cacal

cacal is the program used to calibrate the delay, phase and gain of the antennas. Only the delay calibration is essential: bad delays cause decorrelation across the passband, resulting in reduced S/N. The calibration of phase and gain is done mainly to aid in the on-line interpretation of the data (VIS display) and diagnostics of faults (ASSISTANCE). The off-line calibration using AIPS or miriad will redetermine the phases and gains.

Using CACAL

The standard approach to calibration of the array is as follows:

Check that the Tsys field on CAOBS reads ON.
Enter the command coabs> enable tsys if not.
Track a strong calibrator (e.g. PKS1934-638).
Wait until the telescope is on source and the attenuators and sampler statistics have settled down.
In an xbones terminal, enter the command
xbones% cacal
Answer a to the question "Automatic or Interactive?"
Check the results on the VIS display - it takes a few cycles for the data to be displayed. If all went well ASSISTANCE should tell you "ALL SYSTEMS OK" and VIS should show the delays and phases all close to zero and the amplitudes all equal (for each frequency).

Problems with the standard calibration

The standard approach works well for wide band (128MHz) calibration in the absence of interference. If your setup is different you may experience one of the following problems.

Narrow bandwidth problems.

At bandwidths of 32MHz or less you may have trouble calibrating the delays due to insufficient S/N in the narrow channels. VIS will display wildly varying delays in this case. Note that at narrow bandwidth the delay calibration is much less critical, so you may be able to skip it completely. As a rule of thumb: you want the delays to be less than 32/bandwidth[MHz] nanoseconds. Note however that if you don't form a continuum channel but process the channels individually in the reduction you can use the channel bandwidth instead of the observing bandwidth in this equation. So if VIS displays wildly varying delays, but the amplitudes are roughly ok, you can probably skip the delay calibration (use Interactive mode in CACAL). If after all this you still think you need to improve your delay calibration you have the following options:

Find a stronger calibrator. (See the ATCA Calibrator Page for more information. Ensure that your choice is compact at your frequency.)
Cacal will equalized gains on such a calibrator, no absolute calibration is performed.
Reconfigure to a correlator configuration with the same bandwidth, but less channels, e.g. use full_4_32 when the observation will use full_4_1024 This will improve the S/N for the determination of the delays.

If you 'randomised' your delays by calibrating on a source that is not strong enough, you can restore the global delay values using the CACAL Reset option (choose Interactive and then Reset, then D for delays). These values are often good enough for spectral line observing.

If the global values are not sufficiently close to zero (e.g., because you have an unusual frequency) the following recipe will let you use delays determined at your frequency but at a wider bandwidth:

Run CACAL in EXPERT mode (type 'E' at first prompt)
Select Reset
Select S (Specified offsets)
Look up your combination of frequency and bandwidth and enter it and any others with the same frequency but less than 128MHz bandwidth.
Exit (type Q)
Make new schedule with config FULL_128_1 (or _2), 128 MHz bandwidth and the same observing frequency.
Track calibrator and run CACAL (Automatic mode).
Go back to your original schedule and recalibrate Phase and Gain only (use Interactive mode in CACAL).

Interference problems

If you are experiencing problems due to interference (have a look at the shortest baseline on the SPD amplitude display), the following may help.

Determine if there is a range of channels that are ok. You may need to type X, to get an Amplitude display with channel numbers along the bottom. If there is no interference free range of channels spanning about 50% of the band you may want to reconsider your frequency choice.
Enter the range of channels that are ok in the textbox in the cacor gui with: Command: tvch[an] x₁ x₂ [x₁ x₂]
Tasks on xbones (e.g. cacal, vis, camon, pointing etc.) will be using data based on the new channel range, so you can retry cacal.

Background

cacal uses a 16 cycle in-memory history of the visibilities and can average up to 8 cycles to improve the accuracy. (It actually takes medians.) It checks whether the antennas are on-source to determine if the data is valid. For the gain calibration to function properly the Tsys correction has to be enabled in CAOBS (=default state). If two frequencies are used they are calibrated simultaneously. Calibration of delays and phases is done relative to the reference antenna. The XY phase (when available) is used to calibrate the relative phase of the XX and YY polarization. Amplitude calibration is relative to the specified flux of the calibrator (or the average flux).

All calibration corrections are stored in a 16-slot buffer in memory, new frequency/bandwidth combinations are assigned new slots, replacing the least frequently used slots. The default settings are derived from the slot closest in frequency, and with the same or larger bandwidth. There is a separate buffer for global delays and early-normal calibration, these are determined at reconfiguration.

CACAL can also determine the polarization leakage terms (thanks to Bob Sault for the algorithm), this option is only offered in interactive mode and when on 1934-638.

Automatic mode

In this mode CACAL first does the delay calibration using a 3-cycle average, waits for these to be implemented and then goes on to do a phase and gain calibration. It knows the flux density for 1934-638 and 0823-500 across the ATCA observing bands so it can do an absolute calibration, for other sources the gains are equalized only (i.e. it uses the average flux). The whole process takes about 12 cycles.

Interactive mode

In this mode questions are asked at every stage

Calibrate array or Reset calibration? The reset option allows zeroing of delays and phases, and resetting of the gains to 1.0. This may be needed to recover from calibrating on blank sky or with a faulty correlator.
The reference antenna
The number of integration cycles to average (max=8)
Choice of calibrating D(elay), G(ain), P(hase) or A(ll).
The calibrator flux(es) can be specified. Note that two values are needed in dual frequency mode. The value 1.0 will cause the average flux to be used (i.e. it equalizes the gains).
Optional calibration of polarization leakage terms when on 1934-638. At present these are not applied to the RPFITS data, but they can be applied by VIS for on-line display of calibrated Stokes parameters.
When applying the phase corrections, there is an option to apply a 90 degree offset to the Y phases. This should only be used for PHASED ARRAY observations that need circular polarization.
When applying phase corrections, you can choose to continuously calibrate the phase. This should only be used for PHASED_ARRAY observations.

Confirmation to proceed is asked at every calibration step, a N answer terminates the program.

Expert Mode

This mode is for local staff and experienced Duty Astronomers only. It offers a few more calibration options and some dangerous reset options. Expert mode is entered by typing E at the initial prompt. In calibration mode you can:

Specify the frequency chain(s) to use
Specify antennas to ignore during calibration (apart from those already disabled in CAOBS)
Calibrate the global delays
Determine the polarization leakage terms on sources other than PKS1934-638

In reset mode you can:

Remove stored calibration factors for specific frequency/bandwidth combinations (e.g. to avoid using delays determined at narrow band)
Reset the entire offset calibration area, i.e., start from scratch with only the global delays and early normal calibration in place.
Reset the global delays (for use after reconfiguration)

CACAL - Compact Array Calibration