I'm not an EE (my lack of knowledge will be apparent below) , but I know we've got a bunch of talented people on this board, so I'm thankful for any advice. The company didn't respond to my emails for weeks and the board I'm trying to fix is cheap, so I just ordered a replacement. This is more out of curiosity than need, however it would be neat if the thing could be resurrected, if not -- hey, it'll improve my soldering skills. Here we go:

I got a bunch of development boards from digilentinc.com to expose myself to PLDs. One of them was the XC95 board (Reference Manual, Schematics) with a Xilinx CPLD from the XC9500 family. It did work the very first time I plugged it in, the power LED was lit and even a nice little demo for one of the peripheral boards was loaded. After some fooling around however (erasing and reprogramming the device) I probably made a mistake. For some reason I can't remember making a mistake, a program loaded properly and it worked as I intended it to. The next time I powered the board the CPLD (IC3) and the voltage regulator (IC4) got hot. Too hot to touch actually, plus the board wasn't behaving according to its program. In this state it began to eat PSUs. Power LED went dim, then out. The supplied wallwart didn't output the demanded 6v anymore, but instead 0,3v. Next it killed one of my general purpose PSUs while I tried to erase the chip, but the programmer didn't see the board on boundary scan anyway.
I'm not sure if this is the fault of the last program, as it did seem to work while testing right after the programming session. It wasn't until a Powerdown-Powerup-Cycle that it went mad. Can such a thing be programmed in a way to blow itself up? Do we have unrelated hardware trouble? Right after I plugged in the second (good) PSU, the Power LED(LD1) did not glow. Shouldn't that tell us to take a closer look at the upper left corner, IC4 to be specific? Is there a way to tell with a meter if the voltage regulator is good?

Picture of original board. Nothing looks obviously blown or shorted.

Looks like it says LM317 (which makes sense, as it's a voltage regulator) (data sheet)

To check it I think you need to feed proper voltage on the input and then check the output voltage. I'd try to find/borrow an adjustable [lab] power supply with adjustable current limiter to power it. Or to simplify checking, you might want to desolder it and put it into a simpler circuit (like one of the reference ones in the data sheet)

The excessive hotness and dead PSUs probably means there's a short somewhere (excessive current drawn through the regulator) - anything else that got hot?

/Michael

I'd try to find/borrow an adjustable [lab] power supply with adjustable current limiter to power it.

I've got one right here. Looks like the board takes whatever current I set the limit to.

anything else that got hot?

No, just these two components were getting hot AFAICT. Excellent hint Michael, that got me thinking. I removed the CPLD chip and the Power LED glowed. Additional checking with the voltmeter indicated that the LM317 is fine. So the short is in the chip.I wonder what I've done wrong, the program I loaded seemed to work and I admit that I am surprised that one can generate a short with a faulty program... must have something to do with constraints and reserved pins. I'll just get a new chip I guess, as I don't see a way to erase it out of it's development board since I don't have access to a dedicated chip-programmer. Thanks, it is a lot more satisfying to know where the mistake lies.

If it got too hot to touch, the likelyhood is that the FPGA is now an ex-FPGA, having gone to join the choir invisible. One possible cause is that it latched-up into a high-current mode (a more or less common problem with CMOS architectures under the right conditions) and this has toasted the I/O drivers. Did you manage to program a pin as an output and drive it with another output? That can sometimes do it.

It's certainly not impossible to program a device in a way that causes a thermal mismanagement error (TME ). I designed a board many years ago that used a Motorola DSP56001 processor, in a PGA ceramic package. The things were not all that cheap, but very fast for the time, and fairly low-power. The programmer and I were debugging the board one day when the test program we had just loaded stopped responding, and while we were poking around trying to figure out what was wrong we became aware of a 'hot' smell, like a clean frying pan on a hotplate.

The programmer, not being quite as paranoid in such situations as I am, reached over and rather incautiously placed a finger on the DSP. There was a brief sizzling noise, a somewhat more prolonged howling noise, and he ended up with 'OTORL' branded backwards across a now-shiny fingertip.

It turned out that there was a bug in that iteration of the processor, which allowed you to program it into a self-destruct mode where it suddenly began emulating a heating element. We blew up a couple more in the same way before we found out how to avoid it.

Replace the FPGA, find out what the maximum current it should ever take in normal circumstances is, and put a fuse of just over that value in the power feed to the board.

pca

Did you manage to program a pin as an output and drive it with another output?

Possible. Likely. Okay, I admit, your signature hits the nail on the head. I just gained ~$40 (1 chip, two PSUs) worth of experience.

he ended up with 'OTORL' branded backwards across a now-shiny fingertip

Guess I was still lucky then.

Replace the FPGA, find out what the maximum current it should ever take in normal circumstances is, and put a fuse of just over that value in the power feed to the board.

I'll make sure to do that. Don't want to loose another PSU.

thermal mismanagement error (TME ).

Classic post, as always, Patrick.

Yup. I've got a bunch of Xilinx CPLD boards and if really won't appreciate you doing things like externally driving an output. Be very careful about what you're driving as well as they'll get extremely hot if any reasonable amount of current goes through them.

FPGAs are much more interesting IMO. Bit more expensive but you can do a lot more with them. On the larger ones I've got you can fit a CPU along with a bunch of peripherals cores all into a single chip. Take a look at OpenCores for some free cores to use.

FPGAs are much more interesting IMO.

I've got a 200K-gate Spartan FPGA here. Can you tell why I have begun experimenting with the lower cost chips? I'll stay with CPLDs for a while, until I stop toasting them.
Thanks for the link Tevor, I'm sure I'll have a good use for some of these cores later on. The VGA/LCD core looks like an eyechatcher.