I know, I worked through the documentation when writing the java library.
That was the idea t drop it to 50 packets/s - 100sps. I do understand that it can't be guaranteed, but I think it's worth a shot.
re Nyquist
I do understand that if you are trying to sample at <2 times the highest occuring frequency the signal cannot be constructed properly and aliasing occurs. However in my understanding a little more than 2 times the frequency should be fine (eg. audio sampling at 44.1kHz due to the audible range for humans ending at ~20kHz - sampling frequency = 2.2*frequency). This equals 2 (2.2) sample points per wave.
So what I don't understand is the jump from 2.2 samples per waveform to 4/5?
Can you see that sample rates anywhere CLOSE TO the Nyquist limit are going to give extremely aliased and poor reproductions of the sampled signal? Look at some of the diagrams in that white paper. Such as Figure 10 or 11. In Figure 10, now shift the phase so that the samples are at the zero crossings, disaster. In Figure 11, the reproduced waveform from the samples would also be absurdly distorted.
I'm not saying that the 4 or 5 samples per waveform is any kind of carved in stone standard. But 2 or 2.2 samples per waveform is going to result in gross distortion at the upper limit.
Comments