calc_mean_amplitudes

sofia_redux.scan.filters.filters_numba_functions.calc_mean_amplitudes(amplitudes, amplitude_weights, spectrum, windows, channel_indices)

Calculate the mean amplitudes and weights of a Fourier spectrum.

The amplitudes and amplitude_weights are updated in-place.

The amplitude of the spectrum at a single frequency channel (f) is calculated as:

amplitude = sqrt(2 * sum_{i=0}^{window-1}(a[f+i]^2 + b[f+i]^2) / w)

where

w = sum_{i=0}^{window-1}((a[f+i] != 0) + (b[f+i] != 0))

(a, b) are the (real, imaginary) components of spectrum and w are the amplitude_weights. Thus, if windows > 1 then this effectively downsamples the spectrum using a box average. Typically, windows is set to 1. Note that the Nyquist value is purely real, and stored in the last element of the spectrum a[-1] (b[-1] = 0). It’s amplitude and weight will likewise be stored in the last elements of amplitudes and amplitude_weights. The number of frequencies to calculate (nf/2) is determined from the input amplitude.shape[1]

Parameters:

amplitudesnumpy.ndarray (float)

The empty amplitudes array to fill with the calculated amplitudes of shape (n_channels, >=nf/2). Updated in-place.

amplitude_weightsnumpy.ndarray (float)

The empty amplitude weights array to fill with the calculated amplitude weights of shape (n_channels, >=nf/2). Updated in-place.

spectrumnumpy.ndarray (complex)

The spectrum from which to derive amplitudes. An array of shape (n_channels, >=windows * nf/2).

windowsint

The size of the averaging window in frequency bins.

channel_indicesnumpy.ndarray (int)

The channel indices for which to calculate amplitudes.

Returns:

None