While running at BNL, it was noticed that there was a diagonal line when looking at the correlation between two tiles.  Essentially, they would both fire, and with ADCs that are strongly correlated.

Figure 1: Example correlation between two tiles on the left, example correlation between a tile and the empty tile at tile "zero".

In Figure 1, we can see an example of this.  As expected, one can see that pedestal and mip peak for each tile, completely uncorrelated with the signal in the other, except for the points along the diagonal.  In fact, this was even seen in the first channel when running on evens (which we have called tile 0).  Since there is no tile, or fiber, there is no way for there to be a signal.

The entire set of correlations can be seen at: drupal.star.bnl.gov/STAR/system/files/Histos12142017_EPD_Diagonal.pdf
It should be noted here that the last 4 tiles of the bottom SS were removed from the ADC in slot 7 to the empty ADC (given the labels blank0, etc).  In fact, if we select events in which tile 0 in the top supersector had a significantly higher than pedestal value (adc > 250), we see that we can pull this diagonal correlation from all channels, other than those from the "empty" ADC: drupal.star.bnl.gov/STAR/system/files/Histos12142017_EPD_Diagonal_SelectedonTile0.pdf

What was noticed is that the last ADC, which had been empty, did not show this characteristic diagonal correlation.  This is true even with the data being added to it.  (We noted that the empty channels in the previously full ADC do show it, but none of the channels on the empty one showed it either before or after putting the data into the system).  One difference was the timing, the first 4 ADCs were 28 ns behind.  Another was that the logic feeding these ADCs, it looked sort of like the 4-fold coincidence was a 3-fold (at least it was firing more often than the other, and we could not put the level any higher in coincidence), so we moved this cables into a more stable one and verified that both sets now fired precisely the same.  We also removed the 28 ns of extra cable, so everything was a the same timing.

After doing this, the diagonal correlation seems to go away:

Figure 2: Two correlations after our fix which do not seem to show the diagonal line.

In Figure 2, we do not see any evidence of this correlation.  On the left is the correlation between tile 14 on the top and in the middle, so the correlation in the middle is from true cosmic rays.  On the right is the correlation between tile 14 on the top and tile 6 in the middle.  One can see the pedestal and MIP peaks for both, which are mostly uncorrelated as expected.  (The few points in the middle could be from diagonal cosmics.)

The full range of correlations from these channels can be seen at: drupal.star.bnl.gov/STAR/system/files/Histos12152017_EPD_NoDiagonalProblem.pdf