1. Daniel Mitchell (SAO / CfA) For the MWA Consortium

2. The MWA Consortium MIT-Haystack MIT campus (MKI) Smithsonian Inst. Harvard U. U. Melbourne Curtin U. of Tech. Also contributions from individuals/groups at U. Wisc, U. Wash, CalTech U. WA U. Tasmania U. Sydney ANU Raman Research Inst. CSIRO

3. 32T Results (Field centered on Pic A) First 32 tiles: ~ 5% demonstrator Longer tracks in January 2010 RTS “First light” in December 2009

4. Outline Overview of the site Murchison Radio-astronomy Observatory (MRO) RFI Environment Overview of the signal path Antennas, receivers, correlator Overview of the processing pipeline Real-time imaging system

5. Photos: D. Oberoi & D. Mitchell

6. Boolardy Stn., Western Australia MRO Radio Quiet Zone 100 – 150 km radius restricted zone Remote. Limited access to site & data Power constraints Remote operation & monitoring Very low spectral occupancy (time & freq) Shire of Murchison Pop: ~110 in 41,173 km 2 only AU shire without a town

7. RFI Environment Redshifted HI: VHF band (80-300 MHz) Sources of RFI: Satellite and aircraft signals. Distant transmitters enhanced by reflections off meteor trails or aircraft. Distant transmitters enhanced by occasional refractive ducting. Local RFI. See Judd Bowman's presentation on Wednesday morning for an more on MRO RFI. Min/Max over a few days in 2007 (Bowman) ORBCOMM other satellites Terrestrial? Aviation,... Offset due to the galaxy FM radio RADCAL, DMSP F15,...

8. ORBCOMM 137–138 MHz circular. (~2.5 kHz) 26 LEO satellites (3x8 + 2 polar) 3 or 4 x ~3 hrs windows per day Useful for beam characterization Satellite trails and approx. power over 24 hrs

9. MWA Goals radiation gas-kinetic SOHO/EIT Pritchard & Loeb 2008 Testori et al. (2001, 2004) / Wooleben et al. (2005) Chatterjee & Murphy (adapted from Cordes et al. 2003) EoR via redshifted HI Solar & Heleospherical Galactic & Extragalactic Transients

10. MWA Specifications Lots of cheap antennas (512 antenna tiles) Good snapshot beam (>10 5 concentrated baselines) Full polarisation Wide frequency range (31 MHz from 80–300 MHz) Wide, steerable FOV (10-50 degrees) A few arc-minute resolution (1.5 km: 2.5'-8.5') Max BL ~ km (isoplanatic patch size): ~ 2D ionosphere Phase 1 RFI plan: go somewhere quiet and flag {

11. MWA Antenna Tile Beams Have sidelobe structure at -10 to -20 dB All sky Nulls Are time variable Dipoles point to zenith Not fixed to RA/Dec Little dynamic control Are tile dependent Different for each visibility Require a fix in gridding and during deconvolution Are relatively simple Small number of components No moving parts Have resistance to terrestrial RFI Ground plane creates null at horizon Tiles are low to the ground

12. MWA Signal Path

13. Simple median-based Flagging Current real-time RFI arsenal consists of a simple flagging routine to assist calibration. Will run at the start of the RTS, while integrating visibilities for 2 – 8 seconds. Flag based on ratio of median and interquartile range (proportional to SNR). Initial tests used auto-correlations and generated statistics for each frequency channel. cross-corr phase after flagging auto-corrs flags t f MWA memo, Wayth 2008

14. Calibration Solve for the location of, and gain towards, many point sources. Use the gain measurements (Jones matrices) to constrain a primary beam model for each antenna tile. Use the position offsets to constrain a 2D ionospheric phase screen. Use the ~ wavelength 2 dependency of the ionosphere to isolate the models. Update RFI flags

15. Imaging RTS: produce calibrated, regridded, weighted snapshot images 10 5 BLs, 10 3 freqs, 4 pol. products per 0.5 sec: O(10) GB/s too much data to store: real-time processing. 8 second cadence to correct O(1) min ionospheric changes. Deal with the ionosphere in the image plane. Deal with w-terms in the image plane (at the same time). Integrate processed images over 5 – 10 min. Formal self-calibration not possible, but will store all cal. info. forward modeling, matched filtering,... Need and have excellent instantaneous uv coverage.

16. Warped Snapshots (-3.5 hrs to +3.5 hrs) Co-pol (instrument pp)Cross-pol (instrument pq) Significant changes in (l',m') Significant changes in polarized gain MAPS simulation: HA (img center) = -3.5 hrs to +3.5 hrs 2 of 4 FFT snapshot images per freq per cadence All input unpolarized (diffuse & point sources) All polarization comes from the instrument

17. Warped Snapshots (-3.5 hrs to +3.5 hrs) The “true” grid is determined for each snapshot. The grids are further distorted to include estimated ionospheric displacements. Each pixel is weighted for optimal averaging: sky.gain 2 /sig 2 (weighting is done during gridding).

18. Regridded Snapshots (-3.5 hrs to +3.5 hrs) Each weighted snapshot is regridded to a static coordinate system (position & polarisation). Images are accumulated. Deconvolution needs to account for the antenna- dependent weights. Last chance flag RFI in R/T

19. 32T Results (Field centered on Pic A) First 32 tiles: ~ 5% demonstrator Longer tracks in January 2010 RTS “First light” in December 2009