The MegaPipe image stacking pipeline

The MegaPipe image stacking pipeline Point Spread Functions

Point Spread Functions

This page describes the CFHTLS point spread function (PSF) service, how to access the service and how the PSFs were generated. The PSFs are available for both the Deep and Wide fields. Two sets of PSFs were generated, one set with DAOphot the other with PSFex.

How to access the PSF service

The PSF service is intended to deliver PSF postage stamps relevant to a single location on a particular CFHTLS image. So far, the primary usage has been to model galaxy surface brightness profiles in batch mode. The URLs have the format:[image.fits]&x=[x-pixel]&y=[y-pixel]

For example, to get the PSF for the g-band image for the W1-1-1 field at position x=12000, y=8000 the URL would be:

Remember that most browsers convert the plus character (+) into a space ( ). Encode the plus character as "%2B". For example, for the W1+1+1 field, the URL would be:

Typical usage is not via the browser, but via wget or curl. For example:

  wget -O W1-1-1.G.12000.8000.fits ''

By default, the PSF service delivers a PSFs generated with DAOphot. To obtain PSFs generated with psfex, add "&psfex=true" to the URLs above. For example:

The PSFs are delivered as 31x31 pixel FITS images with the PSF centred on pixel (16,16).

You can download the .psf file produced by DAOphot directly from URLs like:

or you can download the .psf file produced by PSFex with URLs like:

How the PSFs were generated


The PSFs are generated with the standalone version of DAOphot. SExtractor and DAOphot are used to select PSF stars. SExtractor is run on the image. Stars are identified by comparing the 2-arcsecond aperture magnitude to the 3-arcsecond aperture magnitude. Stars form a well defined locus in this parameter. Stars near other bright objects are discarded. Here, near means within 40 pixels and bright means brighter than 4 magnitudes fainter than the star in question. A second list of PSF stars is compiled using the routines find and pick. Only stars in both the SExtractor and DAOphot lists are used. The DAOphot routine psf is then used to generate the PSF.

The gain and readout noise are set using the effective gain and readout noise. The fitting radius is set to 1.5 times the average FWHM of the image. The PSF radius is set to 15. A bivariate Gaussian model with a correction look-up table is used. The PSF is allowed to vary across the field with second order variations.

Two passes are made. After the first pass, the images are searched for second sources lying in the wings of the PSF stars. These PSF stars are removed from the list and the PSF is recomputed in the second pass.

Diagnostic images comprised entirely of postage stamps of the PSF star are generated an examined by eye. The images show 3 panels: the original PSF stars and the residuals after the first and second passes. For each CFHTLS image, two diagnostic images are generated. One shows the PSF star postage stamps arranged by magnitude, brightest to faintest, the other shows the PSF stars arranged by approximate location on the original CFHTLS image. The diagnostic images are visually examined for systematic variations of the PSF residuals as function of magnitude and position.

Compared to DAOphot, PSFex has far fewer parameters to be adjusted. SExtractor version 2.5.0 was run on the images to extract vignettes around each source. PSFex version 3.9.1 was run on the resulting LDAC catalog. The PSF varies with x and y as a third order polynomial. The PSF is not over-sampled. The PSFex configuration file is available. The FWHM and ellipticity diagnostic plots and the residual image were examined for irregularities.