Problem with burnt Altera FPGAs

[Pobyms]

I am trying to program a FPGA with my laptop and an altera byteblaster II cable (and the quartus software downloaded from altera´s site). I have programmed some of them before (yesterday) with the same equipment but today I think I have burnt 3 FPGAs. The software can´t detect them and they get really hot with 5V power supply.

I have tested with two different PCBs but it does not seem the problem (both FPGAs are now undetectable).

Do you think that they could be "get back to life"? May the problem be in the cable?
Any similar experience?


Thanks in advance

 

[Big Boy]

Re: Burnt FPGA?

It depend on the problem. It may be the board design. For sure, if they get hot, it's a sign of trouble.

Make sure that the power pins get proper voltages. Do not put 5V on the chip if it only support 3.3V/2.5V/1.8V/1.2V...

 

[Pobyms]

Burnt FPGA?

I checked the voltages and they were OK. I think it is not a board problem because one of the boards have been working for many days (the other one was new).

The FPGAs get really hot (specially one of them), the power consumption may be very high.

Thank for the answer, Big Boy.

 

[banjo]

Re: Burnt FPGA?

Are the FPGAs programmed via JTAG or a serial prom.?
If a serial prom, remove it and repower the board. The FPGA should be blank and run cool with all IOs configured as inputs.
If via JTAG, power up the board, but do not program it. The FPGA should be cool.
If the FPGA is hot when it is known to be unconfigured, then it is most likely bad.
If the FPGA is cool when it is known to be unconfigured, then it is most likely a mistake in the internal logic configuration that is tieing outputs to ground.

The one thing that can kill an FPGA quicker than anything is a negative voltage on a pin. The turns on the protection diodes and these will fuse to a short circuit very very fast. I fried some Xilinx like this several years ago before I redesigned the board.

To check for fried FPGAs via shorted protection diodes, take a good board and measure the resistance of all power rails to ground. Repeat this for the bad board. If you find that the bad boards have a much lower resistance then they are fried. The FPGA core power supply terminals have a pretty low resistance these days. Some Xilinx designs I measured are in the 20 ohms range. However, none of them are zero ohms. Near zero ohms is a bad FPGA.

If using a laptop with a byte blaster configuration, I would tie the laptop and then system board both to earth ground before attaching the cable. Laptops due to being portable, tend to have floating chassis which can introduce the dangerous negative voltages.

--- Steve

 

[throwaway18]

Re: Burnt FPGA?

In general when programming devices in circuit applying volts to the pins when the power to the device is off is bad.

 

[Pobyms]

Burnt FPGA?

I use a byteblaster II cable to program the FPGA. It is connected to the parallel port of the laptop (I have done the same with a PC and the problem remains) and to a JTAG connector in the board.

The problem is this one:
- I program the FPGA without problem.
- Then the FPGA does not work properly but it gets hot.
- If I try to program it again, the software cannot find it.


I am trying to find the causes and I think it could be the cable or an USB module I use to connect the FPGA with the PC. (I have the same problem with two different models of FPGA and boards but they use the same cable and USB module model).

 

[banjo]

Re: Burnt FPGA?

If the FPGA bit file has outputs defined on physical pins that are tied high or low, aka VCC or GND, there will be a bus contention which will cause the device to overheat. If enough of these IOBs are loaded like this, it could damage the FPGA.
I would take a failed board and a new board and compare the resistance readings across all power supply pins connected to the device. If the bad FPGA was fried internally, then you will see the difference on the resistance checks.

Are you 100% sure you are loading the proper bit file? You would think that these bit files are error checked by the FPGA, but not really. Several years ago, we had a situation where an FPGA would get extremely hot after incircuit reprogramming. It was finally determined that due to a software error, the FPGA was being loaded with a bit file for a cross-bar switch part that was also on the board. Wrong bit file can mean massive heat.

I would suggest generating a very simple design with only one output that flashed an LED, or toggle a single output pin. Load this extremely limited design into the parts and verify that it works. This will prove out your boards and JTAG cable. Once this is working, go back and start debugging the full design.

Finally, are you sure the FPGAs can from Altera? We have had vendors try to slip us "gray market" Xilinx FPGAs. These are either factory rejects that have been smuggled out of the fabs, or sometimes direct fakes. After fighting many weird problems, we now demand that the board assemblers buy FPGAs and CPLDs only from our approved distributor.

 

[IBNobody]

Burnt FPGA?

Which FPGA are you working with? The Cyclones?

- Nobody