available. This
section describes what to do when these vectors are not available. The only situations
where the following rules cannot be used are where NTM reporting is enabled and some
rare holes and slots that are not axis aligned.
For most situations the extra vector definitions are not available. The recommended
procedure for filling in this IJK is to find the smallest absolute I, J or K component of the
Surface Normal Vector and match this with a 1, and zeros in the other two components.
Applying this rule gives us the following examples:
Surface Normal Vector or Suggested Extra Vector
Cylinder Feature Direction
[1.0000 0.0000 0.0000]
[0.0000 1.0000 0.0000] or
[0.0000 0.0000 1.0000]
[0.9936 0.0400 -0.1054] [0.0000 1.0000 0.0000]
[1.0000 0.0000 0.0000]
[0.3452 -0.8727 -0.3452] or
[0.0000 0.0000 1.0000
The Primary and Secondary Normal Vectors are displayed by the feature types that use this
construction. These are computed results using the two entered IJK direction vectors.

5.2.6 Defining the Cylinder Feature


The Cylinder algorithm is designed to measure the position of a cylindrical pin or post, or a
threaded stud. This may also be used to measure a 2-D Min Range feature, which looks
like a small-diameter cylinder, or other, very small feature. Some examples of a 2-D Min
Range features are a hemmed edge where the plane is not viewable, or a small feature
designed to hold body trim are some examples of a 2-D Min Range feature.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 79


For the threaded stud, this requires very specific configuration of a fixed sensor or an FMS
single measurement point. See section 5.15.3 for details. For a scanned stud, this requires
setting up a Stud feature, not a cylinder.
This algorithm only reports deviation in 2-D, parallel to the plane in which the stud is
attached. The plane is not measured with this algorithm as the plane may, or may not be in
the sensors field of measurement. Where the plane can be measured, a Range or a
1-D MinRange algorithm can be used to acquire the last dimension.
The Nominal Point describes the point in body coordinates, where the nominal location of
the measurement is referenced. It is entered as an X, Y, Z coordinate and describes a point
at the center of the cylinder.
The Feature Vector describes the orientation of the cylinder. It is entered as an IJK
direction vector along the centerline axis of the cylinder.
The User Cylinder Normal Vector should be set using the rules described in section 5.2.5.
The Primary Normal and Secondary Normal Vector are computed vectors that are
displayed.

Figure 5-10. Cylinder Algorithm used for cylindrical pins, posts or threaded studs

80 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Figure 5-11. 2-D Min Range Feature

Note that the 2-D feature IJK vector is not a surface-normal vector. The 2-D Min Range is
intended to measure very small features.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 81


Figure 5-12. Cylinder Parameters

The Primary and Secondary Feature Normals are computed values.

5.2.6.1 Algorithms that measure the Cylinder Features


- Cylinder (either the standard cylinder or threaded stud)
- 2-D Min Range

5.2.7 Defining the Stud Feature


This feature is to support stud scanning using an FMS configuration. Scanning a stud
supports measuring a stud in 3-D. Do not select this algorithm for cells that have a fixed
sensor configuration.
To use this algorithm, refer to the manual that supports this feature and algorithm type.

82 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


5.2.8 Defining the Patch Feature
This feature is to support stud scanning using an FMS or hybrid configuration. Do not
select this algorithm for cells that have a fixed sensor configuration.
This measurement is a 1-D measurement.

5.2.9 Defining the Hole Feature


The Hole algorithm measures the position of a circular hole. This may be a punched hole,
drawn/extruded hole or a threaded hole.
The Nominal Point describes the point in body coordinates, where the nominal location of
the measurement is referenced. It is entered as an X, Y, Z coordinate and describes a point
at the center of the cylinder.
The Surface Normal Vector describes the orientation of the cylinder. It is entered as an I,
J, K along the centerline axis of the cylinder.
The Feature Direction Vector should be set using the rules described in section 5.2.5.
The Primary Normal and Secondary Normal Vector are computed vectors.

Figure 5-13. Hole Feature

The math data will be used by the algorithm to describe the perspective the TriCam will
have when viewing this feature. If the camera is not normal to the feature, the hole will
look more like an ellipse. If the Surface Normal Vector is not correct, then the algorithm
may be looking for a different ellipse than the one seen in the image.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 83


Figure 5-14. Hole Parameters

The Surface Normal and Feature Direction entries must be defined. Typically, the
Surface Normal will be derived from the CAD or the CMM that is measuring this feature.
The Feature Direction will typically be a [1 0 0], [0 1 0] or a [0 0 1] IJK. When the
CAD can define this, the Feature Direction vector should point in the long direction of the
slot. When this vector is not available, follow the same rules as the hole for entering this
vector. This one in the IJK position should match a zero in the Surface Normal. Where the
Surface Normal has no zeros, choose the one to match the Surface Normal I, J or K that is
closest to zero.
The Primary and Secondary Feature Normals are computed values.

5.2.9.1 Algorithms that measure the Hole Features


- Hole

84 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


- Hole Presence
- Slug Hole

5.2.10 Defining the Slot Feature


The Slot algorithm measures the position of a slot. This may be a punched hole,
drawn/extruded hole or a threaded hole.
The Nominal Point describes the point in body coordinates, where the nominal location of
the measurement is referenced. It is entered as an X, Y, Z coordinate and describes a point
at the center of the cylinder.
The Surface Normal Vector describes the orientation of the cylinder. It is entered as an I,
J, K along the centerline axis of the cylinder.
The Surface Normal Vector describes the orientation of the cylinder. It is entered as an I,
J, K along the centerline axis of the cylinder.
The Primary Normal and Secondary Normal Vector are computed vectors that are
displayed.

Figure 5-15. Slot Feature

The math data will be used by the algorithm to describe the perspective the TriCam will
have when viewing this feature. If the camera is not normal to the feature, the hole will
look more like an ellipse. If the Surface Normal Vector is not correct, then the algorithm
may be looking for a different ellipse than the one seen in the image.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 85


Figure 5-16. Slot Parameters

5.2.10.1 Algorithms that measure the Slot Features


- Slot
- Complex Slot
- Slot Presence
- Slug Slot

5.2.11 Defining the Gap-and-Flush Feature


The Gap-and-Flush algorithm is the only algorithm that can be assigned to the Gap-and-
Flush feature.

86 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


The Nominal Point describes the point in body coordinates, where the nominal location of
the measurement is referenced.
For more detail, see the AutoFit Getting Started Manual, 009-0483.

5.2.11.1 Algorithms that measure the Gap-and-Flush Features


- Gap and Flush

5.2.12 Defining the Sphere Feature


This feature is to support the measurement of spheres. This is a 3-D measurement intended
for using an FMS configuration. This is a 3-D measurement where sensor X is derived
from the radius of the feature and the radius of the measurement.

5.2.12.1 Algorithms that measure the Sphere Features


- Sphere

5.3 Algorithm Setup Basics


These common items are the result of the sensors using light-sectioning triangulation, also
known as structured lighting. Most of the algorithms that use feature lighting also use the
structured lighting in a separate frame, and therefore use these common tools.
There is a basic overview of the teach controls and functions in Chapter 6 (“Teaching
Inspection Points”) of the AutoGuide Getting Started Manual (009-0461). This should be
read and understood before proceeding in this chapter.

5.3.1 Exposure Settings for Structured Lighting


All light sectioning images should have the exposure set such that the laser line is clearly
visible, and yet not so bright as to overwhelm the important information in the image. This
is achieved when there are just a few pixels in a cross section of the image of the laser line
that are at full brightness (saturation). In most cases, this will provide adequate dynamic
range to tolerate the typical variation of process, and the recommended setting for the
exposure.
The region(s) in which the algorithm functions requires the most attention when setting
exposure. This means that the user must be familiar with what portions of the image are
important to any given algorithm. For example, a corner measurement regions are the two
flat regions between the bend and line length. The middle (bend area) in the corner
measurement is not important and therefore can be ignored without concern, even if it is
totally absent or completely overexposed. Using the same understanding for all corner
measurements, the exposure setting can be a compromise if the two sides of the corner have
substantially different returns.
In extreme over exposure cases, blooming can occur, but this only occurs when the image
is extremely overexposed. Avoiding overexposure is the important consideration, and must
be avoided.
See Appendix C for exposure setting times. This is for reference and cycle time
computations only.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 87


5.3.2 Multipath Issues
In some situations, the multipath return can be a problem for measurements using
structured light triangulation. This problem does not occur frequently, and when it does
occur, has many different techniques to resolve the problems that this could create.
Inside corners are the most common form of measurement that can generate at least two
possible light return paths. The objective is to measure the light that is returned directly
from the laser plane, which is the diffuse return. Unfortunately, the geometry of an inside
corner is such that there is also a specular return path, as shown in the drawing below. If
the surface is specular, the unwanted multipath noise will be stronger than the desired
signal. In the inside corner, the image is also known as a “bow tie” image, for its
appearance.

Figure 5-17. Multipath on an inside corner

There are several recommendations for solving this multipath problem, including:
1- Important! Always use a contour sensor when working with the multipath problem.
The contour having the focus surrounding the laser plane, which is useful as multipath
noise is outside the laser plane, and therefore not in focus. The surface sensor
configuration is such that the multipath noise is at the same focus as the signal,
regardless of how far away it appears in the image. The more out of focus the noise,
the wider it appears in the image, which will make it less intense, and the easier to be
properly filtered.
2- Rotate the sensor about the Y-axis, with the receive optic closer to the part. 10-20° is a
recommended start. This will increase the separation of the signal and the noise in the
image, which allows the algorithm to work better. This decreases the focus and
intensity of the multipath noise.

88 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


3- If it is not possible to adequately rotate the sensor about the Y-axis try rotating the
sensor about the Z-axis. This may require about 45° rotation, and separates and
decreases the focus of the multipath noise.
4- If previous steps cannot guarantee that the noise will never be brighter than the desired
signal, then use the gradient threshold to detect the various lines in the image. The
gradient operator is required for painted parts, and very specular types of galvanized
surfaces.
5- Use the laser filter controls to isolate the good part of the laser line (see Section 5.3.4).
A select group of algorithms also provides the option of leftmost, rightmost and longest
line to assist in filtering the unwanted segments.
6- Use the process window to cut out as much of the multipath noise as possible, with the
usual consideration for process variation.
7- In some situations, blocking the unused part of the laser line may be a useful option.
This may be an option where an inside corner exists next to another feature, such as an
edge. Like windowing, this will remove the unwanted components of the image, and
only the feature to be measured will be lit. Just like windowing, you must understand
what is to be measured, and what the expected process variation will be. This
technique should be used only after the other techniques to control the problem have
failed to provide an adequate safety margin. An edge next to inside corner is one
example where this technique is a possible solution. This solution is not recommended
for a simple inside corner feature.
In the image below, a contour sensor was aimed normal to the feature on a typical
galvanized part with an inside corner. Although the algorithm works and passes
repeatability on this sample part, any change in the surface could create accuracy problems.
A simple scratch under the feature to measure, a process variation of change, or additional
ambient light could cause more measurement uncertainty. Process changes include
changes in the galvanizing, sanding or other surface changes.

Figure 5-18. A Multipath (or “Bow Tie”) Image of an Inside Corner

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 89


Figure 5-19 below shows the same image, same exposure and same sensor. The only
change made was that the sensor was rotated the minimum 10 degrees about its own Y-
axis, exactly as described in step 2 above (10 to 20 is recommended). Note how much the
image improved, showing darker regions around the important features (i.e. improving the
signal to noise), CG measurements as indicated by the smoother red lines and fewer stray
CGs. Any surface change or additional ambient light noise in above image without this Y-
axis rotation is much more likely to cause failed or inaccurate results. The improvement
shown in the image below is such that any further steps are probably unnecessary.

Figure 5-19. Image of an Inside Corner after sensor rotation

Multipath in feature lighting using TriLite or SpotLite sources are possible, but rarely cause
problems. Light that goes into a hole and reflects such that the center is not black is the
important exception.

5.3.3 Window Settings


The window should be set up to encompass at least the measurement regions plus the
expected process variation of the measurement. For example, an edge that has a bend and
line length of 5mm and 5mm respectively, and an expected process variation of ±5mm,
must have a window sized to accept the 20mm process variation plus the necessary 10mm
of part needed to execute the measurement. From a very simplistic view, if the sensor was
aimed normal to the surface, the minimal window must be 30mm wide and 10mm high.
One must add more because the current part's location is not exactly at the expected process
mean, plus some room for expected process changes.
The window is also a very useful tool for eliminating possible problems due to ambient
light or certain types of multipath image noise. With an understanding of the requirements
for a successful measurement, it should be possible to create a compromise such that both
of the above situations. Also, remember that smaller windows mean less work for the
algorithms, which mean faster cycle times.
Windows typically should be set so that they are wider than they are tall. The reason for
this is that with a triangulation sensor, a square space in the measurement field will be a
rectangular shape when viewed from the triangulation angle's perspective.

90 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


For the 20° triangulation angle of a typical structured light sensor, the aspect ratio of the
height to width will be approximately 1:3, and therefore a box that is 1/3 high as wide will
have roughly a square shape in the real world. This 1-to-3 aspect ratio does not apply to
the floodlit feature images of holes and slots, only the laser image.
Processing windows are always presented as red boxes in all algorithms.

5.3.4 Laser Filters for Structured Light Images


The Laser Filter is used for the structured light images. Reviewing the process of
interpreting a laser image, the first step is to find the transition points (TPs). Each TP
identifies the start and end of the bright areas of the image that look like a laser image.
Each bright region of light is then reduced to a single center of gravity (CG) that identifies
the measured result of the center. For a given image, there may be hundreds of CGs.
Now that all of the CGs are found, the next task is how to identify the desired laser line
segment. The purpose of the laser filter is to connect all of the CGs into line segments, and
discard the stray CGs that do not belong to anything. For the typical situation, the laser
filter default parameters will work properly. Changing these settings is rarely required, but
sometimes essential for constructing the right answer.
The next steps are algorithm dependant, but for the typical algorithm and a more complex
image, there may be more than one line in the image. Since the laser filter has determined
how to group the CGs into sets of individual line segments, the remaining task is to select
the one or two desired line segments. For algorithms that have a leftmost, rightmost and
longest option, it will use the results to choose the appropriate line. For holes and slots, the
common situation is that there are two smaller line segments, one on each side of the hole
or slot.
As described, the default settings for the laser filter are such that there is no need for any
further adjustments for the majority of images and algorithms. The reasons for adjusting
laser filters include:
1. Severe multipath situations such as inside corners or features next to inside corners
may require careful laser filter parameter settings to eliminate the undesired effects of
the multipath return.
2. Situations where the laser line might appear broken, because of extremely shiny
surfaces, dark spots, may require that the filter be set wide enough. In this situation,
the laser filter should be set big enough to bridge these gaps when they occur.
3. Situations where the laser line feature to be measured exists near another, separate
areas of bright return in the image could be linked together. This situation needs the
laser filter be set narrow enough to remove the unintended segment from the desired
segment.
4. Unusual features that appear to have two different line segments, such as the example
in section 5.3.4.1 discussing the settings for the Edge algorithm.
5. Some algorithms use the laser filter in slightly different ways. These algorithms are
described in individual subsections at the end of this section.

Note: Most images do have any of the mentioned issues. Therefore, the default
parameters for the laser filter will work properly. The exception is the hole and slot
algorithms, where the laser filter must be tuned to link both halves of the laser line

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 91


together. For other algorithms, and most images, tuning these parameters will not
significantly improve the algorithm results. It is only the rare exceptions for most
algorithms where the default settings need tuning.

For all many algorithms, the laser filter parameters are set graphically with an
Adjustment Box. Positioning this box at different points on the laser line will help
understand what points are expected to connect and what points should not. The location
of the laser filter in the image does not change the behavior for all but the corner and edge
algorithms. The square corner of the adjustment box may used to adjust the width/height of
the laser line filter. The circle on one of the corners may be used to rotate the box to the
slope of the filter.

5.3.4.1 Laser Filter for Edges


The edge algorithms also have a few unique things for laser filters. The column, above-
row and below-row parameters are numeric entries that are in pixel units, and there are no
graphic controls. Whenever these numeric values are changed, the yellow box displayed
shows the size of these changes.
There is also a slope control for the laser filter. This is controlled by the green box that is
used to program the bend and line lengths. Although the yellow box shows all other
parameters, the yellow box does not show the slope. For very rare situations, one must be
careful to consider this slope when evaluating the yellow box.
Figure 5-20, Figure 5-21 and Figure 5-22 show the effects of the laser filter controls. For
the edge algorithm, all CGs that are accepted by the laser filter and the left-most, right-most
or longest option are shown in red. All other CGs are shown in blue.
Figure 5-20 shows the small segment in blue, where these CGs were rejected by the laser
filter. Only the CGs shown in red were ever considered by the algorithm.

Figure 5-20. Two images where the Laser Filter settings disconnected the line segments

The left-side image in Figure 5-20 shows the smaller, right-most segment in blue. In this
example the longest or left-most would yield this result. The right-side image in Figure
5-20 shows the same settings, but this time uses the right-most setting. The algorithm
failed to measure as all CGs inside of the green box were ignored. The smaller red segment
was not considered in the processing of the edge.

92 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Figure 5-21 shows the CGs connected as all are now in red because of the different settings
of the laser filter. Notice the yellow box showing the below-row is bigger than it was in
Figure 5-20. Obviously, this produced a very different measurement.

Figure 5-21. A laser filter setting that connects the two segments

Figure 5-22 shows a similar yellow box as was in Figure 5-20. However, the column width
is larger, and the laser filter slope is now changed with the use of the green box control.
With the slope change, the CGs that are within the tilted green box will connect with the
main line. The slope and width connected the farthest right CG on the left line to the left-
most red CG in the smaller right segment.

Figure 5-22. A laser filter setting where the slope control connects most, but not all of the two segments.

5.3.4.2 Corner Algorithm Laser Filter


The corner algorithms have a unique use for laser filters which makes the user interface
also unique. The typical laser filter concepts do not include the typical left-most,
right-most and longest options found in most other algorithms. It is documented in section
5.11 as it is integral to how the algorithm functions.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 93


5.3.4.3 1-D Min Range Laser Filter
Note that the typical laser filter option for the 1-D Min Range algorithm is uniquely
different for finding the flat surface that is appropriate. The purpose of this algorithm is to
identify the first segment of laser line that is long enough and in the proper orientation to be
the correct segment. The first segment from the top of the image down or the bottom of the
image up is what is found using the laser filter. The left-most, right-most and longest
options are also available.

5.3.4.4 Hole and Slot Laser filters


In order to understand the hole and slot laser filter, you must first understand the sequence
of events that make the algorithm work. This sequence is as follows:
ƒ The gray-scale image for the hole is taken.
ƒ The hole/slot position is found in this image.
ƒ The structured-light image is taken.
ƒ Based on the position of the hole/slot, the extents are positioned over this found
hole/slot.
The CGs used by holes and slots are limited by the octagonal inner and outer extents. Any
CGs that are found outside of these extents are not considered when the laser filter is
applied. For most sensor aims, the remaining CGs in the laser line can be grouped into left
and right segments that must be coupled into one line. The laser filter slope and height
(row tolerance) must be set to correctly link these two segments into a single result.
If the slope or height is not correct, the left segment and the right segment are no longer
considered to be a single line, but two. The side with the most CGs will be used, which can
create a bimodal result in the sensor’s Z-axis direction. In the extreme, any ambient or
stray light may create a segment that might show more CGs than the left or right segments,
in which case the Z-axis result will be random.
Although the U/I allows control of the laser filter width (column tolerance), it has no effect
on the way the laser filter operates. This should be set wide enough to make it obvious if
the slope matches the image or not. The slope and laser filter height are critical settings
that must be correct for accurate results.

5.3.5 Bend and Line Extents


Corners and edges will sometimes use line extents. The purpose of the line extent is to
define the segment of the line that a line fit process will use. The bend length is the
distance from the edge or the corner to where the line fit will start and the line length
defines the amount of the line to use. These measurements are referenced against the
theoretical intersection point of the measured feature.
If you are correlating a measurement against a CMM, it is quite important to set the bend
and line lengths to match the touch points of the CMM. Typically, the two points on a
plane are two hit points, which should be at the bend length and the bend+line length
behind the theoretical intersection point. This is the best match possible with the limited
sampling capabilities of the CMM using a touch probe.
Typical default settings are 5mm bend and 5mm line lengths, but there are many reasons to
change these settings to be either longer or shorter. The reasons for deviating from the
5mm and 5mm default include:

94 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


• The correlating device does something different
• There is not enough metal for these defaults
• There is too much variation of the curvature to the "flat plane" section to be
measured.
Make sure that the bend length is at least the maximum bend length expected on the part.
Measuring too close to the curvature of the bend is quite variable and will increase the
measurement uncertainty. For most hemmed edges, this is 2mm, but the area between the
2mm and 5mm marks are frequently not flat enough for accurate results, which is the
reason for the 5mm default. In corners, if there is not enough metal to make for at least the
bend radius (almost never less than 2mm) of bend length and 3mm of line length, consider
switching to a Vertex Edge algorithm. Using line lengths shorter than the bend length is
not normally recommended.

5.3.6 Threshold Settings


With structured light images, the ability to separate what is part of the laser line and what is
background is quite important. There are typically two options provided, Max Pixel and
Gradient. The Max Pixel technique is recommended for all situations excluding those
applications with severe multipath problems. This technique looks for the brightest part of
the image, constructing the answer from these segments. The gradient operator is designed
to handle the extra image noise that may be in close proximity or where the multipath
return may be brighter than the desired return. As this is not a common problem, the Max
Pixel is the preferred selection. See section 5.3.2 for more details.

5.3.7 Perspective Distortion of Triangulation


Most of the Perceptron sensor product line currently is using 20° triangulation. The most
apparent result of the triangulation perspective makes a square object appear to be a
rectangle or a diamond shape (rhombus) in the image. Similarly, if you put a cylinder in
the image, the result will appear to be a section of an ellipse with a height about 1/3 of the
width. If you put a 90-degree corner into the field of view, the image looks far more like
an obtuse angle than a right angle. If presented to the image such that both sides are at a
45-degree angle to the sensor’s Z-axis, this angle between the straight-line segments would
appear to be approximately 137 degrees apart in the image.
Because of this perspective distortion, a range measurement will show the measurement
approach that will appear to be strange looking without the proper understanding of this
distortion. When the measurement approach line is projected into the correct and square
coordinate frame in sensor Y- and Z-axes, the measurement will be exactly 90 degrees to
the taught surfaces.
Perspective distortion may be understood if you place a 90-degree corner in front of a
sensor, and teach a range algorithm. For this 90-degree feature, the result will be a
measurement line that reflects this distortion. This measurement line is shown in green on
the teach page and in blue on the show algorithm page. The measurement line will be
parallel to the opposite side of the 90-degree corner, which is the correct normal. The
orientation of the 90-degree corner may be changed to many different angles, and as long
as the two laser line segments are visible, you can see the same parallel results.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 95


5.4 Algorithm Setup Basics for Hole and Slot images
Holes and slots require the combination of flood lighting and laser (structured) lighting in
two separate image frames, one for each lighting type. The objective of the structured
lighting is described in the previous section.
Almost all holes and slots are either sharp edged or drawn/extruded. In the sheet metal
applications, most holes are punched, meaning that the holes have very sharp edges.
Drawn/extruded holes have round edges, and are used for mechanical fixturing pins. Both
type of holes and slots may be used to locate sheet metal parts in hard-tooled fixtures.
The first objective of the setup is to maximize the contrast of the hole or slot edges,
regardless of hole type and position. The second objective is to maximize the amount of
the edge available for the algorithm. For holes, this means that more than 180 degrees of
the edge must be clear. For slots, this means that the straight edges are the first priority,
and the rounded ends are second. Sensor aim and lighting are the primary tools for
accomplishing these goals. Make sure that recommendations in Chapter 2 are clearly
understood and followed as carefully as the situation allows.

5.4.1 Exposure settings for Feature Images


As the objective of setting up a feature image such that the contrast on the edges of the hole
or slot is maximized, setting up an exposure is a simple extension of this idea. For a hole
with sharp edges, there may be an additional need to slightly saturate the image to assure a
clean area for the edge search process.
If, for example, we have a front-lit hole with sharp edges, and there is some ambient light
in the hole, this could mean that both the inside of the hole and the outside are not clean.
Slightly saturating the image will create a clean outside, and a strong contrast, making the
edge detection robust and accurate. See Appendix C for exposure setting times.

5.4.2 Final Verification of Hole and Slot Measurement


The Show Algorithm function on the Test Measure page is the most important tool for
verifying the setup. The critical understanding is the meaning of the colored dots on the
feature page. The basic repeatability is not proof that the measurement will succeed.
As the display of the feature algorithm is very brief, you may be required to run this many
times and closely observe the results.

Important note about colored dots in the resulting Show Algorithm


image: It is extremely important to read and understand the following paragraphs
concerning setup for holes and slots.

The process of measuring any feature in a floodlit frame is to first find the feature, and once
found, analyze and refine the measurement for final, accurate measurement. After first
locating the feature, the algorithm proceeds to measure the found edges. The areas
searched for are located between the green dots. The results of the found edges are then
shown in red, yellow or blue, the red being those dots having been found and used, the
yellow or blue having been found and rejected.

96 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Yellow or blue dots, depending on algorithm choice, indicates the specific points that are
not used in the final result, while red indicates those dots that are used. While teaching a
feature, the yellow or blue dots that are clearly measurement blunders that should be
rejected are acceptable. However, yellow or blue dots on a portion that one would expect
to participate in the measurement are not. Repeatability problems may result with these
inappropriately rejected dots. The algorithm accuracy is quite different. Inappropriately
rejected dots will degrade algorithm accuracy. One likely result of these inappropriately
rejected dots is the measurement noise in the CMM comparison.

5.5 Common Concepts for All Algorithms

5.5.1 Re-Teaching an Inspection Point with Certain Changes


Algorithms must be re-taught with change in the feature
IJKs or tool position. The user must find and re-teach all
Inspection Points affected by such changes.
Changes in the process windows or exposure and other minor setup changes do not require
another teach step. This statement applies only to the parameters that will alter the result.
Sensor orientation and IJK parameters are such parameters for certain algorithms. The
teach step in Vector performs calculations that can result in significant changes in the
algorithm results.
Therefore, changes to the tool position or feature IJKs require that all Inspection Points
linked to a particular change must be found and re-taught. Use the sorting function of the
Inspection Points folders to find the points linked to the changed feature or tool position. If
there are any questions about a change requiring another teach step, it is best to just do the
teach step again. The stored taught image(s) may be used to perform the re-teach step, thus
avoiding the need for a part in the station.
The sensor orientation and surface normal are more important than other algorithms. For
hole and slot algorithms, the sensor vector that is displayed on the teach page is the result
of calculations However, if there are changes to either tool position or IJK, the vector
displayed and the calculated results used may not be the same.
For holes and slots, the basic rules to understand are:
- The current sensor orientation and the IJK are combined to create the template
shape of the hole and the cutout results. The effects of the current settings are only
seen by the result of the teach step.
- Any change to the IJK or sensor orientation will immediately change the shape of
the hole templates and the displayed sensor IJK. Other changes may not show any
such changes until the teach step is performed, which shows the details of the
algorithm process.
- For holes where the sensor is not aimed normal to the surface, the cutout structure
displayed may not match the algorithm. You must use the teach step to view the
final adjustments to ending positions of the cutouts. These may not be the same as
before a change to the sensor orientation or IJK.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 97


- Changes as small as one degree can be very significant to the algorithm result. The
smallest of changes requires another teach, even though the results will appear the
same as before.

5.5.2 Calibrating to Metal


The calibration process defines the nominal value, which is used to compare all actual
measurements. The nominal may be set against an actual, known part, or computed using a
calibration technique such as ConeStar.
The process of calibrating to metal is accomplished by taking a single measurement. This
is located in the Tool Position Properties. The “Measure…” button in the “Calibrate using
Case Angles” tab performs the calibrate-to-metal function. As this uses only a single
measurement, be sure that the measurements are repeatable on this part before using this
single measurement. A single measurement of a point with bad repeatability will not give a
zero mean on that part. The “Compute Transform” adjusts the nominal position of the
sensor to affect the calibration to metal result.
Note that if all Inspection Points using this tool position cannot use the calibrate-to-metal
result for this single measurement of a single Inspection Point, you must create more tool
positions to create the different calibration results for the Inspection Points that need the
different calibration.
Use the repeatability on the “Test Measure…” button for the Inspection Point that uses this
tool position. Having good repeatability is not proof of proper algorithm function;
however, bad repeatability or a test/calibrate that fails is proof that the algorithm needs
further setup work.

5.5.3 Test Measure Results


All results shown on this page are in the sensor coordinates, where the X=0 is the laser
plane, and positive Z is away from the sensor. This is shown in Figure 3-3, Figure 3-4 and
Figure 3-5. The statistics from the Test Measure will be X, Y, Z result for 3-D algorithms
or a 0, Y, Z result for 1-D or 2-D algorithms in sensor coordinates. This is the result of the
measurement of the actual part presented to the sensor and algorithm.

Note: If there are no or missing information shown after starting the measurements on
the “Test Measurement…” page for an Inspection Point, this means the algorithm has
failed to measure. Rework the algorithm set up until you can successfully test or
calibrate the checkpoint. No error messages are displayed on this page for these failed
measurements.

5.6 The Y-Z Plotting tool in WinSen


The purpose of the Y-Z plots is to indicate necessary actions to make sure a specific
algorithm will yield accurate measurement results. Currently, this tool is only available in
WinSen, but should be built in to some future release of Vector. It is a magnification plot
showing the details of CGs and some other important algorithm concepts in critically
important regions of the measurement. The Vertex Edge and Hemmed Edge algorithms are
extremely difficult to diagnose without using this tool.

98 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


There are a few algorithms that have the capability to show critically important results in a
chart that is only available through the use of the WinSen tool. The algorithms that are
currently set up to use this capability are:
- Range
- Cylinder algorithm configured to measure cylinders such as a fixturing pin
- Cylinder algorithm configured for threaded studs
- Vertex Edge
- Hemmed Edge
- Floating Point Edge (using the edge and range results)
- Floating Point Corner (using the corner and range results)
The above algorithms have Y-Z plots available because what is shown in these plots
translate into specific corrective actions. The other algorithms do not display a Y-Z plot
using the same logic. For these algorithms, there is no Y-Z plot information that translates
into specific actions that improve setup configuration choices.
To illustrate an example where corrective action is required, the Hemmed Edge algorithm
uses a circle fit for the nose of the edge. An incorrect profile due to the shape of the part
would indicate switching the algorithm to Vertex Edge algorithm. Distorted CGs on a part
that has a correct profile for a good circle fit would require a re-aiming of the sensor to
improve the image, which should result in a set of CGs that follow the part correctly.
Another example is for the Vertex Edge. This is an extremely important tool for
establishing a correct vertex angle. Without the tool, one can only guess at an angle.
See the “WinSen for IPNet” manual, 009-0347-01 for the details of how to run this tool.
The meaning of these charts is explained in this manual for each of the above algorithms.

5.7 The Advanced Analysis


The purpose of this is to test various conditions that could mean the difference between a
successful measurement, or one that has blunders and missed measurements. These are all
numerical and configuration tests. Conditions that could be a problem are shown with
either a yellow bell, or a red bell.
This symbol indicates a likely source of trouble, or an error.

This symbol indicates a warning of a possible source of trouble.

Using the bend and line length ratio as an example, if the line length is 1/2 the bend length,
the fit must project to a theoretical intersect where the projection is might be unstable. The
user will see a yellow bell for this. For a line length that is 1/3 the bend length or less, the
projection to the theoretical intersect is much more likely to become inaccurate.
Repeatability tests may not highlight such a problem, and therefore is not adequate proof of
algorithm accuracy.
It is important to understand that a red bell does not prove that the measurement will fail.
If it is possible to fix the problem, one would be well advised to fix it. Without a
correction, there may be more work required to make other things perfect to make up for
the highlighted problem. If a red bell condition is left, and the CMM comparison fails
because of higher than expected variation from Perceptron, this would indicate that the red
bell is a very likely indicator of the root of the problem.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 99


Additional documentation is available in the on-line help, so it is not repeated here.

5.8 Range Algorithm


Feature definition: The Range feature described in section 5.2.2.
The Range algorithm measures the deviation of a single point on a flat plane, as a 1-D
algorithm, similar to a single CMM measurement result on a plane. This algorithm
measures the actual of a surface along a search path, which was defined during sensor
setup. The measurement path is perpendicular to the projected line, inside the sensor's laser
plane. The results are equivalent to a single point measured by a CMM.
1. Place the image-processing window over the selected laser line. Make sure
that the image-processing window encompasses the entire search direction in
the horizontal direction. For many range measurements, it is appropriate to
make the image-processing window wider than its height.
2. Adjust the Exposure level to an appropriate exposure level.
3. Drag the yellow nominal point to the desired point on the laser line. This
should be quite close to the nominal point defined in the body coordinate
nominal definition.

Figure 5-23. Setting up of a Range Algorithm

4. Press the Teach button. The display updates with the returned information
from the algorithm; the line segment which is blue are the CGs returned, the
green intersecting line is the surface normal along which the measurement is
taken. Since the green line will show where the measurements will be taken, it
is appropriate to examine the image-processing window to be sure that it
appropriately set. Typical window settings need to be wider than tall, but the
example shown here is not such a situation.

100 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Figure 5-24. A taught Range Algorithm

The green line shown after the Teach button is pressed is the normal to the
feature shown in the image. It does not appear normal in the image, because of
the perspective distortion of the sensor’s triangulation angle. See section 5.3.7.
5. Only when the algorithm has been configured, taught and tested, and the
Enable box is checked on the Inspection Point Properties page, the inspection
point is ready for reporting.

5.8.1 Search for Laser Line from the Top or Bottom Setting
The Range algorithm searches along the line normal to the taught surface for a feature to
measure. In Figure 5-24, this search goes along the green line, extending as far across the
red image-processing window as is possible. In both images, there is no noise along the
extended green line, which means that the selection of search direction is not important.
Only a very few images will have any noise along this line, the search direction must be set
such that the intended feature is found by the appropriate search direction. For example, if
there is image noise on the bottom along the search line, one should select the Top option.
For a more complete view of the search line, use the Show Algorithm option on the Test
Measure page. The entire search line is shown in blue on this page and may be carefully
examined for possible image problems.

5.8.2 Range Algorithm Y-Z Plot results


See WinSen for IPNet or Vector manual, part number 009-0347-01 for the steps required to
view the Y-Z plot results.
Figure 5-25 shows an example of the Y-Z plot output for the Range Algorithm. This
algorithm shows:
o The CGs used, shown as green dots
o The blue approach vector, which should be normal to the measured surface.
o The orientation of the found surface with respect to the sensor.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 101


Figure 5-25. The Range Algorithm Y-Z plot

If the found CGs are not normal to the surface, this could indicate that the surface is no
longer in the same orientation as the part taught, but is more likely to be a situation where
the algorithm found something incorrect. If the algorithm did find something incorrect, the
windowing or the top-down/bottom-up selection may be incorrect for the given image. The
fixturing, multipath or the actual part may also be creating the problem.
The conditions necessary to review the Y-Z plot for this algorithm must be quite unusual.
However, the Y-Z plot is a useful tool to analyze any problems when these unusual
conditions occur.

5.9 The 1-D Min Range Algorithm


Feature definition: The Range feature described in section 5.2.2.

The 1-D Min Range and 2-D Min Range algorithms are similar in names only. Their
purpose, functions setups and results are very different.

The 1-D Min Range algorithm functions very much like a Range algorithm in that it finds
the closest/furthest distance. Instead of the 2-D difference between nominal and actual
points, which is a 2-D result, this algorithm creates a result that is a distance along a
measurement line. It is intended to measure broad, flatter surfaces, where Min Range is
intended to measure very small features.
Features that the 1-D Min Range algorithm should be used include:
• A surface with a large radius bend, where the result can only be a range, and the
range algorithm may not find the right point.

102 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


• A small, flat feature, where the range algorithm can miss the intended flat surface
• The flat region around a tooling pin, where there may be missing segments to the line
and the Range algorithm will not work properly. (See Figure 5-29.)
With a large radius surface, the 2-D Min Range algorithm frequently can appear to have
repeatability, when in fact the 1-D component of this measurement is actually very good.
For example, if one were to measure a cylinder with a radius of 50mm, the answer could be
1mm left or right, and still be a valid answer, within 0.01mm. The repeatability of the 2-D
Min Range would look very bad in such a situation with this variation. The extra
dimension creates much more confusion, and is not appropriate for reporting.
The 1-D Min Range algorithm reports all deviation along the line as shown in Figure 5-26
and Figure 5-27.

Figure 5-26. A Sensor Aim using the 1-D Min Range Algorithm for a Hinge Insert

Figure 5-26 shows the 1-D Min Range algorithm configuration for measuring a hinge
insert. The algorithm may be configured to measure either the top of the insert or the base
into which the insert is mounted. Therefore, using two inspection points to measure the top
and bottom to be sure that the insert is fully seated with a Relationship difference of the
NTM result. With proper algorithm configuration, (typically “Search From Top” or
“Search From Bottom”) a fully seated insert may measure anywhere on the top of the
insert. For this example, a skewed insert will measure the only measure the point closest to
the sensor. Typically, this is the desired result.
Figure 5-27 shows the deviation result of a large-radius curved feature. This is an alternate
use of the 1-D Min Range.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 103


Figure 5-27. Result of the 1-D Min Range Algorithm on a Curved Surface

The mechanical roughly equivalent of this concept is a knife-edge on a slide mechanism.


The knife-edge will be normal to the approach vector, measuring along the approach
vector. Pushing the knife-edge into the part, the first contact point is the returned
measurement result, however, the contact point on this edge is not measured, only the one-
dimensional deviation. This does not share the repeatability issues that exist with the
standard Min Range algorithm; however, this is only because there is no information about
where the point of contact was made with the surface. This algorithm choice is very
appropriate for parts with large radii, such as a large cylindrical shape or an irregular
surface, or things where only one-dimensional information is desired. The knife-edge
analogy has a mechanical equivalent when approaching one way. However, the approach
from the other has no mechanical equivalent.

Figure 5-28. 1-D Min Range algorithm

104 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Absolutely flat surfaces can be measured, but this is not recommended and the Range
algorithm would be the better choice. However, the Range algorithm requires a sufficiently
large flat such that the measurement point always lands of this surface. Small flat features
such as the one illustrated in Figure 5-29, where the viewable surface is too small to
support a proper range measurement. This is a situation where the 1-D Min Range
algorithm is a better choice than a Range, where the range may not land on the flat surface
where it can be seen by the sensor. The 1-D Min Range can measure properly if the laser
line lands on either the left side, right side or both sides of a locating pin.
The IJK and the measurement must be setup carefully in this example measuring the flat
surface. If these parameters are not exact, a bimodal measurement can result if one of these
two laser segments disappears as shown in Figure 5-29.

Figure 5-29 Example of measuring a small surface with 1-D Min Range

1. Place the image-processing window over the selected laser line and
adjustment boxes.
2. Adjust the Exposure level to an appropriate exposure level.
3. Select the Search from Top or Search from Bottom (as displayed by the
green arrow) for the appropriate approach.
4. Set the Laser Filter Width (shown in yellow). On a curved surface, setting the
laser filter too wide will result in the wrong segments being used. Too narrow
will pick up any additional image noise. For flat surfaces, the laser only need
be set wide enough to eliminate the image noise. The slope and height
parameters are not adjustable like other algorithms.
5. Use the Graphics to set the approach direction. Rotating the green approach
vector does this. Use the handle on the green box to orient the approach.
6. Set the yellow box to describe the laser filter width. Other parameters of this
filter are not adjustable. Setting the laser filter width too wide on a curved
image will select the wrong part of the arc to measure when measuring a
curved surface. A filter that is too wide may also connect bits of noise to the
desired feature, which may not be appropriate. Setting this too narrow will
mean that any breaks in the laser line will not be connected when it is

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 105


appropriate to connect them. This laser filter is unique to this algorithm, in
function and operation.

Figure 5-30. Setup for 1-D Min Range algorithm examples

7. Press the Teach button. The display updates with the returned information
from the algorithm; the line segment which is blue are the CGs returned, the
green intersecting line is the surface normal along which the measurement is
taken.

106 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Figure 5-31. Taught 1-D Min Range algorithm examples

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 107


Figure 5-31 shows a large cylinder and a hinge insert measurement images all
taught with the 1-D Min Range algorithm. The hinge insert images show the
Search-From-Top example, and the Search-From-Bottom example. The
algorithm result in the hinge examples may appear on the left, right or both.
However as long as the IJK matches the actual surface, the algorithm will
produce the desired results. In the last image, this result happens to be on the
far right.
8. Validate the approach vector to make sure that it is proper. If not, repeat step 4
to readjust the approach.
9. Only when the algorithm has been configured, taught and tested, and the
Enable box is checked on the Inspection Point Properties page, the inspection
point is ready for reporting.

108 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


5.10 The 2-D Min Range Algorithm
Feature definition: The Cylinder feature described in section 5.2.5. This algorithm should
only be applied to features with very small radii, (2mm or less).
The 2-D Minimum Range algorithm measures the point closest or farthest to a selected
boundary within the image-processing window of a small feature. This reports the
deviation from the nominal tangent point to the actual tangent point. The typical feature
will have a very small radius as compared to the sensor's field of view. With this small
feature, the tangent line and the approach vector are by default defined in sensor space. An
additional option allows the user to use a normal to metal approach vector, defined in part
coordinates.
The 2-D Min Range algorithm results may appear non-repeatable, but completely accurate
with features that are too flat. The measured point may roam freely on the surface, yielding
an unrepeatable measurement along the tangent line. The result is accurate in the approach
line; however, one should consider the appropriateness of 1-D Min Range algorithm if the
non-repeatable result creates a problem in understanding the output of this measurement.

Figure 5-32. 2-D Min Range Deviation from Nominal

1. Place the image-processing window over the selected laser line and
adjustment boxes.
2. Adjust the Exposure level to an appropriate exposure level.
3. Select the Search for Laser Line (green adjustment box) for the appropriate
approach and Top or Bottom. The orientation of the green arrow is the control
to adjust the approach vector direction. Rotate the box/arrow combination to
provide the measurement results that are required.
4. Set the laser filter, in the rare event that it is needed.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 109


5. Define the Line Selection. Use Longest, Left Most or Right Most. This is
for the rare situation where multiple line segments exist in the image. For most
applications, the Longest option is the only option needed.
6. Determining whether the point returned will be the highest or the lowest with
reference to the tangent line. The user is setting the approach vector at, above,
or below the tangent line.

Figure 5-33. Setup for 2-D Min Range algorithm

7. Press the Teach button. The display updates with the returned information
from the algorithm; the line segment which is blue are the CGs returned, the
green intersecting line is the surface normal along which the measurement is
taken.
8. Only when the algorithm has been configured, taught and tested, and the
Enable box is checked on the Inspection Point Properties page, the inspection
point is ready for reporting.

Figure 5-34. Taught 2-D Min Range algorithm

110 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Figure 5-34 shows the results for a taught 2-D Min Range, showing the CGs found in red,
the CGs used for the measurement in white, and a red arrow indicating the final answer and
the approach direction of the measurement.

Note: The repeatability of the 2-D Min Range may look bad, but the results may be
correct. Read the introduction to the 2-D Min Range and 1-D Min Range algorithms
carefully to understand how this can happen.

5.11 Corner Algorithm


Feature definition: The Corner feature described in section 5.2.3.
The Corner algorithm is designed to conform to the 8-hit or 5-hit strategies of CMM
measurement, as described in Appendix A of the “Correlation/Comparison Process
Between MCMM and OCMM” manual, part number 009-0185. As long as there are no
significant angle changes to either plane that makes up the corner, other CMM strategies
can be successfully compared to the Corner algorithm.
This algorithm locates the theoretical best-fit intersection of two laser lines in the image.
Corners will have either a "V" or an upside-down "V" shape in the image. The section of
the laser line that exists between the bend and line lengths define a line on each side of the
"V". The theoretical intersection of these lines is the measured point.

5.11.1 Teaching the Corner


The laser-filter parameters and the bend- and line-length parameters are combined into a
single rectangular box for each side of the corner. These are controlled using two boxes on
the screen, a green and orange box. Both boxes set the:
• Bend Lengths – by positioning the green and brown adjustment boxes
over the intended lines.
• Line Lengths - by stretching or shrinking the green and brown adjustment
boxes).
• Filter - by selecting the height of the box the row tolerance is set. See
below for more detail on how this is used.
• Slope - the angle (rotate) at which the box is placed on the laser line image
on the screen. The slope of the left and right sides of the corner must be
set carefully to match the expected measurement as close as possible.
The laser filter concept is very different for this algorithm. The slope and filter size
controls the filtering concept. In the presence of image noise, the height of the box should
be narrow enough to reject most of the unwanted noise. For example, the Z-step feature
around a windshield decking has multipath issues, and the row tolerance should be set
narrow enough to help eliminate the unwanted CGs from other sources.
The corner algorithm do the following steps to process the image
- The slope, bend and line lengths are used to find the approximate location
of the corner in the image.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 111


- Using the rough location, the row tolerance and slope of the laser filter is
used to discard any CGs found outside these boundaries.
- The corner algorithm then iterates to refine the bend and line length results
around the theoretical intersect point.
Teaching the corner involves the following steps:
1. Adjust the Exposure level to an appropriate exposure level.
2. Once those features are set, place the image-processing window over the
selected laser line and the adjustment boxes.

Figure 5-35. Setting up a Corner algorithm

3. Adjust the left and right templates to match the desired slope carefully. The
brown and green boxes adjust the slope, above/ below row tolerances on the
laser filters, bend length and line length simultaneously. Matching the slope is
extremely important to the proper functioning of this algorithm. Getting the
bend and line length to the proper values requires iteration on the millimeter
values displayed after the teach button is clicked.
4. Adjust the height of the boxes if any image noise is expected to control the row
tolerance. Figure 5-36 shows no such noise, so the size of the row tolerance is
not very important. For images with multipath noise, a very small height is
recommended.
5. Press the Teach button. The display updates with the returned information
from the algorithm; the line segment which is blue are the CGs returned, the
green intersecting line is the surface normal along which the measurement is
taken. It is very important to read and understand the computed bend and line
lengths output at this time, and if they are not what are desired, the adjustment
boxes must be adjusted to achieve the proper results, as discussed in step 4.
Typically the default situation would require that all bend and line lengths be
set to 5mm lengths, but may need to be set different if there is very little metal
on one side or the other, or there is a very large radius between the two legs of
the corner to be measured. Consider using the Vertex Edge if there is very
little metal on one of the two sides (see Section 5.12.2.).

112 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Figure 5-36. A Taught Corner Algorithm

5. There are limits and alarms set to color code bend- and line-length parameters.
Any red or yellow flags suggest a possible problem. For example, if a line
length is less than 1/3 the bend length for the same side, this will be
highlighted in red. The reason that this is flagged is that the geometric
construction for the theoretical intersect can become unstable without a perfect
image. This should be corrected if the application permits it. If the application
does not permit fixing the red or yellow conditions, make sure the customer is
aware of the possible problems.
6. Only when the algorithm has been configured, taught and tested, and the
Enable box is checked on the Inspection Point Properties page, the inspection
point is active and ready for reporting.

5.12 Edge Algorithms


Feature definition: The Edge feature described in section 5.2.4. This applies to all edge
algorithms (Vertex, Hemmed and Simple Edge algorithms).
The different edge algorithms are all designed to conform to the 4-hit strategy of CMM
measurement, as described in Appendix A of the “Correlation/Comparison Process
Between MCMM and OCMM” manual, part number 009-0185. As long as there are no
significant angle changes to the edge, other CMM strategies can be successfully compared
to the Edge algorithm.
All edge algorithms generate a two dimensional measurement of an edge and measures an
intersection between the straight portion of the plane surface and the endpoint. This
intersection produces a single coordinate result.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 113


Figure 5-37. Part-to-Sensor Angle (or Edge Angle) as it relates to the part

The above figure shows the part-to-sensor angle, which is a rotation about the sensor’s X-
axis starting at the edge and getting more positive as the sensor is aimed into the plane.
This is important to understand for the following chart, and the recommended choice of
Hemmed, Vertex or Simple edges.

Edge Edge Description Part-to-Sensor Angle


Hemmed or rolled edge, where the From 30° to 70° is desired,
Hemmed Edge edge has a radius, typically 1-2mm, however 10-95° may work.
*
but can be more. 45° is the nominal aim.
Edge ends in a straight bend, or From 30° to 90°,
Vertex Edge hemmed edges with less-than-ideal depending on vertex angle
*
aiming can use Vertex Edge. specified.
Sensor cannot view the point on the From 90° to 130° (sensor
edge that needs to be measured. is behind the edge to be
Therefore, certain edge properties are measured)
Simple Edge assumed very consistent. This is not
the recommended algorithm for all
but where the true edge cannot be
seen by the sensor.
This is an algorithm that is 90°
constructed for measuring the edge of
MinRange YZ glass with a specific image. It is
therefore never recommended for any
sheet metal measurement
*
For hemmed edges, or any edge with a radius less than 2mm, and a sensor-to-feature
angle less than 90 degrees, there should be an additional rotation around the sensor’s Y-
axis, as shown in Figure 5-38 below:

114 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Figure 5-38. Recommended additional rotation for hemmed edges around the sensor’s Y-axis for a contour
sensor.

What this rotation is doing is suppressing the intensity of the very end of the roll radius,
given a better result. Any specular condition on the edge will create very bright spot that
can increase the measurement’s uncertainty. Although the laser line section is no longer
normal to the edge, this is not as important as reducing the effects of the bright spot without
this rotation.
For an FMS or any other type of surface sensor, the rotation is reversed due to the geometry
of the sensor. Since the optic line is normal to the sensor case and the laser is typically 20
degrees, rotating the sensor 20 degrees to make the laser normal and then an additional 10-
15° makes the sensor rotation a total of 30-35 degrees, which is reduces the distance to the
part considerably. Therefore, rotating the sensor in the opposite direction is recommended.

Figure 5-39. FMS or Surface sensor rotation, as viewed from behind the edge

The rule being applied here is the surface being normal to the line bisecting the laser and
optic is the worst-case specular angle. The laser normal, or the optic normal is 10° away
from the worst case. Additional rotation reduces the specular reflection on the critical
edge, allowing a small increase in exposure to get more return from the points around the
edge that are critical to a successful measurement.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 115


Figure 5-40. Geometric construction of the edge

Figure 5-40 is a close-up view of the left side of Figure 5-38. This shows the laser section,
the geometric construction of the Perceptron results, and the CMM hit points using a 2-hit
strategy. The laser plane may not be normal to the feature; however, the mathematics in
the Perceptron control accommodates any reasonable situation like this. There may be
some rotation or translation from the desired measurement, and this will yield valid results.
This means that the goal is not to worry about what is sectioned, but to be concerned with
the best measurement result possible. Matching any taught parameters to the CMM
programming is another goal.
Additionally, it is better on small edges to make sure that the edge is on the left side of the
window. Small edges include hemmed edges that are 1-2mm, and some trimmed edges.
For theses small edges, the edge on the right and the plane on the left are not as good as the
reverse situation.
If this is a small edge, and there are no mechanical or ambient conditions to constrain the
measurement, one should mount the sensor such that the image is turned around to look
like the left image below, and try to avoid the right image whenever possible. This is
accomplished by simply turning the sensor around its Z-axis by 180 degrees. This is even
truer with larger field of view sensors, as the effect is more pronounced. This preference is
the result of the CCD imagers that is currently being used in TriCam sensors. Note that this
will cause more problems on some TriCam sensors and not others. The CCDs of a given
TriCam will have slightly different properties from another, which makes the effect of the
image on the right somewhat less predictable.

116 • Setting Up Algorithms 009-0506, Rev. B 009-0506009-0506


Figure 5-41. The preferred image 5-42. A less desirable image.

5.12.1 Hemmed Edge


The Hemmed (rolled) Edge technique is intended for situations where the
edge has an end with a radius, and the sensor can be aimed properly around
the edge, completely showing the end. Proper sensor aim is quite important
to get the desired images. The rotation about the sensor Y-axis shown must
be in the 10 to 20 degree range, where it is 10 to 15 is recommended.
Rotations that do not meet these exact specifications will mean that the
Hemmed Edge algorithm will not function, and you must use Vertex instead.

Note: A properly set up Hemmed Edge algorithm will yield better results
that Vertex, but the Hemmed Edge algorithm will not yield acceptable
results with improper aiming.
Hemmed edges are somewhat easier to measure if the sensor is setup so
that the edge is on the left side of the image and the plane is on the right.
Typically, this means rotating the sensor about its Z-axis 180 degrees and
making the other appropriate aiming changes.

The limits for the orientation of sensors aimed at hemmed edges are:
• X-axis rotation 30 to 70 degrees,
• Y-axis rotation 10-20 degrees for a normal contour sensor, and
• Z-axis rotation -30 to +30 degrees.
Study the results of the “Advanced Analysis” carefully before setting up this
algorithm. Using the WinSen YZ plot functions are also strongly
recommended.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 117


Figure 5-43 Geometric construction of the Hemmed Edge algorithm

Figure 5-43 shows the geometric construction of the Hemmed Edge


algorithm. With a proper sensor aiming, this yields the best results with a
hemmed or rolled edge that is commonly found on closure panels.
Since this algorithm fits a circle to the round edge, there must be a proper
view of this end to yield the better measurement results.

5.12.2 Vertex Edge


The vertex edge is defined as the intersection point between the edge
measurement using a shank line and the plane of measurement. Measuring
hemmed edges is possible with Vertex Edge techniques, and is recommended
for hemmed edges that do not have the sensor located adequately for a proper
hemmed edge style of measurement.
Note that if the sensor has a reliable view of the edge, there is no difference
between a Vertex Edge measurement with a 0° vertex angle and the Hemmed
Edge result. For absolute calibrated systems, the larger the vertex angle
means the larger the offset required. Note the difference in Figure 5-44
between the true edge and the theoretical intersect point. Since most
hemmed edges are 1 to 2mm, the offset shown in this illustration is also
about 1-2mm with the shape of the metal illustrated. This is a typical offset
required with the 45° vertex angle.
Vertex Edge is also recommended for features that are more like corners, but
with very small flat ends that are too small (1-3 mm, typical) for a proper
corner measurement.

Figure 5-44. Geometric construction of the Vertex Edge on a hemmed edge

Figure 5-44 shows the geometric construction of a Vertex Edge algorithm


result. The plane is measured between the bend length and the line length,
and the vertex edge result is the theoretical intersect between this line and the
found tangent line. What is also shown in Figure 5-44 is a tangent line that is
the result of a 45 degree vertex angle specification. Aiming that require 0° is

118 • Setting Up Algorithms 009-0506, Rev. B


recommended, but not always possible. The 0° will yield the same results as
shown in Figure 5-43. The smaller vertex angles are better as the actual end
is measured on all other competing measurement tools. However, the sensor
aim may not permit these smaller angles.
Note the offset required for this algorithm for absolute measurements. Also
note that the amount of variation in the relief on the edge and the size of the
rolled edge will have a significant error in variation with the larger vertex
angles. At 45°, the variation of the edge relief will map to a 1-to-1 error
ratio. Therefore, 45° does not work properly for all situations.
The recommended angles should be between 0° and 30° for correctly aimed
sensors. Angle higher than 30° should only be used when it is not possible to
aim the sensor properly.

5.12.3 Simple Edge


The Simple Edge technique should only be used when any of the other
choices are not possible. The typical application of Simple edge is where it
is not possible to position a sensor to properly view the edge. This is where
the edge angle is more than 90 degrees. It functions on the gray-scale fall-off
to locate the edge, which is not the point that needs to be measured, but
should correlate if setup and part conditions permit.
The gray-scale falloff is more susceptible to image noise than other algorithm
choices, which is the reason that it is only recommended where other means
will not work.
Any change in the appearance of the surface will yield an inaccurate
measurement result with Simple Edge. For the hemmed (rolled) edges
common in closure panels, it is common to see die-strike marks on the
edge. This completely randomizes the appearance of the edge under the
laser line. In some instances, achieving shim-test results less than 0.5mm
may not be possible with Simple Edge, depending on the condition of the
edge and the sensor aim.
The Simple Edge is only intended for edges where the sensor cannot be
aimed at 90 degrees or less. Proper aiming so that Hemmed or Vertex
Edge measurements are strongly recommended.

5.12.4 MinRange YZ
Feature definition: The Edge feature described in section 5.2.4. This
algorithm can also be defined as a Range feature described in section 5.2.2;
however, this is not recommended.
This algorithm is specifically designed to measure certain types of edges on
glass. Using this algorithm on sheet metal or painted parts is not
recommended.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 119


5.12.5 Edge Y-Z Plot results
The Y-Z plot function is critical for validating edge algorithm setups. The
Vertex option requires small angles for accurate results, and this tool is the
only tool that may be used to validate the sensor configuration with the target
aiming and surface conditions.
There are some extremely important concepts shown graphically with the Y-
Z plots for edges:
- For hemmed edges, measure the edge with either the hemmed edge,
or a vertex edge set to the smallest angle possible.
- Validate the aim so that the bright reflections do not interfere with
the edge measurement
- A way to select the hemmed/vertex selection choice.
- Validate the exposure setting.

Figure 5-45 A proper edge result for a hemmed edge

Figure 5-45 shows a result of an edge that has been properly aimed. The
view of the edge shows valid CGs at the edge to be measured, and the profile
matches the part without significant distortion. This image is the result of
setting the sensor aim to about 45° to the edge, and the rotation about the
sensor’s y-axis is such that there is no distortion of the CG profile.
Although the results shown in Figure 5-45 shows the results for a hemmed
edge algorithm, this kind of image will also work for a Vertex Edge
algorithm result with a Vertex angle between 0 and 45°. As discussed
previously, the smaller angles are better, but only when the image results
permit the smaller angles. This image will provide proper answers for either
hemmed or vertex, and can use a very small vertex angle.
Figure 5-46 below shows the individual components that are shown in the
hemmed edge result for the same image that created Figure 5-45.

120 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-46. The anatomy of the hemmed edge algorithm Y-Z plot

The important constructions shown in this figure are:


- The CGs found shown as red and blue dots.
- The circle fit to the red CGs with the blue cross indicating the
center of the found circle
- The red line showing the tangent to the circle fit used
- The red line showing the results of the fit to the plane
- The intersect point between these two red lines that is the final
answer to this algorithm
- The direction of the sensor, which is important for evaluation of
the bright reflections that frequently destroy the edge algorithm
results.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 121


Figure 5-47. A poorly aimed sensor showing the results of vertex with 0° and hemmed

Figure 5-47 shows the results of a poorly aimed sensor using both the vertex
at 0°(top results) and the hemmed options (bottom result) on the same image.
Both show results that cannot be accurate regardless of the configuration
choices. The problem here is that aim of the sensor is such that there is an
extremely bright spot due to the improper rotation of the sensor about its y-
axis with respect to the feature. This is the importance of the direction
indicator shown in the hemmed-edge.
On the upper image in Figure 5-47 showing the vertex edge, note the white
dots which indicate the points used to fit the tangent-to-the-edge line. The
fact that the algorithm has only distorted CGs that do not reflect the actual
shape of the part is critical. Setting the vertex edge angle to a higher value,
such as 45 degrees will only make matters worse.
On the lower image in Figure 5-47 showing the hemmed edge results, note
where the dashed line is with respect to the inappropriately small circle used
to fit the found points. If the fit was scaled to the appropriate size and center
point, the distorted CGs at the bright spot is perfectly in line with the sensor.
This sensor has CGs that move towards the sensor with the inappropriately
bright CGs. Some sensors will do the opposite, moving the CGs away from
the sensor.
Whenever there are CGs that do not reflect the actual shape of the part, the
sensor must be re-aimed to eliminate this bright spot. If the rules outlined in
this manual and the limits and alarms in the Advanced Analysis tab are not
exceeded, problems of distorted CGs are much less likely.

122 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-48. A worst-case y-axis rotation result

Figure 5-48 shows a situation where the y-axis rotation with respect to the
feature is at nearly the worst-case position, much worse than what is shown
in Figure 5-47. The image shown above is the image from WinSen. The
displayed images in WinSen are tuned to be more sensitive. Image problems
that might exist may only be revealed using the WinSen tool. The Y-Z plot
shows that this aim may show some degradation of repeatability, but can
never be accurate, and needs corrective action.
The following image shows the results of some Vertex Edge algorithm
processes and how this tool helps set a proper vertex angle.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 123


Figure 5-49. A 0° vertex angle setting that will not work

Figure 5-50. A 30° vertex angle setting that may work, but is not recommended

Figure 5-51. A 45° vertex angle setting that will work

Clearly, for the best results, one should always aim the sensor into the edge
to allow a 0° vertex angle to work properly, which will allow either the
hemmed edge or the vertex edge to work. For some situations, including
gap-and-flush measurements, this is not always possible. The smallest vertex
edge angle that will work is always the best. For Figure 5-49, Figure 5-50
and Figure 5-51, this shows the result of a sensor aimed almost normal to the
plane. The red vertex-edge line at the left in these images should pass
through several of the white dots for a successful measurement.
The vertex edge angles of 60° and higher are very rarely accurate, even if
they are sometimes repeatable. The normal process variation of the metal is
usually enough to make such a tangent fit with the larger angles
unacceptable. The white dots are quite likely to fit the tangent line easily,
with these higher angles, but you must be assured of extremely consistent
edge profiles before angle settings of larger than 45° should be used.

5.13 Floating Point (FP) Edge Algorithm


Feature definition: The Edge feature described in section 5.2.4.

124 • Setting Up Algorithms 009-0506, Rev. B


The Floating Point Edge (FP Edge) algorithm is designed to conform to the
3-hit strategy of CMM measurement, as described in Appendix A of the
“Correlation/Comparison Process Between MCMM and OCMM” manual,
part number 009-0185.
The FP Edge algorithm starts with the same process as an Edge Algorithm,
so most options for measuring the edge are the same. Given that the FP Edge
algorithm only supports the Vertex Edge configuration, the sensor aiming
strategies and options for Vertex outlined in sections 5.12 and 5.12.2 apply to
this configuration. Only the results are processed differently to more closely
match a CMM 3-hit measurement strategy.

Important Notes:
The Floating Point Edge algorithm only uses the Vertex Edge form of
measurement. The Hemmed and Simple edge choices are not possible.
All of the figures in this section are drawn showing a 0° vertex angle
configuration. See sections 5.12.2 and 5.12.4 to understand the concept of
vertex angle.
In body coordinate reporting, the results will be a 2-D section
measurement result. As with all 2-D sectioning measurements, these
results can be reported as NTF or cutting plane.

The figure below illustrates the first step in the FP Edge algorithm. This
figure shows both a nominal (red) and actual (blue) and exaggerates a
rotation to show how these results are different from the Edge Algorithm.
This figure shows how the theoretical edge is first located (shown as the
yellow dots), exactly as the Edge Algorithm does. Then the plane is
measured as a single range (shown as the blue dots) at a preset distance
behind this theoretical intersection point.

Figure 5-52. Step 1. A section cut view of the floating-point edge algorithm, constructing the
yellow and blue measuring points on the actual part (for all reporting configurations)

The results may be viewed as an edge deviation and a plane deviation result
as shown in the following figure, and show the NTM results.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 125


Figure 5-53. Step 1, Final NTM Results of the Floating Point Edge.

The body-coordinate result will construct the answer as shown as the green
point, is illustrated in the following diagram:

Figure 5-54. Step 2 of the body coordinate reporting

The 2-D sectioning deviation in this mode is shown in solid green in the
above diagram. Note how this is different from the standard Edge algorithm
when there is part rotation. If the orientation of the metal is the same, the

126 • Setting Up Algorithms 009-0506, Rev. B


green point and the yellow point will be the same. Only when the shape or
orientation of the metal surfaces changes, will the results be different.
Note that the reported deviation (shown in the solid green) is quite similar to
the deviation of plane point, shown as the dashed green line. If the CMM
measures this result, these slightly different results are certainly close enough
for very good correlation.

5.13.1 The Report Options for the Floating Point Edge Algorithm
Using the NTM reporting configuration, this is the equivalent algorithm
found in most CMMs and is the more common configuration, where the
NTM measurement process in step 2 measures the plane and edge deviation
with given distance away from the theoretical intersection. Disabling the
Report Both Plane and Edge option is exactly equivalent to reporting only the
NTM plane deviation, and therefore not recommended.
Using the body coordinate mode, the Floating Point Edge Algorithm is a 2-D
sectioning result on the constructed point. Therefore, NTF and cutting plane
reporting options are possible. Using the body coordinate reports for this
algorithm requires some very specialized software for the CMM, regardless
which reporting is selected, NTF or a cutting plane. Be very sure that the
CMM measurement strategy is well understood and matches the selected
reporting option.
For the body-coordinate reporting, only the NTF reporting is reasonable in all
but a very few situations. While it is possible to report body coordinate
deviations using an X, Y or Z approach vector, there is no known CMM
equivalent measurement strategy for an approach vector option.

5.13.2 Teaching the Floating Point Edge


1. Press Continuous Live Image, preferably with the Teach – Full
Image to view the line image.
2. Adjust the Exposure level to an appropriate exposure level.
3. Place the Image Processing Window over the desired section
of the laser line. Size the window to include only the portion of
the graphic that is of interest. Make sure to allow for part-to-part
variation plus the size of the entire feature to be measured.
Sizing the window too small will cause sensor errors.
4. Drag the edge yellow location/line-filter box beyond the end of
the laser line indicator. (Shown in Figure 5-55 with a hemmed
edge.) It need not be accurately positioned, and may be located
aside so to avoid the other graphic items. This only identifies the
left or right orientation of the edge.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 127


Figure 5-55. Setting up a Floating Point Edge algorithm

5. Press the Teach button. The display updates with the returned
information from the algorithm; the line segment which is blue
are the CGs returned, the green intersecting line is the surface
normal along which the measurement is taken.

Figure 5-56. A taught Floating Point Edge algorithm

6. There are two features under the image window: Range


Parameters and Edge Parameters.

128 • Setting Up Algorithms 009-0506, Rev. B


a. Range Parameters allows the user to select Search for
Laser Line From: Top or Bottom, and define the Poked
Column / Edge Offset.
i. The Search for Laser Line From feature allows the user
to select the direction of the search path, when searching
for the nominal. This instruction commands the
processor to follow along the nominal line towards the
base or laser image intersection, in the search for the
measurement point. Search from Top allows the user to
search along the approach vector for the nominal point
from the top of the screen along the approach vector,
while Search from Bottom allows the user to search
from the bottom of the screen to the top. Choose the
direction that will have no noise along the desired search
path. The little green line indicates the path used.
ii. Poked Column Used / Edge Offset either displays or
allows the user to enter in a value in millimeters to be
used as the offset value for the range measurement from
the found edge point.
b. Edge Parameters allows the user to enter in a Vertex
Edge Angle (Vertex Edge only) and Line Selection.
i. Vertex Edge Angle is the approach angle that applies
only to the vertex edge find technique. Setting the
approach angle to zero provides a capability equivalent
to a CMM shanking the edge surface parallel to the
plane.
ii. Line Selection defines which portion of the part's image
(in the image window) is measured; Longest Line, Left
Most or Right Most. Make sure that the desired
selection is displayed on the screen before proceeding.
Most images will only present a single line, in which
case this setting is not important.
9. Leave the laser filter parameters to the default values.
10. The checkpoint should be set to Active returning to the
Configuration screen. This checkpoint is now complete and
will run during a manual or automatic gauge cycle.

5.13.3 Rules of Programming Floating Point Edge


There are some specific sequences that must be followed for programming
the Floating Point Edge. Failure to follow these processes will create
disappointing results.
- The sensor angles and IJKs must be correct before nominal calibration.
- If there are any changes to the IJKs or the sensor angles, the nominal
calibration will be wrong until a new calibration is performed. Either
calibration to metal or computed calibration must be done over after
these changes.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 129


5.13.4 Floating Point Edge Y-Z Plot results
This algorithm is a construction of an Vertex Edge algorithm and a range
algorithm. Therefore, refer to sections 5.12.4 and 5.8 to understand the
components of each part of your result. Many Floating Point Edge results
only report the plane, which frequently means that using the Y-Z plot for the
edge is not important, only the Range is.

5.14 Floating Point (FP) Corner Algorithm


Feature definition: The Corner feature described in section 5.2.3.
The Floating Point Corner algorithm is designed to conform to the 3-hit
strategy of CMM measurement, as described in Appendix A of the
“Correlation/Comparison Process Between MCMM and OCMM” manual,
part number 009-0185.
The first step in the Floating Point Corner Algorithm is similar to the Corner
Algorithm outlined in section 5.11. Please read and understand these
concepts first. The FP Corner algorithm measures two ranges at a preset
distance from the theoretical intersection point of a corner.

Important Note: In body coordinate reporting, the results will be a 2-D


section measurement result. As with all 2-D sectioning measurements,
these results can be reported as NTF, cutting plane or left and right NTM
results.

The figure below illustrates the first step in the FP Corner algorithm with a
cross-section view. This figure shows both a nominal (red) and actual (blue)
corners and exaggerates a rotation to show how these results are different
from the Corner Algorithm. The theoretical intersection point is first located,
exactly as the Corner Algorithm does, which is shown as the yellow dots.
The taught offsets are then used to locate the blue dots, according to a set
distance from the theoretical intersection point. The two nominal points and
the actual points are both illustrated as the blue dots. Both the nominal (red
corner) and actual (blue corner) are shown.

130 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-57. Step 1 of the floating-point corner algorithm, constructing the blue measuring
points on the actual part

The results may be viewed as a left deviation and a right deviation result as
shown in the following figure, and show the NTM results.

Figure 5-58. Step 2, NTM reporting shown

Step 2 of this algorithm for body-coordinate reporting will construct the


result, shown as the green point, is illustrated in the following diagram:

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 131


Figure 5-59. Step 2 of the body coordinate reporting.

The deviation shown in this mode is shown in green in the above diagram.
This deviation may be reporting in NTF or using the appropriate cutting
plane. Note how this is different from the standard Corner algorithm, where
the reported deviation would be the difference between the actual (yellow dot
on the actual part, shown as the blue corner) as compared to the nominal,
shown as the red part.
If the angle of the metal and its orientation is the same, the green point and
the yellow point will be the same. In other words, there will be no difference
between the Corner algorithm results and the FP Corner results, given the
consistent orientation of the metal. Only when the shape or orientations of
the metal surfaces change will the results be different.
Note that the reported deviation (shown in green) is similar to the deviation
of left and right side measurement points shown in blue. Obviously, the less
the rotation of the metal, the less these differences will be. This rotation can
be either one side of the corner, the other or both. If a CMM measures only
one side, this is usually close enough for correlation, but be aware of the
possible difference.
The correlation of body coordinate reporting requires some very specialized
software for the CMM, regardless whether it is reporting NTF or a cutting
plane. Be certain that the CMM has been programmed to match this
measurement strategy.

5.14.1 Using the Report Options for the FP Corner Algorithm


Using the NTM reporting configuration, this is the equivalent algorithm
found in most CMMs and is the more common configuration. The NTM
measurements process in step 2 measures one or both of the plane deviations
with given distance away from the theoretical intersection.
Using the body coordinate mode, the Floating Point Corner Algorithm is a 2-
D sectioning result on the constructed point. Therefore, NTF and cutting
plane reporting options are possible. Using the body coordinate reports for

132 • Setting Up Algorithms 009-0506, Rev. B


this algorithm requires some very specialized software for the CMM,
regardless which reporting is selected, NTF or a cutting plane. Be very sure
that the CMM measurement strategy is well understood and matches the
selected reporting option.
For the body-coordinate reporting, only the NTF reporting is reasonable.
While it is possible to report body coordinate deviations using an X, Y or Z
approach vector, there is no known CMM equivalent measurement strategy.
Both of the I, J, K normal-to-surface vectors are necessary before teaching
the algorithm.

5.14.2 Teaching the Floating Point Corner


1. Adjust the Exposure level to an appropriate exposure level. .
2. Set the Adjustment Boxes on the laser line. Use the square to
adjust the width/height. Use the circle to rotate the box to the
slope of the laser line. The adjustment boxes will set the
following:
• Line Extents / Filter - by stretching the box from left to
right (width). The Filter allows the user to define where the
laser line is, and which part of the graphic to use for final
measurement. Once set, all the CGs aligned inside the line
filter band, and all CGs no farther away then half the width
of the line filter, are accepted. All other CGs are ignored.
The Line Extents should encompass a straight portion of the
graphic while ignoring the radius toward the corner.
• Slope - the angle (rotate) at which the box is placed on the
laser line graphic on the screen.
3. Place the Image Processing Window over the desired section
of the laser line. Size the window to include only the portion of
the graphic that is of interest. Make sure to allow for part-to-part
movement. Sizing the window too small will cause sensor
errors.
4. Set the range value. This is the distance from the found
theoretical intersect where the range measurement will be
positioned for each side.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 133


Figure 5-60. Setting up a Floating Point Corner algorithm

134 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-61. Setting up a Floating Point Corner algorithm

5. Press the Teach button. The display updates with the returned
information from the algorithm; the line segments that are red
are the CGs returned, the green intersecting line is the surface
normal along which the measurement is taken.
11. Left/Right Range Parameters allows the user to select Search
for Laser Line From: Top or Bottom, and define the Poked
Column / Corner Offsets.
a. The Search for Laser Line From feature allows the user to
select the direction of the search path, when searching for the
nominal. This instruction commands the processor to follow
along the nominal line towards the base or laser image
intersection, in the search for the measurement point.
Search from Top allows the user to search along the
approach vector for the nominal point from the top of the
screen along the approach vector, while Search from
Bottom allows the user to search from the bottom of the
screen to the top.
b. Poked Column / Corner Offsets allows the user to enter in
a value in millimeters to be used as the offset value for the
range measurement from the found corner point.

Note: The user MUST first teach the algorithm at least once before using
the Entered Offset feature.

12. The checkpoint should be set to Active returning to the


Configuration screen. This checkpoint is now complete and
will run during a manual or automatic gauge cycle.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 135


5.14.3 Rules of Programming Floating Point Corner
There are some specific sequences that must be followed for programming
the Floating Point Corner. Failure to follow these processes will create
disappointing results.
- The following is true for the 4.3 release, but may be true for future
releases: The IJKs left and right must match the image left/right
configuration.
- The following may be true for future releases, but other options may
make this test unnecessary: The left/right image/IJK check is done on
the Advanced Analysis tab. If the measurement-to-feature angles are
much larger than the sensor-to-feature angles, this is a strong indication
that the IJKs do not match the image.
- The sensor angles and IJKs must be correct before nominal calibration.
Any changes to the IJKs or the sensor angles, the nominal calibration
will be wrong until a new calibration is performed. Calibration to
metal or computed calibration must be done over after these changes.
- Although possible to swap the left and right image parameters during
teach, this is not recommended. Keep the brown box on the left, and
the green on the right.

5.14.4 Floating Point Corner Y- Z Plot Results


This algorithm is a construction of a corner algorithm and two range
algorithms. Therefore, see section 5.8.2 to understand the components of
each plane used to construct the result.

5.15 Cylinder Algorithm


Feature definition: The Cylinder feature described in section 5.2.5.
The Cylinder algorithm is designed to measure the position of a cylinder or
post. This algorithm works through fitting a circle to the laser line image.
Returned is the YZ deviation of the measured circle center from the reference
circle center.
Cylinder may also be used to measure the location of a threaded stud. See
Section 5.15.3 for more information about threaded stud measurements.
The aim of the sensor can be such that the laser is not normal to the feature.
This creates an elliptical section of the cylinder, and the algorithm will
process this correctly. As with all other algorithms, the most important issue
is to get a good image and the longest data sample around the cylinder that is
possible.

5.15.1 Parameters for Cylinder


The Cylinder algorithm parameter choices are shown in Figure 5-62.

136 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-62. Arc Parameters (shown with Threaded Arc Type enabled)

The Type selection has the following choices:


- Standard: This is for a cylindrical pin or a cylindrical shoulder on a
stud. The aiming and metal conditions assume no reflection at the
center of the arc in the image. Most conditions allow the sensor to be
aimed such that there is no need for the cutout notch. Therefore, this is
the recommended setting for measuring cylindrical features.
- Cutout notch: This is also for a cylindrical pin or a cylindrical
shoulder on a stud, however, the pin and the sensor aim show an image
where the center of the pin is too bright and distorted. Rotating the
sensor about its Y-axis so that this setting is not necessary is the first
recommendation. However, if that is not possible or practical, then use
the Notch-Angle settings to remove this segment. Increasing exposure
to extend the left and right segments may also be required when
operating with this situation.
- Threaded: For a threaded stud measurement, this requires the sensor
be aimed quite differently from other cylinder measurements. It uses
the tips of the threaded stud to make the measurement

5.15.2 Setup for Standard Cylinder Measurements


Note: For more information about threaded stud measurements using
Cylinder, see section 5.15.3.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 137


For cylinders that are specular, the sensor needs to be rotated about the Y-
axis 10-20° to suppress the intense center return. This rotation allows the
exposure to be increased, to extend the valid image data on the left and
right sections of the arc.

1. Place the Image Processing Window over the desired section


of the laser line.
2. Adjust the Exposure level to an appropriate exposure level.
3. Set the green Adjustment Box on the laser line. Use the
corner to adjust the width/height. Use the circle to rotate the box
to the slope of the laser line. The adjustment boxes will set the
following:
• Set the laser filter, in the rare event that it is needed. For this
algorithm, a horizontal slope is recommended.

Figure 5-63. Setting an unthreaded Cylinder Algorithm

4. Define the Line filter technique. Typically, the Longest Line is


recommended.

Note: Make sure that the Threaded Arc Type checkbox is not selected.

5. Notch Angles. Only enabled when the Cutout Notch mode is


selected. See the discussion above to try to eliminate the need
for this option. After the “Teach” has been clicked, the image
shows the user what segment of the arc has been "cut out". For
example, a cutout may be used when a segment of the arc could
not be aimed without the center being too bright. Adjust the two
angles. One value should equal the negative of the other. Only
rare ambient lighting conditions will make these different. Open
or close the cutout to the desired segment of the arc. The green
part of the line is used. The primary purpose of the cutout will
be to eliminate the center of a shiny pin, which is quite common

138 • Setting Up Algorithms 009-0506, Rev. B


for fixturing pins. If possible, the sensor should be rotated about
its Y-axis before using the cutout.

Note: Rotate the sensor first, to try to eliminate any reflection problem so
that Cutout Notch is not necessary.

6. Press the Teach button. The display updates with the returned
information from the algorithm; the line segment that is blue is
the CGs returned.

Figure 5-64. A taught Cylinder algorithm

Notice that the notch angles have been enabled in Figure 5-64,
using -10 and 10 degree settings. This is not appropriate for this
image, but it has been configured this way to show the colors
used for what is discarded and what is kept by the algorithm for
the final results.
7. Only when the algorithm has been configured, taught and tested,
and the Enable box is checked on the Inspection Point
Properties page, the inspection point is ready for reporting.

5.15.2.1 Interpreting the Y-Z plot for Cylinder algorithm


The Y-Z plot for Cylinder (not configured for the threaded stud) results
shows the CGs in a circular pattern, regardless of the aim of the sensor. If
the part angles, sensor angles and IJK are all correct, the result should look
like Figure 5-65.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 139


Figure 5-65. A Y-Z plot for the Cylinder algorithm.

What is shown above is the circle fit with blue points, the CGs used for the
fit in green, and the CGs that were found, but not used in the fit shown in red.
The following discussions show problems requiring corrective actions to
achieve the results shown in Figure 5-65.
If there is a calibration problem, where the part angles, sensor angles or the
IJK is not right, then the results might look like Figure 5-66, where the ends
of the circle fit does not match the CGs. Because of this incorrect
calibration, you will also notice that the reported diameter will not match the
diameter of the actual cylinder. The problem must be found and fixed before
the reported results are valid.
The points that do not match in circle in Figure 5-66 shows points that do not
match at the ends. Other mistakes can show mismatches at other points
around the circle, which means that CGs that form an ellipse instead of a
circle indicate something that must be fixed.

Figure 5-66. A Cylinder Algorithm Result with a Calibration Problem

140 • Setting Up Algorithms 009-0506, Rev. B


For the same reasons as the Edge Algorithm, the sensor must be rotated about
the y-axis to eliminate bright centers. When this is not done, you may see an
image such as shown in Figure 5-67. In this image, we see the result of a
sensor without the appropriate y-axis rotation developing a vertical stripe in
the image. Some CGs are not on the arc at the center, some CGs are
computed using a very fat laser image near the center. The lower exposure
required to control this results in a dim laser return at the ends. Rotating this
about the sensor’s y-axis will reduce the intensity at the center and allows for
increased exposure. An increased exposure will include many of the very
important CGs at each of the two ends that help the algorithm measure the
correct radius/diameter and therefore a proper center point.

Figure 5-67. An image of a cylinder where the sensor needs y-axis rotation.

Figure 5-68 shows the result of the above image. What is important is to
consider what is happening near the center, where we must consider:
- There are missing CGs at the extremely bright center.
- There area few CGs near the missing CGs that are distorted, showing
deviation from the fit.
In this example, the missing and distorted CGs should not disrupt the results
as seen in Figure 5-68 However, as the CGs are randomly placed on the
bright vertical line, it could create CGs near the fit that could yield an
incorrect answer. Rotating this sensor around it’s Y-axis is recommended to
reduce the vertical stripe and therefore the missing CGs.
For some sensors and smaller diameter cylinders, the distorted CGs may be a
bigger problem than what’s shown here. This sensor has CGs that move by a
few tenths of a mm towards the center of the arc on a 15mm cylinder because

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 141


of the thick, overly bright laser. This Y-Z plot shows that it is hardly enough
to create a problem in this example. Other sensors may have more distortion,
or will have CGs that are moved significantly away from the center.
Combining a sensor that distorts CGs with a cylinder that has a smaller
diameter, will result in inaccurate results. The solution is the same as the
solution for the missing CGs, the sensor must be rotated around its y-axis to
minimize the bright center and the vertical stripe in the image.

Figure 5-68. A Y-Z plot from an image with a bright center.

5.15.3 Threaded Stud Measurements using Cylinder


The sensor aim for a threaded stud or bolt is quite different from a Cylinder
aim measuring a cylindrical shape. The sensor must be aimed as to present
an image that crosses at least two threads, preferably many more. The sensor
must be rotated about its Y-axis to achieve the appropriate image, with little
or no rotation about the sensor’s Z-axis.

142 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-69. Sensor Aiming for a Threaded Cylinder

The rotation about the sensor’s Y-axis, as shown in Figure 5-69, must be
such that the laser creates an image of at least two threads, but 3 or 4 is
preferable. Rotating the sensor too far creates another problem, where the
laser could fall off the threads. The laser should never image the tip of the
stud, or any part of the shoulder. With a Y-axis rotation that is too extreme, a
small change in position can easily create a situation where the threads are
missed, and the algorithm will fail.

For threaded arc types, start with the laser plane rotated between 45° to
60° relative to the sensor's Y-axis. The final rotation will be indicated by
the image, the number of threads and the expected part variation.

The feature data, diameter and pitch dimensions must be correctly entered
before teaching or measuring.

Converting the number of thread crossings to rotation may be computed with


the following equation:
⎛ 2 • N • Pitch ⎞
ArcTan ⎜ ⎟
⎝ Diamter ⎠
N in the above equation is the estimated number of thread crossings. The
absolute minimum of thread crossings is 2. However a more practical
minimum is 2.5, and a better 3 threads crossed. The 3 threads crossed means
the suggested minimum rotation about sensor Y-axis:
⎛ 6 • Pitch ⎞
R = ArcTan ⎜ ⎟
⎝ Diamter ⎠
Y

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 143


For some common metric threads, the following table may be used:
Minimum Y-axis
rotation for number of
threads crossed
Metric Diameter Pitch 2.5 3
Standard (mm) (mm) crossed crossed
M6 x 1.0 6.0 1.00 40° 45°
M8 x 1.25 8.0 1.25 38° 43°
M10 x 1.5 10.0 1.50 37° 42°
Using the above equation in reverse, we have the number of thread crossings
as:
Tan (RY ) • Diameter
NumberThreadsCrossed =
2 • Pitch
An SAE bolt might be specified as ¼ - 20, which means the thread diameter
is 0.25 inches (6.35mm) and there are 20 threads per inch. Converting the
1/20th of an inch to metric yields a pitch of 25.4 / 20, or 1.27 mm. This
formula shows that 3 threads crossed requires a 50° rotation for the ¼ - 20
example.
Note that each ½ step on the number of threads crossed is significant. For
example, if the above computation shows 3.8 threads crossed, you will see 4
crossings, or 8 tips more frequently than not. However, it will not be unusual
to see only 7 thread tips, 4 on one side and 3 on the other.
1. Place the Image Processing Window over the desired section
of the laser line.
2. Adjust the Exposure level to an appropriate exposure level.

Figure 5-70. Setting a threaded Cylinder Algorithm

3. Select the Threaded Arc Type. When this option is selected,


specify the:
a. Diameter (of the stud/bolt), and
b. Pitch (distance between threads).

144 • Setting Up Algorithms 009-0506, Rev. B


Both values are in mm units.
4. Press the Teach button. The display updates with the returned
information from the algorithm; the line segment that is blue is
the CGs returned, but not used. The red points indicate the
points used to construct the answer. If the location of the red “+”
does not look like the image below, the sensor angles or the
feature vector might be wrong. Use the exposure control to
maximize the number of red dots shown.

Figure 5-71. A taught threaded Cylinder Algorithm

Note the red dots on the thread tips in Figure 5-71 as these are
used to create the final algorithm results. The blue are CGs that
are also found, but are not used in the final answer.
5. The checkpoint should be set to Active returning to the
Configuration screen. This checkpoint is now complete and
will run during a manual or automatic gauge cycle.

The repeatability of the threaded arc will be noticeably higher in Z than


other algorithms. If the rotation is 60°, the actual cylinder center
repeatability will be a factor of cos(60°), or ½ the sensor Z-axis
repeatability.

Do not expect the threaded arc measurements to be as repeatable or


accurate as other algorithms. All measurement devices, including the
TriCam sensor, cannot measure threads as accurately as other feature
types.

5.15.3.1 Interpreting the Y-Z Plot of the Threaded Cylinder


The Y-Z plot results of the Cylinder algorithm configured for a threaded stud
shows all of the CGs found in red, and the CGs used in green. Figure 5-72
shows what the Y-Z plot should look like with a proper calibration. This

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 145


shows the blue circle fit passing through the green thread tips, the desired
results.

Figure 5-72. Threaded Stud algorithm results

Figure 5-73 shows a serious calibration problem. The image is reasonable, but
the blue circle points do not match the thread tips.

Figure 5-73. Threaded Stud Results with Bad Calibration

Figure 5-73 shows the result if the IJK does not match the part, or there is a
calibration problem with the sensor orientation. For FMS systems this could
mean that one of the following problems occurred:
- The robot-to-part is not calibrated properly
- The tool-finder process failed or was not done properly

146 • Setting Up Algorithms 009-0506, Rev. B


This kind of problem will sometimes cause poor repeatability, or missed
measurements, but will certainly create an accuracy problem. It is important
to find and fix the problem before using the results.

5.16 Hole Algorithm


Feature definition: The Edge feature described in section 5.2.7.
The Hole algorithm measures the deviation of a punched, extruded (drawn),
or threaded hole. This algorithm locates the center of a hole by acquiring
two images, one using the feature lightning and the second using the surface
sensor's structured laser lighting.

The hole algorithm may measure successfully when there is a partial slug
or other obstacle in the hole. Consider using the Slug Hole or Hole
Presence algorithm to detect this kind of defect.

5.16.1 Overview of the Function of the Hole Algorithm


The following is an overview of how the hole algorithm proceeds. All axes
descriptions in the following are in reference to the sensor-case coordinates.
The hole algorithm will proceed with the following steps:
1. Measure the X and Y position of the hole using the feature image
(using the flood lighting) and the feature algorithm parameters.
The location of the hole is measured with a fitting of the ellipse
to the image. The shape of the ellipse is determined by the
feature data and sensor position.
2. The laser is used to identify the final Z-axis position of the hole.
The results from the first step are used to position the inner and
outer extent limits around the found hole. These points are the
points used in the laser image to identify this Z-axis surface
location.
3. Use the laser filter to filter out any further unwanted segments of
the laser line.
4. Combine the results of all of the above to get the 3-D
measurement.
It is important to understand the algorithm overview before proceeding to the
teach process.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 147


5.16.2 Parameters for the Feature Image

Figure 5-74. Hole Parameters

Important: If the feature data and sensor orientation are not properly
defined, the teach process cannot be completed. This information will be
used to define the shape of the hole as it appears in the sensor's
perspective. Sensor orientation may be derived from the any of the many
processes to calibrate the sensor orientation. If this information is
incorrect, the displayed ellipse will not match the image.

148 • Setting Up Algorithms 009-0506, Rev. B


5.16.3 Teaching the Hole Algorithm
1. Lighting of the feature image is extremely important for the
proper functioning of the algorithm. See chapter 2. for lighting
details. For front-lit holes, make sure that there is no severe
difference in light around the edges of the hole. Scratches,
dings, uneven metal or uneven lighting can cause changes in
appearance, and will likely require further lighting adjustment.
2. Set the exposure for the feature image. This needs to be set such
that the dynamic range of the light and dark edges around the
hole to be measured is at the best contrast possible. If adequate
dynamic range cannot be achieved by setting exposure, then
adjust the light source positions or sensor positions until it is
possible.

Note: For front-lit drawn holes, over exposing the surface may be required
for proper contrast of the edges near the bottom of the draw radius. Your
attention should be focused on the contrast at the bottom of the hole when
setting the exposure.

3. Select the Hole Type: Dark (light edges surrounding a dark


inside, for a typical front lit hole) or Light (dark edges
surrounding a bright inside, for a backlit hole).
4. Select the Search from: The search direction is used to indicate
where the search should start which must always specify the area
with the lesser amount of image noise. For the front lit case
(Dark hole type), select Inside when the hole center has less
noise than the surrounding area. Scratches, dings, bumps or
uneven lighting can produce image noise on the surrounding
area. Select Outside when the area the hole is relatively clear of
image noise and the hole center is more noisy. The backlit hole
which has the bright center, the typical selection of this
parameter is to search from the outside because this is usually
quite dark. Only the rare situations with strong ambient light
would change this recommendation for backlit holes.
5. Select the Edge Type: Select Sharp to indicate short light-to-
dark transition around the edges of the hole. Select Front-Lit
Extruded (used on holes that have a drawn, or extruded, radius
around the edges and lighting that is on the same side as the
sensor).

Note: The important concept of measuring an extruded (drawn) hole is


that the measurement should be as close to the bottom of the draw radius
as possible. This means that using front lighting on extruded hole, the
center of the hole must be as dark as possible, and the edges must be as
well lit as deep into the hole as possible.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 149


Using the Front-Lit Extruded hole option with a light hole type is not a
valid combination. A light hole selection means that the hole has been
back-lit.

Note: Any image noise in the center of the front-lit extruded hole is not
acceptable.

6. Position the template over the hole. Make sure that the position
and the size of the template matches the image very accurately.
7. Select the Cutouts: None (disables the cut-out option), One
(allows the user to eliminate a single wedge/slice of the image),
or Two (allows the user to eliminate two wedges/slices of the
image). Cutouts may be used to eliminate any edges that are not
appropriate for the measurement. This can include scratches,
deformed metal that is consistently wrong on one side, or
winking metal. The cutouts cannot be so big as to eliminate
more than 180° of the edge. Although cutouts may be used for
eliminating sections of bad lighting, the recommended practice is
to fix the sensor aim and lighting so that the cutout is not needed.
8. Select the Edge Filter: Outlier Elimination or Use All Points.
The Outlier Elimination feature will identify the largest outlier
in the circle fit. If it is too much of an outlier, it eliminates this
point, tries another fit, and looks for the next outlier point. This
process is repeated until the worst outlier is acceptable, or there
are too few points remaining to continue. Use All Points should
be used only when the user has a very clean feature image. This
option reduces cycle time, but fits the best circle to points
without removing any points that do not fit well with the circle.
Using the Outlier Elimination enhances accuracy for most
situations.
9. Select Hole Option: Complete Hole in Window or Hole
Partially in Window. Complete Hole in Window feature
enhances the robustness of the algorithm, avoiding other objects
that may be mistaken for the desired object, however if any part
of the hole leaves the image-processing window, this will result
in a no measure result. Hole Partially in Window must be used
when it is possible to have a hole that may have some amount of
edge leave the window in normal process variation, yet still have
enough of the hole for accurate results.
10. Select the number of points sampled to find a hole. (16, 32, 48
or 64) The user selects the desired number by selecting the radio
button. More points will enhance accuracy and robustness at the
expense of processing time.
11. There are two image windows on the screen: a Feature and
Laser image. The Feature image shows the setting of the hole
using the flood lighting or back lighting, while the Laser image
displays the laser line across the part.

150 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-75. Setting up a Hole Algorithm

In the above figure, the hole template is correctly sized and


positioned. This means that the setup of the feature definition
and the sensor orientation is most likely correct. See section
5.2.7 if the ellipse in the image does not match the ellipse drawn.
You have control of the overall size of the template and its
position, but other factors of the drawn ellipse are a result of the
computation of the sensor orientation and the entered feature
data.
Note also the fact that the drawing of the laser parameters shown
in the right image of the above figure does not make sense until
the feature image is correctly taught.

Figure 5-76. Example of bad hole template placement (outside edges not touching template)

12. Figure 5-76 shows an example of a bad template match. This


image shows that the height is slightly large, and the width is too
small. One can improve this by adjusting the template size

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 151


further, but once the problem is recognized, finding and fixing
the problem with the feature data and sensor angles is a strongly
recommended. What is shown is approximately a 20° error.
Now use the Teach button. If the template is correctly defined
and matches the image, the teach action will move the template
exactly over the hole, and the size and placement of the laser
extents in the laser image will now be correct. The correct
placement of the laser extents is required to be correct for the
next steps.

Figure 5-77. A taught Hole Algorithm

Section 5.16.4 describes how to interpret the critical red and blue
dots. You must examine the results carefully for the measured
dots as this indicates many important concepts. The green dots
shown in Figure 5-77 are the inner and outer dots showing the
search extents to find the hole edges. For example, if the Search
From mode is “Inside”, the first edge found looking from an
inner green dot to the outer green dot. The first edge found
between the green dots shown as the red dot is one of the edge
samples used to measure the hole.
The image in Figure 5-77 should be set for search from inside to
out because of the possibility of finding inappropriate edges
around this hole if the Search From was set to “Outside”, where
there is more image noise on the lit surface.
13. Again, now that the teach has located the template into the found
position, make sure that the template matches the hole edges and
resize the template until it is exact. Repeat step 11 if any
adjustments are done. Make certain that this is done without
errors before proceeding. If the shape of the hole does not match
the template, review the feature data and sensor orientation.

Note: You must have successfully taught the feature image before
proceeding on to teaching the laser. Without proper placement of the
feature, the display graphics on the laser image will not be properly

152 • Setting Up Algorithms 009-0506, Rev. B


located. If the teach failed to locate the hole, keep working the setup
before proceeding any further.

Examine the teach results carefully at this point. The red points
are edge points found and used. Yellow dots are those points
found but rejected by the algorithm. If the yellow dots are
rejected for good cause, such as a found edge that is not
appropriate for the measurement, this is exactly what the
algorithm is supposed to do. If there are yellow dots at points
that should have been kept, this indicates a probable setup
problem. Similarly, red dots on points that should not have been
accepted represent a problem.
14. Now that the feature has been taught, the laser image must now
be taught. First, adjust the Exposure level in the laser image on
the right side to an appropriate exposure level.
15. The Inner and Outer Extents are used to isolate the area around
the hole that the laser light will be used to measure the surface
location. No information that is not in between the inner and the
outer extents is used for the algorithm. As the teach process will
properly locate the extents over the hole, make sure that this is
successfully completed before trying to adjust the extents. Use
the arrows to size the extents over the segments of the laser to be
measured. The extents and the image-processing window will be
the first filter applied, and will use the surface around the hole to
be measured. Make certain that these are positioned
appropriately to filter out any unwanted parts of the laser line.
On the image shown, there is a drawn hole located on metal
behind a surface, all of which is inappropriate for the sensor Z
measurement. The inner extent must be set large enough to
exclude any part of the laser located on the draw radius.

Figure 5-78. Setting up the Laser image after the Feature image has been successfully taught

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 153


16. Adjust the slope of the laser filter carefully to match the laser
line that exists between the inner and outer extents. As this is
used to join the useful parts of the laser separate from any other
noise, the slope must be exactly matching these two segments
that are typically found on the left and right side of the hole. In
order to set this properly, the laser filter must be expanded to at
least the width of the outer extents. If there is any light that
exists beyond the expected laser lines, you must narrow this filter
to exclude these segments.
17. Only when the algorithm has been configured, taught and tested,
and the Enable box is checked on the Inspection Point Properties
page, the inspection point is ready for reporting.

5.16.4 Interpreting the Hole Teach Display


After teaching, the results display is critically important to understand what is
shown. For a perfect hole, the following drawing shows what that might
look like.

Figure 5-79. Perfect hole

If the sensor orientation or the feature vector is not exact, Figure 5-80 shows
what this might look like.

154 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-80. Teach Display with Incorrect Sensor-to-Feature Information
(Note the template mismatch and the yellow dots on the right side of the hole.)

Two factors show that there is a problem in Figure 5-80. First is the green
template, which does not correctly match. Second are the yellow dots on the
right side of the hole. With the Outlier Elimination option turned on, these
yellow dots indicate that these points were rejected in the best-fit operation,
as the elliptical template does not match the image. Most of the dots around
the edges are red, indicating that this result might be reasonable. At best, this
means the points on the right side are always rejected. It is more likely that
the right and left sides will alternate being rejected, creating a bi-modal
result, destroying repeatability. As the size of the hole appears different with
part variation, the top, bottom, left or right will alternate the rejection
process, which will not be accurate, even if repeatability tests succeed.
The root-cause of the problem must be found and fixed before proceeding.

Figure 5-81.,Teach Display with Bad Sensor-to-Feature Information


(Note green dot and yellow dot placement.)

If the sensor orientation or the feature vector is even further off than that
shown in Figure 5-80, which is illustrated in Figure 5-81, all of the problems
can be much worse. If it is worse than this, the teach function will fail. In
the worst case extreme, the ellipse will shrink down to a line.
The following two figures illustrate the common less-than-perfect hole
image. The dots shown on the teach display must be used to help make the
appropriate decision about using cutouts, or depending on the algorithm to
automatically decide what is bad.
Figure 5-82 illustrates a perfect hole with a section that is bad where no
cutout is used. In this example, the outlier elimination process has selected
two of the four worst blunders, shown in yellow, in the lower-left quadrant.
The yellow indicates that these points have been appropriately removed from

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 155


the final construction of the answer. As mentioned before, only the red
points are used to construct the best-fit answer, in this case, 30 of the 32.

Figure 5-82. Hole with Bad Section

Figure 5-83 shows a single cutout, which is used to permanently remove the
bad section from the measurement. In this case, this reduces need to rely on
using outlier elimination. With this cutout, more usable samples are
available, enhancing accuracy and reliability.

Figure 5-83. Using a Cutout to Remove a Problem

Figure 5-83 illustrates how the cutout has moved the measured points away
from the problem area, relocating them to the expected good region around
the hole.
Note that that all of the edge points in Figure 5-83 are now all red, indicating
that they are used in the answer and the bad segment is no longer being
considered. Be sure to never cut out more than 180 degrees, as the results
cannot be accurate. If there are two cutouts, the sum of the two should never

156 • Setting Up Algorithms 009-0506, Rev. B


exceed 180 degrees, and it is recommended that this sum be less than 160
degrees.

Note: Cutouts for holes that are off-axis must be verified by the taught
result or the Show Algorithm result. If the view is about 20 degrees or
more from normal, the cutout display distortion is severe enough to require
the validation of the measured points, which will not match the user
interface display.
Be very sure the size of the template matches the image exactly, as
illustrated in Figure 4 79. What is shown in this figure is a perfectly
circular hole, where some images may be an ellipse. The template should
also match the elliptical shaped image exactly. A template that does not
match the hole size and shape will create accuracy and reliability problems.
The teach page or the Show Algorithm option may show random missing
points around the edges with an incorrectly sized or shaped template.
Always perform the teach step after any change to the feature or the tool
position. See section 5.5.1 for details.

5.16.5 Interpreting the Hole Show-Algorithm Display


The Show Algorithm view must be examined for at least 10 cycles, and even
more is recommended. Alternatively, you can do the same by poking the
Teach button 10 times, but this is slow, and frequently not recommended on
a correlated system. The idea is to examine the image for the inappropriate
placement of the green dots and inappropriate use of the yellow dots. All this
test can do is show significant problems. Although these problems usually
show up in the repeatability values, and it is only with one part in one
location, it is still important to examine this display carefully as this can have
a significant impact on accuracy.

5.17 Slot Algorithm


Feature definition: The Slot feature described in section 5.2.10.
The Slot algorithm measures the deviation of a punched or drawn (extruded)
slot. The slot must be a standard shape, either with two half-circles at each
end and straight sides, or a rectangle with all 4 sides straight. Just like the
Hole algorithm, the sensor uses both the surface sensor's structured laser
lighting and feature lightning, to obtain the three degrees of freedom result.
For more complex forms that do not fit the simple slot description above, or
one or two of the sides of the slot cannot be used, use the Complex Slot
algorithm.

Note: The Slot algorithm using the semicircular ends assumes that the slot
is a simple standard slot, where the end radii are exactly ½ of the slot
width. If the slot is not of this description, or a simple rectangular shape,
the Complex Slot algorithm is recommended.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 157


The slot algorithm may measure successfully when there is a partial slug
or other obstacle in the slot. Consider using the Slug Slot or Slot Presence
algorithm to detect this kind of defect.

5.17.1 Overview of the Function of the Slot Algorithm


The overview of the slot is exactly the same as the hole. Refer to Section
5.16.1.
1. Lighting of the feature image is extremely important for the
proper functioning of the algorithm. See Chapter 2. for lighting
details. For front-lit slots, make sure that there is no severe
difference in light around the edges of the slot. Scratches, dings,
uneven metal or uneven lighting can cause changes in
appearance, and will likely create a need for further lighting
adjustment.
2. Set the exposure for the feature image. This needs to be set such
that the dynamic range of the light-to-dark edges around the slot
to be measured is at the best contrast possible. If adequate
dynamic range cannot be achieved by setting exposure, then
adjust the light source positions or sensor positions until it is
possible

Note: For front-lit drawn slots, over exposing the surface may be required
for proper contrast of the edges near the bottom of the draw radius. Your
attention should be focused on the bottom of the slot when setting the
exposure. It must be as dark as possible.

3. Select the Slot Type: Dark (edges that are light against a dark
inside, for a typical front-lit slot) or Light (edges that are dark
against a light inside, for a back-lit slot).
4. Select the Search from: The search direction is used to indicate
where the search should start which must always specify the
clearer part of the slot to be measured. For the front lit case
(Dark slot type), select Inside when the slot center is very clean,
but there are scratches, dings, or bumps on the surrounding area
or Outside when the outside edges are defined, and the surface
surrounding the slot is relatively clear of blemishes, scratches or
other noise.
5. Select the Edge: Sharp (when a rolled edge is very small as
compared to the sensor's field of view) or Front-Lit Extruded
(used on slots that have a drawn, or extruded, radius around the
edges and lighting that is on the same side as the sensor).

Note: The important concept of measuring an extruded (drawn) slot is that


the measurement should be as close to the bottom of the draw radius as
possible. This means that using front lighting on extruded slot, the center
of the slot must be as dark as possible, and the edges must be as well lit as
deep into the slot as possible.

158 • Setting Up Algorithms 009-0506, Rev. B


Important: Any image noise in the center of the extruded slot is not
acceptable.

6. Position the template over the slot. Make sure that the position
and the size of the template matches the image very accurately.
The template may be rotated to match the image. The
orientation of the center line bar in the middle is not important.
Figure 5-84 shows the line across the short length, but this may
be turned 90 degrees. The functioning of the algorithm will
function the same regardless of the orientation of this line.

Figure 5-84. Setting up of a Slot Algorithm

7. Press the Teach button. The display updates with the returned
information from the algorithm; the line segment which is blue
are the CGs returned, the green intersecting line is the surface
normal along which the measurement is taken.

Figure 5-85. A taught Slot Algorithm

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 159


8. Note that the resulting display in Figure 5-85 has a few blue dots
on the right side of the image. This shows that the blue dots
were rejected, leaving only the red dots to construct the final
answer. This may be the result of a part that has an unusual
defect, contrast problems, or a misaligned template.
In this specific image, the important problem that needs
correction is the image contrast, showing some dark regions due
to some specular areas around the right edge that are dark with
this lighting. There is also some light coming from the metal
behind the slot, which combined with the dark edge, creates
almost no contrast at the real edge. If this is just an unusual part
defect, then, and only then is the current setup acceptable. If
lighting improvements are not possible, or this is a normal part
defect, then consider using the Complex Slot algorithm with
better controls, but a more difficult-to-use interface. Refer to
Chapter 2. or more details about proper lighting.
The blue dots in the above image must be understood before
moving on to the next steps. In this example, extra hardware is
required to move the stereo lighting further away from the
sensor. If the repeatability of the above image is achievable, it is
quite possible to see accuracy problems in production because of
the lack of contrast. Having a small amount of template rotation
that does not match the part is another reason for seeing extra
blue dots. The adjustment is extremely sensitive, however,
taking a few moments now to be confident of the measurement
results is time well invested!
9. Poke the teach button a few times to be sure that there are no
blue dots as shown in Figure 5-85, or at least the appropriate dots
are rejected.
10. Now that the feature has been taught, the laser image must now
be taught. First, adjust the Exposure level in the laser image on
the right side to an appropriate exposure level.
11. The Inner and Outer Extents and the image-processing window
are used to isolate the area around the slot that the laser light will
use to measure the surface. Any information that is not in
between the inner extent and the outer extent and the window is
not used for the algorithm. As the teach process will properly
locate the extents over the slot, make sure that this is
successfully completed before trying to adjust the extents. Use
the arrows to size the extents over the segments of the laser to be
measured. The extents and the image-processing window will be
the first filters applied to locate the surface around the slot.
Make certain that these are positioned appropriately to filter out
any unwanted parts of the laser line. In Figure 5-86, there is a
drawn slot with some metal to the right that is inappropriate for
measuring the sensor z-axis location of the slot. The inner extent
must be set large enough to exclude any part of the laser located

160 • Setting Up Algorithms 009-0506, Rev. B


on the draw radius and the outer extent must be small enough to
exclude the metal on the right of the image.
12. Adjust the laser filter to match the laser line that exists between
the inner and outer extents. As this is used to join the useful
parts of the laser separate from any other noise, the slope must be
exactly matching the segments that are typically found on the left
and right side of the slot. In order to set this properly, the laser
filter must be expanded to at least the width of the outer extents
or wider.
If one of the two segments cannot be used, using the process
window to eliminate the incorrect segment is the best option. If
the process window cannot be used to eliminate the incorrect
segment and the incorrect segment has a different slope, then the
laser filter is the only remaining option.

Figure 5-86. Laser Setup

13. Only when the algorithm has been configured, taught and tested,
and the Enable box is checked on the Inspection Point Properties
page, the inspection point is ready for reporting.

Always perform the teach step after any change to the feature or the tool
position. See section 5.5.1 for details..

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 161


5.18 Complex Slot Algorithm
Feature definition: The Slot feature described in section 5.2.10.
The Complex Slot can be used for more complex shapes than the Hole or
Slot algorithms can process. These shapes include:
- A slot with only one curved side (D-shaped),
- A regular polygon, including 3-, 4-, 6- or 8-sided figures, or
- A slot where only two or three sides can/should be measured. Figure
2-14 shows such an image where the sensor and lighting is less than
optimal.
This allows the user to enable and define each of the four segments
separately, curved or straight (blue lines on the algorithm teach screen). This
is done by dragging each line (Line A, B, C or D) and sizing accordingly.

The rules for teaching the Complex Slot in Vector are completely different
than the rules for IPNet. Be sure to understand these differences.

Complex slot may be used on simple slots that have substantial visual
defects, however should only be used when all other attempts to make the
simple slot to function have failed. Reasons for abandoning the simple slot
include:
1) There may not be enough control on the simple slot algorithm parameters
to create an accurate measurement. This could include the need to move
the slot side measurement to collect accurate results.
2) One or two of the four sides is not visible or cannot be used reliably,
3) The shape of the slot is not 4-sided and does not match the simple slot.
Complex Slot is not an appropriate choice for the following situations:
1) If the shape is a circular hole. The best-fit algorithm is not designed to
properly process circular and elliptical shapes in the image. Although
the algorithm can be repeatable, it will not be accurate.
2) For very short slots where the straight sides cannot be used.
3) Measuring a long, deformed slot with only the two rounded ends. This is
only valid with perfectly formed slots with an excellent image of these
two ends. The two straight sides should be used whenever possible. A
slot feature is a very important process tool where the straight sides are
used for fixturing pins. It is rare for the curved ends to be used.
4) If you are choosing the complex slot as a means to offset proper lighting,
or other correctable problems, this is a poor reason to choose the
complex slot. Proper tuning of a Complex Slot requires extra work
beyond the simple slot, and the root-cause problems must be fixed,
making the selection of Complex Slot a poor work-around strategy.

The Complex Slot algorithm may measure successfully when there is a


partial slug or other obstacle in the slot. Consider using the Slug Slot or

162 • Setting Up Algorithms 009-0506, Rev. B


Slot Presence algorithm to detect this kind of defect. The algorithm is even
more likely to measure with 1 or two sides marked as “Not Used”.

When choosing less than four sides for the standard slot with “Not Used
option, the measured location will change with the configuration of the sides.

Figure 5-87. Actual locations with the different sides that are “Not Used” shown in light gray

Figure 5-87 shows the different location of the actual (see glossary) with
different configurations of sides that are not used. The nominal should be
modified to the location of the center of the red indicator. If it is not, offsets
may also be used.
The teach process also displays this result, and will be illustrated with a red
“+” mark in the image defining the location of the actual.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 163


Figure 5-88. Complex Slot parameters

1. Select a line segment. An active line segment will appear with


gray boxes on each end of the color-coded line matching the
color of the text in Figure 5-88. A text indicator will appear next
to the Teach button informing the user which line is active.
2. If a Curved line is needed, click on the pull-down arrow and
select Curved. This will add a third gray box in the center of
the blue line. Place the cursor over this box and drag until the
desired shape is created.
For a two or three-sided shape, deactivate a line segment by
selecting the appropriate Not Used button(s). This will cause
the selected line segment to disappear.
3. Repeat each line segment until the template is set.
4. Place the window around the slot template.
5. Select the remaining Complex Slot parameters:
a. Slot Type: Light (edges that are dark against a light
background, typically a backlit hole/slot) or Dark (edges that
are light against a dark background).
b. Search From: Inside (when the hole/slot center is very
dark for front-lit image and clean, but there are scratches,
dings, or bumps on the surrounding area that create image
noise outside the feature) or Outside (the area outside the
feature are relatively free of image noise compared to the

164 • Setting Up Algorithms 009-0506, Rev. B


inside area). More often than not, backlit images should use
the Outside option.
c. Edge Type: Sharp (when a rolled edge is very small as
compared to the sensor's field of view) or Front-Lit
Extruded (used on holes/slots that have a drawn, or
extruded, radius around the edges).
d. Slot Option: Complete Slot in Window or Slot Partially in
Window. Complete Slot in Window feature enhances the
robustness of the algorithm, avoiding other objects that may
be mistaken for the desired object. Slot Partially in
Window must be used when only a partial hole/slot is in the
field of view. This occurs if the hole/slot moves out of the
selected sensor window or there are random edges inside the
hole/slot, or only part of the hole/slot is visible.
e. Select the appropriate Number of Points from the list and
define Min. Points by typing in the minimum number of
points. Increasing the Number of Points improve the result,
but require more time to process the image. Increasing the
Min. Points may cause more missed measurements, but
reduces the chances for measurement blunders. For a worst-
case example, if one configures a 180° cutout, and only
requires 16 of 32 points, it is possible to report a result using
only 90° of an arc. Measuring a hole with an arc less than a
160° sample can produce many measurement blunders. It is
almost certain that a 90° arc will produce measurement
blunders.

Figure 5-89. Template set for Complex Slot

6. There are two windows on the screen: a Feature and Laser


window. The Feature window defines the area used to find the
slot center, while the Laser window defines the area used to
find the laser line segments for the second phase of the
algorithm.
7. Position the line segments over the feature carefully to match the
parts of the image that define the measurement.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 165


The Complex Slot in IPNet made important assumptions about adjacent
and opposite sides in the configuration. The IPNet rules are not the same
for Vector. The rule that now applies is that the segments must make a
closed form polygon with the individual line segments. The “Not used”
segments must also be part of the closed form structure.
8. Press the Teach button. This will move the position of the 8-
sided figure that is used for the inner and outer extents around
the feature. These extents are used to try to maintain a consistent
distance away from the feature to measure the laser surface
samples.
9. On the Laser window, use the Inner and Outer Extent arrows
to position the template over the object in the image. The red
processing window may be used to further control which parts of
the laser may be used. Although not recommended, it is
sometimes necessary to use the window to eliminate one side of
the laser or the other.
10. The red dots and blue dots mean the same for Complex Slot as
they do in the simple Slot algorithm. For the same reasons, you
must tune up the sides to maximize the red dots, indicating that
they are used in the algorithm. Note that Figure 5-90 needs to
have two of the sides modified slightly. You must carefully tune
each segment separately as to minimize the number of blue dots,
indicating the maximum measured segment is used to create the
best-fit final answer.

Figure 5-90. Taught Complex Slot algorithm

11. Note where the red “+” is located. The gray-scale image on the
left shows this point at the bottom. This is the measured result.
In Figure 5-90, notice that this point is located at the center of
the bottom line, not the center. The XYZ nominal point should
be adjusted to this location.
12. Now that the feature has been taught, the laser image must now
be taught. First, adjust the Exposure level in the laser image on
the right side to an appropriate exposure level.

166 • Setting Up Algorithms 009-0506, Rev. B


13. Set the laser extents and laser filter just as it is done for the Hole
algorithm. This is done exactly as it is for the Hole and simple
Slot algorithms. See section 5.16.3, steps 11 and 12.
14. Only when the algorithm has been configured, taught and tested,
and the Enable box is checked on the Inspection Point Properties
page, the inspection point is ready for reporting.

Note: These must be validated with the Teach or the Show Algorithm
functions. Make very sure that you understand the significance of the dot
colors when teaching this algorithm.

Using short segments on straight sides is sometimes used to get


repeatability. This may not be accurate, and should be avoided. There
have been many failures with this configuration, frequently highlighted
with a CMM comparison. A short straight segment is never recommended
where only 2 of 4 sides are used on a standard slot.

Always perform the teach step after any change to the feature or the tool
position. See section 5.5.1 for details.

5.19 Gap and Flush Algorithm


Feature definition: The Gap and Flush feature described in section 5.2.11.
The Gap and Flush algorithm documentation isn’t ready yet.
See the AutoFit Getting Started manual (009-0483) for information on Gap
and Flush.

5.20 Hole Presence Algorithm


Feature definition: The Hole feature described in section 5.2.7.
This algorithm does not return measurement results, and is only intended for
inspection. A Surface (or Feature) sensor is used; no laser line processing is
required. Sensor orientation and IJK vectors must be set properly; therefore
VeriStar or TetraStar is suggested if all other sensors use this calibration
technique.
Used for detecting:
• Changing areas: the increasing/decreasing diameter of a hole
returns a value of 0 to 1, e.g.: a nut welded onto the back of a
punched hole that can partially block the hole. The area is
computed using the measured radius, not the hole center size.
• Material Left Over - checks the background of the hole, i.e. the
amount of light inside the hole. Flaws of this type can be due to
a punch not completely punching through material, leaving the
slug in the center still attached.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 167


• Punch or die wears out - affects the background area, edge
points are affected. The algorithm reports rough edges on the
hole segment used for measuring as the shape parameter.
• Missing hole – due to a broken punch.
There are 4 parameters evaluated by this algorithm, all of which are in the
range of 0.00 to 1.00.

All parameters are between 0 and 1. Lower values closer to 0 is for a


stronger presence. Higher values closer to 1 indicate absence of the hole.

The four options used for changing areas of a hole, are:


• Edge Points – increases as the number of points returned on the
measured hole edges decreases.
• Hole Center Background - as light increases in a hole with a
dark center, this value goes closer to 1, commonly due to the
presence of the slug.
• Shape – to the extent that the edge points do not form a perfect
circle on the part, the shape factor increases. This can be thought
of as the edges not closely matching the expected circular shape.
Note that any lighting defects are not good for this parameter.
• Area - computed from actual radius and nominal radius
differences.

168 • Setting Up Algorithms 009-0506, Rev. B


Figure 5-91. Hole Presence parameters

Setup Hole Presence Parameters:


1. Select the Hole Type: Light (edges that are dark against a light
background, typically a backlit hole) or Dark (edges that are
light against a dark background).
2. Select the Search: Inside (when the hole center is very dark and
clean, but there are scratches, dings, or bumps on the
surrounding area) or Outside (the outside edges are defined, and
the surface surrounding the hole is relatively clear of blemishes
and noise).
3. Select the Edge: Sharp (when a rolled edge is very small as
compared to the sensor's field of view) or Front-Lit Extruded

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 169


(used on holes that have a drawn, or extruded, radius around the
edges). Most extruded/drawn holes or slots are made with a
process that makes it difficult or impossible to leave a slug.
Therefore the “Front-Lit Extruded” selection is not a normally
used configuration with this algorithm.
4. Select the Cutouts: None (disables the cut-out option), One
(allows the user to eliminate a single wedge/slice of the image),
or Two (allows the user to eliminate two wedges/slices of the
image).
5. Adjust the Exposure level to an appropriate exposure level.
6. Select the Total Points per Hole, which will be the number of
edge samples taken. The more points that are measured will
increase the cycle time.
7. Select the Minimum Points per Hole, which defaults to half the
total points per hole selected. If the number of points that are
measured and kept after filtering is lower are fewer than this
minimum, the result is a grey (no-measure) marker for the
inspection point in question.
8. Press the Teach button.

Figure 5-92. A taught Hole Presence algorithm

9. Decide which of the four variables are most important to the


inspection procedure. This can be accomplished by using
representative targets of good and bad samples and reviewing the
feedback results. Read and understand the equations that are at
the end of this section.
10. Adjust the weight factors as appropriate to achieve the desired
results.

Note: The total of all weight factors must equal 100!

11. Press the Test link, to go to the Test screen. As this is not a
measuring algorithm, there is only a test result. This is not
useable information.

170 • Setting Up Algorithms 009-0506, Rev. B


12. Select the Limits and Alarms link – these limits and alarms are
only applied to the final “Result”.
13. To verify results, go to Checkpoint Analysis - X-Bar and R or
the Pareto report for the performances.
This algorithm does not measure the feature deviation and is designed for
inspection use only! However, sensor angles and feature IJK are
important and must be correct for the algorithm to process the image.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 171


5.20.1 Hole Presence Result Computations
The overall result for this algorithm is computed using the following
equations:
HolePresenceResult =
EdgeWeightFactor% * NumberEdgePointsParameter
+ HoleCenterFactor% * HoleCenterBackgroundParameter
+ ShapeFactor% * ShapeParameter
+ AreaChangeFactor% * AreaParameter

The following equations describe the four parameter values used above:

NumberEdgePointParameter =
NumberEdge PntsUsed
1 .0 −
TotalPntsForHole
where TotalPntsForHole is 16, 32, 48 or 64.

HoleCenterBackGroundParameter =
AverageGrayLevelInsideHole
MaxGrayLevel
for front-lit holes with dark centers, or

HoleCenterBackGroundParameter =
AverageGrayLevelInsideHole
1.0 −
MaxGrayLevel
for back-lit holes. MaxGrayLevel is always 255.

ShapeParameter =
SixSigmaOf EdgePntsUs edInFit
ActualRadi us

AreaChangeParameter =
ActualRadius 2 − NomRadius 2
±
NomRadius 2
where the sign is plus (+) for “Increasing”, and minus (−) “Decreasing”
selection. The NomRadius value is defined by the hole size as it appears
with the taught image. As with other parameters, the AreaChangeParameter
value is limited to the range of 0 to 1. Any negative result is forced to zero.

172 • Setting Up Algorithms 009-0506, Rev. B


5.21 Slot Presence Algorithm
This algorithm does not return measurement results, and is only for
inspection. A Surface (or Feature) sensor is used; no laser line processing is
required. Sensor orientation and IJK vectors must be set properly; therefore
VeriStar or TetraStar is suggested if all other sensors use this calibration.
Used for detecting:
• Changing areas: the increasing/decreasing diameter of a slot
returns a value of 0 to 1, e.g.: a nut welded onto the back of a
punched slot that can partially block the slot.
• Material Left Over - checks the background of the slot, the
amount of light inside the slot. This can be due to a punch not
completely punching through material or a broken punch.
• Punch wears out - affects the background area, edge points are
affected.
• Missing slot – due to a broken punch.
As in the Hole Presence algorithm, there are 4 parameters evaluated by this
algorithm, all of which are in the range of 0.00 to 1.00. The lower values
mean that the hole is present, and the higher values indicate absence
conditions.
The four options used for changing areas of a slot, are:
• Edge Points – increases as the number of points returned
decreases.
• Slot Center Background – as light increases, this value goes
closer to 1.
• Shape – a perfect slot shape = 0.00; as the image distorts, the
shape factor increases.
• Area – As the area of the slot changes from the original value,
this value increases. Areas that are exactly nominal are 0.00.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 173


Figure 5-93. Slot Presence parameters

Setup Slot Presence Parameters:


1. Select the Slot Type: Light (edges that are dark against a light
background, typically a backlit slot) or Dark (edges that are light
against a dark background).
2. Select the Search: Inside (when the slot center is very dark and
clean, but there are scratches, dings, or bumps on the
surrounding area) or Outside (the outside edges are defined, and
the surface surrounding the slot is relatively clear of blemishes
and noise).
3. Select the Edge: Sharp (when a rolled edge is very small as
compared to the sensor's field of view) or Extruded (used on
slots that have a drawn, or extruded, radius around the edges).

174 • Setting Up Algorithms 009-0506, Rev. B


4. Select the Slot Shape: Straight or Curved (defines the shape of
the slot edges).
5. Adjust the Exposure level to an appropriate exposure level.
6. Check the corresponding radio button to select the desired Total
Points for Slot.
7. Press the Teach button.

Figure 5-94. A taught Slot Presence algorithm

8. Decide which of the four variables are most important to the


inspection procedure. This can be accomplished by using
representative targets of good and bad samples and reviewing the
feedback results.
9. Adjust the weight factors as appropriate to achieve the desired
results.

Note: The total of all weight factors must equal 100!

10. Press the Test link, to go to the Test screen. As this is not a
measuring algorithm, there is only a test result. This is not
useable information.
11. Select the Limits and Alarms link - value results returned.
12. To verify results, go to Checkpoint Analysis - X-Bar and R, or
to the Pareto report for the performances.
This algorithm does not measure! It is used for inspection only. However,
sensor angles and feature IJK are important and must be correct.
The equations for the Slot Presence algorithm are exactly the same as the
Hole Presence algorithm. The radius used is the radius of either end of the
slot, all other concepts are the same.

5.22 Slug Hole Algorithm


Feature definition: The Hole feature described in section 5.2.7.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 175


The Slug Hole algorithm returns a measurement and a presence/absence
result. The hole algorithm is robust enough to frequently create a valid
measurement even with the slug left in the hole. The Slug Hole is designed
to fail when the hole is not completely clear.
Most extruded/drawn holes do not have such a defect, therefore the “Front-
Lit Extruded” option would not be an expected option using Slug Hole.

Figure 5-95. Slug Hole Parameters

Slug Hole has the same concepts as the Hole and the Hole Presence
algorithms. See sections 5.16 and 5.20 for details.
The Hole Presence Threshold parameter allows the user to define which
holes should be considered to have a no-measure condition. When this
threshold is exceeded, the checkpoint shows an error 46596 “PRESENCE
VALUE EXCEEDS THRESHOLD”. This means that the X, Y, Z, diameter
or presence value is not reported, and will not be part of the SPC calculations
in the reporting.

176 • Setting Up Algorithms 009-0506, Rev. B


5.23 Slug Slot Algorithm
Feature definition: The Slot feature described in section 5.2.10.
The Slug Slot algorithm behaves exactly like the Slug Hole algorithm on slot
measurements for the same reasons.
Teaching the Slug Slot algorithm is exactly the same process as teaching the
Complex Slot and the Slot Presence algorithms. Slug Slot uses the Complex
Slot algorithm setup. If a simple slot is required, a structure of 2 round and
two straight ends must be built. See 5.18, 5.20 and 5.21 for details.

5.24 Sphere Algorithm


Feature definition: The Sphere feature described in section 5.2.12. The
diameter is defined in the feature and is critically important to be accurate for
this algorithm to function properly.
This algorithm has a unique teaching concept on the “Aiming Aid” tab. This
will show the measurement result of the diameter of the sphere as sectioned
by the laser plane in real time. These measurements are shown on a graphic
that also shows the lower and upper limits appropriate fro a valid 3D
measurement result of the sphere center. A measurement that is close to the
equator must be avoided, and so must a measurement too close to the pole.
The Aiming Aid will show the current measurement on a scale that has the
upper and lower limits and the recommended diameter on the sphere.
The laser must section the sphere such that the center is on the opposite side

5.25 Inspection Point Test Measure


In order to use this option, the user must have completely and properly setup
the Checkpoint Teach screen. This is navigated through the Model,
Checkpoint List, Checkpoint, Teach menu options.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 177


Figure 5-96. Test Measure screen

On the Test Measure screen:


1. Select the number of Cycles.
2. Choose the “Show algorithm” if desired. See note below.
3. Press the Start Test button (the Stop Test button when running).
4. Verify that the number of cycles run matches the desired number
of cycles. If it is less, an error occurred and the teach parameters
should be adjusted again and the checkpoint re-calibrated.

Note: The user may want to select the check box to show the algorithm.
This will show the image, some of the parameters used by the algorithm to
construct the measurement.

178 • Setting Up Algorithms 009-0506, Rev. B


5.26 Relationships
A Relationship is a mathematical formula calculation or comparison that is
set up between one or more Inspection Points. The Relationship equation
calculates results, and is reported similar to other Inspection Points. There
are also limits and alarms applied to the results.
Relationships contain the ability to use a mathematical formula, calculation,
or comparison that is set up between one or more checkpoints. This formula
can be limited and alarmed just as a checkpoint can. The Relationship can
include deviation results, numeric constants, nominal values from other
Inspection Points, offset values from other Inspection Points, or the results
from other Relationships.

Important Note: Relationships values are always the reported


value. If it is on the left side, and the left-side reporting has been chosen,
this may mean that the value is negative of what is needed. If the
checkpoint is configured for NTM, the NTM result is the value.

5.26.1 When to use Relationships


The Relationship concept is a powerful tool that can be used for basic
reporting or limit and alarm functions. Some examples of the proper use of
Relationships might include:
- Basic opening width computations, where 2-D or 1-D features are
parallel
- Basic angle computations where the angle is not more than a simple
trigonometry function.
More complicated applications may be best performed using Visual Fixturing
instead of Relationships. Some examples of concepts that are best performed
using Visual Fixturing instead of Relationships include:
- 3-D computations, where the Relationships would become too
complicated to understand and are likely to be incomplete. Deck lid
waterfall computations and motor compartment holes are some
examples.
- Point-to-plane distance computations
- Suspension geometry computations
- Frame-to-body mounting point inspection, particularly where there
are more than 4 mounts
- Situations where the features are not parallel. Computing the width of
a windshield opening from A-pillar measurements sometimes requires
the header measurements to be correct. Visual Fixturing is a better
solution for this application.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 179


Visual Fixturing is a powerful tool that is much more capable with more
complicated 3-D geometry. If the solution requires only a few simple
equations, Relationships are the right answer - but make sure that the scope
of the project will never expand beyond these equations. Answering the
question regarding what information is needed by the customer will help
make the correct choice of Relationship or Visual Fixturing.
If the application requirements need to consider rotation, or features that are
more complex than simple parallel lines, chances are that Visual Fixturing
will be the tool that will provide the best results and will be easier to
configure.

5.26.2 Creating Relationships


5.26.2.1 Free Form
To create:
1. Add a feature, and configure it to be a Relationship Feature
2. In the properties, define the name and the equation(s) to be used. Up
to three equations can be defined (for R1, R2 and R3). Equations
must be separated by the “;” character. The F1-F20 factors will be
defined later.
3. If a problem has been found in the equation, the OK button will not
close the dialog until the problem has been fixed.
4. Note that the next step locks the equation from any further editing.
Make sure that it is right.
5. Go to the Relationships folder in the specific part where this newly
created Relationship will be used. Click New Relationships, and
select this Relationship from the list shown.
6. Click the properties for the newly added Relationship. Configure the
appropriate reporting characteristics for R1, R2 and/or R3. The F1-
F2 as the deviation (“Measurement”), nominal or offset must also be
configured now. Also make sure that the “Enable” is properly
selected.
7. Click the “OK” button.
The free-form Relationships support the following math functions:
Function Comments
SQRT(x) Returns the square root of x. If the value of x is less than zero,
the result is undefined.
SQR(x) Returns x2.
SIN(x) Returns the sine of x, where x is in degrees.
COS(x) Returns the cosine of x, where x is in degrees.
TAN(x) Returns the tangent of x, where x is in degrees.

180 • Setting Up Algorithms 009-0506, Rev. B


Function Comments
ASIN(x) Returns the arcsine of x in degrees. If x < -1 or x > +1, the
results are undefined.
ACOS(x) Returns the arccosine of x in degrees. If x < -1 or x > +1, the
results are undefined.
ATAN(x) Returns the arcsine of x in degrees
MAX(x,y) If x>y, MAX returns x, otherwise it returns y.
MIN(x,y) If x<y, MIN returns x, otherwise it returns y.
DEGTORAD(x) Returns π * x / 180.
RADTODEG(x) Returns 180 * x / π.
The following is an example Relationship. This Relationship is used in an
example where Visual Fixturing would provide better information, but is
used for explaining the capabilities of a complex Relationship. This has been
restructured to align the elements of the MAX and MIN functions, only for
the purpose of readability. In Relationship equations, this spacing is not
present, and the names need to be replaced with F1-F20 items that are
configured to point to the appropriate inspection point or Relationship names.

This example computes the equation 2 * XaxisRange 2 + YaxisRange 2 .

R1=2*SQRT(
SQR( MAX(LBOX_1.X,
MAX(LBOX_2.X,
MAX(LBOX_3.X,
MAX(RBOX_1.X,
MAX(RBOX_2.X, RBOX_3.X)))))
-MIN(LBOX_1.X,
MIN(LBOX_2.X,
MIN(LBOX_3.X,
MIN(RBOX_1.X,
MIN(RBOX_2.X,RBOX_3.X))))))
+ SQR( MAX(0-LBOX_1.Y,
MAX(0-LBOX_2.Y,
MAX(0-LBOX_3.Y,
MAX(RBOX_1.Y,
MAX(RBOX_2.Y, RBOX_3.Y)))))
-MIN(0-LBOX_1.Y,
MIN(0-LBOX_2.Y,
MIN(0-LBOX_3.Y,

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 181


MIN(RBOX_1.Y,
MIN(RBOX_2.Y, RBOX_3.Y)))))));

Note that the MAX and MIN functions only permit two arguments.
However, there is no limit on function nesting. These functions are nested to
provide the ranges..
The second note relates to the Y-axis and negating values. Since many
manufactures have left-side reporting of deviations negated, the Y-axis
values on the left side must be negated for this equation. This is
accomplished with the “0-” leading the left-side Y-axis reporting. There is
no unary minus in Relationships, so you must either use a “0-” or a “-1*” to
negate single terms.
The current Relationship equations have only 20 factors (F1 through F20),
where there are 24 named items in the above equation. A reference like
RBOX_3.Y must be referenced using one of these factors. Although it is
possible to break this 24-term equation into two pieces, three equations
would be easier to read. One could use 12 factors in two equations and third
equation to create the final result from the first two parts.
For another example, computing the actual distance between two XYZ points
would look like:
R1= SQRT( SQR((F1+F4) - (F7+F10))
+ SQR((F2+F5) - (F8+F11))
+ SQR((F3+F6) - (F9+F12)))
- SQRT( SQR( F4 - F10)
+ SQR( F5 - F11)
+ SQR( F6 - F12));
Where the following F# items are configured as:
ƒ F1-F3 are the XYZ deviations for the first point
ƒ F4-F6 are the XYZ nominals for the first point
ƒ F7-F9 are the XYZ deviations for the second point
ƒ F10-F12 are the XYZ nominals for the second point
If any one of these points is a cross-car reporting point (i.e. left-side y
report), then the deviation(s) should negate the deviations that could be found
in F2 and/or F8.

5.26.3 Relationship Limits and Alarms


Limits and alarm functions for Relationships behave in the same way as the
limits and alarms function for other Inspection Point.

5.26.4 Editing Equations


One must delete all Relationship uses in the part before any equation editing
will be permitted. Once the equation edit is done, the Relationships may be

182 • Setting Up Algorithms 009-0506, Rev. B


reinserted into the part, and the R1-R2 and F1-F20 factors must be
reconfigured.

5.26.4.1 Work-around for Relationship Offsets


If offsets are required for a Relationship, the current implementation is such
that one cannot edit the equations or directly input offsets into the
Relationship point. Therefore offsets cannot be used in Relationships
without this work-around. This work-around uses a Color feature is defined
to specifically offset a Relationship result. The idea is to use the Color
feature XYZ nominal points, which can be accessed by the Relationships.
Therefore the offsets may be entered into the Color feature nominal XYZ
without disabling and the painful reconstructing all of the Relationship
structures.

009-0506, Rev. B 009-0506009-0506 Setting Up Algorithms • 183


Appendix A. Error Codes

A.1 Error Code Listings


The following is a list of numbers, and corresponding error codes and
descriptions. Many messages are internal faults, which should never happen.
These internal fault numbers and text are not shown here. If any unlisted
error should appear, or if there are any doubts about the meaning of a
message, call Perceptron, at 734-276-2186.

Message
ID Message Text Description
1 FRAME A RW ERROR During initialization of the frame memory, a basic write-
then-read pattern is test is performed. If this message
appears, the frame “A” memory is failing the pattern test.
Replace the PCI/DSP card (495-0243) containing the
failed memory.
2 FRAME B RW ERROR See FRAME A RW ERROR description.
256 INVALID WINDOW Window is too narrow, or specified beyond the imager
size.
259 INVALID NUMBER OF CGS Same as too few CGs.
260 INVALID TOLERANCE The laser filter was too large.
262 INVALID SENSOR NUMBER During the download or request for a sensor to execute a
function, the sensor was not recognized.
263 INVALID ROW An algorithm asked to rectify a point outside of the image.
For example, a corner could construct the answer at the
theoretical intersect such that it is above or below the
image. A partial hole or slot in an image could also
construct the answer outside of the image.
264 INVALID COL This occurs for the same reasons as INVALID ROW.
266 INVALID LINE This applies to slots with straight edges, Edge algorithms
and the Min Range algorithm. The algorithm attempted to
find a point on a line that was not valid.
270 INVALID BEND LENGTH Vertex Edge has a bend length that needs to be larger.

184 • Appendix A. Error Codes 009-0506, Rev. B


Message
ID Message Text Description
285 INVALID FRAME GRAB The row and column camera configuration does not match
PARAMETERS the current camera attached to the DTC. The WinSen,
tool identified the camera configuration upon executing
the “Load Rect Data”. Re-program the rectification data
with the intended camera attached. See the WinSen
manual, part number 009-0347 or 009-0347-01. WinSen
also has an additional reason for generating this error that
is not possible in Vector software.
528 TOO FEW TPS For a laser image, not enough laser line could be found in
the image. Check for the part missing, the incorrect part,
an exposure that is too low, etc.
529 TOO MANY TPS For a laser image, too many points were found for a likely
answer. Check for a bad sensor, too much ambient light
or other source of image noise.
530 TOO FEW CGS For a laser image, after the CGs were filtered, not enough
laser line was found to continue processing.
532 TOO FEW POINTS For a laser image, not enough CGs could be used for the
desired algorithm
533 TOO FEW GRAYS For the Simple Edge Algorithm, the structure of the edge
does not have a profile that can create a proper answer for
that edge.
534 EXPOSURE LIMIT EXCEEDED This is an autoexposure fault, where the proper exposure
could not be found. Check the image.
545 NO VALID OBJECT For all Range measurements including Floating Point
measurements, no laser line could be found along the
search vector. Check the Test Measure page with Show
Algorithm turned on to evaluate the search vector. A
blank spot in the line along the search vector could explain
this result. For corner measurements, nothing that looked
like the desired edge could be found.
546 NO LINE FOUND For all Range measurements, including Floating Point
algorithms: No line found along the search path.
For other laser-based algorithms: The expected line could
not be found. Check the laser filter settings, exposure,
ambient light, multipath, etc.
For the Slot algorithm: Make sure that this message is not
in reference to the laser image. More likely to apply to the
feature image, where no edge line found where the
template found the object.
547 TOO MANY LINES When a laser-based algorithm finds too many lines to
track in a single image. Check for image noise, and laser
filter settings.
548 NO CIRCLE CENTER Only expected on the Cylinder, Hemmed Edge algorithms,
and some configurations of the Hole, Slot, and Complex
Slot algorithms. The data could not successfully find a
valid circle fit, where required.

009-0506, Rev. B 009-0506009-0506 Appendix A. Error Codes • 185


Message
ID Message Text Description
549 EDGE TOO SMALL The plane measurement has a length smaller than 1/3 the
taught line length.
551 TOO CLOSE TO WINDOW With the edge algorithm, the found edge point should not
get close to either side of the processing window. For
example, if we expect an edge on the left and the plane on
the right, part of the laser image connects to the left side of
the window.
553 BAD FEATURE VECTOR A bad feature vector on a Hole or the Cylinder algorithm.
The sensor orientation combined with the feature vector
does not make sense. Check the information on the
Advance Analysis tab. The limits for this condition
should show the problem and more documentation.
554 TOO MANY OBJECTS More than one valid feature was found in the processing
window. Check exposures, windows and thresholding
options. Also check for multipath and ambient light
causing problems.
768 NO PLANE FIT For hole and slot algorithms, a plane was measured with
the laser, but there no XYZ point could found on this
plane. Check the sensor orientation and feature data
carefully. This message will appear on older releases of
CSGMain.bin with the laser line being at a very large
angle from normal, typically more than 60 degrees.
771 SLOPE DIFF TOO SMALL Corner measurement found a result where the two sides of
the corner are nearly parallel or flat, and the answer is not
repeatable.
774 SLOPE DIFFERENCE TOO LARGE For corners, holes, slots and range measurements, found
laser object does not match the slope of the taught laser
object. Check the image and laser filter parameters. For
Vertex edges, this refers to the actual edge points as
compared to the taught vertex angle, not the plane
measurement between the bend and line length points.
Currently, all edge algorithms do not create this message
for the plane.
4096 UNKNOWN COMMAND An unrecognized command was generated by Vector to
the image processor in the sensor. The programming of
the sensor’s DSP either has been lost or is not compatible
with the Vector software. Try restarting services or using
WinSen to load all sensor binaries. If that does not fix the
problem, check the versions on both CSGMain and
Vector. Call Perceptron with the version numbers for both
pieces of software for further instructions.
12323 ERR INVALID RECT SERIAL Rectification data serial number does not match what is in
NUMBER the sensor's flash memory. If serial number correct,
reprogram the entire sensor using WinSen.
46337 RELATIONSHIP BAD INPUT An inspection point result required for a Relationship
MEASUREMENT result has failed for this cycle.

186 • Appendix A. Error Codes 009-0506, Rev. B


Message
ID Message Text Description
46339 RELATIONSHIP CALCULATION An inspection point that is required for the Relationship
FAILED result is not active.
46340 Result is Not A Number (NaN). An invalid function argument parameter in a Relationship
will create this error. The square root of a negative
number is an example that can produce this error. The
arcsine and arccosine functions given a parameter that is
outside of the range of -1 to 1 will also produce this error.
102001 Could not be fixtured One of more of the fixturing translations or rotations has
exceeded its limit. Examine the Diagnostic Client
exception log for a report. If a report exists, it will show
more details about the failure.

009-0506, Rev. B 009-0506009-0506 Appendix A. Error Codes • 187


Appendix B. Mechanical
Considerations

B.1 Introduction
Carefully review mechanical considerations when installing a system to
eliminate instability that exaggerates the effect of temperature or vibration on
system measurements. Adding as little as 0.1 mm of unintentional variation
can have a negative impact. This variation can add noise to collected data,
causing a measurement to exceed system parameters, and throw all collected
measurements into question.
The earlier in a project you can address mechanical considerations, the more
likely you are to save time and money and increase customer satisfaction.
The further into a project you are, the more expensive it is to correct
mechanical problems. As a project evolves through the normal project stages
of:
- design,
- implementation,
- correlation, and finally
- customer usage
fixing a design flaw gets increasingly expensive. Often, the required fix will
disrupt previous work before the mechanical problem can be corrected.
Problems found during the customer use phase will be extremely hard to fix
and can destroy customer confidence.
If a system is not working, the customer is not likely to understand the
difference between the mechanical issues and Perceptron technology. Even
if a third party designed and built the station, Perceptron is the first to be
asked to find and fix any problems.
Avoiding mechanical problems helps insure that customers see design
integrity that supports data quality. Proper design instills the perception of
quality, and is a significant factor in shortening the time to buyoff. Customer
confidence in the data helps him identify process problems sooner, making
the return on investment in proper design obvious. Although many things
contribute to this confidence, mechanical design is what many customers see
and understand.

009-0506, Rev. B Appendix B. Mechanical Considerations • 189


B.2 Automotive Industry Experience
Many people with experience in the automotive industry are familiar with the
vibration and temperature control requirements for Coordinate Measuring
Machines (CMMs), which is driven by the very small tolerances for such
equipment, as small as 10 microns or less.
CMMs are installed in temperature-controlled rooms with an extremely rigid
mechanical grounding isolated from surrounding vibration. The CMM bases
are usually made of very thick granite isolated on a foundation that may be
several meters in thickness.
These requirements make sense in an environment that uses Renishaw
contact probes mounted on a long movable arm that can have a very low
resonant frequency at its full reach. Such equipment cannot tolerate any
external vibration.
These limitations, however, do not apply to standard Perceptron installations,
although there are certain similarities. People who have a CMM background
will be some of the toughest people to convince of some of the reliability of
Perceptron installations where the foundation is nothing more than a standard
concrete floor in an automotive factory.

B.3 Mechanical Factors


This section includes a discussion of three important mechanical
considerations.

B.3.1 Abbe effects


The Abbe effect, which describes the mechanical effects of a long arm
traveling on a way, a configuration common to many mechanical CMMs, is a
common problem in measuring systems. The basic concept and terminology
is shown in Figure B-1.

Figure B-1. A simplified Abbe effect with a single error.

This shows how angular errors are amplified with lever arms. The effect is
named after Dr. Ernst Abbe (who worked with Dr. Carl Zeiss in the late 19th
century). Originally used to describe a certain type of mechanical lever arm
effect, the term Abbe error has been extended to describe many more
situations where long lever arms do harm. These lever arms can appear
everywhere.

190 • Appendix B. Mechanical Considerations 009-0506, Rev. B


Some examples of Abbe effects in our application include:
− Dismounting and remounting a long-standoff sensor is unlikely
to yield the same results. The size of the mount is smaller than
70 mm on a TriCam sensor. Therefore, to maintain 0.1mm with
the remount means that the sensor’s surface must repeat with
(70/1000) * 0.1mm, or 0.007 mm (7 microns). This is clearly
not a reasonable expectation from the sensor mounting surfaces.
− A door fixtured by in/out locators located low with inadequate
separation increases Abbe error at the door’s header. The
mechanical fixture and Visual Fixturing results have identical
problems for identical reasons.
− Distortion in a concrete floor influences a robot most at full
extension. Figure B-11 shows a full extension configuration
where the arm is horizontal.
− A sensor mounted on long tubes and/or long standoff amplifies
the distortions of its mounting base.
− Temperature changes to the mounting structure of a long
standoff sensor, or a sensor held in position with long tubes.
− Changing a resolver position on a robot causes small angular
changes on the readings which are amplified through the distance
of that joint. The Abbe offset in this situation is the distance
between the joint and each measurement point. Therefore, the
Abbe error changes at each point. A new correlation is
frequently the best way to recover from a robot repair that
changes the resolver reference.

B.3.2 Vibration
Vibration affects measurements in the here-and-now. Two readings, taken
only a few seconds apart, can reflect very different vibration conditions. This
means that the impact of vibration is immediately seen in system
measurements. For a stamping applications where the floors vibrate, if you
see good repeatability when the machines are not working, and degraded
repeatability when the machines are, this means that you need to find and fix
what is moving.
High-frequency vibration is good for vision-based systems because the small
high-frequency vibrations suppress the laser speckle (the effect of the slight
roughness) of the surface being measured. Theoretically, anything slower
than the speed of the imager is bad.
While a theoretical approach to the problem would suggest that any vibration
above the sample rate of 30 Hz is generally good, practical experience has
shown that resonant frequency above 10 Hz is adequate for great results. The
amplitude of a 10 Hz vibration is small enough and damp down quickly
enough not to be a concern, except only the most extreme cases.

009-0506, Rev. B Appendix B. Mechanical Considerations • 191


B.3.3 Mathematical Modeling of Vibration
The concept of vibration can be expressed mathematically.
The resonant frequency of a mass at the end of a spring can be represented
as:

1 k
f =
2 m
where
f is the natural resonant frequency,
k is the stiffness or spring rate, and
m is the mass.
This formula clearly shows that increasing the mass at the end of a spring
lowers the resonant frequency. Also, decreasing k, the stiffness or spring-
rate, lowers the resonant frequency.
This equation requires proper units of force and mass. The spring rate in
units of force per unit of distance—for example, Newtons per meter. With
this measurement of force, the mass must be kilograms, which match the
spring-rate force units for this formula. Therefore, a spring-rate of 1000
Newtons/meter with a 5 kg mass would have a natural resonant frequency of
7.07 Hz.
The importance of this formula is for comparative purposes, not for
theoretical modeling. Some examples might be:
- Adding a second sensor doubles the mass on a long wall-mounted tube,
what effect does this have on the torsion forces applied to the tube?
If the resonant frequency was marginal with one sensor mounted on the
tube, you can expect that adding another sensor will cause problems,
since the resonant frequency will be lower by a factor of 0.707.
- Adding another sensor to a robot gripper is usually not problematic with
most FMS systems. The arm is rigid enough to handle the mass of the
arm and a full payload. The mass of the sensors and typical mounting
hardware is a small fraction of the mass of the arm.
- Adding a brace or fillet to a system clearly improves the stiffness, and
therefore raises the resonant frequency.
Although it is practical to model a situation with this equation, it is important
to understand how certain changes make an existing problem worse, or
eliminate it.

B.3.4 Temperature
Temperature variation can cause significant measurement swings over time,
but the variation between succeeding individual readings is very small. Over
time, however, the changes in measurement data caused by variations in
temperature can be greater than those caused by vibration.

192 • Appendix B. Mechanical Considerations 009-0506, Rev. B


B.4 Common Problems
We recommend that installers be conservative in the original design, when
the problems are easy and inexpensive to fix. Using appropriately stiff
structures does not increase the cost of a project as much as extensive
troubleshooting does. If you cannot implement a conservative design, then
installation is the next place to consider reducing mechanical design
variation.
The examples below show how an understanding of the underlying principles
can make simple design changes that have a major impact on the quality of
the installation.

B.4.1 Design Issues


B.4.1.1 Mechanical ground issues
Make sure that the sensors and the part fixture are on the same mechanical
ground; that is, that they are solidly connected. If the part and the station are
rigid, they should move together. Proper mechanical connection of the
sensor and part means it is feasible to measure accurately even in the
presence of vibration. Most concrete floors in automotive assembly and
stamping plants are adequate mechanical grounds.
Figure B-2 shows an example of how not to ground the sensors and the part
fixture. The design uses the floor, which creates a longer structure, on a
small attachment to the concrete floor. Grounding like this is more likely to
exhibit low-frequency resonance. Note the structure is a large mass at the
end of the spring created by the small mounting structure, and that this
structure is susceptible to temperature variations because small changes on
the floor are amplified at the point of measurement.

Figure B-2. A mechanical ground reference using a small footprint on the concrete.

009-0506, Rev. B Appendix B. Mechanical Considerations • 193


Figure B-3 shows a better arrangement. The designer has used a base plate
as part of the structure and added a fillet at the structure base.

Figure B-3. Improved mechanical grounding using a brace, a larger


footprint and a more direct connection.

To correct the problems in Figure B-2, change the fixturing so that it more
closely resembles Figure B-3 by:
• Locating the structure to the top or the side of the plate, or
• Increase the size of the structure and use fillets or bracing to make the
structure stiffer.
Note this change also reduce any influence of the concrete due to time or
temperature. Using a brace on the outside of the mounting point in Figure B-
2 is also acceptable, assuming a stable concrete floor and the mounting
footprint is big enough.
Under certain conditions, it may make sense to add vibration dampers
between the floor and the mechanical ground, although this is an extreme
solution. If you do decide to use this approach, remember that the
mechanical ground reference for the support structure is now mounted on the
dampers. This means that it must be sufficiently rigid to hold the sensor-to-
part Relationship constant. A professional engineer should be used to select
the correct dampers.

B.4.1.2 Top of the structure may move


Rectangular, I-shaped, or L-shaped structures mounted to floors need bracing
or filleting to prevent the top of the structure from swaying in relation to the
lower part. The braces should be apparent from every view of the structure.
Figure B-4 shows a sensor support structure without any support structures.

194 • Appendix B. Mechanical Considerations 009-0506, Rev. B


Figure B- 4. Improper structure with no braces present (front view only).

Figure B-5 shows the same structure as Figure B-4, but braces have been
added at the top. Bracing can also be added at the bottom. Typically, braces
at the bottom are added to the outside, and requires more grout plates.

Figure B-5. Proper bracing elements installed at the top forming a truss
structure.

B.4.1.3 Part Fixture issues


A very large mass at the end of a spring is likely to cause unwanted low
frequencies of vibration and possible unwanted temperature variation.

009-0506, Rev. B Appendix B. Mechanical Considerations • 195


Figure B-6. An improperly designed part fixture

Figure B-7 shows the needed corrections to the improperly designed part
fixture from Figure B-6. The structure in B-6 is improper because of the
large mass at the end of tooling with a small footprint, with a bolt pattern that
is too small. With this structure, the spring-rate is too low and there is a lot
of mass at the end of a lever arm.
There are also questions about the stability of the design of Figure B-6 over
temperature variation for the unfixtured results, which will be the process
fixtured results if the process locaters are used to hold the part. It is possible
that only the Visual Fixturing (product fixturing) results will be valid with
temperature variation.

196 • Appendix B. Mechanical Considerations 009-0506, Rev. B


Figure B-7. Corrected part fixture design.

This is the same problem as seen in Figures B-2 and B-3, only involving the
part fixture instead of the sensor structure. Therefore, the solution is the
same.

B.4.1.4 Overly-long tube structures


A variation of the mass-spring problem is using a long, unsupported tube to
mount a sensor. This structure is susceptible to both low-frequency vibration
and temperature problems. It can also affect the surrounding structure. Look
at the wall tube in Figure B-8 with its long standoff tube to see how
susceptible it is to torsional forces. Adding more sensors to the tube
aggravates this problem.

Figure B-8. Tube structure is too long because tunnel is too wide

009-0506, Rev. B Appendix B. Mechanical Considerations • 197


In an environment where temperature can vary significantly, these longer
standoffs are susceptible to temperature variation.
Consider vibration at the If the support for the wall tube is significantly far away from the supporting
feature on a real part, not at cross-clamp, the effective spring rate is even less.
the sensor to get the most Either situation can lower the resonant frequency to unacceptable levels. In
accurate understanding. general, use smaller structures whenever possible and keep the mounting
tubes short.
The vibration at the feature to be measured may be the problem, not the
sensor’s mounting structure.
If the long reach is unavoidable, you must add more structure. See Figure
B-9 for one solution to this problem—simply reducing the tunnel width.

Figure B-9. A more appropriate tunnel size for part shown

Another approach to consider is to add support using another wall-mounted


tube. This solution typically requires two tubes and three cross-clamps, so it
is not optimal, but it does work.

B.4.1.5 Too many sensors on a single tube


Another problem is similar to the overly long tube structure discussed in the
previous section. If there are many sensors mounted on a single tube with
long ball tubes, there is not enough rigid structure to keep the stiffness high
enough to avoid the low resonant frequency vibration. In Figure B-10, this
structure can create a low spring rate with the torsion forces applied to the
tube. Further, the use of the smaller footprint and additional cross clamps
create additional spring. If the structure behind the footprint is a 4x4 inch
structure or a 15mm thick plate, the forces are such that this structure is not
adequate.
Possible suggestions include:

198 • Appendix B. Mechanical Considerations 009-0506, Rev. B


- Use the correct wall mounts, and add one or two in the middle.
- Add a single wall mount to the side, with another tube and cross clamps
to suppress the torsional spring rate.
- Move these sensors to different or additional structures, removing some
of the mass of this structure.

Figure B-10. Too many sensors on a single tube.

B.4.2 Specific Design Recommendations


The application of these rules can serve as insurance against real and
perceived problems:
• It takes good data to create usable information. Vibration and
temperature problems can affect the accuracy and quality of the data a
system collects.
• Structures that are not stiff enough, which may have large mass or low
spring rates, can cause low frequency vibration that can be a problem in
any measurement system.
• Solid mechanical connections between the part fixture and the sensor
structure reduce low-frequency vibration, temperature variation, and
system instability.
You can see this condition during the installation process. If the
connection between the tunnel and the part fixture is not solid, the laser
tracker results for points on the two features will be inconsistent.

009-0506, Rev. B Appendix B. Mechanical Considerations • 199


• Use steel in the sensor mounting structure whenever possible. Using
aluminum should be avoided wherever and whenever possible.
• Concrete expansion joints in the mechanical ground should always be
avoided. The coefficient of expansion on most concretes is close to that
of steel, but it is subject to distortion with time and temperature changes.
Concrete is more elastic than most people imagine.

B.4.3 Designing for Special Features


• For part fixtures – Take the time to understand how flexible this area is.
Under certain circumstances, vibration can move this region of a part
excessively. For example, car door headers are too flimsy to locate
properly, which means it is subject to vibration when the door is fixtured
by the lower door structure.
• In stamping facilities, see if it is possible to remove the system from the
area affected by the vibration of the stamping presses. These facilities
generate low-frequency vibration, which is particularly difficult to deal
with anywhere above the bottom floor. Using dampers to suppress
vibration is a reasonable alternative; however, do not guess at the damper
selection. A wrong choice can make things worse.

B.5 Installation Issues


Even if a design is well thought out, small construction flaws can completely
defeat even the best designs. Once the system is delivered, and before
mounting the first sensor, be sure to inspect the measurement system
hardware so that you are completely confident about the quality of the station
construction.
Look for:
• Missing or defective welds
• Bolts that are not tight
• Improperly specified or poorly installed floor anchors
• Bad concrete
For Surface Plate that have reinforcing stringers welded to the back, check
these welds for being present and proper quality.

B.6 Mechanical Considerations for FMS


Flexible manufacturing systems are special cases, as sensors mounted at the
end of a robot arm can have unique issues. Like most installations, problems
are caused by low-frequency vibration, but, with a robot arm, one has no
control over the spring rate (See B.4.1.3 Part Fixture issues for a discussion
of the standard calculations that cause low frequency issues).

200 • Appendix B. Mechanical Considerations 009-0506, Rev. B


Robots have a unique problem with vibration because their resonant
frequency is a function of both mass and arm configuration. These problems
are most extreme when the arm is fully extended.
Adding stiffness need not involve building structures using titanium and
carbon fiber, but you should be careful when using aluminum, which can
aggravate temperature issues.
The proper rules for FMS design are:
• Keep the mechanical grounding between the robot and the part fixturing
solid.
• Aluminum robots have more vibration and temperature issues than steel
ones do. Make sure that you use temperature compensation when using
an aluminum robot.
• Use steel to mount the sensor to the robot gripper, but remember to keep
the mount as light as possible.
• Consider using a steel cage around a sensor in situations where the
sensor is vulnerable to collisions. Generally, the mass of the sensor,
mounting and protection is only a fraction of the robot’s rated capacity.
• Keep the distance between the part and the robot base as short as
possible.
• If there is a human operator, always consider all safety factors.
• If vibration dampers are required, note that robots are more vulnerable to
the issues described in Section B.4.1.4 Overly-long tube structures
than a fixed system because the robot arm resonance is more difficult to
control and the mass of the measurement assembly can be much higher.
Also note that robots create their own vibration when they start and stop.
• Accelerometers are a perfect tool for analyzing system resonant
frequency and vibration amplitude if you have any concerns about the
effect of vibration on a new system.
• If necessary, custom engineering services that analyze vibration
problems may be necessary to solve an intractable problem.
Figure B-11 shows a robot at full extension. This exaggerates any possible
problem of the robot base, the effects of the robot mounting and any possible
issues with joints 1, 2, or 3. All of these effects are amplified with the Abbe
offset. Also, all temperature effects are maximized with full robot extension.

009-0506, Rev. B Appendix B. Mechanical Considerations • 201


Figure B-11. One of the worst-case robot configurations for FMS.

With the horizontal configuration, as shown in B-11, if there is any elasticity


in the floor, or the anchors are not completely secure, this will be
exaggerated at the robot’s tool center point (TCP). This configuration
maximizes the force applied to the robot mounting.
The natural resonant frequency is at the lowest possible at full extension,
with the maximum mass at the lowest spring-rate possible. However, most
robots are stiff enough that this possible source of low frequency vibration is
not an issue, assuming that the mounting is proper. The link configuration of
the robot is such that the temperature variation is at maximum, which is an
issue for applications with temperature variation. Finally, any backlash in
joint 1 has the maximum effect with the configuration in B-11.
For all of the above reasons, it is recommended that some testing be
performed with this robot configuration. A simple target temporarily located
so that the sensor can measure this point would allow testing for backlash in
joint 1. A simple repeatability test using this target where the arm is
retracted and extended several times will highlight some of the worst
problems of the robot base, although a good repeatability does not prove that
there are no problems.
For FMS, shortening the distance between the robot base and the part is
helpful, but may cause other issues such as clearance problems, and
sometimes safety issues.

202 • Appendix B. Mechanical Considerations 009-0506, Rev. B


Appendix C. Exposure Index
Table
The exposure times in the following table are for reference only, and apply to
sensors with the 101A or 101B cameras. The exposure values are expressed
in seconds of exposure or exposure counts for WinSen. When collecting
images though WinSen, it is important to use this table to convert the Vector
exposure value to the WinSen exposure value.
Actual Actual
Time WinSen WinSen
Exposure Exposure
Displayed Exposure Exposure
time time
in Vector Value Value
(seconds) (seconds)
0.0005 -6 0.00025 14 0.5000
0.0010 -5 0.0005 15 0.5333
0.0020 -4 0.0010 17 0.6000
0.0040 -3 0.0020 18 0.6333
0.0080 -2 0.0040 20 0.7000
0.0167 -1 0.0080 21 0.7333
  (0.0167) 23 0.8000
0.0333 0 0.0333 24 0.8333
0.0667 1 0.0667 26 0.9000
0.1000 2 0.1000 27 0.9333
0.1333 3 0.1333 29 1.0000
0.1667 4 0.1667 37 1.2667
0.2000 5 0.2000 44 1.5000
0.2333 6 0.2333 52 1.7667
0.2667 7 0.2667 59 2.0000
0.3000 8 0.3000 67 2.2667
0.3333 9 0.3333 74 2.5000
0.3667 10 0.3667 82 2.7667

009-0506, Rev. B Appendix C. Exposure Index Table • 203


0.4000 11 0.4000 89 3.0000
0.4333 12 0.4333 119 4.0000
0.4667 13 0.4667 -- 5.0000
Doubling the exposure time is equivalent to a “stop”, which gathers double
the amount of light energy. (See the definition of stop in the glossary.)
Although most algorithms work over a range of ±2 or more stops, some may
be tighter because of the surface condition or the taught parameters. Typical
part-to-part surface changes will be at least one stop. Note that exposures 0-5
only have single stop resolution for each step.
Note the exposure gap above noted at the red “”. The exposure jump
between exposure values 5 and 6 has a 2-stop jump (factor of 4 time), where
it is not possible to get the expected 0.0167 second exposure time.
The following chart shows the exposure settings in logarithmic scale up to
one second. The log scale is shown as to understand the “stop” equivalents,
or where the amount of light collected doubles. Each full number increment
represents a doubling of light. Note the large two-stop increment between
exposures 0.0167 and 0.0333 seconds, is noted at the red “” as it was
above. The 0.0167 seconds display is actually 0.0080, as shown in the table
above. At the other end of the exposure spectrum is where there is very little
difference between exposures. For example, exposures number 21 and 22,
only show approximately 1/10 of a stop. The higher exposure numbers have
more resolution than the lower. For a full doubling of the light from
exposure number 21, exposure 30 is required for a one-stop equivalent.

204 • Appendix C. Exposure Index Table 009-0506, Rev. B


Vector/DTC exposures on log scale up to 1 second

5.0

4.0

3.0

2.0

1.0
Time (logarithmic)

0.0 0.000
0.050
0.100
0.150
0.200
0.250
0.300
0.350
0.400
0.450
0.500
0.550
0.600
0.650
0.700
0.750
0.800
0.850
0.900
0.950
1.000
-1.0

-2.0

-3.0

-4.0

-5.0

-6.0

-7.0
Vector exposure times displayed

009-0506, Rev. B Appendix C. Exposure Index Table • 205


Appendix D. Tools to Upgrade an
IPNet to Vector

D.1 Overview
There are a few pieces of software that now help the process of moving an
IPNet system configuration to a Vector platform by automatically
transferring most of the important configuration parameters for the
Checkpoint/Inspection Point. The version 4.3 release or later is required to
perform this conversion.

D.1.1 What the conversion tools do


The major tasks that the tools do for the IPNet-to-Vector transfer include:
- Transfer the sensor position calibration. This can be manual angles,
VeriStar or TetraStar calibration. This eliminates the need for the laser
tracker work to recalibrate the sensor position.
- Extract the XYZ and IJK feature data available and transfer what
information is available.
- Transfer most of the taught parameters from IPNet into the Vector,
making most teach requirements a single click.

D.1.2 What this tool does not do


The tools do not do the following:
- Define the newly required IJKs that were not required for some features
in IPNet. Holes, slots and cylinders are examples.
- MinRange (now called 2-D MinRange in Vector) IJK features are a
unique problem in IPNet. This will require special conversion.
- The nominal calibration is not preserved. Recalibration is necessary to
establish these means.
- Converting and transferring FMS calibration is currently not available.

206 • Appendix D. Tools to Upgrade an IPNet to Vector 009-0506, Rev. B


- The bend lengths for all corners and edges must be redone. Clicking the
teach step will use the current part in station which eliminates the IPNet
bend length parameters.

D.1.3 Overview of the IPNet-to-Vector configuration transfer process


The major steps for IPNet-to-Vector setup configuration transfer are:
- Decide what process will be used to recalibrate the means. Typically,
this is another CMM comparison of multiple parts.
- The feature names, XYZ and IJK data is transferred into Vector using the
standard “Sensor Lay-in” data file formats. The import of the file is used
to create the features in Vector.
- Some basic manual operations are required to configure some concepts
in Vector that do not exist in IPNet.
- Using the created features, and the IPNet configuration xml file, a
program is used to create all of the inspection points and import almost
all of the taught parameters.
- All algorithms must be taught again, most of which does not require
much manual intervention. Given that the exposure, window settings
and many other parameters are transferred from the IPNet configuration
xml, to the Vector parameters, many algorithms will be a simple click of
the teach button. As discussed above, the bend length parameters are
destroyed when one teaches to the current part.

D.2 Calibration Options


The current conversion tools do not preserve the calibration. One of the two
options listed here requires measurements from the IPNet. Therefore, a plan
to calibrate the Vector must be chosen before the process starts.
The first of the two options is obvious, using a new CMM comparison. This
may take hours or days of time, and the station will be down during this
period.
The second option to consider is use a part in station during the entire
conversion process. It is recommended to leave this part in station during a
shut-down period. If it is not possible to leave the part in station, the second
option requires a properly conducted and documented gauge R&R test. The
same part must be reinstalled after the conversion process, and there must be
confidence that it will reinstall to yield the same results with the second
installation. If the conversion process cannot be conducted during a shut-
down period, there will also be some station down time.
The process of using this reference part would be the following steps:
- Use the current IPNet to measure and document a part in station with a
minimum of 10 cycles. This documents what the means and
repeatability of all Checkpoints.
- Perform the conversion steps below.

009-0506, Rev. B Appendix D. Tools to Upgrade an IPNet to Vector • 207


- Reinstall the reference part in the measurement station, if required.
- After the conversion, measure the part 10 times with the Vector.
- Create offsets to move the means back to the original IPNet mean values.
- Measure the part again 10 times, and validate the means are now the
same. If they are not, adjust the offsets and measure until this is
successful.

D.3 The Step-by-Step Conversion Procedure


Once the planning phase is complete, and the current IPNet configuration
xml is exported, then one can start using the conversion tools. The following
provides the more detailed step-by-step procedure:
- Creating the “Sensor Lay-in” data file
• Make sure that you have a current version of the IPNet configuration
xml file.
• Make sure that the Vector is starting with an empty configuration.
• Find all sensors, and review the sensor list for completeness.
• Transfer the IPNet configuration file to the Vector hard drive.
• Execute the “IPNetToLayin.exe” program found in
D:\Perceptron\Utils. This program will use standard windows tools
to locate the IPNet configuration xml, and defining the output as a txt
file.
• Fill in the platform and part space names. (It is also possible to use
standard Vector software to define these names a few steps later, but
the names will omitted from the txt file that is being created.)
• Click “Convert”, and wait for the “Conversion successful” message.
• Click “Close”.
• Validate the created txt Lay-in file. If there is a reason to validate
the contents of this file, the best tool is a spreadsheet. (Standard text
editors do not typically line up the columns.)
- The steps to import the Sensor Lay-in data file are as follows:
• Go to Plant Setup.
• Right click the cell, and then left click “Import Sensor Lay-in”.
• Navigate to the txt file that was created in the previous step.
• Check to be sure that the platform and Part Space are now all
correctly defined.
• Note that the Min Range algorithm IJKs must be incorrectly
defined on the IPNet to make that algorithm work. These need
to be converted to the proper standard before the algorithm will
work on the Vector platform.

208 • Appendix D. Tools to Upgrade an IPNet to Vector 009-0506, Rev. B


- Create the model(s) in the platform.
- Assign all of the created features into the appropriate models.
- Create the required Part Types in the Part Space.
- Assign all features as Inspection Points in the Part Type. Note that all
Inspection Points are now assigned to phase 0.
- Run the IPNetPostLayin.exe program found in D:\Perceptron\Utils.
• Navigate to the same IPNet configuration xml.
• Make sure that the correct cell and Part Space is
selected.
• Left click “Convert”.
• Any Checkpoints found in the IPNet configuration xml
that do not have a matching Inspection Point in the
current Vector configuration will create a warning.
• If any Tools required are not found in the Vector
configuration, this will create a warning. This may
happen if a sensor is not found. A sensor that has been
disconnected will create such a warning.
- Refresh the Tool Position displays. Validate the calibration types in the
list using the “Details” option.
- Teach all algorithms.
- Establish the nominal calibration through the chosen process.

009-0506, Rev. B Appendix D. Tools to Upgrade an IPNet to Vector • 209


Glossary of Terms

1-D Algorithm
An algorithm where the measurement result is expected to lie along a given
line. As this line may be skew to the 3-D coordinate frame, the result may
have X, Y and/or Z components, but only with the NTF reporting. The
Range algorithm is the common example. A 1-D algorithm can be reported
as NTF XYZ coordinate deviations, an approach vector deviation or a single-
value NTM deviation. See section 3.5 for which algorithms are classified as
1-D algorithms. For CMMs, all moves should occur inside of the
measurement line.

2-D Algorithm
Also known as a Sectioning Algorithm. A 2-D algorithm will yield
measurement results that are expected to lie in a plane. As this plane may be
skew to the 3-D coordinate frame, the results may have X, Y and/or Z
components with the NTF option, similar to the 1-D algorithm. The Corner,
Edge and Cylinder algorithms are all examples of a 2-D algorithm. The
floating-point corner and edge with the “both” option are also 2-D. A 2-D
algorithm may be reported as NTF XYZ coordinate deviations, cutting plane
deviations or two-value NTM deviation. The two values would be left and
right for corners, and plane and edge for edges. See section 3.5 for which
algorithms are classified as 2-D sectioning algorithms. For CMMs, all
measurements must occur in the plane to properly construct the 2-D result.

3-2-1 Net
The 3-2-1 Net (also commonly called a 3-2-1 nest) is the configuration of
fixturing points that controls the six degrees of freedom with the minimum
number of points. Locating points are in 3 planes, with 3 points forming a
triangle in one plane, 2 forming a line in another, and 1 comprising a single
point in the last plane, making six dimensions in all. The planes should be
reasonably orthogonal, that is, they should not be parallel, and ideally are at
right angles to each other. See PLP Locators.

210 • Glossary of Terms 009-0506, Rev. B


Abbe error
The error that results from the small positioning error that is magnified by a
long lever arm. See Appendix B.

Actual
The as-measured result of a measurement typically expressed in part
coordinates. A deviation is then the difference between the actual and the
nominal.

Algorithm
In this manual, an algorithm is a sequence of instructions that will allow the
computer to process an image. The algorithm generates the geometric
construction of the actual or “as-measured” results of a single measurement
result of a TriCam sensor. The taught parameters are used for processing the
image and generating the geometric construction of the results.

Anchor Point
An anchor point will always generate a zero deviation in the unfixtured
results. As the fixtured results will generate a non-zero deviation for a part
that is not at the nominal position, an anchor point may be used for analysis,
or for limits and alarms. For a door example, say that we use the same
feature for fixturing and an anchor point for the x-axis fixture. If this
measurement is 5mm aft, the fixtured results should be nearly zero for the
inspection point, but the anchor point would show 5mm forward. This alarm
may be used to signal that the door locators are no longer working as
expected.

Approach Vector
An approach vector measurement is only possible in 1-D algorithms,
specified as an X-, Y- or Z-axis approach vector or an I, J, K. For example,
an X-axis approach vector is a measurement whose results always have the
deviation values zero in the Y- and Z-axes. This emulates a Surface Plate or
CMM measurement where two of the three axes are locked at the specified
coordinates as the measurement is taken as the probe approaches the
measurand surface in the selected axis. Therefore, all of the measurement
deviations are is in the selected axis.

Auto Mode
Refers to the Vector's state when an external event controls its operation.
Typically, the external event is a digital I/O signal processed by the PLC.
Contrast to Manual Mode.

009-0506, Rev. B Glossary of Terms • 211


Blunder
The term now commonly used to describe a measurement mistake. Other
terms formerly used for blunder include error, outlier or flier. It is
appropriate to remove measurement blunders from any form of data analysis.
In the context of a hole measurement, an edge point that is far away from the
actual edge is a blunder, which, if not properly rejected, results in a
measurement blunder.

Body Coordinates
Customers in North America frequently use this term to describe what is
know in Vector as Part Coordinates..

Canned Relationship
A canned Relationship is one of the predefined, commonly used Relationship
results.

Cartesian Coordinates
Means whereby the position of a point can be defined with reference to a set
of axes at right angles to each other. Cartesian coordinates are usually scaled
in millimeters.

Centers of Gravity (CGs)


A set of measurement points that represent the exact center of the laser line at
several points. Many CGs are used for constructing an actual measurement
result. A CG may be expressed in row/column or as a Y/Z point in sensor
coordinates. Also see Transition Points.

CMM (Coordinate Measurement Machine)


A CMM is a machine using servos and a touch probe to make precision
measurements. It is frequently used as the correlating device to calibrate
other measurement systems.

Contour Sensor
A sensor that measures a 2-D sectioning or 1-D algorithms. A contour sensor
is the preferred sensor for these algorithms. See section 3.1 for a description
of the benefits of the contour sensor as compared to a surface sensor.

Coordinate Frame
A coordinate frame is a Cartesian coordinate frame. Points in a Cartesian
coordinate frame are expressed as XYZ, and direction vectors expressed as
IJK. Features are defined as XYZ points and IJK vectors in the user’s
coordinate frame, and measurement results are reported in the user’s
coordinate frame.

212 • Glossary of Terms 009-0506, Rev. B


Critical Checkpoint
A checkpoint in the system that has been assigned an alarm condition and a
PLC output signal.

Cutting Plane
A cutting plane measurement is only possible in 2-D algorithms. This is
always specified as an X-, Y- or Z-axis cutting plane, indicating which axis is
locked to a given plane for the entire measurement process. A cutting plane
result is the only answer that is seen on Surface Plate, or CMMs that are
measuring a feature with one of the three axes set at a given coordinate for all
touches. For example, an A-Pillar may have a Z-axis cutting plane
measurement, where all measurements are at the selected Z-axis height,
showing only X and Y deviations. The Z-axis deviations will always be zero
for a Z-axis cutting plane as in this A-Pillar example.

Cycle Time
The period of time from starting one machine operation to where another
start may be properly performed.

Data
A general term used to denote any facts, numbers, letters, and symbols, or
facts that refer to or describe an object, idea, condition, situation, or other
factors. The act of interpreting data yields information. In the Vector, the
raw measurements are data, and information allows a corrective action to be
performed.

Datum points
Another name for the 3-2-1 net points, PLP locators or RPS points. This is
the most common terminology in GD&T fixturing schemes.

Degrees of freedom
In the context of a three-dimensional space using Cartesian, or rectangular
coordinates, there are three possible translations (along X, Y, and Z) and
three possible rotations (about X, about Y, and about Z). These are referred
to as the six degrees of freedom.

Deviation
The difference between an actual part and a nominal part. There are many
ways to express deviations, including NTM or XYZ coordinates.

Digital TriCam Sensor


Measures a feature in two or three dimensions (2-D or 3-D). The image
created by the laser line on the contour of the part surface is used to

009-0506, Rev. B Glossary of Terms • 213


determine the changes in location of edges, corners, and surfaces. There are
two categories for the TriCam sensors, either Contour or Surface. See
Contour Sensor and Surface Sensor in this glossary for more details

Digital SpotLite
The Digital SpotLite is a proprietary LED flood light that floods a feature so
that it may be measured. See chapter 2. for the proper application.

Digital TriLite
The Digital TriLite is a proprietary LED flood light that floods a feature so
that it may be measured. See Chapter 2. for the proper application.

Drawn Hole/Slot
Frequently used for fixturing, a drawn hole or slot is a hole that has very
rounded edges. When a pin is driven into the hole for locating the part, the
rounded edges help center the pin, and the bottom of the hole makes positive
contact with the pin. This is also known as an extruded hole or slot. See
Figure 2-1 for a drawing of a drawn hole, and Chapter 2. for details on
lighting considerations.

Equivalent Offset (or Equivalent Deviation)


When an offset is entered that does not lie in the reporting line or reporting
plane, these are the compute equivalent X, Y and Z offset values for the
values entered, or what the entered offset’s effective values will be. An
equivalent offset can be computed from either an invalid offset or a truncated
offset. If one entered the original offset directly into Vector, the resulting
offset results will be only the equivalent offset values. This is the result of
the assumption built into Vector that the source of the offset is completely
valid.

Extended Offset
A truncated offset that is extended such that the equivalent normal-to-feature
offset’s reported components match the entered values. For two correlated
machines measuring a given feature, either one or both reporting truncated
results, this is the recommended offset choice to repair the damage created by
the truncation.

Extruded Hole/Slot
See Drawn Hole/Slot.

214 • Glossary of Terms 009-0506, Rev. B


Flexible Measurement System (FMS)
A cell configuration of a robot with a sensor mounted to the gripper for the
purpose of flexible measurement. In contrast, most Vector systems are fixed
sensor designs, with no robot.

Field of Measurement
A smaller area than the field of view that is useable for measurement
purposes. For example, the area that is too far away from the sensor standoff
will yield out-of-focus results that are not usable for accurate measurement,
so these regions are not considered part of the field of measurement.

Field of View
The region that the sensor can image. In a contour sensor, this is considered
the 2-D area in the laser plane. In a surface sensor, this is the full 3-D area
that can be imaged.

FP
An abbreviation used for Floating Point throughout this manual. The
Floating Point Corner and Floating Point Edge are frequently abbreviated as
FP Corner and FP Edge throughout this manual.

GD&T
Geometric Dimensioning and Tolerancing. GD&T define a set of standards
(ASME Y14.5M – 1994) that are commonly used in industry for dimensional
measurement and tolerancing of parts. Briefly, the GD&T includes
specifications for:
ƒ The standards for the definition of fixturing of parts.
ƒ The concepts of measurement of certain part features. For
example, a hole position and size.
ƒ The definitions of the specification limits for these measurements.

Hard Fixture Point


A hard locating pin or surface that is used to locate a part in an AutoGauge
station. The locating pins are either round or diamond shaped, and will
locate holes or slots that are on the part. These can be combined with
sensors. The Visual Fixturing can therefore use mechanical locators and
sensor measurement results to define the actual part location.

IJK Vector
A vector expressing only direction, but not location. An IJK can be used to
define a line or a normal to a plane. It takes both an XYZ point and an IJK
direction vector to fully describe a line or a plane in a given 3-D coordinate

009-0506, Rev. B Glossary of Terms • 215


frame. The length of an IJK vector is expected to be 1.0000. The length of
an IJK vector is calculated by the formula I 2 + J 2 + K 2 . See section 1.3

Image Window (Image-Processing Window)


Denoted by a red boundary that is set by the operator in Vector image
algorithms. It selects the specified region of the image to process and ignores
the regions outside this boundary.

Inside corner
A corner where the sensor is aimed such that the theoretical intersect point is
further away from the sensor than the planar sides. Such a corner can
generate unwanted noise in the image from multipath reflections.

Inspection Point
A checkpoint in a Vector system is a reporting entity that uses a sensor to
measure a parts deviation at a specific location. Checkpoints have all of the
taught parameters required to perform this operation.

Interface
A hardware and/or software link which enables two systems, or a system and
its peripherals, to operate as a single, integrated system.

Invalid Offset (Invalid Deviation)


An offset that is neither in the expected reporting line or plane nor meets the
criteria for a truncated/extended offset is considered invalid. When an out-
of-line or out-of-plane offset is found, it is a likely due to some other
measurement, configuration or computational problem. Therefore, the root-
cause of the creation of the invalid offset should be found and fixed.
One means of generating an invalid offset is by adding extended offsets and
complete NTF offsets in multiple correlations. Although this may give a
proper result, combining offsets in this manner is not recommended. A
simple example of an invalid offset would be a non-zero Y-axis offset
entered into an X-axis approach vector measurement.

Lambertian
A surface that is perfectly matte or dull, which completely scatters light is
said to be lambertian. The classic example of lambertian is the surface of an
eggshell. A lambertian surface is the opposite of a specular or mirror-like
surface. Most surfaces are a mix of both specular and lambertian.

216 • Glossary of Terms 009-0506, Rev. B


Laser Tracker
A portable measurement system using an interferometer and encoders to
measure the position of a retro-reflector in a 3-dimensional coordinate frame.
SMX, Leica and API are common suppliers of this technology.

Light Emitting Diode (LED)


A solid-state device used for signal indication on the manual control and the
I/O module. Also, LEDs are used in TriLite and SpotLite to provide feature
lighting.

LVDT
Linear Variable Displacement Transformer or Linear Variable Differential
Transducer. This is an electronic equivalent of a dial indicator.

Manual Mode
Refers to the automation's state when the operator controls machine/process
operation. Contrast to Auto (automatic) Mode.

Matchboxing
A problem in car or truck bodies where the roof may move with respect to
the underbody, mostly in the in-out direction. This term came from the
original demonstration of this concept, where a matchbox cover was used as
a visual aid to illustrate the problem. A body that has a matchbox problem
will have difficulties fitting the glass into the opening, or the deck lid into the
deck lid opening, etc.

Measurement Uncertainty
Uncertainty is commonly (and incorrectly) called “measurement error”, and
is used to describe how accurate any given measurement might be. As the
correct measurement is unknown, the error value is therefore unknown.
Uncertainty is therefore a more appropriate term. For larger samples sizes,
this is expressed as a standard deviation or statistical variance. For more on
the basic understanding of measurement uncertainty, see
http://physics.nist.gov/cuu/Uncertainty/basic.html.

Mechanical Ground
The physical connection between the segments of a measurement system.
Just like an electrical ground, this defines the baseline from which everything
is referenced. For many systems, this is the concrete flooring. This is also
known as the “metrology base”.

009-0506, Rev. B Glossary of Terms • 217


Multipath
With structured light, the expected return is from the diffuse or lambertian
return of the structured light on the part. In other words, the light section
striking the path is the desired signal. With a sensor measuring an inside
corner, there may be a secondary noise return from a specular reflections or
the multipath. In this situation, there are at least two returns, that when
combined generates an image that looks like a “bow tie”, another name for
this condition.

Network
An arrangement of two or more interconnected computer systems to facilitate
the exchange of information in order to perform a specific function.

Noise
An extraneous signal capable of interfering with the desired signal. Noise
can be generated in electrical circuits with improper grounding or signal
levels that are too low. For optical sensors, ambient light or light from a
multipath configuration are common sources of image noise.

Nominal
The reference target for a given measurement. Typically, the nominal is set
to the as-designed target. The nominal (as-designed) measurements are
subtracted from the Actual measurements (as-measured) to generate a
deviation.

Normal to Feature (NTF)


This is a common measurement technique that, when implemented on a
CMM, the probe remains in a plane or a line that is normal to the feature that
it is measuring. The results of this are always XYZ deviations in the user’s
coordinate frame. Corners, edges, cylinders are all 2-D sectioning algorithms
where the measurement results should line in the plane normal to the feature
line. A range algorithm is a 1-D algorithm where the measurement results
line in the line normal to the plane that it is measuring.

Normal to Metal (NTM)


This measurement technique only yields dimensional data that shows the
deviation results normal to the surface or feature. These results are not
expressed in any coordinate frame. For a range measurement, the magnitude
of the NTM deviation is simply the square root of the sum of the squares of
the normal-to-feature (NTF) X, Y and Z deviation components. For an edge,
the NTM deviations are with respect to the surface and the edge shown as
two separate answers, specifically the plane deviation and the edge deviation.
The typical convention is that positive deviations for NTM results indicate
the metal “growing”, and therefore typically indicate that the deviation is
towards the sensor.

218 • Glossary of Terms 009-0506, Rev. B


Orthogonal
Intersecting or lying at right angles, perpendicular. Although perpendicular
and orthogonal are synonyms, and are used interchangeably, perpendicular is
typically used to describe the orientation of lines and planes, and orthogonal
is used to describe vectors or coordinate frame axes being perpendicular.

Part Angle
Defines to the system the location of the part relative to in car position, or the
body coordinate framework. This is the same concept as the Model Angle in
previous Perceptron products.

Part Coordinate Frame


Also known as Body Coordinates, CAD coordinates or World Coordinates,
this frame is the absolute measured X, Y, Z coordinates. By absolute, we
mean the total distance from the measurement coordinate system’s origin to
the measured location of the feature in the X, Y and Z axes. Body Coordinate
measurements match typical CMM methods of measurement that report the
deviations along each body coordinate.

PLP Locators
PLP is an abbreviation for plane-line-point, a common technique for
fixturing a part, defining the datum points. In the simplest form, PLP
location is the same as a 3-2-1 net fixturing scheme. Three or more points
define the plane, two or more points define the line, and one or more points
define the final point. PLP locators are the set of pins and pads that use
holes, slots and surfaces to define the part’s fixturing. PLP can also be used
for Principle Locating Points, which are simply different words for the same
concept.

Out-of-Line measurements
When measuring a plane, the results are typically expected to lie in a line,
typically normal to the nominal plane. For a CMM, a single-hit measurement
is measured along this line, and therefore all results are expected to lie inside
this line. The results may fall outside of the plane if they are improperly
programmed, or truncated. If the CMM uses PC-DMIS ™, an out-of-line
result can be the result of using the ‘RMEAS’ command. This term only
applies to 1-D measurements.

Out-of-Plane measurements
Similar to the out-of-line condition, the out-of-plane condition may happen
with 2-D measurements.

009-0506, Rev. B Glossary of Terms • 219


Process Fixturing
Fixturing that reflects how a part was made. In a door example, there are
typically a hole, a slot and a surface that reflects the references the assembly
process used to make the part. Similarly, for a frame or a body-in-white, a
hole, a slot and a surface is typically used for the assembly process. Using
measurements in the process fixture is important when a process adjustment
is required.

Product Fixturing
Fixturing that reflects how a part will be used. For a typical door, it will be
hung by making the flush adjustments, then two up/down locators and finally
a single for/aft adjustment to create the desired gap. Given this, the product
fixture will reference these outer surfaces and features. Measurements to this
standard will reflect how the door will appear on the final product. For a
body-in-white, the product fixture concept breaks down into functional areas
such as deck lid opening, suspension mount points, drive-train mount points,
etc.

Range
In reporting or simple data analysis for a given sample of data, the difference
between the maximum and minimum values in these data. A range algorithm
is also 1-D measurement strategy, equivalent to a single CMM measurement
touch point.

Reference Positioning System (RPS)


Another name for fixturing definition. (Sometimes RPS is the abbreviation
for Reference Point System.) See definition for PLP locators.

Residual
The differences between the actual and the predicted values based on a best-
fit math model. With adequate sampling of actual values, small residuals
mean that the math model is properly fitting the data and the actual values do
not have a significant amount of noise. High residual may indicate blunders
in the actuals, too much noise in the actual measurements, or an improper
math model.

Sensor
Devices that are used to measure particular attributes. Perceptron sensors are
used to determine the dimension or location of features.

Sensor Angles
The orientation of each sensor's axes relative to a coordinate frame relative to
gravity and the horizon. The part angles define the relationship of the Body
Coordinate to gravity, and therefore the sensor angles define the orientation

220 • Glossary of Terms 009-0506, Rev. B


of the sensor to the part coordinates. The sensor angles may be computed
from the ConeStar process, FMS calibration or directly measured with an
inclinometer and protractor.

Sectioning Algorithm
As another name for the 2-D algorithm, please read the “2-D algorithm”
entry in the glossary. This name is derived from the sectioning view that is
common in CAD drawings.

Shanking
CMM measurements typically use the calibrated red ruby ball tip for
measurements. For sheet metal measurements of small features, the white
shank, which is not calibrated, is sometimes used. The shanking operation
can use the shank at various depths, creating more measurement uncertainty.

Specular
A surface that is a very glossy or mirror-like is said to be specular. A mirror
is the near-perfect example of a specular surface, where light is only reflected
at the mirror angle and not scattered. Black automotive paint is also very
specular, although not as efficient as a mirror. A specular surface is the
opposite of a lambertian surface. Most surfaces are a mix of both specular
and lambertian.

SpotLite
The controlled lighting used to floodlight a hole or slot to be measured with a
Perceptron surface sensor. See also TriLite.

Stop (Exposure Stop)


Photographers will use the term f/stop and “Exposure Value” to describe the
amount of light gathered. The use of the stop term is a slight variation on
these concepts, as the sensors always have a fixed aperture that cannot be
changed. Doubling the exposure time gives the equivalent exposure value as
opening the aperture by 1 f/stop, and is said to increase exposure by one stop.
Doubling the amount of light on a give subject is also equivalent.

Standoff
This is the distance between the sensor and the center of the field of view.
With TriCam sensors, this is the normal distance between the sensor’s
window and the Z=0 location, which is also the center of the field of view.
For the contour sensor, the standoff is the distance between the laser window
and the center of the field of view. For the surface sensor, it is the distance
between the receiving optic window and the center of the field of view,
which is also the best point of focus.

009-0506, Rev. B Glossary of Terms • 221


Structured Lighting
Lighting that has a distinct geometric structure is referred to as structured
lighting. Perceptron sensors project only a plane of light. This projected
plane will generate a line when striking a part, which is then viewed at a
triangulation angle for measurement. Feature lighting (described in Chapter
2. ) is sometimes referred to as unstructured lighting for its lack of geometric
structure. Generic structured light may be multiple points, multiple planes,
circles or other geometric forms.

Surface Plate
A surface plate looks just like a CMM, excepting that the measurement
equipment is all hand driven. There are no motors and no programming
involved. This tool is the origin of the cutting plane and approach vector
concepts that still can be found in CMM programming. Actual surface plates
are becoming quite rare and obsolete.

Surface Sensor
A sensor that measures 3-D. A surface sensor also uses a structured light
source, which means that it is capable of measuring the 2-D and 1-D features
as described in Figure 3-6. As the receive optics are adjusted for the 3-D
measurements with a certain depth of focus, this means that objects far or
near may be out of focus. In contrast, a contour sensor is focused only on the
laser plane, which is important for that purpose (see section 3.1). Surface
sensors are only recommended choice for 3-D measurements, an FMS
sensor, or the easier 2-D and 1-D measurements in a fixed sensor
environment.

Transition Points (TPs)


A Transition point defines the start and end of a laser line image in a given
image column. Since the image of the structured light is a line, each column
has a transition point pair, between which is the section of the line crossing
the column. A single Center of Gravity (CG) is the computed point defining
the accurate center using the lit region between these transition points.

Translation
A movement along an axis (or axes) such that there is no rotation is a
translation movement. Mechanical ways provide translation-only moves.
Three axis translation stages are typically built with three orthogonal set of
ways.

Triangulation
The location of an unknown point, as in navigation, by the formation of a
triangle having the unknown point and two known points as the vertices.

222 • Glossary of Terms 009-0506, Rev. B


Perceptron sensors have the projector angle fixed, and the camera can be
thought of as an angle measurement tool. One known side length and two
angles create a known triangle.

TriLite
The controlled lighting used to floodlight a hole or slot to be measured with a
Perceptron surface sensor. See also SpotLite.

Truncated Offset (or Truncated Deviation)


The truncated deviations result from reporting where reporting axes with data
are disabled or ignored. Using comparison results to create offsets from
truncated deviations means that the offsets will be truncated. This results in
the loss of significant information that sometimes needs to be restored or
have computations applied to make it valid. See the PRCPAnalysis Manual
(P/N 009-0429) for a more detailed description of offset problems.

Variation
Change in the value of a measured characteristic. Sources of variation can be
grouped into two major classes: common causes and special cases.
Computing the range or sigma of the deviations are common means of
evaluating variation. Remember that using the sigma to evaluate
distributions that are not normal is not valid.

Visual Fixturing™ Calculator (VFC)


Each fixture defined in a system has a VFC, which includes all of the
definition for each fixture.

Y- Z plot
See section 5.6 for the definition and purpose of this tool.

009-0506, Rev. B Glossary of Terms • 223


glossary definition, 214

E
Edge Algorithm, 116
Edge, Nominal Point and Surface Vectors, 79
Error Code Listings, 187
Exposure Settings, Structured Laser Lit Images, 90
Index Exposure Tables, 205

F
Feature Lighting, 8
Feature Vectors, 75
Features, Defining, 75
1 Field of Measurement
1-D Min Range Algorithm, 105 glossary definition, 216
Field of View
glossary definition, 216
2 FP
glossary definition, 216
2-D Min Range Algorithm, 112 FP Corner Algorithm, 133
FP Corner Report Option, 135
FP Edge Algorithm, 127
A FP Edge Reporting, 130
Actual FP Edge, Teaching, 130
glossary definition, 212 Free Form Relationship, 183
Advanced Analysis, 102
Algorithm Setup Basics, 90
Algorithm Setup Concepts, 100 G
Ambient light, 65 Gap and Flush Algorithm, 170
blocking the light path, 67 Gap-and-Flush Feature Definition, 89
changing sensor aim to control, 68 Glossary, 211
changing the light sources, 67
How the sensor helps, 66
using Auto-Exposure, 72 H
Hemmed Edge, 120
B Hole Algorithm, 150
Hole and Slot Setup Basics, 99
Bend and Line Extents, 97 Hole and Slot, Final Verification, 99
Hole Feature Definition, 86
Hole Parameters for Feature Image, 151
C Hole Presence, 170
Calibrating to Metal, 101 Hole Show-Algorithm Display, 160
CGs Hole, Interpreting the Taught Results, 157
glossary definition, 213
Complex Slot, 165
Coordinate Frames, 37 I
Corner Algorithm, 114 Introduction, 5
Corner teaching, 114
Corner, Nominal Point and Surface Vectors, 77
Creating Relationships, 183 L
Cross-car Configuration (Y-left/Y-Right), 73
Cylinder Algorithm, 139 Lambertian
Cylinder Feature Definition, 82 glossary definition, 217
Cylinder Parameters, 139 Laser Filters, 94
Lighting, 13
Backlighting, 23
D Diffuse, 22
Equal-but-OppositeRelated Documentation, 28
Deviation

009-0506, Rev. B Index • 225


Specular, 18 Test Measure, 180
Stereo, 13 Test Measure Results, 101
LVDT Theory of Triangulation for the Contour Sensor, 36
glossary definition, 218 Theory of Triangulation for the Surface Sensor, 36
Threaded Stud with Fixed Sensor, 145
Threshold Settings, 98
M Truncated Deviation
glossary definition, 224
Matchboxing
Truncated Offset
glossary definition, 218
glossary definition, 224
Mathematical Notation, 7
Mechanical Considerations, 191
MinRange YZ, 122 U
Multipath Issues, 91
Upgrading an IPNet to Vector, 207

N
V
Nominal
glossary definition, 219 Vertex Edge, 121
NTF
glossary definition, 219
NTM W
glossary definition, 219
Window Settings, 93

P Y
Patch Feature, 86, 90
Y-Z Plot Results for the Edge, 123
Perspective Distortion of Triangulation, 98
Y-Z Plotting Tool, 101

R
Range Algorithm, 103
Range Feature Definition, 76
Related Documentation, 5
Relationships, 182
Re-teach, When it is Required, 100

S
Scanned Stud Feature Definition, 85
Sensor Angles
glossary definition, 221
Sensor Position and Orienation, 40
Setting Up Algorithms, 73
Shim tests, 72
Simple Edge, 122
Slot Algorithm, 160
Slot Feature Definition, 88
Slot Presence, 176
Slug Hole Algorithm, 178
Slug Slot Algorithm, 180
Specular
glossary definition, 222
Structured Lighting
glossary definition, 222

T
Table of Contents, i
Teaching the Hole Algorithm, 152

226 • Index 009-0506, Rev. B

You might also like