Back in 2023, I created a subject to get help about a weather station that I inadvertently broke by applying twice its expected voltage at its power source
As I said at the time, I replaced it by another model of the same family and it had been running fine since then. The base principle is that it reads outdoor temperature from a TX35-IT sensor reporting on 866MHz and receives forecasts and time synchronization signals from a network of radio emitters across the territory.
Sadly, this station stopped working as well earlier this year, but this time this was not due to my own stupidity, but rather, it was not receiving the forecast any longer.
Reading online I discovered that I was not the only person in France affected by this and that if I was to restart the station, it would stop displaying the outdoor temperature as well.
Yes, you read that right, despite being totally local to my home, the weather station is incapable of reading the external sensor reliably if it has not seen the other signals in a timely manner!
This reminds me of all-in-one printers that no longer scan documents because of a missing yellow cartridge…
Clubic, a French tech website, started an investigation into the matter which you can read here:
Here is a summary of what happened:
- The stations are using the POCSAG protocol for their messages. Yes, you read that right, the pager protocol from the 90’s
- The weather station manufacturer had an agreement with the German operator (eMessage Germany) for its WetterDirekt products, that was never officially extended to the French subsidiary.
- The French subsidiary (eMessage France) went bankrupt in January 2026
- Following a justice ruling, another company was allowed to buy the assets in February 2026, with no obligations to continue providing existing services.
- The manufacturer has no intention to enter a new commercial agreement, effectively ending the life of those weather stations, redirecting to their “newest” WiFi based ones.
We could argue for hours that it’s quite strange to sell products relying on a third-party radio transmission provider without any commercial agreement, but that would not change the current situation.
Luckily, some curious and dedicated people had been capturing those very messages when the network was still on, and an effort has been jointly made to decode what’s found in those messages. The result of their investigation can be found here: https://forum.tetrahub.net/decodage-meteo/meteofrance-pocsag-t5935.html
Note: you need to have a login/password to read this, I believe this was done to keep AI scrapers at bay.
I came along this effort only recently, with the intent to build an emitter with a small enough range that it only covers my home.
I settled on an ESP32 with a RFM69C module as I already had it from a leftover RFlink32 experiment.
After identifying some missing details in the protocol, I was thrilled to see the weather station coming back to life with weather forecast via OpenMeteo
You can see it’s boot and synchronization sequence in this video:
This may seem convoluted but it matches what is in the user manual. The difference is that it now takes less than a minute to be fully synchronized while it could take up to 12 hours when the real signal was live.
This tiny project might also prove useful when the German signal is cut off, as announced for 31/12/2028: https://www.tfa-dostmann.de/en/produkte/wetterdirekt-stations/
The entire project is here: https://github.com/obones/CelebWeather
You can find the code but also documentation for the protocol as it is known today.
As I’m used to more recent devices, it felt weird working with 4-bit based “words”, but it sort of makes sense because it appears the CPU inside theses stations is a 4-bit one.
Talk about a blast from the past!
A warning must be given: this module emits on a frequency that is not free for general use (466.206250 MHz), so one must make sure it does not interfere with regular users on the same frequency. In my case, I made sure not to use RFM69HC module where the H denotes high transmit power, and I used on purpose a wire antenna of the wrong size, 8.5cm instead of the recommended 15cm. From empirical observation I noticed that the signal is only received for up to 4 meters through concrete walls, but if anyone has reference material on this subject, I’m all ears.
I hope you appreciated this report, do not hesitate to ask for clarifications if need be.