Handling large spectral bandwidths

When using a single spectral window, and observing spectral lines, the bandwidth of the ATCA can be problematically narrow, or the channel widths can be too coarse. This is most likely to be an issue when observing spectral lines of extragalactic sources at high frequencies.

A partial work around for this situation is to use the two ATCA frequency bands to observe in adjacent somewhat overlapping windows, and using the same bandwidths and channel counts for the two windows. With the ATCA this is possible with some correlator configurations when observing with bandwidths of 16, 64 or 128 MHz. Miriad allows you to stitch these two spectral windows together in a reasonably straightforward fashion (although some quirks do become apparent). In so doing, it handles any overlap region between the two windows in a sensible fashion. The steps to achieve this are as follows:

  1. Make sure you set a rest frequency in atlod (or during the subsequent processing). If you really want to view the spectrum in frequency (rather than velocity), then you might set a dummy rest frequency as the average of the centre frequencies of the two spectral windows.
  2. During the flagging process, flag any channels in the overlap region between the two spectral windows which you do not wish to contribute to the final output stitched spectrum. These would be channels where the bandpass gain is quite low in one, but reasonable in another band. Task uvflag's edge parameter is the easiest way to flag these channels.

  3. Perform the calibration as described above, keeping both bands within the one file.

  4. Subtract any continuum, using task uvlin as normal. Task uvlin handles multiple spectral windows within a single dataset.

    Task uvlin processes each frequency band separately. You will want to give some thought when setting the range of channels to use in the fitting of the continuum. If the spectral of interest in near the edge of the individual spectra, you will want to set the channels to use in uvlin's fitting process carefully. You may wish to use a zeroth order fit, and just fit using channels at the edges away from the overlap region. Alternatively you may wish to run uvlin after the two spectra are stitched together instead. Some experimentation (and certainly thought) will be required.

    Note that when handling multiple bands, both Miriad's line and uvlin's chans parmaters number channels consecutively from 1 to the total number of simultaneous channels. The distinction between simultaneous bands is ignored in this numbering. But when fitting to the continuum, the boundaries between bands act as a break point.

  5. It is nearly time to stitch the two spectral windows together. To do this you will need to use the ``velocity linetype'' One of Miriad's quirks is that you cannot apply on-the-fly bandpass calibration and use a ``velocity linetype'' within the same task. You will need to make a copy of the dataset with the bandpass (and presumably other calibration) applied directly to the data. If you have used uvlin, you will have done this already. If not, create a copy of your dataset using uvaver.

  6. With a dataset with calibration applied, you can now stitch the two spectral windows together. See Sections 5.4 and 16.8 for information on using velocity linetypes. Do the copying and linetype conversion with uvaver (or possible uvlin if you have not already used it) to generate a new stitched copy of the data. By using a velocity linetype, when creating an output spectrum, Miriad copies channels by their velocity. Any channels that overlap between the two spectral windows are averaged together.

Miriad manager