Self-Balancing and Continuous Learning Without Internet Connection

My setup: I am hoping to use my tempest in a remote location without access to internet and will grab the data from the UDP broadcast. I will perform the initial setup at a location with internet, but only once and for a short period of time.

My question: I understand from a couple post (here and here) that the Tempest enters a self-balancing state when it is powered on which is fine, but are the variables for this calibration calculated and stored on device and without access to WeatherFlow servers? I also see that there is continuous learning applied to the values - am I correct in assuming that this is server-side and will not affect my reading accuracy? Is there any transparency into what exactly happens to get these values?


In short

when you power up the Tempest, it does a balance of the 4 wind sensors (needs a bit of wind from the four sides). This is done pretty quickly and the result is stored locally.
The other Cl values are done server side every x days and the result is pushed back to your hub.
Since your system will live off grid it can’t be done. You’ll have the factory settings by default.

Thank you, @eric. This was my rough understanding as well. The CL factory defaults are suffice for my usage.

A follow up question: If the device and hub have been online for a year and have new CL variable values, are those values reset in the hub on power cycle or stored in non-volatile memory? No matter what the CL variable values are, I will just need them to be consistent across all of my devices (no matter if they are setup in my lab today or in my home a year ago).

After confirmation by the coder, they are retained in hub and will be used after the reboot.

Will resetting the hub clear them?

No, the only reset you can do on the hub is radio reset.
If you want to go back to default, a technician has to reset a default CL and send it to your hub and from there set ‘no update’ to keep it this way if ever any contact is made with server.
I can’t do so and best way to go is via a ticket to customer support as this will leave trace of your request.

Understood - thank you for your help, @eric!

1 Like