No, I want the pad between the bottom of the board and the aluminum housing.
Both chips are BGAs, and both have about as good a PCB layout copper thermal path as is possible, on multiple layers. But the bottom of the board is 60C, and I want to pull that down.
-- John Larkin Highland Technology Inc www.highlandtechnology.com jlarkin at highlandtechnology dot com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom timing and laser controllers Photonics and fiberoptic TTL data links VME analog, thermocouple, LVDT, synchro, tachometer Multichannel arbitrary waveform generators
Hi, Klaus,
At 0.276" thick, it just might squish down properly in my 0.25" gap. I'll try some and maybe post some thermal images. Thanks.
-- John Larkin Highland Technology Inc www.highlandtechnology.com jlarkin at highlandtechnology dot com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom timing and laser controllers Photonics and fiberoptic TTL data links VME analog, thermocouple, LVDT, synchro, tachometer Multichannel arbitrary waveform generators
You might want to check out Bergquist, but the stuff isn't cheap.
try
.highlandtechnology.com jlarkin at highlandtechnology dot com
It looks nice. It's a bit hard to understand the data sheet. But it looks like you need a 60% 'deflection' to get the rated 3.2W/m-K (I'm guessing the data sheet quotes the 'best' numbers.) At only .25" out of .276" that about ~10% deflection (?) or only ~20% rated thermal conductivity (?) And for more 'deflection' are you going to be stressing the pcb.
I'm not sure at all about those numbers though. It'd be nice if they called defelction compression if that's what it is.
George H.
Probably an old AMD Sempiron or Athelon CPU. There are YouTube videos showing meltdowns (but no flames).
Intel CPU's will survive running with no heat sink because it has a built in thermal sensor, that slows down the clock and lowers the supply voltages when too hot. Eventually, it shuts down the power supply, reboots the server, and/or sends alarms: I know it works because I've tried it. The problem is that some system builders don't do a very good job of attaching the heat sink, add too much silicon grease, don't properly secure the heat sink, or sometimes ship a heat sink with one out of four retainer clips missing. The result is lousy CPU cooling. So, the CPU runs at reduced speed and the customer things that Windoze is "slowing down".
The big problem is how do you know if you've done it correctly? I use a digital thermometer with a tiny thermistor probe at the tip. I run the machine for about 10 minutes and then measure the CPU temperature and the heat sink temperature near the CPU. They should be within a few degrees C of the same temperature. I varies by CPU type, heat sink size, fan size, location of the thermistor, etc. If there's a big difference, the heat sink isn't attached correctly.
In my never humble opinion, the thermal problem is not limited by device or heatsink roughness, but rather by flatness. The heat sinks are delivered straight from the extruders, with no secondary milling operations to straighten out flatness issues. Take a steel rule and place it on the bottom of a typical heat sink to check flatness. Some CPU's and GPU's (video chips) have mirror surface finishes, which are great, until ruined by a thick thermal pad or thick layer of silicon goo. These would work much better without thermal goo, but only if the associated heat sink were also polished to a mirror finish and reasonably flat. Flatness is far more important than polish [1] My test for flatness was to smear on some light oil, and squeeze the CPU (or power xsistor) together. When pried apart, if there were any blobs of oil left on either surface, it's not flat. I've also used a Moire pattern test, but that didn't work well (for me) with small areas such as GPU's. To the best of my knowledge, you can't buy flat and polished aluminum heat sinks.
[1] Note that abrasive lapping of the aluminum surface with carborundum is a lousy idea because it imbeds abrasive material in the aluminum surface.-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
If you did that measurement with a closed enclosure, the internal heat sink won't do anything. With no place for the heat to go, it will just build up on the heat sink. Basically, you're trying to convert thermal conduction into radiation. The heat sink needs to be attached to the case which will do the radiation.
Revised suggestion: Take a sheet of aluminum and slide it into the extrusion in the closest slot above the PCB: You may need to make a cutout to clear the power connector. Jam a block of aluminum, copper, graphite, or whatever between the aluminum sheet and the two chips. See if that helps. Don't worry about flatness or smoothness at this point. Later, you may need to cut the aluminum sheet lengthwise and add an insulating spacer to thermally isolate the two chips (to keep the hotter chip from heating up the other chip).
How many watts is each chip consuming? I can use that, and the 60C temp to make a rough guess as to how large a heat sink might be required (assuming decent thermal conductivity to the chips).
be
It could also mean that the FPGA package is a lousy heat conductor and that most of the heat is remaining in the die.
If all else fails, add a cooling fan on the cover. It looks like you have plenty of room for the fan, ducting, plenum, and outlet ports. Hopefully, you don't have components on the board that require thermal tracking each other.
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
We have a few hundred units in the field already, and I don't want to change the PCB layout. My customer is worked up about the FPGA temperature and, frankly, I'm not. Some retrofit gap-pads might solve the "problem."
We do a number of products in this same extrusion, so having a way to move some heat would be useful in the future, too.
We have matched impedance traces on the bottom, so I don't waht to crush the board against metal, even with a thin insulator.
Here's a thermal image of the FPGA. It has a bad hot spot in the center.
A heat sink on the top of the FPGA may help, discussed in another post.
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
Incwww.highlandtechnology.com jlarkin at highlandtechnology dot com
The folks in the thermal pad business are, in general, chronic liars. I'll get some and try it.
But most anything is a better thermal conductor than air. I guess I could try a small baggie filled with plumber's putty or something, just to see what happens to the FPGA temperature.
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
I'm not sure you are barking up the right tree. You say you want to cool the PCB, but aren't concerned with the chip temperatures.
I see you picturing something that I don't think is realistic. You seem to be describing something that has significant thickness will conduct heat well, but not be electrically conductive. Unless you use something expensive like ceramic, I don't know of anything like that.
It looks like you have a to fill of over 0.1 inches. This might take as many as three layers of the thermal pad you seem to like. I read the data sheet on the stuff and they don't specify a compression level, so maybe it will conduct heat without being squeezed. It is cheap enough, so it should be easy to try. What temperature do you want to achieve?
I'm betting that using 100% thermal pad doesn't do much for the temperature of the PCB and less for the temperature of the chips. Remember, it may seem like the board conducts the heat away, but it is really an insulator for the most part. But it is pretty much the *only* path for the heat from these chips other than a small amount radiated and convected.
Actually I'm not sure why you want to cool a 60°C board. Is it because you think the chips are too hot? If so, I think you will need to add an Al slug to the top of the chip with thermal connection to the chip and the case. If the PCB won't handle the flex, add a support below the chips which can add more cooling. If you are concerned with electronic effects from the underside slug, how much spacing do you require? If the 0.060" thermal pad is compressed to 0.030" will that be enough? Maybe you can add a non-metalic spacer underneath.
If you think the top side slug will be insulated too much by the plastic case, use the top side as a mechanical support for the thermal slug on the bottom. One or the other has to work.
This is very much analogous to electronic circuits. Make a spice simulation with appropriate resistances and current. You will quickly find out where the problems are and how well solutions will work, assuming you have some reasonable data to work with.
-- Rick
Did you ask if that spec is under compression? How much do they need to be squeezed to make decent contact with the heat sink?
-- Rick
What you say is true, but he isn't trying to get every last drop of thermal resistance out, he is trying to improve on the thermal conduction of the PCB. But I agree a 0.25" gap is hard span without metal for the bulk of it.
He says he is worried about electronic effects between the slug and the traces. So a few 10's of mils thermal pad may help here (with the emphasis on *few*).
As to general situations, I spent a little time hot rodding PCs some years ago when it was all about optimizing your heat sinks. I never heard anyone say get rid of the thermal grease. It was however, all about minimizing it by spreading it very thinly using a razor blade. It was also very important to get a polished surface using a glass plate as the polishing surface. That was pretty extreme stuff however, not really suited for mass production.
I would be very interested in seeing a test of more conventional apps where the surfaces were not polished and extremely flat. I bet in those cases thermal grease makes a *huge* difference.
-- Rick
The reason to pull down the PCB temp is to also pull down the temp of the chips. I sure don't care about heat stressing fiberglass.
Practically anything solid is going to be orders of mag better a heat conductor that what I'm using now, which is air.
The compressible pad things have thermal conductivities in the range of 1 to 5 w/m-k. Air is 0.025.
The bottom of the board is running around 60C, when the extrusion is maybe 30. That's a 30C rise. It would be great to cut that down to 15, or even 10.
The FPGA designer wants the Cyclone die temp to stay below 85C, and the customer is mumbling about ambients of 40. Dropping the PCB temp rise by 15 or 20C would sure help.
In general, it would be nice to have a material around that can be plopped between a PCB and a chassis to cool off hot spots.
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
I used thermal conductive epoxy to attach a heat sink to the top of the horizontal output chip in a tektronix 2465 scope and it lowered the case temperature by 10 °C. It is in a (not too effective) forced air cooling environment though.
Regards, tm
How about epoxying an aluminum block to the top of each chip and connecting it to the cover with a flat head screw from the outside?
Regards, tm
65°C as a case temp is not so bad. Too bad you can't educate your customer. I think an inch or so of thermal pads on the top will help. The IR camera won't see a temperature nearly so hot through that much padding! lol
-- Rick
The gappad VO is what I used. Digikey has the best price for this stuff and sells it in small chunks.
For what you want to do in spreading heat this would work fine and should reduce the 60C nicely. You should have no problem getting down into the 40C range but you will have trouble getting much below that.
That would work, but it could really stress the BGA balls. The cover slides in/out horizontally.
I'd rather try to get the heat out through the bottom of the board, where nothing is going to move much.
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
I just needed to ask to make sure I understood what you were doing.
That is an assumption really. The "air" is the entire space given that the entire board is what is radiating to the air. When you add a square inch or so of thermal pad you might not be changing things much at all. A test is a great idea. I'd like to see your results.
Apples and oranges. The air is not limited to the small area you would likely use for the thermal filler. Air also convects. But the proof of the pudding...
With proper filling of the space that could be a lot less difference than 15°C even.
The IC case temp is not the same as the die temp. The thermal coefficient for the case should be in the maker's data sheet. You can figure out a die temp very likely, or an approximation.
It would be nice to have a lot of things too... how would you make this stuff? Usually they use a milled copper or aluminum slug that has the right height at each chip to make good contact through some grease. But that all depends on how "good" you need.
That is one of the better sounding ideas I've heard yet. The only mod is to add a hole to the lid which from the looks of the extrusion may be slid into the grooves. The screw would be needed to make any real contact which is what the exercise is all about.
-- Rick
The cover is relatively thin, no? Just pull the cover up a bit when sliding it in, there has to be some slop in the groove to let it slide easily. Then tighten the cover down. No need to make it bend - you just want a good connection between the two. No stress on the BGA unless the mechanical engineers can't deal with tolerances.
-- Rick
e:
eed
ke
to
u seem
ct
ething
ake as
the
so
ough,
eve?
,is
*only* dbecause
cting
I use a similar looking box, the cover is the same material as the box, it doesn't bend at all and the fit is real tight
-Lasse
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.