SPI Tolerance 2
Bill=Galloway%HW=Stor%Sys=Hou at bangate.compaq.com
Bill=Galloway%HW=Stor%Sys=Hou at bangate.compaq.com
Wed Jun 1 14:12:14 PDT 1994
Bcc:
1) In my previous mail message I suggested a tolerance of +/- .25%
as a tolerance for synchronous negotiation period. I did not
intend to put an upper limit on the period. More correctly I
should have suggested +infinity / -0.25%. I arrived at 0.25%
(2500ppm) because this tolerance is achievable by crystals,
ceramic resonators, and phase locked loops. It probably is not
achievable by RC circuits.
2) I am not opposed to a fixed tolerance such as 1ns (as Bruce Adler
suggested) if that is what the committee wants.
3) I agree that the correct reading of SPI today means a minimum of
100.000000...ns not 99.99999999...ns. I cannot negotiate for a
period of 96ns because that is not defined in SCSI-2, and I doubt
that many targets would agree to a period of 96ns.
4) I believe that most target and initiator implementations today
use a crystal oscillator that is some small integer multiple of
the synchronous period. With these designs, the synchronous
period can only be changed in gross steps. For example changing
the divide from four to five would increase the period by 20%.
I do not believe that it is acceptable to slow down by 20% to say
that I will NEVER be less than 100ns.
5) Because of 3) and 4) above I believe that most synchronous SCSI
devices violate the SCSI-2 specification today. I am trying to
change the specification so that the devices will not be broken.
If we get the specification changed, I can not use that to go
beat up my drive vendor but maybe I can prevent the problem from
happening again.
Gerry, If you believe that 100ns is the absolute lower limit then
your Seagate 2Gb drive is broken unless you have found a 25Mhz
oscillator with 0ppm tolerance.
-
Bill Galloway P.O. Box 692000
Compaq Computer Corp. MS 090504
(713) 374-6732 Houston, TX 77269-2000
billg at bangate.compaq.com
More information about the T10
mailing list