The primary idea was to create a default noise table used in simulation.

Actually, the plain simulation uses real noise values , which :

1. are not easy to deal with since it introduces independant ladder behaviour (meaning not uniform detector)
2. could be an issue for simulation request

There will be 2 tables :

1. for simulation in run 5 : using ssdStripCalib table
2. for simulation in run 7 : using ssdNoise table

The goal is to determine nominal (expected) noise values. Instead of applying brutaly rms = 3 adc for all the strips (which may be the nominal noise), I looked at the values for run 7 to get an

estimate.

The level for such a default table will be on a ladder basis, ie 1 value per ladder. (later one can optimize on a wafer basis).

Method :

• using all the entries in db, get a rms distribution for all strips in a given ladder, for all cumulated file.
• fit the distribution with a gaussian and take the mean as the rms value for all these strips 

(in the following plots I removed entries where rms == 0 and rms == 255)

 

example 1 (where it works) : ladder 11p

 

Here the result of the fit is a mean  <rms> = 42.89 adc.

So the real noise (divided by 16) would be <rms> = 2.68 adc

 

example (where it does not work) : ladder 1n

 

Here we have 2 peaks (one could fit by a double gaussian but this is not the goal of this study)

Looking at the wafers level :

We see that wafers 9 and 10 are responsible for the first peak (<rms_1> = 63 adc) and the other wafers for the second peak (<rms_2> = 130 adc)

--> Possibilities :

1. take one or the other mean value, or take the mean of <rms_1> and <rms_2>
2. determine the mean rms on a wafer level (which ihere is more meaningful)
3. look at the time dependence to see if the peaks disappear, that means for each entry in db (there are ~170 entries) , fit the distribution for each ladder, the take the mean of this distribution and fill another histogram containing the mean per ladder per file

(at the end it will make 170 (entries) * 2 (side) * 20 (ladders) = 6800 fits)

 

Part II :

I slightly modified the macro.

1. I evaluate the mean <rms> wafer by wafer (it was easier than I though, just do not use NTuple ...)
2. The fit is done using TSpectrum : it allows to find peaks and get the position (in our case the mean <rmz>). The the previous plot would be easy to solve
3. now we can choose which peak (based on it's height) to keep

An example : for wafer 3 of ladder 1 (side P)

 The output of TSpectrum is :

 

  side P     ladder = 0    wafer  = 2

 peak  0 : x =  55.5 , y =  6885 

 peak  1 : x =  34.5 , y =  1569 

 peak  2 : x =  175.5 , y =  6 

 peak  3 : x =  168.5 , y =  5 

 peak  4 : x =  152.5 , y =  6 

 peak  5 : x =  115.5 , y =  37 

 peak  6 : x =  189.5 , y =  5 

 peak  7 : x =  213.5 , y =  3 

 peak  8 : x =  207.5 , y =  4 

 peak  9 : x =  224.5 , y =  2 

 peak  10 : x =  255.5 , y =  3302 

 peak  11 : x =  127.5 , y =  16 

• <rms_1> = 34.5   adc
• <rms_2> = 55.5   adc
• <rms_3> = 255.5 adc

If I choose the highest contribution (<rms_3> has to be removed), then the mean rms for this wafer would be :

---> <rms> = 55.5 adc (<rms> = 3.46 adc)

 

 

 Part III : mean of rms <rms> vs run id (time dependence)

NOTE : these rms values have to be divided by 16.

For each wafer , I plot the rms for all the strips for a given run : I get the following plot (here for wafer 2 ladder 11 side P)

 

 Then i can fit this distribution and get an evaluation of the noise increase during the time :

This example shows a stable wafer as the decrease is less than 0.5 adc

On the contrary, there are others wafers where the increase is much bigger and also present some structures up to a given run ; it indicates a change during the pedestal run (air compressor failure)

The fit gives :

According to the slope of the fit , we have :

• beginning of the run : <rms >= 44 adc
• end of the run : <rms> = 116 adc
• so the increase is > 250 %

However, this wafer present a clear step for run ~ 90.

Following are the list of runs between id = 85 and id = 95.

 # 85   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070520.000113.root 

 # 86   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070520.042201.root 

 # 87   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070520.053337.root 

 # 88   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070520.113014.root 

 # 89   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070520.171331.root 

 # 90   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070521.035010.root 

 # 91   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070521.041543.root 

 # 92   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070521.091805.root 

 # 93   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070521.132234.root 

 # 94   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070521.235235.root 

 # 95   :/star/data06/SSD/pedestal_calibration/run7/ssdNoise.20070522.042054.root 

 

Looking at the SSD mailing list, I saw that ladder 3 had its HV changed the 21th of may : here ,

which can match with the sudden increase

 ---> The result of the fit for all wafers on p-side is here 

 

•  For the n-side ladders : here for wafer 2 ladder 11 side N

And for wafer 2 ladder 2 side 3 N (these 2 last plots are the N-side from the plots for the P-side )

---> The result of the fit for all wafers on n-side is here

 

part IV :

This plot shows the slopes (as calculated by the fit) for all wafers in a given ladder (here for ladder 3 side P, noted 2P)

Then I evaluated the increase of <rms>, knowing the slope and the max number of pedestal files (170)

So for wafer 3 , we have :

• an increase of <rms> by ~78 adc (according to this plot)
• which match the blue curve (section part III) where at the beginning of the run, the <rms> for this wafer is ~50 adc and at the end of the run, the <rms> is approximatively 120,130 adc
• then the real noise increase for this wafer is 78/16 = 4.875 adc
• we can also deduce a global increase per ladder : here the overall increase for ladder 2 would be ~ 70 adc (4.375 adc)

 And finally a pdf file for all ladders P : here