6
\$\begingroup\$

I really like the specs of the Infineon OptiMOSPower-Transistor IPT007N06N, but how would I keep this beast cool?

Usually I use regular aluminium heat sinks (with water cooling or a fan) for MOSFETs in this power category, but this has to be soldered onto a PCB. Would I create a larger copper area and then attach a heatsink to this, or is there a way to keep the PCB itself cool? What is the usual the strategy here and are there any sample designs?

Datasheet of IPT007N06N (pdf 1.2 MB): https://datasheet.octopart.com/IPT007N06NATMA1-Infineon-datasheet-43351903.pdf

\$\endgroup\$

4 Answers 4

6
\$\begingroup\$

Depends on your power levels. From the datasheet you'll get a temperature rise of 62C/W on a minimal copper area, but this goes down to 40C/W with a larger cooling area. A multilayer plane (if you can fit it) with plenty of vias for extra area and copper makes 6cm^2 not all that much. If you look through datasheets for other TOLL-88 packages (or similar, Infineon has weird packages), you'll usually find recommended PCB layouts. At 0.75mOhm rdson, 25C ambient, and Mercury in retrograde you could put 56A through that part before hitting 120C on your junction. The datasheet says you can go as high as 175, but I never take power semis even close to that. (With the minimal copper area you're looking at ~45A to hit 120C).

enter image description here

If this still isn't enough for you, then start looking into heatsinks. The whole point of these packages is to save space, and heatsinks do not do that...

Of course, my analysis only works if you are not switching frequently as it does not take turn-on and turn-off losses into account. It assumes your max ambient is 25C, etc, etc. I'll leave calculating your actual power consumption to you.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ I read a Texas Instruments app note, stating that you cannot take these thermal resistance values literal. You can only use them to compare to other packages since it is a standardized test. \$\endgroup\$
    – Bonnevie
    Jul 27, 2020 at 14:06
  • 1
    \$\begingroup\$ Oh no, of course not. That's why I built in 50C of margin. The actual value/measurement has too many variables. It isn't going to be super far off though. Whats 20% error amongst engineers? Any chance you know what app note that was (for my own edification)? \$\endgroup\$
    – Stiddily
    Jul 27, 2020 at 18:08
  • 1
    \$\begingroup\$ Ah okay then we agree ^^. It is called TI SPRA953C (ti.com/lit/an/spra953c/…) \$\endgroup\$
    – Bonnevie
    Jul 28, 2020 at 6:24
  • \$\begingroup\$ Thanks for that! It's good to see how they are actually measuring these. \$\endgroup\$
    – Stiddily
    Jul 28, 2020 at 11:54
3
\$\begingroup\$

You can buy heatsinks that can solder to the PCB copper but, for this device, if you look at the thermal characteristics in the data sheet on page 4, you will see that directly attaching a heatsink to the case is probably you best bet. The typical value of thermal resistance from junction to case is 0.2 degC per watt. Something like this ought to be an option: -

enter image description here

Available from Farnell. There are plenty of options to consider and plenty of offerings from the usual heatsink suppliers.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ Just a quick note on this. As someone who has had to specify heatsinks and other thermal materials, make sure the adhesive that's used on the heatsink you choose is suitable for the package material. You can get thermal attach tapes which are specifically engineered for low surface-energy materials (like plastic packages). If you get the wrong type, they are liable to drop off! \$\endgroup\$
    – SiHa
    Jul 27, 2020 at 13:14
  • \$\begingroup\$ This probably works a negligable amount compared to the bottom tab due to the high Rtjp of the plastic. If that is what you want lookup Infineon DirectFET. It does work for low power chips, like memory. \$\endgroup\$
    – Jeroen3
    Jul 27, 2020 at 13:57
  • \$\begingroup\$ There's no way you're getting a Tjc of 0.2C/W on the plastic side of a package. I've only seen top-side Tjc specified when there's a metal contact on top, or in an isolated TO-220 package. \$\endgroup\$
    – W5VO
    Jul 27, 2020 at 22:19
3
\$\begingroup\$

Cooling from the bottom:

I know from some "Gansystems" applications, that they attach heatsinks on the other side of the PCB. This will require multiple vias in the component footprint to get the heat down to the other side of the PCB. But this strategy is viable, and maybe you have a lot of space on the back.

Cooling from the top:

If you desire 1 heatsink per 1 component then a gluetype heatsink will do. If you require 1 heatsink for multiple transistors the transistors might not be mounted exactly equal in height or planarity. Then go for a heatsink with screw/bolt mount into the PCB with a heatpad to get proper thermal connection.

\$\endgroup\$
1
\$\begingroup\$

The only answer I've found (and used extensively) is to bend up the legs of a (say) TO-220 device, and invert on PCB using through-hole. Requires hand soldering in production, but can take 100 W off a power PCB per FET.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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