Aluminum oxide insulator safety precautions?

AlO2 is pretty benign. It's not toxic, so is as dangerous as any other dust... sawdust, glass, sand.

It's not a very good heat conductor. AlN is much better. BeO is even better, but BeO is seriously toxic.

John

Thanks John.

I was considering AlN until I saw that they were about $8 each (TO-247, 100pc lots)! I was hoping that AlO2 was good enough to be worth the 75-cents they cost. :-)

At $8 each for the AlN, hard-anodizing the heat sink becomes the better solution. And you can't beat the thermal impedance of anodizing.

You wouldn't happen to have a source for AlN TO-264 insulators at a decent price, would you? TO-247 insulators can work, but the MOSFETs are TO-264.

John

If you have the option to hard anodize, then that's the best bet, as long as it can take whatever voltage you're applying. Use good heatsink goo, and you should be fine.

Failing that, what's wrong with mica?

Good Luck! Rich

Max voltage is 55VDC and we're using 1 mil Type III anodizing (0.5mil penetration, 0.5mil surface).

Unfortunately, at the low volumes I'm working with (heat sinks) it's pretty expensive to hard anodize. I did six for $30 each, with six MOSFETs on each sink. If I can find a decent solution for a lot less than $5 a MOSFET (with low thermal impedance as a priority), I'll take it!

I thought mica had a pretty high thermal impedance? I did a quick Google check and found a thermal conductivity number of 0.71W/mK.

Hmmm...using the 15.06W/mK value for the AlO2 insulators I purchased, that makes the AlO2 a bit over 21 times more conductive. So, if I can find a mica insulator at least 21 times thinner than the 80mil AlO2 insulators I have, that means that the mica would be a better choice? That means the mica must be about 3.5mils or thinner (and still insulate to 55V at that thickness).

John

The advantage of mica is that it's usually very thin. 3 mils is common, and 80/3 is 27, so the mica wins over the 80 mil alumina. 1 mil hard anodize would of course be a lot better than either, and would easily stand the voltage.

I did get a quote, and some sammples, on some AlN insulators, and I recall numbers closer to $2. I'll see if I still have the info.

Are the fet drains common? The best insulator is no insulator!

John

I just found 3mil mica from Keystone, so there's hope! But I remember that mica had even worse thermal ratings than Sil-Pads (and equivalent insulators) and those pads had wayyyyy too high a resistance for my uses.

Uh-oh, I just realized that if typical mica was about 3mil thick and that mica's specs weren't good enough for the power levels I want to reach with these MOSFETs (acting as an electronic load) then the AlO2 I have may have too high a resistance too.

LOL, even with the cost, hard anodizing is looking better.

Alas, each drain has current sense resistor for the "servo loop" controlling each MOSFET (as it acts as an electronic load). :-(

$2 or so is a great price. If you had access to that AlN info, that would be great! Thanks.

John

Sil-pads are garbage. Even with huge mounting pressure, they never meet the specified thermal conductivity. And they're thick, so thermal resistance is high.

Move the resistor to the source!

John

Hmm...good idea. I could do that (dealing with some high common mode voltages, manageable though) but I'd still need to electrically insulate the heat sink and prevent any touching. An enclosure can solve that problem but there's a LOT of heat to remove, cost of enclosure, etc.

LOL, there's never an easy way!

I think I'm seeing why hardcoat anodizing is used. It's keeps me from dealing with pads and grease, speeds up assembly, already provides the insulation and it looks good. Probably worth calling around for better prices.

I'll still test out those AlO2 pads I have and perhaps some AlN pads too (if you find that info for the ~$2 pads). Gotta' get me some learnin' about those things and their effect, even if the anodizing looks like the best solution

I did try Dremeling out a hole for a thermocouple in one AlO2 pad earlier today. Hah! I thought that 2 of my grinder/cutter bits were carbide but all they did was leave a steel-color smear on the surface of the pad. Time to go shopping for something a little harder. :-)

John

There are big differences in safety when these things are respirable dusts. Sawdust, and probably alumina (aluminum oxide or corundum) are typically in the "nuisance dust" category, although alumina at high concentrations and/or long exposures is known to cause scarring leading to Shaver's disease. Silica (silicon dioxide), however, is a major component of sand and glass and is a much more serious problem. When the particles are fine they cause permanent scarring, enough of which leads to slicosis. Those who work with mineral dusts wear respirators rated to remove this stuff. (NOT the silly little "nuisance dust" masks sold to painters and woodworkers.)

Yeah , BeO is scary stuff! For airborne dust, it's 50 times more toxic than Arsenic. Before I got into pottery and started reading up on all this toxicity info, I acquired some TO3 BeO pads. Now I think I'll just let them sit in their little plastic bags in their little plastic parts drawer....

Best regards,

Bob Masta D A Q A R T A Data AcQuisition And Real-Time Analysis

Scope, Spectrum, Spectrogram, Signal Generator Science with your sound card!

We do hard anodize and as I recall it's not real expensive. It's probably worth it when you consider how many fets and associated driver/sense circuits you may be able to eliminate. Make sure the heatsink is *flat* first... most extrusions are wavy and need to be fly-cut flat if you want the lowest theta.

You could also consider heat spreaders: a small, chunky aluminum or even copper plate between each fet and the heatsink. The fet is in direct contact with the spreader, and the spreader is insulated from the sink. The bigger insulator footprint reduces theta.

What is really cool to measure is true junction temperature. To do that, characterize the substrate diode drop vs temp at some small current, 10 mA maybe. Then arrange to power up the fet and then suddenly, not too many microseconds, disconnect the regular stuff and reverse bias 10 mA, and measure the substrate diode again. You can even plot the cooloff curve and estimate various thermal time constants.

John

Apparently it's fine as long as you don't grind it up or anything. You can still buy BeO slabs fabbed to your specified dimensions. I can't imagine how they manufacture and slice up the stuff, at least how they would do that in the USA.

John

Good tip on the heat sink flatness. The mounting surface is milled to a Rmax of 6.3 (I'm assuming mils). Not a mirror finish but it's not bad. I couldn't find any other info on how much my theta-case-to-sink could be lowered with a smoother surface.

I was wondering earlier about using a spreader (100mms square) or individual spreaders.

I'm using six TO-264 FETs on a 100mm square sink mounting surface with less than .5" between each FET and between each FET and the edge of the sink. A bit over 34% total coverage. Do you think that a spreader (or spreaders) could lower the total theta-case-to-sink enough to be worth the theta-case-to-sink penalty from the extra greased surface (of the spreader)?

Whoa, that's a great idea! :-)

OK, so I'd measure the body diode voltage drop at a low current level at several different temperatures representative of where the MOSFET would be operating (probably by having another MOSFET on the sink running as a load and waiting several minutes for thermal stability for each step).

I could set up a LM317 for the constant-current 10mA supply. I'd then use two MOSFETs to switch the main current in/out and the LM317 supply out/in. A little bit of care in the timing to not blow anything up and I think I could do this. And I have a nice shiny new Natl. Instruments DAQ module just itching to plot the body diode voltage!

Thanks again for your help John.

It's possible to work safely with any damn thing if you throw enough money and paranoia at it. In petroleum refining, heavy hydrocarbons are catalytically 'cracked' and lighter ones are polymerized to make gasoline. The catalyst for the first process is silica; no problem. The catalyst for the second is *hydrofluoric acid*. I like to think that any bean counter who suggests cutting back the multilayered percautions around the polymerization plant is shown a safety video that gives him nightmares.

Aluminum oxide is an extremely hard material. I remember trying something similar back in grad school, and getting that same grayish metal deposit on the previously-white ceramic.

Time to shop is right, for new carbide bits to replace the ones you've dulled. If any thing is going to work with the Dremel I think you'll need diamond tooling, but even then you may just rub off the diamond grit before it cuts a hole through the aluminimum oxide.

Regards,

Mark

I never drilled through the AlO2 pad as the Dremel bits were wayyyy too expensive to even consider using, but I got some testing done.

I couldn't test the case temperature of the FET thru the AlO2 pad but I did measure the drain lead temperature (as it exited the case) and using that, along with case temp. w/o the pad, derived the approx. junction temperature for the FET with the pad and without.

It appears that the TO-247 Aavid AlO2 insulators I'm using have approx. 0.165 degrees-C/W thermal resistance using a good grease. This is about what a K-10 Silpad gives me. Not good.

Double checking my results by using the thermal conductivity of AlO2 (from Aavid) and that of a K-10 pad verifies the results above.

AlO2 conductivity = 15.06 W/m-K, 80 mil thickness K-10 pad cond. = 1.1 W/m-K, 6mil thickness

AlO2 pad is 13.3 times thicker than K-10. AlO2 pad is 13.69 times more thermally conductive.

The numbers balance out almost perfectly.

I would love to try AlN as it's, IIRC, about 15-20 times more thermally conductive than AlO2. Gotta' find it for a decent price though (

I think you're right. I went shopping for bits and the diamond ones were $15 each! IMHO, not worth the risk of just stripping the diamond off before I get through the AlO2.

Using the drain lead temperature of the MOSFET with and w/o the pad in place and the case temperature w/o the pad, I think I was able to extrapolate the junction temperature for both scenarios pretty accurately. Well, at least accurately enough for my purposes. :-)

John

No, No, No! You still need the goo. I prefer the white pasty stuff to just plain silicone grease - in fact I like it so much that when I buy a new CPU and heatsink, I clean their goo off the heatsink and use my white stuff. One number that comes to mind is Dow Corning DC-340, but the little tube of the stuff I have has lasted so long that I can't read the label any more. )-;

Good Luck! Rich

Thanks Rich! I should have said "grease AND pads". I was definitely going to use grease. :-)

Currently I have Aavid's phase-change UltraStick and Wakefield's non-silicone Series 126 grease.

The UltraStick has slightly better thermal specs but is hard to get thin enough to get those specs. The Wakefield grease is that great white pasty stuff you mentioned and goes on incredibly thin. A little messier than the UltraStick, but it's probably what I'll stick with.

I'll check out DC-340 too. I only have a small amount of the Wakefield grease so another tub of something else won't be a problem.

Thanks again! John

I don't trust the phase-change crayon stuff. It's likely to leave air pockets, and I'm not sure if it will ever really flow thin. I've measured the white Dow grease as squashing down to below 100 micro-inches under moderate pressure, about my resolution limit.

The phase-change pads are definitely garbage.

John