Insulating Heat Sinks From Their Transistors-Why?

Thread Starter

tempest411

Joined Dec 29, 2012
6
Hello,

I was just searching around a bit trying to find opinions on whether the more expensive heat sink compound is really any better than the cheap stuff. In a tiny store like Radio Shack they have three different types to choose from, so that must say something, but what?

However...I can't help asking myself why in the world they design this stuff so you can bolt up the heat sink in a nice solid metal to metal fashion. If the name of the game is heat removal, then nothing beats a metal to metal junction. And it seems furthermore that the compound you're supposed to do use also has heat insulating properties as well as electrical. I thought the grease was supposed to a way of 'countering' the fact that there wasn't a good a decent connection between the transistor and the heat sink in the first place. If the plate on my four TDA7294 transistors are all at the same potential, can I just mount the heat sink directly and insure that the heat sink itself doesn't touch any other part of the enclosure? I think heat transfer would go up by nearly 80% this way!!

Thank you for your opinions,

Rick
 

mcgyvr

Joined Oct 15, 2009
5,394
First there are multiple types of thermal compounds/pads,etc..
Thermally conductive/electrically insulating
Thermally conductive/electrically conductive
Thermally conductive/adhesive
,etc..

despite what you think a metal/metal surface has "inclusions/bumps/holes" on a microscopic level. These holes/bumps create air pockets and air sucks at conducting heat. So the thermal compounds are applied to fill in those holes.
Most people apply it WAY to thick which hurts thermal performance. Usually it should be applied as thin as you can possibly get it.. But we are human.
The thermal compounds typically specify an "application thickness" that is more than likely never met by hand application.

Hence thermal pads from bergquist,etc.. they are like a "sticker" that you simply apply and can rely on the stated thermal resistance more than you can with a hand applied grease application from assembly to assembly.
 

takao21203

Joined Apr 28, 2012
3,702
Normally you would push down the component and move it a little if you have excess thermal grease.

Not it is not insulating of course.

If you have solid black aluminium you can actually attach it with silicon putty and don't need insulation- no holes are drilled, no screws.

Otherwise or for high voltage, you must use silicon pads.

Another way is to use VGA coolers and mount the parts upside down like DPAKs.

If you have a component with only moderate cooling requirements- add it there normally, plastic mounted upside. Thermal performance is reduced, but still some cooling.
 

takao21203

Joined Apr 28, 2012
3,702
NEVER..NEVER rely upon black anodizing for insulation.
No, normally not. I was looking at these Pentium II sinks, and they really have a thick solid anodization.

If I mount parts with silicon putty + no pressure, what can cause puncture of the anodization?

Normally I'd use a secondary cooler, not to sacrifice thermal conductivity with a silicone pad. These really need good pressure, or are bad for transfer.

Many cooler grids only have a thin anodization, it has a shining appearance.

These Pentium cooler grids have a very thick anodization.

I'd not really try it for a circuit except low voltage if I am forced otherwise to use 2 sinks.

There is a silicone pad anyway. No way to drill holes, only method is to make a PCB + use screws to force pressure on the assembly. This is how I came up with the idea to try "No more Nails" for mounting.

It is often enough for secondary parts to mount them 180 degree with the plastic side (I use VGA coolers mainly).

Never relied on the anodization. I do not really recommend it too, actually.
 

Attachments

Last edited:

mcgyvr

Joined Oct 15, 2009
5,394
^^ first that is a black enamel paint not anodizing.
second that black surface is FULL of holes/bumps,etc.. and as such you MUST use some form of thermal compound.
third there is nothing wrong with "energizing" a heat sink as long as you intend it to be energized.
And in the real world..not your playing around/DIY world.. what works for one might not work for the next part. There are inconsistencies/voids in plating all the time. So one might not short out but the next one can.
 

mcgyvr

Joined Oct 15, 2009
5,394
This is how I came up with the idea to try "No more Nails" for mounting.
seriously.. come on. You are such a hack. :p
There are plenty of appropriate thermal adhesives out there.. Liquid nails/no more nails is for carpenters.

I've seen your silicone caulk to hold multichip LED's in place too.. hack
 

crutschow

Joined Mar 14, 2008
34,201
...................................
If the plate on my four TDA7294 transistors are all at the same potential, can I just mount the heat sink directly and insure that the heat sink itself doesn't touch any other part of the enclosure? I think heat transfer would go up by nearly 80% this way!!
The "plate" tab of the TDA7294 is connected to -Vs so the heat sink will also be at that potential. So as long as the heat sink is isolated that will work. But use thermal grease as others have stated.
 

takao21203

Joined Apr 28, 2012
3,702
seriously.. come on. You are such a hack. :p
There are plenty of appropriate thermal adhesives out there.. Liquid nails/no more nails is for carpenters.

I've seen your silicone caulk to hold multichip LED's in place too.. hack
It is actually heat resistant. It is just having the right consistence. The LEDs are hold in place securely. With pliers you can pull it off if need be.

I had the idea some months ago to turn a DPAK LM2576 upside down, and attach a VGA cooler with epoxy. Isn't so easy to get the consistency right, and messy as well.

Then I tried the No more Nails.

Using it with a bridge rectifier, which needs one of these Pentium heatsinks. 2 Months, no issues.

Found a cheaper alternatively some weeks ago, only 1/5 the price. Using it today for the first time with $100 worth of LEDs.

You can not drill holes into these VGA coolers properly, must rely on special thermal adhesive glues (expensive).

It works for me! Silicone putty as such is a hitech material.

Enamel painted Pentium Heatsinks?? Not to sure if there are any.
 

mcgyvr

Joined Oct 15, 2009
5,394
It is actually heat resistant. It is just having the right consistence. The LEDs are hold in place securely. With pliers you can pull it off if need be.

I had the idea some months ago to turn a DPAK LM2576 upside down, and attach a VGA cooler with epoxy. Isn't so easy to get the consistency right, and messy as well.

Then I tried the No more Nails.

Using it with a bridge rectifier, which needs one of these Pentium heatsinks. 2 Months, no issues.

Found a cheaper alternatively some weeks ago, only 1/5 the price. Using it today for the first time with $100 worth of LEDs.

You can not drill holes into these VGA coolers properly, must rely on special thermal adhesive glues (expensive).

It works for me! Silicone putty as such is a hitech material.

Enamel painted Pentium Heatsinks?? Not to sure if there are any.
yes its a painted heatsink.. anodizing doesn't chip..look at the edges. its paint.

and proper thermal adhesive is only a couple bucks you cheap-o. 2 part artic silver is $5 and will easily do 100+ LED's and won't look absolutely ghetto.. and is thermal compound at the same time and low-outgassing..
 

takao21203

Joined Apr 28, 2012
3,702
yes its a painted heatsink.. anodizing doesn't chip..look at the edges. its paint.

and proper thermal adhesive is only a couple bucks you cheap-o. 2 part artic silver is $5 and will easily do 100+ LED's and won't look absolutely ghetto.. and is thermal compound at the same time and low-outgassing..
So I should have asked on a forum about that.

Don't think it is good for 100 LEDs.

Not that it would make a difference when each LED costs 30 dollar.

Oh and the LEDs are not used to show off LEDs. They just have to produce light.
 

takao21203

Joined Apr 28, 2012
3,702
Hello,

I was just searching around a bit trying to find opinions on whether the more expensive heat sink compound is really any better than the cheap stuff. In a tiny store like Radio Shack they have three different types to choose from, so that must say something, but what?

However...I can't help asking myself why in the world they design this stuff so you can bolt up the heat sink in a nice solid metal to metal fashion. If the name of the game is heat removal, then nothing beats a metal to metal junction. And it seems furthermore that the compound you're supposed to do use also has heat insulating properties as well as electrical. I thought the grease was supposed to a way of 'countering' the fact that there wasn't a good a decent connection between the transistor and the heat sink in the first place. If the plate on my four TDA7294 transistors are all at the same potential, can I just mount the heat sink directly and insure that the heat sink itself doesn't touch any other part of the enclosure? I think heat transfer would go up by nearly 80% this way!!

Thank you for your opinions,

Rick
Yes you can if they are supposed to have electrical contact from the backplane.

You get however a "hot chassis".

Be careful here with high voltage.

You also should not rely on that electrical contact unless you know what is going on there.
 

Thread Starter

tempest411

Joined Dec 29, 2012
6
Thank you very much for the replies. I think what I'm going to do is a bit of testing. First, I'll mount the thing with insulators and grease and operate it taking measurements with an IR temp gun and also check with a meter and see that all the transistor mounting plate abutting the heat sink are at the same potential so that if I did go the route of metal to metal from chip to heat sink that I wouldn't be inadvertently shorting something out. Yes I know I would have to be mindful of contact between the heat sink and the chassis. But once the enclosure is together a short would be nearly impossible. Speaking of which, any suggestions for a reasonable priced enclosure? There doesn't seem to a be a large selection available. I don't need a pristine Cherry cabinet with milled aluminum plates, just a simple metal (preferably aluminum) 'box'.

Thanks again everyone,

Rick
 
Last edited:

mcgyvr

Joined Oct 15, 2009
5,394
An IR gun is NOT sufficient. Thermocouples or other contact/accurate measurement devices MUST be used. The need for an emissivity value with an IR gun is the problem. No to mention the spot ratio.
NONE of the standards organizations allow them to be used for any testing where accuracy is required.

Also if you want any good suggestions on enclosures you are going to have to provide a little more info..physical size/doors/openings/mounting locations/venting,etc..
 

THE_RB

Joined Feb 11, 2008
5,438
An IR gun is NOT sufficient. Thermocouples or other contact/accurate measurement devices MUST be used. The need for an emissivity value with an IR gun is the problem. No to mention the spot ratio.
NONE of the standards organizations allow them to be used for any testing where accuracy is required.
...
IR gun is fine for most hobby stuff if used properly. I put white paper stickers on anything metal before measuring it, that covers the emissivity issue and gives a good reading, within the gun's accuracy of 1'C. (I also have contact measuing thermometers).

Re the spot ratio that doesn't matter provided the gun is very close (an inch or less) to the target and the target is >0.5inch in size. Again, tested compared to contact thermometer.

One big issue is that the IR gun will see inside certain parts so you you may be reading the hottest part of the internals, like the semi die or the resistive heater etc. Things like white 5W ceramic resistors are transparent to IR so the gun reads the temperature of the wire coil inside, not the ceramic case temp.

That effect can be a good thing with semis, a lot of flat pack semis you can read higher than the case temp showing a value closer to the die temp than you could get with a contact thermometer.
 

takao21203

Joined Apr 28, 2012
3,702
I use a cheap LCD thermometer for this purpose with 10K NTC.

1.5V button cell, runs now from a depleted 1.5V AA battery.

The tiny sensor broke off, replaced with small NTC.

It actually freaks and the firmware hangs up from EMI that comes from SMPS. Solution: to inline a 330uH RF coil. The only way was to shorten the battery to reset it.

Or use a PIC + serial display :)

I made smaller one's with 3-digit only, they are cheaper, so the price is only $3.98

Works days from a 3V button cell. Hi-tech.
 
I built a Maplin 225WRMS amp in the 1980's and mounted the output transistors on two heatsinks and in a wooden chassis. Using two separate heatsink's meant there was no electrical connection between npn and pnp transistors.
 
Top