Ok, so let's use the PCB as a heat sink. The problem is you're mostly conducting the heat through the BGA balls and have an air gap between the BGA case and the PCB. Fill the air gap with something thermally conductive. Some ideas:
- Draw a bead of RTV (high temp automotive engine head gasket type) on 3 sides of the BGA chip. On the open 4th side, use a blunt end syringe full of diluted silicone grease to inject the goo UNDER the BGA chip. For dilution, use xylene, toluene, Goop-Off, acetone, or any organic solvent that does not have much water in it. Reagent grade alcohol will work, but the 90% stuff will leave water behind. You want the stuff to evaporate on application. When the gap is filled and the solvent evaporated, seal the 4th side with more RTV.
- Same as above except use some kind of thermally conductive liquid. Fluorinert comes to mind, but is rather pricey. Maybe 100% antifreeze or your favorite water free salad oil instead. Same ritual as above except there's no evaporation. However, there is a risk of a leak, which should be tested with an UV dye tracer and a UV flashlight. You can find UV dye at the local automotive store. Get the one type made for engine oil or A/C leaks (i.e. no water):
- Same as above except fill the gap with silica (quartz) sand. We used to do that for "potting" discrete A/D converters that required thermal tracking between components across the PCB. You can find silica potting sand at any plant nursery except that the stuff is really coarse. For the amount you need for testing, just beating on it with a hammer should produce enough silica dust to fill the air gap. Use a face mask when handling the stuff. Seal 3 sides of the BGA chip with RTV, fill the air gap with sand, and seal the 4th side. No need for a leak detector.
- Use an FBGA socket. There's more mass closer to the package, which hopefully will act as a heat sink. It's probably one big impedance matching nightmare, but it's still worth a try.
trace.
Good plan, but I think filling the air gap with something thermally conductive will have a greater effect.
True.
If you wait long enough, everything else sags except the implants. At my age, I'll take anything I can get. Perception is everything and if that proves insufficient, just close your eyes and use your imagination.
My office purge has been procrastinated until Monday. It's too nice a day to waste at the office. Good luck on the audit. Be sure to leave something obvious and disgusting to distract the inspectors and give them something to complain about.
Thanks.
Cyclone III family package. FBGA package. Thermal resistance specs at:
32.4 degC/W from junction to ambient. 8.4 degC/W from junction to case. 15.6 degC/W from junction to base.Using only the BGA package as a heat sink, in still air. However, that's not the case here because some of the heat is being removed by conduction through the PCB. Without the PCB heat sink and with 1.5 watts dissipation, the BGA case alone would run at: 25C + (1.5W * 32.4 C/W) = 73.6C
Since the case was measured at 60C, the PCB is acting as a heat sink, the junction temp at 1.5 watts is: (1.5W * 8.4C/W) + 60C = 72.6C
However, there are other devices nearby raising the ambient temperature and a PCB heat sink with an unknown temperature. I can't go much further some measurements. Got any numbers for:
- Average operating temperature of the PCB near the FPGA?
- Average operating temperature of the PCB away from the FPGA? With those, I can guess(tm) that thermal resistance of the PCB, which can the be compared to the various rubbery and foamy things to see if they will offer much of an improvement.
Agreed. With 8.4C/W on top, and 15.6C/W on the bottom, the hot spot will probably be even hotter UNDER the FPGA.
Dunno. Unfortunately, there's no room under the middle of the chip or I would suggest you leave a hole in the PCB dead center so that a syringe could be used to inject some manner of thermal compound as I previously suggested.
Work out the cost benefit ratios. Cleaning the office will gain how much new business and cost you how many hours of work? Fixing the thermal problem will salvage an irate customer, that offers existing business at how many more hours of work?
Anyway to run this thing with the FPGA turned off? Then, measure the average PCB temperature and the ambient room temperature for the approximate temperature rise produced by everything else. If you can't turn it off, maybe slow down the clock temporarily?
Also try a thin black LDPE (low density polyethylene) plastic trash bag. IR goes right through it. Start at 2:44 for the garbage bag. As a cover, there will be some thermal radiation through the plastic, but the hot air will remain inside the package.