During pp500 GeV Run 12, the luminosity went to record levels.  This has caused some issues with the ZDC's in the RICH and RHIC scalers, mostly because the singles rates have reached more than 1 MHz, leading to large and non-linear singles corrections.

There are a number of types of coincidences and singles scaled for the ZDC.  In Run12, a new electronics unit was put into place, the TCIM, which allowed for much of the discrimination and coincidence logic to be moved out of NIM modules and into presumably more stable custom-built, and remotely settable, electronics.  There are also killer-bitted channels, in which the counting is forced to be dead for X bunch crossings (typically 4-9) after a count fired.  When you have MHz-level counting rates, this will lead these to undercount, and the killer bit logic is really in place only to avoid afterpulsing that occurs with the large pulses that happen in heavy ion runs.  For pp, one should really look at the counts without killer bits.

There are also different channels for the RICH (read out on the platform) and RHIC (read out through the DAQ room) scalers.

In general, the singles correction (assuming uniform population of filled bunches) is:

C = RHIC Clock (in Hz) * (Number of filled bunches/120)

Corrected = -C*ln( 1 -  (EW-E*W/C)/(C+EW-E-W) )

Corrected/EW ranges from ~70% at the beginning of a fill to a value closer to 1 later on in a fill, as calculated at http://www.star.bnl.gov/protected/common/common2012/lumi500GeV/singlescorrections_scalers.txt

With the full STAR trigger scaler system, you can do this bunch by bunch more accurately, but only on the signals (14/8) that correspond to the counts in the datastream.

We were seeing inconsistencies in the singles corrections, and so decided to raise the thresholds on the discriminated signals sent to the RHIC and RICH to bring down the singles rates, without affecting the signals recorded in the DAQ.  This was done at 20:27 on March 24, and reverted at 17:40 on March 26.  The issue is that the singles correction became larger, and, more importantly, to bring the singles rate down to ~300 kHz the threshold had to be raised to 225 GeV or so, leading to an overweighting of the counting towards bunch crossings in which signals from two collisions in the same bunch crossing added in the ZDC's.  This could be seen by a falling ratio of the singles-corrected value in RICH ch. 15 to the fully singles-corrected ZDC counts in the trigger system (monitored by RICH channel 14). 

This did not solve all issues, since in the RHIC scalers ch. 4 (coincidence done in NIM in the DAQ room) counted much lower than ch. 9 (coincidence done on the platform in the TCIM electronics), even though these should have counted at the same rate.  During the access on March 28, we found that the ZDC East was rediscriminated with an updating discriminator, which can lead effectively to killing the second count in the ZDC East if the previous bunch crossing had signal.  This was changed to match the ZDC West during this access.  This is fixed starting after 4 pm on March 28.  At the same time, the width of the rediscriminated pulse on the ZDC East and West was shrunk from ~90 ns to 80 ns.

For the two days from March 28, there still appears to be noise that leads to intermittent ~10% loss in count rates for the first hour to two hours of fills16643 and 16650.  In an access at 11 am on March 30, a dangling cable was removed, and the rediscriminated pulse width widened back to 90 ns.