[sebhc] Good news, bad news -- Robin succeeds; Message too long (>20000 chars)

Barry Watzman Watzman at neo.rr.com
Mon Mar 27 15:11:21 CST 2006


Story time.

Sometime in 1980 or 1981 (don't remember exactly when, but Don Moffet was
ZDS President at the time) we had a "panic" because Z-89's were crashing and
locking up in final test for no apparent reason.  There was circumstantial
evidence that the serial ports, specifically the 8250's, were involved
because sometimes the machine was found to be not locked up completely, but
rather just seemingly dead because the 8250's had spontaneously reset and
had to be reinitialized (finding out just this much was a major
accomplishment).  Production lines were shut down, we had engineers from
National Semiconductor working on the problem, we had meetings in Moffet's
office twice a day to review progress/status, and the engineers working the
problem were on 7-day a week schedules.

Interestingly, there were a few reports of similar occurrences in the field,
but not the number that we would have expected, and almost always in
brand-new machines.

Bottom line:  The problem was caused by arcing inside the CRT from "crap"
(could be microscopic, invisible crap) that was initially present inside a
new CRT.  If the crap got into the wrong places (the 2nd anode voltage in
the 89 is up near 10,000 volts) it caused an arc.  The arc could be fairly
violent, and could distribute itself both via the CRT wiring and also
wirelessly, by electromagnet radiation, and it could reset the serial port,
the entire computer or anything else, as well as cause a "glitch" in memory
or the CPU.  But it was transient and generally non-destructive.  And the
arc destroyed the "crap" inside the CRT, kind of like a bug-zapper, so the
problem was self-resolving ... it happened in new units, but each hour of
operation reduced the chances of it's ever happening to that unit again
(although shipment could redistribute un-zapped crap inside the CRT and
cause a brief return of the problem).

After a period of much hand wringing, shipments were resumed.  I don't
remember if we did anything other than, perhaps, to slightly lengthen the
burn-in duration.

Barry Watzman
Watzman at neo.rr.com


-----Original Message-----
From: sebhc-bounces at sebhc.org [mailto:sebhc-bounces at sebhc.org] On Behalf Of
Dan Lanciani
Sent: Monday, March 27, 2006 3:37 PM
To: sebhc at sebhc.org
Subject: Re: [sebhc] Good news, bad news -- Robin succeeds; Message too long
(>20000 chars)

|The cause of the lack of serial comms was actually that the 8250 itself
|was faulty. Talk about bad luck; I had originally discounted the serial
|card having a hardware fault because I tried swapped serial cards
|between my two H89s and this had not fixed the fault. As it turns out,
|BOTH the serial cards had duff 8250s in the LP port circuit !

Back when I first bought a (new) 4-port serial card for the H8, 3 of
the 4 8250s were bad.  I don't remember the exact symptom, but it wasn't
a complete failure.  I think they locked up after a few characters.  Heath
did not believe me.  They were willing to send one replacement but no
more.  I wrote several letters explaining that the good 8250 could be moved
to any of the four channels and would work fine, so the problem was likely
not anywhere else on the card.  They pretended (at least I hope they were
pretending :) to be dumb and claimed that instead this proved that I must
have made a mistake assembling the card.

Later I think I saw the same problem on an H89.  So either there was 
a bad batch of 8250s or their design was pushing some timing parameter
to the edge and only the "best" chips worked.

				Dan Lanciani
				ddl at danlan.*com
--
Delivered by the SEBHC Mailing List


--
Delivered by the SEBHC Mailing List



More information about the Sebhc mailing list