I haev many reasons for objecting to it, which basically fall into two groups. The first is the obvious one. It's wasteful. OK, I'll sometimes replace a module rather htan a part (e.g. I'll replace a trasnformer, even though I am quite capable of rewinding one, and would do so if a suitable replaceemnt was unavailable). But I find it totally ridiculous to replace a complete PCB because one cheap IC has failed. Ah, but you're going to say to me that finding out which IC has failed takes time, knowledge, and experience. I fully agree with this, which brings me to my second point. Board swapping, unless done after a lot of measurement and thought, is not necessarily going to solve the problem. There are 3 sub-problems. Firstly, a fault elsewhere in the machine could damage the new board. An obvious exaple of this is a PSU problem that blows up chips. Secondly, the new module mioght not be totally compatible with the old, you may have to make changes elsewhere (I've been bitten by this..). And thirdly, and most seriously, even if you find that replaceing a particular board seems to cure the problem, you may not have really cured it. A classic example of this is that you have 2 modules, A and B, with a signal going between them. The timing of A drifts and the system fails. You replace B, and by chance the new one has a looser spec. The machine works, but A continues to drift, and it fails again a few days/weeks later. OK, a new A now gets it going, but that would have worked with the original B, which in fact was never faulty. Seen it happen. So if you're going to swap boards, you have (IMHO) to make plenty of measurements anyway. And then you pretty much know which component has failed. It's probably less effort to just change that one component. A second rant of mine is about people who are trained to do a particlar job. Such people generally have no real knowledge of the device (or the subject in general) and are incapable of logical though. I remember one such idiot who wanted to replace every PCB in an expensive graphics display system, when 10 seconds with a voltmeter showed that the 5V PSU was sitting at 4V. He was escorted off the site by our system manager and told never to return, BTW. But I digress. Hmmm.. I know pletty of such knowedgable people who can't get a job in the electronics industry, either design or repair. Is this somehow less wasteful than having them doing field service work? I have never found 'that's the way things are done' to be a sufficient reason for carrying on doing them that way. Indeed, and then it's a pity that proper serice docs don't exist. I have tried to remedy this for the HP desktop calculators, but it's a long job... Oh don't get me started. The worst is the Torch XXX service manual. It contains nothing of any use at all. It contains a block diagram that's fundamentally incorrect (it's the video memory, not the main memory, that's shared between the main processor and the service processor), it contains a 'glossary' that would insult a 2-year-old. HPO desktop calculator service manuals are curious. They have some useless stuff in them (a typical repair procedure for the HP98x0 machines is to swap out the 4 processor boards, if the machine then works to put the original ones back one at a time until it stops working). But they also have some odd bits of information that would appear to be useless to a boardswapper, but are useful to us now. For example, the 9825 service manual, at least the version I've seen, contains almost no schematics (expected). You do get the PSU schematic, and you get the schematic of the printhead test fixture. That's the one part, of course, that's not part of the machine, so it couldn't be reverse-engineered from a real 9825. -tony

At 07:06 AM 10/26/05 -0700, Sellam wrote: Actually I HAVE done the analysis (in APL yet!) and there definitely IS a place for board swapping. Especially when you consider the down time of the system and the cost of extra systems when you HAVE to have a system in operation (specificly in Air Defense!) I ran a simulation for the Canadian DND (Department of National Defence) which showed that a 3 level repair concept was THE most cost effective and yielded the maximum UP time. There are a LOT more factors involved here than what Tony realizes. You have to include things like stocking ALL of the parts and manuals at the more (and multiple!) forward areas, also training all of those personel in specailized things like repairing 7 layer circuit boards, providing very advanced test equipment including in our case numerous optical test beds mounted in air ride semi trailors. All together we looked at something like 40 different factors for a couple of thousand items AND all of the interactions. (That's why we used APL. It's great for matrices!) I think one thing that Tony forgets is that VERY few people are as skilled as he is. I know the repairmen in the Canadian Army are much more skilled than the ones in the US Army but even they wouldn't attempt the kinds of things that Tony routinely does and four years is simply not enough time to teach the skills necessary to do component level repair, much less troubleshooting, on EVERYTHING. And in many cases those specific failures happen so seldom that no one is likely to maintain the knowledge and skills for every possible repair. * Field: PM and board swapping by system operator based on BIT. Depot: board and LRU level repair at special facilities. One in North America and one in Europe for high and medium failure rate items Factory level: for items with critical alignments, very difficult repairs and items with low failure rates. If you're interested, here what I worked on <http://www.army-technology.com/projects/adats/>. I was responsible for all of the Electro-Optical Guidance System (the barrel looking thing with the windows sitting between the missiles) and a large part of the Turret Electronics. Now that you've peaked my interest I'm going to go see if I still have that analysis around here. It was quite a piece of work if I do say so myself. Joe

At 05:46 PM 10/26/05 -0400, you wrote: You're right. The last place that I worked before returning to college was a relatively small food processing plant. We had three lines there and it cost $600 per hour per line when they were down. It only takes a few seconds to swap circuit cards so we had plenty of spare cards. After the line was up and running we took the cards back to the shop and repaired them at our leisure. I made them REAL happy when I build some test fixtures so that we could test and adjust the cards in the shop. After that we could just pop out the old card and drop in a new one and the adjustments were close enough that everything worked with no futher adjusting. Joe

Tony Duell wrote: You know, that always surprises me - I'd expect board swapping to rarely cause problems for reasonably modular systems (I can believe it with such as DEC hardware though, where you so much as cough near it and something breaks ;) Depends on the nature of the fault I suppose. For field faults on current hardware I would expect board swapping to hardly ever make things worse (my annoyance there would be if failed boards were just tossed rather than being fixed back at base). For restoring classic machines that may have been kept in bad conditions or not powered up in years, it's likely a different story! cheers Jules

OK, one that bit me (it's not a board-swap, it's a module swap, but...). The fan in this PC's PSU decided to fail (bad bearings). I didn't have a fan in stock, so foolishly, I grabbed a 'spare' IBM-brnaded PC/AT PSU extracting the fan from that PSU and popping it into the existing PSU. Result : the machine didn't power up at all. It turned out that the 2 PSUs were of different makes (although both IBM labelled), and both met the IBM spec _which specifies a minimum load on some of the outputs). The 'old' supply had been happy with less load than that, then 'new' one wasn't. And of course my modern hard drive didn't put much load on the 12V line. It would have been quicker, and a lot less hassle, to just swap out the fan. -tony

Don't be silly. It's _me_ you're talking about. The machine was (then) over 10 years old. Well, technically (!) I already had that information. The IBM PC/AT Techref gives the PSU spec (although no schematics). Thing is, the PSU was supplied by various vendors, all _met_ the official spec, some exceeded it rather. My system did not have enough load to be withing spec (I should have added one of those IBM load resistors :-)), but it worked with the original PSU (which exceeded the ofifical spec), not with the one I swapped in. OK, in sesne it was my fault in that my machine was out of spec. But it's the sort of thing that can easily happen when you replace a module with one that's not quite identical (but appears to be). That, agreed, is what board-swapping usually means, and what I really object to. If you're going to swap boards, you _have_ to do enough tests to be sure the board in quesiton is the defective part (and the fact that the machine appears to work when that board is replaced is not enough, as I've explained many times in the past). And to be honest, if you've done enough tests to _know_ which board is faulty, you don't need to do much more to find the faulty component on that board. -tony

I hate to say this, but _random_ boardswapping, which is what I am really complaining about -- that is, when you easter-egg parts until the machine seems to work again -- is the last thing I hope would be used on critical systems. Unless you know what/where the fault is, you can't know you've fixed it IMHO (I've explained the problems many times before). [...] Even worse when you're trying to make one good machine out of 2 defective ones and you haven't a clue as to which boards are good or not. Note that I have no problem at all with 'test rigs', which often are working examples of the machine in question that you can plug defective boards into for testing. A good example of this is for the HP41 calculator. Due to the mechanical construction of that machine (the inter-board connectors are clamped by the case screws), it's impossible to have the machine in a working state where you can probe pins of the chips. HP serice centres had a 'test calculator' which, from the picture in the serviec manual was a mechanically modified HP41. An afternoon in the workshope with a battered 41 produced a workable substitute. Actually, this is a great advantage. A lot of 'training' is based on known or common faults, and if you're not careful the serviceman will get totally lost when faced with something that's not been considered. Since we don't get the manugacturer's traing, we (or at least I) approach such problems from the viewpoint of standard electronic principles and are less likely to get totally confused (well.... that's another story :-)) Hence my comment that I don't have a modern PC because I can't afford the test gear I'd need to maintain it. Hmmm... I'll go a bit further than that. There is a suprising amount you can do on a modern machine with limited test gear intellegently used... Hmmm.. I have a fairly well-stocked junk box, but it contains components, not boards. Suppose a serial port fails. OK, it's going to take me a little longer to find the defective 1488 (or whatever) rather than just swap the board _but_ it takes less time than me going over to PC-world or whererver to get that replacement board. Of course the fact the PC-world doesn't sell boards for any of the machine I own is a secondary issue :-) -tony

I fail to see how, at least without doing further tests and measurements (some of which can be quite complicated). I've yet to see a system where symptoms alone with determine the faulty module with 100% reliability. You know as well as I do that it doesn't work that way for most, if not all, of the machines that _we_ work with, and probably never has. The sort of failed servoids I've met just look at the symptoms, pull a part, put a new one in, and hope the problem is cured. AFAIK the defective part is _not_ analysed futher. When was the last time you met a droid in a PC shop who sent the defective board back to %far-eastern-country? Well, actually, I do refuse to use anything I don't properly understand, but that is another issue. OK, I'll go further. Not only do I not have the skills, documentation or test gear to maintain a modern PC properly, I don't know anybody else who can, or how to find such a person. And I am not going to trust my data and my (in)sanity to some idiot who can't understand why a logic analyser or 'scope might even be useful Actually I am trying to think of anything I own and depend on (or even use actively) that I am not capable of repairing.... I'd have problems with a wristwatch, but pocket watches and clocks wouldn't be a problem. Nor, for example, are real cameras. -tony

Alas yes... I can't attack it like I do with the RK's, etc :-) That depends on _which_ computer :-). I can certainly repair the CPUs in my minis, PERQs and HP desktop calculators Yes, OK, I agree with you. I think want annoys me is that I know I should be able to fix a motherboard to compoennt level. I can do SMD rework (it doesn't bother me at all now). I can trace signals with a 'scope and logic analyser. And I have the obijection to replacing more than I need to. Yes, there is always that :-). I am having peoblems finding space for the PX8 I bought the other week... -tony

Yes, exactly that. I got one with what seems to be a PSU problem -- it powers up OK from a bench supply, but if you turn it off it won't turn on again without a hard reset. The always-on backup power line is not behaving correctly. One SMD sener has been blasted off the board, so %deity knows what other damage has been done. But the rest of the machine, in particular the custom gate arrays, all seem to work, so I think I can repair it. If you are seriously looking for one, they are not that uncommon on E-bay. I got this one there fairly cheaply because it was non-working, even though the PF-10 3.5" floppy drive was included (this, BTW, is an interesting device inside, it's one of the few 40 cylinder 67.5 tpi 3.5" units, and the drive elecrtronics (read/write chain, etc) and controller (765 + 63A03 microntroller + etc) are all on one PCB). -tony

On Mon, 31 Oct 2005 01:13:21 -0500 (EST) der Mouse <mouse at rodents.montreal.qc.ca> wrote: Correct. As Tony has said, he has a PC-AT with one of those CPU-update modules that makes it an 80486 system. -- http://sasteven.multics.org/MacSE30/MacSE30.html

Tony Duell wrote: Yes, but I suppose in the field a company's not aiming for 100% reliability. If the amount of customers annoyed/inconvenienced by incorrect boardswaps is less than those annoyed by the time it takes for a 'proper' tech to diagnose the fault with their hardware then the boardswapping approach makes more business sense and keeps more of their customers happy. I know it's a shame it's like that - if I were in that field I'd much rather be tooling around in a van doing component-level fixes than doing it in a warehouse somewhere :) Yes, that's the unfortunate problem with computing today (or any electronic consumer goods). People seem happy to just throw perfectly fixable stuff away because it's cheaper for a business to just pay someone to make more than it is to pay someone to fix the broken item. That seems incredibly stupid in the longer term, but who thinks long term any more... On the plus side it means people like most of the inhabitants of this list can get some pretty cool broken stuff for free and fix it for almost nothing except time... :) apart from your brain ;-) see above :) cheers Jules