Frequency Switching and Multi-Frequency Synthesis

Inexperienced observers should make sure that they consult with an AT expert before undertaking a frequency switching observation.

As described in the Introduction, the ATCA provides four IF signal paths for each antenna pair. The local oscillator system supplies two local oscillator signals, so that you can observe two frequencies simultaneously (e.g., observe several molecular lines simultaneously). The two frequencies can be anywhere within the tunable frequency range of the observing band and rapid frequency switching within an observing band can be used to obtain more than two bands. Observing frequencies can be set to the nearest MHz only.

This gives rise to the possibility of exploiting this agility to improve the $uv$ coverage of the array. As the antennas are at fixed physical distances from each other, by changing the observing frequency, you can alter the number of wavelengths between the antennas and hence modify the effective baseline lengths. In multi-frequency synthesis this is done rapidly (several times a minute), thus acquiring data with more complete $uv$ coverage than might otherwise be obtained. Substantial reductions in observing time are possible for dynamic range limited fields, so the user is encouraged to use multi-frequency synthesis.

A switch between frequencies from cycle-to-cycle is a process taking the order of 10 seconds. A switch between bands, with the rotation of the antenna turret, takes less than 20 seconds. Note that hardware limits restrict turret rotations (to move from the 3/6cm feedhorn to 13/20cm and vice versa) to occur at intervals of 15 minutes or more.

The ATCA has a special fast frequency switch mode. This allows you to switch frequencies within a single scan, thus avoiding the scan startup overhead of 3 cycles. Note that at present there is still a one-cycle switching glitch in the system, so you lose one integration cycle at every switch. To use this mode, specify sctype=FREQSW in the schedule file and make a file in the $ATCA_FREQSW directory with the name source.frq, where source is the same as the source name of the FREQSW scan. Frequency switching files should comply to the following format (comment lines start with a hash (#)). Data lines, which should contain 6 fields, separated by spaces:

1. Frequency 1, integer, (MHz)
2. Frequency 2, integer, (MHz)
3. Number of cycles to spend on the frequency pair, integer
4. F1 LO inversions. To find the value, run Lo_chain & check the ``phase sign", e.g. 4800 MHz at 128 MHz BW gives +1.
5. F2 LO conversions
6. Field name, must begin with $ (which is used to mark the start of the name, and is not itself included). The field name must be an ASCII string, no longer than 9 characters

The file must not contain blank lines. An example frequency switching file follows:

# $ATCA_FREQSW:EXAMPLE.FRQ
#
# Example FREQSW file to step through
# a set of 6cm frequencies, with a static second frequency
#
5278 8640 3  1 1 $FREQS001
5286 8640 3  1 1 $FREQS002
5294 8640 3  1 1 $FREQS003
5302 8640 3 -1 1 $FREQS004
5310 8640 3 -1 1 $FREQS005
5318 8640 3 -1 1 $FREQS006
5326 8640 3 -1 1 $FREQS007
5334 8640 3 -1 1 $FREQS008
5342 8640 3 -1 1 $FREQS009

If you are intending to switch more than once per minute, it is advisable to have a 2nd FRQ file with slower switching rate for setting up. This will allow samplers and attenuators to settle down for each frequency combination. Thereafter they will track system temperature changes.

Use the envflg field in ATCASCHED when frequency switching. This saves the attenuator and sampler settings used at a particular frequency so that they can rapidly revert to the appropriate state when changing frequency. The use of more than two frequencies involves time-sharing. Thus it will not increase the effective observing time or the sensitivity. Given the resulting gaps in hour angle coverage for any one frequency, you will have to judge whether the $uv$ coverage is usefully improved.

When both multi-frequency synthesis and two or three configurations are being used and $uv$ coverage to 6km is required, the best choice of configuration is a combination of 6km with 1.5km and 750m arrays with all arrays using the 6km antenna. The 128MHz bandwidth provides 1% frequency spread at 3cm and 8% at 20cm; the latter is a significant improvement in $uv$ coverage if all channels are retained throughout the data processing. Note that the flux density will vary significantly between different frequencies, and the spectral index may vary across your source. This may be of interest in itself, or it may make a meaningful combination of the dual frequency $uv$ data difficult, especially when high dynamic range imaging is required. The task MFPLAN in MIRIAD can be used to plan multi-frequency synthesis observations. When it comes time to reduce your data, MIRIAD provides algorithms to solve for both a basic flux density image and a spectral index image.

For more information about frequency switching interested observers should refer to the following documents:

-

The MIRIAD Users guide, particularly `Multi-Frequency Synthesis Observing Strategies'. This gives a very good description of frequency switching and the ramifications it has on the reduction of the data.

-

Multi-Frequency synthesis with the ATCA, Sault 1992.

-

Multi-frequency synthesis techniques in radio interferometry, Sault & Wieringa 1994.

-

Multi-Frequency Synthesis, Conway & Sault 1995.