Task sdsmooth performs smoothing along the spectral axis using a user-specified smoothing kernel. Currently, 'Gaussian' and 'boxcar' kernels are supported.

The default Kernel shape is 'Gaussian'. The width of the function is specified with the kwidth parameter, in number of channels.

Standard data selection parameters are used: antenna, field, spw, scan, timerange, pol, and intent.


Weights are propagated to smoothed spectra following:

$W^{k}_{\rm out} = \frac{ (\sum_{i=1}^{n} c_{i})^{2} } { (\sum_{i=1}^{n} c^{2}_{i} / W^{j+i}_{\rm inp}) } $    ,

where $W$ is the input/output weight, $c$ is the spectral smoothing kernel originated from a channel $k$ where the weight is evaluated, and $n$ is the width of the spectral smoothing kernel in channels. In the case of Gaussian, $n$ corresponds to FWHM in channels.