NetNews Usenet Archive 1993 #1/NN_1993_1.iso/spool/comp/sys/3b1/4231

Text File | 1993-01-10 | 3.1 KB | 59 lines

Newsgroups: comp.sys.3b1
Path: sparky!uunet!utcsri!torn!spool.mu.edu!introl.introl.com!booth!roger
From: [e-mail] (Roger Abrahams)
Subject: Re: rs232 computer connection
Message-ID: <[e-mail]>
Organization: Private System
References: <[e-mail]> <[e-mail]>
Date: Sun, 10 Jan 1993 15:32:59 GMT
Lines: 48

In article <[e-mail]> [e-mail] (Rob Stampfli) writes:
>In article <[e-mail]> [e-mail] (Michael R James) writes:
>>I have my 3b1 configured to answer on the serial port
>>with a 486 clone running OS/2 on the other end.
>>Problems:
>>1) I'm not getting 9600 baud over this connection, the line
>>appears to very slow with a lot of lag.  But file transfers do work.
>>
>>2) Shutting off my 486 crashes the 3b1, although I can disconnect
>>the serial cable and then shutdown with no effect (on the 3b1).
>>
>>Any ideas what my problems might be?  Maybe this isn't new but
>>the FAQ doesn't address these problems.
>
>This is from memory, so it may not be 100 percent accurate.  There was a
>design bug found in an early version of the 8530 chip that was used in some
>later versions of the serial and combo cards.  The card was redesigned to
>cut costs, and the redesign uncovered the bug.  I'm not sure of the revision
>number of the affected cards, but I know the "E" series was clean, so the
>problem was introduced after that, I think around "G" or so.  If you have
>a board that has one silver box instead of two crystals on it, it has the
>potential for being affected.
>
>The bug manifests itself by allowing the card, when in use, to "ring",
>causing many high level interrupts of the 68010 in quick succession.  This
>has the effect of consuming all the processor's time, effectively locking it
>up from the UNIX-PC's perspective.  Of the "affected" batch of cards, only
>some actually exhibit these symptoms.  The rest are "on the edge", but work.
>The solution is to replace the 8530 chip with a newer one where the bug has
>been eliminated.  The old 8530 is still good for service in devices that
>don't stress it in this particular way, like older serial cards or DOS boxes.
>(But, obviously, don't just swap chips with the older cards, since they
>don't have the corrected 8530s.)
>
>If you see your Unix-PC just lock up (LEDS aren't even blinking), try this:
>Disconnect the RS232 connections, one by one, from the serial cards.  It
>might take the machine several seconds to recover from processing the queued
>interrupts, but when you disconnect the right cable, the machine will
>eventually come back to life in most cases.  (If it has been locked up for
>an exceeding long time, a stack may be blown and a reboot required.)  If,
>on the other hand, the machine starts working again, you have found your
>problem.  AT&T used to send you a free good 8530 if you complained that you
>had these symptoms.  I doubt the offer is still good.
>--
>Rob Stampfli  [e-mail]      The neat thing about standards:
>614-864-9377  HAM RADIO: [e-mail]  There are so many to choose from.