Measurement Accuracy
There are a number of factors influencing the measurement accuracy. The main factors are measurement time (Tau), decimation, temperature changes of the tinyFPA and the noise floor of the tinyPFA
Input Frequency As the tinyPFA down converts the input signals to a constant IF frequency the measurement accuracy is proportional to the input frequency. All accuracy data is specified for 10 MHz, input signals at 1 Mhz will show a factor 10 less accuracy.
Measurement time The accuracy of measured phase is roughly proportional to the measurement time. Every decade decrease or increase of measurement time results in a decade decrease or increase of the accuracy. Thats is why the measurement accuracy is specified as better than 1e-12/(Tau in seconds). So when you measure phase every ms you can expect the accuracy to be better the 1e-9
Decimation Decimation is by default set to 10. This ensures the response to phase changes with frequencies close to the inverse of the measurement period is comparable to a PhaseStation. Setting the decimation to 1 reduces the measurement noise but does introduce some damping for phase change frequencies close to the inverse of the measurement period. This is best visible when measuring the ADEV. When decimation is set to 1 the ADV at the start of the plot will be a bit too optimistic. For best ADEV measurements the decimation should be set to 10.
Temperature changes The nanoVNA-H4 HW is not optimized for very high accuracy phase measurements. Temperature changes in the order of some degrees can create a phase change in the order of 1 ps. For measurements requiring sub ps accuracy the nanoVNA-H4 should be given ample time to reach a stable temperature and be kept at a constant temperature.
Noise floor The noise floor of the tinyPFA is the fundamental limit of the accuracy of phase measurement. With decimation set to 1 and a measurement period (Tau) of 1 second and kept at a constant temperature the noise level of phase measurements it in the order of 1e-13. When doing ADEV measurements of very stable sources one should always keep this noise floor in mind as the hard lower boundary
Example of noise floor measurement with all settings at default:
