I previously tried something like that with temperatures, but I didn't obtain reliable results as it appeared that the cycles don't have a consistent duration and this prevents automating things straightforwardly without "massaging" the data too much.
Beautiful and clear analysis CAN - Thank You!
I'm surprised the temperature cycles are not consistent. The PID is simply looping a program cycle:
Starting at 600°C
Ramp to 900°C 1 minute
Hold at 900°C 15 minutes
Ramp to 600°C 1 minute (cooling takes longer than that, but this is about the time sequence)
Hold at 600°C 15 minutes
and repeat for 1000 cycles, or ~51.6 hours.
So each cycle should be 32 minutes in duration, and the power bands on your chart shows 32.7 near as I can tell, pretty close.
Temperature will lag the power steps of course, due to the thermal inertia of the system. That is dependent to some degree on ambient temp and airflow, and may be causing some timing jitter in the temp data. But there should still be a temperature cycle corresponding to each power cycle, of roughly equal length.