PPC version 2014: Operates on 1D spectra extracted from compact sources

Full technical description is summarised in POSSUM Report 62

Item

Description

Issues/Comments

Status

Stokes I, Q, U Spectra

1D spectra extracted at the location of compact EMU sources.

May need to be updated to deal with resolved sources.

Complete

I, Q, U Noise Spectra

1D spectra of RMS noise from a square aperture of 50 beam areas. Use MADFM estimator to be robust to source contamination.

May need to be updated to deal with resolved sources.

Complete

Model Stokes I spectrum

1D model Stokes I spectrum used to create fractional polarisation spectra q=Q/I, u=U/I.

Constructed using a 2nd order polynomial fit. If fit fails default to flat spectrum at median flux.

Complete

lambda_0, freq_0

Weighted average wavelength squared and frequency squared.

Used to convert FDF back from fractional units to Jy.

Complete

RMSF,
width

Rotation Measure spread function and width of main lobe.

Property of wavelength sampling and weighting (assumed 1/variance^2).

Complete

FDF

Dirty Faraday Dispersion Function

Real, Imaginary & Polarised Intensity Faraday Dispersion Function. Created using fractional Q/I and U/I spectra, weighted by variance.

Complete

FDF Measurements

RM of peak in intensity FDF, fractional polarisation of peak (+ debiased version), polarisation angle at lambda_0, polarisation angle at lambda=0 (derotated), complexity flags.

Complexity flags measure deviation of P spectrum from a constant and residual in Q and U after subtracting a Faraday thin component at the peak RM.

Complete

Quality Flags

Flags indicating something wrong with the data or processing.

Stokes I model fit failed, unusually high noise, large numbers of channels flagged, RM unusually high, detection on the edge of the FDF or data cube.

In Progress

Questions, discussion points and notes.

• Need to supply unit-check (artificial) data with the PPC prototype for verification purposes.
• Do we need a RM-spectrum in annulus/region around EMU source.?
• Is a noise 'map' in I, Q, U & V calculated along channels useful?
• Complexity: If using spread in CC to determine complexity, need to have 20-30 channels across RMSF main lobe. Better approach is to look for variations is P vs lambda^2.
• Specification of algorithms/metrics for quality control. Key step in moving from Level 5 to 6.
• Can we use EMU MFS spectral index and curvature as a model for Stokes I data?
• Input POSSUM data cubes (300 channels) won't be checked by EMU. Need to create a quality control process for these (reports 45 and 51).
• How do we detect sources that are between 5 and 10 sigma (90% of possible detections)?
• Would bright sources be leakage dominated? How to incorporate leakage in uncertainties?
• If we provide a RM-synthesis cube, how can we correct for spectral index over the whole image? What do we divide by for the 'noise' pixels. Larry says we should use the fitted spectrum from MFS and divide every off-source pixel by that spectrum.