gap-pad thermal conductivity

I made this

formatting link

formatting link

to measure the thermal conductivity of a 3Gshielding gap-pad material. They spec it at 6 w/mK without any details.

I used a modified TO-220 resistor (tab cut off, thermocouple added) as the heater. A micrometer squeezed the gap-pad to tweak thickness.

I measured 2.7 w/mK at zero compression, and 5 at 50% compression. This soft stuff needs to be compressed. Between the change in conductivity and the change in thickness, the compressed thermal resistance is about 1/3 of not compressed.

I could buy one of these

formatting link

but it would cost about $40K.

My machined copper parts are done, so I'll try the AlN next.

That $40K machine won't measure AlN directly. It's only spec'd up to 8 w/mK, and AlN is more like 170.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

Isn't this a bit of overdesigning? Why not simply mount the actual device as it will be used, add the enclosure, apply power and measure the actual temperature?

Your real goal is to reduce the device temperature. Why not do so?

Reply to
Steve Wilson

The board in question is bolted to a machined, water-cooled baseplate, with thermally conductive insulators between. Many parts run at high power density so need to be cooled, but also need to have low capacitance to ground. I want to *design* the electrical and thermal situation, not experimentally iterate PCBs and machining, so I need to know the material properties. In this case, I need to plan for the proper compression of the gap-pad, and control capacitance but not bend the board too much.

And measuring stuff is fun.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Not very convincing. You have a number of thermal gaps between the IC and the water. You need to measure the device termperature in the actual application. If the temperature is below your limit, all is good. If it is over, no amount of thermal calculations will save you.

You love to measure. Your plots on devices have been very instructive. But temperature is a different animal. Find out if it is go/no go first. Then fool with measurements if they will help.

Otherwise don't waste your time.

Reply to
Steve Wilson

I don't need to convince you. All I want is purchase orders and checks.

The hacked customer demo unit got too hot at about 40 KHz. The rev A PCB makes it to a pulse rate of 1 MHz. They only need another 5:1 or so.

You have a number of thermal gaps between the IC and

But I do the thermal calculations before we design the product, not after.

No, it's not.

Find out if it is go/no go first. Then

I'm an engineer. I fool with measurements and math and design stuff that works.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

That's a very long way to go.

the first 30 dB of RFI/EMI isolation is easy. Then it gets progressively harder. The last dB is very tough. Then you give up.

Thermal is the same.

Reply to
Steve Wilson

The fancy machine only makes sense if you are making these measurements every day.

You will have great trouble measuring the AlN. You could measure one piece, then stack two pieces, but likely you will only end up measuring the interface between them, not the AlN itself. 170 W/mK is really high.

Jon

Reply to
Jon Elson

Thermal stuff is slow though. I made this TEC mount.

formatting link
I think the TEC is one of these.
formatting link

I couldn't get the cold plate below 0 C. I wonder if it was stuck trying to freeze the water? And the hot plate was then up to 45 C. (1.9 amp of current) So according to the chart about

1 watt of heat load... I guess that seems reasonable. I wanted more!

Let us know when you measure AlN.

George h.

Reply to
George Herold

No. My mistake was ignoring skin effect and toasting the inductors at

1/3 of their rated RMS current. Now I have numbers and aluminum nitride insulators to cool them down.

I don't design by soft analogies. And I don't give up.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Sure you do. Let us know how it goes.

Reply to
Steve Wilson

I'll have the rev B board in a few weeks. I have some pretty good gap-pad stuff, and a bunch of aluminum nitride insulators from two sources, one US and one Chinese. It should work.

The inductors have big topside pads with a lot of big thermal vias, down to copper pours on the bottom of the board. The board will press against the AlN, zero clearance to the baseplate, with thermal grease. Board flex will press everything together.

Coilcraft says that I can let the inductors get to 150C, but I hope that won't happen.

formatting link

I'll use the 5-turn version. Since I'm cooling the ends, the middle turn will be the hot spot. Thermal epoxy between turns would help, but would be messy.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

progressively

Wishfull thinking. What calculations have you done.

Wishfull thinking again. You do not hope a design will work.

No evidence of engineering going on here.

Reply to
Steve Wilson

Measured inductor temp rise on existing board, pulsing 1KV at 1 MHz.

Measured thermal resistance end-to-end of the inductor

Calculated via thetas

Calculated theta of bottom side pours through Aln to baseplate. Ignored fringing, so worst case.

Calculated inductor temperature rise per MHz pulse rate.

Calculated capacitances from various pads to ground.

Spiced circuit to see what those capacitances do.

Just measured the American and Chinese AlN thermal conductivy. The Chinese is better. I think the American stuff may be the machinable AlN, which is not as good a heat conductor as the pure stuff.

It will work. I just hope it will be below 150C at the center turn of the inductor.

POs and checks are evidence of engineering.

Show us something you've designed.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Not now. Maybe later.

Reply to
Steve Wilson

l.

as

8

ice

ual

But that does involve convincing other people, which in turn involves intel ligently designed measuring set-ups that clearly measure what they are clai med to measure and don't included embarrassing drop-offs (like the unknown heat coming out of the "other" side of a resistor).

"Only"? The obvious twiddles have got you a factor of 25 but you need anoth er factor of five ...

is

Which is how you started off with a "hacked" unit that fell a factor of 125 short of what you need? Not a great advertisement for your thermal calcula tions.

ut

You do need to think harder about how you measure it. A pre-calibrated ther mocouple-based instrument may be available off the shelf, but it's grossly sub-optimal as a measuring tool for what you ought to be doing.

If you were enough of an engineer to engineer your measurement systems, the stuff you "designed" would work better, and would need to go through fewer iterations before you could evolve it into something that you could sell.

--
Bill Sloman, Sydney
Reply to
bill.sloman

ly

It's also going to add capacitance between the turns.

Filling the gaps with thermal epoxy doesn't look that difficult. If you tur n the inductor upside down the paper label on the top gives you a saucepan into which you can pour thermal epoxy, and in which you can subsequently ba ke the epoxy until it gels.

If it goes through a low viscosity phase during the cooking process, you'd have to mask the gaps with masking tape, which would make the process even fiddlier, but you'd do it before adding the just-mixed epoxy and hardener, so it wouldn't make it any messier.

Dispensing a well-controlled volume of just-mixed epoxy with a throw-away s yringe shouldn't be all that messy, but sloppy people can create a lot of m ess when they don't actually have to.

--
Bill Sloman, Sydney
Reply to
bill.sloman

As I understand it, machinable ceramics are micro-cracked, which doesn't make them impure, but would lower the thermal conductivity.

Why 150C? Is there a high permeability core embedded in the inductor? Since it doesn't wrapped around the windings, it wo't make much difference to the inductance. If it's the enamel on the wires, it shouldbn't matter.

They are evidence of testing, which could equally be part of the process of evolving a design. Engineering is about anticipating problems - which doesn't mean expecting them, but rather building in features that get around predictable problems.

Not something John Larkin does - he shows us snippets out of work in progress, which would be, by definition, examples of failures in the design process, if there was one going on.

--
Bill Sloman, Sydney
Reply to
bill.sloman

  • What is this BS using very light grey text? I someone trying to hide information, in hopes of selling it? This crap of having obscene (unreadable) text is becoming rampant.
Reply to
Robert Baer

Argh!

With pasta the single thing you need to be able to read on the packet is the cooking time. So what does one supermarket do? Uses 0.1" high light yellow ext on dark yellow background.

Makes my blood (cf water) boil.

Reply to
Tom Gardner

Here's the AlN test.

formatting link

formatting link

formatting link

formatting link

It's probably not a very accurate measurement, maybe 50% low, but it shows that the Chinese parts are as good as, actually better than, the USA-made ones, which is all I need to know now.

Even 77 w/mK is damned good for an insulator.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

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.