-initial_signals: - rats: desc todo - monkey: desc todo - humans: desc todo - proccessed signals: - bua: 1. Bandpass raw signals with a butter filter of order 4 and frequencies [300, 6000] Hz. 2. take the absolute value 3. lowpass the resulting signal at 300Hz 4. interpolate the result (to downsample) at 1000Hz (just to reduce data size) - neurons: generate a continuous signal at 1000Hz and counting the number of spiketimes in each bin. No smoothing. - data is zscored - fft computations on 1s hann windows, with 50% overlap, linear detrending - welch: mean norm of all window values on time dimension - coherence: on pairs of ffts by computing (fft1 * conj(fft2)).mean("time_dim") I forgot to divide by welch !!!