I am thinking this way
Device due to interference in a full-metal package, which uses a radiator, during long work, it reaches 60 ° C, the safety overtemperature stop is set at 80C .
The original solution.
Use crystal any as
this , and this is the main and practically the only source of instability, see Frequency Tolerance @+25°C 50ppm, Frequency Stability over the Operating
Temperature +-50ppm and any manufacturer not mentioned long time stability.
New solution with HRTIM
Firt source error is HSE crystal for MCU as clock source , for example 8MHz as stable as above,
second source error all inside MCU, above all PLL and Delay line in HRTIM ,I do not know what is thermal and above all long time stability of this . I do not know whether these errors will tend to add up theoretically, if the frequency error is large and it is still increasing upwards, collisions may occur output LPF (6th order) and output power will heat LPF , theoretical :thinker:
I probably complicate my life.
I have on the table a big 10kW induction melting furnace , where STM32F334 with HRTIM directly drive full MOSFET drive and in fact it also determines the frequency, and I can not see the problem.
Probably because frequency is in order 100kHz and frequency is online tuned on resonst freuency. May be because the value of the HRTIM counter is large and not 340 as on 13.56MHz output PWM.
I can try program STM32F3334 discovery HRTIM on this frequency, problem is I do not have the equipment to accurately measure the clock parameters. I have old HP 5334A with precise time base option and HF divider, but with this I measure the max frequency