The controller of the oven of my vintage crystal oscillator

Initial design decisions for the controller:
The time constants of the model are very high and an analog implementation would be very bulky (large capacitors), so a digital implementation is selected.
Differentiation of such a slow changing temperature cannot be done reliably since noise will cover it. The controller will not use differentiation, so settling time will be large and some overshoot may be present.
Zero final error is required for accuracy, so integration of the error will be implemented.
In order to keep low the dissipation outside the oven and have linear control over the supplied heating power, a pulse width modulation (PWM) scheme will be used. The PWM frequency can be quite low because the time constants of the oven are quite large.
After these design decisions, a Proportional – Integral (PI) controller will be implemented digitally. PWM control at 225 Hz will be supplied by an AVR running at 14.7456 MHz using a 16 bit counter.
After building, the oscillation frequency is observed for different temperature set points. A clear minimum is found near 65 degrees Celsius. This is a characteristic of AT cut crystals. The minimum is the optimum operating temperature because small temperature changes around this point have very small effects on output frequency.
This is the last video about the evaluation of my vintage quartz crystals.