August 1st

This page is to prepare the future work on the ssd embedding.

Brief summary

This section briefly summarizes the ssd soft in simulation.
More details can be found on : STAR SSD- Softwarepage

Simulation

SSD Database Manager (SDM)

Calibration, Condition : it simulates the detector response (pedestal,noise,dead strips...)

SSD Lazy Simulator (SLS)

Converts the geant hits into sequences of amplitudes on strips : it simulates the charge collection and the charge sharing between the strips.
Out put : sls_strip table

SSD Pedestal Annihilator (SPA)

Read the strips amplitudes, add noise and pedestal.
Substract the pedestal and apply a DAQ Cut.
Output : spa_strip table.

Hit reco.

SSD Cluster Finder (SCF)

Scan the strips on both sides to built the clusters.
Output : scf_cluster table.

SSD Cluster Matcher (SCM)

Association of the clusters found on the 2 sides of each wafer. The charge matching is used to solve ambiguous packages.
Output : scm_cluster table

Status & Exemples

SSD Database Manager (SDM)

Actually, for the simulation concerning the ssd, it requires a timestamp to the pedestal files (the same case for the data : a correct timestamp refering to the day number is required to pick the right pedestal file)
The simulation has a specific timestamp which refers to a 'default' pedestal file that stores for each strip a noise = 3 ADC and a pedestal = 100 ADC.
(Again the same as for the data reco --> see this message : SSD hypernews)
In this case, the sdm_Maker is not use.

Exemple of the sdm_Maker

Until now, 2 tables (stored in root files) are created with this maker. Noise and pedestals values (in electrons) are randomly generated.
To remind a MIP particle creates 25000 electrons in 300um of silicon.
The conversion factor of electrons into adc count depends of the electronic chain and is : conversionFactor = (NAdcChannel)/(ADCDynamic*NElectronInAMip)
conversionFactor = 2^10/(22500*20) = 0.002275
The values setted in the sdm_Maker are :

Pedestal : gaussian distribution with a mean = 100000 electrons and a dispersion of :
1. 10000 electrons for each wafer
2. 5000 electrons for each chip
3. 50 electrons for each strips (P or N)
Noise : gaussian distribution :
1. with a mean = 1400 electrons and a dispersion of 50 electrons for the P-strips
2. with a mean = 2200 electrons and a dispersion of 70 electrons for the N-strips


Figure 1 : Pedestals in electrons	Figure 2 : Noise in electrons

Then, the same values in adc : vv


Figure 3 : Pedestals in ADC	Figure 4 : Noise in ADC

The main remark is that those values are far from the pedestal and noise values for 2005. RunV analysis by Lilian .
Thus a realistic simulation must take account of this and modify the parameters.
From the data, the pedestals are fluctuated between 100 and 150 ADC for the P and N sides, whereas the simulation generates values from 140 to 300 ADC.
The same is for the noise :

in the data, 16*noise is represented in the page indicated before so the noise goes from 3.125 to 5.
We see also the P and N sides don't present such big differences as it was setted up in the simulation : the noise are roughly the same for the strips on P and N side
each ladders have a different behaviour from the other so the uniformity of the simulation does not reflect the reality.

I then plot, as an exemple ,the pedestal and noise distribution for data ( 1 pedestal run ):


Figure 5 : Pedestals in ADC of 1 pedestal run in 2005	Figure 6 : Noise in ADC of 1 pedestal run in 2005

Try a more realistic pedestal & noise values

Here are the new set of values for the sdm simulation that I try:

Barrel pedestal = 80000 electrons = 180 ADC
Noise P = 2000 electrons = 4.5 ADC
Noise N = 2000 electrons = 4.5 ADC

The others values remain the same.


Figure 7 : Pedestals in electrons	Figure 8 : Noise in electrons

Then we see that the pedestal goes to [40000 ; 110000] electrons ; that gives a pedestal [90 ; 250 ] ADC.
The noise goes to [1800 ; 2200] electrons; that gives a noise [4 ; 5] ADC.
These values reflects, cf the web page, a little bit more the real pedestal and noise values .
We can also said that the default pedestal (100 ADC) and noise (3 ADC) are far too from the realistic pedestal ans noise values.

Comparison

In this section, I compare the results of 2 differents calibrations on the physical values of the SSD, such as the size (strips) of the clusters, their signal to noise ratio and the charge matching of the reconstructed hits.

Default calibration

This calibration is the one which the simulation is actually done :

All the ladders are active
For each strip, the pedestal is set to 100 ADC and the noise is set to 3 ADC


Figure 9 : Size of the clusters	Figure 10 : Signal to noise ratio	Figure 11 : Charge on N side vs Charge on P side

Modified calibration

Ladders 7 and 8 are off.
For each strip, the pedestal and noise are set according to the previous section 'Try a more realistic pedestal and noise values'


Figure 12 : Size of the clusters	Figure 13 : Signal to noise ratio	Figure 14 : Charge on N side vs Charge on P side

Conclusions & Oulooks

The maker that simulates the pedestal & noise (sdm_Maker)is obsolete (written in 2003) and is not optimized for CuCu collisions
It is not used too for the simulation that have been done.
The noise and pedestal of all the strips are automatically set
At this point, we have 2 choices for a realistic pedestal and noise simulation:
1. Update the sdm_Maker in order to describe the status of the SSD in 2005.
2. Take directly the pedestal and noise file from the data as input for the simulation
However, the impact on the changes of the pedestale file (default or 'tuned') seems to be only for the clusters reco, eg realistic pedestal and noise values seems to remove more fake clusters (signal induce by the noise of the strips, not by the paasge of a particle).
The size of clusters get also more realistic, as we remove almost those noise clusters : by comparing figure 9 and 12 , we see that the clusters with only one strip appear less time.
It also seems not to affect the hits reconstruction because most of the noise clusters that we could before formed cannot be associated since this case appears mostly on 1 side of the wafer.
Then the association of clusters on both side give no hits.

Comparison the data and the simulation

To compare :

The size of the clusters (large cluster protection in simu AND data is tested)
Noise, ADC and s/n ratio

The next plots show thesize of the clusters for all the ladders ( in blue = P side , orange = the N side)
For each plot :

In the X axis : number of strip : 1 to 4
In the Y axis : the number of entries for this ladder, eg the number of clusters, normalized to all the entries and show in log

Simulation (cucu200 GeV)

Before cluster Matching


Figure 15 : Ladders on p side	Figure 16 : Ladders on n side

Simulation

After cluster Matching


Figure 15 : Ladders on p side	Figure 16 : Ladders on n side

Data

cucu62 GeV

Daq files : st_physics_6073023_raw_2030001.daq
Pedestal file = ssdStripCalib.20050314.80908.root


Figure 19 : Ladders on p side	Figure 20 : Ladders on n side

cucu200 GeV

Daq files : st_physics_6049030_raw_1040001.daq
Pedestal file = ssdStripCalib.20050218.81219.root


Figure 21 : Ladders on p side	Figure 22 : Ladders on n side

First conclusions

For these 2 datasets, the prevention for clusters bigger than 4 strips is actually on and the number of clusters with 1 strip is comparable to the number of clusters with 2 strips.
It seems also that the size of the clusters from the hits actually stored into the StEvent container looks the same (no more differencies as I found before)

Jonathan Bouchet

Last modified: Mon Aug 21 14:44:16 EDT 2006