Here is the macro that compares relative variation of ADC slopes and LED peaks on a day-by-day basis SCalSlopeVSday.C. It takes as arguments the led channel number and (nstb,row,col) that specifies a single cell.  The most time consuming part of this macro is projecting the ADC of all of the events to a histogram, given the cell number:

----

TTree* t1 = (TTree*)f1->Get("Tr_adc");
sprintf(cellcut,"nstbADC==%d&&rowADC==%d&&colADC==%d&&led==0",nstb1,row0,col0);
sprintf(ledcut,"nstbADC==%d&&rowADC==%d&&colADC==%d&&led==1",nstb1,row0,col0);
sprintf(cellname,"Celln%d_r%d_c%d",nstb1,row0,col0); 
cout<<"processing "<<cellname<<" on day: "<<day<<endl;
cout<<"cellcut: "<<cellcut<<endl;
c1->Clear();
h1where->cd();
t1->Project("h1","ADC",cellcut);
t1->Project("hl","ADC",ledcut);

----

The List of data files is stored in adcTrlist80_98.txt. Currently it contains adc files from 10 days (from day80 to day89) and there are about 10 runs combined for each day.

script MkSCalSlope takes the list of preselected cells (led_chan*.txt ) from led channel 1 to 63, and generate the job description file SubSCalSlope.bat to process all of the cell in all of the 63 led channels. The result for led channel 1 is here. Note that I have modified the description file in order to submit the jobs to a condor long queue. Short queue only has a 3 hour limit, after that the jobs will be evicted. Long queue has a 5 day limit and lower priority.

The purpose of this analysis is to estimate the size of slow component of gain drift on a large time scale, especially to see if there is common mode of this drift. Because this is the part that will enter the common mode variation of LED events when we are trying to relate this common adc variation to the change of LED pulser intensity.

From the result of led channel 1, we see alomst all of the small cells show a ~15% drop in the ADC of LED events while the slopes largely remain stable around 1, except the cases where the slopes have large error bars ( didn't have a threshold behavior ). SInce most small cells show a stable LED adc counts within a run ( see Chris ), they are more reliable to be used in this analysis to estimate the pulser intensity variation. Furthermore, it's also necessary to exaime the common mode of slope variation for these group of small cells.