[Amps] LDMOS HEAT SPREADERS

[Manfred Mornhinweg]

John,

> I wonder if a high power transistor could mount on a chill block sold 
> for puck type very high power devices. For instance,
> 
> http://catalog.chtechnology.com/viewitems/extruded-bonded-fin-heat-sinks/liquid-chill-block?&plpver=10&forward=1

Let's take the first of those as an example. It's 2*2 inches, with a 
0.008°C/W rating (but they don't say at what water flow rate and 
pressure). If you solder a typical high power LDMOSFET to it, you will 
be using roughly one eight of the mounting surface, which will make the 
thermal resistance increase nearly, but not fully, 8 times. It might 
leave you with 0.05 to 0.06°C/W of thermal resistance at the unspecified 
flow rate. And that's still jolly good! It means that when dissipating 
600W, which is a typical ballpark value for one of these LDMOSFET 
devices working in ham service, the transistor's mounting surface will 
heat roughly 30-35°C above the water temperature. No practical 
air-cooled heatsink and spreader can be that good!

So this confirms again that water cooling is a good way to go, when 
using devices that give off a lot of heat from a small surface.

> Of course, getting the RF into and out off the device with this cooler 
> in the way is tricky.

I don't think this is a really big obstacle. I would solder the FET and 
two copper strips to the cooler, in one operation. Being brazed, it 
should have no trouble surviving soldering temperature. The two copper 
strips would run under the gate and drain terminals, and be full width. 
And then I would solder the FET and those straps to my printed circuit 
board.

Given the numbers above, it might actually be possible to use a single 
BLF188XR at full legal limit in linear service! That's not a possibility 
with air cooling, at least not with a good reliability.

Water cooling does have its complexities, but I think it's worth using 
it, when using these high power devices.

Is there a place where one can buy single quantities of such coolers, 
off the shelf? I mean, without having to contact a sales department, and 
trying to convince them to sell a single one to an experimenter.

Steve,

>> With water cooling, why not sit the transistor over a pocket and
>> run the coolant directly against the flange? That's a method I saw
>> in production test at the Philips (as it was) factory.

The thermal resistance of a metal-water interface depends strongly on 
the flow velocity, and thus on turbulence. If you want to remove a 
kilowatt or so of heat from a surface as small as that of an LDMOSFET's 
flange, you would probably need a very high water velocity. That means a 
high pressure pump, noise, and risk of cavitation, erosion, and so on.

I haven't done the maths for this specific case, but just from feeling I 
would say that it's not practical. It's better to solder the LDMOSFET to 
a suitably machined chunk of copper, that provides much more water 
contact surface, along with the shortest and widest possible heat path 
through the copper. So you can use far lower water velocity, using just 
a simple, cheap, low power, quiet pump.

Manfred

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