The question was whether we could improve the EPD performance to improve the minbias-hlt70 (good) trigger rate.

Prior to the EPD changes (<2124017) the EPD performance looked like:

Figure 1: TAC values prior to any EPD changes from run 2149015.  Note - the red line is what would be expected through the DSM, if the slew correction were also removed the blue and red histograms would agree.

Looking at the spread of the means from this run, we find:

Figure 2: TAC means for all the QT32C Channels in the EPD for 21249015, this distribution has a mean of 1384 and an RMS of 40.

The worry is that the west side is ~2 ns earlier than the east side.  So first I set the TAC offsets back to zero (pedestal 21249017, data 21249018).  The fill was ending so the statistics weren't great.

Figure 3: TAC values with zero TAC offset run 2149018.

Figure 4: TAC means for all the QT32C Channels in the EPD for 21249018, this distribution has a mean of 1640 and an RMS of 103.

First I tried to gently pull the average for all channels to 1600 (pedestal 2124027, data 28-29).  This decreased the rate of hlt70/all from about 15% to about 12 - 13%, also the peak of the fastest TAC distribution was more like 2200 (https://online.star.bnl.gov/epd/Signal/21249/21249028.tac_evsw.png).  I then tried to push the average to 1100, unfortunately I didn't use the zeros as my input so these files were nonsense (pedestal 2124030, data 31).  I then did this properly (pedestal pedestal 2124032, data after) and the results improved the hlt70/all rate to about 17%. 

Figure 5: TAC values for run 2149035 after the change.

Figure 6: TAC means for all the QT32C Channels in the EPD for 21249035, this distribution has a mean of 1315 and an RMS of 25.

So this definitely looks like an improvement.  Checking the data, looks like each channel is firing ~same rate as the fastest TAC, so there are no obvious methods to improve things further.  Then I thought perhaps to try and optimize a little the TAC windows for the EPD.

We started with:
E and W TAC Max, min = 2750,300
Tac Diff = 4928,3262

First I tried to increase the TAC Difference by in both the min and max since we are cutting off the "feet" of the TAC difference curve.  This attempt was 21249039 and crashed and burned horribly.  I probably should have realized that adding ~1.7 ns or so on each side was a bad idea.  We then tried to increase the window by 10 (21249041) and that decreased the hlt70/all rate.  So returned to the original values.

At this point MCR wanted to play with the squeeze so it is best to leave the EPD be for the moment.  I thought about changing the E and W max as well - but this can wait for the morning.