From:	SMTP%"janderson@fnal.gov" 18-APR-2001 19:17:21.14
To:	GRANNIS
CC:	
Subj:	PRELIMINARY bit error rate results

Gentlemen:

	Please let me stress that the following results are 
PRELIMINARY, not final.  However, they're pretty darn good, so 
you'll be interested in them.

Bob Angstadt has revised the bit error rate software for testing the 
readout from the AFE to the SASEQ, including quite a bit of code 
to test for 'bad events' where the SASEQ doesn't issue a trigger.  
Various species of these exist, and in all of them no data is 
transferred from AFE to SASEQ, so that these should not be 
included in the bit error calculation.  We've exercised the code by 
introducing known errors in the readout as opposed to the pattern 
he's looking for, and it catches the known errors as expected.  
However, we haven't yet put in intentional small, random errors to 
see how the code responds to those.

None the less, using two different SASEQ boards in the DAB3 
South test stand, we've run 1E9 bits thru one SASEQ and, so far, 
about 1E8 bits thru the other with ZERO errors.  A very short run 
thru the DAB3 North test stand with one of the same two SASEQs 
yielded much less spectacular results (in fact, they were lousy).  
However, we also found a power supply problem at that test stand 
today, so likely the two observations are linked.

Sounds sorta too good to be true, so I won't claim total success, 
but the code has been worked on for a couple of days now.  I also 
have an independently written piece of code (written by me, not 
Bob), that also found zero errors looking only at the VSVX data 
over about 1E7 bits the other night.  

I should also point out that this is not a total test because only 
Chip ID and address information is checked.  That's only every 
other byte and skews the test by only looking at the FIFO Write 
Clock generated from one edge of the DVALID, not the other edge.  
It will certainly catch 'extra' or 'missing' DVALIDs but won't tell us if 
the BER for falling-edge data is the same as that for rising-edge 
data.

We've also seen that the SASEQ 'bad event' rate, where no trigger 
is issued, seems to be about 0.8%, so that certainly could play 
into other problems people have reported.  The software at Lab3 
certainly has some knowledge of SASEQ bad events, but I do not 
know how picky it is at rejecting them.  This large of a failed trigger 
rate only serves to raise the priority of getting an AFE hooked up to 
REAL sequencers, not just SASEQs.  

Regards
John