thermal baggie things

One of the designs I cleaned up was an HF power amp that used a cast aluminum heat sink. I don't recall the casting process used. I sectioned it with a diamond saw. It looked like it was made from beach sand. No amount of polishing would ever get it smooth enough to act as a reasonable heat sink. My fix was to polish the surface with a fine aluminum oxide abrasive. That made it flat, somewhat polished, but also filled all the surface voids with something almost as thermally conductive as extruded aluminum. However, it wasn't perfect, so I added a really thin layer of silicon grease applied with a plastic razor used as a spatula. The metal razor blades would gouge the aluminum, while the plastic blades did not.

...

'll try

it

Yeah well the thinner pad the better, (obviously).

Air convection can be ok. Here's a crazy idea, orient the box with the pcb vertical and stick a partition between the pcb and Al extrusion, with long slots top and bottom, and set up an air circulation. (I wonder if that would work? .25" is not much room)

George H.

ed text -

The PCB is an insulator. I think this won't work.

There are so many variants of BGA that it doesn't mean much. You can have BGAs that use boding wires, and thus are worse than leaded components, that at least have some thermal mass due to the lead.

Have you contacted the factory and asked to talk to a packaging engineer?

The PCB is still a far better thermal conductor than the 1/4" air gap below it.

Both chips have lots of vias to the ground and power planes. That spreads the heat out laterally and through the board thickness. The ADC has a bottomside power pad, which vias to a big bottomside copper pour. The problem is, once the heat is spread out across the PCB surfaces, it has no path to the outside world except the big air gaps.

What factory?

I know that the PCB is about 60C and the two main chips are about 70C. I want to pull all of that down.

Oh it will have an effect, we used to stick these on top of SO-8 LTC regulators and squeeze them against machined blocks of aluminum.

Cheers

Some experiments:

Actually, in the photo it looks like the PCB is mounted to the bottom with standoffs. John will do what John wants to do regardless of what would make sense to the rest of us. Certainly the solution he is suggesting is the short path and it will gain him something. We don't know if he is shooting to lower the temp by 5°C or 20°C. He might get a

Since he is mounting the board on the bottom, it might be practical to turn the board over and use a metal spacer between the heat sink and the FPGA/ADC with heat sink grease or a thermal pad, adequately compressed. It looks like the end plate is reversible so it won't change any metal parts other than the mounting holes if they are not symmetrical.

Also notice that the ADC doesn't even show up while John is saying it is also at 65°C. I find that a bit odd.

That is a very good idea actually. All this is because they are worried about the temperature of the chip, but they have no idea what the chip temperature is. When someone said he should contact the factory he said "what factory?". Sometimes the people who make the parts actually know something about using them... sometimes. But certainly even just adding a heat sink inside the case with no attachment to the walls will do better than nothing. That will get the heat into the air more easily and lower the case temp of the FPGA.

Don't bother me with the facts!

You don't need to push the blade, you drag it. The point is that the razor blade is very flat.

to

The factory that makes the chip. That is where you find packaging engineers.

Alternatively, there is no shortage of companies in the south bay that can Xray chips. Once you know the internal construction, you can determine the best thermal management.

Or you can just ask questions on usenet if you don't believe in science or using facts.

Another problem is the big arse heatsinks are kind of heavy and put stress on the PCB. Ship a PC and the shock can move the heat sink.

The trend towards water cooling does take the stress off the mobo. Also it gets the big arse heat sink out of the way of the RAM, which like to be close to the CPU.

The intel theremal sensing scheme just feeds a current to a diode, running it at 1x and 8x, then looking at the delta.

When you use a random diode on a chip for temperature determination, you use a constant current, put the chip in an oven, sweep T and measure V. The current needs to be low since it is carrier injecting when the chip is powered up.

Any significant to the hot spots by the pads? Obviously it would help to know the chip.

While we're at it, that plastic blade from Home Depot is actually metal. Just the handle is plastic.

I get plastic scrapers since often they do the job and you are less likely to damage the surface. Wally World has them. In a pinch, you can use the cheap plastic scrapers if you are going to snow country. Often bay area store don't have items for winter weather.

Of course, there is the issue of assembly errors. I ended up debugging a customers temperature problems which ended up being someone not being able to count the layers of kapton used.

to

You have to have Gold or Platinum-level support from Altera to get any human to give you attention. We don't.

It's a discussion group. It would die out if nobody ever introduced topics.

Asking for suppliers of thermal interface material isn't unreasonable. I still haven't found a source for the gel-filled baggies, and I was hoping somebody knew.

That is a nice result. Should at least double the MTBF.

Maybe that is where the ADC is. He said the ADC runs as hot as the FPGA. If so, the ADC may be putting out more heat total than the FPGA. That looks like it might be making a major board hotspot.

Why is it FLIR anyway? I know what it stands for, but what direction would you be looking other than forward? Isn't that just a matter of where you point the durn thing? I guess it is a holdover from the use in aircraft and in this case it is just a company name. So this is really just an IR camera... lol

Cool, thanks for the follow up. George H.

There ought to be commercial products, thermal sandwiches in a few common thicknesses, a few sizes of discs and squares.

The P4 and above heat sinks that I've seen have mounting screws that go through the PCB, and attach to threaded standoffs on the case. There are a few odd PCB's that use a different arrangment, but all of the PCB's in my palatial office (I checked) use this arrangement.

I prefer the fish tank approach, where the entire PC is immersed in a tank of anti-freeze or mineral oil. Works nicely, but does tend to be rather messy.

Methinks it's also patented by Intel. Meanwhile, TI came up with the TMP006 infrared thermopile, which sits under the CPU and measures its temperature. No silicon grease or mechanical contact required.