Hi Jerry,

here are a few suggestions (obviously not a complete list) for what to
look at to quantify the performance of AFE8 boards. I think all of these
tests are critical, but you can of course argue about the details and
specs. The numbers are based on the belief that the CFT trigger will
function with a 1 p.e. threshold at 0.5% noise rate.

SIFT tests without using the SVX:

- noise rate: measure rate at which discr. fires vs. SIFT threshold
  setting, with low VLPC bias voltage (maybe around 4-5V) and no light on
  the VLPC. Preferably this should be done with all 512 channels at once,
  and two boards operating next to each other. For optimal operation of
  the CFT trigger, the rate should be well below 0.5% at an input charge
  threshold corresponding to about 1 p.e. at gain 35000 (see next point).
- to understand what charge a threshold setting corresponds to, one could
  bias the VLPC to nominal bias voltage, illuminate with 2 p.e. and
  measure rate vs. threshold setting. These curves could be compared
  (channel-by-channel) to expectations from stereoboard (SB) data to
  determine calibration factors for each channel (variation of these
  factors across channels in a SIFT should be negligible?).
  There should be at least one set of SIFT threshold settings
  for the whole board satisfying the noise rate spec described above and
  at the same time giving a signal rate >= x % (for all channels). Here x
  should be the fraction of events you expect to pass a charge cut
  corresponding to about 1 p.e. at a nominal gain of 35000. Ideally x
  should be calculated channel-by-channel using SB data collected under
  the exact same conditions (VLPC bias, LED voltage+width, same LED system
  etc.), to correct for VLPC or LED non-uniformities.

Tests with SVX readout:

- measure the error rate of the SVX readout under operating conditions as
  realistic as possible (2 boards next to each other, with VLPCs biased
  and light on the VLPCs, temp. control system in operation, with and
  without zero suppression, high trigger rate etc.). SVX readout should be
  error-free (i.e. no missing or short events, all channel addresses
  transmitted correctly) and running without having to reset/redownload
  for some significant amount of time.
- Assuming the SVX can be used to measure poisson-spectra with p.e. peaks
  resolved and assuming that the spectra are consistent with SB spectra
  for the same channels and operating conditions, one could use the SVX
  readout to measure the SIFT threshold directly in p.e. (for that
  particular cassette), and therefore skip the calibration step described
  above (even though you'd still want to scale to a gain of 35000).
  I'd probably do the SIFT-standalone test nevertheless to have a
  consistency check.

We also agreed that it would be important to test the AFE at lab3 with a
more realistic DAQ system, i.e. with Sequencer/VRBC+VRB+VBD+L3 rather than
SaSeq+Bit3. We'll set this up using the cassette testing DAQ in parallel
to Paul continuing his SaSeq tests. As a side effect this should
allow Paul to take data >20 times faster than with his current setup.
Ideally you would of course want to see results from this test before
going ahead, but it's not completely clear how quickly we can get this DAQ
to work, so...

Cheers,
Volker