Log entry time 18:04:05 on September 3,2009

Entry number 287719

Since we had several hours I wanted to try the "improved" ADC18
library and CRL mods.  The results were, on balance, worse than
what we have been using, so I have rolled back to our old code.  

Another IMPORTANT CHANGE that should be noted is that, per Luis,
the CRLs set the timing board (rampdelay=0, integ=13200).  This 
will avoid a mixed state or wrong timing -- assuming we run HAPPEX
without oversampling it's a good, practical solution.

Regarding the "improved" ADC18, some details of the tests and 
results are in ~apar/dev_bob/adc18_trial_3sep09.txt

Briefly, the tests showed that the new code got rid of all "bad1" 
events at the start of the run (improvement), but the big problem 
is that if the VME crashes (I crash it deliberately) the ADCs 
don't "right" themselves.  The old code causes 1 to 5 "bad1" at 
start of a run, but is somehow more robust when CODA crashes. 

Going forward, it's imperative that people pay attention to the
correlation of Detectors to BCM and of Lumi to BCM.  It's also
important to NOT comment-out the messages about "ADCX unexpected"
in Pan.  They mean something.

My decision is this:

1. We have to do the experiment.  Must be practical / conservative.

2. The old code has been working ok for 1+ week.  We leave it.

3. No more DAQ tests until Christmas UNLESS there are problems
or unless we need oversampling.  Test Crate tests will continue.
Full-system tests await Christmas shutdown.

For Luis: the new code would not fix the few-channel DAC noise
residual.  That's a different problem.  Please check the gain
settings and the 16-bit ADCs (we power-cycled once).