Recent imaging results from TKSP commissioning

Jess Broderick
LOFAR TKP Meeting, Amsterdam, June 2011

Transcript

1. Recent imaging results from TKSP commissioning Jess Broderick (University of Southampton) Martin Bell, Valeriu Tudose, Dan Calvelo, Dan Plant, John Swinbank, Jason Hessels, Evert Rol, Bart Scheers, James Miller-Jones, Rob Fender and the LOFAR Transients KSP
2. (J. McKean, Introduction to LOFAR, Leiden Busy Week May 2011) LOFAR imaging pipeline
3. casapy, awimager (J. McKean, Introduction to LOFAR, Leiden Busy Week May 2011) LOFAR imaging pipeline
4. (R.J. Van Weeren, Introduction to LOFAR data reduction, Leiden Busy Week May 2011) LOFAR imaging: current status
5. (R.J. Van Weeren, Introduction to LOFAR data reduction, Leiden Busy Week May 2011) LOFAR imaging: current status
6. (R.J. Van Weeren, Introduction to LOFAR data reduction, Leiden Busy Week May 2011) LOFAR imaging: current status • The casapy imager cannot make flux corrected, widefield images • If you solve for direction-dependent effects in BBS, you can only apply the corrections in a single direction. • Need to wait for the new LOFAR imager (awimager) to become available (in a month or two)
7. (R. Pizzo, Latest Imaging Results, Leiden Busy Week May 2011)
8. (R. Pizzo, Latest Imaging Results, Leiden Busy Week May 2011)
9. ASTRON press release, 'LOFAR makes deeper images of Universe than ever before', 1 June 2011 Part of an image centred on 3C196 HBA data processed by P. Labropoulos (EoR KSP) Not deconvolved Angular resolution 8 arcsec Dynamic range (full image) > 200,000:1
10. Transients KSP observing plan • PSR 0329+54 zenith field (HBA; Bell, Broderick) • SS 433 (HBA and LBA; Broderick) • Cygnus X-1 and Cygnus X-3 - within one HBA tile beam of one another, and so can be observed simultaneously with 24 MHz bandwidth each in high band (HBA; Tudose, Calvelo, Plant). • GRS 1915+05 • M31 (failed HBA observation from two weeks ago) • LSI +61 303 • In addition, we can search for transients in repeated LOFAR observations of other fields (e.g. 3C196) • We are obtaining new data every ~weekend at the moment. Timely data reduction and analysis is an issue – many of our raw datasets have already been removed from the LOFAR cluster (but most are stored in the long-term archive).
11. • Use the data from these fields to develop a transients imaging pipeline (trip). • Images are then passed on to the transients detection pipeline (trap; Evert's talk).
12. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. B0329 Pulsar Field (possible transients!) *11 x 12 hour observations. *3 x 6 hour observations. *150 MHz. *7 – 38 stations *25 deg2 FOV. *Cadence ~ weeks and months. *Noise ~ 15 mJy. *One Full Year of monitoring *Pulsar *Subtracted source
13. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. Station Locations / uv Plot
14. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. Progress. (Bell, PhD thesis)
15. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. Two month transient search Stacked Aug - Sep December
16. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. ILT J0320.3+5512 (Bell#1) Stacked LOFAR Summer Stacked LOFAR December
17. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. ILT J0320.3+5512 (Bell#1) Stacked LOFAR Summer Stacked LOFAR December WENSS (325 MHz)
18. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. ILT J0320.3+5512 (Bell#1) *Requested follow-up WSRT observations at 350 MHz and 1.4 GHz – not approved *Liverpool telescope observations in (Jan) reveal no new optical source or any source within the field that is significantly variable. *Weak detection of source in latest LOFAR observations
19. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. ILT J0331.5+5712 (Bell#2) WENSS (325 MHz) LOFAR December *Not detected in April 2010 observation. *Observed 'on' in all subsequent observations. *Triggered WSRT ToO.
20. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. ILT J0331.5+5712 (Bell#2) *Requested WSRT ToO observations at 350 MHz and 1.4 GHz. *Liverpool telescope observations reveal no new optical source or any source within the field that is significantly variable. *High time resolution data acquired with LOFAR and Lovell MkI (credit B. Stappers and J. Hessels). Still 'on' in Dec
21. April. 2010 MeerKAT Meeting The Transient Radio Sky and LOFAR. Martin Bell, Rob Fender and the TKP. Optical and Radio Follow-up. Courtesy D. Bersier Liverpool telescope (optical). WSRT 1.4 GHz Courtesy V. Tudose
22. Recent testing for the 0329 field • cimager vs casapy (Broderick, Bell) (casapy is much faster, but is it also more accurate?) • sky models - four sources vs many sources (Broderick) (recall that a good model is very important) • removal of remote stations to improve the calibration (Broderick) (VLSS/WENSS/NVSS models not ideal for the longest baselines) • station beam option enabled in BBS (Broderick) (directional gains and station beam must eventually be properly taken into account) • aoflagger vs MADflagger in NDPPP (Calvelo, Plant) (are we flagging out transients along with RFI?)
23. L2010_21641 (December) 240 SBs cimager SUBTRACTED DATA column robust=0 weighting ~2.9 mJy/beam rms resolution 100 arcsec x 90 arcsec 40 arcsec pixels Bright (~40 Jy in VLSS) source subtracted cimager vs casapy Bell #2
24. L2010_21641 (December) 240 SBs casapy imager (gridmode 'widefield') CORRECTED DATA column robust=0 weighting ~2.5 mJy/beam rms resolution 120 arcsec x 120 arcsec 40 arcsec pixels cimager vs casapy
25. cimager vs casapy
26. cimager vs casapy – what happened to Bell #2? • Bell #2 is not visible in the image made with casapy. • A persistent imaging artefact with the cimager would be very unusual, but this cannot be ruled out given that there are other obvious differences at the edges of the images. • Apart from the choice of imager, the only other differences in the analysis were (i) the data column imaged, and (ii) the angular resolution. Neither are significant in the disappearance of Bell #2. • Further tests are necessary with some of the other 0329 datasets.
29. New HBA data – different phase centre (close to Bell #2) L2011_24933 (April 3, 10:05- 16:05 UT) 240 SBs, 139-186 MHz 16 core stations (each with 2 HBA ears except one) and 5 Dutch remote stations casapy imager (widefield gridmode) uniform weighting; core stations only (max uvdist 3.75 km) ~2.5 mJy/beam rms resolution 140 arcsec x 100 arcsec 20 arcsec pixels 40 source sky model (all sources > 0.5 Jy in WENSS, assume spectral index of -0.8). Apply station beam model and correct in direction of source closest to phase centre Bell #1 not detected
30. Green = WENSS contours
31. Green = WENSS contours
32. Too much residual sidelobe contamination (even with uniform weighting) to tell if Bell #2 is on or not
33. Too much residual sidelobe contamination (even with uniform weighting) to tell if Bell #2 is on or not
34. Summary and future work – B0329 field • Image quality reasonably good despite the current calibration/imaging limitations and sky models derived using ~arcmin resolution catalogues. • Is Bell #2 a real source or some sort of artefact? Constrained CLEANing could help answer this question (have we put CLEAN components on a sidelobe?) • Perhaps reduce all or some of the previous 0329 datasets again with more updated calibration parsets (especially as initial tests of the flaggers by D. Calvelo and D. Plant suggest that we were flagging far too much data). • We need to do more comparisons between the outputs of the cimager and casapy, especially as the latter was only implemented in the pipeline ~1.5 months ago.
35. A LOFAR image of SS433 / W50 • 2 x 4 hr runs (2010 August 25 & 26). • UT range 18:23 – 22:23 in both cases. • 15 core stations (each with 2 HBA ears) + 4 Dutch remote stations. Nominally 561 baselines. • Frequency range 115-163 MHz; 248 subbands; 256 channels per subband. • Processed data compressed in time and frequency; integration time per data point 5 s and only 1 channel per subband (channel width 183.1 kHz). Dubner et al. 1998 – 20 cm
36. A LOFAR image of SS433 / W50 ~25 km
37. • L2010_09967 - averaged image from 246 sub-bands. Pixel size 60 arcsec, image 512 x 512 pixels. Robust weighting (r=0). Rms ~ 60 mJy/beam. Note: SS433/W50 not included in sky model (derived using VLSS and NVSS)
39. • L2010_09967 - averaged image from 243 sub-bands. Pixel size 40 arcsec, image 1024 x 1024 pixels. Robust weighting (r=1). Rms ~ 40 mJy/beam. Better model
40. New SS433 data – HBA and LBA • Two HBA datasets – 2 x 2 hr observations (April 10 and 11) • Remote stations flagged out in an attempt to improve calibration solutions • Sky model constructed using VLSS and a 74 MHz VLA map courtesy of James Miller-Jones. SS433/W50 still excluded from the model. Station beam model turned on. • New LBA dataset obtained in May; data reduction at very early stage only (testing demixing on a few subbands).
41. L2011_25130 HBA L2011_25130 (April 10, 08:15- 10:15 UT) SBs 0-121 (119- 143 MHz) El. 15-30 deg 19 core stations (each with 2 HBA ears except one) and 7 Dutch remote stations casapy imager (widefield gridmode) uniform weighting; core stations only (max uvdist 2 km) ~35 mJy/beam rms resolution 280 arcsec x 250 arcsec 20 arcsec pixels W44
42. L2011_25133 HBA L2011_25133 (April 11, 04:00- 06:00 UT) SBs 0-121 (119- 143 MHz) El. 39-42 deg (transit) 19 core stations (each with 2 HBA ears except one) and 7 Dutch remote stations casapy imager (widefield gridmode) uniform weighting; core stations only (max uvdist 3.75 km) ~30 mJy/beam rms resolution 350 arcsec x 135 arcsec 20 arcsec pixels Same sky model and calibration parsets SS433 peak flux about 1 Jy for both datasets
43. Cyg X-1 and Cyg X-3 data • Very hard to get decent images of these fields at the moment. • Nearby bright sources (especially Cygnus A) causing problems. Demixing doesn't work at the moment for any A-team sources within 10-15 deg of target. • Solving for directional gains can take many days.
44. Cyg X-1 and Cyg X-3 data • Very hard to get decent images of these fields at the moment. • Nearby bright sources (especially Cygnus A) causing problems. A-team demixing doesn't work at the moment for any A-team sources within 10-15 deg of target. • Solving for directional gains can take days. 1 SB (144 MHz), no calibration applied
45. Summary – SS433/W50, Cyg X-1 and Cyg X-3 • Progress has been slow for these datasets. • Likely problems with sky models – lots of extended emission in these fields. Model with e.g. PyBDSM or shapelets? CLEAN component models? • Demixing does not work at the moment for A-team sources within 10-15 degrees of target. This is a major issue for the Cyg X-1 / Cyg X-3 data. However, we should be able to use demixing for the SS433 LBA data (and initial tests suggest that it may work well).
46. Summary – TKSP imaging commissioning • Progress has been somewhat sluggish since December. Nonetheless, the images for the 0329 field are fairly promising. Hopefully we can take a big step forward when the new LOFAR imager becomes available. • A good sky model is essential. However, over-reliance on sky models must surely be a worry at some level (e.g. are some sources being artificially created in the final image?) • Correcting for directional-dependent effects can be very slow. This is not good news for ~real-time, widefield transient searches. The current limitation of having to pick a single direction in BBS for the relevant corrections is similarly a problem. • We need to start focusing more on how we are going to efficiently and accurately flag, calibrate and image small chunks of data. Evert Rol has written a modified version of the imaging pipeline to do this, and Martin Bell has conducted some initial tests.
47. Summary – TKSP imaging commissioning • Other things to test or implement in the pipeline in the coming months: combining subbands in uv plane prior to deconvolution, bandpass solution, ionospheric corrections, polarisation, self-calibration. • If you (or your student) would like to help with the commissioning, please volunteer!