Bad channel/Magic value handling

Bad-channel handling has been available for some time in the various mapping functions, although it tends to go unseen (except when you forget to do an INTERPOLATE for sparsely sampled data). A version of this has now been introduced for single-spectrum data. Channels in spectra set to this value are not displayed on plots, or used in computing quantities such as maximum intensity or line width. Channels which end up being unspecified, as the result of a SHIFT operation say, are now set to the bad channel value, rather than zero as before. Bad channels can be interpolated over with smoothing commands such as HANN, SMOOTH, CONVOLVE, BIN etc.

You can determine the current setting of the bad channel value by doing:

>> print badpix_value

and if you don't like it, you can change it by doing:

>> badpix_value = -100

for example. The one place where bad channels in spectra are not handled well is in spectra in a map cube -- fewest problems will result if you smooth over such bad channels before ADDing to the map, or choose a value of badpix_value close to zero (I tend to use 1.e-5 by default). Although we did have some teething problems with this feature, it seems to be working pretty well now -- It is particularly useful for dealing with DAS data, where the end channels of each subband tend to be set to $10^4$ or some such similar value, and totally destroy the auto-scaling of plots.

Note that if you change the value of badpix_value as shown above, then previously ``hidden'' channels in your data will now be displayed, and/or used in reduction operations.