David Nebenzahl wrote:
So I made up a SCSI cable for the scanner I got off the street

(sigh) SayNoMore. SCSI cables need to have controlled impedances.
Even at ASYNC speeds, you can get all sorts of reflections that
can cause havoc.

(literally), an Agfa SnapScan 1236. (Had to cobble up the cable since my
SCSI host adapter has a high-density socket but the scanner has the old
25-pin socket.) The cable's a bit ugly, with something of a rat's nest
of wires soldered together in the middle, but it works. Scanner works fine.

And you know it works at DC, right. How do you know it works
"at speed"?

But since using the scanner, I've found that sometimes--not always--the
scanner seems to be causing a hardware fault that causes Windows blue
screens. Sometimes I can boot up and do dozens of scans; other times the
BSOD appears some time after booting, or when I access the scanner driver.

At first I wasn't sure whether the blue screens were due to the scanner
itself or the scanner driver, but I'm pretty sure it's the former, as I
never get them when I unplug the scanner.

I'm sure part of the problem can be traced to Windows 2000's rather
****-poor SCSI handling. While it is definitely better than Windows NT,
which *really* sucked in that way, it's still problematic.

I have *7* SCSI busses in my W2K box. I don't see SCSI errors
*or* hangs. I'm currently running W2KS SP4 but have run W2K SP4
in the past with comparable performance.

And the fact that I don't have a terminator on the scanner might not be
helping, either. But as I said, when it works I can do many scans with
no problems.

Yeah, and if you have sex without a condom your lady friend probably
*won't* get pregnant -- MOST of the time. :-/

Do you put 66MHz memory on your 133MHz bus?
Do you run with your 5V power supply at 4.4V?
Do you overclock your CPU?

(sigh) If you want things to work as they are designed to work,
then *use* them as they were designed to be used!

SCSI cables can be incredibly finicky. The Standard even describes
*where* certain signals should be located *within* the cable
itself for optimum performance. I don't think this is a case
of picking nits just because there was nothing on TV that
afternoon... (i.e., if they went to this length to codify the
standard's requirements, there is a bonfide reason for doing so!)

I should say that this isn't a critical problem for me. I have another
scanner that works about as well (a Microtek using USB) that works about
as well; the Agfa just happens to be a bit faster and not as annoyingly
noisy. I'm mainly curious why I'm having these problems.

Any guesses? (Educated ones get more points.)

Oh, almost forgot:

o OS: Windows 2000, SP 2

Upgrade to SP4. While you can still find it on MS's website.

o Advansys SCSI host adapter ("fast/wide"), but I'm using regular old
single-ended SCSI II. The Advansys is basically a clone of the old
Adaptec 1542, but with two adapters (16 SCSI IDs) on one card. There are
two CD-ROM drives (Plextor) on the same bus as the scanner.

If there are "BIOS settings" (like the 1542's SCSIselect)
try setting the adapter to run ALL targets in ASYNC mode.
But, fix the cable and terminator!

o Motherboard is an Asus something-or-other, almost 10 years old

Look for bad caps.