Date   : Tue, 06 Jul 2004 17:25:27 +0000
From   : Jules Richardson <julesrichardsonuk@...>
Subject: Re: Acorn Winchester unit

On Tue, 2004-07-06 at 16:24, Mark Usher wrote:
> As far as I can remember the biggest problem with connecting SCSI drives is
> that the SCSI drive uses parity which the Acorn host adapter doesn't do.

Hmm. I know my modern Adaptec board says to disable parity checking if
there's any device on the bus that doesn't support it. That takes care
of the read side of things on a modern system - but how things work on a
write I don't know. I suppose it's assumed that a host controller
*always* generates parity information when writing, and it's up to the
target device whether to inspect it or not.

I wonder if parity information can be generated in hardware "at the host
end" just by monitoring the I/O line? 

E.g.: 

Assume there's a parity generator circuit connected to the 8 data lines
at all times.

If I/O is deasserted by the target (i.e. a write op, host to target),
then couple the parity generator's output to the parity line.

If I/O is asserted (i.e. a read op, target to host), then set the output
of the parity generator to a high impedance state and let the device
control the parity line. Don't do anything with it, of course, as we're
out of free input lines to the parallel port :)

I only have bad timing diagrams here. However, it looks like for a
write, the target only assumes data is valid on the bus when the host
asserts ACK, so providing there's enough time between the target
asserting REQ and the host asserting ACK for the parity generator to do
its stuff (before the target decides that the host has gone bye-bye and
gets upset), then it should work.

Maybe a better approach would be to only control the parity line when
I/O is deasserted and REQ is asserted, just in case there's some odd
condition where I/O can be deasserted indicating a write, but the
target's still trying to drive the parity line...

A 74HC280 chip looks to be the critter to use. Presumably cost ruled out
Acorn generating parity info? (plus the cable run was very short and
under Acorn-controlled conditions inside the hard drive cabinet)

cheers

Jules