Table of Contents
Dealing with low signal to noise in your bandpass and/or phase calibrator really just means doing extra averaging before finding bandpass/gaincal solutions. Here are some suggestions to improve the signal to noise in your observations, though at the cost of losing some information (fine changes in time/frequency/polarization).
The data reduction script generator has a flag that you should use for calibrating low signal to noise data sets. When running generateReducScript
, use the lowSNR
option. This will increase the fraction of the bandwidth used doing the gaincal pre-bandpass.
If your bandpass calibrator is too faint to get a good channel-to-channel solution for amplitude and phase, you have two options
- Try to use your phase calibrator
- In most cases, your bandpass calibrator is brighter than the phase calibrator, but more time is spent on source for the phase calibrator
- This is unlikely to work, but it's easy to test and worth the attempt
- Try using your bandpass calibrator, but average over (more) channels
- For instance, try solint='inf,128ch'
- For mixed mode data, it is preferable to specify the averaging in frequency units. Use the function: au.bandpassPreAverage('uid_XXX.ms'), which will compute the amount of pre-averaging to perform based on the source flux density, system temperature and time on source. - T. Hunter
To deal with low signal to noise in observations of the phase calibrator, we can average over some combination of time, frequency, and polarization.
- You may be able to get away with only changing the time averaging in gaincal
- Instead of using integration length solutions (solint='int'), use scan length solutions (solint='inf')
- If your phase calibrator is very faint, it may be necessary to combine spectral windows during 'p' solves in order to bring up its signal-to-noise enough to be usable.
- In the gain calibration step (usually #13 or 15), insert a new gaincal to derive a phase_offsets table containing the spectral window phase offsets using your (bright) bandpass calibrator scan with
- gaintype = 'G'
- calmode = 'p'
- solint = 'inf'
- gaintable = '(bandpass table)'
- Be sure to include this phase_offsets table anywhere that the spectral windows will be combined
- Note that the resulting values in the .fluxscale file for the narrow spws may be biased upward by a large factor (x2), but they are not subsequently used as long as you setup the later applycal commands correctly. - T. Hunter (4/2015)
- If the flux calibrator is also too weak in the narrow windows, then you will need to set refspwmap in the fluxscale command. - T. Hunter (4/2015)
- The spectral windows can be combined by in gaincal using
- The resultant solution will be placed in spectral window 0
- In future calls to gaincal and applycal, you will have to use spwmap to have these solutions applied to the other spectral windows (for example: spwmap=[,,[0,0,0,0]] when applying bandpass, phase_offsets, and a combined-spw-solution).
- In the applycal calls, you will also have to set gainfield appropriately (for example: gainfield=['','0','2','2'] for the bandpass, phase_offsets, phase_inf and flux_inf tables to the science target)
* It is essential to restore the BeforeGainCalibration flags if you have previously run applycal with solutions that were mostly flagged.
Therefore, it is good practice to insert a flagmanager(mode='restore') at the beginning of Step 15 before you starting editing the gaincal and applycal calls, so that you do not forget to do this restore.
- You can also try averaging together the two polarizations
- In gaincal, use gaintype='T' in your phase-only solutions (it is already used in amplitude-only solutions)
- The phase offset between the two polarizations must be removed, as above, with a call to gaincal on the bandpass calibrator