On Sat, 16 Mar 2002 10:05:26 +0000 (GMT)
Geoff Blake <geoff@palaemon.co.uk> wrote:
> As the day job is/was an RF (electronics) engineer, I know a little
> about this :-)
> Andrew is absolutely right, except that the whole system is dictatated
> by the speed of the slowest device. If you have 160MHz (Mb/sec) disks
> connected to a SCSI bus running at 5MHz (why?) then you only need to
> treat it as a 5MHz system.

> At 160MHz, the "length" of a data bit on the ribbon cable is probably
> around 1 metre allowing for some velocity factor (in case there are
> other RF engineers out there) and the reflections from even a short
> (say 10cm) length of unterminated cable will cause significant
> distortion of the data edges. 
> At 5MHz, the data "length" is around 30 metres and 10cm is somewhat
> less significant.

Basic transmission line theory. Spot the engineers :)

I was a little unclear, I meant in the context of a full modern 160Mb/sec
system. The 160Mb/sec SCSI doesn't run at 160MHz because it is wider than
the 8 bits of the old sparcs - I think 160Mb/sec is ultra-wide ie. 32bits,
so it would only need to run at a sensible 40MHz. Could you imagine trying
to get reasonable propogation characteristics at 160Mhz from unscreened
ribbon cable? If you want to go faster than this there are (mainframe)
devices that talk SCSI-over-FDDI but then you're back to a serial data
stream.

NetBSD dmesg reports my SCSI (on my SS2) to be clocked at 5Mhz wheras
OpenBSD reports it clocked at 7 point something MHz. It should be SCSI-2
(fast) == 10Mhz. Am I missing something?

-Andrew

-- 
sparc sun4c stuff : http://www.lostgeneration.freeserve.co.uk/sparc
personal email    : bob at lostgeneration dot freeserve dot co dot uk