Potting a switch mode LED driver PCB to improve thermal withstand?

[Flyback]

Hello,
We have a 510W LED driver PCB (three 170W channels) and we cannot mount it on a heatsink a there’s no room in the enclosure for a heatsink.
There’s also no room for fans in the enclosure.
Also, we cannot have vents in the enclosure sides.
In order to try and make the thermal situation for this SMT PCB as best as possible, we believe that we should pot it?…because then at least the hot components will better interface to the air inside the enclosure.
We believe the (60mm by 70mm) PCB can be coated in rubbery, low thermal impedance potting compound…..rather like the way a lollipop is kind of grown over a lollipop stick.

What do you think of this idea?
(obviously we will do everything we can with adding as much thermal spreading copper to the PCB also)

 

[picbits]

What is the duty cycle for the LED ? What is the mass and thermal conductivity of the potting compound ?

If you are generating a constant 510 watts of heat, you're going to fry stuff very rapidly !

 

[ronsimpson]

My guess is that 510 watts goes into the LED and 50 watts is lost in the PWM.
Even if the LED is external to this enclosure, 50 watts is a large amount to get rid of.
If 500 watts is all inside the enclosure then.............ouch.
We don't know the size of the enclosure, what it is made of..........anything. It sounds like there is more parts inside.

 

[Flyback]

The total losses in this three channel LED driver when driving the maximum power led load is as follows..

AON6284 FET conduction loss = 0.23mW.
AON6284 FET switching loss = 0.19W (Switching frequency = 200KHz)
SBRT25U80 diode conduction loss = 0.105W
MSS1210-333 Coilcraft inductor loss (total) = 0.438W (by coilcraft web loss calculator)
R(sense), hi-side, 1812 = 0.310W
R(sense), lo-side, 2//1206’s = 0.172W
…Obviously the above get multiplied by three for the total loss’s to the components of this SMPS….therefore , the total component loss in this SMPS is just 4.335W
…That’s efficiency of around 99%…sounds unrealistic, but that’s the figures.

 

[picbits]

Oops - I read that as you were going to mount the LEDs in the same enclosure - my bad.

4W could fairly easily be dissipated by a casing of a controller. Is the casing of the controller made of metal ? Possibly potting the board to the casing (using this as a heatsink) could be an option ? Even cutting small grooves in the casing can help with the heat dissipation by increasing the surface area

 

[Flyback]

that's a good idea to put the PCB into a metal can and then pot it in there so it conducts heat to the metal can , then on from there...we will see if there is room

 

[picbits]

If the metal can has a stud on it for mounting then if there are any excess heat issues, this can then be bolted onto another heat dissipating object to further help.

 

[schmitt trigger]

Do you have a particular compound in mind? If so can you share its name/description?

We use to pot brick converters with thermal compound for the exact same reason your are mentioning. I am not at liberty to disclose what we used, but at least I can check the properties of your proposed compound and give an opinion.

 

[Flyback]

well, I will look some up, but I believe even plain old rubber would be a better thing than just the pcb in air?

 

[schmitt trigger]

No, it depends on its thermal conductivity, and rubber itself actually has poor conductivity.

The compounds that I've seen are polymers mixed heavily with alumina or at least silica fillers.

 

[Flyback]

yes but rubber is more conductive than air?

 

[picbits]

Depending on the application - if you have ventilation holes in the unit then convection can be more beneficial than conduction through rubber.

 

[schmitt trigger]

Eventually, all the assumptions and theories will have to be tested, and the tests determine whether to validate or discard that assumption or theory.

A couple of years after the fall of the USSR, my company was able to hire a pair of Russian scientists with PhDs whose experience was thermal management. They would make these incredibly complex theoretical analysis, using some mathematical operators I had never seen before. Yet, they always validated them with actual testing.
The problem with heat transfer is that it is a three-dimensional field, and has to be analyzed accordingly.

One of them had actually worked for spacecraft electronics. He had a saying: here in earth it is simple, one can always rely on dumping the heat to the air. In space, one only has radiation...but sometimes one has positive radiation gain, for instance surfaces pointing to the sun. Very interesting.

 

[kubeek]

Even just 170W is a huge amount of heat that needs to be dissipated. The case the LEDs will be inside needs to be the heatsink, there is simply no way around that (mabe except for water cooling).
Look at how big are heatsinks in for example audio amplifiers, where they usually have a fan as well. Now imagine that the transistors in the the amp can withstand 175°C junction tempreature, and your LEDs will likely have the limit at 100 to 125°C.

You didn´t mention how big the led pcbs are or how big the enclosure should be, but I think it will be a huge chunk of extruded aluminium to get rid of that much heat. If you can get a shape that could easily bond with a heat transfer block inside, then you could use just a block of aluminium to transfer the heat between the case and the board.

But it seems that you are way ahead of what needs to be done. Have you acutally managed to run a signle LED board, just mounted to some adequate heatsink, without the LEDs overheating? I.e. if you manage to keep the cool side of the PCB at 25°C do the LEDs stay within limits? First try that and then go thinking about some potting and cases and whatnot.

 

[Flyback]

a single 165w led mcpcb is about 3cm by 3cm by 1mm thick...it is stuck to the watercooled block like that that you mentioned