CycloneII "freezes"

I use a CycloneII (EP2C5T144) as a global control of a high power electronics board, designed a few years ago.

Now there is notest on some boards that the FPGA "freezes" during operation. After a power cycle it works Ok for a time until it again freezes.

* Investigations done:

1) When "freezed":

All power rails OK.

All user IO seems to be at Tri-State/High-Z: IO-levels are in accordiance with the external Pull-Up or Pull-Down Ressistors.

Measured some extra signals:

ConfigDone: High-3V3

nConfig: High-3V3

nCE: Low-0V (Pull Down via 10K)

nStat: High-3V3

Clk: OK, 32Mc, 3V3

MSEL1,2: OK, 2b'10: Fast AS-Mode

Signals to EPSC1: all signals High-3V (3V ok??)

JTAG: FPGA reconized via JTAG (ref. "Auto Detect")

2) Try to "unfreeze":

Recycle the power of the total board: OK. FPGA running.

Loading the SOF-file via JTAG: no errors, still "freezes".

Pulling nConfig down: still "freezes".

3) At the molent the FPGA "Freezes":

Checked the voltage rails: OK, no dips, no overshoot.

* Extra Info:

EP2C5 is configures in ASmode with an EPCS1.

It's a power electronics board: so there can be expect some noise. If the FPGA freeses it is always when the board is at max load.

No use of any tri-state IO on the logic design.

No use of the nDEV_CLR and/or DEV_OE pins.

*Some tips?

Does someone an idea why the FPGA-IO goes to global Tri-State/High-Z?

Does someone hase some ideas about extra checks, measurements?

Thanks,

Kind Regards, SDPR