> On 2020-01-29, at 15:08, Pasi 'A1bert' Ojala <a1bert_at_iki.fi> wrote: > >>> No... circuit seems to be identical. >>> >>> Though they changed the masks which was expensive they didn't want to >>> introduce new part numbers. >> No, but the revision would be expected to change, wouldn't it? It looks like a negligence to me. Might even be deliberate like with 8521s labeled as 6526s. > Seems like a yield improvement revision combined with cleaning up some layout errors. Some of the top metal traces of the original are bound to be susceptible to defects. > > Lot numbers are used for this kind of yield/testability/reliability changes that are not visible to the end user. > > The manufacturer knows which lot is produces with which masks. I think you are right. And in this case it's probably only those changes around the pads, which don't affect the operation in any way. What I am afraid though is that they were also introducing other, less "safe" changes without bumping revision number too. When testing BeamRacer we had a case when two VICs of the same revision # behaved differently timing-wise. Small differences, totally not noticeable in regular operation, but getting important when we are acting in single-digit nanoseconds ranges. -- SD!Received on 2020-05-30 00:30:58
Archive generated by hypermail 2.3.0.