Subject: Re: SCSI terminator (Re: Internal drives in a 3/80)
To: NetBSD/sun3 Discussion List <port-sun3@NetBSD.ORG>
From: Greg A. Woods <woods@most.weird.com>
List: port-sun3
Date: 05/29/1999 13:22:35
[ On Saturday, May 29, 1999 at 08:03:33 (+0900), Makoto Fujiwara wrote: ]
> Subject: SCSI terminator (Re: Internal drives in a 3/80)
>
> >                                   Sat May 29 07:56:59 1999
> >                                                makoto@ki.nu
> magnus> The "correct" way to connect the drives would be:
> magnus> motherboard (1) -> short SCSI lead ->  drive (1)
> magnus> motherboard (2) -> short SCSI lead -> (terminator) -> drive (2)
>  
> What I saw in the old days are:
> 
> motherboard (1) -> short SCSI lead ->  drive (1)
> motherboard (2) -> (terminator) -> short SCSI lead -> drive (2)
> 
> Although either which (1) or (2) has the (terminator) doesn't matter.
> 
> I can't remember whether (1) or (2) should have another
> terminating register inserted in the hard disk drive.

The old Sun Hardware Reference says somewhat confusing things:

14) I occasionally get strange SCSI errors and other disk flakiness with
    the internal drives on my 3/80.

        The SCSI bus is routed internally to three locations (two disk
        connectors and the external connector), which sometimes causes
        problems. According to one report, a Sun "in-line" SCSI
        terminator on one of the internal disk connectors will solve the
        problem.

32) Should internal SCSI drives be terminated?

        It depends. For older systems (Sun-2's and Sun-3's except the
        3/80, non-desktop SPARCs), the only difference between internal
        drives and external drives is which way the cables go. They
        should be terminated as normal for SCSI.

        For the 3/80 and desktop SPARCs (SPARCstation 1, 1+, 2, 4, 5,
        10, 20, IPC, IPX, etc.), there is some confusion. In some cases,
        SCSI terminating resistors are present on the motherboard
        itself, and internal drives are regarded as being close enough
        to that terminator not to need their own terminators. In other
        cases, there may be a terminating feedthrough connector in-line
        with a drive. Or none of the above may be present.

Strictly speaking, and in deference to the SCSI specifications, the
internal drives are on "stub" connectors (i.e. < 10cm), and neither
should require either termination.  However unless the internal
connector cables are really short then the their length, in combination
with the traces on the motherboard, may exceed the maximum length the
spec. allows and if anything else is marginal then the bus won't work
very well.  Adding additional termination may be a problem, though it
may be somewhat offset if the drive implements on-board active
termination and is supplying its own termination power (i.e. newer
drives might be OK with on-board termination enabled, and may even need
it).

In the newer desktop "pizza-box" sparcs (5's, 10's, and 20's, etc.), at
least one internal cable is often much longer (often to reach the
CD-ROM), and since they're also possibly using the "FAST" option, proper
termination is much more critical.

-- 
							Greg A. Woods

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