SPI Tolerance

[Tak Asami (Asami]

Gerry Houlder writes:
> I agree with Jeff Stai's position. The existing words put 100 ns as a lower
> limit (on the period) that cannot be exceeded, and the upper limit is
> unlimited.
 
I agree with what he says, in terms of what SPI document specify today.
I am not so sure if that was the right thing to do.

If 100.0nsec is the absolute minimum cycle time on the SCSI bus, then
assuming you are using +/-250ppm (there you go again) crystal as the
reference clock, YOU MUST NOT DESIGN a target or an initiator that operates
at 10.000MHz clock, or 100.0nsec.  You must use 9.998MHz crystal so that
in the worst (best?) case of +250ppm frequency, you are still 10.000MHz.
 
If you used a 10.000MHz crystal, then half your product have just violated
the standard.
Or, if you did it *RIGHT* (try to find a 9.998MHz crystal), your nominal 
performance is somewhat less than 10MXfr/sec.
 
Is this what we wanted ultimately?  I'll shut up if it is.
 
Tak Asami ================================================================
Western Digital Corp
I/O Product Engineering
Tel: (714)932-7621
Fax: (714)932-6496
E-mail: asami at dt.wdc.com