PS/2 Memory Expansion Adapter troubleshooting

[excession]

Hello everyone,

First post here. I've spent the last week or so getting my old PS/2 Model 60 running. After replacing the battery and the electrolytic capactitors in the floppy drive, it's happily booting from the ESDI fixed disk. (There are still some FDD issues but they probably deserve another thread!)

My issue is currently with the Memory Expansion Adapter that is installed. QBMCA identifies it as an FEFE (IBM 2 MB 16-BIT MEMORY ADAPTER) which matches the information here (it's the first one in that list) and the Reference Disk seems to see it and configure it just fine, saying there is 3072 KB of installed memory. However, the Reference Disk only shows 1024KB of available memory. When booting, I get a 160000 FFFE 201 and a 164 error.

My understanding is that this is a memory size error, accompanied by a memory test failure which says it's the first bit failing at address 0x160000. I've tried reseating the card and all the SIMMs, as well as swapping the SIMMs around and I don't think the address of the error changes. That is a bit of a surprise, and I also expected that if the error was with the card it would show up at just after 1 MB...?

With the card removed and the machine reconfigured from the reference disk, it boots fine with 1 MB of RAM but I was really hoping for the full 3 MB.

As a starting point, I metered all the through-hole and SMT capacitors on the memory expansion card and none are shorted, which I believe is the common failure mode for tantalum capacitors.

Any pointers for next steps, or thoughts that I've missed something stupid?

Cheers

James

 

[modem7]

First post here.

Welcome to these forums.

When booting, I get a 160000 FFFE 201 and a 164 error.

160000 = address 1,408 MB
FFFE = 1111111111111110 decimal = bits 15 to 1 in error, bit 0 good

Based on my experience with IBM's 201 error on the IBM 5170, when one sees FFFE, it should be treated as FFFF (all bits in error). For example, if you remove all SIMM's, the error would probably change to 100000 FFFE 201 (i.e. missing data bit 0 is showing as good!).

..., and I also expected that if the error was with the card it would show up at just after 1 MB...?

If the problem cause is a SIMM:
- It is possible for an address to be faulty part way in the SIMM's address range.
- In a card with multiple banks, each bank starts at a different address. Example at [here].

The card you point to has 4 banks of 16 data bits, with each bank consisting of two SIMM pairs (one SIMM is data bits 0 to 7, the other SIMM is data bits 8 to 15).

Assuming a fully populated card of 256K SIMM's:
Bank 0 (two 8-data-bit SIMM's for an 16-bit data bus) = Address 100000 (1 MB) to 180000 (1.5 MB)
Bank 1 (two 8-data-bit SIMM's for an 16-bit data bus) = Address 180000 (1.5 MB) to 200000 (2 MB)
Bank 2 (two 8-data-bit SIMM's for an 16-bit data bus) = Address 200000 (2.5 MB) to 280000 (2.5 MB)
Bank 3 (two 8-data-bit SIMM's for an 16-bit data bus) = Address 280000 (2.5 MB) to 300000 (3 MB)

The failure address of 160000 is part way through bank 0.

However, if only one of the two SIMM's in bank 0 is at fault, the bit error pattern should show the bits in the other SIMM as good (e.g. the "00" in FF00, 00FF, ...)

I know that some PS/2 RAM cards use special IBM SIMM's. Maybe your card is in that category, but someone has fitted 'standard' SIMM's.