26/05/11, Zadar Shallow & deep integrations with the MWA Gianni Bernardi Harvard-Smithsonian Center for Astrophysics (cooperative effort with D. Mitchell, L. Greenhill, S. Ord, N. Shravan, R. Wayth & the whole MWA collaboration)
26/05/11, Zadar The bigger… the better!!!
26/05/11, Zadar MWA is (will be) the largest N- array (128 elements, 512?) 800m u,v plane filled inside 1.1km 150 MHz < k < 0.4 Mpc -1 filling suppresses artifacts in continuum fg subtraction dipole layout areal fillingantenna spacings 5x5m tiles
26/05/11, Zadar MWA is (will be) the largest N- array (128 elements, 512?) 800m u,v plane filled inside 1.1km 150 MHz < k < 0.4 Mpc -1 filling suppresses artifacts in continuum fg subtraction dipole layout areal fillingantenna spacings 5x5m tiles
26/05/11, Zadar The Real-Time calibration and imaging System (RTS) (Mitchell et al., 2008, IEEE, 2, 707) (Ord et al., 2010, PASP, ) (GB et al., 2011, MNRAS, 413, 411)
26/05/11, Zadar Deconvolution/source subtraction via forward modeling (or further development on image based deconvolution) (Bowman, Morales & Hewitt, ApJ, 695, 183) (Geil, Gaensler & Whyithe, 2010, 2010arXiv G) (Pindor et al., 2011, PASA, 28, 46) (GB et al., 2011, MNRAS, 413, 411) in the MWA case, visibility data are not stored – causing a possible limitation in the deconvolution accuracy; once the uv plane includes time and position dependent primary beam and ionospheric correction the synthesized beam is position dependent and there is no standard deconvolution method applicable (no Clean, no Cotton-Schwab method); it is a valuable method to evaluate the statistics of the residual visibilities (in the light of EoR detection);
26/05/11, Zadar Flow chart Select a subset of visible sources from image data Generate the FM (the synthesized beam) for each source using current best parameter estimate (position, flux) Simultaneous fit for all the source parameters through a non linear minimization Convergence? Subtract sources from sky model Are there unmodeled sources? No Yes Add to sky model Yes No Done
26/05/11, Zadar Deconvolution can actually be represented by matrix algebra: for M sources and N image pixels, the following system of linearized equations is solved at each iteration: get a new parameter estimate x i: 3M vector of parameter estimates Jacobian matrixweight matrix N vector of data points
26/05/11, Zadar A 512T application: initial image (101 sources + thermal noise) Peak: 87 Jy rms: 105 mJy/beam DNR (apparent) ~ 800
26/05/11, Zadar Non linear minimization for the 15 brightest sources (5 iterations)
26/05/11, Zadar Non linear minimization for the 50 brightest sources (5 iterations)
26/05/11, Zadar Non linear minimization for all the 101 sources (5 iterations) rms ~ 25 mJy/beam final DNR ~ 3400 (Source subtraction for the EoR: ~1200 sources down 1 mJy in a 20° FoV - MWA will be confusion limited before that)
26/05/11, Zadar 400m …5% prototype for MHz MWA 32 tiles (32T)
26/05/11, Zadar 32T Pipeline First Light 20°x 20° Zenith 103 MHz 134 MHz 180 MHz Observations from January x 7hr tracks with real-time calibration, peeling, imaging, resampling and averaging. HEALPIX projection of Pic A field. Pic A peeled to reveal many secondary sources. Several simplifying assumptions relative to the 512T version, but all of the pieces are in place.
26/05/11, Zadar Essential features of calibration and imaging are in place… What does it still need? sky models beam measurements/characterization imaging improvements
26/05/11, Zadar Sky survey: observing specs & strategy Coverage of the whole southern MHz MHz Meridian survey: 136 fields + 34 calibrators Each field is observed close to meridian (within a few minutes of transit) for ~5 min observations of similar HA -> similar primary beam for each scan -> similar synthesized beam and sidelobe structure MHz bandwidth & ~ 30’ angular resolution
26/05/11, Zadar Calibration and imaging calibration and imaging performed through the Real Time System running on CPU/GPU in an off-line mode; choice of a few calibrators per pointing direction to determine passband and gain solutions (both direction independent). Gains are computed over ~ 8 MHz bandwidth (after bandpass fitting). The converged solutions are transferred to fields where there is not enough signal-to noise-ratio for selfcalibration. A few calibrators are sufficient to cover the whole sky for δ > -70°: CenA ( MHz) PicA ( MHz) HydA ( MHz) HerA ( MHz) VirA ( MHz) TauA ( MHz)
26/05/11, Zadar Calibration and imaging (cont’d) each 8 sec-0.6 MHz snapshot is re-sampled into the Healpix frame and weighted by the (model) tile primary beam. The short baselines are down weighted using a Gaussian taper with σ = 15 u; snapshot images are co-added up to 5 min; mosaicking is performed in the traditional way, by combining the various pointings weighted by their primary beam (the tile beams are considered to be all the same); dirty images need (offline) deconvolution (both point sources and diffuse emission). We have been doing work in this direction (deconvolution through forward modeling);
26/05/11, Zadar Sky coverage
26/05/11, Zadar Selected areas (undeconvolved images)
26/05/11, Zadar
Zoom in (1) Galactic centre CenA CenB PKS
26/05/11, Zadar Zoom in (1.1) SNR: G GRS & GRS SgrA star
26/05/11, Zadar Zoom in (2) Vela + Puppis FornaxA PicA HydA
26/05/11, Zadar Zoom in (3) PKS
26/05/11, Zadar Moving up north… HydA
26/05/11, Zadar (cont’d) HerA 3C253 SNR W44
26/05/11, Zadar (cont’d) HerA VirA
26/05/11, Zadar Deconvolving real data: an example Source J is modeled in the same way that the pointing is processed via the RTS (beams, cadence, frequency) Convergence after 2 iterations. Positional error ~ 15’, flux error ~ 10%
26/05/11, Zadar Primary beam measurements the sky drifts overhead while the tiles point at zenith; ~30 bandwidth 185 MHz; snapshot images (one every 5 min) are used to measure the beam response towards the J (which is ~ MHz);
26/05/11, Zadar Primary beam measurements J
26/05/11, Zadar Fitting a simple primary beam model The beam is accurate at a 5% level
26/05/11, Zadar Extending the beam work: zooming in to HydA field HydA Observations span slightly more than 5 hours (total) over MHz: 21 tiles available HydA provides the direction independent calibration of the array Snapshot images co-added Multi-frequency synthesis (but in the image plane)
26/05/11, Zadar
Conclusions: o the MWA calibration and imaging pipeline has been tested. It is working well, it will be refined – significant choices will have to be made; o real science data will be ready to come out in the next months; o a first all sky survey (catalogue of point sources brighter than 10 Jy, images of the A-team source) will be available in the next months; o MWA will expand to 128 tiles in the next year, expanding its science targets; Thank you!