On Sun, 13 Dec 2015 11:27:47 -0500, Neon John Gave us:
Use hard anodized Al for the sink. That surface is non conductive (if done to mil spec).. Then use silver filled epoxy for the attach. That is what chip makers use to attach the die to the substrate.
Better might be to use a paper thin anodized Al sheet clad onto a copper mass for the sink.
On Dec 12, 2015, Jeff Liebermann wrote (in article):
An ordinary vertical mill with a saw blade will do. The key is to embed the specimen in epoxy, cut, and polish the cut face using wet/dry sandpaper followed by metallurgical polishing film (diamond dust in mylar film) face up on a sheet of float glass.
Joe Gwinn
Kapton has a high thermal resistance, maybe OK if it's really thin.
The phase-change stuff that I've tried was terrible, because it was thick and didn't really flow out from under the part. There's probably better stuff around. If it really liquifies and flows, it's probably OK.
If you need low theta and insulation, you might consider a thin slab of lapped AlN and grease or epoxy. I'm seeing prices like $1.50 for a TO247-size piece, 10 or 15 mils thick. Of if the voltage is low, under 100 volts maybe, grease over 1 mil thick hard anodize, as the Decadent guy suggests.
If thing get hardcore, use a copper heat spreader and grease, and bolt the copper to an anodized aluminum heat sink. That will do good lateral heat spreading, which can be the bottleneck with aluminum. The spreader magnifies the area of the insulator, too.
I want blocks of solid isotopically-pure diamond. I wonder why diamond is so hard to make.
Double-check that; your browser may be doing autocomplete behind your back. Those are the same supplier, but very different pages. Hydrothermal is a pressure-cooker crystal growth scheme, while the less expensive comes from a Czochralski type plasma scheme, which has more impurities and flaws (this might be OK, if you wanted a star ruby). I'm not sure how either scheme would work out in a garage or kitchen.
Bingo. After flushing my cache and checking the URL, I get a cheaper page. However, 10 mm is now $25 and far away from $2/ea.
Thanks, but growing crystals and semiconductor processes are some of the many areas of electronics about which I know nothing. I found a bit more on growing various crystals in a microwave oven, but nothing for rubies.
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
Oops. Different mistake this time. The $25 is the "lot" price for $25 pieces. The unit price is $1.28, which is less than $2.00. Sorry(tm).
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
On Sun, 13 Dec 2015 09:58:25 -0800, John Larkin Gave us:
Maybe they'll make some fine grain diamond in a graphene sheet matrix.
I think that sounds pretty good, actually.
Not thermally related, but DARPA got the world record for highest frequency amplifier. They use HEMT and they are using a vacuum to get Terahertz operation.
Or low-temp solder? Even room-temp solder?
Possibly destroys Al heatsinks progressively when warm, same as mercury does.
Do copper slabs display mercury rot, LME, like aluminum has?
We used a silver filled sheet epoxy for bonding PZT8 piezo ceramics to aluminum. But first the two surfaces were abraded against each other with a carborundum abrasive using a figure 8 pattern and constantly rotating the ceramic. The aluminum, sheet epoxy and ceramic were mounted on a heated clamping assembly with 20 psi of clamp pressure. These were 2"round ceramics on a 4" round aluminum piece. Mikek
Lapping in the surfaces can help a lot. One tip (that I learned from Usenet) is that lapping two surfaces makes them both spherical (with some very large radius in this case), but using three surfaces and lapping them pairwise forces all the radii to infinity, i.e. all three surfaces become very flat.
The reason for this, I think, is that the constant change of direction and orientation forces the surfaces to be spheres (i.e. independent of orientation), and the odd number of surfaces means that the surface can't be either concave or convex. If two of the surfaces were convex and one concave the two convex ones would flatten each other out in the middle when they were lapped together, which would force the third surface to become flatter as well. That means that the three surfaces all tend to become spheres of infinite radius, i.e. planes.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Mental block, lapping was the term I wanted :-) Any radii created by lapping was much less of a problem than the difference in expansion of the ceramic vs aluminum. After the transducer had cooled it had a slight concaved surface. It wasn't a problem in use, just an observation. It also produced a big spark when it cooled. Mikek
Yes, one problem with a too-thin bond line is that the shear stress in the cement is maximized. Something a bit thicker can accommodate the thermal expansion a bit.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
We didn't have delaminating issues, our big thing was efficiency of the finished transducer. A thicker bond line, generally caused lower efficiency. We were driving a 2" diameter ceramic with 250 Watts continuous and 1000 Watts pulsed. I miss that job, we had lots of fun! Mikek
On Mon, 14 Dec 2015 07:42:51 -0500, Phil Hobbs Gave us:
The reason for abrading the surfaces is so that the epoxy will have better adhesion. The reason to do one with the other is for a mating/coplanarity improvement.
The main reason, however is because if left smooth, the epoxy can detach very easily. The abrading makes tiny hooks for it to grab onto, which increases the pressure required to create a detachment event.
Lapping solves all those problems at once. Fine valve grinding compound is one approach--it gets the surfaces close but leaves a bit of 'tooth'.
For mounting TECs, it's best to use diamond grit because emery is Al2O3 lik e the TEC end plates, and so doesn't polish them very well (it does work ev entually).
Cheers
Phil Hobbs
Not me. With RF power transistors, I have one advantage over transducers and such. RF transistor flange mounts are bolted to the heat sink. There's no need for the heat sink compound to have any adhesive properties. Therefore, I don't need to solder, bond, or glue anything together. So low-temp solder won't be on my short list. However, that doesn't stop people from trying it. Here's an example: Looks like only the ends of the flange mount are soldered.
Solder also isn't that wonderful a thermal conductor, even with silver added: I'm looking for thermal paste with the highest possible conductivity, that doesn't ruin the budget and is manufacturable by mere mortals. For example, I won't be looking into brazing or soldering the copper transistor base to the aluminum heat sink because the process will probably destroy the transistor junctions and warp the heat sink. It also makes replacing RF power transistors rather difficult.
Dunno.
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
PowerMac G5 quad core, methinks. I think it acted like solder paste, not the usual heat sink goo, but I'm not sure. Apple also did a Delphi liquid cooling in the quad core G5 machine which uses some kind of exotic thermal paste. Indium-gallium sounds expensive.
These look like solders, or maybe solder paste:
W/m*K Indium (pure) 86 Gold leaf 314
I would think that gold leaf would work better. However, gold leaf is a nightmare to handle (and subject to assembly line shrinkage). Look like it's more expensive than gold:
0.05 and 0.1 mm thickness should work, especially if it's easier to handle than gold foil.Good idea. I've never played with indium foil. Is it malleable?
For RF, the best I've done is 300 watts CW through about 120 sq-mm flange mount or: 300 / 120 = 2.5 watts/sq-mm I think I can do better with todays GaN power devices.
One sun is 1000 watts/sq-meter = 0.001 watts/sq-mm. 2500 suns * 0.001 watts/sq-mm = 2.5 watts/sq-mm. Hmmm... that number looks familiar.
That's the same idea behind the carbon flake thermal "paste".
I forgot to mumble something about press fitting the transitor copper flanges into the aluminum. The idea was to eliminate the thermal grease and get a metal-to-metal air tight fit. It would also add the sides of the transistor flange to the surface area. It didn't work because I never could totally eliminate all the air gaps and the xsistor manufacturers (TRW, CTC, etc) never could maintain their case tolerances. Oh well.
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
It often makes sense to use more semiconductors and spread the heat out. That's easy at low frequencies, harder at RF.
Transistors often cost less than heat sinks!
The problem with gold foil is that if you clamp it between two pieces of metal, it disappears--it wants very badly to form an amalgam at the surface.
Very. Softer than Sn63 solder by a lot. It's often used as the O-ring in cryostats.
Yep, but solar cells have to run cooler, because you lose power output really badly otherwise--the series resistance goes up and the open circuit voltage drops.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
I think too thick is a problem when I'm trying to squeeze every last possible watt out of the devices. The problem with Dow 340 is that it continues to flow for weeks after installation. Try two small pieces of steel with some Dow 340 in between. Use some 4-40 set screws to compress the sandwich (same as in most RF power device mounting). Torque to whatever your device specifies, but don't use spring (Belville) washers or lockwashers to maintain tension. After a day or so, re-torque the screws. You'll probably find that they've become loose and require some tightening. This continues until you get to metal on metal, which is where the sandwich should have been in the first place before too much silicon grease was added.
You can do the same experiment with large metal packages (i.e. TO-3) but the problem is different. Instead of the silicon grease oozing out from under the device, the mounting flanges compress, the device bends, and the silicon grease collects in the middle. However, thermal cycling will eventually pump this accumulation out from under the device, resulting in an air gap.
I measured the capacitance between the device and an anodized (insulated) heat sink. I have the numbers here somewhere but as I recall, they were close to zero because the ideal amount of thermal goo was where metal hit metal. I think my accuracy was limited by the accuracy of my torque wrench, but I don't recall the relationship.
Yep, that works and is what I try to do with CPU heat sinks. However, most CPU sockets prevent that from working. So, I settle for smear the goo on thick, wipe as much as possible off with a plastic blade, attach. Using the CPU internal temp measurements and a CPU "exercise" program, this works a few degrees better than anything else I've tried.
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
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.