new (to me) power fet package

On a sunny day (Tue, 22 Feb 2022 07:35:01 -0800) it happened snipped-for-privacy@highlandsniptechnology.com wrote in snipped-for-privacy@4ax.com:

Yes looking at teh datasheet is says .6 degrees C / W

Note 5 says 6W on a peeseebebe

...

The virtue of that package is the many source connections and presumably a lot of source wirebonds. Good for high-current pulses. Dpaks have that one dinky source pin.

An xray would be fun.

Someone could start a hugely profitable web site that included, among other things, part x-rays.

tirsdag den 22. februar 2022 kl. 18.40.16 UTC+1 skrev snipped-for-privacy@highlandsniptechnology.com:

maybe send one to this guy,

That's how inverters in some EVs work. I'm not clear on why the package has to be so small though, or how you clamp it down.

Anybody know how the automotive folks order parts? Do they call up digikey for special 100k, million and 10million qty prices? Do they haggle parts from $0.238476 to $0.238475 when it comes to price?

I believe most often the automotive subcons (Delco, Bosch, etc.) or their subcons deal directly with the manufacturers for such large quantities. I recall hearing about clauses in many manufacturers' distributor contracts allowing for direct sales for large enough volumes.

Pricing used to be absolutely brutal, maybe not at the micro-cent level but certainly at the penny level in my memory, and I doubt that's changed.

Yes, 250W if you can keep the case at 25 degrees. Good luck with that.

But I don't think they're really suggesting using it that way. It's just a data point for thermal calculations.

Sylvia.

I seem to remember an IRF app note on how transistor testing is done, everything is submerged in a special liquid that boils at 25'C

tirsdag den 22. februar 2022 kl. 16.35.20 UTC+1 skrev snipped-for-privacy@highlandsniptechnology.com:

seems similar, only marginally smaller than a TO263-8

I remember a presentation on a cooling method that could achieve very high thermal densities involving tiny nozzles that would maintain a thin layer of liquid on chips. That gave higher thermal transfer because allowing the liquid to boil resulted in a layer of insulating gas around the chip most of the time. The only problem they had was the coolant was an organic substance that had the property of dissolving most inks including the markings on the chips which were inked in those days. Now everything is laser etched and nearly invisible to start with, so not a problem anyway.

IR invented extreme fantasy fet power specs, and everybody else had to be competitive.

The boiling liquid, among other things, keeps the source lead from unsoldering itself or outright melting.

Some fet specs have a footnote "package limited", which means the specified current will melt the wire bonds. Thanks.

I believe they try to hit the peak,

onsdag den 23. februar 2022 kl. 01.09.11 UTC+1 skrev John Larkin:

I don't think that was the case. Notice the large temperature delta from the liquid saturation temperature to the peak of the curve. The coolant has to be liquid at something above room temperature to allow handling. Such a large temperature delta results in excessive junction temperatures. The goal was not to obtain the highest possible heat transfer, but rather a means of providing adequate heat transfer for every part on the board. The chips did not have droplets of coolant on them. They were sprayed so as to maintain an even, thin layer of liquid. The evaporation at the surface of the liquid provided an adequate means of cooling. They simply needed to prevent the liquid from being too thick so as to add excessive thermal resistance between the evaporative cooling and the device being cooled.

As it was explained, immersion was less effective because the bubbles forming essentially create an insulating layer which is less effective at cooling. A thinner layer does not form bubbles and so optimizes cooling. In essence, the issues of nucleate boiling are side stepped.

Now that I look at this with fresh eyes, I realize they were cooling chips in packages. Packages present significant thermal barriers. So maybe this approach is not adequate for the high thermal densities in use today. Or maybe it would work very well with appropriate packaging. They do use water cooling on CPU and GPU chips. They don't encounter boiling because the coolant is constantly in motion and never reaches the boiling point. However, the other advantage of the "spray" approach was the high density since they used a thin sheet to to distribute the fluid to the microscopic nozzles in the sheet. So that might be very effective for laptop use or other dense packaging applications. The context I saw it was military applications. I think they also were researching use of a diamond substrate as a heat spreader which would conduct heat outside the enclosure where it could be carried away by fluid. I don't think the two methods were intended to be combined. Seems synthetic diamond is not so expensive. I expect it was not single crystal diamond, but who knows? This was DOD research, like the Internet.

The automotive manufacturers will do things like removing the chip package, measuring the silicon area and telling the manufacturer what they are going to pay for it. Even if a subcontractor is designing and making the product they will still be heavily involved in such details when they think it necessary.

John

One of our big customers demanded that we document our costs and accept a small profit.

And a government facility claimed that they had already calculated our costs (including the engineeering) and demanded a price cut. They were lying of course.

We told both, get it somewhere else, but don't use our IP.

Quite right too! That would have been my response :)

If the engineering time was the dominant cost, I would have told them "pay peanuts, get monkeys".

Potential customers often insist on NDAs, and we are happy to oblige. We have valuable IP, and they usually don't.

Lol. Not sure what's funnier here, the bogus cost calculatios or anybody in government even pretending they can estimate or control expenses.

We did ask them to send us their spreadsheet.

It was a bizarre episode. I think the dollar amount of the contract sucked in some "expert" procurement droids. After we didn't cut the price, they split the project into two POs which apparently got both below some threshold.