Coherent Noise Task Update

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 45

Study of Coherent Noise in Silicon

Strip Detector at CMS


Atiq ur Rahman
Dr. Ashfaq Ahmad
National Centre for Physics, Islamabad

4/2/16

Coherent Noise Task

Outlines
Motivation
Framework
Measurement of Noise
Parameterization of coherent noise
Conclusion

4/2/16

Coherent Noise Task

Overview of the Silicon Trackers

4/2/16

Coherent Noise Task

CMS Tracker overview

4/2/16

Coherent Noise Task

Module Types

4/2/16

Coherent Noise Task

Motivation
To measure coherent noise in silicon strip detector at
CMS.

Develop a criteria/method to flag noisy modules in


Silicon Strip Tracker.

The corrections derived is intended to be used during


track reconstruction

Would reduce the chance of reconstruction of noisy


clusters and help to get more precise AOD data for
sophisticated analysis.

4/2/16

Coherent Noise Task

Framework
Using CMSSW_6_2_12-patch1
DQM and SiStrip packages
Root
Identification criteria

Fec:crate/slot/ring/CCU/MODULE/LLD/I2C etc .
4/6/2/85/26/10 vs DetID
Data: Pedestal Run

4/2/16

Coherent Noise Task

Some Important Definitions


Pedestal
ADC counts for each strip of APV chip in the tracker at stand alone
mode of silicon strip tracker without particle trigger.

Noise
The oscillation of these pedestals is called noise. Mostly noise has a
Gaussian distribution.

Common Mode
The common mode is the type of shift in the signals of strips due to
some EM pick up. It affects all the Analogue to Digital Converter
(ADC) signals in an Analogue pipeline voltage (APV) chip in a
coherent way.

Common Mode Subtracted Noise


The standard deviation of signal after common mode subtraction.
4/2/16

Coherent Noise Task

Noise Calculation

For each strip

To find the coherent noise, we look at the

difference of noise in adjacent strips.

4/2/16

Coherent Noise Task

Noise Calculation

Cont

Normalized Noise difference

Normalized noise difference is assumed to have

a nominal value when there is no correlated


noise and points deviating from this line are
considered to be in correlation or anti-correlation
as a first test.

4/2/16

Coherent Noise Task

10

Noise Modes
Depending upon the different trends of noises

which we encounter in Silicon strip tracker, we


define noise modes
Normal mode
Wing suspect
Dead or low noise mode
Weird mode
Normal is one which is normal trend in tracker,
wing mode is wing shaped noise trend , Dead
mode is none responding and having zero noise
mode

4/2/16

Coherent Noise Task

11

Normal Mode

4/2/16

Coherent Noise Task

12

Correlation coefficients for normal mode

4/2/16

Coherent Noise Task

13

Wing Mode

4/2/16

Coherent Noise Task

14

Correlation Plots for Wing Mode

4/2/16

Coherent Noise Task

15

Parabolic or quadratic fit

Size of A decide whether the parabola is wide

or narrow.
Positive A shows it is upward.
B is the general slope in equation of line.
C shift in the parabola i.e up or down

Xmin=-(B/2A)
Ymin= (Xmin, Ymin) is vertex of the parabolas

4/2/16

Coherent Noise Task

16

Parameterization of correlations coefficients

4/2/16

Coherent Noise Task

17

Map for Parameter A

4/2/16

Coherent Noise Task

18

Map for the Parameter B

4/2/16

Coherent Noise Task

19

Map for the Parameter C

4/2/16

Coherent Noise Task

20

Map for the Y(min)

4/2/16

Coherent Noise Task

21

Map for X(min)

4/2/16

Coherent Noise Task

22

Y(min) vs A

Only the large value of A do not ensure that the


channel will have large correlation. The
Parameter C play a decisive role in addition
to A . We deicide about the correlated channel
4/2/16
Coherent
Noise Task not with C or A individually.
23
with only
Y(min)

Y(min) vs X(min)

Lager values of X(min) is not hurting


our results
4/2/16

Coherent Noise Task

24

A vs X(min)

All the channels having large value of A parameter


always lies in Physical range of the APV. So we
should not worry about the non-physical strips.
Basically , such channels have a almost linear
distribution of correlation coefficients. That is why we
4/2/16
Coherent
Noise
Task
25
are getting
the
large
X(min).

Y(min) vs C

4/2/16

Coherent Noise Task

26

A vs C

4/2/16

Coherent Noise Task

27

X(min) vs C

All the anomalous values of the parameters


mostly coming in the Physical range of APV
Strips.Coherent Noise Task
4/2/16
28

Channels in the tail of Y(min) after the chiSquare cut As we have very small window
for y(min) in the distribution Y(min)>-.12
Channels in the Tails of Y(min) for Left_fit

4/2/16

Coherent Noise Task

29

Channels in the Tails of Y(min) for Left_fit

4/2/16

Coherent Noise Task

30

Channels in the tails of Y(min) for Left_fit

4/2/16

Coherent Noise Task

31

Channels in the tails of Y(min) for Left_fit

4/2/16

Coherent Noise Task

32

Solution to noisy and dead channel problem


As we were getting the noisy as well as the

correlated channels in the tails of the y(min)


which were contaminating our results for
correlated channels. I have flagged the noisy or
dead strips as not fitted to our general
quadratic(as well linear model). We have
obtained the distribution of chi square and got a
good cut for channels.
In rest of the channels we get only correlated

channels in the tails of Y(min).

4/2/16

Coherent Noise Task

33
33

4/2/16

Coherent Noise Task

34

Chi square distribution before and


after cut

The channels in the tail of Left Chi-Square


before the cut are presented in later slides. It
can flag even a small missing of strips!
4/2/16

Coherent Noise Task

35
35

The channels in the tails of the ChiSquare


we have got almost all the noisy and dead
channels in the tails of chi-square which were in
the tail of the Y(min) before the chi-square cut.
The Chi square definition is highly sensitive for
the channels which have small number of noisy
strips.
The quadratic fit model is a general fit and linear
fit is a special case of quadratic fit with parameter
A equal to zero.
All the non-Physical X(min) have nominal values
of parameter A and Y(min). They do not hurt our
results. This indicate that they have astonishingly
linear distribution of correlation coefficients.
4/2/16

Coherent Noise Task

36
36

Map for Chi-Square

4/2/16

Coherent Noise Task

37
37

Channels in the tails of chi-Square for


Left_fit

4/2/16

Coherent Noise Task

38
38

Channels in the tails of chi-Square for Left_fit

4/2/16

Coherent Noise Task

3939

Channels in the tails of chi-Square for Left_fit

4/2/16

Coherent Noise Task

40
40

Channels in the tails of chi-Square for Left_fit

The above noise profiles have chi-square value greater than two for
left_fit.

4/2/16

Coherent Noise Task

41
41

TIB
Does this y(min) is stable in the other Sub detectors?
Stability of the Y(min) in SST

4/2/16

Coherent Noise Task

42

TEC+

4/2/16

Coherent Noise Task

43

TOB

4/2/16

Coherent Noise Task

44

Conclusions
The Y(min) is used to decide about the correlation of the channels. No
any individual parameter can be used to quantify the correlation. A and
C are both sensitive but large value of A may not decide that this
channel is sensitive for correlation. Similarly , the only value of C do
not ensure us about correlation only.
But in the case of terribly linear distribution of correlation coefficients, we
can say that the C parameter becomes same as Y(min).
The chi-Square value for the quadratic fit gives a good criteria to
eliminate the noisy channels. The channels whose chi-square value for
the quadratic fit is greater than two are all noisy.
The values of Y(min) will be used for fine tune of cluster charge.
This criteria can be used to correct the CMS data and can reduce the
chance of noisy clusters.
Here is our twiki page https
://twiki.cern.ch/twiki/bin/viewauth/CMS/SiStripNoiseCorrelation
4/2/16

Coherent Noise Task

45

You might also like