Python • Re: very noisy SoC temperature measurement (RPi5)

Having ascertained that a Raspberry Pi's SoC temperature reporting has noise, what are you going to do with it?

I've written a daemon in C that does some filtering on the fly and spits data in a file located in /tmp.
I'm testing it for now and will drop it on gitlab or github at some time.
Currently it has moving average, 1st order IIR, 2nd order IIR, Wiener and Kalman options at launch.
Wiener and Kalman have low time lag and preserve transients better, but they're probably overkill.

There's also a chance that a 100 Hz Python loop will generate enough load to warm the CPU a couple of degrees. The considerable initial CPU load bringing up the Python interpreter and compiling the program code may have an effect, too.

Yes, even compiled stuff have a thermal impact. Awakening a process every 10ms for some floating point task has a higher cost than once per second, for sure. But "measuring modifies what is measured" has been a reality known in the lab way before quantum physics, there's nothing new here.

There's much less massaging done to processor sensors on smaller SoCs.

Our Raspberry Pi 5 has a hard time keeping its cool. A massage would definitely do it some good

Statistics: Posted by pragma — Sun Feb 11, 2024 10:33 am

---