SPI Tolerance

Tak Asami (Asami asami at dt.wdc.com
Fri May 27 18:16:15 PDT 1994

Bill Galloway of Compaq says
> There is no tolerance specified for the negotiation period in SPI.

and he be right.  In fact, there is no timing tolerance whatsoever 
specified in the SPI or, for that matter, SCSI-2.
I believe the timing specs in the documnet were supposed to be interpreted
as "minimum".  In other words, when it says 100nsec, they mean 100.00000nsec
minimum.  Your 99.776nsec adaptor is technically in violation of the standard.

Having said that!!!!
You are right, that you can buy 10.000MHz +/-250ppm crystal, but not 0ppm!!!
You can buy a 9.75MHz +/-250ppm crystal so you can be 100nsec worst case,
but that probably is a special custom frequency and will cost you $$$.

Yes, I agree with you that SPI should have timing tolerance specified.
Most of the time when we look at SCSI timings, we look at them as
"nominal" timings rather than min/max.  Let's document the spec as 
(nominal timing) +/- (tolerance).

And I think 250ppm is a good limit to start with.

I can submit such comment in the next round of SPI public comment period, 
but am interested in hearing how SCSI community at large feels about this.

Tak Asami ================================================================
Western Digital Corporation
I/O Product Engineering
Phone: (714)932-7621
Fax:   (714)932-6496
Internet: asami at dt.wdc.com

More information about the T10 mailing list