A2091+Rev 6 A2630 = CRASH. Please help!
Klaus Weber
For the last two years I used a Nexus PIO-SCSI-Host-Adapter
(Polling, i.e. not even interrupt-driven IO) and had no problems
with it. Finally, I decided to buy a DMA-Host-Adapter (A2091) to
improve overall system performance. After installing the A2091, I
experienced system crashes during Harddisk writes every now and
then. At first, I thought the RDB format of the Nexus H-A was
slightly incompatible with the A2091, so I backed up my Harddisk,
reformatted/-partitioned it with the A2091 and restored all files.
Even during the restore process I experienced 10-15 system crashes
(with about 70-80 Megs of data to be restored, to give you an idea
about the frequency of the crashes). This was strange, because I
had booted from my original 2.04 Workbench disk, and everything
should have worked fine.
To make a long story short, it seems that the crashes-during-write
problem is a hardware problem.
Symptoms:
- System crashes when writing to the Harddisk, if the Host-Adapter
uses DMA. System works fine if the H-A uses PIO.
- The A2091 works fine in another A2000 with an Rev 9 A2630, so I
would say the A2091 is O.K.
- The System works fine (and slow :-<) if the A2630 is removed.
- The System works fine if the A2630 is present, but the A2630 RAM
is disabled (by setting J303).
- The System _seems_ to work fine if the Rev 6 A2630 is replaced by
a Rev 9 A2630 (I could not test that as thoroughly as I wanted
to, so I would say I'm only 90% sure, but maybe I was just lucky).
- The system _seems_ to work if
* The Rev 6 A2630 and the A2091 are the only cards in the system.
_and_
* The A2091 is put in the first Zorro-II slot.
The system crashes again if
* Other Zorro-II cards are present.
and/or
* The A2091 is put in another slot (i.e. more to the left). The
higher the slotnumber, the more crashes I get (it's hard to
measure though :-<)
From what I have tried, I would say this is a timing or noise problem
with DMA to the Rev 6 A2630 RAM, but then, I'm not a technician.
Configuration:
A2000:
- Rev 6.2 Motherboard
- V37.175 ROMs
- J900 is set.
A2091:
- PCB Rev 4.0
- V6.6 ROMs
- Serial #: 0100892312769001111833
- Memory jumper set to 0 Meg.
- JP 5 (Options): all jumpers removed.
- Harddisk is correctly terminated.
A2630:
- Rev 6 Board
- Rev 6 ROMs (2.0 compatible, both MouseButtons for Bootmenu, AMIX
Option)
- populated with 4 Meg RAM
- Backside says: EXCEL M 94VO 21-89 [next to the RAM]
3 Stickers: * RWK 8
* RWK 7
* RWK 4 (?, not sure)
- Frontside:
* Sticker which says "West Chester build 0611" (what does that
mean, anyway ?) [next to U702]
* "ZD 6" (?) stamped on it. [next to U704]
- J200 is moved in direction of the RAM (away from the CPU).
I think this is what the A2630 manual calls "on pins 2-3",
although I tried both settings (did not cure the problems).
Other cards (the system crashes even if they are not installed, so
this should not matter):
Nexus: Rev 1.1 board (8 Meg option).
A2060 ARCNet Card.
---
If there is an easy solution for my problem (throwing my A2630 out
and buying a new one is not my preferred solution), or even just an
explanation what is going on here, I would appreciate to hear it.
Please send your answer (also) via email, to make sure I get it.
Thanks in advance,
Klaus
--
Physical: Klaus Weber c/o Schlitte, Graf-Spee-Str. 23b, 2900 Oldenburg _
Virtual: Klaus...@arbi.informatik.uni-oldenburg.de _ //
\X/
"Ich weiss von Schafen, die Dich locker austricksen wuerden. Ich hab schon
Pullover mit einem hoeheren IQ gehabt."
Wanda, Otto, "Ein Fisch namens Wanda"
armst...@cgc.com
Quite odd, I have the identical configuration, minus the Nexus, 2060 and my
drive is teh quantum lp240s. I have had no problems at all. All the rom
levels are identical also. Have you tried placing your level 6 roms in the
other 2630(the one with 9) and tried it? It might be the memory in your card
that is causing the failure. Since the transfer is DMA, it might just be
hitting a bad location. (Like I know anything about it...)
> Thanks in advance,
> Klaus
>
>
> --
> Physical: Klaus Weber c/o Schlitte, Graf-Spee-Str. 23b, 2900 Oldenburg _
> Virtual: Klaus...@arbi.informatik.uni-oldenburg.de _ //
> \X/
> "Ich weiss von Schafen, die Dich locker austricksen wuerden. Ich hab schon
> Pullover mit einem hoeheren IQ gehabt."
> Wanda, Otto, "Ein Fisch namens Wanda"
Mike Armstrong
armst...@cgc.com
Klaus Weber
In <1993Apr23....@arbi.Informatik.Uni-Oldenburg.DE> I wrote:
[I'm quoting that stuff, because i made some additions and clari-
fications. Sorry to waste some bandwidth...]
!!! I'm not quite so sure anymore that the A2630 is to blame, see
!!! notes below.
>
>
> Configuration:
>
> A2000:
> - Rev 6.2 Motherboard
> - V37.175 ROMs
> - J900 is set.
>
> A2091:
> - PCB Rev 4.0
> - V6.6 ROMs
> - Serial #: 0100892312769001111833
> - Memory jumper set to 0 Meg.
> - JP 5 (Options): all jumpers removed.
> - Harddisk is correctly terminated.
>
> A2630:
> - Rev 6 Board
> - Rev 6 ROMs (2.0 compatible, both MouseButtons for Bootmenu, AMIX
> Option)
> - populated with 4 Meg RAM
- Board is _not_ overclocked. :->
> - Backside says: EXCEL M 94VO 21-89 [next to the RAM]
> 3 Stickers: * RWK 8
> * RWK 7
> * RWK 4 (?, not sure)
> - Frontside:
> * Sticker which says "West Chester build 0611" (what does that
> mean, anyways ?) [next to U702]
>
> * "ZD 6" (?) stamped on it. [next to U704]
> - J200 is moved towards the RAM (away from the CPU).
> I think this is what the A2630 manual calls "on pins 2-3",
> although I tried both settings (did not cure the problems).
>
>
> Other cards (the system crashes even if they are not installed, so
> this should not matter):
>
> Nexus: Rev 1.1 board (8 Meg option).
>
> A2060 ARCNet Card.
- Rev 2 Board.
> ---
>
Barry Tigner <tig...@paun01.pa.msu.edu> sent me file called
"ATechTips", which contains several do-it-yourself repair
instructions. I did the following modifications to my mother-
board (they were described in that file):
First step:
- "Replace U602 and U605 (74LS245) with 74ALS245."
-> ** NOTE ** My 6.2 board had the ALS types already installed.
see below...
- "Make sure that 74XX24(4|5)'s between Agnus & Ram are F types
the symptoms include spots appearing on screen"
-> they are F types.
- "put 3300 ohm resistor between #20 & #11 of U605"
-> did that.
As I said, U602 & U605 already were 74ALS245's. I replaced them
anyway (soldering sockets in of course) and added a 3K3 ohm
resistor between #20 & #11 of U605. I then tried if the problem
would appear again. Much to my surprise (and dislike, I might
add :->), the problem was much worse than before! With the old
chips installed (see below for what I mean with "old" and "new"
cheips) the problem showed up randomly and was not easily repeat-
able. With the new chips, the computer will crash within the first
two seconds after starting the write operation. Further investi-
gation revealed that U605 is the critical chip:
U602 U605 crash easily repeatable ?
old old no
old new yes
new old no
new new yes
I exchanged the chips several times to make sure that the change
of behaviour is really dependant of the type of chip installed
as U605.
Second step:
After taking another close look at the "ATechTips" file, I found
some more modifications mentioned later in the file. So I made
these to see if they would cure the problem:
- "For the A2000-A2500, put four 3.3K ohm pull up resistors on U605
(74ALS245)
pin11 to pin20; pin12 to pin20; pin13 to pin20; pin14 to pin20"
-> did that, too.
This did not seem to change the behaviour described above (Computer
crashes fast if new chips installed, etc).
SUMMARY OF MODIFICATIONS:
- U602 and U605 socketed. Tried several combinations of 74ALS245's.
- added 3.3k ohm pull-up resistors to pin 11, pin 12, pin 13 and
pin 14 of U605 (using Pin 20 as Vcc source)
SPECIFICATIONS of the 74ALS245's used:
"old" (as installed by CBM):
MALAYSIA 9001DF
* SN74ALS245AN [* is a strange company logo :->]
"new" (the chips I bought today):
_
| / 74ALS245 A - 1N [that's a large S between the
| \ FHK6883 vertikal bar and the text :->]
| _/ 9035VB
Perhaps someone knows if there is a difference between those two
types. I know they _should_ be identical, but since they behave
differently.
As a change on the motherboard clearly make a difference in behaviour
I'm not sure anymore if the A2630 is to blame.
In <1993Apr23....@cgc.com> Mike Armstrong <armst...@cgc.com>
wrote:
> Quite odd, I have the identical configuration, minus the Nexus, 2060 and my
> drive is teh quantum lp240s. I have had no problems at all. All the rom
> levels are identical also. Have you tried placing your level 6 roms in the
> other 2630(the one with 9) and tried it?
Not yet. I will try that as soon as I can though.
> It might be the memory in your card
> that is causing the failure. Since the transfer is DMA, it might just be
> hitting a bad location. (Like I know anything about it...)
Well, IMHO that's unlikely, because the system is working stable
without a DMA-Hostadapter. If the memory would be faulty, I would
expect errors even if the CPU is accessing a bad location.
>
> Mike Armstrong
> armst...@cgc.com
Thanks for your suggestions.
This is again quoted from my previous posting:
>
> If there is an easy solution for my problem (throwing my A2630 out
> and buying a new one is not my preferred solution), or even just an
> explanation what is going on here, I would appreciate to hear it.
>
> Please send your answer (also) via email, to make sure I get it.
>
Thanks in advance,
Klaus
--
Physical: Klaus Weber c/o Schlitte, Graf-Spee-Str. 23b, 2900 Oldenburg _
Virtual: Klaus...@arbi.informatik.uni-oldenburg.de _ //
\X/
"Sparen kannst Du auch, wenn Du nichts mehr hast."
Andreas 'atte' Gelhausen
Klaus Weber
it seems that I have solved my problems with my A2091 and rev 6 A2630. :-]
SUMMARY OF MY PROBLEM:
- A2000, Rev 6.2 motherboard, V37.175 ROMs
- Rev 6 A2630 W/ 4 MB and V6 ROMs
- Rev 4 A2091 W/ 0 MB and V6.6 ROMs
I experienced random system crashes when _writing_ to the harddisk.
Reading worked fine. The system works fine, if a non-DMA-host-adapter
(e.g. PIO) is used. This is _not_ a lockup of the SCSI-Bus, but a "real"
system crash.
SOLUTION (see disclaimer below)
I replaced U605 (74ALS245, on A2000 motherboard) with an 74HCT245. The
chip I used is labeled:
TOSHIBA 9131H
74HCT245AP JAPAN
U605 is located left of the A2000 coprocessor slot. I _strongly_
recommend the use of an IC-socket.
The computer has written continuously to the harddisk for several (>=6)
hours now (and, in fact, is still writing while I type this :->). I have
not experienced a single system crash since I swapped the chips, to
I think the problem is solved.
DISCLAIMER
This modification has worked for me. It may or may not work for you.
In fact, since U605 is specified as an 74ALS245 part, there is a good
chance that it won't work (you have been warned!). I disclaim all
responsibility for loss of data or damage done to any parts of your
equipment, as a direct or indirect result of the modifications described
above. You make these modification entirely on your own risk.
If you have questions about the problem or its solution, feel free to
email me.
Thanks to all who responded.
Regards,
Klaus
--
Physical: Klaus Weber c/o Schlitte, Graf-Spee-Str. 23b, 2900 Oldenburg _
Virtual: Klaus...@arbi.informatik.uni-oldenburg.de _ //
\X/
"NEVER EVER mess with a PCB jumper you don't understand, even if it's
labelled 'SEX AND FREE BEER'."
Dave Haynie in <35...@cbmvax.commodore.com>
Michael Mickatroid
Dear Klaus
I just soldered together the A2000 remake v6.2 I then added a R4.0 A2091. Alles gut. I then added a Rev 6 2630 and was greeted with what passes for Guru Meditation these days. HDD access was so garbled I never made it to workbench (I am using a Zulu scsi HDD replacement so replacing the scrambled mess on DH0 was easy.)
I too became suspicious of an interaction between the cards and searched for them by model number plus 'crash' to immediately find your post. I could not find the HCT chips you used easily and used HC. I believe they are slightly faster but that both are significantly slower than the original ALS part.
Replacing U605 did nothing but replacing the second ALS245 right next to it as well fixed my problems too. Hopefully it has not messed up the bus for other cards. Looking at the bandwidth of the HC though I cannot imagine how.
Thanks so much for sharing the details 30 years ago.
M