Date : Wed, 24 Nov 2010 02:19:55 -0000
From : profpep@... (Mike Pepper)
Subject: Harddisc fakery using a microcontroller
> People had always surmised since the 90s that the host adapter board
doesn't work directly with a scsi drive due to lack of ATN implementation.
Yet here we are with the very board working quite happily with a scsi drive.
>
> I don't think though the circuit diagram for the host adapter is included
in the filestore service manual.
>
The diagram is certainly in the "Winchester 110 and 130 Service Manual";
I've just been looking at it. I have a Seagate 80Meg half height drive here
running on an Oak copy of the Acorn HBA. and the same circuit works with the
Viglen one too, (I repaired one for a group member a while ago).
There were two issues I remember, from my last serious work on this,
(setting up the above mentioned HDD). One was the lack of the ATN line, and
the other the lack of a parity line. I seem to remember being told, (perhaps
by JGH), that the parity could be 'fudged' if needed, (by generating it from
the actual data). The lack of ATN seems partly because of it's beginnings in
the era of SASI, (I had a couple of Xebec HDD's from Acorn systems that had
the edge connector rather than the 50 pin as on a later SCSI), and partly
because the Acorn Host Adapter doesn't try to do full multiple target SCSI.
I do believe 2 devices are possible, was the tape drive on some of the file
servers on the same SCSI bus as the HDD? I think that the lack of ATN is not
too critical, if all one wants is a disc or tape drive.
Worst bit of getting my drive going was getting the correct incantation to
do a mode selsct on the drive to tell it that it had 256 bytes/block, not
the default 512. All SCSI drives are supposed to respond to that command,
from my experience at the time, some do, and some don't. Earlier IBM's and
Seagates did. I couldn't get Connor's to respond, though the manual said
they did. These days, it's easier to get a copy of the old Western Digital
SCSI tools, and change the block size on a PC, (though not so purist).
The Acorn 50 pin connector is a scary thing. It has 5V on pin 34 which the
spec says should be grounded, it grounds 23,25, and 27, (23, 27 are
reserved, and 25 should be open). They aren't the only culprits though,
quite a few devices do similar things. My old Archive DAT autochanger had
the problem, (it is common to find Akai samplers with the Termination Power
Fuse blown by rogue SCSI CD-ROM drives, and then they won't start up without
a terminator on the bus, even if no devices are connected. A good solution
is a 'Polyswitch' to replace the PCB mounted, wire ended fuse). The host bus
card for the Master 128, as used in the AIV system looks a bit better, from
it's diagram, though I've never had my hands on one, (BTW - can anyone tell
me what the 16MHz oscilator on the drawing is for? I can't see it as part of
the SCSI system).
It's a pity the old NCR SCSI chips are hard to get now - years ago RS used
to sell them. I used one or two, and they had most of the work done for you,
and interfacing became more a matter of loading chip registers. I did get
one to talk to a 1MHz bus, but my assembler skills weren't up to adapting
the ADFS to use it.
If you want to be really hardcore, there is a copy of the SCSI-2 spec here:
http://ldkelley.com/SCSI2/index.html
or a 500 page PDF here:
http://www.t10.org/ftp/x3t9.2/drafts/s2/s2-r10l.pdf
(Sorry for the scrawl, my mnemoic memory just wenrt into replay mode).
||\/||ike