Production testing

Pretty similar:

We build most test sets using a rackmount PC, standard interfaced instruments (DVMs, scopes) and usually some custom interface board or box. If the DUT is a VME module, we'll sometimes make the special interface board VME, too. We usually program the PC in the DOS version of PowerBasic.

Only simple stuff, like all-analog boxes, gets manually tested with a custom fixture.

Lately we've been building a lot of BIST (self-test) into our products. That helps a lot in production, and our customers like it, too. Sometimes we charge for it, sometimes not.

John

We found ICT was necessary when the complexity of the boards got to the point the yield at final test was essentially zero without it (couple of thousand components on a board was a real problem). Before that it was possible to do functional test. After ICT we got a much better handle on where the problems were, too. We still don't do ICT on simpler boards (it's expensive and locks the design down too much). Final test is still done manually with test jigs.

My guess is where the in-house cost of design, development, debugging and documentation (4 dimensional..) would exceed the cost of a known workable ATE plus cost of programming etc. The tradeoff is extremely inexact as the company may not have the many kilobucks for the ATE but has lots of tech time to spend.

Though we had the ICT system sitting in the corner, unused, the programming and fixtures weren't cheap (nor is maintaining an ATE system). Your analysis is pretty much what we went though, though there were a few things we learned after converting the first board. The most important lesson was that our quality went *way* up. Functional test was missing a lot of errors. We also found a few supplier problems, which were easy to clean up after we knew what they were. On the down side, we've since found it almost impossible to ECO the boards (ICT programming cost) so the product doesn't get many hardware updates. Compliance costs are to blame here, too, though.

Depends somewhat on the skill level of your staff, but I build test circuitry into my boards. In my experience in small companies: test jigs get lost, no one documents how they were made or altered, after a few years the boards are essentially untested. If the circuit is high value, I build test hardware into it and describe how to use it in the circuit's technical description. If it's low cost, it's probably got a micro in it and you can at least flash an LED if it scans something awry at its inputs or fails a checksum test.

Modularity is good. (For isolating faults.)

When does off the shelf ATE

Difficult to say when ATE becomes worthwhile as it depends on volumes and what kind of faults you experience. For example in the bad old days, hand stuffed boards could have a resistor of the wrong value inserted. That's not so likely these days, though it's not unknown for an SM machine to consistently put the wrong value resistor on a whole heap of boards if someone loads the wrong reel in. The better SM machines do a certain amount of checking as they place components.

Or perhaps your major problem is the hand made wiring looms or a rechargeable battery or cosmetic scratches rather than PCB circuitry... another issue is that if the product is not a stable design, or likely to only have a short life cycle, it may not be worth creating a jig. You'll have to decide what's best for your product yourself.

My last place, we had runs of a couple of hundred boards at a time and the Production people seemed to think it was worth using an ATE machine with a jig for each board type. I had my doubts; it seemed an expensive bit of kit which needed a skilled technician servicing it all the time, if only to interpret the results or replace worn parts (lots of mechanical wear & tear on those things).