SPIRE Spectrometer HPDP: re-gridded hyper-spectral cubes

Mapping observations re-gridded to remove high resolution spectra

Rosalind Hopwood, ESAC SPIRE Instrument and Calibration Scientist

12th December 2016, v1

Description of HPDP

There are 22 SPIRE Spectrometer mapping observations that suffer from one or more individual spectra that have higher than expected spectral resolution. In the Herschel Science Archive, the final standard product generation (SPG) produced hyper-spectral cubes (cubes) reports the highest resolution of all the spectra included when projecting the cube. This value is stored in the top-level metadata, with the FITS keyword ACTRES. The best spectral resolution expected for high-resolution SPIRE Spectrometer data is 1.18448225 GHz, which is representative for nearly all the spatial-pixels of any spectral cube where one or more of the constituent spectra have degraded resolution. This is due to the large number of spectra projected into any given cube and because the degraded spectra are always towards the outer ring of either of the Spectrometer bolometer detector arrays. For both of these reasons the 22 mapping observations in question can be re-gridded without change to the majority of their spatial pixels, although one or more pixels (of low quality) can be lost at the edge of the map.

As there is minimal change or loss of data when re-projecting mapping observations with the degraded resolution spectra omitted, and as the value of ACTRES in the associated standard pipeline products is not representative of the associated cubes as a whole, the mapping observations that suffer from one or more degraded spectra have been re-gridded to be provided as HPDPs. The new cubes, therefore, all report an actual unpadded spectral resolution of 1.18448225 GHz in their metadata.

The spectra used to create the new cubes were reduced with HIPE version 14.1, using spire_cal_14_3.

The cubes are extended-source calibrated in standard Spectrometer surface brightness units of W/m2/Hz/sr and presented in the same format as their HSA counterpart. There are four cubes per observation affected: Naive projected, convolution projected (CP), and apodized versions of these. All 22 observations were taken in high-resolution (HR) mode.

Each product is provided as a gzipped FITS file, which are named with the convention:

OBSID_correctedActualRes_PROJECTIONCube_SSW.fits.gz

OBSID_correctedActualRes_PROJECTIONCube_apod_SSW.fits.gz

where OBSID is the observation identification number, PROJECTION is either “naive” or “cp”

The files size ranges from 1.1 MB to 28.7 MB.

Purpose

These data have been made available to avoid confusion over their science readiness, and spectral resolution, and the need to re-project the cubes before they can be analysed.

Caveats

There are two Spectrometer bolometer detector arrays (the long wavelength, SLW, and short wavelength, SSW, arrays). Not all of the observations reprocessed suffered degraded resolution spectra for both arrays. Where only one array was affected, only one re-gridded HPDP cube is offered. A list of the reprocessed observations and which array was affected are below, where True indicates a HPDP exists.

OBSID SLW SSW
1342192173 False True
1342192174 False True
1342192175 False True
1342204898 False True
1342204920 True True
1342214827 True True
1342214841 True True
1342214846 False True
1342228703 True True
1342243631 False True
1342243632 False True
1342243633 False True
1342243634 False True
1342243635 False True
1342243636 False True
1342243637 False True
1342245851 True True
1342262908 False True
1342262909 False True
1342262913 False True
1342262916 False True
1342265845 False True

Postcards

The folder postcards contains a simple comparison of the cubes in the HSA and the regridded ones. A 2-D image was extracted from each cube by integrating over the frequencies from 1400 to 1450 GHz for SSW and from 600 to 650 GHz for SLW. The postcard shows both images together, using the same scale and colour table. There is no visible difference in most cases as the effect is really small and only involves pixels at the edge of the coverage.