CPLDs

[gregboyle4]

I have used the Xilinx CPLD's for about 15 years in many different designs and have never had one fail.  This is both with (0 to70C) and (-40 to 85C) parts.  Keep up the great work!

Recently we had a group out source (overseas to Taiwan) one of our designs that uses an XC9572XL part.  This out sourcing was not known to us and that is a different problem that we are solving.  We have always, only bought from an authorized dealership.  We are checking with the vendor on the traceability of the parts.

We had 26 boards built and 6 of them have Xilinx failures (25%).  I have never seen a xilinx fail, let alone with a 25% rate in one batch.  Of the failed units, some don't work out of the box and others fail below 0C.

What seems to jump out is how the temperature range is marked on the part.  In the attachment for item 1, all the industrial temperature grade parts I have always used had the speed (10) and the temperature grade shown in the drawing.

The parts we received from overseas has the temperature grade as an "I" without the horizontal bars as shown in item 2 of the drawing.

Question.  Does Xilinx print out the "eye" for -40 to 85C in both styles as shown in item 1 and 2 of the drawing, or could item 2 markings be suspicious.

The parts in question have a marking of  "XC9572XL  TQG100AWN1113   F4227511A   10I"

Best Regards

Greg

[austin · Accepted solution]

g,

A big problem is with bogus parts (they are not Xilinx parts at all, but are remarked to look like ones), or Xilinx parts removed from old boards, and remarked to be more expensive parts.

If they are not from an authorized distributor, we can't help you.

If they are, then you need to contact that distributor (and let us know, as well).

Sloppy and different marking is common with these bogus parts.

[gregboyle4]

Thanks for the quick reply.

Greg

[jpine]

We had a problem like that with another CPLD brand (lets call it brand A).  Turned out the parts were fine, but the original design used asynchonous methods that didn't work over temperature and new parts.  Recoding for synchonous design solved the problem.