There is a new program version V1.2.
The evaluation of the linear regression can now be switched off so that the frequency is calculated solely from the start and end times of the TDC. The associated variable 'F1_reg_modus' can be set to '0' or '1' manually or by serial command.
The ser. command looks like this:
.1Z to turn on and .0Z to turn off the calculation.
An Adev curve without regression calculation shows the reduced resolution of the measured values.
Initially, all calculations were carried out with 'double' values, which could no longer improve the results with longer measurement times > 3 s. As a new compiler option, there is now a #define INT128, which takes care of the acquisition and preprocessing of the time stamps down to the last bit. 'Double' values are only used for the final evaluation.
A maximum of 50 kHz is provided as the sampling rate for the time stamps, which at a clock frequency of 133 MHz uses 50% of the controller's capacity. If you activate the 2nd processor core, the time stamps can be recorded without restriction by the rest of the program.
In order to use the 2nd core, #define CORE1_AKTIV must be inserted in the source code.
At 133 MHz it is around 100 kHz sampling rate.
If you double the processor clock (3k3 resistor between GPIO1 and GPIO2) to 266 MHz, you get a sampling rate of around 200 kHz. The lower curve shows the comparison between these two operating modes.
Apparently, the higher sampling rate does not bring the hoped-for success. This as an experience.