[Amps] Ldmos heat spreaders

Manfred Mornhinweg manfred at ludens.cl
Fri Aug 19 11:51:38 EDT 2016


Ian, Steve, and all,

> I've only found up to 6mm think would this be okay ?

It's less than optimal, but better than having no spreader.

Here is a useful online calculator for heat spreaders:

http://www.novelconceptsinc.com/calculators-slab-thermal-resistance-constriction.cgi 


It turns out that for a typical modern LDMOSFET and the typically used 
spreader sizes, 6mm thickness is about 30% worse than the more usual 
10mm. You can use the online calculator to find the exact results for 
the actual sizes and conditions you might use.

The other calculators on that web site are also very useful to calculate 
the entire thermal path from an LDMOSFET to the air. And that's an 
exercise every builder of solid state amps should do, to find out what 
combination of FETs, spreaders, heatsinks, joining, and fans he really 
needs. Many ham solid state amps fail simply because the heatsinking 
isn't good enough, and the FETs overheat.

 > Or could I get away
> with 2 layers to get 12 mm with heat transfer paste ?

A single 12mm layer is MUCH better. But two layers soldered together 
would be almost as good as a single 12mm layer. As long as the surfaces 
are very flat, of course.

> Heat transfer is more about the minimum number of interfaces and a 
> proper fit - than the best-best material.  Which means, a machined 
> glossy-dead-flat aluminium surface is going to work better than a poor 
> copper surface.

Steve, I partially disagree here.

I do agree on the interfaces being extremely important, and the most 
important one is that between the FET and the spreader. Soldering the 
FET to the spreader is essential when the application has high power 
dissipation, like 600 watts from each device, which is usual in class AB 
linear service. Thermal paste, even the best one, just isn't good enough.

But once the thermal barrier of grease under the FET has been 
eliminated, the next biggest thermal resistance comes from the material 
immediately under the FET, and up to a few cm away from it. It is VERY 
important that the material used there has the absolutely best thermal 
conductivity. And that's why copper heat spreaders are universally used 
in such amps. If you do away with the copper spreader, and try to 
replace it by an aluminium heatsink that has an extra thick base plate, 
it just won't be as good.

Since it's the material very close to the FET that matters most, the 
size of the spreader isn't terribly important. With a heatsink that has 
a thick base plate, the difference between a copper spreader covering 
the entire heat sink, and another that's only one third as large, is 
rather small. But the difference between that small spreader, and no 
spreader at all, is big!

It all interacts with the quality of the interfaces, of course, you are 
very right there. A large copper spreader is useless if it's so warped 
that a thick layer of grease is needed.

> You CAN get the MOSFET soldered to an aluminium heatsink - they use an 
> ultrasonic soldering tool and the results are superlative.

But it makes sense only when weight saving is more important than the 
thermal resistance! That's the case in some non-linear aeronautical and 
space applications, but not in a class AB linear amplifier for 
terrestrial use!

Try those calculators on the novelconceptsinc website. Some playing with 
them is very enlightening.

> There is also the option of water-cooling an alloy block.  This is a 
> little more fiddly in the cooling department, but quite a lot simpler in 
> the mounting department.

I love that method, but the block should be copper, not any alloy! I 
know, pure copper is an unpleasant material to work with on a milling 
machine! But it provides 3 times better thermal conductivity than a 
typical aluminium alloy, and that decides the matter. Aluminium is fine 
for more distributed heat sources. But when you need to remove 600 watts 
of heat from a 30x10mm surface (and that's being optimistic), while 
keeping that surface below 100°C, copper is the only option.

Manfred

========================
Visit my hobby homepage!
http://ludens.cl
========================


More information about the Amps mailing list