Date: Tue, 10 Sep 2002 22:11:52 -0400 From: Ed Fomalont Reply-To: aips2-naug@donar.cv.nrao.edu To: aips2-naug@parallax.aoc.nrao.edu, aips2-naug@donar.cv.nrao.edu Subject: [Aips2-naug] AIPS++ VLBI write-up Hi all, Below is our suggestions of the VLBI tasks that are needed in AIPS++. After writing this we found a 1997 version of VLBI in AIPS++ which covers alot of this ground. After talking with Joe, he thought that revising this memo to include the following would be useful. 1. Additional VLBI tasks not covered in the earlier memo five years ago. VLBI has advanced considerably since then. 2. Priorities over next six months to a year so that AIPS++ can be used for standard VLBI observations as soon as possible. That seems like a good idea and we can discuss this at the meeting. Cheers, Ed PS This memo is long and rough and will be drastically cut. I wouldn't read every word in it! Read the very last part which is a suggested list of three things to implement to start AIPS++ VLBI processing. ------------------------------------------------------------------------ VLBI Processing in AIPS++ draft Sep 10, 2002 Fomalont, Brisken, Claussen VLBI processing in AIPS++ will soon be tackled in full force. In order to produce a relatively complete and robust reduction package, a summary of the needed VLBI tasks are given below. Some are already implemented in AIPS++ and some tasks are more important than others. SUMMARY of VLBI Tasks: 1. Read in correlator u-v data: A. Supporting data needed: Correlator model (geometry, clock, weather model) Antenna parameters EOP parameters (polar motion, rotation rate, etc) Antenna gain curves System temperatures Meteorological data Orbit information for spacecraft Phase-cal tone information on-line flags subarray information B. Some plotting and editing of above data needed C. Data concatenation and data splitting into parts needed 2. Apply apriori corrections: Apply antenna gain, system temperature, phase-cal tones and flags. Apply auto-correlation amplitude corrections Apply external ionospheric model Apply parallactic angle corrections Apply proper weighting of data if not already done 3. Apply other corrections: Ability to update parameters such as: source position, antenna position, EOP parameters, better weather model using meteorological data, clock offset. Use of GSFC Calc algorithms useful. 4. Flagging of data: Need several plotting/editing tools. 5. Calibrations: A. Fringe-fitting. Extremely important program. Must have various options of combining frequencies and execute relatively quickly. Detection of weak sources is important, needed good algorithms for global detection. Methods to determine the coherence time for the fits would be useful. B. Solution table editing. The VLBI solutions often contain spurious solutions. Flexible editing of incorrect solutions are needed. C. Phase referencing connection. The interpolation of the VLBI solutions between calibrator solutions must be done carefully in order to keep phase coherence during the interval. This requires sophisticated smoothing and interpolation schemes for the phases. D. Multi-band delay. Finding the linear phase slope between all frequencies. Can be done with fringe-fitting for better sensitivity, or as a stand alone program. E. Incoherent fringe-fitting could be useful. F. Bandpass Calibration on strong sources. G. Calibration. Further time calibration with short integration times needed. Flexible combination of frequencies and polarizations useful. Improving amplitude calibration with certain constraints. H. Interpolation. Several interpolation methods between solution times, especially important for phase referencing. Multi-source phase referencing. I. Polarization. True polarization calibration / self-calibration technique needed. Variable Faraday rotation problem. 6. Imaging/processing programs: A. Imaging and Self-calibration. The bread and butter of VLBI imaging. Difmap is a good system to emulate. B. Fringe-rate mapping. Determining position of source using the observed residual rates C. Delay mapping. Determining position of source using the observed residual delays. Could combine rate and delay mapping D. Mini Calc/Solve capabilities. Using rates, delays (and maybe phases) to determine improved source positions, antenna positions, clock correction, better tropospheric model, earth rotation parameters, (many more - see astrometry below) E. Model fitting in the u-v plane. Useful for small-diameter sources where one can determine accurate source parameters. Could use used in conjunction with deconvolution using the point-spread function. E. Closure phase analysis would follow from model-fitting. 7. Astrometric Analysis: Interface with the Goddard Calc/Solve software. To duplicate this software in AIPS++ is a big effort. Needed for interfacing are: Apriori parameters such as Correlator model, antenna parameters, Earth Orientation parameters, phase-cal tones, subarray information, ionospheric models Fringe-fitting and multi-band delay solutions, preferred one baseline at a time. Conversion of earth-centered solutions to baseline-based solutions. Make output binary-style format for reading into Calc/solve. Assessment of the VLBI processing in AIPS++ (September 10, 2002) 1. Read in correlator u-v data: The basic program to read in the u-v data has been written. Much work is still to be done to incorporate the many secondary data needed to successfully image and reduce VLBI data. Plotting and editing of the u-v data is similar to VLA data and is in good shape. Some editing and plotting capabilities of the secondary data are need, but not of great important. Data concatenation and the splitting of data sets are available, but probably need some improvement. 2. The application of the apriori information into the data base is coming along. What is missing is the application of gains, system temperatures and phase-cal tones. 3. There is no ability yet in AIPS++ to update the data for corrections. These are also needed for VLA data analysis and is being worked on. 4. The flagging of data using graphical displays is in good shape. Some additional flexibility might be needed for VLBI data. 5. Calibrations: A. The VLBI calibrations are similar to the VLA calibrations and is well along the way. Most important is fringe-fitting, and a first pass program has been written. Fringe fitting, especially global fringe fitting to reach the best detection sensitivity, is a difficult task and will require lots of testing and experimentation. The AIPS fringe-fitters are not particularly satisfactory. The fringe-fitter used at Haystack (FOURFIT) is a good system to emulate. B. Often, there are obviously bad fring solutions over a long run, especially when trying to detect weak sources. Some editing capabilities are needed to remove the bad solutions, with the choice of flagging the associated data or interpolating from good solutions. This editing can be useful for VLA data as well. C. The smoothing and interpolation of fring results, particularly the phase, is a critical part of phase referencing. With relatively weak calibrators, large phase rates and large non-linear phase changes, different smoothing and interpolation techniques are needed. D. Multiband delay solutions are required for astrometric use and for good detection sensitivity. This should be provided in a basic fringe fitter. Careful attention to the execution speed is necessary since this algorithm can be very time consuming. E. Non-coherent fringe-fitting is not important at this stage. F. The bandpass algorithm used for VLA reduction is fine. G. The calibration package used for the VLA reduction is fine. But, VLBI reductions require strict scan-based calibration since the phases vary quickly with time and often there is missing data for some antennas during a scan. Amplitude only calibrations also need work. H. The interpolation of the fringe and calibration solutions in time are sometimes tricky because of the possibility of lobe ambiguities. This aspect of AIPS++, also for VLA data, needs more flexibility. I. AIPS++ has the potential for robust polarization calibration of VLBI data because of the use of proper full-polarization matrices and the similar treatment to all calibrations. 6. Processing programs A. The imaging and self-calibration algorithms are similar to those for the VLA and are in good shape. Some work can be done on iterative imaging, calibration, editing of solutions and data with Difmap as a good system to emulate. B. and C. Fringe-rate and delay mapping would be useful and not difficult to program. D. Some capability to determine better antenna positions, source positions and weather models would be very useful. None of this is available in AIPS. E and F. Model fitting in the u-v plane is very useful for VLBI sources. At present, nothing is available in AIPS. Closure phase analysis can also be included in model fitting. 7. Astrometric Analysis. In order to interface with the GSFC Calc/solve system, the following necessary tasks listed above are needed. Most have been included in the above discussion. The conversion to baseline-based parameters is not difficult. Some astrometric analysis, none of which is available in AIPS, would be useful for both VLBI and VLA data. These would include fitting the phases/delays/rates from fringe fitting to models which contain 1. Antenna position errors 2. Source position offsets 3. Tropospheric zenith path delay 4. Clock correction This would cover most of the ground. Other terms like EOP, nutation, etc will let GSFC do. ------------------------------------------------------------------------ Most important software needed for a beginning VLBA package. 1. Finish loading in VLBI data. Incorporate secondary parameters. 2. Finish fringe-fitter. Needs lots of testing for robustness and speed of execution. 3. Improve interpolation schemes when applying scan-based calibrator solutions to data, i.e. phase referencing. _______________________________________________ Aips2-naug mailing list Aips2-naug@listmgr.cv.nrao.edu http://listmgr.cv.nrao.edu/mailman/listinfo/aips2-naug