A600 1MB memory card showing up as 512kB?

[hjalfi]

I recently acquired a really crusty old A600 and have been making it work. It came with a AMPRO6001 trapdoor memory module, which had been badly mauled by the Varta battery (accurséd be their name...).

After removing the battery and lots of cleaning, it actually doesn't seem too bad. However, when I connect it to me A600 it appears as 512kB of 'graphics RAM' (I take it that's chip RAM). AFAICT this memory module is supposed to be 1MB. It looks like it's completely dumb, with no autoconfig ROM or memory mapping logic, and according to the memory map it should appear at the hard-coded address range of 0x100000-0x200000.

I've traced through the address and data lines and they all seem to be connected to the obvious places. I've cleaned all the contacts, too.

What could be wrong here, and how can I diagnose it? Does anyone know of a bootable bare-metal memory tester, memtest86-style (it's hard to test arbitrary address ranges when AmigaDOS is living in random bits of it...)?

[Jpor]

Originally Posted by hjalfi

I recently acquired a really crusty old A600 and have been making it work. It came with a AMPRO6001 trapdoor memory module, which had been badly mauled by the Varta battery (accurséd be their name...).

After removing the battery and lots of cleaning, it actually doesn't seem too bad. However, when I connect it to me A600 it appears as 512kB of 'graphics RAM' (I take it that's chip RAM). AFAICT this memory module is supposed to be 1MB. It looks like it's completely dumb, with no autoconfig ROM or memory mapping logic, and according to the memory map it should appear at the hard-coded address range of 0x100000-0x200000.

I've traced through the address and data lines and they all seem to be connected to the obvious places. I've cleaned all the contacts, too.

What could be wrong here, and how can I diagnose it? Does anyone know of a bootable bare-metal memory tester, memtest86-style (it's hard to test arbitrary address ranges when AmigaDOS is living in random bits of it...)?

You should have 2MB showing I total (1MB Chip on the A600 + 1MB extra RAM from the trapdoor card).

Yep look up AmiTestKit.

Can be downloaded here:

https://github.com/keirf/Amiga-Stuff/releases

[hjalfi]

Hey, keirf! I know that name!

After copying the ADF to a floppy using his GreaseWeazle hardware (but my FluxEngine software!) and finding the memory tester, I see the following.

https://imgur.com/a/IKJCLBz

I'm going to guess that whichever chip is handling bits 12-15 isn't getting selected properly when address bit 0x080000 is asserted.

Unfortunately it's DRAM, which means the 20-bit address is split up into row and column and probably divided by two anyway to allow 16-bit accesses, so actually figuring out what line is causing problems is going to be non-trivial. I don't know yet whether the eight 4-bit RAM chips are divided into upper-half-megabyte/lower-half-megabyte, which might mean a bad chip, or if they represent even and odd 32-bit words, which means the chips would have to be fine for the low half megabyte to test correctly.

Looks like it's back to the continuity tester tomorrow...

Thanks!

[cpiac64]

verify ras and cas

[Lord Aga]

How do you see it's 512Kb? Do you see it somewhere reported as a 512Kb expansion, or do you just see "1.5MB free" on the Workbench bar?

[hjalfi]

I don't see it reported anywhere as an expansion card --- it seems to be completely dumb, and just adds DRAM chips to the chip memory bus. When it's plugged in, RAM appears at 0x100000 up.

I've double checked that all the DRAx, DRDx, CASx, RASx, OE, WE, Vcc and Vss are wired through to the RAM chips. The CAS and RAS wiring is kind of weird but seems plausible --- I need a circuit diagram here, I think. Does anyone know of one?

I think this is pointing at a bad RAM chip. Unfortunately they're not socketed or I could start swapping them. I don't really fancy the idea of desoldering eight 20-pin DIPs.

I suppose I could just enjoy my free 512kB expansion board.

[Lord Aga]

My point is, if you just see 1,512,345 free on the bar, that is normal, because the OS eats the rest (depending on the eye candy). And you do have 2MB in total and a working 1MB expansion.

[cpiac64]

check connector on a600 www.amigapcb.org

[hjalfi]

Originally Posted by Lord Aga

My point is, if you just see 1,512,345 free on the bar, that is normal, because the OS eats the rest (depending on the eye candy). And you do have 2MB in total and a working 1MB expansion.

Oh, I see --- no, Workbench is reporting about 900kB free. I'm using Sysinfo or ATK to get the real data.

[zipper]

Just type on Shell: Avail

[hjalfi]

I applied a logic probe to all the chips on the board while it was running. I discovered two interesting things:

(a) I see activity on all lines (except /OE, which was constantly asserted, which I think is normal);
(b) if you accidentally short two data lines together, the Amiga gets extremely unhappy and crashes. Who knew!

So I think this is pointing more and more to a bad chip. I have some sockets on order and once they arrive I'll try desoldering them. Once I have at least some socketed I can try swapping them around to see if the fault moves. That should make it trivial to identify which chip, if any, is bad.

[cpiac64]

read this

https://www.victortrucco.com/Commodo...0/ExpansaoA600

and check all signal

[hjalfi]

Very belated update: there was a bad chip. Socketing the two suspects and swapping them instantly isolated the problem. I ordered a 5-pack from ebay, replaced it, and it now works fine.


I did record it all on video but apparently I never got round to editing and uploading it all.