Super-resolve synthesized channels, then average in frequency

[iancze]

For very coarse frequency channels, it may no longer be a valid approximation to simply evaluate V(u,v, nu) at the nu corresponding to the center of the channel. Instead, we should synthesize say 4 or more channels at finer dnu across this channel, and then Hanning smooth them to a single channel.

[iancze]

Some more notes on this appear on the MPoL issue tracker. When modeling a disk, you need to first identify the frequency resolution that "Nyquist samples" the kinematics of the problem. If you have very cold gas (narrow lines), this could be a much finer frequency resolution than your target resolution.

Then, you need to convolve the FFT'd cube by the spectral response function (corresponding to your correlator setup) and resample it at the appropriate channel centers.