Transistor heatsink insulator alternatives to Mica or Sil Pad

[Mephitus]

I am working on a linear supply that uses some classic 2N3055's. As my component budget is slim enough for me to literally count pennies, I was trying to think of potential insulator alternatives to having to buy mica or sil pad insulators. Such as using silicone parchment paper and thermal compound. Searching the forums and using google is not returning anything that would allow me to electrically isolate the TO-3 package while still allowing efficient thermal transference.

Have any of you experimented with any alternatives or have some recommendations I can use to experiment with? I do have a few spare 2N3055's I could sacrifice  in experiments if anyone has some ideas.

[SeanB]

How about just isolating the heatsink inside the case and use a PC CPU cooler fan ( and the heatsink from the CPU itself) as the heatsink with the 2N3055 devices directly mounted. That way you only need some plastic isolators, which can be nylon or other non conductive plastic strips, and which will handle 100W of power dissipation.

[edpalmer42]

It would be difficult to mount a 2N3055 in a TO-3 case to a CPU heatsink.  But the basic idea is good.  Mount the 2N3055 on a single-sided heatsink without any isolation and then bolt the heatsink flat against the circuit board.  If there are any traces under the heatsink, either switch to a vertical heatsink or use a sheet of heavy plastic between the heatsink and the board.  Lots of ways to do it.

[Seekonk]

I have some anodized aluminum pads that are supposed to electrically isolated.  Never did trust them.

[Mephitus]

How about just isolating the heatsink inside the case and use a PC CPU cooler fan ( and the heatsink from the CPU itself) as the heatsink with the 2N3055 devices directly mounted. That way you only need some plastic isolators, which can be nylon or other non conductive plastic strips, and which will handle 100W of power dissipation.

I may just end up doing something like this. Using an isolated heatsink within the case. I have a (smallish) blower fan that I was going to use that will move MORE than enough air to cool it effectively.

[SeanB]

TO3 fits CPU heatsinks nicely, you just need to drill the 2 mounting holes and the 2 lead holes from the bottom, then from the top butcher a larger hole,10mm, down through the fins ( or on a solid round P4 one just drill back till there is 3mm left from the bottom) so you can get the 2 leads though already soldered to wires. But better to slap 2 TO247 packages on there, they do fit.

[edpalmer42]

Yes, a couple of TO247 or TO3P packages would fit very nicely on the flat surface of many CPU coolers.  Lots of cooling capacity there!

[edavid]

Sell one of your spare 2N3055s and buy the Silpads @ 8c each

http://www.taydaelectronics.com/hardware/heatsink/transistor-rubber-washer-to-3.html

[Richard Head]

To answer your original question, Polyimide (Trade name "Kapton") is excellent and much cheaper and better than Silpads (which are expensive). I ones I've used have heatsink compound pre-applied and it's very easy to peel them off the card and place them on the heatsink.

[davelectronic]

I have used a few cpu heatsinks with exellent results. Choose a heatsink that has its slot cut out through the fins, your pins go through there. Drill 2 x blind holes for mounting holes, use self tappers for these holes. Really makes a super cooling solution up to about 100 watts. The older amd heatsinks often have a channel already cut through the heatsink fins.
I used 3.5mm self tappers, and drilled 3.2mm holes for mounting holes. Be sure to use heat shrink tube on your transistor pins, and take your base and emitter wires out the top of he channel.

[edpalmer42]

In another thread I talked about using this:

http://www.shoppingsquare.com.au/p_400895_NonStick_Oven_Guard_For_Clean_Cooking_

inside an oven for a quartz oscillator.  I got mine at a chain store that sells cheap stuff.  It cost me about $US4 for a 13"x18" sheet.  It's very similar to a Silpad except it doesn't have as much silicone.  If you wanted to experiment, you could look for this or something like it.

Ed

[T3sl4co1l]

Like your avatar, IT STINKS!

I mean, 2N3055 is bad enough... but also, by limiting yourself to an archaic package, you're stuck on heatsinking options.   Punched mica or Sil-Pad is basically all.

I suppose if you're that desperate, you could... solder on some water-cooling tubes from spare junk, or something.  (Solder it on the bottom side, in the center of the device. That's where the die is attached.)

Note that a mica insulator is nearly useless unless you also have thermal grease.  Sil-Pads work on their own (and shouldn't be used with grease, in fact).  You have to have something compliant to fill the gap, otherwise you have very few points of metal-on-metal contact with a teensy, tiny bit of air inbetween that might as well be vacuum.

And if hardware is such a burden, it might be worth considering other options?  The linear supply can be controlled from the AC side using a TRIAC (and some precautions for the inductive load of the transformer, and some derating at low output voltages because it won't be able to supply full output).  It's a terribly hacky substitute for a variac, but if you've got to make do..

If you have some inductors* handy, you can do better, either with SCRs and a big inductor (doing the same sort of thing, phase control), or with a switching power supply (and a smaller, ferrite or powdered iron choke).

"Hybrid" switchers (i.e., using a big iron transformer to get low voltage, then controlling the last step-down switching stage to get good regulation and efficiency) are easily enough made, though you really ought to have an oscilloscope to make sure it's behaving correctly.

You still need big transistors (and faster ones than the 2N3055), but the heatsink requirement goes down, possibly even away.

*Inductors, meaning, a magnetic component that can store energy.  Transformers have a lot of inductance, but very little energy storage.  You need an air gap in the core to store energy.  If you don't mind hacking the core apart, you can reassemble it with a gap, and then you'll have an inductor that can be used for filtering and switching supplies.

Tim

[Mephitus]

Well, good news on that front. I was able to get ahold of some free samples of LT3081, LT3082, LT3086, and LT3092.
So I can make a MUCH better bench supply with those. A proper 0v-24v 3a supply. I will make a post once I am able to finalize the design.

[Gyro]

Like your avatar, IT STINKS!

I mean, 2N3055 is bad enough... but also, by limiting yourself to an archaic package, you're stuck on heatsinking options.   Punched mica or Sil-Pad is basically all.

I agree on the limited heatsink options but that "archaic package" is still a damned good one in terms of thermal transfer. I wish a few more recent devices were available in TO-3 so you could actually use more of their claimed datasheet capabilities. I still think single screw mounting sucks - that's why people start resorting to clamping bars to hold devices down.

[pmbrunelle]

TO-3 is good, TO-247 is modern-day rubbish with uneven clamping.

[SeanB]

TO247 is good if you simply use longer screws to hold a TO3 package pushing down on the plastic so it has even clamping.

[T3sl4co1l]

None of them are very good.

TO-3 (TO-204, etc.) is usually a steel package, and not usually very flat.  Although the all-metal design does typically allow slightly higher junction temperatures (think I've seen 225C before).

TO-247 is much easier to use: PCB mounted, single screw to heatsink, insulated hole (no freaking shoulder washers!).  TO-3P is an, I think Japanese creation, which is very similar.

TO-264 is much larger, but the same basic thing.

But none of them are exceptional, because you have the same fundamental problem: getting over 100W of heat out through a small footprint, very likely through an insulator.  All insulators suck.  Either the conductivity is poor (mica, Kapton, Sil-Pads, etc.), or the material is very thick (AlN, BeO and such), or the flatness sucks (TO-3 in particular, but extruded heatsinks are an offender too).  And all the good ones need grease, which is still more assembly work.

It really doesn't make much difference in packaging, because the insulators are all awful.  Make your choice on which is easier to connect to, and to assemble.  TO-247 and Sil-Pad is hard to beat.  TO-3 can't be put into a PCB, not in a serviceable manner, because it has to pierce the heatsink.

Really, the underlying truth, more illuminating than making tedious statements about crappy packages, is -- dissipating power with semiconductors is expensive and sucks.

In power dissipation applications (active load, battery conditioner?), you can offload a lot of dissipation into a stack of resistors, saving considerably on heatsinking requirements and power transistors and mechanical assembly.  This is a lot harder to implement for amplifiers, so you're kind of stuck in that case (think of it as a "cost of doing business" to get a good, low noise, high speed amplifier).  That's where the magic of class D amplifiers comes in; if you don't need the speed or the noise requirement (which is largely the case in slow, audio applications), they solve all of this.

BTW, TO-247 doesn't exhibit the flex that TO-220 is notorious for.  It is still uneven, but not as bad, and smooth, flat, greased mating surfaces do fine.  You can also secure them with clips (required for the unholey MAX247 variant!), which can be even easier for assembly (there are clip-on types that don't need a screw at all).

For more than 100W per package (or maybe 150W for TO-264), consider using multiple devices in parallel (mind the current sharing) or larger devices (ISOTOP / SOT227 or various industrial modules).  TO-220 and 247 are so damn cheap, they're hard to compete with, as long as you can connect them in parallel effectively; just keep piling them up.

Tim

[Melgatronic]

I use Kapton Tape Polyamide  high temperature, for the MosFet with upper 200 V voltage and no problem.