This README describes the BLAST public data release =================================================== Authors ------- E. Chapin (echapin@phas.ubc.ca) M. Rex (madamson@physics.upenn.edu) Revision History ---------------- 20-Nov-2007 Initial version coinciding with first data release (EC) 26-Nov-2007 Modified flight times (inconsistent with web page) (EC) 29-Sep-2008 Updated to correspond with 06 data release (MR) 20-Nov-2008 More updates for 06 data release (EC) 03-Apr-2009 Give better explanation about use of PSF (EC) 14-Apr-2009 Add description of beam-smoothed data products (EC) 22-Apr-2009 Fixed typo in equation for smoothed flux maps (EC) Introduction ------------ The Balloon-borne Large-Aperture Sub-millimeter Telescope (BLAST) is a 2-m telescope that conducted the first wide-area (> many square degrees) sub-mm surveys at wavelengths 250--500 um. Built and flown by an international collaboration headed by the University of Pennsylvania (P.I. Mark Devlin), the telescope uses a prototype of the SPIRE camera for the Herschel satellite. Despite parts of this band being available to ground-based telescopes from high-altitude sites such as Mauna Kea (e.g. JCMT) and Chile (e.g. future ALMA site), BLAST surveys are currently un-matched in sensitivity and area given the comparatively negligible atmospheric water vapour at 38 km altitude. The BLAST data and associated papers that are currently available are listed at: http://blastexperiment.info/results.shtml Ultimately we plan to have all of the data hosted both by the CADC and IPAC. See links from the above page for access to the most recent data releases. Data were collected during two long-duration flights: the first was from northern Sweden to Canada in 2005 (BLAST05), lasting 4 days; and the second was a circumpolar flight in Antarctica in 2006, lasting 11 days (BLAST06). See "The Balloon-Borne Large Aperture Submillimeter Telescope: BLAST" for further details on the instrument and flights (linked from the results web page). Only processed map data are being made available to the public (primarily because the raw data is extremely difficult to work with --- BLAST is an experiment designed for a specific task rather than a common-user instrument). Observations are taken simultaneously in three broad imaging bands centred over the approximate wavelengths 250, 350 and 500 um with spectral bandwidths that are about 30% of the central frequencies. It is therefore important to account for the BLAST filter profiles when interpreting results derived from these maps. BLAST nominally has diffraction-limited beams with full-width half-maxima (FWHM) ~ 30, 42 and 60 arcsec. The BLAST05 flight suffered problems with its optics causing severe asymmetric side-lobes and a degradation in overall point-source sensitivity. In order to extract fluxes from maps it is important to make use of the effective point-spread functions (PSFs) that accompany the data. This problem was fixed for BLAST06. Data Product Details -------------------- 'Reduced': o Maps from the bolometer and pointing data have been produced using algorithms that have evolved from techniques used to make maps of the Cosmic Microwave Background (CMB) anisotropies. Three different map-makers which tackle the problem in slightly different ways have been developed, named SANEPIC, Almagest and Optibin. The method that was used for a particular data set is given by the keyword 'RECIPE' in the FITS header. A description of SANEPIC is now published --- see "SANEPIC: A Map-Making Method for Timestream Data from Large Arrays" on the BLAST results page. The units of the maps are given in the FITS headers (FITS keyword 'BUNIT'). *** NOTE: BLAST is insensitive to __ABSOLUTE__ brightness (zero-point), therefore the maps can have an arbitrary mean (usually close to zero). o The effective PSF for each map is stored as an image extension with the same pixel scale and orientation as the main image. It accounts for the fact that the beam is asymmetric, and the fact that the field is observed at a number of different position angles (relative to the telescope) --- always use the specific effective PSF that accompanies a map since they are __different for each observation__! The effective PSF should be used to convert between surface-brightness and point-source flux density units when required. The centre of the PSF is given by the CRPIX values in the FITS header in the extension. For example, given a surface brightness map M(x,y) [Jy/sr], and the effective PSF map P(x,y) [1/sr], a smooth map can be calculated using the following formula: S(x,y) = M(x,y) * P(x,y) ------------------- Sum ( P^2(x,y) ) x,y where (*) is the cross-correlation operator (not to be confused with convolution which induces flips over the x- and y- axes!). Each pixel in S will then contain an estimate of a point-source flux density in [Jy] centered over that location. For improved measurements of point source flux densities including the noise map see Eq.3 in Truch et al. (2008) ApJ 681 415. o We have adopted a calibration in which the reported brightnesses correspond to the __average brightness across the filter passband__. Given a filter transmission profile (see 'Filter' below), as a function of observing frequency, T(nu), the reported band-averaged flux-density (or surface brightness) is given by: S_av = integral( T(nu) * S(nu) * dnu ) ------------------------------- integral( T(nu) * dnu ) where S_av is the value of a pixel in the map, and S(nu) is the source spectrum (as either a flux density or surface brightness spectrum depending on the units you are working in). o Details on the beam shape and absolute calibration for BLAST05 are given in "The Balloon-borne Large Aperture Submillimeter Telescope (BLAST) 2005: Calibration and Targeted Sources" (linked from the results web page above). A similar report has now also been prepared for the BLAST06 data. 'Noise': o The map-making algorithm produces an estimate of the r.m.s. noise in each pixel along with the map itself, based on a propagation of statistical errors measured in the raw detector time-streams to the final map (which is some linear combination of the raw data). In reality the noise between different pixels in the map, P_i, have non-negligible covariances (i.e. is often significant). This noise map is therefore only the (square-root) of the diagonal of this covariance matrix. However, on small-scales relevant to the detection of point-sources these pixel-pixel correlations are probably small-enough that they can be ignored. See "SANEPIC: A Map-Making Method for Timestream Data from Large Arrays" on the BLAST results page for further details. 'Hits': o A map showing the number of data samples per pixel. BLAST data are sampled at 100 Hz, so that the total integration time in a pixel is the (hits)*(0.01 s). This map is useful for masking out the poorly-sampled edges of the observation. 'Decon': o This data file contains a more processed version of the 'Reduced' data. The type of processing is indicated by an additional suffix in the RECIPE keywords, which may be one of the following: "_FFT","_RL", "_WHT". o As with 'Reduced' data, 'Decon' maps may also have the effective PSF included as an image extension, which account for the effects of the additional processing. o Data from BLAST05 suffered optics problems that severely reduced the angular resolution of the telescope. We used a direct Fourier deconvolution technique to suppress the side-lobes and recover information from the maps down to scales approaching the designed diffraction limit (at the expense of increasing the noise, and introducing systematics that are difficult to characterize). The deconvolved maps ('Decon') have the same units as the regular 'Reduced' maps. Read "The Balloon-Borne Large Aperture Submillimeter Telescope (BLAST) 2005: A 4 deg^2 Galactic Plane Survey in Vulpecula (l = 59^deg)" for further details (on the BLAST results page). All maps that use this technique are flagged with the "FFT" suffix for RECIPE. o Future data will also be processed using a Richardson-Lucy deconvolution, and will be flagged with the "RL" suffix for RECIPE. o New extra-galactic data sets designed to detect point sources have been processed using a filter to suppress any residual noise on scales significantly larger than the PSF. Images processed with this technique are flagged with the "WHT" suffix for RECIPE. This filter can also attenuate sources slightly, so it is important to use the effective PSF for measurements of flux densities in the map. 'DeconNoise': o Similar to 'Noise' for the 'Reduced' maps, these files give an estimate of the r.m.s. noise per pixel (the sqrt diagonal of the pixel-pixel covariance matrix). 'Smooth': o The variance-weighted correlation between the signal maps and the effective point-spread functions. This operation calculates the maximum likelihood flux density [Jy] of an isolated point source centered over each pixel in the map (see description of PSF in 'Reduced' above, and Eq.3 in Truch et al. (2008) ApJ 681 415). See the values of the provenance FITS keywords (PRV*) to determine whether the raw reduced maps, or spatially filtered maps were used to produce this final data product. Note also that in some cases 'Smooth' maps may have had the number of pixels in the x- and y- directions truncated to even numbers (and may therefore differ by 1-pixel from the dimensions of 'Reduced' and/or 'Decon' maps). 'SmoothNoise': o Similar to 'Noise' for the 'Reduced' maps, these files give an estimate of the r.m.s. noise per pixel, which in this case is in units of [Jy]. 'Filter': o The BLAST filter transmission profiles (arbitrary normalization). The first column is wavelength (um) and the second column is transmission. These profiles must be used to interpret BLAST map brightnesses! (see 'Reduced') Acknowledgements ---------------- The papers that should be cited for a specific BLAST data set are given in the FITS header comments under 'ACKNOWLEDGEMENTS'.