You have a problem.  I don't see how a DPAK can get rid of this much heat (you
didn't say what the high temp limit is of the environment) unless you use a
substrate (PC board) that can carry the heat away as that is what this package
was designed for.  The TO-220 uses the tab which you can bolt to and with a
copper block probably can conduct the heat to the chassis BUT then eventually
the chassis (which I presume doesn't have fins or significant air flow across
it, must itself get hot and loose heat by radiation (black box radiation) and
convection.  The more external surface you have (unpainted metal is better) the
better you will be.

How to determine how well you are doing?  Thermocouples placed on the tab, on
the chassis and in the air surrounding the box.  At 85C, you are pushing a
boulder up hill - you need to keep the junction temp below, what, 150C?  Every
10 degrees C doubles reliability in my book so get that die temp down by any
means you can.

I don't have the spec on this MOSFET but get ones with as low Rdss on as you can
as this is what is creating the heat.  If you go from 0.02 ohms to 0.04 ohms
then obviously the dissapation doubles.

I personally don't like DPAK - it can't get rid of the heat - you have found
this out the hard way.  Sorry.  Heatsinks inside chassiss don't do much but they
do help even if the heatsink is nothing but a strip of aluminum under the TO-220
tab bolted to the chassis wall.

You MUST instrument with thermocouples and a chart recorder in a temp chamber to
measure the improvements.  I like them best as they are the smallest sensor and
so can be wedged in places other sensors cannot.

Good luck,

Bob Landman
H&L Instruments LLC


----- Original Message -----
From: Lum Wee Mei <[log in to unmask]>
To: <[log in to unmask]>
Sent: July 28, 1999 3:59 AM
Subject: [DC] D3PAK Thermal Dissipation


> I have a power mosfet part number APT20M45SVFR from Advanced Power
> Technology. The initial design was done on FR4 substrate. Although the
> board is tested functionally OK, we have a problem of heat dissipation
> from 6 of these poer mosfet.
>
> These 6pcs of power mosfet are D3PAK package. Presently, the heat is
> conducted through the surface by contacting with the chassis box with
> help from thermal tape. However, this method does not seem sufficient.
> We improvised it by providing a wide exposed trace to the drain and
> attached copper block to dissipate the heat onto the chassis. However,
> our thermal group said that this method is still not sifficient to get
> rid of the heat.
>
> The component supplier said that this D3PAK has to be used with alumina
> substrate to provide efficient heat dissipation. My questions are :
>
> 1. Is alumina/ alumina nitrate substrate really necessary?
> 2. If no, what other cheaper and effective method can we use to
> dissipate the heat. Don't suggest heat sink as space is a constraint!
> 3. Although each Mosfet can handle up to 300W, we only expect to reach
> 120W. How do you determine the heat dissipation on this board if it is
> enclosed within a chassis box?
> 4. We are using a 28V with 4Amp to drive these 6 mosfets in two groups
> of 3. Is 120W too high or too low?
>
> Hope someone out there can help me out of this HOT situation : (
> Thanking you in advance.
>