Understanding Amiga Test Kit memory error messages (A1000 RAM issues)

[Turrican_3]

Hey there,

I was recently lucky enough to find an A1000 in very good shape (all things considered)

It boots Kickstart and Workbench 1.3 fine, but Amiga Test Kit almost immediately halts and shows memory issues as per image attached.

Removing the front RAM expansion though makes the memory tests go on without any apparent issues.

So I'd like to know if this is indeed a confirmation of the expansion being the cause of the errors (which also causes instabilities with Workbench, read Guru meditations and corrupted graphics) and also if the exact location of the problematic bits might help to locate/troubleshoot the supposed faulty ICs.

Is it safe to assume that, should a RAM IC be proved faulty, swapping position should lead to a different location of the aforementioned Xs on Amiga Test Kit?

[Daedalus]

RAM issues aren't always straightforward to diagnose, mainly because RAM chip layouts differ. Sometimes you get one chip that's responsible for each bit, sometimes you get one chip responsible for 4 bits, so a pair of chips make up a byte, and sometimes you get one chip responsible for an entire byte. Details like that can't be determined by software, so what needs to be done is to use the information that ATK gives you and combine that with the schematics to decipher where the fault likely is.

With your testing there it does indeed look like the RAM expansion is the problem. If you look at the address range being tested when the failure occurs ($40000-$7FFFF), that's the second 256kB of chip RAM, which does indeed correspond with the RAM expansion. Bits 1 and 3 are giving the error (on 16-bit buses like the 68000-based Amigas, the 16 bits are repeated in the 32-bit reading so you can ignore that), so you need to see which chips are responsible for those bits. Swapping some chips around could indeed help locate the faulty chip if you don't have the expansion schematics, and if the same bits fail in all combinations, it indicates an issue with the connectors or PCBs in either the expansion or the motherboard.

[Turrican_3]

Thanks Daedalus, you always offer interesting and insightful replies.

I guess it's time to bring the expansion to the office then (I do not own any proper desoldering devices right now) and start swapping some chips - there are 8 of them for the record IIRC, but I forgot to take note of the exact chip.

Will report back as soon as possible.

[Vypr]

Theres an A1000 schematic here: http://www.devili.iki.fi/mirrors/4x4.hopto.org/
That appears to have the expansion board schematic, it's the first zip file in the list and the expansion schematic is circuit2a and circuit2b
At a guess I would say the prime suspects are U1L and U2M

[Turrican_3]

Great, I'll have a look at that too!

Thanks.

[Daedalus]

If the chips aren't socketed, I'd look at the simpler things first: Check all the traces on both the expansion and the mainboard that are related to those bits on the expansion slot/card. Also, check the connectors and make sure they're undamaged and clean and free of contaminants. Also, if bits 1 and 3 happen to be near each other at any point, look for shorts between them.

[Turrican_3]

Yep, no sockets unfortunately, and it seems there's no space to fit them either (they're on the same side of the connector so I'm afraid installing them would lead to mechanical instabilities in the best case scenario)

Since we're here, may I ask why fellow Vypr suggests it might be a failure of both U1L and U2M ?

I can barely read schematics, hence I would have guessed a failure on a chip connected to numbers 1 and 3 at the bottom, that I believe are actual bit numbers (or 12 and 14 in case order is reversed, for the life of me this is something I *always* forget :-\), but obviously I'm not sure.

[Daedalus]

Hmmmm, I think that's the main motherboard RAM - the designations match the motherboard RAM, and it includes some of the RAM control logic that wouldn't be needed on the expansion (which is little more than a carrier for the RAM chips).

[Vypr]

Yes, it looks like the schematic I'm pointing to is the daughterboard, not the RAM expansion.

So I found an image of the RAM expansion and it appears to use the MB81464 which, according to the datasheet, is 65,536 words in 4-bits.
Given the results you are getting then my first surmise of a bad chip may be in error, I was assuming the chips were 1 bit, if you had a faulty chip I would expect to see the result as '-------- ----XXXX' but since two of the bits on that chip are reading properly thenit could be a bad join
I would first check the solder joints on the chips at each end of the expansion it will be one of those two that is having problems, paying particular attention to pins 2,3,17 and 18 as these are the Data I/O pins.
If it helps you can find the pinouts of the memory chip on page 129 of this document:
http://bitsavers.trailing-edge.com/c...s_Databook.pdf

[AmigaHope]

It could also be the result of a bridge on the address bus where the motherboard chips are responding to expansion addresses, but the problem doesn't show itself until the expansion is plugged in and two chips try to respond to the same request.

[Turrican_3]

Awesome replies guys as usual.

Weekend is coming so I should find some spare time to fiddle with this beauty, will report back.

[Turrican_3]

Data 00 -> D03 U7+U8
Data 01 -> D04 U7+U8
Data 02 -> D02 U7+U8
Data 03 -> D01 U7+U8

Data 04 -> D03 U5+U6
Data 05 -> D04 U5+U6
Data 06 -> D02 U5+U6
Data 07 -> D01 U5+U6

Data 08 -> D03 U3+U4
Data 09 -> D04 U3+U4
Data 10 -> D02 U3+U4
Data 11 -> D01 U3+U4

Data 12 -> D03 U1+U2
Data 13 -> D04 U1+U2
Data 14 -> D02 U1+U2
Data 15 -> D01 U1+U2

CASU0 --> U3+U1
CASL0 --> U7+U5

CASL1 --> U8+U6
CASU1 --> U4+U2

RRW --> all RAM chips
RAS --> all RAM chips

DRA0 --> A0 (all RAM chips)
DRA1 --> A6 (all RAM chips)
DRA2 --> A5 (all RAM chips)
DRA3 --> A4 (all RAM chips)
DRA4 --> A7 (all RAM chips)
DRA5 --> A3 (all RAM chips)
DRA6 --> A2 (all RAM chips)
DRA7 --> A1 (all RAM chips)

U1..U8 are MN41464-15 and I have followed this pinout for the A1000 connector.

I've checked continuity between the RAM expansion connector and the chips, this is what I've found.
(at first I was a bit surprised because I expected a full 16 bit address bus, but this clearly showed I'm not very familiar about the low-level architecture of the Amiga: I did some research and found out that it is actually Agnus that accesses RAM and it does it in a multiplexed fashion, or at least this is my understanding)

Since I'm not sure about the actual location of bit 1 and 3 I've also checked all RAM chips for shorts between the 4 data pins and there seems to be none.

Is there anything else that could be checked regarding this part of the system?
(I'm moving to the mainboard now)

[Turrican_3]

Nothing unusual (at least to my untrained eye) on the motherboard apparently.

Since I was there I also verified the powersupply and everything seems to fit specifications found in the A1000 Assembly Level Repair Manual.

I wanted to disassemble the daughterboard to properly see the mainboard in its entirety, but I was a bit scared to damage it (it wouldn't disconnect easily), and considering I'm not even sure this machine actually has issues I thought it was better to avoid risking to damage it.

So I guess I'm stuck now. :-\

[Turrican_3]

Ok I think I have some very interesting news.

I desoldered U1/U2/U7/U8 and swapped chips, now Amiga Test Kit still gives two bit errors, but in a different position!

I still have to check if it's really the same one that's giving errors, but I guess this behaviour heavily hints towards a faulty chip, right?

[Vypr]

Yep, I think that pretty conclusivly rules it's a faulty chip if the fault follows the chip.

Replacements should be easy to get hold of, found thison eBay for a decent price:

https://www.ebay.co.uk/itm/5pcs-MB81...QAAOSwd-pankzM

[Turrican_3]

Thanks again mate, as soon as I have the replacement chip(s) I will report back but now I'm quite confident I can solve this issue and bring back the A1000 to its full original potential.

[Turrican_3]

Ok it seems the problem has been solved: I've found a couple of spare DRAM ICs, located and replaced the one that was giving errors and now Amiga Test Kit is able to perform its memory tests flawlessly!

Thanks everyone for the support and advices.