Recently, I've been having a lot of 'fun' with an AS-2518 MPU board. It's reasonably tidy, but came in a machine a friend of mine bought non-running. Initially, it had the LED stuck on, but after having slowly worked through a bunch of problems (resistor in reset circuit cracked, bad RAM, etc) I've reached the point where it's giving me flash codes. However, it only ever gets to four flashes, the fifth one is eluding me. It currently has the original ROM set for Mata Hari in it. I've tried known good replacements for every socketed IC, swapped the PIAs, etc, but nothing changes. Giving it the ROMs out of Cosmic Princess results in the same behaviour, fails at the fifth flash, though with those ROMs sometimes the LED is stuck on instead of a flash, or flickers then goes out. Very strange.

Specific weirdness - fifth flash is U11 PIA failure. But, that's the PIA that's driving the LED that's flashing, so address lines, chip select etc. must be OK, right? I've metered out various connections, e.g. data bus, seems OK too.

Any ideas? I really want to get rid of this board...

Yep - tried 6820s, 6821s, swapping the two existing PIAs, but haven't sacrificed any livestock yet. I've cycled through about ten PIAs in my desperation.

This is on the bench - the board behaves the same way in a machine, though. Re failure on a single output, I wouldn't expect the self-test to look at outputs? Re the LED flashing, sure, I was only using that as proof that the addressing/CS was correct. It proves at least that the PIA is accessible to the MPU at the right address. Presumably the self-test is failing to read back the contents of one of the registers (I think that's all the self test does), which suggests that one or more of the data lines are bad. That certainly would correspond to the symptoms - provided that bad data line wasn't used when setting the PIA up to flash that single output, the LED would behave fine. I have metered out the data lines, and in theory they must be reaching the other PIA (U10) OK, as that one passes self-test. Maybe I missed something when I did that, but it seems unlikely. Bah. 2518s always give me problems!

Ha, given that I mainly do these repairs because I enjoy bringing dead boards back, that would be a bit counter-productive!

Yep - I've actually been pondering designing/making a low-cost alternative. It could be pretty budget, e.g. no silk screen, but if built with good quality sockets etc. there's no reason why they wouldn't be as reliable as any other board. Would probably change a few things, e.g. 74 series logic, different clock, redesigned reset, single non-volatile RAM - just common-sense stuff, no bells and whistles. The problem with the Allteks is the exchange rate. US$ are too expensive in NZ these days! Of course, I have so many other projects oh that I'll never get time to do it.

Definitely MPU200, I think. I'm not sure whether the CMOS RAM difference is an issue; it might be, but so long as AS-2518/MPU100 code completely ignores the extra RAM, it might be fine. I don't know any more than you do though!

Re the RAM generally, I'd probably use a single NVRAM module to replace all the RAM (6810 and 5101), so that either a 6802 or 6808 could be used. Could afford to be a bit wasteful of RAM, in order to end up with a simpler design.

Single EPROM - definitely. Downside being that there'd have to be ROMs burnt specifically for this design, though this was something I was also thinking was the best approach. Mind you, I have an EPROM burner, and many people don't, so maybe I'm being a bit selfish...!

Resistor/cap networks - sure, why not. Anything to simplify, tidy, shrink. Mind you, just losing most of the reset and clock components would help a lot. Of course, you could go to PLCC 6821s too, but that might be going too far.

Board assembly - probably easiest just to farm out the production. There are places that you can send your design, and have it assembled professionally - SMD, ROHS reflow soldered etc. But as to cost, no idea. I'm so lazy, no research on that yet. Maybe I'll have a look tonight if I get a chance.

DIY - sure, would dictate design to some extent though. You'd probably have to go for through-hole components rather than SMD, but maybe that's not a major issue.

Yes, it's quite likely it's something like this. Someone's done some work on this board in the past (U11's socket has been replaced). It looks like good quality work, but who knows. I'll investigate further tonight.

Really? Cripes! Are you sure? I thought this one was ex-Italy! Maybe that's just because someone's written Genaro on it. I so lose track. I think it was the other one that ran fine for a few hours then died again, too. So mixed up...

Interesting. I've always assumed that the MPU200 used two 5101s to get a full byte wide data bus; finally got round to verifying that. I can imagine that the 4-bit models assume that the unused bits would always be low (I think), and if they aren't, unexpected things could happen. If that needed to be forced, it could be a bit tricky, but it could be done with a bit of logic I guess.

Yes, sounds like that'd be required. Or take the coward's way out, and build two variants based on the same board - one AS-2518/MPU100, and one MPU200.

If I wanted to be lazy, it would be possible to use a single 64kx8 non-volatile RAM IC, and just have it selected when nothing else is. That way, all otherwise-unused memory locations work as RAM. That would work OK, but isn't very nice design, and doesn't address the 4-bit 5101 bus for AS-2518 style boards.

The only argument I have against a PLD is serviceability. It means that in many cases, once that goes faulty, the board is junk. As opposed to the original boards, using all off-the-shelf logic, which were repairable by anyone with the ability to solder and source parts. These days, some of the bits are harder to get, of course. My thoughts were that for AS-2518s, they use so little logic that I'd be OK with using common 74LS parts. If I were trying this with something like a System 11, I'd probably give up due to the large number of logic ICs, and consolidate in the way you're describing.

Additionally, I guess there's a question around the use of the word 'redesign'. My intention was to avoid any unnecessary changes, making the board as similar to the original as possible. There are a few things that just make no sense at all to leave as they were, i.e. the clock design (hard-to-find IC, not very accurate), and reset (too many components, replace with a simple reset monitor). Both of those could be replaced with very common off-the-shelf alternatives, so they seemed like a good idea.

So many different approaches that can be taken to this. In all the time I've spent thinking about it over the past couple of days, I could have repaired the original one, almost certainly!

Ha! I'm mortally offended! Using something faster than the core MPU just to generate the display timing!! What's the world coming to...

Seriously, I don't mind the PLD approach. I'm just a bit traumatised by all these boards I've encountered in the past that have made me go 'ah crap, undocumented custom IC, no point in putting any effort into this piece of crap, mumble mumble'. You are right, publishing the equations etc. would mitigate it well enough. At least then most people could find someone vaguely local to burn a replacement for them.

Yep. I don't often see MPU-200s. There must be far more of the earlier ones around. Maybe it wouldn't be worthwhile supporting those. Oh well, I'll ponder. It's not as if I'm thinking about doing this anytime soon.

This thread has moved on a bit, but I figure I should post a conclusion of sorts anyway. Last night I had a chance to spend some time going over U11 in detail. Looking carefully under the socket, I could see many pins where there was no solder on the upper part of the board, and some pins where there was too much solder that could have been touching multiple tracks (it's pretty tight under there, lots of tracks run very close to the socket pins). So I spent quite a while doing something like pinhole surgery. Made sure that solder flowed through in all cases, and removing extra solder with solder wick. I didn't want to remove the socket, as the board was already looking a bit questionable, I was pretty sure that'd do more damage. Once I'd done that, I metered out every pin, and every other location that was supposed to have continuity, did some more tidying up, and then powered it up. I was still pretty amazed when I got six flashes (no 7th flash on the bench, 6 is good enough for me on these). Then, into a similar machine, though still with the Mata Hari ROMs, and off it went. It seemed pretty reliable, so I've given it back to Dave for a quick test in the real machine. My testing hasn't been thorough, so I won't be too surprised if it still has minor issues, but I think it's pretty much there. Finally...

Yep, being in New Zealand, the situation is the same here. Tax is lower, but shipping kills it too.

Definitely something I've pondered - do absolutely everything in a FPGA. Not too sure about implementing NVRAM in a FPGA, might need that to be external still.