[SunRescue] Proper use of VME scsi controllers?

[Greg A. Woods]

[ On Wednesday, February 9, 2000 at 19:33:32 (+0100), Peter Koch wrote: ]
> Subject: [SunRescue] Proper use of VME scsi controllers?
>
> Active termination??? On a SCSI-I bus? What a waste!
> High quality cables and connectors? Flat ribbon cables    
> with pressed connectors work fine. Sun used them for
> the internal cabling of the 3/xxx chassis, so it must
> be ok!

For internal cabling, yes, this is probably OK.  I'm not an electrical
engineer with detailed knowledge of signalling but I do know that even
one bad connector can fubar even an RS-232 port never mind a 5MHz
parallel data bus.  Noise on the lines of a single-ended bus is very
bad.  Lack of good echo cancellation is fatal.

> Termination power? Sun applied GND to pin 26...

They only made this mistake on a few specific models.  Don't worry if
you don't have one of those models.  Test it if you don't know.

> After all it seems like pure luck, that my machines     
> work at all!

SCSI will often work outside the spec.  It's just not guaranteed to do
so.  This is the mistake most PC people make:  "Well, it works fine for
me so it should work fine for you too!"  Even some so-called engineers
can be fooled as is evidenced by the blatant rule-bending done in the
official SCSI standard.

Sometimes the difference of a centimetre or so in cable length can make
an un-terminated bus work sufficiently well to fool the naive.  Don't
bet on it though -- always follow the spec. if you want reliable
operation.

If you really depend on reliable operation then ditch everything and go
with (high-voltage) differential busses -- there is no real alternative
until you get into optics and FibreChannel.

> BTW: 5 MB/sec async are a joke! I never got more than
> 1.5 MB/sec from a Sun3-SCSI (si) controller and no more
> than 1 MB/sec from the Sun2-SCSI (sc) controller.

The speed of the data transfer is only barely related to the speed of
the signals on the bus.

The data actually flows across the bus at 5MHz, or rather the data
signals on the bus can change state at rates up to 5MHz.  The size of
the blocks transferred, and the gaps between the blocks (whether
introduced by slow controllers or slow target devices, or whatever) are
what govern the overall perceived speed of the data transfers.

> If you plug the "sc" into a 3/160 and attach a MFM
> harddisk with an ACB-4000 (this was sold by Sun!!!)
> you get no more than 400 kB/sec!!!!

Yes, it was a piece of crap.  However it worked pretty good for the time
and was a lot cheaper than all of the alternatives then available.  In
technology these days what was economic a few years ago can now be a
major deficit, even for a hobbyist.

-- 
							Greg A. Woods

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