PDA

View Full Version : Testing RAM



Erik
August 12th, 2003, 03:28 PM
I've got a whole ton of S-100 RAM cards of varying types (static vs. dynamic) and sizes that just don't work for one reason or another.

I've determined that these don't function by plugging them into a known-working Altair depositing and examining various addresses throughout the card. Some don't accept deposits, some don't return what was deposited (stuck bits, but usually across the whole card) and some don't appear as RAM to the machine.

Does anyone have any thoughts on how to diagnose and repair these cards?

Erik

NathanAllan
August 13th, 2003, 03:38 AM
Erik,
It may sound strange, but I have heard of all kinds of things like ram, video cards, harddrives and even monitors being fixed by setting a pretty strong speaker magnet near them overnight. A guy called Nimda(I think) fixed a graphics card he had, and then a friend of mine fixed his monitor like that. I went looking and found stories (not very detailed ones) about ram and other things being fixed the same way. It's worth a shot, I figure. And it already doesn't work, so what's to lose? :) Hope this helps.

Nathan

bobstek
August 13th, 2003, 08:16 AM
Erik -

I remember any number (okay, a small number of...) RAM test programs. I think there was one included in NorthStar DOS, but there were several that ran under CP/M - one of the CP/M archive sites should have some. They usually did multiple tests - walking bit left, walking bit right, checkerboard, all 1's, all 0's, etc. After identifying bad addresses, the usual solution was to swap with known good RAM chips to more precisely determine which chip(s) were bad - usually in a binary search fashion - i.e., replace half a bank and run the tests again, etc. Most if not all of my memory boards were socketed, so the task was tedious but do-able.

One of my favorite trouble-shooting 'tricks' was just to reseat all IC's. They were known to 'walk' out of their sockets due to thermal cycling.

Bob
Saver of Lost Sols

Erik
August 13th, 2003, 08:50 AM
I remember any number (okay, a small number of...) RAM test programs. I think there was one included in NorthStar DOS, but there were several that ran under CP/M - one of the CP/M archive sites should have some. They usually did multiple tests - walking bit left, walking bit right, checkerboard, all 1's, all 0's, etc. After identifying bad addresses, the usual solution was to swap with known good RAM chips to more precisely determine which chip(s) were bad - usually in a binary search fashion - i.e., replace half a bank and run the tests again, etc. Most if not all of my memory boards were socketed, so the task was tedious but do-able.

The biggest issue for me would be that the machine is running MITS BASIC so the CP/M stuff won't work. I haven't yet gotten as far as running CP/M on the Altairs.

I do have a smallish machine language RAM test routine that I could probably use as a basis for a front-panel entered tester. . .


One of my favorite trouble-shooting 'tricks' was just to reseat all IC's. They were known to 'walk' out of their sockets due to thermal cycling.

Most of these boards are socketed so I can take a shot at some swaps. I've re-seated chips on most already and done some bank swaps to isolate the bad RAMs. Of course, getting replacement chips for some of these will be harder then finding the bad ones.

Some of the boards are suffering from problems that are probably unrelated to the RAM itself.

Thanks!

Erik

barryp
August 19th, 2003, 12:24 AM
Of course, getting replacement chips for some of these will be harder then finding the bad ones.

What are the chip#s? If not 4164 or 6264 or 41256 or 62256, then what? Banks of eight or nine?

Also the problem could be with the line drivers or buffers or...

Erik
August 19th, 2003, 06:54 AM
What are the chip#s? If not 4164 or 6264 or 41256 or 62256, then what? Banks of eight or nine?

Also the problem could be with the line drivers or buffers or...

These are much older boards. There are a couple of newer boards in my collection with 4116s that I can find replacement chips for easily, but most of the rest have less common static RAMS such as 2102s and the like.

Fortunately I have multiple examples of many of the boards so chip-swapping is an issue, if I can figure out which are the bad chips!

For the boards I have docs on there are sometimes helpful hints (like which column contains which bit), but for others the docs are sparse or non-existent.

Erik