Keep in mind, that you're not willing to do any testing on any parts before simply soldering them into place, and then, if they happen to fail during whatever checkout you deem necessary, you'll invest time in figuring out what's the matter.

I'm willing, in this case, to forego testing because I know which distributor sold me these parts and I know something about that local distributor and his sales staff and their practices. I suspect that they're rejects from Miniscribe, which I know bought parts from Hamilton-Avnet and which I know used 8748's in some of their products of the time. I also suspect that these were routed to me because they knew they would be used as stock and not for a project of any immediacy where they'd likely turn up as defective. They were, of course, right in that assumption.

I've plugged a number of these parts into circuit boards that were commercially built at a time contemporaneous with the manufacture of the chips, in which some parts worked fine, and other failed. Those that failed, failed in other applications as well, and not only in reset-related modes. I don't know why this particular batch behaves badly, but, knowing that they do doesn't, in this case make me very curious.

I did find it odd that they seemed to start up in RC-reset cases, and only then went on to demonstrate other failure modes, while failing absolutely in the cases which used the MAX1232 or that comparator-based reset generator which fed the MCU a 25-clock-cycle-long reset pulse no matter how long the reset button was pressed.

I'm convinced that I'll have to figure out some way in which to test any 805x parts that I intend to stock so I know they operate properly. When I actually have correct analog specifications for things I can test, I'll consider that, but, in the meantime, if it takes 30 minutes to test a part before it goes into my stock, I'll do that, because it involves 5-6 parts and not 100. I'd point out, though, thta this batch of 8751's is the first time I've encountered a batch of parts where more failed than passed simple power-on testing. Since I started out with microcontrollers/microprocessors, I've never thrown out a batch of parts before. I've had one 8008, one 6502, two 68701's, one 8748, and about a dozen various Pentium chips fail, along with some 3-4 memory IC's, and about two dozen MC1489's and one lone MC1488. I've still got all those, with the exception of the 8008, which I made into a "high-tech Tie Tack" and wore to a number of events. I'm not going to hang onto those 8751's, though, as I don't want them wandering back into stock. I'm convinced that there's a reason why the whole batch was rejected wherever they were rejected, and I don't want that flaw to come around and bite me.

If I have to write a test routine for 805x's, I'll do that, but it won't be anything to test I/O analog performance, nor will it extensively test reset behavior, since I can rely on the Maxim/Dallas parts, with their built in watchdog and reset circuitry, to work properly with pretty nearly any external reset signal that meets their digital requirements.

The thing that standsin the way of working a self-test routine, or, for that matter, any sort of tester, is that it requires that one first construct a test regimen. I suspect a tester will require that one build a circuit within which an instruction is jammed into the program store of one MCU by some other, perhaps much faster processor, and then have the MCU execute the instruction, with the other processor "looking over its shoulder," to ensure that it behaves as expected.

I may or may not be willing to do this, and will decide later, depending on how the future unfolds. With more and more 805x-core parts going "internal" it will be more difficult to test them. If one can't really test them, it may not be worth the effort to attempt it.

RE