Widefield, high sensitivity VLBI

1
Wide-field, high sensitivity VLBI

• surveying and astrometry with mas resolution

• Adam Deller

• VLBA Astrometry Symposium
• July 2009

2
High resolution interferometry
20 mas

• Traditionally, narrow fields for studying single
  compact objects (pulsars, AGN, masers)
• Astrometry is the current killer app
• The VLBA is the currently the premier instrument
  for
• precision VLBI astrometry

A typical VLBI image
3
VLBA developments

• Factor of 4 increase in continuum sensitivity
  through the bandwidth upgrade to 4 Gbps
• Allows fainter astrometry targets
• Additional benefit for the use of fainter, more
  nearby in-beam calibrators (and hence better
  astrometry)

Pradel et al. 2006
3
4
VLBA astrometric capabilities

• 4 minute baseline sensitivity _at_ 4Gbps is 0.7 mJy,
  summing all bandwidth
• Thus calibrators as faint as 5 mJy can be used as
  in-beam calibrators - and brighter calibrators
  can solve for even shorter term
  atmospheric/ionospheric variability

4
5
Finding in-beam calibrators

• No comprehensive catalogue of the radio sky at
  high resolution exists (reasons later)
• The nearest equivalent is the geodetic source
  list maintained at astrogeo.org, with 4000
  sources (bright enough for use as primary
  calibrators, but density lt1/sq. deg.)
• Thus every astrometry project must typically find
  in-beam calibrators with a
• dedicated survey

5
6
Finding in-beam calibrators

• This typically involved selecting candidates from
  low resolution VLA surveys, testing compactness
  with higher resolution/frequency VLA
  observations, and finally VLBA follow-up Tedious
  slow!

Question The VLA and VLBA primary beams are the
same size so why are VLBA observations
sotime-consuming that a VLA pre-filter is
required?
6
7
Why no VLBI surveying?

• Resolution is a curse imaging the full VLBA
  primary beam (0.25 sq. deg. _at_ 1.6 GHz) with 2x2
  mas pixels (synthesized beam 10 mas) requires a
  600 Gpixel image 2.4 TB, which is almost
  entirely noise!!
• Plus the correlated data for8 hours _at_ 4 Gbps
  totals60 TB - infeasible

7
8
Directed surveys

• Forming small images around multiple fields of
  interest is possible, however
• Requires a uv shift to be performed, correcting
  the antenna-based delay difference for each
  desired phase centre
• Can be done post-correlation, but the
  intermediate data volume is tremendous - 60 TB/8
  hour VLBA track, as with imaging the full
  field (time/freq. resolution)

8
9
Directed surveys

• Such post-correlation shifting has been used to
  test wide-field VLBI imaging (e.g. Lenc et al.,
  Middelberg et al.)
• However, the most efficient implementation
  (minimizing I/O) is within the correlator, before
  data must be written to disk
• Such a capability is in the final stages of being
  tested in DiFX, the software correlator
  integral to the upgraded VLBA

9
10
Multiple phase centre cost (1)

• Phase shift adds a negligible overhead to
  station-based cost of correlation
• However, the baseline-based XMAC must be
  duplicated for each phase centre
• Station-based processing for VLBA (10 stations)
  outweighs baseline-based by 31
• Therefore theoretical overhead of N fields is a
  (N-1)/3 slowdown to correlation speed

10
11
Multiple phase centre cost (2)

• Alternative implementations exist where the
  rotation is done more analogously to
  post-correlator rotation, after subintegration
• Zero station-based cost, greater baseline-based
  cost, but less frequently
• Sacrifice time resolution (but still to an
  acceptable level) and computation reduced (factor
  of several lower overhead per field?)

11
12
Wide Area VLBI Res. Radio Survey

• The VLA FIRST survey covered 9,000 sq. deg. to
  an rms of 150 ?Jy _at_ 5 resolution, detecting
  800,000 sources (20/pointing)
• At 4 Gbps, VLBA sensitivity is comparable to the
  original VLA, and hence duplicating FIRST at VLBI
  resolution would take around the original VLA
  time (3000 hours)
• Hugely useful for understanding nature of a
  source in general studies

12
http//sundog.stsci.edu/
13
Outcomes of WAVRSS

• 800,000 uv datasets and images 12 TB correlated
  data, 6.5 TB image data
• Expect many non-detections 30 hit rate (Porcas
  et al. 2004) still yields 240,000 VLBI images
  (optimistic? CDFS 20-25)
• Provides an excellent grid of reference sources
  for astrometry (expect 1.5 detected sources gt 5
  mJy per pointing, total 60,000 calibrators)

13
14
WAVRSS astrometric accuracy

• The density of known calibrators is low 1 per 4
  sq. deg. - only 1 per 20 pointings!
• 1min/pointing -gt lengthy interpolation
• How to calibrate phase with such infrequent
  solid calibrator scans?
• Must bootstrap newly detected calibrators
• Absolute accuracy of final positions depends on
  existing calibrators - I expect 1 -- 10 mas

14
15
Other applications

• Multiple VLBI fields/pointing has plenty of
  applications beyond selecting in-beam calibrators
  (either WAVRRS or targeted)
• Globular cluster observations, with many
  astrometric targets in a single pointing
• Star formation region studies (searching for
  compact radio emitters for astrometric analysis)
• Discriminating AGN from starbursts in deep radio
  surveys (not really astrometry related)

15
16
Implementation status

• Now verification using the CDFS dataset
  (Middelberg et al.) already mentioned
• Hot off the press small shifts verified, bug
  affecting SNR with large shifts (probably
  precision related)

16
17
Conclusions

• The combination of higher sensitivity and new
  correlator flexibility will allow much more
  efficient inbeam calibrators searches than
  previously possible
• A wide survey to provide a database of 50,000
  VLBI calibrators is feasible
• These capabilities will be available from
  early 2010

17