SKA Meeting Dwingeloo June 2005

1
SKA Meeting Dwingeloo June 2005
2
Peeling
Tom Oosterloo
3
Peeling of WSRT data using Miriad

• Sensitivity of radio telescopes depends on how
  well one can
• calibrate off-axis sources
• E.g. at low frequencies (LOFAR) different
  regions of the image seen
• through different parts of the ionosphere
• ? cannot assume a single calibration for the
  entire field
• ? need different calibration for different
  directions
• Peeling (MeqTree Jan Noordams talk)
• Important for future telescopes such as LOFAR
  SKA,
• but also needed now for WSRT (and other current
  arrays), even at L band,
• due to pointing/tracking errors, small
  differences in dishes, .

• have to use existing package for WSRT
• explore peeling using WSRT to compare to MeqTree

4
WSRT can make nice images!
NGC 6946 H I Boomsma et al.
NGC 5055 H I Battaglia et al.
5
But
½ degree
NGC 6946, L-band continuum
6
Large position-dependent errors
Result of standard selfcal on entire field (this
is a cleaned and selfcalled image!!!!) Dynami
c range 50 (!!!)

• due to
• pointing errors
• small differences between dishes
• different spectral response for
• off-axis sources (17-MHz ripple)

M81 M82, L-band continuum
7
Peeling, the Miriad way
Normal assumption 1 set of gains for entire
field
Clearly, this does not work for WSRT even at L
band (let alone LFFE.) Image plane effects
forces use of patches separate calibration for
different regions of the image
gain g and bandpass b position dependent
8
Peeling subtract sources and calibrate what is
left
1 degree
selfcal on entire field
3C343 3C343.1 WSRT L band
9
subtract central sources ? only off-axis source
left
10

• Peeling
• apply extra gains to data
• subtract off-axis source
• undo extra gains

11
off-axis source ( errors!!!) gone
selfcal centre again ? good calibration for
central field
12
By solving for extra gains for off-axis source ?
off-axis errors gone dyn range gt 10000
13
Model completeness
Only 2 sources included in amp selfcal
All sources included in amp selfcal
14
Worse case M81 M82 - L band - 2 x 160 MHz
bandwidth
selfcalled cleaned image dynamic range 50!!!!
15
selfcal on what is left
16
complication bandpass
image deconvolved using wide-band clean ? assume
smooth spectrum !!
17
The 17-MHz ripple.
Spectrum of off-axis source looks like
this Large bumps in spectrum every 17
MHz standing waves in dish
Wide-band clean in Miriad assumes smooth spectrum
(??) ? leads to deconvolution errors
18

• image deconvolved using wide-band clean
• assume smooth spectrum but 17-MHz
• bumps are present

do peeling also for bandpass off-axis spectrum
becomes smooth ? 17-MHz rings disappear
Both gain g and bandpass b position dependent
19
M82 well subtracted ? better calibration for M81
20

• 2nd iteration
• can subtract better M81 model to
• calibrate M82
• better removal of M82
• better calibration of M81

M82 well subtracted ? better calibration for M81
21
Final result
dynamic range 50
noise 15 ?Jy dynamic range 10000 (1822 cm
single deconvolution)
22
(No Transcript)
23
other example NGC 3079
different errors around different sources
UHF flaps stuck
24
Problem solved with peeling
25

• Peeling effective for calibrating off-axis
  effects WSRT
• Solve for extra time bandpass gains for each
  patch
• Do in full polarisation

Peeling works OK in Miriad but it is a lot of
work!!!! (scripting)
At L band just about doable but for LFFE
unpractical (many patches) there are
limitations (e.g. clean based) ? need
automated approach to patches and model
subtraction parametric models
? MeqTree based applications for WSRT