SPI Tolerance

Jim McGrath jmcgrath at qntm.com
Wed Jun 1 18:03:10 PDT 1994

        Reply to:   RE>>SPI Tolerance

I think I have to agree with Jeff Stai - the appropriate response to Bill's
problem is not to create yet another spec (a tolerance), but rather to have
the devices end up negotiating a slighty lower speed to make everything work.

One point that I do agree with is that we are often very simple minded in
setting our timing expectations.  We should always make some allowance in the
standard setting for common frequencies, tolerances, and physical
imperfections.  Thus we should have kept a minimum cycle time, but maybe
defined it to be 97 ns (to allow for tolerance spread from 100 ns). 
Afterall, the 100 ns number in the standard does not really give any new
information, since it is already the nominal timing number for the fastest
speed - the purpose of an overall minimum is to incorporate any tolerance,
deviations due to cable, power supply, etc....


More information about the T10 mailing list