Does the ESP-32S thermal/ground pad need paste/soldering?

[electronicsguy]

I bought the DIYMall ESP32 dev board: https://www.amazon.com/gp/product/B01MT ... UTF8&psc=1
and the ESP-32S package: https://www.amazon.com/gp/product/B01LX ... UTF8&psc=1

The ESP-32S has a thermal/ground pad in the center. The DIYMall dev board pcb too has a small square in the footprint. I only found one discussion about this, but the Question remain unanswered:
viewtopic.php?f=2&t=1273&hilit=thermal+pad

Does the pad need some kind of mating with the PCB. Do we need to put any thermal paste there? or does it need to be soldered to the PCB - which by the way I have no idea how to do, if so.

Could someone please answer this?

[ESP_Sprite]

No, no need to. Our own devkits don't have a pad for this either, and the ESP32 doesn't get *that* warm.

[WiFive]

No, no need to. Our own devkits don't have a pad for this either, and the ESP32 doesn't get *that* warm.

Pls add recommendation to the wroom32 datasheet.

[esch_esp]

I have tested both cases.
If the thermal pad is connected (by soldering) to the pcb, I read about 62 (probably this means 62 degree celsius) from the internal temp sensor.
If the thermal pad is not connected to the pcb, as it seems is the case for the development boards I got so far from aliexpress, the internal temp reads to about 140.

[electronicsguy]

I have tested both cases.

@esch_esp Thanks. Could you explain how you managed to solder it to the pcb? Also, have you tried thermal paste instead of soldering?

Also, what were the conditions for testing? Could you mention what benchmark/tests you were running when measuring the temperature? And what was the ambient temp?

[esch_esp]

Hello,

To get the thermal pad connected, I used one of the standard adapter boards from ali e.g. https://www.aliexpress.com/item/5Pcs-ES ... 178.bnub8x&. Then drilled a 1.5 mm hole through the thermal pad position on the adapter board.
By that it was easy to solder the thermal pad from the back. The proper way of course would be to have a via in the board design ..
So far I have not tried thermal paste. But also did not expect that much heat dissipation.
I did not do any benchmark tests. Just running a webserver based on a local filesystem, including an adapted version of tinyshell. Debugging information is transferred over tcpip to an ajax based log window. The idf is version 2, freertos. This of course if not much load on that CPU yet.
Ambient temperature around 22 degrees celsius.
Touching the metal case of a board which has no thermal contact and shows around 140 with the finger, the value goes down by about 5 values. So there is an effect, but not that high. I was surprised, that frequency scaling still is not supported via the API, that could be the answer for some heat issues. I am running my application at 240 MHz.

[electronicsguy]

@esch_esp thank you for the detail answer. very helpful. How did you change the frequency from 80MHz to 240MHz? Is 240MHz the core and bus frequency?

[esch_esp]

The frequency can be set in the makefile configuration "make menuconfig->Component config->Esp32-specific->cpu frequency".
It should be the core frequency. Changing it from 240 MHz to 80 MHz has an effect on the temperature. The temp 140 reading goes down to about 125 ...

[johnabel]

I have tested both cases.
If the thermal pad is connected (by soldering) to the pcb, I read about 62 (probably this means 62 degree celsius) from the internal temp sensor.
If the thermal pad is not connected to the pcb, as it seems is the case for the development boards I got so far from aliexpress, the internal temp reads to about 140.

I don't expect that chip to go that hot. A reading of "140" could be any ADC value representing any temperature. Most electronics stop or fail above 125 C! You wouldn't be able to put your finger on for long at that temperature - a quick way to check if that's real or not.

[esch_esp]

Good point ! I agree the chip should not be working at 140 C. The esp32 chip is shielded in a metal housing.
You will not feel the heat produced in the die on the metal housing and that metal part, which seem not to be attached to the thermal pad, already gets warm.
I am reading these values with the function "temprature_sens_read()". It would be good to know what is read out with the
function ? Should it represent ADC values or a temperature close to a defined unit (C/F) ?