Coolrunner-II CPLD XC2C512 issue seen at high temps with usage at ~100%

[hguy]

I am using a Coolrunner-II CPLD. Part# XC2C512-7FT256I. I am using 503/512 macrocells (99%).

Is there a limit to how full the CPLD can be and still have guaranteed performance?

I did do a timing simulation using max delays and everything worked.

I do not recall seeing this issue with parts i used in the past.  Maybe an issue with these newer CPLDs we received?

I am already using the fastest speed grade (-7)

[drjohnsmith]

Well full, well done,
No limit as far as I know about performance over speed,

There are very strict constraints upon how many simultaneous IO can toggle at one per bank, and how much power each bank can source / sink.

What problems you having ? has the design worked for a while and stopped ?
are you purchasing form an authorised distributor , there are many grey / fake suppliers around.

Give us some more details we might be able to help.

[hguy]

We have been building these boards for a few years. No trouble until recently.  I am worrying about possible counterfeit part as well.  Not sure how I could tell that though.  As for switching I/O, the CPLD is used for driving resets to devices on the board in an orderly fashion.  It also has a UART interface (19200) back to the system that controls it.  We do use the UART to command the resets on/off.

[joancab]

What is the issue?

[hguy]

The issue is, at higher temps I cannot command the CPLD, via the UART interface, to release the rest of the board out of reset.  As we lower the temperature things start working.

[drjohnsmith]

Can we have some background please.

has this been working for a while and its just this batch of chips that have failed ?

It has all the hall marks of grey imports,
was it an authorised distributor ?

[joancab]

When it comes to temperature one has to bear in mind there can be 10 or 20C difference on a PCB and the value that really affects things (the die temperature) is most of the time, un-measurable. 

There is also variability in components regards to temperature. Not all chips will stop some function at 92.3C +/- 0.1C, I mean it could be that batch that is more sensitive to temp, or, for some reason (including apparently disconnected matters like PCB materials with different thermal conductivity) this batch of boards run the FPGA hotter.

Not discarding, as @drjohnsmith says, the source not being the best. If the above is discardable, then maybe time to retrofit a heatsink... what if a previous batch board runs, say, 5C or 10C hotter?

[hguy]

Previous boards have been built and there was no issue at hotter temps.  So yes it has been working for awhile and it seems we have an issue with these batch of chips possibly.

Trying to check to find out what supplier these parts came from.

[drjohnsmith]

Your right,
it is sounding more like the chip batch,
did you purchase from an authorised distributor would be the first question ?

also check things like the power supply, that could also be a suspect chip.

[lowearthorbit]

Scrap material is often re-marked as the fastest speed grade,

Maximizes the profit by the seller.

Yet another reason to only buy from the authorized distributor.

lowearthorbit