QTPy ESP32-S3 worryingly high temp (66ºC/150ºF) / high current (140mA) when idle

[todbot]

CircuitPython version

Adafruit CircuitPython 7.3.0-beta.2 on 2022-04-27; Adafruit QT Py ESP32-S3 no psram with ESP32S3

Code/REPL

import _bleio

Behavior

I'm experiencing worryingly high-temps (66ºC/150ºF) and high-current draw (140mA) on QT Py ESP32-S3 when sitting idle in the REPL. Right out of reset it'll draw ~50mA but then jump to 100mA. Upon import _bleio current draw immediately goes to 140mA and keeps rising, with Vreg temperature feeling dangerously hot.

Description

This seems CircuitPython ESP32-S3 internals-related, but may be a general ESP32-S3 issue.

I'm experiencing worryingly high-temps (66ºC/150ºF) and high-current draw (140mA) on QT Py ESP32-S3 when sitting idle in the REPL. Right out of reset it'll draw ~50mA but then jump to 100mA. Upon import _bleio current draw immediately goes to 140mA and keeps rising, with Vreg temperature feeling dangerously hot.

These below photos were taken about 10 minutes apart. There's a thermocouple kapton'd on top of the Vreg.

On a Feather ESP32-S3 w/ no PSRAM and Adafruit CircuitPython 7.3.0-beta.2 on 2022-04-27; Adafruit Feather ESP32S3 No PSRAM with ESP32S3 current draw is also 140mA on but the board doesn't get nearly as hot: about 38ºC (100ºF) at the Vreg near the BAT pin.

On a QT Py ESP32-S2, idle or connected to WiFi is around 50mA and the highest draw I can get is 100mA when running a simple HTTP JSON API server using ampule and a simple hammer of it with while [ 1 ] ; do ; curl http://192.168.42.174/status ; done
Is the ESP32-S3 just a more power-hungry chip, is there some chip setting that's wrong, or is there an issue with the vregs?

Additional information

No response

[dhalbert]

Filed on my request after Discord discussion. Thanks!

[ladyada]

@brentru check yours? mine gets warm but not 150

[todbot]

I was worried is a one-off issue too and checked the other QTPy ESP32-S3 that I have and both behave similarly.

[ladyada]

what if you use arduino?

[todbot]

I just tried the esp32 2.0.3-RC1 Arduino core. And:

• 20mA - ROM bootloader mode
• 40mA - TinyUSB bootloader mode (not used but interesting)
• 100mA - Running void loop() {Serial.println("Hello world!"); delay(500);}
• 156mA - Running "BLE_uart" example but not connecting

So not a CircuitPython issue exactly but wth?

And grasping the board between thumb and index finger gives me a little burny feeling when its drawing 156mA.

[ladyada]

what about a wifiscan example?

[todbot]

Good call.

• 110mA - Running "WiFiScan" example Arduino sketch

Thermocouple on vreg reading about 45ºC (112ºF).

[ladyada]

note your temps should be in C - 150*C is silicon limit, do you mind converting/editing comments above?

[todbot]

Done. Apologies for that.

[ladyada]

ok just to verify you're getting the same power draw on the Feather and the QT Py S3, its just the QT is warmer right?
there's not a lot of ground plane on the QT to dissipate the generated current, and it is a dual core 240 mhz processor. so being toasty isnt that out of bounds

@UnexpectedMaker are you seeing similar power draws? i dont have one of your boards with me right now to check :)

[UnexpectedMaker]

@ladyada I see approx 145mA draw when importing _bleio - but temp wise on my TinyS3, it's not getting above 45deg C after about 5 mins.

[dhalbert]

I am also seeing similar power consumption, but not excessively high temperatures. The regulator on top is at most 50C, and the ESP32-S3 is about 45C. I have essentially the same measuring equipment as @UnexpectedMaker: exact same multimeter, and similar USB monitor. I also tested with a Thermoworks Thermapen Mk4 cooking thermometer, which has more sinking ability.

My QT Py has header pins installed, which are getting warm and probably acting as a good heat sink. @todbot, it appears you do not have header pins installed.

[ladyada]

@todbot assuming the current is OK (seems like its high for just an import but like, not a hardware issue) - can you check the temp rise of the other 2x S3 QTPys u have? it may be our RF isnt tuned well enough and/or im wondering if its a one-off

[todbot]

I'm thinking this is just a vreg heatsinking issue with the QTPy just being so small that there's not much copper on the board and the ESP32 being known for having 130mA BLE "by default" that people complain about it. (I did some searching, people have been complaining about this since at least 2017)

The Feather ESP32-S3 I have has similar current consumption, but its vregs don't get that hot (40ºC at most)

I tested all 3 of my QT Py ESP32-S3s:

• All three behaved the same: 140mA when running "BLE_uart" Arduino sketch
• Headers do seem to act as heatsinks, it was 55ºC. (Pic 1)
• The no-header one was 60ºC in a vice (vice as heat-sink?)
• After 20 mins, the free-floating with headers one went to 170mA and 60ºC (Pic 2)

Ambient was 23ºC, 43% humidity, no fans.

Janky test setup for reference:

"Pic1" - Both headers & no-headers in a vice after 5 minutes

"Pic2" - headers free-floating after 20 mins

[UnexpectedMaker]

@todbot I think it's less about the LDO and more about the lack of ways for heat to exit the board. My TinyX boards run hotter than my Feather format boards, with the same specs. The Feather boards have 3x more PCB surface exposure to air, allowing more heat to get into the copper pours, and then dissipate out off the board.

The QT Py is smaller than my TinyX boards and double sided as well, so even less air contact area for heat to dissipate. That might be reflected in components that get hot, like the LDO and the ESP32, but 170mA is not taxing the LDO on the QT Py much. It can do 500mA sustained and peak to 600mA.

[todbot]

I was assuming it was the LDO producing most of the heat, but yeah totally. From reading other ESP32 issues/forums, looks like the ESP32 heats up too.

[brentru]

Tested: Qt Py ESP32-S3 running WipperSnapper Sketch (1x AHT20 sending measurements to AIO every 5m), arduino-esp32 2.0.3-RC1 core.

Avg. QT Py ESP32-S3 Temp (NO headers): 44degC
Avg. Environmental Temp/Humid: 18degC/48%

[ladyada]

@todbot yah your boards probably have less-than-ideal tuning which causes some RF self-heating, but also - its a small board, with a powerful chip. the regulator is doing a lot and there isn't a ton of space for it. your board isnt damaged and wont be damaged from use, its just a toasty lil boy :)

[dhalbert]

@todbot We'll close this for now, and note the temperature characteristics in the guide. We appreciate your original writeup: it was definitely worth investigating, especially since it might have been a CircuitPython issue. Thanks.

[ladyada]

@todbot you could attach a smol heatsink if you like
https://www.adafruit.com/product/1493

[todbot]

Agree on closing this issue. Everything seems to be working as intended. I was all excited to try BLE on ESP32 and was startled it was so hot when idly grabbing it. But after reading up, I see ESP32 BLE is much less developed than NRF. Not surprising, I guess.

@ladyada ahaha that heatsink is adorable! I'm sure I'd fat-finger it to oblivion. And right now I'd rather use an NRF52840-based approach.