I often do designing power modules and always wonder if there is some kind of rule how to properly cool SMD devices. I always look at final products from other companies and some are placing heatsink on bottom of PCB, while others pleace them on top - directly on SMD components with plastic packages.

Many power IC and MOSFETs do have large thermal pads on bottom, and thermal transfer should be really good to the bottom. But adding about eg 2mm FR4 of PCB is also good thermal insulator... Comparing to like 1 mm or less of epoxy resin on IO package, I dont think it could be worse...

To be more confusing, some power IC do not even have proper thermal pad - e.g. TPS63070 that I am working with right now.

What is better? Adding heatsink on bottom or on top? Does it depend on how thick Cu layers are?

  • 3
    \$\begingroup\$ If using the thermal pad, does the PCB have via's and a large ground plane to help dissipate the heat? Are there via's used to help pass some of the heat to the opposite side? What is the normal mounting direction of the PCB? Airflow in the enclosure? All these questions help answer that... \$\endgroup\$
    – Ron Beyer
    Commented Sep 26, 2019 at 13:50
  • 4
    \$\begingroup\$ I'm not sure that you should even look for a general rule. The device manufacturer knows about the thermal characteristics of their products, so follow the manufacturer's recommendation. \$\endgroup\$ Commented Sep 26, 2019 at 14:06
  • \$\begingroup\$ Try and get away with just the PCB as a heatsink if you can. Heatsinks are a PITA. \$\endgroup\$
    – DKNguyen
    Commented Sep 26, 2019 at 14:28

3 Answers 3


Here's a somewhat general rule. When a SMT transistor or IC has got a thermal pad at the bottom, it's designed to heatsink to the PCB. (The components without thermal pads heatsink to the PCB through pins.) The thermal resistance from the die to the thermal pad is much lower than to the top of the IC. In case of ICs, the thermal pad is as the ground potential, and you can connect it to a ground plane which is a good heat sink. In case of transistors, the thermal pad often can't be connected to a ground or power plane.

So we heat sunk the IC to the PCB. Then how do we dissipate the heat from the PCB?
A PCB can dissipate some power by itself. If the PCB is large enough, and the power which it has to dissipate is not too large, and the ambient temperature is not too high, then the PCB itself may be enough. If the design is dense, and the PCB can't dissipate enough power by itself, then additional heat sinking is needed.

You can heat sink the PCB to a metal enclosure. You can add SMT heat sinks. You can bolt-on heat sinks. There are different ways to do this, which depend on different requirements and situations.
This is where it becomes tough to speak about general rules.

additional reading:
How does power dissipation for surface mount components work?
Optimize heat sink design - connect cooling pad on PCB backside by vias


The general rule is, there is no general rule.

In every case, you need to understand the complete thermal path from die to ambient.

Some SMD parts have thermal pads, in which case, you can use lots of copper vias to conduct the heat to the other side of the PCB where you can deal with it more effectively.

Other SMDs use only their lead frames to get rid of heat, in which case, you can do much the same, with the additional requirements for electrical isolation.

Applying a heatsink directly to an epoxy package is the least desirable — but sometimes the only available — option.


You probably don't need a heatsink if only the thermal junction to board thermal resistance is 13C/W and the part supports an operating temperature of 125C.

Assuming the highest output voltage and current, which is 2A and 9W, this would amount to 18W of power through the device. With a lower bound efficiency of 80% this would mean a figure for worst case power dissipation would be around 3.6W. A 3.6W dissipation would amount to a 50C temperature rise above the PCB temperature.

This means if the PCB temperature were 40C, the part would be 90C, which would still be well within it's operating temp. This is also assuming no heat lost to air, the air contributes an additional 20% heat loss through the top.

There are also some caveats to this figure, one being that some of the inefficiency is due to the inductor and some power will be dissipated there, so this worst case scenario should be less than 50C.

You should use a heatsink if the board temperature is going to be high, or if the thermal ambient environment is bad (like higher than 50C). Keep in mind that a heatsink has 4x less of a thermal pathway than the pins, so the best way to get heat out of the part is through the pins, and a good thermal PCB design.

If not, then follow the recommended board layout, and use good thermal PCB design: enter image description here
Source: http://www.ti.com/lit/ds/symlink/tps63070.pdf


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.