Sun SCSI termination (& esp0 data transfer failures)

[Paul Evan Matz]

In article <1990Aug13.011541.6424 at rice.edu> devil%TECHUNIX.BITNET at cunyvm.cuny.edu (Gil Tene) writes:
>X-Sun-Spots-Digest: Volume 9, Issue 302, message 9
>
>- Which Suns don't provide termination power on the SCSI bus ?
>  (Does the Sparcstation 1,1+,SLC,IPC ?)
>

The SparcStation 1 and 1+, Sun3/50 and 3/60, all do not supply term power.
And, I think I recall that some 3/60s even have their TERMPWR pin (pin 26)
wired to ground.  This can cause serious smoke if you connect a device
that sources term power; The 386i seems to be the only Sun (non-VME) that
sources term power, as this was required to power the external terminator
plug that was supplied with that machine.  Don't know about the SLC or
IPC, or any of the VME-based machines;  Anyone out there know???

>- If the Sun DOES provide termination power, can setting the terminators
>  on the drive to use the drive's power HURT anything (two power sources) ?
>

A device that has termination installed on it locally usually does not
power it via the TERMPWR lead, but instead via its own local +5V supply.
So, another device suppling TERMPWR will not cause a problem.  The problem
occurs when two devices are strapped to source TERMPWR;  The current
practice lately is for devices that supply TERMPWR to do it through a
diode, so that the two sources of +5V don't "fight" with each other, as
would normally happen if two power supplies were tied together in
parallel.  There is a reference to this issue in Sun's SCSA (Sun Common
SCSI Architecture) documents.  Their basic position is that any device
that supplies TERMPWR must do it through a diode.  This may also be a SCSI
II requirement.

Most devices are strap selectable, so one should be able to pick just one
device to do this function, and it is only really necessary when one uses
external terminators.  It sure would have been nice if the initiator (CPU)
had been the source, though.  But, doing this may have power supply
capacity & EMI implications that are better left alone.

The issue of using device or external termination has gotten more
complicated by seemingly increased sensitivity of the SparcStation's SCSI
bus.  Recently, we connected a 386i expansion box containing a 327M WREN,
150M Archive Viper, and WangDAT to a SparcStation 1+.  After a few days,
the machine started to get console error messages of the type:

      esp0:  3.0 data transfer failure
      etc....

The apparent reason for the problem turned out to be very interesting.
Because the Sparc doesn't supply TERMPWR, it wasn't immediately clear how
to use the external terminator, so instead, the last device inside the
external box was terminated.  This seemed OK since that is historically
how Sun3 shoeboxes where terminated.  I've always wondered about that
segment of ribbon, between the last device and the SCSI out connector on
the shoebox, though.  1-2 Ft. of ribbon past the termination does violate
the intent of the SCSI spec.

Well, anyway, I guess the SparcStation is more sensitive than a S3,
(especially considering that creative SS1 SCSI bus implementation) because
once the termination was moved from the last device to the external
terminator on the SCSI out port, the error messages did stop.  (In this
case, it was necessary for one of the devices to source TERMPWR;  Both the
Viper and WangDAT are strap selectable.  The WREN may be, but I don't have
the manual, so I don't know).

So, folks, consider yourself warned!  Those of you planning to use old
shoeboxes for disk storage on SS1s should consider buying a cache of
external terminators.  [Most new "pizza-box" expansion boxes are using
"micro D" PCB type SCSI In/Out adapters that allow termination right at
the end of the bus, so new buyers, not to worry].

By the way, has anyone out there got a tip on suppliers of external
terminators (:-) ?

Cheers!
 ___________________________________________________________________
|Regards,                           PPG Biomedical Systems          |
|Paul Matz                          One Campus Drive                |
|914-741-4685                       Pleasantville, NY. 10570        |
|{..}!uunet!philabs!ppgbms!paul     ppgbms!paul at philabs.philips.com |
|___________________________________________________________________|