Thermal conductive epoxy: powdered diamond, only $1200 for 5g

My experience is from the marine radio business, where a few dollars difference in selling price often made the difference in a sale. Well, not so much at the retail level, but certainly when selling to fishing fleets and distributors. So, we tried to use one RF power device in as many places as possible. That means compromises and learning to use mismatched and inferior components. At the time, there were only a few RF power manufacturers, all with parts that were allegedly compatible, but certainly not interchangeable. So, I had to design boards for one specific manufacturer's power devices. Changing vendors was not a trivial exercise. Today, due to automation and larger quantities, such parts are more consistent, predictable, and more available.

For RF, the limiting factor on obtaining rates power output is getting the heat out of the device. Spreading the heat around over a larger number of devices certainly would help. However, the cost and output of RF power devices is not a linear relationship. Doubling the number of 1/2 power RF devices would more than double the cost, as well as introduce added losses in power combiner devices. The trend is in the opposite direction, to put more dies into the power transistor package. For example:

Yep. Transistors get the benefit of automation. Heat sinks, much less so. However, with todays higher efficiency switching amps (Class D/E/F) and envelope tracking RF power amps, the total dissipation is less, allowing a smaller and cheaper devices and heat sink. Progress lurches forward.

In the case of heat sinks, who cares if the surfaces are planes or spheres? In fact, spherical sections would be *slightly* (read: insignificantly) better.

That's a feature! But sometimes I think that a very thin, unfilled oil is better if the surfaces are very flat.

Try two small pieces

Some sorts of spring washers are necessary, especially if there's FR4 in the stack. FR4 cold-flows, so fasteners will loosen up. Wavy washers are nice for PCBs, because they don't dig into the board and make many contacts.

I got the 100 uinch measured gap with Dow 340, with just seconds of pressure.

Do you think so? Why would the air preferentially displace the silicone?

The smear-and-mount thing guarantees air gaps, in my opinion. I don't know how to prove that.

At low frequencies, the tradeoff favors more cheap parts, like more D2PAKs instead of 500 watt TO247 types. And pick-and-place is a lot cheaper than hardware and screwdrivers.

It would make the devices interchangeable, which would be nice from a servicing point of view.

Cheers

Phil Hobbs

It does--I've seen lots of pictures from my old colleagues in the Packaging Research Department at IBM. (It was a weird place to be doing MIM tunnel junctions, but the people were nice.)

Thermal expansion pushes the paste out, but once it's out, there's nothing to pull it back in as the temperature drops again. So eventually it pumps out. Putting paste reservoirs underneath helps, because they're sealed from the atmosphere, so there actually is a restoring force.

Cheers

Phil Hobbs

If it amalgamates faster than it flows, all it'll do is harden the metal at the high spots. Dunno which is faster.

You got that right.

Cheers

Phil Hobbs

Really? They have been building high power solid state transmitters for decades. Harris was building a 5 KW AM broadcast transmitter with FETs in TO3 cases in the late '80s. It is needless to say this but they had to redesign the modules, when they stopped manufacturing transistors in the TO3 package.

Now, you can buy a UHF TV transmitter that will put out as much power as you can afford. They build standardized modules that are fed from a power divider, then the outputs go into a power combiner. The transmitter has an embedded computer system to monitor the system. It will report the total power out, reflected power and it will tell the operator if one of the RF trays fails, along with which tray is bad.

Huh? Read again, looks like 5,000x better than thermal grease.

That's millikelvins, not K.

Solid metal is a bit higher than silicone oil. By ~three orders.

Room temp solder overheats transistors? Or 50C solder? "Low temp solder" is below 100C. Indium-gallium type melts in your hand. Using normal tin-lead would be silly. But low-temp solder might not be stable against aluminum

Yep. Notice the 85 to 90% efficiency on AM. Basically, it's a collection of Class D switchers running 2.5kW at the carrier frequency. That much better than the old daze of class A and AB where one was lucky to get 35% efficiency.

TV efficiency is not so good running in the 20-25% range: See chart at bottom of page. Drop another 5-15% depending on the output filtering and/or combinging.

Ok, back to partying. It's a tough job, but someone has to do it.

Nope. Thermal conductivity, in SI units, is in Watts per meter-Kelvins W/m-K W/m/K

See:

I can see where you got that idea. Many typesetters leave out the multiply sign resulting in W/mK or milli-Kelvins.

Yep. Silicone transformer oil is about 0.1 to 0.2 W/m-K. Aluminum is about 200 W/m-K. However, mix some powdered metals or alumina with the smallest amount of silicon grease needed to not make a huge mess, and the resulting thermal conductivity might creep up to about 0.5 W/m-K. Phil Hobbs found some stuff that claimed 3.5 W/m-K. That's about it for the greases. Metals are more interesting. W/m*K Diamond 1000 h-BN 600 c-BN 740 Silver 406 Copper 385 Gold 314 Aluminum 205 Graphite 200 Carbon 150 SiC 120 Brass 109

None of the above. I was thinking of trying Alumiweld or one of the other aluminum brazing or soldering rods. There's some debate as to whether it's soldering or brazing. The problem is that the brazing or soldering temperature for aluminum is high enough to damage the transistor and warp the heat sink. Still, it's tempting.

Ok, back to the party...

Don't herrings smell after a while? Still, I bet the contract doesn't specify odour limits.

NT

Yep. One of my favorite munchies is "raw" (actually lightly pickled) herring with onions and pickles. I like it smoked (which is where the red color comes from), but it's difficult to find in California. This dish produces the ultimate dragon breath, which is useful for ending meetings quickly, and repelling undesirable visitors to my office. The effect lingers for hours so a little goes a long way. Tastes good too.

I wasn't very serious about using a red herring in my next project. I've been doing mostly design reviews, and analyzing science fiction product proposals. I doubt if any company would trust me to design anything that actually needs to work. Still, if they're foolish enough to do so, I would probably include some obviously bad idea or useless feature, that invites instant criticism and eventual deletion. The managers, having satisfied their compulsion to kill off some part of the project, will then not bother me.

Yup, it just squishes like putty.

--sp

Yup, essentially forms metal joints at room temperature on contact pressure. To create a gas-tight O-ring, you just overlap the ends and squeeze. If only other materials worked this way :)

Right, the anodize isn't safe much over 100 volts.

Most of the AlN people seem dedicated to not selling parts. I did have good response, and some nice samples, from

Or DeBeers disappeared him.

IMO brazing is soldering using a bronze filler.

solder a copper plate to the heatsink using the special solder and then use ordinary, or low temperture solder to attach the transistor to that.

Al foil seems like a good fill to me. Or gold leaf if surface is flat to a few microns anyway. But that's only my idea, you would have to prove it out in a lab.

I don't know what the noise about gold is for - it has worse thermal properties than silver and cost heaps more