Open brettviren opened 10 months ago
This replaces the slightly incorrect decimate-then-convolve with proper filtered downsample and linear convolve for FRxER. There is still a very slight unexplained difference between 512ns with 64ns resampled FR, 500ns with 100ns native FR and 512ns to 500ns resample.
I have one more confusion pop up in my mind w.r.t. the fact that ER is peak-normalized. I make this draft again until I can clarify and maybe make some changes.
This PR adds support for: