I’ve just received my XIAO ESP32S3 Sense from Mouser. When I first uploaded the camera streaming example from the Seeed wiki, I noticed that the ESP32S3 board heats up very quickly. I implemented the reading of the internal temperature sensor with an external temperature probe attached to the bottom of the board. I found that the ESP32 is overheating rapidly, with the internal temperature exceeding 80°C and the external temperature reaching over 95°C+ before I disconnected it. With the camera board disconnected, the measured internal temperature was around 60°C and external was about 55°C. With the camera board but without the attached camera, the temperature reached 67°C. With the camera itself attached to the camera board but not initialized, the temperature hit 75°C in a short time. When I initialized the camera with streaming, the temperatures skyrocketed past 80°C, as I described earlier.
Sorry, but I believe the XIAO ESP32S3 Sense board design is flawed. I don’t really know which specific IC generates so much heat, but I suspect the voltage regulator, based on the rising temperature even without the camera attached. It’s undeniable that the ESP32 generates a lot of heat, but this amount of heat is concerning. Without schematics, I can’t pinpoint the problem. With such a tight hardware layout and grounding, it’s impossible to discern which IC contributes to heating up the board. All I can say is that I can’t run camera streaming indefinitely, even with a copper heatsink.
I really hoped that the board would be cool with all the work you’ve done to design it, but I can’t use it. It will destroy itself without any modifications or a huge cooling solution.
Yeah, but why it is so hot even without initialized camera?
The ESP32 was idling while board goes to 75°C and rising.
Just the difference between “the idle ESP32 board without cam/SD addon” and “the idle ESP32 with cam/SD addon and ov2640 attached” is 20°C+. In both mentioned scenarios, the Wifi AP and printing of the internal temperature into the USB serial port every 1 second were active.
The heat has got to be coming from processor power, its not like an internal short circuit or something so the problem is in the code, its trying to do too much, so i dont think it is hardware specific
I just got my hands on another board supporting a camera out of the box, and the temperature was acceptable (under 50 °C) even after hours of camera streaming. It uses ESP32-S3 on the module instead of the chip directly on the board, but this info only points out there might be a flaw in the hardware design of the XIAO with ESP32-S3.
This post is not an advertisement, just an example of how cold an ESP32-S3 with a camera can run. I just added a periodic printout of internal temperature to the unused main task loop (in the camera stream example) with a period of 10 seconds. I measured the external temperature too and it was about 6 °C lower. https://1drv.ms/i/s!Ai_4aucnnOOOj8lSNrHBmKb-CROZjA
Sadly, I am convinced that a different camera can’t change the XIAO ESP32-S3 Sense board is overheating. I used different cameras of the OV2640 type. The XIAO ESP32-S3 Sense always overheats and is hot even without it. The camera works as expected and gets only slightly warm duting prolonged use. So there is no point in a purchase of another camera type. I am more interested in global shutter cameras. I work with fast robots like small autonomous racing cars and drones.
I started playing with Xiao ESP32-S3 sense. I see same poor thermal performance. Even with 1x1x0.5cm heatsink from stepper motor driver the temps are quite high. Just streaming i2s audio and not having sense board attached causes temps to be around 55C. Adding sense board and limiting fps for camera image to 1fps makes temp hover around 60-62 (800-600 res). Doing anything above 1fps makes temps skyrocket well above 65 which is the rated temp for this xiao module. For me the camera on this board is not usable for this reason.
I will still use the board a bit more to see what can be done with it but this thermal issue is a dissapointment for me. This may be due to compactness and power delivery chips (xiao has some buck converter which may run hot and sense board have two LDOs which need to burn some of the voltage into heat). The board mentioned by @electry has some dedicated PMIC doing all the power delivery so it may be better thermalwise.
If anyone has those issues and knows of some smart solution (other than slapping and intel cpu cooler on this) I am all ears.
