Transcript Slide 1
Overview of Calibration and Dynamic Range Challenges Paul Alexander AAVP 2010 Dynamic Range Paul Alexander Required Dynamic Range • Sensitivity: 2000 m2/K at 150 MHz; 300 MHz BW; station beam ~ 1 degree • In 24 hrs integration s = 0.2 mJy • ~ 1 500 mJy source per sq degree at 150 MHz 2.5 × 106 • EoR signal ~ 10 mK in presence of ~1000K foreground. Image at > 10s 1 × 106 Take this estimate with a “pinch of salt” – limited by foreground subtraction Suggests 107 : 1 AAVP 2010 Dynamic Range Paul Alexander Required Dynamic Range • Note do need to think about source confusion • In 24 hrs integration s = 0.2 mJy • Source density ~ 1.5 x 105 sources per sq degree Required baseline ~ 100km at 450 MHz AAVP 2010 Dynamic Range Paul Alexander Achieving high dynamic range now What do we know we have to include in an analysis: Include Discussion Antenna-based complex gains Standard calibration and self calibration – iterative Removing sources in global sky model Removing bright sources from UV data even with local phase solution is relatively robust RFI and “bad data” excision Can be critically important: • still largely done by hand for GMRT, eVLA and LOFAR • Expert algorithms not well developed Bandpass calibration AAVP 2010 Well defined, but often more problematic than it should be – software limitation Dynamic Range Maturity ? Paul Alexander Achieving high dynamic range now Include Discussion Debugging the system We learn a great deal about our instruments over time and correct often 2nd order errors Position-dependent effects Hugely Important relatively recent advance • Time dependent pointing errors – antenna models may be limit • Position-dependent phase screen – critically important for the ionosphere – modelling? • Many algorithms (peeling. Aprojection ...) Full stokes imaging AAVP 2010 A position-dependent effect – polarization response changes across FoV Dynamic Range Maturity Paul Alexander Achieving high dynamic range now Other known issues • Algorithm approximations mean analysis has known problems and errors which are not necessarily well dealt with Wide-field imaging approximations (faceting, w-projection) Deconvolution errors and artefacts – still an art using human judgement to drive non-linear algorithms Time-averaging and bandwidth smearing poorly dealt with (but also useful in very wide fields). AAVP 2010 Dynamic Range Paul Alexander AA Pros and Cons • AA is operating at low frequency Ionosphere! • Physical stability (wind etc.) Good, study details • Unblocked aperture Inherent • Smaller beams are better >60m collectors • Narrow band is important AA is Wide Band but many channels • Calibration capability Excellent, by channel • Trade DR for sensitivity AA v. flexible AAVP 2010 Dynamic Range Paul Alexander Designing for dynamic range Stable, known antenna patterns are key • • AA advantages • AA’s mechanically stable • Unblocked aperture • Direct measurement of field But • Need to calibrate 105 elements per station • What accuracy of element calibration is needed • Model dependent calibration – how many parameters can we solve for? • Station beam is time dependent – transit experiment for individual elements • Multiple independent elements for AA-low AAVP 2010 Dynamic Range Paul Alexander Designing for dynamic range Element-level calibration options and issues • Importance of phase versus amplitude – how accurate? How often? • Where are the main errors introduced in the RF chain - • If copper is used for signal transport – active measurement of cable lengths? • Deployment issues – position, orientation, misalignment • If digitisation at the element – accuracy of clock distribution • Temperature variations – large ambient fluctuations – fibre better than copper? • Expert health monitoring system at element level – flag failed or failing elements • AAVP 2010 Noise injection? Dynamic Range Paul Alexander Pathfinders • The design decision must be informed by the pathfinders and precursors • SKA community mu go beyond – “waiting to see what we will learn” SKA team must pose the questions that we want to be answered • Get answers either from experience of the pathfinders or doing explicit experiments and measurements Obvious area of immediate cooperation between all the experiments and the AAVP team AAVP 2010 Dynamic Range Paul Alexander Other design issues • Sufficiently good ionospheric model • Station size and UV coverage – competing issues • Larger stations – smaller station beam easier ionospheric model? Lower cost & less processing • Smaller station size better – more stations better UV coverage, better imaging capability • Hierarchical beam former • Hierarchical beamformer in which data decimated reduced accuracy of station beam AAVP 2010 Dynamic Range Paul Alexander Algorithm issues (AA emphasis) • Maturity of approaches is not there yet • Much use of human intervention still required • Transitioning to totally automated pipelines will be a major challenge • Expert system for RFI excision? • Are wide-field imaging approaches sufficiently accurate? • Is our underpinning understanding of interferometry based too much on “experience” rather than a formal understanding of the underpinning formalism? • Relying very much on a tiny group of real experts who have both the “experience” and the formal analysis AAVP 2010 Dynamic Range Paul Alexander Magnitude of the task Grid science reduction and visualisation Science proposal Observation definition M&C database Global and local sky model Data excision Correlator Data excision Collectors Monitor and Control system Cloud store Data product distribution Calibration loop Imaging processor Science product archive Data Visibility routing processors Local science reduction AAVP 2010 Dynamic Range Paul Alexander SKA1 Data Rates and Configuration • AA Line experiment 50 AA-low stations • 100 sq degrees, 10000 channels over 380 MHz bandwidth 3.3 GS/s • Issues • What data rate can we process? • Trade UV coverage (Ns) for FoV and hence survey speed (W) • Line vs continuum requirements • What is the longest baseline • Single or multi-pass algorithms increase data rate and buffering AAVP 2010 Dynamic Range Paul Alexander Reducing the data rate • Relax criteria for dump rates and frequency resolution – My criteria based on uniqueness in UV plane – Can the criteria be relaxed and still achieve high dynamic range? • Dump times and frequency resolution baseline dependent Design correlator for worse case upgrade path AAVP 2010 Dynamic Range Paul Alexander AAVP 2010 Dynamic Range Paul Alexander