Calculating thermal resistance for multiple FETs on a heat sink?

[John]

I think I'm doing this the right way but I wanted to make sure.

I have 10 FETs being mounted on a single heat sink. To calculate the
maximum theta-sa (thermal resistance, sink to air) for the heat sink
can I merely add together the theta-jc (junction to case) and theta-cs
(case to sink) for each FET and then divide by 10 to arrive at the
"effective" theta-js to use for calculating the max theta-sa that the
sink can have?

Example:
IRF1405
theta-jc = 0.45 degrees C/W
theta-cs = 0.50 degrees C/W (flat greased surface)***

Added together, the theta-js for one device is 0.95. Dividing this by
10, we get 0.095 degrees C/W.

Assuming that the devices are spaced out well on the heat sink, can I
use that 0.095 degrees C/W value for my theta-js for the other
equations?

Thanks!


***Why is the spec 0.50 degrees C/W when the thermal resistance of a
good compound is about 0.05? If the theta-jc takes care of the
resistance all the way to the case, shouldn't the resistance from the
case to the sink be just the resistance of the compound itself and not
this 0.50 number? Where is the extra resistance coming from?


-- remove SPAMMENOT for e-mail responses --

 

[Jon]

John,
This will work, assuming that the power is equally distributed among
the 10 FETS, and that you use the total power for your temparature rise
calculations from junction to case. There are some 2nd order effects
that are not taken into account when you simply add the thermal
resistances together: There is a "spreading resistance" effect caused
by heat flow across the heat sink. The safest thing to do is to mount
the transistors, apply known power, and measure the case temperature
rise for the worst case transistor. This will give you the true worst
case thermal resistance junction to case. Then base your heat sink
thermal resistance requirements on this value.
Regards,
Jon

 

[John]

This will work, assuming that the power is equally distributed among
Thanks Jon! Lots of prototyping for me to do but knowing that I can
get myself into the neighborhood with these calculations will
definitely help shorten the proto cycle.

-- remove SPAMMENOT for e-mail responses --