Australia Telescope National Facility

The subreflector on each of the ATCA antennas can be raised and lowered to allow the signal to be focussed at the receiver feed. However, changing the subreflector position will change the path length (obviously) which will change the geometric phases and delays, so, it should only be done when required, usually when the frequency changes. Extra care is needed with calibration with programs that change focus during the observation.

A good time to do this is at the start of the observing and it can be done while driving to source.

Default focus positions have been determined for each band.

Note: As the antenna tips over, it deforms slightly, and the detected focus position changes by up to 1mm. This is normal and will not significantly affect the data.

The rms digitiser quantisation levels (as displayed in the sampler tab of cacor; see Figure 3.2) should be close to 20 (Range: Approximately 15-25). To set the levels, attenuation can be added or removed. The values in the Atten section of the window show the settings of the cacor attenuators. They have a range of 0-15 dB.

Figure 3.2. cacor sampler level display

For 16cm observations, strong RFI in the band will often make the sampler levels vary significantly from cycle to cycle, making it difficult (and often impossible) to set the attenuators appropriately. It is therefore recommended to skip this step when observing at 16cm. However, in rare cases the attenuator settings shown in the Atten section could be quite low (all values far below 7), and at those times it could be useful to reset them either automatically or manually as described below.

For all other bands, if the levels are not close to 20, the ATTS button in the cacor GUI can be activated to enable cacor to automatically add or remove attenuation.

This can also be done in the cacor interface:

Wait for the levels to settle to around 20, then turn the adjustment off again with:

Fairly obviously, the algorithm is unable to “improve” the levels if the attenuators are at their limits (0 or 15).

As noted above, strong RFI in the 16cm band makes it difficult to set the attenuators appropriately. High values of attenuator settings, even up to the maximum value of 15, are not harmful. However, you can also set the attenuators manually to values that are known to be good, as listed in Table 3.2. To set the attenuators manually, you will need to use the att command. For example, to set the CA03 attenuators as per Table 3.2, you would need to give the commands:

See attf ca0# n m for more complete information about how to use this command.

If the sampler levels are very low (approx 0), there is probably a problem which will need to be checked.

There are another set of attenuators in the receivers (as opposed to the correlator) which can also be used to set the levels.

These are controlled from caobs.

cacor generates an average visibility (either the mean or the median across the spectrum) that is used for T_sys and array calibration. In general, the median is much more stable and allows for a more reliable calibration. Otherwise, care needs to be taken to ensure that the range of cacor channels used for calculating the average is as wide as possible within the useable passband, but also excludes interference.

The default behaviour is to compute the average across the spectrum. To ensure that the more stable and reliable median is being used, it is recommended to enable the tvmedian option:

At higher frequencies and when the tvmedian option is enabled, it is usually appropriate to use the central fifty percent of channels. The easiest way to do this is to type

Alternatively if tvmedian is disabled (not recommended), the information below can be referred to.

At cm frequencies, interference means that it is often better to select non-default channels, when tvmedian is disabled (not recommended). See Table 3.3 for information on recommended channel ranges for 2048 MHz continuum bands. If these channel ranges are not generating the anticipated visibility in vis, inspect Appendix C to determine the cause of the problem.

The tvchan command sets the range of channels that are used for calibrating the array and for T_sys calculation. The same range is used to compute the average (or median) visibility for display in vis. All unflagged channels are written to the data (RPFITS) file regardless of what the tvchan setting is.

(Note: Channels that are flagged are written to the file as zeroes. See Section 3.3.5 for more information).

Current settings are displayed in the cacor GUI, as shown in Figure 3.3.

Figure 3.3. cacor display showing the currently used tvchannels.

It is also possible to check the current setting with the cacor command

Setting the range is done with the cacor command:

Table 3.3 lists the suggested tvchan ranges for the standard continuum bands. On any given day (or interference environment) other ranges may be more appropriate.

(There is a wide range of valid frequency settings at mm frequencies).

To calculate the delay of an individual IF, cacor checks the phase in each channel in the tvchan range (just defined) and calculates an average linear slope from the phase measurements as a function of frequency; this slope is the delay.

If the noise on the phase is high (due to a faint source or poor weather, for example) or there is some strong unavoidable RFI in the tvchan range, then the average slope may not accurately represent the real slope due to the delay.

To reduce the noise, or to average out the effect of the RFI, setting delavg in the correlator averages together (or takes the median of) the corresponding number of channels in the tvchan range before calculating the slope to determine the delay.

Here too it is useful to use the median option rather than the default mean (computed within each group of delavg channels):

Higher values of delavg can be useful for determining the delays. However, care needs to be taken when selecting this value! If the phases are wrapping quickly, averaging over a large number of channels will give a meaningless average phase, and therefore incorrect delays will be determined.

The current value of delavg is in the status bar of the cacor interface.

Figure 3.4. cacor showing current delavg setting

Setting delavg is done with the cacor command:

Figure 3.5.		Figure 3.6.
An spd plot demonstrating slow changes in phase. The phases are not wrapping at all in the A pol, and wrapping approximately every 250 channels in the B pol. Averaging over 32 channels would still provide enough points across the band to achieve a useful linear fit to the data, so delavg can be set to such a large value.		An spd plot demonstrating rapidly wrapping phases. The phases in both pols of this graph are wrapping across a small number of channels. It is possible to check just how quickly the phases are wrapping by viewing a smaller number of channels: spd> chan f# 400 500 The value of delavg will need to be set to a small number in this case.

Occasionally, there will be channels that, because of interference (or other issues), you may wish to completely excise from the data (i.e. not write those channels to the data file).

To check the extent of currently flagged channels:

Figure 3.7. cacor display showing the current number of channels that are unflagged.

There is no easy way to discover which channels have been flagged, and flagging will remain even with a change of project (or frequency etc.) so flagging should be checked at the start of observing. If there are more channels flagged than anticipated, flagging should be cleared then reflagged if necessary.

This is done by:

To flag the set of known bad continuum channels

This needs to be done at both f1 and f2.

To flag explicit channels, use the command:

Again, this needs to be done at both frequencies.

If you know exactly which channels have been flagged, they can be unflagged with the funflag command: formats are similar to the fflag command i.e.

Once the array is working, the system can be calibrated.

Calibration will not fix a broken system.

The calibration involves calculating and applying the delay offsets, checking the pointing (for mm observations), setting the phases to zero degrees and correcting the visibility amplitudes to the appropriate flux levels.

Figure 3.8. vis display showing output during ATCA calibration

3.3.6.1. Calibrate the Array Delays

Delay offsets need to be determined and corrected at the start of the observation. Failure to correct for delay offsets may result in decorrelation, and loss of data.

In the 64MHz zoom configuration, delay calibration is more complex as it generally needs to be done in two steps. (First delay calibrate a zoom channel, then in the continuum. See Section 3.3.6.1.1 for more information)

To calibrate the delays, track an unresolved source with large flux density (e.g. PKS1934-638 or PKS0823-500 for cm observations; for mm observations use PKS1253-055, PKS1921-293, PKS2223-052 or PKS0537-441). PKS1934-638 and PKS0823-500 do not have enough flux density to calibrate at mm frequencies.

Correlation should be apparent before you proceed. Phases in spd should be stable, though they may be wrapping rapidly and delays should be flat in vis. Check that you have correlation in both frequencies and both polarisations.

If there is significant interference (see the amplitude plot in spd), you will need to carefully craft the tvchan range (or, if bad enough, actually flag interfering channels) to avoid the interference in the calibration data.

If working at 3mm or with significant interference, it may be necessary to average over several channels (to improve the signal/noise ratio) with the command:

caobs > corr delavg #

where 8 is a good initial value for # - however, if the delays are large (>50nsec), you should try a smaller value to start with.

If you still don't have correlation, check Section 3.5.6 at the end of this chapter, and if this doesn't resolve your problems get help.

Assuming that the phases in spd are stable, enter the command:

caobs > corr dcal

dcal calculates and applies the required correction.

This command should correct the delays so that the phases across the band on the spd plot are flat (or fairly close to flat) and the delay offsets as seen on vis are close to 0 nsec. It may be necessary to do this again.

(Early versions of the CABB software required an “a” switch at the end of the dcal command for the correction to actually be applied. The delays are now applied automatically by the command. Any switch other than an a will result in the delays not being applied.)

3.3.6.1.1.  64MHz Delay Calibration

Delay calibration is slightly more complex in 64MHz mode.

The issue is that typical delays can cause the phases to wrap more than once within a single (64MHz wide) continuum channel. As is the case when delavg is set too high in 1MHz mode, a valid delay solution cannot be determined in this circumstance. The solution for this is to obtain an initial solution for the delays using zooms, which will unwrap the phases sufficiently to be able to determine the delays using the continuum channels. The process is illustrated in Figure 3.9 to Figure 3.13.

To use the zooms for delay calibration, set calband to z z:

cacor > calband z z

Uses zoom data in both frequencies. For a hybrid correlator config, use f z

Do the delay calibration as normal. At this stage a large value of delavg is often acceptable.

caobs > corr dcal

As always, ensure that there is a working system before doing the dcal

Figure 3.9.		Figure 3.10.
An spd plot demonstrating typical phase variations across a zoom band. In this case the delays are large enough that the phases wrap multiple times across the zoom band, and thus are also wrapping multiple times within the corresponding 64MHz-wide continuum channel.		An spd plot showing the zoom band phases after delay calibration. After performing delay calibration on this zoom band, the phases have been flattened across the 64MHz range.

Once the majority of the delay error has been corrected for, the whole continuum band needs to be used for a second delay calibration:

	cacor > calband f f
	caobs > corr dcal

Pay attention to the value of delavg at this step to ensure that it is sufficiently low. There are only a handful of channels to work with in the continuum bands in this mode!

Figure 3.11.		Figure 3.12.		Figure 3.13.
Continuum band phase variations before delay calibration. This is the continuum band corresponding to the zoom band shown in Figure 3.9. It is clear (especially for the A polarization) that before any delay calibration, the average phases within each 64MHz continuum channel are not useful to determine a delay solution.		Continuum band phase variations after zoom-band delay calibration. After performing delay calibration on the zoom bands, the phases have been largely flattened. The delays determined across a 64MHz frequency range are not sufficient to fully calibrate the continuum band, but they are sufficient to allow a refined delay calibration over the full continuum frequency range.		Continuum band phase variations after final delay calibration. After performing a further delay calibration on the continuum bands, an excellent solution has now been obtained. This is apparent from the flatness of the phases across the full frequency range.