Gentlemen:

	Given that most were unclear about the readout time of the 
AFE to the sequencer, I decided to work out the numbers on a 
spreadsheet for distribution to all.  The deadtime of the AFE is a 
combination of the digitization time plus the readout time.  Things 
are further complicated by choices in the clock speeds used.

DIGITIZATION

	VSVX:  Performs eight loops of reading eight bytes of data from 
MCMs; runs at 61 MHz (16.5nsec/tick).  State machine introduces 
a couple extra clock ticks every eight words.  Total:  Approximately 
80 ticks of the clock, or 1.32 usec.

	SVX:  256 ticks of the 18.9 nsec/tick clock, plus about 
300nsec of pipeline collapse time, or 5.14 usec.

READOUT

	VSVX: one chip ID/Status byte pair, plus 64 address/data byte 
pairs, plus three ticks for setup/token pass.  Net: 68 * 25nsec = 
1.70 usec at 40 MHz readout rate.

	SVX:  At 10% occupancy, have 1 ID/Stat + 7 address/data, 
plus 1 for token pass.  Thus, 9 * 25nsec = 225nsec.  Have 8 SVX 
on board, for total of 1.80 usec.

SUM:  5.14usec + 1.70 usec + 1.80 usec = 8.64 usec total dead 
time.  At 40 MHz readout, with a full Digitize cycle, no chance to 
make the 8.00 usec limit specified.


If the readout clock could be made to run at 53 MHz, this would 
reduce the total time to 7.78 usec, which is where the 8 usec 
number must come from.  Well, there's no chance that the readout 
clock will run at 53MHz.  The timing inside the SASEQ plus the 
variance measured between different SVX chips says that is just 
not possible.

It seems to me that for the trigger fibers (but probably not the 
preshower) the SVX never gets anywhere near full scale.  I can't 
recall anything that obviously prevents the system from running a 
shorter Digitize cycle in order to buy back deadtime. I think it 
would be worth talking to Mike Utes about this to see if the 
Sequencer can be programmed for a shorter Digitize cycle, to 
compensate for the slower Readout clock.

Regards
John