Weird problem with WP 9.1sp1 and XC95144XL

Hi,

I've got a really strange problem with a XC95144XL. It's a simple design, all schematic based, and the device is not behaving -- but only sort of. Pins that are unused are turning into weak outputs, some outputs work, and other don't (they may be high or low). I have data buses with 6 out of 8 pins working. Change something and recompile, reload and the results will be different eg. same data bus but now 5 pins work correctly and 3 are high or low. Most of the internal logic seems to be working OK -- this only seems to affect IO pins.

I've checked:

All pins (this is a TQ100 if that matters), including VCC/GND. A few pins ar "fast", most are not, and the 3.3 V supply looks fine. The download from a Parallel IV cable is verifyable, and repeated downloads work (or fail to work) the same. All the Fitter / Placer warnings make sense. The equations in the reports are all correct.

If I add some unneeded logic ie. add a counter, clock it and feed one output to an unused pin then the symptoms change.

Yes -- I do have it setup for the correct device.

Any bright ideas (or even some really silly ones)

While I hopefully wait for those I'm downloading 8.2 again.....

--
TIA,
Gavin Melville
gavin.melville@acclipse.co.nz
Reply to
Gavin Melville
Loading thread data ...

If you have an older design iteration archived, you could download that to check all the hardware is OK ?

You have checked the fitter pin-report to make sure it is not moving pins about ?

There have been other postings here about Xilinx CPLD flows, which suggest their PLD regression testing is, shall we say, "Casual".

-jg

Reply to
Jim Granville

Reply to
Benjamin Todd

Not for that device -- the design got too big. The smaller one was done with 8.2.03, and I've downloaded that now -- I'll try it tomorrow.

Yes I have. In some ways it's stranger than that -- take two pins that should be a CPU databus (but are actually permanently low) -- if I ground (internally) either one, the other works, however if I pull either to VCC the other stays grounded.

As a user of about 10 different versions from DOS Viewlogic forward I've never noticed that before. I've always found the software to be rock solid, well tested and clearly documented ;-)

Reply to
gavin.melville

The problem I had with a 9536 was "solved" by setting FSM extraction to no in the Xilinx tool. I also got a report that enabling WYSIWYG mode helped but I did not verify that myself since I had already gotten the design to work.

/Andreas

Reply to
Andreas Ehliar

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.