SKA Calibration and Imaging Workshop 2006 - Abstracts

Sanjay Bhatnagar - Parametrization of the Measurement Equation

Imaging with interferometric telescopes can be described as solving for the various unknowns parameters of the Measurement Equation (ME),namely those that model the sky brightness distribution, instrumental and atmospheric effects at the level appropriate for a given imaging dynamic range. Depending upon the required imaging dynamic range, various simplifying assumptions can be used to develop efficient imaging algorithms. Examples of simpler parametrization are the use of a pixel basis to model the sky brightness distribution, the assumption of identical and ideal primary beam patterns for all antennas, assumptions of direction independent atmosphere, etc. Errors due to such simplifications in most existing imaging and calibration algorithms are typically below or comparable to the sensitivity limits of existing telescopes. With increased sensitivity of the upcoming and future telescopes, these errors will limit the achievable imaging dynamic range, requiring more sophisticated parametrization of the ME.

In this talk, I will discuss the progress made so far in developing efficient algorithms to better model the sky brightness distribution as well as correct for some direction dependent effects.

Tim Cornwell - Recent progress on xNTD and NTD

Phased Array Feeds promise a way to expand the field of view of conventional parabolic antennas at low cost. This promise is recognized in the SKA Reference Design in which the high resolution/large field of view capability may come from PAF-enabled telescopes. The NTD program at the ATNF has been working towards a practical application of PAFs in interferometry. We have built a two element interferometer at the Marsfield site in Sydney, with the goal of making measurements of PAFs in the frequency range 0.7-1.8GHz. Results from this setup have been hard to obtain, not least because of the hostile RFI environment at Marsfield - digital TV being the most recent and troublesome source. Nevertheless, we so now have some first test results - interferometry on a satellite, OPTUS B, and on a number of radio sources, including M87. These measurements have been made using a Vivaldi PAF, provided by ASTRON. I will also describe the progress on a novel PAF based on a new approach, and recent work specifying the calibration and imaging pipelines.

Bill Cotton - Ionospheric Calibration and Wide Field Imaging at 1 Metre Wavelength

The effects of ionospheric phase errors on VLA observations at 1 meter wavelength (327 MHz) are examined and techniques developed at longer wavelengths are applied to model these ionospheric effects. The spatial dependence of the ionospheric calibration requires coupling of calibration and imaging. Several techniques are also examined to improve the fidelity and dynamic range of wide field imaging of this data. These include automatic setting of CLEAN windows and automated centering of bright sources. The problems addressed here will be more serious for SKA with longer baselines and higher sensitivity.

Ger de Bruyn - Full polarization wide field WSRT calibration/imaging at a million to one dynamic range

I will show recent WSRT images at frequencies ranging from 0.15 to 1.5 GHz on very bright sources affected by spatially dependent calibration errors. The impact of pointing variations, frequency dependent primary beams and more subtle errors is illustrated. Dynamic range is a strong function of distance to the pointing centre. Closure errors in the WSRT appear to be small, stable and removable. At 21cm a dynamic range of more than 1 million has been achieved on 3C147 and 3C84. The improvements resulting from multi-source selfcal, a.k.a. peeling, is illustrated.

Mattieu de Villiers (SKA SA / KAT), Marianna Ivashina (ASTRON), Wim van Capellen et al. - Phased focal plane array modelling

This presentation is of the recent FPA modeling work performed during a 3-month visit by Mattieu de Villiers to ASTRON, working with Marianna Ivashina and Wim van Capellen. The current status of the FPA simulation activity is addressed and a simulation system designed to assist in answering a few of the currently unanswered FPA design questions is described. Calculation of various efficiencies, temperatures and beam qualities is possible over a wide range of design parameters.

Mike Garrett (JIVE), Emil Lenc (Swinburne), Seungyoup Chi (Kapteyn Institute/JIVE), Olaf Wucknitz (JIVE), Steve Tingay (Swinburne) et al. - New Scientific results from wide-field VLBI

We present some new results that apply wide-field techniques to Very Long Baseline (VLBI) Data. In particular, we present deep Global VLBI 1.4 GHz imaging of the Hubble Deep Field North (HDF-N) that reach a noise level of 7 microJy per beam in the central part of the field. We report one new detection in the central region and several new detections in the adjacent Hubble Flanking Fields. These sources are very likely powered by AGN processes. In addition, we report on the first unbiased low frequnecy (327 MHz), deep wide-field VLBI survey of the radio sky that covers an area of ~ 1 square deg. Around 10% of the target sources are detected at a noise level of ~ 1 mJy per beam. The results have implications for the extension of low frequency instruments to very long baselines (e.g. the extended LOFAR project). The calibration techniques employed, and details of the high resolution correlation capabilities of the EVN correlator at JIVE will also be presented.

Kumar Golap - Implications of brute force solution to problems facing long wavelength long baselines interferometers

In this discussion we review briefly the challenges of calibrating and imaging with long baselines at low frequencies. We then look at VLBI style fringe-fitting and discuss its applicability and implications to low frequency, full beam imaging.

Gerry R. Harp (SETI Institute) - Wide-field imaging studies at the ATA

As a large-N, small-D array, the Allen Telescope Array is an excellent testing ground for calibration and imaging algorithms proposed for the SKA. The ATA field of view is 2 degrees at 1.4 GHz. This permits, for example, interferometric imaging of most of the Andromeda galaxy with a single pointing. The wide FOV and fast tracking also permits high speed mapping of the entire sky in single dish mode. In this presentation we review the current status of the ATA with a special focus on calibrating the primary beam shape of a small dish.

Brian D. Jeffs, Karl F. Warnick and James Nagel - On Interference Cancellation with a Focal Plane Array

One of the promising aspects of using array feeds in radio telescopes is that off-axis interfering sources can be adaptively nulled while observing a space object of interest. We have constructed a seven element focal plane array for use on a three meter dish to study array gain, aperture efficiency, sensitivity, and interference attenuation levels in a controlled experimental environment. Successful cancellation was demonstrated to attenuate a moving broadband interferer which had previously completely masked the signal of interest. Beam pattern rumble due to interferer motion was shown to increase the effective noise floor level.

Adaptive beamformers are known to suffer from beampattern distortion, particularly when interferers approach the nominal mainlobe. We consider the issue of computing power spectral density (PSD) estimates at the output of a beamforming array in the presence of strong moving interference. A bias corrected PSD estimation algorithm is introduced which uses subspace projection methods to cancel interference at the beamformer. A correction step eliminates both pattern-distortion-induced (pattern rumble) PSD bias and spatial response errors over the long-term PSD averaging window.

Richard Lord - KAT RFI Environment and Preliminary Mitigation Strategies and Results

Although the effects of RFI are potentially disastrous to radio astronomy, great faith is placed on the radio-quiet qualities of the chosen KAT site, known as "K3". The RFI measurement data of the K3 site (and others) was analysed extensively in order to arrive at an estimate of the required ADC dynamic range, i.e. the number of bits required to adequately sample the received waveform without saturation or distortion. Indeed, measurements done to date indicate that for 90% of the time, even 1-bit quantisation should be sufficient. If one incorporates the inherent RFI attenuation characteristics of the interferometer, it would seem that a lot of science should be possible, even without any (or just very rudimentary) RFI mitigation strategies in place. This talk will describe the RFI environment and preliminary RFI mitigation strategies that apply to KAT. Furthermore, initial results obtained from injecting RFI into a voltage-based simulator will be presented.

Daniel Mitchell - Direction Dependent Beam Calibration for the MWA-LFD

Each of the 500 MWA-LFD tiles is a 4x4 array of crossed dipoles. We are investigating the effectiveness of a parameterization scheme that uses properties of each tile's dipoles to characterize its beam shape. I will describe our current work on the calibration of the tile beams, and how we intend to deal with the additional effects of the ionosphere. I will also present some simulations of the technique and discuss how we intend to parallelize the processing.

Ronald Nijboer - UVBricks: Predicting Patches of Sky and applying Image Plane Effects

For the new generation of Radio telescopes like LOFAR and SKA we need to develop new ways of reducing our data. On the one hand because we need to be able to cope with data sets of ever increasing size, on the other hand because the data needs to be corrected with ever increasing accuracy. UVBricks may be a way to efficiently predict patches of sky containing extended sources and / or groups of point sources, while applying image plane effects.

A UVBrick is a gridded cube of uv-data having dimensions u, v, frequency, and optionally polarization. It is obtained by applying an FFT to a similarly gridded cube of Image (or rather Brightness) data. In order to predict Visibility data the UVBrick needs to be interpolated.

Image plane effects are described by a multiplication of the Sky Image with some function. After Fourier transformation, this yields a convolution in the uv-domain. Since gridded convolutions can be calculated in the same way as an interpolation, the two operations can be efficiently combined. This would then yield a general way to apply image plane effects like w-term correction, antenna pointing errors, and ionospheric errors.

Jan Noordam - LOFAR Calibration Overview

no abstract

1. N. Pandey, N. Udaya Shankar (Raman Research Institute, Bangalore, India) - Hierarchical RFI Mitigation System at the Mauritius Radio Telescope:: We present a hierarchical post correlation RFI mitigation system for a large volume of visibilities at low frequencies and resulting RFI statistics. Interference detection is carried out in visibility and image domains using a conjunction of signal processing techniques : Thresholding, Fourier filtering followed by a CLEAN like operation, Hampel filtering, Model fitting, Decision based algorithm, using RFI statistics and information that sky signal is correlated day to day but interference is most likely not. Its successful application on ≈20,000 hours of astronomical observations for the survey at 151.5 MHz with the Mauritius Radio Telescope show ≈99.7% automatic RFI detection, while the remaining is detected by semi-automatic methods. Such an approach of applying linear and nonlinear techniques judiciously at various stages of processing is likely to play an mportant role towards achieving the ultimate goal of automatic data flagging, an inevitable requirement for processing data produced by large next generation telescopes.

1. N. Pandey, N. Udaya Shankar (Raman Research Institute, Bangalore, India) - Automatic Evaluation of Quality of Astronomical Data from Synthesis Telescopes:: We present a novel framework to automatically classify visibility data depending on its quality and usefulness for imaging. The suitability of a data file for imaging is expressed as a function of key parameters. To a visibility file, for each key parameter, a numerical index called the Quality Factor (QF) is assigned. QF assesses the extent up to which the observed visibilities satisfy the expected requirements for the parameter. Thresholding is applied for files which rank poorly in certain QFs, but are within acceptable quality limits based on other QFs. Its successful application to automatically classify ≈20,000 hours of astronomical observations for survey at 151.5 MHz with the Mauritius Radio Telescope, shows an excellent agreement of ≈92% with results of traditional approach based on human expertise. With suitable modifications, an astronomical tool based on such framework can potentially be employed on data sets from other interferometric arrays and also to monitor the performance of an observatory itself.

Rob Reid - Automatic Model Construction in smerf

Getting your model right is a common goal, but your model will receive extra scrutiny if you are claiming resolution sharper than the Rayleigh limit. Smear fitting happens to be a deconvolution method that routinely produces such sharp images by modelling with dozens of elliptical gaussians. To allow its users to avoid the criticism that smear fitting simply enshrines the models they give it, the amount of user interaction must be minimized. smerf, an implementation of smear fitting, comes with modcons, a script to automatically produce models using as many elliptical gaussians as it sees fit. I will discuss my experience in writing modcons, and the heuristics that went into it, in a way that hopefully will also apply to model construction outside the context of smear fitting.

Rob Reid and Tony Willis - Correcting wide field image errors with complex primary beams

The primary beam of an interferometer is the product of the voltage pattern of one antenna and the complex conjugate of the other antenna's voltage pattern. If all the antennas in an array have identical voltage patterns, then there will only be one primary beam, and it will be real. In less than perfect arrays, however, different baselines have different primary beams, and the discrepancies in amplitude and phase grow worse with increasing distance from the beam center. We present an alternative to peeling that uses measured voltage patterns to correct image artifacts in the Canadian Galactic Plane Survey.

Bob Sault - Imaging and calibration of the MWA-LFD

The Miluera Low Frequency Demonstrator is a 80-300 MHz interferometer array of 500 tiles. The telescope is currently being planned and built, ready for deployment in Mileura starting in late 2007. A realtime pipeline will calibrate and image the correlation data, whereas analysis will be mostly off-line. This talk will give a description of the science requirements for the imaging and calibration system, along with a sketch of the system that we are building to achieve this.

Oleg Smirnov (ASTRON), Cormac Reynolds (JIVE) (requests 30 mins) - An Overview Of SKA Simulations Using MeqTrees?:::

The MeqTree system is a "simulation and calibration" toolkit being developed by ASTRON and outside collaborators. It provides a number of unique simulation capabilities which are especially critical to the next generation of radio telescopes. The recent MeqTree Workshop hosted by ASTRON/JIVE/SKADS produced a number of interesting simulations and thought-provoking images. We will give an overview of MeqTree capabilities, and present some real simulations, including moving sources, transient sources, antenna-dependent primary beams, effects of pointing errors, simulated ionospheric disturbances and their impact on LOFAR patch imaging, and other interesting image-plane effects.

Rau V. Urvashi and T.J. Cornwell - Multi-scale multi-frequency synthesis imaging

Multi-scale techniques are required in multi-frequency synthesis imaging to accurately deconvolve extended emission whose flux varies across the observing band. The full continuum sensitivity provided by new broadband reciever systems can then be taken advantage of. We will describe an extension of the Conway/Sault MF-CLEAN algorithm using the match filtering technique of the Cornwell/Holdaway MS-CLEAN, and demonstrate its application to the imaging of the LMC at 20cm. We will also discuss the inherent ambiguity between scale size and source spectrum arising during imaging, due to the mapping of visibilities measured at different temporal frequencies onto different spatial frequencies.

Wim van Capellen (presented by someone from ASTRON) - FPA's for APERTIF

no abstract

Sebastiaan van der Tol - Ionospheric calibration from an array signal processing perspective

no abstract

Maxim Voronkov - Calibration and simulation strategy for multi-feed interferometers

Interferometers with multi-feed receivers like the xNTD and KAT pose new challenges for the calibration procedure. The bandpass and gain calibration has to provide a separate solution for each feed. Although it can be done by observing a secondary calibrator in a usual way, the overheads of this approach are prohibitive. A full beam self-calibration may help, but it requires the bandpass shape to be stable at the time-scale of days and the software to cope with a large fractional bandwidth. The feed legs effect is able to put a significant stress on the data reduction software if the azimuthal telescope mount is used. The influence of this effect is less for a better uv-coverage and may be even tolerated for uv-coverages, the xNTD and KAT are expected to provide. The pointing errors are likely to be a minor issue as they can be solved for and corrected on-line. In addition, their effect on the image quality can also be reduced by improving the uv-coverage. We discuss these effects in detail and comment on the other problems like the full polarization imaging, cross-coupling of the phased array feeds and the beamforming. We also outline the role which can be played by computer simulations in the future study.

Stefan Wijnholds - Reducing the impact of station level spatial filtering limitations

Station beam side lobes are highly sensitive to station calibration errors. This makes the attenuation experienced by side lobe sources effectively unknown preventing us from subtracting these sources by using prior knowledge from catalogs. This increases the number of parameters in the calibration and imaging routines considerably which makes these routines more demanding or even intractable. In this study we investigate this problem using LOFAR as a practical example and propose and study several techniques to reduce the impact of these station level spatial filtering limitations.

Sarod Yatawatta - Subspace techniques for astronomical data enhancement

Astronomical data have rich diversity in terms of frequency, observation time, array configuration etc. This diversity can be exploited in various ways to enhance image quality, the most obvious being averaging. We propose the application of subspace decomposition for exploiting this diversity. We present results of application of subspace decomposition in separation of artifacts and noise from astronomical images. We also investigate the possibility of extending this technique for source subtraction in the uv plane.

Sarod Yatawatta (Groningen/ASTRON) - Self-calibration using Shapelets and Friends

Shapelets have been in existence as an invaluable tool in astronomical image analysis. We investigate their usefulness in modeling complicated source structure in the self calibration of radio data. We also present how shapelets are incorporated into MeqTrees?. In a calibration pipeline, it is important to build the Sky Model automatically using clean components. We present ways in which we could automatically group clean components and model them as Shapelets or its friends i.e., Gaussians, Point Sources or even Images.