Subject: Re: New to the club...
To: None <port-sparc@netbsd.org>
From: Greg A. Woods <woods@weird.com>
List: port-sparc
Date: 06/18/2001 17:23:50
[ On Monday, June 18, 2001 at 22:53:51 (+0200), Oliver.Bruckauf@gmx.de wrote: ]
> Subject: Re: New to the club...
>
> If you don't obey to these rules you introduce reflections on your bus. For
> that matter any stub causes reflections, so even the Sun way of doing things
> is stretching the limits already. If you have reflections on your bus you
> might be lucky that they eliminate themselves or not so lucky because they
> amplify themselves depending on the rest of your configuration.

Indeed.  If you want to talk about the physical limits instead of the
limits set down in the specifications then it's already almost a fact
that SCSI single ended buses just barely "magically work" as is.

I.e. the specification definitely stretches the limits of any practical
implementation in hardware.  If you build hardware right to the limit of
the specification then there's no real guarantee that it'll actually
work.  The limits in the spec are stretched beyond the physical limits
because few implementations will hit all the specified limits
simultaneously and usually extending beyond a few physical limits will
not immediately cause problems.

(What I meant was that a 10cm stub doesn't stretch the limits of the
spec.  That *is*, IIRC, the limit given in the spec.  :-)

And of course 10cm stub is from the bus wire to the interface chip, so
that's *not* 10cm of ribbon, but somewhat less depending on the disk.

I don't know why they didn't simply outlaw single ended buses a long
time ago -- differential is the only sane way to go, and luckily we've
now got a fair penetration of LVD devices in the market.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>     <woods@robohack.ca>
Planix, Inc. <woods@planix.com>;   Secrets of the Weird <woods@weird.com>