Re: [N8VEM-S100:982] i960 / V96BMC ~BLAST timing

[Rob Doyle <radi...@gmail.com>]

On 7/16/2012 6:04 PM, John Monahan wrote:
> We could go with a daughter card Rob, we did one with our SMB (a far
> simplier situation of course). The only reason I like the over the top is it
> has a better potential to be a bus in itself with perhaps other cards joined
> up later.   Width is not really an issue. Could we not have alternating
> ground lines.  It's total length would potentially be less than seen on a
> typical old 386 PC motherboard.  I would anticipate having something like a
> 74AS645 and 74AS533 to sharpen up the signals as they arrive on the daughter
> board.
>
> No I have no tools. Suggestions?

The '386 ran with a 33 MHz memory bus speed.

Interlacing grounds will help with cross-talk.

At slower speeds you can think of interconnections as wires
where everything happens simultaneously.

At these speeds the interconnections are transmission lines.
Signals get launched from the source and travel down the
transmission line to the destination.  There will be a
reflection at every impedance discontinuity - that includes
the destination.

You'll need to ensure that the PCB trace, connector,
ribbon cable, everything else all have the same impedance.

I wouldn't know how to do that without a four layer card
and microstrip [1] transmission lines.

I guess PATA disks ran at that speed but even then,
the maximum distance between source and sink was 18 inches.
You should be able to stay within that, if you're careful.

Currently we use IBIS simulators [2] but the '386 precedes
that.   One would probably have to resort to SPICE.

[1]http://en.wikipedia.org/wiki/Microstrip

[2]<http://en.wikipedia.org/wiki/Input/output_Buffer_Information_Specification>

Rob.