last updated Feb 28 2021 2PM Herb Johnson for context, return to the "s100_conn.html" Web page at retrotechnology.com Introduction ----------- [Late Feb 2021 I was contacted by a former service tech of the late 1970's- early 80's who worked on many MITS/Pertec 8800 series computers. They co-founded a servicing company that worked on mini and microcomputers. They had a big national industrial client, who operated many sites and facilities. Here's what they said about specific problems of the MITS/Pertec Altair backplane 100-pin connectors. The problem described in these notes, is damage to the "ears" or ends of MITS (S-100) connectors on the backplane. That problem also occurs, on backplanes of other brands of S-100 systems. It's apparently debatable, whether the "ears" should be screwed to the backplane, with spacers underneath. Stresses due to insertion and removal, seems to produce intermittant bus signals to the boards, may lead to corrosion, and so on. Understanding these cause and effects is a little challenging. So it's good to find detailed "field experience" from the earliest days of Altair/S-100 systems. These remarks are slightly edited by me, specific edits are in []'s. The person chose to remain anonomous. - Herb] Conversation with a Pertec service tech ======================================= As a long time service person of [MITS/Pertec Altair] units in actual production environments, I thought I should mention what I found to be the most common problem with the S-100 bus in the Altair systems I serviced. One client had MITS/Pertec people try to fix this problem; finally they had to replace it. It was simple to fix when I received it but I never passed along the details. [Regarding my background and experiences.] I was handed the service on MITS systems from another company that sold them fully assembled and loaded with whatever software the customer required for their application. I only serviced one Front panel model and the rest were basic 8800b models with the Pertec Hard drive [5mb fixed and 5mb removable packs]. [On MITS/Pertec motherboard models,] I never worked on the 4 board motherboards, though I did have one that I was taking care of. It only needed software work so I never delved into the hardware on that unit. All of the units I worked on had the card guides and I wouldn't allow a broken guide to be used. For the units [used in a 24/7 application], downtime was not much of an option so careful work was the watchword. It was rare for me to have to adjust the power tap on the main transformer as all of the systems were assembled before I got to work on them. Only when updating, or when using my own setup was that necessary. The MITS/Pertec bus problem This may sound stupid, and might be only in specific production runs. But the connectors that are mounted to the motherboard will cause VERY intermittent trouble. There seemed to have been quite a few fixes to try to work around, but the actual problem [I found, is:] *The connectors CANNOT be screwed to the motherboard, either before or after soldering them to the board.* This sounds stupid, I know, but this is a Proven fact, as verified on 10 different configurations. I never figured out what the actual defect was. And,the problem can actually hide for years before appearing. You can imagine how bad it could get, when removing a S-100 board and then replacing it, would make the problem go away for a while. The results of the "error" would also change. I have seen hardware changes, software changes, and miscelaneous logic-gate replacements [attempt] to fix this. Quite a few being called a "possible" fix. There were a few service people who carried an eraser, to clean the board edge connection [contacts, also] in an attempt to correct this. I have seen many people get caught by this. I'm certain you can see where a lot of this fixes would lead. Such as, wearing the gold from the board, causing wear and contamination to the springs in the motherboard connectors, etc. etc. etc. I won't get into the weird signal mods that occur. - regards Herb responds ------------- On "screwing the connectors to the backplane". The 100 pin connectors probably have "ears" at either end, and are drilled for a screw to be inserted to mount them on the PC board. That's where the screws would be placed. But, there's often a space between the ear and the PC board. Seems to me, if there is no spacer inserted in that space, then if the "ear" is screwed, it puts a torque on the end of the connector. That stress could distort the connector and cause an issue. For instance: it puts a torque on the pins to the motherboard - possibly a cold solder joint would be pulled and fail. I'm an engineer, and those are effects I can imagine. And in fact - I see S-100 motherboards these days, with broken ears. Inserting the S-100 cards slightly offset, puts a lot of force on those ears. If they are not reinforced underneath, they will snap off (especially after 30 years). So I'm not making up this idea of torque. (This was called to my attention by Jon Chapman, who saw the result but perplexed by the cause.) Finally - symptoms. I understand the symptoms can be screwy, because of intermittant or high-resistance connections on signals. But you probably have some idea of 1) the kinds of symptoms you saw, 2) the ways you could confirm the problem was on the backplane (as in, I smacked the case and the problem changed), 3) specific means you tried to isolate the problem (as in, oscilloscope monitoring each pin, etc.). - regards, Herb The tech replies, with details -------------------------- The screws that go through the "Ears" are the exact screws that I am talking about. The true cause of the problems I never bothered to check because it seemed to be so varied. I did see split connectors, warped connectors and even (once removed) a slightly warped motherboard. Replacing the connector could often make the problem worse, although this could easily have been due to the stress put on the motherboard itself. As getting the machine up and running; when the company owner was looking over my shoulder, "up and running" was the only objective, I didn't go much further than identify what the problem was and getting it going; and service call over. All of this debugging was only for my own records. The initial tracing of the problem was simple. Reseat all the cards and "Poof", the problem is gone, only to resurface later. The true symptoms were all over the place but the only one I fully traced was the Ready line from/to the CPU card. It seemed that there was a capacitance value added to the line, which was one of the more critical timing signals. I believe the [quad input NAND gate] was an input to the CPU card from the bus. It also had other logic associated with it but I am fairly certain that one of the inputs was directly tied to the bus itself. (It also "might" have had a gate between that input and the actual buss. It was quite a while ago.) Also, I wonder how many people actually had a 100 MHZ scope to work with at that time. That's what showed me, for instance, that a card extender was quite useless for tight-timing problems as it introduced too many timing problems of its own. I never found a problem in a system by using it. Replacing the [Ready line's] quad input NAND logic IC fixed the problem, but then it returned a month or so later. The scope showed the problem but, strangely, moving the card in its slot didn't seem to change it. Moving the card to a different slot made other problems appear, but changed that specific one, as far as the scope trace was concerned. Replacing the connector "carefully" fixed the problem fully, until the next connector did the same thing. Within 1 year, every motherboard with connectors that were serew-mounted, showed "some" problem but always tough to find. Every motherboard with non-mounted connectors was OK, as far as connectors went. I also remember getting angry after reading in an assembly book that mounting the connectors was NOT recommended. I forget which book/manual I read that in now. It was an eye-opening find and made me wonder how many people actually read all of the addendums and every page of the manuals. I hadn't [read them all myself] until then, so I can blame no-one for such an oversight. I also am quite certain that I only saw one unit with the "spacers" under the ears. However the connectors seemed to have extra plastic on the ears to allow direct mounting. Maybe that was a newer connector or maybe I am wrong. I would lean toward the "I am wrong" explanation, especially after all this time. Strangely, at the time, among the few people I talked with about this [at the time,] who had real experience with the MITS systems, considered [my conclusion] to be common knowledge. Why I had to [find] it myself] before it was mentioned, has always been a slight mystery to me. Maybe original job security? Also, If I remember correctly, the screws did not just pass through the motherboard but also through the case, making the mounting problem even more critical. I imagine that if all the connectors were properly mounted, with correct spacing and torque specs, THEN soldered, that the problem might not exist. But I have no way to determine that now. I am certain, that some of the connectors could not be soldered after the motherboard was installed so that is another question. Also, I have seen one unit that ALL the "ears" were cut off. This was a very dependable unit and I only had to work on it once. You could call that machine a "First" overclock. (I should qualify that. This user had a CRT display hooked up as the terminal and the max baud rate of the serial card UART was 9600. He wanted faster response and so modified the card to allow 19200 baud. The UART handled this for a while but started sending bad data. I just had to bring it back down to 9600 to fix the problem. I still consider that my first experience with someone "Overclocking" a chip.) Diagnostic experiences I do remember, that the once or twice that I actually "played" with the mounting screws on customer units, but that always seemed to make it worse. I do seem to remember some motherboard warping on my service unit, but it was very small. I ignored it at the time. Actually, as I ran the service machine with no covers all the time, I believe that the entire frame was slightly warped after a time. I never recalled that detail until just now. "Did I ever remove the entire motherboard?" As these were all "In Use" systems and repair was job one, I only removed the entire motherboard on one unit. That was my test/diagnostic system that never left my shop. It had so many cards in it and I swapped them so often that this exact problem wasn't the biggest issue. It was usually not run for more than 8 hours before cards were swapped. [So that] motherboard was fully removed only once, when the PROM burner card shorted and literally fried a land right off the motherboard. I manually added a wire and threw it back together, with a couple new connectors. A few swap arounds and it was working well enough for the day. (Note: This system was not used to test the MITS systems, it was for testing special customer's equipment.) Service strategy I have to admit, I was always an Anti-board swapper. I "Never" attempted to swap a card from one system to another and didn't carry cards with me. As and example, even that bad-chip unit [with the Ready problem] was worked on and fully repaired at the customer site. The original hardest part was just figuring out which card was causing the problem. Of course, I had no "Known Good" cards to swap in anyway. I didn't trust my "service" machine because of all the swapping I did on those cards. There were so many mods on the PROM card, that were different between machines, that swapping would have been tough, anyway. (Remember, there was the Phantom mod, to remove the proms from the address space after booting, and I had at least 4 different versions of just that mod among just the machines I saw. There were several other mods on other cards as well so swapping was just never considered by me.) I strictly followed the rule "If it ain't broke, Don't fix it!" and I don't believe I ever fully removed a motherboard on any customer unit, only mine. Of course, seeing that just reseating the boards usually affected the problem. So I couldn't say what would have an effect, [because] you needed to remove the cards to access the [connector/backplane mounting] screws. In such a case, who could say which fixed it. I would have to wait to see if the problem re-appeared, just to see if what I did had the proper result. [That wait and those board-removals are] why it took me quite a while to prove to myself what the problem was. I will note that rack mounted units could sometimes be fixed just by sliding the unit out to look at it. At first, I thought it must have been cables being moved or something like that. That was strange at first, but [those experiences] led me in the right direction. - regards