Date   : Tue, 17 Apr 2007 20:21:23 +0100
From   : adsb@... (Andrew Benham)
Subject: Acorn SCSI board arbitration

On Tue, 17 Apr 2007 17:17:19 +0100 (BST)
"Richard Gellman" <splodge@...> wrote:

> SCSI does, but Acorn's interpretation of SCSI is interesting. Now I
> can't remember fully from memory the circuit so the following may not
> be entirely accurate, but I do remember the data part of the SCSI bus
> is double-tied (tied to both +5V and 0V by differing resistor
> values), and the output from the BBC Micro/Master is gated by various
> parts on the controller board. When it is enabled, and 0's on the
> data bus will drive those bus lines to logic 0, rather than
> open-collector.
> 
> Now, should another initiator on the bus drive its SCSI ID to logic 1
> at the same time as the BBC Micro is driving it to logic 0 as a
> by-product of its own arbitration the two will meet in a head-on
> collision.

The SCSI bus is terminated by resistors to +5V and 0V, but these are
to stop the lines from ringing.

The SCSI bus data lines are bi-directional.  The output drivers are
open-collector (i.e. can only pull down to 0V, else are open-circuit
and the line is only pulled up to +5V by the pull-up resistor).  When
the data line is being used as input, the output driver is not asserted.

Now the SCSI data bus lines use negative logic - i.e. logic 1 is 0V,
logic 0 is +5V.  So during the arbitration phase, each device sets
its ID line LOW (logic 1), all other lines are HIGH.  So there's a
potential problem if other devices on the bus are not open collector
outputs.   *** I think, it's getting late and there's a lot of SCSI
spec to read ***

What I'm not sure is whether the Acorn host adaptor design allows for
reading the data bus whilst asserting its device ID, which is needed
for the arbitration phase.  Again. it's getting too late for this!

-- 
Andrew Benham         adsb@...       
Southgate, London N14, United Kingdom

The gates in my computer are AND OR and NOT, not "Bill"