Heatsink Isolation

[09d08]

I have two rather large bare aluminum heatsinks and want to use them to cool four IGBTs . Since the IGBTs are supposed to handle rectified mains and dissipate a high amount of power I need a way of attaching them to the heatsinks which provides electrical isolation as well as good thermal performance.

I thought of silicone pads so I attached a TO-247 BJT to one of the heatsinks using one just to test it but the side of the transistor with the markings reached around 50 degrees while dissipating around 12 watts. This didn't happen when I repeated everything with thermal paste instead of the pad. The IGBTs would probably get damaged after a while if I were to use pads so I need suggestions. I've thought of partially anodizing the heatsinks but I would like to avoid that if possible.

Thanks in advance

[cur8xgo]

I have two rather large bare aluminum heatsinks and want to use them to cool four IGBTs . Since the IGBTs are supposed to handle rectified mains and dissipate a high amount of power I need a way of attaching them to the heatsinks which provides electrical isolation as well as good thermal performance.

I thought of silicone pads so I attached a TO-247 BJT to one of the heatsinks using one just to test it but the side of the transistor with the markings reached around 50 degrees while dissipating around 12 watts. This didn't happen when I repeated everything with thermal paste instead of the pad. The IGBTs would probably get damaged after a while if I were to use pads so I need suggestions. I've thought of partially anodizing the heatsinks but I would like to avoid that if possible.

Thanks in advance

just in case have you run the numbers on isolating thermal solutions to make sure that makes sense? Off hand I thought the performance hit of isolated thermal pads was not bad at all

[wraper]

Silicone pads are very different, thermal conductivity may differ by an order of magnitude. In any case, mica (+ thermal paste) or ceramic insulator should provide better performance.

[Ian.M]

Worst case: if you are really pushing the limits of what power the semiconductor package can handle,  you may need a thick copper heat spreader directly under the semiconductor package to increase the area of mica (or other insulator) the heat is transferred through, to decrease its thermal resistance.

[digsys]

I ran extensive tests ages ago for similar reasons. As others say, there is a HUGE difference between the types etc - These notes may be of help -

[cur8xgo]

for instance berquist sil-pad 600 series 0.35C/W $0.23 in stock at digikey...digikey has zillions of thermal pads greases etc... this is just one random one

https://www.digikey.com/product-detail/en/bergquist/SP600-54/BER103-ND/201890

[09d08]

just in case have you run the numbers on isolating thermal solutions to make sure that makes sense? Off hand I thought the performance hit of isolated thermal pads was not bad at all

My testing methodology was not that scientific, I just mounted the same transistor first with paste then with a sil pad and measured the temperature of the top side. The measured temperature didn't rise above ambient with the paste but reached 50 degrees with the pads. It probably has to do with the quality of my sil pads but there still seems to be a performance hit.


Thanks for all the suggestions, I have a couple copper plates lying around so I'll try to find a large mica sheet to form a heat spreader. Finding better sil pads also makes sense but ordering from digikey is not an option for me and local shops don't have brand-name sil pads. Thanks again.

[Brumby]

I've thought of partially anodizing the heatsinks but I would like to avoid that if possible.

Don't rely on the anodizing to provide reliable isolation.  There has been at least one report on the EEVblog where this was done and after a period of time, it failed.

[Siwastaja]

Worst silpads are about 1W/mK. The best one are around 20W/mK. A huge difference. Yet, the best ones are still 10x worse than aluminum or 20x worse than copper. Making it thinner helps, but for safe mains operation, you can't go arbitrarily thin - the thickness protects against small burrs and surface damage from piercing through insulation. So I wouldn't go below 0.5mm thickness.

Invest in high-thermal-conductivity (> 10W/mK) silpads, do the math for your contact area. If this still isn't good, one way to solve it is to solder the device to a larger-area copper heat spreader, then mount this spreader through the silpad.

[BravoV]

Apart from thermal conductivity, dielectric breakdown voltage also important especially at high voltage, say compared to lower voltage stuffs like in CPU vs mains related circuit.

Usually from 1st tier manufacturers, they came various thickness, with the same material, its obvious that thicker pad has lower thermal conductivity than thinner one, but has higher dielectric breakdown.

[09d08]

I've just cut small pieces of copper to use as heat spreaders and the front side of the transistor rises only 6 degrees above ambient while dissipating around 20W (the highest power the IGBTs will be dissipating while in actual use). The thickness of my silicon pads also seem adequate for 310V. Thanks for all the help.

[max_torque]

My testing methodology was not that scientific.....

.... The measured temperature didn't rise above ambient with the paste

I think you're test was probably flawed in some way!  (Or you have just broken the laws of physics for heat transfer.......)   

[09d08]

It

My testing methodology was not that scientific.....

.... The measured temperature didn't rise above ambient with the paste

I think you're test was probably flawed in some way!  (Or you have just broken the laws of physics for heat transfer.......)   

You can only get that much accuracy out of a $4 chinese infrared thermometer.

[Siwastaja]

You can only get that much accuracy out of a $4 chinese infrared thermometer.

You are aware of the surface emissivity, right? For example, you can't measure a shiny aluminum block with a infrared thermometer, cheap or expensive, Chinese or western. A piece of masking tape, for example, allows the IR readout of such material.

[Jeroen3]

You can use aluminum oxide isolators.
AAVID THERMALLOY sells them under 4180G for example. But thickness and shape differ.
It isolates and transfers a lot better than mica or sil, at the price of fragility and cost of both bom and assembly.

https://www.shopaavid.com/Product/4180G

[Siwastaja]

You can use aluminum oxide isolators.
AAVID THERMALLOY sells them under 4180G for example. But thickness and shape differ.
It isolates and transfers a lot better than mica or sil, at the price of fragility and cost of both bom and assembly.

https://www.shopaavid.com/Product/4180G

15W/mK, really, for what? You can get silpads of similar thermal conductivity and completely bypass the issue of fragility and assembly steps of applying thermal paste.

If a modern high-tech 10..15W/mK silpad isn't going to do the job, then I'd be seriously looking at spreading the heat so that it can be solved with said silpad. If you can accept down to about 5W/mK, you have a massive, proven market to choose from. Electrically isolating, safety-approved, robust thermal coupling just seems to be limited at around 10..15W/mK, if you need higher, it'll be very expensive and difficult, or risky.

Beware of brittle and/or very thin materials, which work fine in lab conditions and may have all required approvals on paper, but break down in production due to minuscule surface imperfections that either shatter the material (ceramics), or puncture through it (silpads or tapes too thin).

Increasing the surface area is often the easiest way.

[rsjsouza]

You can use aluminum oxide isolators.
AAVID THERMALLOY sells them under 4180G for example. But thickness and shape differ.
It isolates and transfers a lot better than mica or sil, at the price of fragility and cost of both bom and assembly.

https://www.shopaavid.com/Product/4180G

15W/mK, really, for what? You can get silpads of similar thermal conductivity and completely bypass the issue of fragility and assembly steps of applying thermal paste.

Do you have a reference for that? Aavid (pages 101 and 103) and Heinkel (page 7) do not go far with SilPads. Aavid's 15W/mk looks quite good in comparison.

[Siwastaja]

15W/mK, really, for what? You can get silpads of similar thermal conductivity and completely bypass the issue of fragility and assembly steps of applying thermal paste.

Do you have a reference for that? Aavid (pages 101 and 103) and Heinkel (page 7) do not go far with SilPads. Aavid's 15W/mk looks quite good in comparison.

https://www.digikey.com/products/en/fans-thermal-management/thermal-pads-sheets/218?k=&pkeyword=&sv=0&pv70=377&pv70=311&sf=1&FV=ffe000da%2Ca940029%2Ca940033%2Ca940034%2Ca94003a%2Ca94003b%2Ca94003c%2Ca94003d%2Ca94003e&quantity=&ColumnSort=0&page=1&pageSize=25

[wraper]

15W/mK, really, for what? You can get silpads of similar thermal conductivity and completely bypass the issue of fragility and assembly steps of applying thermal paste.

Do you have a reference for that? Aavid (pages 101 and 103) and Heinkel (page 7) do not go far with SilPads. Aavid's 15W/mk looks quite good in comparison.

https://www.digikey.com/products/en/fans-thermal-management/thermal-pads-sheets/218?k=&pkeyword=&sv=0&pv70=377&pv70=311&sf=1&FV=ffe000da%2Ca940029%2Ca940033%2Ca940034%2Ca94003a%2Ca94003b%2Ca94003c%2Ca94003d%2Ca94003e&quantity=&ColumnSort=0&page=1&pageSize=25

Those are not suitable for this task. They are for low voltage applications.  Used to fill a gap (like between RAM and heatsink on GPU) and are unable to bear any mechanical stress. So you cannot use them as insulator.

[rsjsouza]

15W/mK, really, for what? You can get silpads of similar thermal conductivity and completely bypass the issue of fragility and assembly steps of applying thermal paste.

Do you have a reference for that? Aavid (pages 101 and 103) and Heinkel (page 7) do not go far with SilPads. Aavid's 15W/mk looks quite good in comparison.

https://www.digikey.com/products/en/fans-thermal-management/thermal-pads-sheets/218?k=&pkeyword=&sv=0&pv70=377&pv70=311&sf=1&FV=ffe000da%2Ca940029%2Ca940033%2Ca940034%2Ca94003a%2Ca94003b%2Ca94003c%2Ca94003d%2Ca94003e&quantity=&ColumnSort=0&page=1&pageSize=25

Thanks. Quite an expensive SilPad!