Path: utzoo!attcan!uunet!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!ub!dsinc!bagate!cbmvax!daveh
From: daveh@cbmvax.commodore.com (Dave Haynie)
Newsgroups: comp.sys.amiga
Subject: Re: board<>3000 RETRACTION-NEAT NEWSGROUP
Message-ID: <15325@cbmvax.commodore.com>
Date: 22 Oct 90 21:10:27 GMT
References: <10213.AA10213@caleb>
Reply-To: daveh@cbmvax.commodore.com (Dave Haynie)
Organization: Commodore, West Chester, PA
Lines: 55

In article <10213.AA10213@caleb> jdp@caleb.UUCP (Jim Pritchett) writes:

>From the net traffic on the subject over the years, it seems that some
>A1000s would perform well with 2 or even 3 well designed SOTS cards.
>However, most would fail to work properly under the same conditions
>with the same cards.  A very few A1000s were reported to fail with only
>ONE SOTS!  This is why CBM went to the single SOTS rule.  However, the
>first time that I heard anyone claim that this was official policy was
>about 1.5 to 2 years later!

Early Amiga specifications did discuss the logic for supporting multiple
SOTS boxes, but it never gave enough information to do this properly.  They
gave some simple guidelines for making two boxes work together, but even
these were flawed.  The original documentation also had 4 Meg PICs as the
largest version; anyone interested can read about this original stuff in the
early "A2000 Hardware Manual", which was part of the 1.0/1.1 "Phone Book"
series.  They also talked about 86 pin PICs.  All of this was work in progress
done during the A1000 development, indicated as such ("subject to change 
without notice", though in practice, notice was given via Tech Support, 
DevCons, etc.), and was corrected very shortly after the A1000 shipped.

The reason chained SOTS boxes aren't supported is that it is IMPOSSIBLE to 
correctly build multiple SOTS boxes.  The requirements for SOTS boxes logically 
make this so.  The first rule for a SOTS box is that it must buffer the A1000
bus as closely to the edge connector as possible.  This has the unfortunate 
effect of adding two TTL delays, one in, one out, to each additional SOTS box.
If you follow the published timing specifications, which are for the Zorro II
bus, the second box in such a setup won't be able to meet the specified times
for SLAVE*, BERR*, OWN*, XRDY, and a number of other signals.  Double buffers
on the clocks will dangerously skew these clocks with respect to the 
motherboard clocks.  Published data setup times won't be met by the second
box.  The list goes on and on.

The traditional developer solution to this was to pass the signals through
unbuffered.  This is the main reason there's alot of flakiness between
different SOTS boxes in different orderings, etc.

Had the SOTS box design worked like this:

A1000 <-> Buffer <-> SOTS <-> SOTS <-> SOTS <-> Termination
	   Box					    Box

It would have been reasonable to support a chained SOTS protocol, and at least
would have had a chance to work. Unfortunately, that was not the way they did 
things.  And I was busy with the C128 at the time :-).  So the chained SOTS box
concept is and forever will be flawed.  That doesn't mean it can't work in 
any specific situation, but it doesn't mean that it can't be guaranteed in 
any arbitrary hookup by following all Amiga bus timing specifications.  Because
it is NOT supported.

>                                                Jim Pritchett
-- 
Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests"
   {uunet|pyramid|rutgers}!cbmvax!daveh      PLINK: hazy     BIX: hazy
	Standing on the shoulders of giants leaves me cold	-REM