Wemos Temperature Node (how not to design a temperature sensor)

Hello people,

I wanted to share my semi-fail, hoping others will not make the same mistake.

Last week, I wanted to create a temperature sensor for my livingroom. I had a couple of Wemos D1 Mini laying around and also a AM2320. Here is the result after an evening of Inventor modeling, soldering and printing:

I severely underestimated the ESP heat generation. I knew it gets warm, but I assumed the generous amount of venting holes and my ‘thermal barrier wall’ between the ESP and sensor would be enough not to influence the sensor. Surprise: It’s not. Sensor reads 31C when it’s 24C in the room.

I fixed it for myself by adding deepsleep 15s/5min. This dutycycle seems to have a negligible effect on the sensor. But for people who want a faster, more real-time, update interval, consider thermal management carefully. Either move the sensor away from the wemos, or design it so that the ESP heat is lead away from the sensor somehow.

That’s a shame. It’s a nice looking case.

I use two of these glued back to back.

The ESP is mounted on the far back lid and the DHT22 on the front. So there is a vented air gap, double thickness plastic than another vented air gap between them.

Took a while to come up with this version but I finally have something that suffers no self heating issues with it running 24/7.

Thanks. It works fine I guess for most purposes. Now that I have switched from ESPHome API to MQTT, I can have have it update every 2min. Its on for about 7-8s, then sleeps 2min. Impact on temperature compared to 5min sleep is (maybe) 0,1C. But the temperature in my room might as well have changed.

I will keep your design in mind for when I’ll want to make a real-time sensor. It also looks very nice.