Re: How many drives can a sun3 scsi bus really support?

Robert,

Looks like I got confused about who I was writing to.  Note, even with the
newer improved connectors, the count is still important.  The connector
still dominates the factors that degrade performance.  You have described a
set-up with 11 connectors/

There is not enough potential benefit from changing out the DB-50 connector.
Unless you are very experienced and have the correct tools, you would
destroy the multilayer board attempting such a change.  The only "internal
scsi" connector I know of in a 3/60 is the P4 connector which is used for
the color monitors.  Although they have the same pinouts, I have never seen
any documentation that they use the same controller, or that the P$
connector uses the scsi protocol.  Note that there is no documented
restriction on the scsi address space, so there is no place for the video
card to be on the scsi buss.  If they did use a second scsi controller for
the video buss, there is no configuration option to be able to put a disc
drive on that buss.  It is basically hard coded to be used for video cards
only.

One inexpensive way to get a high number of drives onto your machine is to
get a used AT or ATX case and have an internal cable made by your local
cable monger or a company like CSC for $30-50.  You can also find scsi
server cables in the used bins at used computer stores.  This reduces the
connector count.

Note the engineering for the external data buses were not designed for high
box count.  It was expected that there would be one or two boxes connected.
If high device counts were desired, it was expected that larger cases would
be used.  There is nearly no perceptible impedance  bump produced by an
internal connector.

best
Jim

----- Original Message ----- 
From: <[email protected]>
To: <[email protected]>; <[email protected]>
Cc: <[email protected]>
Sent: Sunday, November 02, 2003 8:11 PM
Subject: Re: How many drives can a sun3 scsi bus really support?


> My sun3 has one db50-hd50 cable for the primary cpu-to-drives
> run.  Sadly, it is 5 feet long.  The boxes used are now 4 x sun411
> lunchbox drives (shortly to be upddated to a 911 box with 4 drives).
> Originally, I could only get a maximum of 3 drives interconnected
> with two 8 '' hd50 ribbon cables (special things I had made up).
> The original terminator was a passive hd50 kind, which, as it turns
> out, seems to be the culprit.  I replaced it with a DEC active scsi
> terminator from an Alpha 7000 machine that I scrapped a while back,
> and that seems to have cured the problem.  Now, the cable length
> is 5 feet + (5x4" box cables) + (4x8" cables) = 112" (9'4"), which is
> working.  I have not tried to add an additional cdrom and its cable
> (12" total), although I am expecting that should work.  In my particular
> case, it appears that the terminator block was the culprit.  The 3/60
> is now up on sunos 4.1.1U1, and hopefully will provide me a platform
> to boot one of the drives in NetBSD and get up to current on that.
>
> This kind of trouble caused me to wring my type 3 keyboards in the
> air, one time, and chuck out a truckload of sun3 vme crates.  But,
> when I ran across this lonely, forgotten 3/60, I got bit by that sunbug,
> again.....(:+}}..... (as it turns out this particular 3/60 had the good
3.0.1
> prom in it so I can run the luncbox 150mb tape drives, which is nice).
>
> Anyone know if there is a way to rebuild the cpu board to have either
> a Centronics style output or an HD50 output plug instead of that
> awful db50 thing?  I was thinking it might be possible to header off
> the internal unused 50 pin scsi dip thingy with say the cable out of
> a 411 box.  That would provide the hd50 output to make things a bit
> more happy.
>
> Thanks for the discussion and noggin'-scratchin'...(:+}}.....
>
> Bob Keys
>