1. dataset: 6 runs on day080, 6 runs on day095

I combined runs that belong to the same day, and plot adc spectrum for each one of the 18 cells in group 1. Then I compared the slopes of a exponential fit to the region above threshold:

                www.star.bnl.gov/protected/spin/yuxip/adcCellslopes.pdf

Fit was from ADC channel 600 to 2000. There are two frames for each cells, corresponding to the above two datasets respectively.

2. Following is the math that shows slope1 / slope2 = gain1 / gain2:

Assume the real energy spectrum obeys exponential distribution;

                 N = N0*exp(-k*E), with the factor k unknown, N is the number of events observed in energy bin E. Using E = ADC*gain as in our convention, then

                                   N = N0*exp(-k*ADC*gain).

                 Plotting N as a function of ADC, then the slope above threshold (shoulder) region is equal to k*gain. Compare the slopes of two datasets, we can show that

slope1 / slope2 = gain1 / gain2. The ratio gain1 / gain2 can be equivalently calculated by the number of triggers above a certain threshold ADCth, as a fraction of the

total number of events, namely

                                  ln(N/N0) = -k*ADCth*gain.

                 By fixing the same ADCth for each dataset, and compute ln(N/N0) in which N the the number of events above ADCth, we can show that

                                 ln(N1/N10) / ln(N2/N20) = gain1 / gain2

3. Here is a comparion of the gains.  www.star.bnl.gov/protected/spin/yuxip/adcCellslopes_1_095080_6.pdf

For each cell, there are 4 types of gain factors.

                  gcorrOld:  gain correction factors calibrated for run12080001

                  gcorrAdc:  calculated by the above method using slopes. ( from adc spectra of 6uns on day080 and 6 runs on day095)

                  gpredicted: by following the mean ADC variation of LED events. (with 6 run from day095 combined, then compared to the baserun 12080001 )

                                       gpredicted = nomianl gcor for baserun*( 1st pink point / mean of the rest 6 pink points), where the 1st pink point is the mean LED adc of

                                       baserun

                  gdynamic:  following the mean ADC variation of LED events, but taking into account of the change of pulser intensity, which was described by the blue

                                        dots.  gdynamic = nominal gcor for baserun* (mean of 6 blue points / mean of 6 pink points) , 6 pink points are the ones from day095

  My conclusion is that for those cells which have enough number of triggers above threshold, therefore a relatively good exponential fit, the gain factors calculated

from slopes of adc spectra are more close to gdynamic.

ps. Here is the LED channel map: www.star.bnl.gov/protected/spin/yuxip/ledmap/

       and the related macros:  www.star.bnl.gov/protected/spin/yuxip/macro/. This can be explained later, if necessary