1. A follow-up of LED and pion mass vs day (from day80 to day98): www.star.bnl.gov/protected/spin/yuxip/LedVSday89_90_All.pdf

     only imposed an energy cut on pion candidates: abs(E12-30)<10GeV, including both small and large cells  www.star.bnl.gov/protected/spin/yuxip/dPionDay_All.pdf

     pion mass varies within 2%, while LED drops monotonically all the way to day98.

2. I tried a simple correction scheme to eliminate LED intensity drop when making time dependent gain corrections.

    In the previous t-dependent correction method gcorr(event X) = gcorr(baserun)*NominalLED(baserun)/LED(event X). Now in the method that I tried, the NominalLED is updated

    on a day-by-day basis, gcorr(event X) = gcorr(baserun)*NominalLED(day Y) / LED(event X), where event X belongs to one of the runs on day Y. NominalLED on day Y was read

    from the blue dots in the first graph.

   For example, I analyzed 5 runs on day89, base on the calibration made for day80. In the following graph pion energy is from 20 to 40 GeV.

   a. Small cells day80 (baserun),                                                                                                                                 mpi(small) = 0.132 GeV, width = 0.023

        Large cells, day 80                                                                                                                                                  mpi(large) = 0.150 GeV, width = 0.037

   b. small cells, previous method with t-dependent gain correction but no correction to LED intensity.         mpi(small) = 0.137 GeV, width = 0.028.

       large cells,                                                                                                                                                                  mpi(large) = 0.152 GeV, width = 0.042

   c. small cells , with t-dependent gain and LED intensity corrections.                                                                 mpi = 0.130 GeV width = 0.023

       large cells                                                                                                                                                                   mpi = 0.144 GeV,width = 0.042

   d. small cells, without any correction                                                                                                                         mpi = 0.129 GeV width = 0.024

       large cells                                                                                                                                                                    mpi = 0.156 GeV width = 0.047

   Because of limited data being analyzed, there is no clear evidence that the new method has made any improvement. It's also necessary to look at more days later than day89.

   I also have a LED channel by channel correction algorithm, which updates the nominal LED by looking at the daily variation of all the cells wthin the same LED channel. But it turns

   out to produce much worse results than not doing anything with the LEDs.

3. I looked at the pion mass and LED variations within a fill, as there was indications that the actual gain might be correlated with fill.

    Here is the run-by-run pion mass and LED variations for all the runs on day90. X axis is the relative run#, where 0 indicates baserun 12080001. From 1 to 10 are the 10 runs in fill

    15372,  while 14 to 27 are the 14 runs that belong to fill15376. We see that pion mass (red dots) decreases linearly within a fill, and LED events is correlated with pion mass. This

    indicates that LED can be used to make run-by-run corrections.

    Considering the overall decrease in the ADC of LED events and the relative stable day-by-day pion masses, I simpy set the nominal LED to be the mean LED of the 1st run of

    fill15372. Here is the recalibrated run-by-run pion mass for this fill. It shows that we can get more uniform gains simply by ignoring the day-by-day/fill-by-fill LED variation.