FPGA overheat and not working problem

[eehadi]

Hi, I have a new PCB includes a Spartan 3 FPGA, it worked well first day, then it had probably a connection problem that sometimes didn't work but when I moved board inside case or push some places on it, it ran again, in that case indicator LEDs were counting, stopped by fault but kept their values and now the FPGA is not working, it overheat but can be programed and some outputs (LEDs) are on and some are off and don't change. I tested all supply voltages (3.3, 2.5, 1.2, GND) and they were OK.
Is the FPGA damaged? if yes How can I test? if no what's the problem?
sorry for bad English

 

[pbernardi]

How much current each LED sink? Are the LEDs connected directly to the FPGA?

FPGA pins have a max. current allowed, if LEDs current exceeds this limit, it might be possible that the FPGA I/Os are damaged.

 

[eehadi]

Thanks. the power supply current indicator is not such accurate but it showed 0.1 when there was no problem and it shows 0.3~0.5 A now and FPGA overheats.
how can I measure LEDs current while they're placed in board? LEDs are connected to FPGA pins through a 100 ohm resistor. if pins are damaged, whole pins (also unused) would been damaged? I tested FPGA with a test program, that asserts some pins high and some low and it worked but blinking LEDs processes and concurrent part that connects input to output don't work and output remains in a fixed state.

 

[pbernardi]

Well, supposing that LED Vdrop is around 0.3V, then you would have normally (3.3V - 0.3V)/100 = 0,03A per pin. I don't think you need to measure the current, but as good practice, you should estimate the currents you're working with.

On Spartan-3 datasheet, they say it should support 100mA as absolute maximum current per I/O. How many LEDs are involved?

Anyway, I don't think it is a good practice connect LEDs directly to the FPGA. Use a transistor to help you to sink the current. And if FPGA overheats, it certainly is damaged.

 

[eehadi]

Thanks. I will test the board for detects tomorrow. what type of defects in PCB can make this problem?
There is 7 LEDs connected to FPGA ports, I think I can calculate LED current by measuring series resistor voltage drop. I wanted a minimum size board so I didn't use transistor to drive LEDs, btw they are little SMD LEDs that I guess they don't sink many current.

 

[ads-ee]

As you refer to the board as a "new PCB" I'm assuming this is a board you've had built from scratch? Was more than one board built? Do they all behave the same? If this is the only board...

I would first remove the resistors and check if the FPGA gets warm, that will indicate if the LEDs were/are causing a problem (I suggest the resistors as you can probably get more of them to replace the ones you remove). If they are the problem check the datasheet.

Given pressing on the board was resulting in changes in behavior (working/not working)...Was RoHS solder used? It is more prone to solder joint cracks under thermal cycling and let's not forget the whiskering that can short BGA pins together.

Is the Spartan-3 a BGA or a quad flat pack? If it's a quad you'll be able to check the solder joints of the FPGA. If it's a BGA well...xray the board...use Boundary scan...reflow the part...from least drastic to most drastic.

good luck.

Regards