Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!usc!apple!coherent!dplatt
From: dplatt@coherent.com (Dave Platt)
Newsgroups: comp.sys.mac.hardware
Subject: Re: Internal HD scrambled by SCSI device
Message-ID: <74129@coherent.coherent.com>
Date: 21 Sep 90 18:31:44 GMT
References: <1990Sep19.175620.15824@morrow.stanford.edu>
Reply-To: dplatt@coherent.com (Dave Platt)
Organization: Coherent Thought Inc., Palo Alto CA
Lines: 72

In article <1990Sep19.175620.15824@morrow.stanford.edu> rex@pangea.Stanford.EDU (Rex Sanders) writes:
> Setup:	Mac II with 80 Mb Rodime internal HD
> 	new DPI 44R (Syquest) cartridge disk on external SCSI
> 		internally terminated.
> 
> Problem:  Turn OFF power to the DPI, but leave everything attached.
>           Power up Mac II.
> 	  BOOM - 80 Mb internal HD scrambled.
> 	  Spend 2-3 days recovering files using SUM II.
> 	  Could not recover from backups - that's what we attached the DPI for!
> 	  Repeated problem (accidentally) twice.
> 
> 
> Called DPI - they were very apologetic, but said this was a "feature" of
> SCSI.  Said that some manufacturers had (expensively) engineered around
> this problem.  NO MENTION of this problem in (very slim) DPI manual.
> 
> I'm certain I've powered up a number of Mac IIs with non-DPI SCSI
> devices turned off, including scanners, slide makers, CD-ROMs, and
> external hard disks.  I've never seen this problem before.
> 
> Is DPI, um, correct?  How could BMUG recommend a drive with this
> "feature"?

One of two things could be occurring here... one is easily correctable,
the other isn't.

[1] The DPI's SCSI interface (the actual bus-transceivers) may be
    "jamming the bus" when the drive's power is turned off.  This is
    often due to the overvoltage/surge-protection circuit in the
    transceiver... it ties the bus-lines to the +5 power supply rail via
    a diode, in order to shunt off any spikes on the bus.  When the
    drive is powered off, the +5 rail sinks down to ground, and the
    protective diode effectively short-circuits the bus.

    A design of this sort _is_ in compliance with the SCSI-1 standard,
    because the standard explicitly forbids having powered-off devices
    on the bus.  It's generally acknowledged that this is an annoying
    design, however, and most manufacturers are switching to protective
    circuits which do allow the device to be powered down without
    shorting the bus.

    This problem is not easily correctable... it's inherent in the SCSI
    transeciver design.

[2] The DPI's internal terminators are designed to accept power from the
    DPI's power supply.  It's quite possible that these terminators
    _cannot_ also accept power from the SCSI bus if the DPI is turned
    off... that takes extra parts (one fuse, one diode, total parts cost
    less than half a buck).  If this is the case, your SCSI bus will be
    misterminated if the DPI is off.  This will seriously scramble data
    being sent over the bus;  if your hard disk doesn't have
    parity-checking enabled, it could end up exhkuyrtinq gArvadge undt
    kerup!ingue@thu dusk.

    This problem is easily correctable.  Remove the internal terminator
    resistors from the DPI (they'll be a set of two or three SIP or DIP
    resistor packs, plugged into sockets near the 50-pin ribbon cable
    header).  Buy an external terminator, and plug it into the spare
    50-pin CHAMP connector on the DPI case.  External terminators will
    accept power from _any_ device on the bus... and the Mac II provides
    terminator power.

    If the DPI uses DB-25 connectors for its SCSI jacks (ugh!), then you
    may have grave difficulty finding an external terminator which will
    fit them.  It's easier to buy a short DB25-to-CHAMP50 cable, plug it
    into the second DB-25 port, connect a CHAMP50 external terminator to
    the other end of the cable, and insulate the other side of the
    terminator with black electrical tape.

I rather suspect that you've run into situation [2], and that switching
to external termination will solve your problem.