search

Kladdy / neutrino-dnn

Deep-learning applied to neutrino property reconstruction
0 stars 0 forks source link

Add bandpass filter to remove frequencies that are not relevant #2

Closed Kladdy closed 3 years ago

Kladdy commented 3 years ago

Right now, the bandpass filter is 80-800 MHz (?), SNR improves filter to lower freqencies, ie 80 - 500 MHz, or 80 - 300 MHz. Implement Christians bandpass to test and see if it improves performance. The filters are found under _direction/tools/bandpassfilters. The filters are of type "10th order butterworth lowpass filter for 300MHz and 500Mhz"

How frequency filtering is done:

import numpy as np
filt = np.load("500MHz_filter.npy")
data = np.load("data/SouthPole/data_000001_LPDA_2of4_100Hz_4LPDA_1dipole_fullband_0000.npy")
data2 = np.fft.irfft(np.fft.rfft(data, axis=-1) * filt, axis=-1)

In load_data:

data = np.load(os.path.join(datapath, f"data_01_LPDA_2of4_3sigma_{i_file:04d}.npy"), allow_pickle=True)
data = np.fft.irfft(np.fft.rfft(data, axis=-1) * filt, axis=-1)
data= data[:, :, :, np.newaxis]
Kladdy commented 3 years ago

run50 showed that this did not improve anything