PowerPoint Presentation

Advanced MWA tile beam models Randall Wayth ICRAR/Curtin University

1MWA Primary beamBackground:Beam amplitude bootstrapping in MWA GLEAM Survey False Q seem by Emil in high freq, large zenith angle obsQ/I = (XX-YY)/(XX+YY)Cal solutions transferred from calibrator close to zenith.Current model of tile is analytic: = array factor * dipole pattern. No mutual coupling.Team: Adrian Sutinjo, John OSullivan, Emil Lenc, Shantanu Padhi, Tim Colegate, Budi Juswardy, RW

GLEAM backgroundMeridian drift scansNight-time observing8-10 hours of RA per campaignWeek-long campaigns7 DECS, 1 DEC per night, all freqs 80-230 MHz

Year 1:GLEAM 1.1:Aug 2013(RA 19 3 )GLEAM 1.2:Nov 2013(RA 0 8 )GLEAM 1.3,1.4: 2014-A(6-16, 12-22)

GLEAM 1: 2013/14+900 h6 h12 h18 h12 hGLEAM 1.1Aug 2013GLEAM 1.2Nov 2013GLEAM 1.3Feb 2014GLEAM 1.4May 20140-90

GLEAM 1: 2013/14+900 h6 h12 h18 hGLEAM 1.1Aug 2013GLEAM 1.2Nov 2013GLEAM 1.3Mar 2014GLEAM 1.4May 2014Complete!As at Dec 20130-90

MWA Primary beamsBackground: when calibration soln from 3C444 (DEC-17) are transferred to the DEC -27, -14, and 1.6 scans at 216MHzClearly the magnitude of XX and YY are off. (phase OK)

Fitting a simple primary beam modelBased on work @189MHz in Bernardi et al, 2013.Inter-port (mutual) coupling model

Known (=delays)Unknown (=dipole complex gain)Known (=LNA impedance, diagonal)Known (=impedance matrix, via sims)MWA LNA impedance

Example Z_tot matrix, 216MHz

NS-NS interactionsEW-EW interactionsNS-EW interactionsExample Z_tot matrix, 155MHz

Example Z_tot matrix, 118MHz

216 MHz: zenith vs ZA=14 degs

How about other freqs? 186 MHz

No gradient across tileModest gradient across tileHow about other freqs? 155 MHz

No gradient across tileNo gradient across tile216 MHz cuts through az=0 beams

Whats going on?Dipole is 74cm across = wavelength at ~200 MHzBelow this freq, short dipole approximation is increasingly validAlso, magnitude of coupling decreases with freq

Whats going on?The phase delay gradient forces side-to-side interactions on the X bow-ties as opposed to end-to-end interactions on the Y bow-ties.Such asymmetry does not occur when pointing the telescope in the diagonal plane as the interactions are symmetric with respect to the X and Y bow-ties.end-to-end vs side-to-side asymmetry is reduced with increasing wavelength.

Direction of increasing delay for az=0, za > 0Predicted False Q for simple model

The way forwardThree tier model:Analytic dipole model with impedance matrix from simulations or measurements(basically what has been presented in this talk)Average dipole response based on simulations with impedance matrix. One lookup pattern per freq.Relatively straightforwardIndividual dipole pattern for each dipole per pointing per frequency. ExpensiveHow well can we do with 2nd tier?

216 MHz, ZA=14 degBlue line: predicted false Q for simple model using incorrect (old) model for calibration

Red line: expected false Q for 2nd tier model using 3rd tier model as truth.(zero is good) How well can we do with 2nd tier?

155 MHz, ZA=14 degBlue line: predicted false Q for simple model using incorrect (old) model for calibration

Red line: expected false Q for 2nd tier model using 3rd tier model as truth.(zero is good) Tile: bottom lineMutual coupling does affect the tile beam, especially at higher freqs (>= 200 MHz)At high freqs, far from zenith along cardinal axes (0,90 degs az) mag and phase is quite differentCal solutions for amplitude are only valid at that pointing (this is not new, but model is)False Q due to transferring near-zenith amp cal to off-zenith meridian data.Mutual coupling also affects the relative response of dipoles, even at the zenith. This can taper the tile response, hence affect beam width and sidelobes.

Questions?

Optical H-alpha image: Luigi Fontana. http://www.astrobin.com/27523/B/

Optical H-alpha image: Luigi Fontana. http://www.astrobin.com/27523/B/Rosette NebulaOrion NebulaBarnards loopImpedance matrix: 32x32

Labels: 1-16 N-S Y dipoles17-32 E-W X dipoles200 MHz az=45o el=59.8o tile