This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TMP117: Placement

Part Number: TMP117

Could you please expound on the recommended use of the thermal pad used in most of your TMP local temp sensors (for instance TMP117). I am looking to measure the temperature of a solid surface. The datasheet seems to suggest measuring through the PCB, but aren't FR4 and solder mask poor thermal conductors?

There is also mention of a copper plane on the bottom of the board. Since FR4 is a poor thermal conductor, should the copper plane be connected to the exposed pad with thermal vias to help conduct the heat through the PCB to the exposed pad? Should the solder mask be removed from that copper plane so it is in more direct contact with the surface being measured?

Ideally the sensor and its PCB would be upside-down on my solid surface so that only the top of the sensor is touching my solid surface and so the PCB doesn't effect the surface's temperature gradient. Is that better or worse than sensing through FR4?

Any other advice you could offer is greatly appreciated.

Thank you,


  • Hi Garrett,

    Please note that this discussion of thermal pad only applies to the DRV (aka WSON) package variant of TMP117. The YBG (aka DSBGA) package variant of TMP117 does not have a thermal pad.

    Thermal pads are meant to extract heat from high-power IC. To do this effectively, they should be soldered to a metal pad that can carry thermal energy away. The solder aids in thermal transfer. To increase the effectiveness of the PCB heatsink, there should be thermal vias to couple the metal pad thermally to other metal layers in the PCB. This allows the PCB heatsink to be larger than the device package dimensions, and to even thermally couple to an external heatsink. This kind of PCB design is covered in the following application report:

    Temperature sensors are not high-power IC and do not need help dissipating heat. However, we can use the same thermal path that was created for heat dissipation to thermally couple our sensor to the object we wish to measure. The one difference is that the sensor only benefits from thermal conductivity of the materials used; it doesn't benefit from added thermal mass in the path. Thermal mass slows the change in temperature like a capacitor. The additional mass usually comes with additional area, which helps in the heat dissipation application. For the sensor application, it's important to note that thermal mass slows temperature change for both heating and cooling. The sensor doesn't benefit from additional area; you cannot build a thermal antenna to gather more signal.

    To directly answer some of your questions, yes, FR4 and mask are poor thermal conductors relative to metals. The mask should be removed from the pad even if the pad is not soldered. It should be the paste layer of your PCB design that decides whether the pad gets soldered in automated assembly. We have to recommend that the thermal pad not be soldered on TMP117, because we have observed a shift in temperature error due to the mechanical stress of soldering. If you can tolerate the small error introduced by this shift, soldering the pad does aid in thermal conductivity through the thermal pad. Your last question about the top of the sensor requires some further discussion.

    The sensor can be deployed to measure an object above or below it. Both approaches are viable, and likely perform about the same in the DRV package. When measuring an object below, we can use thermal vias (as discussed above) to effectively transfer through the PCB. The thermal pad gives more efficient thermal transfer, as expressed in the theta jc (bot) specification, but requires the added thermal mass of the vias to use this path. The top of the package is described by the theta jc (top) specification, but allows us more direct access. One downside to this access is that we have to be careful when clamping or attaching to not damage or stress the package. This imposes assembly limitations. These thermal metrics reflect market standards written for heat dissipation, so they don't reflect sensor applications. They also have their own problems, as described in the SPRA953 application report. However, they still give some insight into the relative characteristic of these thermal paths.



  • Ren,

    Excellent. Thank you for pointing out that table. I believe that answered all my questions.