Bandpass Calibration

Bandpass calibration is necessary to correct for complex gain variations as a function of frequency. The shape of the bandpass is determined by the baseband filters as most other components have a flat response as a function of frequency. A typical bandpass amplitude profile has a steep rise, a flat top over the central 75% of the total bandwidth, and a steep decline (see Figure 3.1). The rise and decline, approximately the first 1/8$^{\rm th}$ and last 1/8$^{\rm th}$ of the total number of channels, should not be used. These edges of the bandpass are the effects of the particular filter that is used to set the bandpass. Across the bandpass, there is a linear phase slope of a few degrees per MHz. The channel-to-channel amplitude variations are of order 0.1-1%. Some antennas show a larger, few percent, amplitude ripple with an approximate width of 3 MHz (see below).

In the case of strong maser emission and/or narrow bandwidths, one can often do without a bandpass correction. This will depend, however, on individual circumstances. For all other observations one will have to do a proper band pass calibration by observing a strong, not necessarily unresolved source. Often the flux density calibrator will do.^3.2

The quality of the bandpass calibration determines the spectral dynamic range which one can achieve. This is defined as the ratio of the peak flux density of the strongest continuum source to the rms noise level in the spectrum. In an ideal world, the bandpass is stable and a single bandpass calibration of sufficient duration will be adequate. Sufficient duration implies that the signal-to-noise ratio reached on the bandpass calibrator in a channel for a particular baseline is at least as high as that on the target object. It is advisable to split the bandpass observation up in a few scans so that the integrity of the bandpass correction can be verified. Spectral dynamic ranges of 500:1 (100:1 at P-band) can thus be obtained.

Unfortunately, the bandpass is not stable as a function of time and more effort is required to achieve a higher spectral dynamic range. If one makes regular bandpass observations, one can see how a sinusoïdal ripple with a period of about 3 MHz seems to move across the bandpass at a rate of typically 0.4 MHz per hour. The origin is a standing wave in the 20 mm waveguide which connects an individual antenna to the main circular waveguide. Temperature variations affecting the waveguide result in the observed drift. Some antennas have a larger amplitude ripple than others and one can considerably improve one's spectral dynamic range by deleting the data taken with the worst offenders.

In order to minimize the error in the bandpass calibration, one will have to observe a bandpass calibrator more often, up to once every 15 minutes for the most demanding cases. Assuming the bandpass calibrator is as strong as the object under investigation, one has to spend at least the same amount of time on each of them. To date, the bandpass correction cannot be interpolated in time within AIPS, the best option being the application of the nearest bandpass solution in time to a particular scan. Deleting the worst antennas from a dataset and doing a very careful (meaning often) bandpass calibration can result in maps with spectral dynamic ranges of up to 5000:1 (lower at P-band).

When observing Galactic HI each bandpass calibrator should be observed at frequencies offset by a few MHz at either side of the frequency of the target field in order to avoid contamination by Galactic HI emission. This works well and takes out the gross shape of the bandpass. Offsetting in frequency does compromise the determination of the detailed bandpass shape and will limit the achievable spectral dynamic range to perhaps 200:1.