Professional Documents
Culture Documents
Xiang 2014
Xiang 2014
Xiang 2014
frequency of stripes
Xiansong Xiang*a, Yancong Lua, Chunlong Weia, Changhe Zhoua
a
Shanghai Institute of Optics and Fine Mechanics, 390 Qinghe Road, Shanghai, China
ABSTRACT
A new method of single-track absolute position encoding based on spatial frequency of stripes is proposed. Instead of
using pseudorandom-sequence arranged stripes as in conventional situations, this kind of encoding method stores the
location information in the frequency space of the stripes, which means the spatial frequency of stripes varies with
position and indicates position. This encoding method has a strong fault-tolerant capability with single-stripe detecting
errors. The method can be applied to absolute linear encoders, absolute photoelectric angle encoders or two-dimensional
absolute linear encoders. The measuring apparatus includes a CCD image sensor and a microscope system, and the
method of decoding this frequency code is based on FFT algorithm. This method should be highly interesting for
practical applications as an absolute position encoding method.
Keywords: absolute position encoder, single-track, frequency encoding, stripe
1. INTRODUCTION
Position encoders are key parts in computer controlled manufacturing, and the encoding method employed decides the
accuracy, resolution, and error tolerance of a position encoder.
In general there are two kinds of position encoders, incremental type and absolute type. A significant difference between
incremental position encoders and absolute position encoders is the pattern form on the track.
The track of an incremental position encoder is marked with periodic stripes, and the period of stripes is a constant,
therefore distance can be measured by counting stripes. With the periodic patterns, the use of Moire effect and
quadrature detections gives incremental position encoders very high accuracy and resolution. The disadvantage of this
kind of encoders is that when power failures or mechanical shocks occur in the manufacturing process, encoders must go
back to a certain reference position (like zero position) in order to redetermine the current absolute position.
As for absolute position encoders (we discuss single-track situation only, since multi-track is hardly used), the patterns
on the track are aperiodic stripes arranged according to a pseudo-random binary sequence, m-sequence usually. The
sequence consists of a large number of “0”s and “1”s called codebits. On the track, “0”s and “1”s are placed at periodic
positions, each position has a stripe representing a “1” or has no stripe representing a “0”. The basic unit of the track is a
codebit, several such consecutive codebits form a codeword that is unique in the whole sequence. So it is feasible for the
encoder to find out the absolute position with a codeword owing to the property of pseudo-random binary sequence.
Although the accuracy and resolution is lower than incremental encoders, absolute encoders are more popular because
their convenience and ability against accidents.
As was stated above, conventional absolute position encoders use pseudo-random binary sequence to encode position.
There are the two main problems of such absolute position encoders.
First, the absolute position information is stored in every codebit as “1” or “0”, if a small area of the track is damaged or
contaminated, every ambiguous codebit will lead to a bit of information loss. One reversal codebit changes the codeword
completely, and the decoding result will be totally wrong.
Second, this kind of position encoding method only provides a coarse position value with the photodiode arrays as
reading head, in the order of the width of a codebit. As for more precise measurement, an additional incremental track is
necessary in general, which makes the system much more complex than a single-track system.
Holography, Diffractive Optics, and Applications VI, edited by Yunlong Sheng, Chongxiu Yu, Changhe Zhou,
Proc. of SPIE Vol. 9271, 92711I · © 2014 SPIE · CCC code: 0277-786X/14/$18 · doi: 10.1117/12.2071456
fs = N / w , (1)
the maximum of frequency detection is Nyquist frequency:
f N = fs / 2 , (2)
and the frequency resolution is given by
fdiv = 1 / w . (3)
2.3 Reading head
The most important part of the encoding method is the detection apparatus, especially the reading head, it decides the
way that patterns are arranged and the accuracy the position encoder can reach.
The reading head should be able to not only find out the period of stripes so as to provide a coarse determination of
position, but also recognize every stripe to achieve high resolution at the same time.
Conventional reading heads are photodiode arrays, the number of detection unit (photodiode) is the same as codeword
length, which means n photodiodes matches n-bit codewords. To recognize frequency of stripes, according to the
formulae stated above, much more detection units (sampling data) is necessary, photodiode array is not suitable for
frequency decoding system.
Charge Coupled Device (CCD) is an ideal device. A CCD sensor has thousands of pixels, can provide thousands of
sampling data in one time, which meets all the requirements of the reading head. CCD technology is now widely used in
point-and-shoot digital cameras, camcorders, high-definition TV, satellites and medical endoscopes, there are no
difficulties in bringing it into the field of encoder reading.
3. EXAMPLE
3.1 Parameter set
Based on the patterns designed above and using CCD as the reading head, here is a practical example of linear frequency
encoding system with realistic parameters.
The reading head is a CCD with at least N=1024 pixels along the track. A microscope system attaches to the CCD,
which adjusts the sight range of CCD to cover a distance of w=550um on track.
The track is 5cm long, and divided into 100 districts. Each district is L=500um long equally.
So the value range of spatial frequency is from 0 to 256/275 um-1 according to Formula 7.
Since sight range of CCD w is a little longer than district length L, there must be always a district longer than
L/2=250um in the sampling image, which determine the frequency internal should be at least 1/250um-1 according to
Formula 8.
The stripe frequency of the first district is set as
f1 = 2 / 11μm −1 . (12)
For the n’th district
Δf = 2 / 275μm −1 . (14)
Correspondingly, the period of stripes
d 1 = 1 = 5.5μm
f1 , (15)
1 275
dn = = μm
f1 + (n − 1)Δf 48 + 2n . (16)
By the end of a district, a blank area is reserved, which is longer than a stripe period, marking the end of the district.
According to this rule, the number of stripes can be calculated with
⎡ 949 + 40n ⎤
M = ⎢ ⎥
⎣ 11 ⎦ . (17)
Figure 2 is the diagram of such apparatuses. The key parts include a CCD image sensor as reading head, a microscope
with stable magnification, and a track with stripe patterns.
2.5 Parameters
Discuss the parameters in linear frequency encoding system.
Assume that the reading head has N detection units and the output covers a distance of w along the track, and the serial
stripes within it spread over a distance of v at least.
According to Formula 2, the value range of frequency is limited by Nyquist frequency
N
fn <
2w . (7)
According to Formula 3, in order to separate two peaks of adjacent districts in the frequency spectrum, the interval
between every two frequency value should be above frequency resolution
1
Δf >
v . (8)
In the data sampled by reading head, a border of district must be recognized within district, so the sight range of reading
head w should be longer than district length L to cover a border.
These are the basic limitations of parameters.
2.6 Decoding
A CCD reading head provides a grayscale image of the track as sampling data. The purpose of decoding is to extract
absolute position information of the reading head from the image.
Stripes are the reference marks for location in the image. A stripe’s absolute position (left edge) can be described as
district position plus the stripe’s relative position within district. Assume district order is n, stripe order in district is m,
then stripe’s absolute position X is
X = (n − 1) × L + m − 1
fn . (9)
The order of the district n can be obtained from the district frequency
n = 1 + (fn − f1 ) / Δf . (10)
The stripe order can be counted from the blank area that indicates the end of district. Assume the last period in district is
replaced by blank, the amount of stripes in a district is
M = [fn L − 1] . (11)
3. EXAMPLE
3.1 Parameter set
Based on the patterns designed above and using CCD as the reading head, here is a practical example of linear frequency
encoding system with realistic parameters.
The reading head is a CCD with at least N=1024 pixels along the track. A microscope system attaches to the CCD,
which adjusts the sight range of CCD to cover a distance of w=550um on track.
The track is 5cm long, and divided into 100 districts. Each district is L=500um long equally.
So the value range of spatial frequency is from 0 to 256/275 um-1 according to Formula 7.
Since sight range of CCD w is a little longer than district length L, there must be always a district longer than
L/2=250um in the sampling image, which determine the frequency internal should be at least 1/250um-1 according to
Formula 8.
The stripe frequency of the first district is set as
f1 = 2 / 11μm −1 . (12)
For the n’th district
Δf = 2 / 275μm −1 . (14)
Correspondingly, the period of stripes
d 1 = 1 = 5.5μm
f1 , (15)
1 275
dn = = μm
f1 + (n − 1)Δf 48 + 2n . (16)
By the end of a district, a blank area is reserved, which is longer than a stripe period, marking the end of the district.
According to this rule, the number of stripes can be calculated with
⎡ 949 + 40n ⎤
M = ⎢ ⎥
⎣ 11 ⎦ . (17)
Figure 2 is the diagram of such apparatuses. The key parts include a CCD image sensor as reading head, a microscope
with stable magnification, and a track with stripe patterns.
_ microscope
3.2 Simulations
In order to examine the sampling and decoding process of this system, a simulation is performed as below. Consider the
extreme case, the sampling simulation is at the beginning of the last district, since the last district has the densest stripes,
close to the sampling limit of reading head and also shares half of sampling points with the prior district.
Figure 3 shows the track with stripes of the distance between 4.946cm to 4.954cm, the last district starts from 4.95cm as
we can see.
L L
Assume the CCD reading head is centered at this starting point of the last district, taking 1024 sampling points of 550um
long on track, the result will be like Figure 4 (only about 160 pixels from 430 to 590 are shown). In practical situations,
sampling data are usually binarized to a sequence of “1”s and “0”s, correspondingly in Figure 4, a black bar represents a
“1” and a white bar represents a “0”.
11111111111111,111111111111111111110110111,111111111111111111111111111111111111
Figure 5 is the frequency spectrum from the whole 1024 sampling data of Figure 4. Although there are two districts with
the same length in sampling data, their main peaks of frequency can be distinguished clearly with no overlap. So any
district can get its frequency value according to this FFT process, then convert it to district order with
n = 1 + (fn − f1 ) / Δf . (18)
20
After getting the district frequency, the sampling data can be restored to a sinusoid wave by correlation calculation of
them. The simulation result is shown in Figure 6 (only about 60 pixels in the middle of 1024 are shown). The black
square wave represents the origin stripe patterns on track, the gray bars represent the sampling data like in Figure 4, the
two sinusoid waves (turned to upside-down for contrast with square wave) represent the restored functions of the
sampling data for two districts respectively. With the restored sinusoid waves, every pixel can be converted to the
absolute position value accurately.
A A AA 11 A i1 II II AA A I1 f AP A AA A 11
U U U V U U U U U VVV V U , v v U V V V ki ni V
487 492 497 502 F,87 512 517 522 527 532 537
4. OTHER APPLICATIONS
4.1 Two-dimensional absolute position encoder
Two-dimensional absolute position encoders can be achieved by the similar method. Figure 7 shows a two-dimensional
track with stripes of linear frequency encoding. In two orthogonal directions, the two sets of stripes do not influence each
other in data sampling. So in each direction sampling data can be obtained and processed as in one-dimensional system
independently, rows for y-axis position and columns for x-axis position.
As the reading head, a CCD image sensor perfectly suits the two-dimensional track, since it provides enough sampling
pixels in each direction, and every pixel is used in decoding process of two directions.
.............. - - IMII
=
MI
NMI MI
MI MI
.........C...........==.
111111111111111111111111111111111111111111111111111
C..........M..
M =
...........=.........=IM
............ =
.IIIII.........=I.I....IIMI...MM
11111111111111111111111111111111111111111111111M1111
II.II.III.II...1111III...I1M1.1M.
IIIIIIIIIIIIIIIMI.I..II=MM
IIIIIIIIIIIIII=I.I....II.IMMM.
..........IIIII=I.IIIIIIIIIM.M.M.
n..I.........I=.........IM....M=..
IIUIIIIIIIUIIMI.IO.III=IUMIMI.....
uuumnuuuunnn.............
1 ..............111...I..I...IM1....M....NI...
IIIIIIIIIIIIIII1111111111111111111111111111111
111111111I111111M.1....1.1.MI.........M.......MI....
11111111II11111M.I....U1.M.........M.......M...
1I11I111111111M..........M........M.....MM
1I111I111111111=...........M.........M......M
111111111111111=.........M.......M...
11111111111111M...........M........M
111111111111111=..........M.......M
1I1I1111111111IM.........IM.........M.......M.
44
.4.
.....,
4am
,.....
4aco
........
v-i li
Figure 7. Track of a two-dimensional absolute position encoder.
1-4
E,
J L 9J
Figure 8. The track of an absolute angle encoder.
ACKNOWLEDGEMENTS
The authors acknowledge the support of National Science Foundation of China (61127013).
REFERENCES
[1] Yeatman, E. M., Kushner, P. J., Roberts, D. A., “Use of scanned detection in optical position encoders,” IEEE
Transactions on Instrumentation and Measurement 53(1), 37-44 (2004).
[2] Engelhardt, K., Seitz, P., “High-resolution optical position encoder with large mounting tolerances,” Applied
optics 36(13), 2912-2916 (1997).
[3] Engelhardt, K., Seitz, P., “Absolute, high-resolution optical position encoder,” Applied optics 35(1), 201-208
(1996).