1
\$\begingroup\$

I'm designing a mosfet PWM device with a rather high current. According to my LTspice simulation the mosfet has an average power dissipation of 20W. The mosfet had a TO-263S package ( https://fscdn.rohm.com/en/products/databook/datasheet/discrete/transistor/mosfet/r6020pnjfratl-e.pdf ). What would be the best setup/design to get rid of this heat? I have two of those mosfets next to each other, so the heat sink should be able to dissipate 40W. Can i just connect the heat sink to the top of the case or are there better alternatives?

EDIT: Datasheet talks about a maw PD of 305W, how can this ever be dissipated out of such a package? Are there special tricks to do so?

\$\endgroup\$
7
  • 1
    \$\begingroup\$ The best TO-263 heatsink I've seen (this one) is 5.6 degC/watt. So 20 watts means a temperature rise of 112 degC and very little headroom especially when the device itself is 0.4 degC/watt so, realistically it will warm by 120 degC and, in a controlled ambient of 25 the max temp is 145 degC. But even this won't account for local ambient heating of another 10, 20 or 30 deg. The MOSFET junction max temperature is 150 degC so, IMHO, wrong device package for your application without extensive fanning. \$\endgroup\$
    – Andy aka
    Commented Nov 25, 2020 at 16:54
  • 1
    \$\begingroup\$ I started a proper answer for this and it was both too long, too vague, and had me hitting the books too hard. Do a search on "thermal management for surface mount devices". There's a lot of information out there on the web. Basically, you need to get the heat from the FETs to the board, from the top of the board to the bottom (with lots of thermal vias), from the bottom of the board to a heat sink (with gap pad material, probably), and from the heat sink to the air. Expect to do a lot of work, and expect to either need a fan, or an external heat sink as big as a paperback book. \$\endgroup\$
    – TimWescott
    Commented Nov 25, 2020 at 17:46
  • \$\begingroup\$ This site is helpful for once you get the heat off the board and onto a heat sink. \$\endgroup\$
    – TimWescott
    Commented Nov 25, 2020 at 17:47
  • \$\begingroup\$ Oh -- and expect that you'll want to use thicker copper than the typical 0.5 or 1oz that you get if you don't ask for anything special. You'll be trading off the minimum feature size for copper thickness, but the power levels you're considering imply at least 2oz copper, if not 4 or 8 (so, no 0.5mm pitch packages on that board!) \$\endgroup\$
    – TimWescott
    Commented Nov 25, 2020 at 17:49
  • \$\begingroup\$ @TimWescott PCB will be 2oZ of copper. I'm thinking about a large copper pour with thermal via's and connect a heatsink to this pad. Although i have no experience with such heatsinks and i don't know what is the best solution to attach this to the copper pour \$\endgroup\$
    – Michiel Reyntjens
    Commented Nov 25, 2020 at 18:53

3 Answers 3

Reset to default
2
\$\begingroup\$

Surface-mount devices, in general, use a different method of heatsinking from through-hole ones.

You'll notice that one of the terminals of your TO-263 (also called D²PAK, incidentally) package is a very large tab on the back, unlike the others. This tab is bonded tightly to the die, and intended to be used as a heatsink.

What you do for this is use a large copper fill on your PCB (I'm not personally familiar with the calculations that go into how to size it), and use that as your heatsink. You can via-stitch this to a large fill on the other side of the board too, for additional power dissipation. The copper conducts both electric current and heat, and can fairly effectively radiate it out to the surrounding air.

If this still isn't enough, you can get extruded heatsinks intended to be mounted to such a copper fill, but generally if you're thinking of doing that, you may want to consider a different package.

\$\endgroup\$
13
  • 1
    \$\begingroup\$ Unlikely that copper fills alone would be able to dissipate 20W. \$\endgroup\$
    – Wesley Lee
    Commented Nov 25, 2020 at 16:41
  • \$\begingroup\$ @WesleyLee It is a lot to ask, but I expect it could be done with sufficiently thick copper (though you may be looking at 3 or 4 ounce copper, which gets expensive fast), a very large pad, and the via stitching I mentioned. Still, this isn't an application where I'd use a D²PAK; I'd look for a TO-220 or TO-247 variant that can be mounted to an extruded heatsink. \$\endgroup\$
    – Hearth
    Commented Nov 25, 2020 at 16:43
  • \$\begingroup\$ I know copper fills and via stitching is a common technique to do so. But i don't think it will be able to dissipate 40W( 2 mosfets close to each other). I'm pretty sure i will need a heatsink but i have no idea on how to have a good heat conductive connection between mosfets and heat sink? \$\endgroup\$
    – Michiel Reyntjens
    Commented Nov 25, 2020 at 16:44
  • \$\begingroup\$ @MichielReyntjens As I told Wesley, I would recommend you look for a different package, or a different part if you can. Surface-mount heatsinks can be bonded to the copper fill, but that seems like more trouble than it's worth, at least to me. \$\endgroup\$
    – Hearth
    Commented Nov 25, 2020 at 16:46
  • \$\begingroup\$ SMD thermal jumpers could help move that heat further away. \$\endgroup\$
    – rdtsc
    Commented Nov 25, 2020 at 16:47
0
\$\begingroup\$

Standard copper foil (1 ounce per square foot, the thickness of 35 microns or 1.4 mils) has thermal resistance of 70 degree C per square.

If the footprint is 1 square cm (0.4" on a side), then a 3 by 3 grid gets you to 1.2" square, with thermal resistance additional of 70/8 = 9 degree C temperature rise.

At 40 watts, the rise is 360 degree C.

Just to spread the heat out to 1.2 square inches.

I think you should plan on using a different package.

\$\endgroup\$
0
\$\begingroup\$

The TO-263 package can be advantageous e.g. for high frequency switiching due to lower parasitic inductance compared to the common TO-247 package. Also, space or part requirements can restrict the package choice.

For SMDs, the heat is usually transported to the other side of the PCB using thermal vias and the heatsink is connected on that side. For high power dissipation like in your case, thermal vias might not be sufficient for the heat transport to the other side of the PCB. In that case, it might be an option to use copper coins or inlays which are basically massive copper blocks instead of vias. 20W of power dissipation should be within the realms of possibility with such a solution.

A nice article about this topic can be found here: High efficient heat dissipation on printed circuit boards

Regarding your edit, those 304W power dissipation are possible if you can manage 25°C case temperature. It's rather theoretical. It is basically calculated directly from Rthjc and the maximum junction temperature e.g. 304W*0.41°C/W+25°C=150°C

You could think of extreme measures like liquid (nitrogen) cooling to make such an insane power dissipation at least a little realistic.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.