SPI Tolerances

[JEFFE at fdc.mhs.compuserve.com]

                               - - Mail - - 
         June  3, 1994         9:25am   MAIL IS -
   FROM: Jeff Epstein                    Viewed Ntfy Urgent    
     TO: SCSI Reflector                  
SUBJECT: SPI Tolerances                  
   COPY: Richard Moore                   
         Jeff Epstein
         Mike Kosco

>From our point of view at Future Domain:

After all those arguments we still agree with the original complaint. The
failure to allow real-world designs to be "compliant" and still achieve a 10
MB/s transfer rate -- which was the intention of Fast SCSI -- is a deficiency
in the spec, and the spec should be repaired. Why should we have to negotiate
to an artificial (and possibly illegal) rate of X-4 ns, just so the other
manufacturer's device knows its OK to use his data transfer rate of X ns +/-
some tolerance that only he knows? The other side of the coin leads to the
same conclusion: If I have a 40 MHz crystal with a finite tolerance, then the
spec as it is written means I'll have to settle for a rate of 8 MB/s, since
dividing the clock by four risks exceeding the limit by some minute quantity.
This is a 20% penalty, not a 4% penalty!

Richard Moore, Future Domain Corporation
internet: richard at fdc.mhs.compuserve.com