ATCA Multi-Frequency Observing Strategies -- MFPLAN

   Given that you are going to perform a multi-frequency experiment, you will have determine the appropriate frequencies, and how many frequencies will be observed with each configuration. As the multi-frequency synthesis techniques in MIRIAD do not cope well with large spreads in frequencies (greater than about 30%), your frequencies should be confined to one of the ATCA observing bands (i.e. either 3, 6, 13 or 20 cm). In the following, we will assume that you want to derive the best result at one frequency band, and so all frequencies will be confined to one ATCA observing band.

There are some pretty simple rules to follow for the number of frequencies to observe per configuration:

• For detection experiments and the like, u-v coverage is generally not important. The time wasted in switching frequencies means time lost integrating. It is best to observe at two frequencies, which are measured simultaneously.
• When observing in the 13 and 20 cm bands, the width of the bands (relative to the 128-MHz continuum bandwidth) is comparatively small. There appears little benefit switching frequencies during the observation, as there are few places to switch to, and there is plenty of interference. Again it is probably best to observe at just two frequencies, which are measured simultaneously.
• The 3 and 6 cm bands are quite wide relative to the 128 MHz continuum bandwidth. It is advantageous to switch frequencies at approximately 5 minute intervals. Either 4 (usually) or 6 (occasionally) frequencies are appropriate. (i.e. 2 or 3 pairs of frequencies). As you will want to observe you calibrators in with the same frequency setup, a typical observing strategy would be to cycle through the frequency setup on your program source two or three times, and then cycle through them once on the secondary calibrator.

The above are the number of frequencies to use per configuration. In general, if you have multiple configurations, you will observe with different frequencies for each configuration. When two or three configurations are being used, and u-v coverage out to 6 km is required, the best choice of configurations is a combination of 6.0 with 1.5 and 0.75 configurations ( not two or three 6.0 sets).

Given the number of frequencies to observe per configuration, and the configurations to be used, the task mfplan can find frequencies which optimise the u-v coverage. In doing so, it does not consider tangential holes -- effectively it assumes that the frequencies are switched sufficiently rapidly that these are not a major consideration. A description of the parameters of mfplan follows:

• freq: This gives a number of pairs of frequency ranges. Each pair gives a range over which observation is possible. Generally you would set this to avoid observing near known interference sources. This is important in the 20- and 13-cm bands, but less so in the 6- and 3-cm bands. See the technical memo `A survey of the interference in the 13 and 20 cm bands at the ATCA' by Mark Wieringa and Ravi Subrahmanyan.

• nfreq: The number of frequencies to observe (for each configuration). Generally this will be 2, 4 or 6. As the ATCA can observe two separate frequencies simultaneously, this means that it would cycle around 1 to 3 settings.

• nants: The number of antennas to use. This can be set to 5 or 6, depending on whether you want to optimize u-v coverage over 3 km or 6 km.

• config: ATCA configuration identifier, such as 3.0a, 3.0b, etc. See the help file for the possible configurations. Several configurations can be given. The observing frequencies determined by mfplan will be optimized for the best u-v coverage from the combination of data from all the configurations.
• device: Standard PGPLOT device, where a plot of the u-v coverage will be produced. The default is no plot. This plot assumes that you are observing at the pole. Nor does it allow for frequency switching (i.e. it assumes all frequencies are observed simultaneously).

Task mfplan can take a while to run if there are many frequencies or configurations. As it iterates towards a solution, it prints out a ``temperature'' (a control parameter in the optimisation, which is not of much interest to you) and a ``goodness''. This ``goodness'', the parameter that mfplan is optimising, is the ratio of a measure of the goodness of the actual u-v coverage to the ideal u-v coverage that could be achieved given the number of baselines and number of frequencies. The goodness will always be less than 1, because mfplan has no control over the baselines (you, or the Time Assignment Committee and the ATCA itself, dictate these), and the frequency band is also constrained. If the number of configurations and frequencies are the same, the goodness measures are directly comparable. However the goodness measure does not give too much insight into the relative merits of using different numbers of configurations or frequencies.