Typical characterisation for ground-based radio interferometry data

This is applicable to data using the Griessen/Calabretta/Valdez definitions, if there are any differences that is my mistake! It also assumes AIPS processing or equivalent, hence complications like the array coordinates can be ignorred.

Other examples are possible, e.g. the mapping for solar or planetary observations.

*Filling - is this relative to Sensitivity Bounds or to Support? I assume Support. It might be better not to force it to be a factor so that No. visibilities can be included, unless there is a better place.

#Some radio interferometry data prodivers will have applied astrometric, flux scale, bandpass, Vlsr etc. corrections; others simply provide a package which includes the means to do so.

(1) For milli-arcsec accuracy in precessing coordinates it may be necessary to make leap second and epoch corrections, which is probably best done using a dedicated radio astronomy package e.g. AIPS UVFIX. In any case high accuracy double precision is needed, no worse than e.g. the highest resolution output of 'coco'.

(2) Coordinate conversion needs to be done to the nearest second, see (1). VLBI requires much higher precision pre-correlation but the VO should not need to worry about that. Pulsar data also need much better precision but there is a separate PVO being developed. In these instances the temporal Location etc. would be GPS, local maser etc. time/freq. standards. An observation can run over many days or months for a single image.

(3) Band may be a letter but to be VO compatible it should be the centre freq. + width or the upper and lower limits. Usually accuracy is much better than finest possible resolution, might need to include for SETI and some specialised projects.

(4) If visibility data, this could have a range of potential products.
Or may be a single product/set of products e.g. datacube
Polarisation products (order matters) :
Visibilities - 1 - 4 from circular (LL RR LR RL) or linear (XX YY XY YX);
Images (may be one per polarization or a cube) - as visibilities or Stokes I Q U V.

(5) The systematic position uncertainty is the sum of the uncertainty in the phase-ref source position which could be updated by the VO, and uncertainties in the phase transfer, antenna positions etc. which should be provided as numbers.

(6) This (and all the bounds) are inclusive extrema for Registry use etc., e.g. the lowest noise in the data, might be worse in some parts of the image, or if different weighting is used.

(7) Typically, an observation can be spread over 3 days (bounds). Each day, for each source, a track is spread over 12 hr (support) within which the source is observed for 4 mins out of each 6 mins (could be much less e.g. 4 mins per hr) (filling).

(8) If several spectral lines are observed in one frequency band the velocity coverage in the source rest frame will be different for each line. However if e.g. two lines are observed in two bands the frequency coverage will be different but the velocity coverage may be the same.

(9) Occasionally a correlator mode may be used which discards, e.g., every other channel giving 50% filling factor

(10) Field of View limit to 50% sensitivity (F50) (e.g. for images to be extracted from visibility data) is given by the smallest of:
F50~PB (Primary Beam to Full Width Half Max) of largest antenna in the array
F50~1E+4 Bmaj/Tint (Time smearing due to Earth rotation and phase rate in integration time)
F50~2(freq/dfreq)Bmaj (Frequency smearing in channel width used for imaging; multi-channel data can be used to produce a single image).
In each case the F.o.V. is in the units of the PB or Bmaj and Tint is in sec.
These are really continuous non-linear functions e.g. MERLIN version which express how the observable deteriorates as a function of distance from the centre of the F.o.V. (i.e. rms(F10)~1.1xrms).

(11) A long observation (one or more tracks) is usually designed to be averaged in time. However for variable sources one can measure the visibility amplitudes or spectra on shorter timescales. The minimum is the sampling interval (integration time) Tint
The noise in visibility data scales as root(time on source), i.e. rms(tsegment)=rms(full obs). root[(t x Nt)/tsegment]
For images there is an additional constraint of the minimum hour angle coverage needed to make an image (min, synthesis time) which the data provider would need to specify.

(12) The bandpass correction is usually automatically applied where needed. However for wide-band continuum images an estimate of the source spectral index may be needed. Leave to data provider/user at present!

(13) Image dynamic range D(n) if n channels are averaged as a function of dynamic range for whole band, Dmax e.g. D(n) ~ Dmax(n.dfreq/Dfreq)^2
where Dmax = peak flux density in image/rms noise
Weighting data to improve resolution increases noise, either data-provider provided or rms(weighted)/rms ~ Bmaj/Bmaj(weighted)

(14) Natural beam Bmaj ~ c/(freq.Bmax); restoring beam can be varied by factor ~2 up or down by weighting giving Bmaj(weighted). For uv ellipses of high eccentricity Bmax, Bmin and Bpa may be given.

-- AnitaRichards - 22 Jun 2004

Examples

Note that I have tried to fill in all fields but in practice only certain fields would be relevant for any specific data retrieval. Other details (e.g. bandpass calibration source used) are part of the data history and could be got elsewhere in the model (e.g. source schedule). However all fields may be required under some circumstances e.g. the bandpass calibration source name is needed if the bandpass correction hasn't already been applied.

Typical MERLIN archive data

Spectral data for Mrk 231 megamasers (2 OH lines in the same band after processing)

MERLIN+VLA HDF OBSERVATIONS
Deep field observations; typically calibrated visibility data would be available for extracting images at chosen position. Sources usually very faint so separate imaging of small time/frequency/polarization sections of data not practical, but some information about resolution in all domains is needed for F.o.V.

-- AnitaRichards - 24 Jun 2004