DFT Digest

Unfortunately for me, due to scheduling, I wasn’t able hang out at ITC this year as much as I wanted - and as a result, didn’t get to have many in-depth conversations with folks. I was busy doing a little bit of catching up with a lot of people. The ol’ quantity vs. quality trade off.
However, one short encounter keeps coming back to mind: I talked very briefly with Ken Butler (a colleague from my TI days several years ago). He’s doing some stuff with outlier test these days, and asked me if I had done any more with mixed-signal test improvement, and ways of determining coverage of analog tests, referring to a test quality initiative I was involved in back then.

Well, I haven’t, but a recent press release I read reminded me of our conversation (“Cadence and Advantest Address Zero-Defect Testing Requirements For Automotive Electronics”).

The release is about a partnership between Advantest and Cadence - ATPG technolgy paired up with PAT (Part Average Test) facilities on the ATE. Actually, it was the first time I’d heard the term PAT, and after looking around, saw that it is a form of outlier detection - throwing out die that don’t fit in with the crowd. One way to get to “zero-defect” screening. Maybe we’ll explore these techniques on this blog someday.

So I can see how one could apply some of the newer digital ATPG technologies aimed at small delay defects could help out with zero-defects - better defect coverage is a good thing. But what about the analog circuits?

Back in my TI days, there seemed to be a flurry of activity around mixed-signal DFT, with some new BIST approaches, inductive fault analysis research - IEEE 1149.4 analog boundary scan was being developed at the time. Our own test quality initiative promoted a very manual, but fruitful analog test coverage evaluation that took up to a few days of going through schematics, and deciding whether or not “that transitor over there” was covered by any of the existing tests on the ATE, and if not, what test could we add?

That got us down to about 50 DPM or so, but not zero.

Since then, I’ve gotten away frrom the mixed-signal area, so I haven’t been paying attention. But has anyone done anything to automate that process? Is it possible?

Someone? Anyone?