I was able to drastically reduce the rate of invalids by removing the heatsinks, scrapping clean the white adhesive and using gasoline, and the apply thermal paste (I couldn't quickly find thermal epoxy here in Mexico) and then gently press the heatsinks back on.
I think TheSeven is right about the nasty plastic packaging of the FPGA, it is the major bottleneck for its heat to dissipate. Isn't this chip available with a metal top?
I think TheSeven is right about the nasty plastic packaging of the FPGA, it is the major bottleneck for its heat to dissipate. Isn't this chip available with a metal top?
Glad to hear you got it working better! Is the heatsink then attached to the PCB in some way? I don't think the thermal grease will hold onto it very well (for long at least). Or maybe I misunderstood what you used...
Nope, no metal packages available for the Spartan 6 FPGAs, unfortunately. Looking at the published thermal specs for these FPGAs (UG385, page 341), though, I don't think the case is that bad. Obviously metal would be better, but we're just trying to move less than 10 W here. In rough numbers, keeping the thermal resistance from case to ambient under about 5 degC/W should be good enough. This is possible with the stock heatsinks and a good fan next to the board. The key then is that we have two parallel thermal paths: through the case and through the PCB. Adding those in parallel means that the total thermal resistance from junction to ambient is reduced a lot. This is why airflow over the board itself is very helpful! If you want to eliminate whole board airflow, you need to drastically reduce the thermal resistance of the heatsink.
