CDF COT - October 2001 Shutdown

COTASDQ overview:

The COTASDQ calibration procedure is a set of front-end code which allows Run Control to take calibration runs with varying control voltage (CV) settings of the ASDQ boards, writing single-channel hit occupancy, time, and width information to text files for easy (PAW,FORTRAN) analysis.

Different runs are taken with the following CV settings:

• NOM --> nominal (default) calibration CV settings
• ATN --> nominal settings, except attenuation ON
• QEN --> nominal settings, except dE/dx DISABLED
• TRE --> nominal settings pulsing EVEN cells only
• TRO --> nominal settings pulsing ODD cells only
• NUL --> null pulsing with TRE/TRO set to 0 mV (off)
• DTH --> discriminator threshold scan (from 700 to 170 mV)

COTASDQ problem list:

The quantities defined below are extracted from the COTASDQ data and used to form cuts which separate problem cells and wires. The resulting list of cells and wires is automatically generated, with each entry headed by the brief mapping:

   superlayer-cell (quadrant,repeater,TDC crate-slot-input): #wire/channel

• N --> average hit number (occupancy) per pulse
  
  N is found for the nominal run (both TRE/TRO on), same and opposite cell on only (TRE or TRO), and the null run (TRE/TRO both off). Additionally, occupancy information from YMon (normalized to unity for each superlayer) is added. These values are written in this order for wires failing one of the cuts. For example:

     SL1-042 (B,189,12-17-1): #00/24 N = 1.112 (1.008,0.007,0.106,1.113) 

  is a wire with nominal N=1.112, same-side N=1.008, opposite-side N=0.007, null pulse N=0.106, and YMon normalized occupancy N=1.113. There are several hit number categories for the calibration pulse data designed to separate and prioritize distinct problems. For nominal and same-side calibration pulses (where we expect N=1.000):
  
  
  • DEAD: N < 0.990
  • WARM: 1.010 < N < 1.100
  • HOT: N > 1.100
  
  (there are sometimes glitches which cause a crate to lose a calibration event, which is why the DEAD cut is not N < 1.000). The following cut:
  
  
  • NULL: N > 0.100
  
  applies to opposite-side and null pulses, which should be N=0.000. In general, real chamber and front-end noise will have N > 1 for nominal and same-side pulses, and N > 0 for opposite-side and null pulses. Crossed u-coax channels will have mixed results between same and opposite side pulses. There is also a special category which usually has ( 1 < N < 2 ) for nominal and same-side, but almost exactly 0 ( < 0.010 ) for opposite and null runs, and so is not noise upstream of the TDC's. It occurs predominantly for TDC channels 15, 39, 47, 67, and 83, and so is labelled "TDC Noise".
  
  There are "low" and "high" occupancy cuts on the YMon data which are 3.5 times the sigma fit from each superlayer.
  
  
• T --> average start time for nominal runs
  
  The average times per superlayer are computed from the data, and both single-wire (loose, 3.5 sigma) and cell-average (tight, 3.0 sigma) cuts are placed (see plots). Very loose cuts are imposed before cell-averaging so that single bad channels do not pull the result. Problems are reported in the error list, along with the superlayer average and cut limit (in parentheses) and the deviation from the average. For example:

     SL2-118 (C,469,06-06-1): #05/29 T = 457.9 (442.4+/-7.0) +15.6 

  where this wire has a time which is 15.6 ns above the superlayer average of 442.4 ns, with a cut of +/- 7.0 ns.
  
  Most of the "bad time" cells are grouped into a few rogue TDC's. I have measured calibration and returning output pulses at the repeaters for these cells, and cannot see the roughly 10 ns deviation. Ron Moore suggested at one time that this could be a via problem. Of course, there's no evidence that the t-zero calibration doesn't successfully deal with these.
  
  
• W,A,Q --> average width for nominal, ATN, QEN runs
  
  Cuts on the nominal width are made as with time (see plots). Additionally, width cuts for the ATN and QEN runs are also placed to make sure that the ASDQ is receiving the proper control voltages. Most cells/wires showing "bad QEN or ATN" are on the border of the cuts, but there is at least one ASDQ which was found to have a bad ATN line set. ATN and QEN cuts are only made for cells and wires passing the nominal width cuts.
  
  
• D --> discriminator threshold 50% point
  
  Using the discriminator threshold scan, the DTH point where the ASDQ is firing for 50% of events is calculated. Cell and wire cuts of 3.0 and 3.5 sigma are made (see plots) as with the time and width cuts. This is meant to check for ASDQ connectivity with the motherboard (DTH will be twice as high as normal if no injected pulse is split down the wire). However, most failures (especially the bad DTH cells) are due to oddly-shaped calibration pulses, and are strongly correlated with bad widths.
  
  Also from the threshold scan, the "oscillation point" is calculated, where the hit average starts jumping up to the maximum N = 8 at low threshold. In particular, the algorithm searches for the lowest threshold value closest to N=1.100, which catches the beginning of the rise in occupancy. No cut is made on the value; our nominal DTH is so close to bottom that all of these "high oscillation point" wires would show up as "hot" in the cuts on N. However, these values are plotted, along with D (the 50% points), in the cell summary plots.