Intersil HM1-6508-9 memory ic

[SeanB]

Look at the 8080 datasheet and the recommended layout and test circuit........

en.wikipedia.org/wiki/Intel_8255

[TerraHertz]

Well it _was_ probably just one flaky chip. Now it's two.
First, mark that reversed chip - it's very likely dead if they powered the board up like that.

Second, those sockets look like the flat-spring kind, not machine pins. Very prone to having the contacts bent and making intermittent contact. Inspect every pin with a magnifying glass. Replace any sockets that look sus. Ditto with ICs - someone needs to extract every IC and inspect for bent legs. Then clean all pins by rubbing lightly with an ink eraser. The pins are tin or solder plated - notorious for developing oxide films that make intermittent contact over time.
This is actually a dreaded syndrome I remember well - flaky memory boards that used cheap sockets for the RAMs.

 [Edit: also do the same for all the logic. Same type of sockets, same tinned IC legs. There's no guarantee the fault isn't an intermittent in those circuits, not the RAMs.]

As for identifying the original bad chip...
Do they have any programs that will fit in 1K?
Take all the chips out but the 9 in one row (top or bottom, experiment to find out which works.)
There may be jumpers or something to indicate how many K are installed. No documentation at all?
Still getting parity errors? If so, swap to another 9 chips. Repeat till you have a set of 9 that always works.
Then swap in other chips in one position, one at a time, till you find the bad one.

About doing a hack with newer RAMs-
Since the board obviously has battery backup of the data, there'll also be some circuitry that disables writes when supply rail droop is detected (if it's done badly), or on a power-fail warning signal from the PSU (if it's done right.)
If you decide to use a newer SRAM to replace the lot, you'll need to ensure that it is a low power in standby type, with a write-enable that is compatible with the data protection scheme used on that board.
That's going to be interesting with the 9-bit words.  The cache RAMs people are suggesting don't tend to be designed with power-down data retention in mind.
Or you could just use two 8bit 64KB or 128KB CMOS RAMs, and waste most of the the storage space. <gasp>

[trevwhite]

Wow, everyone is giving me so much great information about this board. Thanks.

I will have another proper look at the board over the next few days. As I can now identify the edge connector pinout I may well be able to trace things a little easier.

I may well try and change all the IC holders. I do not want to damage the board. Removing those holders will be a pain.

[jeroen74]

To remove the holders, it best to crush the plastic, then remove each pin by heating it up, then pull it out when hot with tweezers. Or the brute force method, an hot air blower

[TerraHertz]

Wait, wait! I didn't suggest changing ALL the IC sockets. Only any ones that have obviously bent contacts. Which there may be none of.
How tight do the IC's feel during insertion? If it takes a sensible amount of force to insert or remove them, the sockets are probably fine. There were a few makes of sockets that were really horrible, but these ones don't seem to be any of the really bad ones I know of.

From my experience, trying to replace a large number of sockets on a board has a very significant risk of doing serious damage to the PCB. So best to avoid. Especially if the socket pins were bent over or trimmed on the solder side.

Any you do replace - yes, if you can, remove the plastic case first, then lift out each pin with a soldering iron on the component side, then use a solder sucker on the rear side very carefully to clean the through holes out. Without the socket preventing airflow from the component side of the board, and the pins in the holes keeping some solder in there by surface tension, the solder-sucking can be done much quicker and with less heating, reducing risk of lifting pads and tracks off the fiberglass. But many socket types don't allow for easy removal of the plastic body, without putting a lot of force on the pins. Which can separate the copper from the fiberglass.

With very old boards, it also helps to put a spot of fresh flux-cored solder on the solder-side pads before using the sucker. This removes the oxide layer, allows better heat flow and makes the solder flow easier during the suck.

[amyk]

I don't think it's S100. Not enough contacts. Probably a proprietary interface.

TerraHertz: You're right about using cache RAM, didn't notice they have several mA of standby draw (rather high for CMOS) whereas the original's standby is in the uA range. In that case, something like a CY62256N would be more suitable.

[trevwhite]

Hi.

My problem I suppose is that I cant really test very easily at present. I would have to spend a day with the machine. Trying all these things. Maybe that is what I need to do before anything else.

That is why I was thinking of replacing all the sockets. Because a few hours doing that is easier than guessing if is one of them. Replace the lot and see where we get. The confusing thing is that I tested all the memory. I wrote a small program that wrote a 1 then read back. Wrote a 0 then read back. It did this for all 1024 addresses.

All the memory passed. But when they got the board and had a play they found that is did not work still. And then when they moved memory around it worked slightly differently. A small program would load. So it seemed like some of the memory was bad  and was being moved around. But I tested it all?

Maybe dirty IC holders are not helping? What else could cause such a scenario?

[PA0PBZ]

The confusing thing is that I tested all the memory. I wrote a small program that wrote a 1 then read back. Wrote a 0 then read back. It did this for all 1024 addresses.
All the memory passed. But when they got the board and had a play they found that is did not work still.

If I read this correctly you tested the board in another machine?

And then when they moved memory around it worked slightly differently. A small program would load. So it seemed like some of the memory was bad  and was being moved around. But I tested it all?

Maybe dirty IC holders are not helping? What else could cause such a scenario?

If you tested the board ok in another machine, it could be that there is an error somewhere else in the address or data lines.

[trevwhite]

Hi. No I did not test the board in another machine. I took each IC out and hooked it up to a dev board. I then wrote a small program that addresses the IC and did write/reads to each location. Testing that I could write a 1 and a 0 to each.

When the fella took the board away he tested it in the original machine. It did not work. They moved some memory around and it worked a little. They could load a small program in but nothing good enough to make it useful.

So it might just be dirty/unreliable ic holders and dirty ic pins?

[TerraHertz]

When the fella took the board away he tested it in the original machine. It did not work. They moved some memory around and it worked a little. They could load a small program in but nothing good enough to make it useful.

So it might just be dirty/unreliable ic holders and dirty ic pins?

Possibly. That type of fault can produce very confusing results during swapping, since every time you pull the board out you're slightly moving the intermittent contacts.
On an old machine such a fault can be anywhere, starting with the solder joints of the edge connector to the backplane, the contacts on the edge connector socket, dirt on the edge connector itself, a through-hole plating gone intermittent, IC socket contact springs gone weak, legs of the ICs having grown oxide layers, etc.
Even an actual faulty IC - which can be temperature dependent or just plain intermittent.

Oh, and of course in a machine shop there's the chance of iron filings/swarf getting to the board, which can make horrible vibration-sensitive intermittents. Some IC pins and sometimes the socket pins may be iron cored - which means they are magnetic, which means magnetized iron filings may like sticking to them.

Take a closeup photo of the board, then take ALL the ICs out (and those jumper headers.) Inspect every socket pin and every solder joint under magnification. With the solder joints you're looking for hairline cracks, especially around the leads of through-hole components, that suggest dry joins. ANY joint that looks suspect, resolder - it does no harm.
Visually check the edge connector socket in the rack with a flashlight. No bent pins, no corrosion, no foreign material stuck down there?
Lightly clean the gold edge connector pads with an ink rubber. (Not too much, don't want to wear the gold off!)
It's also possible to clean the edge connector socket pins, for example with a bit of cardboard soaked in alcohol, held in tweezers or a surgical clamp. But be extremely careful not to snag on the pins and distort them. You have to look at the contact shape, and decide how to do it. Bear in mind that if you wreck one pin on the backplane, you've definitely killed the machine.

Check every IC pin is straight.  Any that are bent, straighten with fine flat-nosed pliers. Then lightly polish the flat faces of the IC leads with the ink rubber. Blow the PCB off with compressed air, both sides. Any dirt, scrub it with isopropyl alcohol and a stiff toothbrush. Dry with compressed air, put all the ICs back per the photo.
Did it make any difference?
If still flaky, then you have a candidate for attempting electronic diagnosis.

[amyk]

Hi. No I did not test the board in another machine. I took each IC out and hooked it up to a dev board. I then wrote a small program that addresses the IC and did write/reads to each location. Testing that I could write a 1 and a 0 to each.

When the fella took the board away he tested it in the original machine. It did not work. They moved some memory around and it worked a little. They could load a small program in but nothing good enough to make it useful.

So it might just be dirty/unreliable ic holders and dirty ic pins?

Figure out the pinout of, and test the whole board at once via the edge connector. If any bits fail their address can lead to the clue as to where on the board the problem is.

[trevwhite]

Hi All.

Update on the board. I think I have fixed it now. Thanks everyone. I went back and checked the board again. There were bad connections, bent pins, broken tracks. I replaced the memory ic that was the wrong way round. I found tracks broken from the battery backup. The battery is connected directly to the memory for its power down mode.

I cleaned up all the ic holders. I think they will need replacing at some point but for now some switch cleaner did seem to do the trick.

Thanks for all the help. There is a bloke in a factory who is so happy to have his machine back working.

Trev

[PA0PBZ]