[Amps] FET cases / heat sink

[Manfred Mornhinweg]

Chris,

> I know that Manfred and others on this list have advanced
> thermodynamics knowledge

Well, mine isn't really advanced, it's just somewhat better than that of 
most hams. I still need to use online calculators for more complex 
things such as calculating the thermal resistance of a heat spreader.

> There is prevalent thought that unless one can significantly derate
> device power handling, bolted device to copper heat spreader
> connections are  quite inferior - soldering is best.  The device case
> to copper connection seems the weakest link.

It's not the weakest link, but it can indeed contribute a very 
significant amount of thermal resistance, while a good solder joint is a 
lot better. In a really "serious" amplifier, the weakest link is 
actually the internal thermal resistance, in those few millimeters from 
the silicon chip to the mounting surface of the FET. When using a good 
spreader, heatsink, fans, or specially if using a good water-cooled 
block, the device's internal thermal resistance can be higher than the 
sum of all other thermal resistances!

In a rather casual setup, such as bolting a transistor onto a generous 
aluminium heatsink, the internal thermal resistance of the device can 
still be about a third of the total, at least in the case of smaller 
transistors.

> Why don't the manufacturers add another $3.00 worth of copper to the
> case size (and mass) to better spread the heat over more area (and
> slow the thermal time constant)?

I can only take a guess at this, as I'm not a manufacturer. My guess is 
this: The datasheet looks better when the internal thermal resistance is 
lower, since that determines the official datasheet power dissipation 
rating! And the internal thermal resistance would INCREASE rather than 
decrease, if the manufacturer added more copper between the silicon and 
the mounting surface, due to the longer path!

So, mounting the silicon chip on a thicker copper base is clearly a bad 
idea, from the ratings point of view. Instead using a copper base with a 
larger area, but not being any thicker, would slightly improve the power 
dissipation rating. But the improvement provided by this would be rather 
small.

So the manufacturers opt for using a very small copper base, make it 
thin so that the thermal path is short and thus the internal thermal 
resistance is low (and the power dissipation rating in the datasheet is 
high), and they leave it to the system designer to keep that small 
mounting surface cool.

Such transistors are really designed to be soldered to the cooling 
system. The manufacturers provide two versions of the cases: One that 
MUST be soldered, and another that has screw notches, so you have the 
choice of soldering it or mounting it with bolts and thermal grease. 
The latter option really only makes sense in high efficiency 
applications, or in pulsed service at low duty cycle, so that the power 
dissipation is only a small fraction of the value rated for the FET.

And once we accept the fact that these FETs are intended to be soldered 
down, it suddenly makes sense that the manufacturer provides just that 
minimal, small and thin piece of copper: The thermal resistance of a 
solder joint is very low, and now the system designer has total 
flexibility in designing the optimal cooling system! This is good, 
because these FETs can work into the VHF and even UHF range, and they 
work at very low impedances, and the mounting surface is also the source 
connection. So it's necessary to optimize the heat spreader or cooling 
block not only for cooling, but also to provide the lowest possible 
impedance between the source connection and those parts of the circuit 
where it matters. If the FETs were provided with a big copper base built 
in, this would interfere with such optimization!

It's certainly not that the manufacturers are trying to save on copper. 
Three dollars currently buys you more than a pound of pure copper! The 
added cost for a bigger copper base would be insignificant.

> Wouldn't this greatly improve the functioning?  Wouldn't bolted
> mounting then work better?

Yes, a bigger mounting surface would indeed improve dissipation when 
using bolted mounting. But it wouldn't make a significant difference in 
the thermal aspect when using a soldered mounting, and it would make it 
harder to get low impedance connections to the source.

Manfred

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