AllegroWiki

User Tools

Site Tools

Trace:
allegrodrc:selfcal

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision		Previous revision
allegrodrc:selfcal [2020/11/27 15:10] – created intemaallegrodrc:selfcal [2021/01/12 12:58] (current) – intema
Line 1: Line 1:
 +<note>**#VIO: NEEDS UPDATING**</note>
 +
 ====== Self-Calibration ====== ====== Self-Calibration ======
  
Line 4: Line 6:
  
 ==== The Need for Selfcal ==== ==== The Need for Selfcal ====
- 
-> The gain calibration interpolates the phase from the ''%%phase_inf%%'' solution of the gain calibrator. Of course, this approach can only be seen as an approximation for the true phase of your science target. 
  
 In the classical calibration scheme, observations of your science target are always bracketed by observations of a gain calibrator. This calibration source will in most cases be an unresolved point source, for which we get uniform visibilities with a phase of zero (the amplitude is not important for the moment). This enables us to follow the varying phase with time with the aim of zeroing the phase for our observations. In the classical calibration scheme, observations of your science target are always bracketed by observations of a gain calibrator. This calibration source will in most cases be an unresolved point source, for which we get uniform visibilities with a phase of zero (the amplitude is not important for the moment). This enables us to follow the varying phase with time with the aim of zeroing the phase for our observations.
 +
 +{{ allegrodrc:selfcal0.png?400|The gain calibration interpolates the phase from the ''%%phase_inf%%'' solution of the gain calibrator. Of course, this approach can only be seen as an approximation for the true phase of your science target.}}
  
 Generally, the averaged phase offset of a entire scan (CASA: ''%%phase_inf%%'') will be linearly interpolated onto the science target scan from the two neighbouring calibrator scans. Obviously, this scheme is not perfect, as the atmosphere is not necessarily changing linearly with time. Moreover, the atmosphere towards your science target can be different from the atmosphere towards your gain calibrator. But for most purposes, the linear interpolation is a good estimate and allows to retrieve science ready datasets with typically only a minor degree of phase decoherence. Generally, the averaged phase offset of a entire scan (CASA: ''%%phase_inf%%'') will be linearly interpolated onto the science target scan from the two neighbouring calibrator scans. Obviously, this scheme is not perfect, as the atmosphere is not necessarily changing linearly with time. Moreover, the atmosphere towards your science target can be different from the atmosphere towards your gain calibrator. But for most purposes, the linear interpolation is a good estimate and allows to retrieve science ready datasets with typically only a minor degree of phase decoherence.
Line 17: Line 19:
 Prerequisites for self-calibration are a decent signal-to-noise ratio (typically around 20 or better) and a well-behaved source structure. The latter point is because must be able to get a good visibility model for your source structure. Very extended structure over the full field-of-view is definitely not suited for self-calibration, but point sources or moderately extended structures are definitely worth a try! Prerequisites for self-calibration are a decent signal-to-noise ratio (typically around 20 or better) and a well-behaved source structure. The latter point is because must be able to get a good visibility model for your source structure. Very extended structure over the full field-of-view is definitely not suited for self-calibration, but point sources or moderately extended structures are definitely worth a try!
  
-> The goal of self-calibration is to determine and correct the residual phase offsets of your science targets from nominal zero after the application of the phase solution from the gain calibrator.+{{ allegrodrc:selfcal1.png?400|The goal of self-calibration is to determine and correct the residual phase offsets of your science targets from nominal zero after the application of the phase solution from the gain calibrator.}}
  
-Self-calibration starts with a dataset that would be already science ready. So phase offsets between spectral windows or similar instrumental effects are already gone. If the interpolation of the gain calibrator phases would be perfect, all phases would be zero. However, since that is hardly the case, the goal of self-calibration is to correct the residual phase offsets that are still present, because...+Self-calibration starts with a dataset that would be already science ready. So phase offsets between spectral windows or similar instrumental effects are already gone. If the interpolation of the gain calibrator phases would be perfect, all phases would be zero. However, since that is hardly the case, the goal of self-calibration is to correct the residual phase offsets that are still present, because ...
  
-  - ...the atmosphere is changing most likely not changing in the same fashion as the interpolation scheme +  - ... the atmosphere is changing most likely not changing in the same fashion as the interpolation scheme 
-  - ...the atmosphere towards your science source can be different from the atmosphere towards your gain calibrator.+  - ... the atmosphere towards your science source can be different from the atmosphere towards your gain calibrator.
  
 With CASA, self-calibration is almost a no-brainer in terms of //how// to do it (but still you have to make sure that what you do makes sense in an astronomical sense). You have to have a model of your source in the ''%%model%%'' column of the measurement set. That is typically generated by a first clean of the calibrated data. With CASA, self-calibration is almost a no-brainer in terms of //how// to do it (but still you have to make sure that what you do makes sense in an astronomical sense). You have to have a model of your source in the ''%%model%%'' column of the measurement set. That is typically generated by a first clean of the calibrated data.
Line 28: Line 30:
 Then you iterate through the following steps to get a gradually better and better image quality: Then you iterate through the following steps to get a gradually better and better image quality:
  
-<HTML><blockquote> +  - ''%%gaincal%%'': Your gain calibration will find the gain correction factors $g_i$ for each antenna for the time interval specified by ''%%solint%%''. For antenna-based correction, each baseline has to be corrected by the gain factor $g_ig_j$ that consists of the individual antenna gains $g_i$ and $g_j$. This factor is given by $$g_i\,g_j = \frac{V^\mathrm{data}_{jj}}{V^\mathrm{model}_{ij}}$$ \\ How convenient of CASA that it already wrote the visibility model into the model column! CASA then finds the $N$ gain factors $g_i$ by fitting the visibilities of $N(N-1)/2$ baselines. 
-<HTML><ol></HTML> +  - ''%%applycal%%'': Applying the calibration table to the data will update the ''%%corrected%%'' column. Then you're ready for the next iteration. 
-<HTML><li></HTML><HTML><p></HTML>''%%gaincal%%'': Your gain calibration will find the gain correction factors $g_i$ for each antenna for the time interval specified by ''%%solint%%''. For antenna-based correction, each baseline has to be corrected by the gain factor $g_ig_j$ that consists of the individual antenna gains $g_i$ and $g_j$. This factor is given by<HTML></p></HTML> +  - ''%%clean%%'': Cleaning your image will automatically generate a visbility model of the structure that you have just cleaned and write it into the ''%%model%%'' column.
-<HTML><p></HTML>$$g_i\,g_j = \frac{V^\mathrm{data}_{jj}}{V^\mathrm{model}_{ij}}$$<HTML></p></HTML> +
-<HTML><p></HTML>How convenient of CASA that it already wrote the visibility model into the model column! CASA then finds the $N$ gain factors $g_i$ by fitting the visibilities of $N(N-1)/2$ baselines.<HTML></p></HTML><HTML></li></HTML> +
-<HTML><li></HTML><HTML><p></HTML>''%%applycal%%'': Applying the calibration table to the data will update the ''%%corrected%%'' column. Then you're ready for the next iteration.<HTML></p></HTML><HTML></li></HTML> +
-<HTML><li></HTML><HTML><p></HTML>''%%clean%%'': Cleaning your image will automatically generate a visbility model of the structure that you have just cleaned and write it into the ''%%model%%'' column.<HTML></p></HTML><HTML></li></HTML><HTML></ol></HTML> +
-</blockquote></HTML>+
  
 ===== Examples ===== ===== Examples =====
Line 41: Line 38:
 ==== Continuum self-calibration ==== ==== Continuum self-calibration ====
  
-<HTML><blockquote> 
   * **Dataset**: 2011.0.00629.S, two protoplanetary disks (DMTau, MWC480)   * **Dataset**: 2011.0.00629.S, two protoplanetary disks (DMTau, MWC480)
   * **Starting Point**: ''%%calibrated.ms%%'', the result of concatenating the two execution blocks ''%%uid___A002_X51d696_X315.ms.split.cal%%'' and ''%%uid___A002_X51d696_Xef.ms.split.cal%%''.   * **Starting Point**: ''%%calibrated.ms%%'', the result of concatenating the two execution blocks ''%%uid___A002_X51d696_X315.ms.split.cal%%'' and ''%%uid___A002_X51d696_Xef.ms.split.cal%%''.
   * **Spectral Setup**: 4 FDM spectral windows with 3840 channels each   * **Spectral Setup**: 4 FDM spectral windows with 3840 channels each
   * **Selfcal Target**: Continuum of MWC480 (''%%field='4'%%'')   * **Selfcal Target**: Continuum of MWC480 (''%%field='4'%%'')
-</blockquote></HTML> 
  
 === Preparing the continuum dataset === === Preparing the continuum dataset ===
Line 83: Line 78:
  
 === Starting Self-calibration === === Starting Self-calibration ===
- 
-> Continuum image of the original data 
  
 Self-calibration starts with a first clean, which generates the initial model : Self-calibration starts with a first clean, which generates the initial model :
 +
 +{{ :allegrodrc:mwc480_cont_selfcal0_image.png?400|Continuum image of the original data}}
  
 <code> <code>
Line 107: Line 102:
 === First Iteration === === First Iteration ===
  
-> Phase solutions for the first self-calibration step +{{ :allegrodrc:MWC480_cont_selfcal0.png?400|Phase solutions for the first self-calibration step}}
->  +
-> Continuum image of the original data (left) and after the first self-calibration (right). The two colour scales are aligned and highlight the improvement.+
  
 As next step follows the gain calibration, which compares the observed visibilities in the ''%%data%%'' column with the source model in the ''%%model%%'' column : As next step follows the gain calibration, which compares the observed visibilities in the ''%%data%%'' column with the source model in the ''%%model%%'' column :
Line 119: Line 112:
     solint = '120s')     solint = '120s')
 </code> </code>
 +
 In this first iteration we use a solution interval of 2 minutes. Since the gain calibrator is observed every 6 minutes, we get three solutions per scan. These can be displayed with : In this first iteration we use a solution interval of 2 minutes. Since the gain calibrator is observed every 6 minutes, we get three solutions per scan. These can be displayed with :
  
Line 130: Line 124:
     figfile = 'MWC480.cont.selfcal0.png')     figfile = 'MWC480.cont.selfcal0.png')
 </code> </code>
 +
 As you see, the solutions generally group around a phase of zero, but there are some significant correction factors to see - particularly in the last two scans. As you see, the solutions generally group around a phase of zero, but there are some significant correction factors to see - particularly in the last two scans.
  
Line 139: Line 134:
     flagbackup = F)     flagbackup = F)
 </code> </code>
 +
 The corrections are now written into the ''%%corrected%%'' column, which will in the next iteration be used for the inversion from the Fourier into the image plane. The corrections are now written into the ''%%corrected%%'' column, which will in the next iteration be used for the inversion from the Fourier into the image plane.
 +
 +{{ :allegrodrc:MWC480_cont_selfcal1_image.png?400|Continuum image of the original data (left) and after the first self-calibration (right). The two colour scales are aligned and highlight the improvement.}}
  
 We will now make a new image, where ''%%clean%%'' will use the data from the ''%%corrected%%'' column : We will now make a new image, where ''%%clean%%'' will use the data from the ''%%corrected%%'' column :
Line 159: Line 157:
     interactive = F)     interactive = F)
 </code> </code>
 +
 The improvement in image quality is obvious. The noise level goes down from around 5 mJy/beam to 0.9 mJy/beam. The improvement in image quality is obvious. The noise level goes down from around 5 mJy/beam to 0.9 mJy/beam.
  
 === Second iteration === === Second iteration ===
  
-> Phase solutions for the first self-calibration step +{{ :allegrodrc:MWC480_cont_selfcal1.png?400|Phase solutions for the first self-calibration step}}
->  +
-> Continuum image after the first iteration (left) and the second iteration (right). The two colour scales are aligned and highlight the improvement.+
  
 Again, we perform a gain calibration to find new correction factors. This time we use a solution interval of the integration time of a single data point ''%%solint='int'%%'', which in this case is 30 seconds (since we've averaged down to 30 seconds when we created this dataset). Moreover, we now apply the first calibration table ''%%MWC480.cont.selfcal0.pcal%%'' so that the newly created calibration table ''%%MWC480.cont.selfcal1.pcal%%'' only contains the additional corrections that are needed : Again, we perform a gain calibration to find new correction factors. This time we use a solution interval of the integration time of a single data point ''%%solint='int'%%'', which in this case is 30 seconds (since we've averaged down to 30 seconds when we created this dataset). Moreover, we now apply the first calibration table ''%%MWC480.cont.selfcal0.pcal%%'' so that the newly created calibration table ''%%MWC480.cont.selfcal1.pcal%%'' only contains the additional corrections that are needed :
Line 176: Line 173:
     solint = 'int')     solint = 'int')
 </code> </code>
 +
 The results are again displayed : The results are again displayed :
  
Line 187: Line 185:
     figfile = 'MWC480.cont.selfcal1.png')     figfile = 'MWC480.cont.selfcal1.png')
 </code> </code>
 +
 The corrections become less, but still they are up to around 50°. The corrections become less, but still they are up to around 50°.
  
Line 196: Line 195:
     flagbackup = F)     flagbackup = F)
 </code> </code>
 +
 +{{ :allegrodrc:MWC480_cont_selfcal2_image.png?400|Continuum image after the first iteration (left) and the second iteration (right). The two colour scales are aligned and highlight the improvement.}}
 +
 So far, we have only corrected phases. In the next step, we are also correcting for variations in the amplitude gain. Again, we clean the image with the visibilities with the improvements from the last iteration : So far, we have only corrected phases. In the next step, we are also correcting for variations in the amplitude gain. Again, we clean the image with the visibilities with the improvements from the last iteration :
  
Line 214: Line 216:
     interactive = F)     interactive = F)
 </code> </code>
 +
 Again, the improvement can be nicely seen. The noise goes down from 0.9 mJy/beam to around 0.5 mJy/beam. Again, the improvement can be nicely seen. The noise goes down from 0.9 mJy/beam to around 0.5 mJy/beam.
  
 === Third Iteration === === Third Iteration ===
  
-> Amplitude correction factors +{{ :allegrodrc:MWC480_cont_selfcal2_a.png?400|Amplitude correction factors}}
->  +
-> Continuum image after the second iteration (left, phase only) and the iteration (right). The two colour scales are aligned and highlight the improvement.+
  
 Now we perform an amplitude gain calibration, where we try to find the average amplitude gain per scan. We do not expect to see any phase variations anymore : Now we perform an amplitude gain calibration, where we try to find the average amplitude gain per scan. We do not expect to see any phase variations anymore :
Line 232: Line 233:
     solint = 'inf')     solint = 'inf')
 </code> </code>
 +
 We display the amplitude correction factors (we are not displaying phase correction terms here, as these are indeed zero) : We display the amplitude correction factors (we are not displaying phase correction terms here, as these are indeed zero) :
  
Line 242: Line 244:
     figfile = 'MWC480.cont.selfcal2.a.png')     figfile = 'MWC480.cont.selfcal2.a.png')
 </code> </code>
 +
 And again applying everything to get the corrected visibilities : And again applying everything to get the corrected visibilities :
  
Line 249: Line 252:
     flagbackup = F)     flagbackup = F)
 </code> </code>
 +
 +{{ :allegrodrc:MWC480_cont_selfcal3_image.png?400|Continuum image after the second iteration (left, phase only) and the iteration (right). The two colour scales are aligned and highlight the improvement.}}
 +
 Cleaning will show us again the improvement : Cleaning will show us again the improvement :
  
Line 267: Line 273:
     interactive = F)     interactive = F)
 </code> </code>
 +
 And indeed, the noise goes down from 0.5 mJy/beam to around 0.2 mJy/beam. And indeed, the noise goes down from 0.5 mJy/beam to around 0.2 mJy/beam.
  
 We started off with a noise level of around 5 mJy/beam and ended up at 0.2 mJy/beam, which is an improvement of around a factor of 25. We started off with a noise level of around 5 mJy/beam and ended up at 0.2 mJy/beam, which is an improvement of around a factor of 25.
  
allegrodrc/selfcal.1606486211.txt.gz · Last modified: by intema
Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki