This is a general overview of tasks/topics for the LBG.
The main long-term goal is to enable the LOFAR system (including the pipeline software) to process long-baseline observations in a (semi-) automated fashion. This is a prerequisite to use long baselines in the KSPs.
To develop the required methods and test the instruments we observe a number of commissioning targets ranging from unresolved and simple to extended and with complicated structure. We start with small fields and later continue with wide-field imaging.
In order to analyse the commissioning observations, we have to work with preliminary solutions. These include parts of the pipeline, AIPS and difmap as well as own software and MeqTrees. This should be seen as a testbed for our calibration methods and not as the final production software.
Results from this work have already provided (and will continue to provide) invaluable input for general commissioning, because long baselines require thorough data inspection.
How do we want to participate in the milestone projects A-team, 3C sources and MSSS? We have to define strategies.
The long baselines can also be very useful to help with short-baseline projects. Examples are foreground removal for EoR, calibration of objects with compact components and polarisation calibration with pulsars. See the crab project.
There are a number of topics where close cooperation with other groups is required for the good of all. Differential Faraday rotation is an example. It has to be included in our calibration strategy and is also of great importance for the polarisation groups. Another topic is imaging.
Do we want a publication policy? I would say no. Membership in the LBG should not automatically imply any rights. Papers should be authored and written by those who do the work. People who contributed should be included. We have to obey the LOFAR publication policy in papers using LOFAR data.
Particular problems of long baselines are reduced signal (most sources are resolved), calibration (larger clock offsets, ionospheric delays, differential Faraday rotation) and imaging (sparse uv coverage, less reliable calibration).
The following is currently an unordered lists of topics we have to work on.
What are the current limitations of the pipeline for long baselines?
We believe that full fringe-fitting will be required for all but the brightest sources. The needs for LOFAR are quite different from other VLBI arrays so that standard fringe-fitting codes (e.g. in AIPS) are not sufficient.
This means we cannot determine phase solutions per subband because of the limited S/N. Instead we have to find a global solution for delays (dispersive and non-dispersive) and rates (of both delay types). Least-squares fits with standard starting values generally do not work, because the chi2 function has many local minima in the many-dimensional parameter space. Instead we have to start with a scan of at least some of the parameters.
The software of Olaf Wucknitz explores some possibilities but is not meant for full data calibration. Developing and implementing the required methods is the major task for the LBG.
As workaround for the moment, AIPS or patched versions of AIPS can be used. Can we modify FRING or KRING to work better with LOFAR data? Olaf Wucknitz is making experiments.
What is the potential of casa? It has not been explored for long-baseline data yet.
Most important is the problem of differential Faraday rotation. Optimally this should be determined and corrected as part of the general fringe-fitting. As preliminary workaround we can use a circular basis, but this relies on a good understanding of the dipole response (element beam).
Faraday rotation allows us to disentangle instrumental leakage terms and intrinsic polarisation.
To study the behaviour of the instrument and to explore possible calibration models, MeqTrees may be a good tool.
This is not a particular long-baseline issue, but we will suffer most from a reduced signal.
On the other hand the differential Faraday rotation on long baselines can be used to determine X-Y delay/phase offsets that are currently not corrected in the station calibration. These offsets affect the conversion to circular polarisation. In the final calibration with the full measurement equation, we can probably solve for these effects, but not yet.
How do we determine and correct the X-Y offset?
AFAIK there are no data weights determined for the data yet. This is unfortunate, because as a result of RFI and the variable dipole gains the sensitivity is a strong function of frequency. Not weighting the data properly is a waste of S/N. On the long baselines we cannot afford this.
The ad-hoc software of Olaf Wucknitz determines weights from the auto-correlations using simple noise statistics arguments. The weights are applied in the fringe-fitting process and also written to the uvfits file to be used in the calibration and imaging in AIPS and difmap.
Something similar should be part of the official pipeline.
For small fields we can calibrate the ionosphere in the normal VLBI-way. What to do for larger fields without bright sources in them?
The calibration of LOFAR data provides incredibly accurate information about the ionosphere. This should be used.
Contacts to the ionosphere groups and to ionospheric physicists are important.
We have to monitor the clocks to find if they behave according to their specifications (variations of ca. 20 nsec). Are the larger but constant offsets still there? If so, can we understand and remove them? If not, we have to correct the data. Can we do this with NDPPP?
This is an issue for long baselines, because the uv coverage is sparse compared to the Dutch array. Nobody else will test/improve the imagers for sparse coverages. On the other hand we have less problems with complicated extended structures, because they will generally be resolved out.
In the case of targets with small-scale structure, long baselines can help to calibrate and iamge the short baselines (e.g. crab).
Where can we process our data? The Groningen cluster may soon be overloaded with other projects. We have two clusters in Bonn (a big one at MPIfR and a small one at AIfA) that can be used for some projects, but not by the entire community. Jülich will probably be available, too.
The cluster at AIfA also has powerful GPUs. If anybody has the time to implement calibration algorithms for GPUs, please contact Olaf Wucknitz.
Enno Middelberg is currently installing the LOFAR software in Jülich. It is already installed on the AIfA cluster, but not tested yet.
The currently best option to transfer data is via one of the LTA (long-term archive) sites. This is currently tried for Jülich and from there to Bonn.
Should we try to establish an alternative correlator (e.g. DiFX in Jülich or Bonn)? This would also provide more flexible options for special experiments.
We have to keep track of our observations and data sets and keep those that will be needed as a reference. For the moment Jülich should provide sufficient space, but this has to be coordinated with ASTRON and everybody who keeps a local copy.
Can the coordination of long-baseline observations be improved? Currently it is unpredictable which stations can be used.
Software updates etc. are often delayed at international stations. How can this be improved?
Suggestion:
This belongs to the A-team but is special. The pulsar is bright and point-like on all baselines, maybe with the exception of extremely low frequencies where we may start to resolve the scattering disk. This makes it a perfect calibrator source for long baselines. At least in the high band the nebula is visible only on short baselines. The calibration can be tested by imaging the nebula using short baselines. This has already produced spectacular results.
The object is also of high relevance for the pulsars and polarisation groups. Olaf Wucknitz found linear polarisation at a level of 1% with the right rotation measure. Others have found higher levels.
It can be used as perfect calibrator for the Sun when it passes in June. At the same time we can study the corona and solar wind.
Short-baseline people are interested in help from the long baselines to calibrate and image this object. Do we see it on long baselines?
Should be interesting on long baselines.
Nice target to show that long baselines are working. Not perfect for commissioning, because of the beating from the two components.
Nice and compact, good for commissioning. Does not look exciting, though.
Nice and compact, good for commissioning. Does not look exciting, though.
Suggestion:
This project is pursued by Olaf Wucknitz together with Jason Hessels. The idea is to produce visibilities with high time-resolution on a small number of baselines by modifying the coherent station addition code used for pulsar observations. In this way the scattering disks can be studied as function of pulsar phase. The crab is a test object for these modes.