E.8 Target source data — edit and SPLIT

At this point, your calibration should be finished. You should now do an initial editing on the target sources, much like that done above for the calibration sources. Run RFLAG on the target source(s) and perhaps use:

> DEFAULT TVFLG ; INP

to reset all adverbs and choose the task.

> INDI n; GETN m  C R

to select the data set on disk n and catalog number m.

> DOCAL 1 ; DOBAND 3  C R

to apply the delay, complex gain, and bandpass calibration.

> BCHAN c1 ; ECHAN c2  C R

to average across a range of channels — not as flexible as ICHANSEL but probably okay here.

> BIF j ; EIF BIF  C R

to edit one IF only, which will suffice for problems that are not IF dependent, such as drop outs, antenna not on source, etc. Choose an IF that is reasonably free of RFI.

> NCHAV ECHAN-BCHAN+1  C R

to average all the channels into one number.

> CALCODE ’-CAL’  C R

to do just target sources now.

> DPARM(6) Δ t  C R

to do no time averaging in the work file set Δ t to the data interval in seconds.

> GO  C R

to start the task.

Again, remember to set it to flag all channels and IFs. You may have to select sub-windows and force the averaging to one times Δ t to edit in detail, or perhaps the default time averaging will be beneficial. In general, the DISPLAY AMP V DIFF is a powerful way to catch bad amplitudes and phases. It will catch drop outs either as bright lines for strong sources or dark grey ones for weak sources.

Since EVLA data sets tend to be large and unwieldy, it is recommended that you separate the data into the separate target sources, applying the current calibration and flagging once and for all. The imaging task IMAGR can do this on the fly, but, especially for observations of spectral-line sources, this is excessively expensive.

> DEFAULT SPLIT ; INP

to reset all adverbs and choose the task.

> INDI n; GETN m  C R

to select the data set on disk n and catalog number m.

> DOCAL 1 ; DOBAND 3  C R

to apply the delay, complex gain, and bandpass calibration.

> CALCODE ’-CAL’  C R

to do just target sources now.

> GO  C R

to write out separate calibrated data sets for each target source.

A task new to 31DEC13 will help calibrate data sets which contain a strong, e.g., maser, line in one or a few channels with interesting but weaker signals in other channels and/or the continuum. After applying the best standard calibration described above, you can split out the maser channel(s) with UVCOP. Make images and self-calibrate following the processes described in Chapter 5 making an SN table containing the full calibration needed to apply to the maser. Then use task SNP2D to convert this into delays and phases to apply to the full, multi-IF data set. This will work so long as the residual phases found in the self-cal are small and is actually required to do the best calibration possible over very wide bandwidths.

Unless TYAPL has been used, EVLA data sets have weights which only reflect the integration time in seconds. Calibration routines do not change these weights when changing the data amplitudes. There is a task called REWAY which computes a robust rms over spectral channels within each IF and polarization. It can simply base the weights on these on a record-by-record, baseline-by-baseline basis. Alternatively, it can use a scrolling buffer in time so that the robust rms includes data for a user-specified number of records surrounding the current one. A third choice is to average the single-time rmses over a time range and then convert them to antenna-based rmses. In all three modes, the task can then smooth the rmses over time applying clipping based on user adverbs and the mean and variance found in the rmses. A flag table (extension file) may be written to the input data file removing those data found to have rmses that are either too high or too low. For these weights to be meaningful, the bandpass and spectral polarization calibration must be applied and it helps to omit any RFI or other real spectral-line signal channels from the rms computation. For the weights to be correctly calibrated, all amplitude calibration must also be applied. For these reasons, REWAY might well be used instead of SPLIT — when TYAPL was not used — running it one source at a time. Thus,

> DEFAULT REWAY ; INP

to reset all adverbs and choose the task.

> INDI n; GETN m  C R

to select the data set on disk n and catalog number m.

> DOCAL 1 ; DOBAND 3  C R

to apply the delay, complex gain, and bandpass calibration.

> SOURCE target1’ , ’ ’  C R

to do one target source.

> APARM 11, 30, 12, 0, 10, 4  C R

to use a rolling buffer of 11 times separated by no more than 30 seconds and then smoothed further with a Gaussian 12 seconds in FWHM. Data are flagged if the rms is more than 4 times the variance away from the mean averaged over all baselines, IFs, and polarizations. Flagging on the variance of the rms from the mean on a baseline basis is essentially turned off by the 10.

> GO  C R

to write out a calibrated, weighted data set for the first target source.

Then, when that finishes

> SOURCE target2’ , ’ ’ ; GO  C R

to do another target source.

It is not clear that this algorithm is optimal, but it certainly should be better than using all weights 1.0 throughout. It will be interesting to compare data weights found with TYAPL to those found with REWAY. The task now displays statistical information to assist the user in determining which weights are excessively high or low.