7.1.1. Scan map pipeline processing issues
The current automatic pipeline processing provides reductions up to PACS photometer level 1
(detector readouts calibrated and converted to physical units) and level 2 products (fits maps). At this point
in the Herschel mission, the goal of the pipeline is to deliver useable, but not necessarily optimal, products.
Instead, the automatic pipelines are optimized for stability, speed and delivering browse quality data. The
pipelines are expected to mature as both the instrument effects and the understanding on how to mitigate said
effects evolves.
The major data processing modules of the pipeline are discussed below:
- Cosmic rays removal: The default deglitching algorithm for PACS photometer is the multi-resolution
median transform (MMT), which uses wavelet scales (Starck et al, 1998, PASP, 110, 193S) to differentiate
between a cosmic ray hit and signal. The method produces reasonable results except for bright point-like
sources (e.g. compact galactic nuclei, etc.) and fast (60"/sec) scan speeds. Under such conditions, MMT
incorrectly masks point source cores. Users are advised to check the coverage map to look for any gaps
in exposure depth at the location of point sources. The optional module, "2nd order deglitching" can be
used in the interactive data processing as a replacement for MMT deglitching. "2nd order deglitching"
uses spatial redundancy and sigma-clipping to reject cosmic ray hits.
- Cross-talk correction: The red bolometer array shows cross-talk between column 0 and 16. That is,
when a source is present in column 16, its flux is also observed in column 0. Investigations on proper
removal of cross-talks are underway. Users are currently advised to mask all signal in columns 0 and 16
to avoid any artifacts from cross-talk in the final maps when observing very bright (point-)sources
- Offset drift correction: Offset drift correction is performed by applying a simple high-pass filter on
the data. This is adequate for point source fields, but not for structured fileds and may result in dark halo
artefacts. Increasing the filter window so that it covers a scan leg length already improves the map in the
case of structured objects.
- Map reconstruction: Map reconstruction with photProject is a simple projection of the data cube on
the map grid, considering the intersection areas of the native pixels with the map pixels. For this purpose
the ra/dec coordinates of the four corners of each detector pixel in the cube are computed on the fly in
photProject.
An alternative, optimal map maker, MADmap (Cantalupo et al. 2009), is provided in the
data processing environment. This is a java implementation of the original MADmap C code. MADmap
uses optimal map-making techniques to produce the best-fit solution for the final reconstructed maps. The
primary advantage of MADmap is that it does not require the use of high-pass filter to mitigate 1/f noise
and thus preserves spatial structures up to the size of final maps. Currently, MADmap is only available
as an interactive tool and requires pre-processing to remove correlated signal drifts prior to optimal map
reconstruction. MADmap additionally requires apriori knowledge of the detector noise properties (as the
inverse of the time-time noise correlation matrix). Investigations on the noise filters are underway from
the in-flight data. Current version of the data processing environment relies of pre-launch estimates from
ground-based tests. Users wishing to use MADmap for their data processing needs should contact the
helpdesks at ESA or NHSC for further information.
7.1.2. Level 2 pipeline products for scan maps generated with HCSS in the HSA archive
There are 3 types of products in the level 2 produced by the pipeline for the PACS photometer in scan map mode.
MADmap as been disabled from the HSC pipeline with HCSS 2.0, hence the second product (HPPMMAP) is
not available in recently processed scan maps observations.
These maps are produced by automatic pipeline scripts and shall only be considered as a preview, and not
for science directly.
- HPPPMAPB & HPPPMAPR stands for "Herschel Pacs Photometer PhotProject MAP Blue/Red"
This refers to maps produced by the photProject task, i.e. a simple projection of each frame (10Hz),
after running a temporal high-pass filter with a width of n=20 (i.e subtracting a median with a width
of 2*n+1 frames). This allows to filter a signicant part of the 1/f noise at the expense of removing
completely ALL spatial scales larger than this width (i.e. typically larger than 1 arcmin), and creating
negative undershooting around bright sources along the scan direction.
To preserve extended emission, the pipeline script shall be re-run with higher width in the high pass
filtering and masking bright source when necessary or an alternative map-making algorithm/tool.
This processing is mostly targeted to detect point-sources with good sensitivity.
Scan maps are in Jy/pixel.
- HPPMMAPB & HPPMMAPR stands for "Herschel Pacs Photometer MADMap Blue/Red"
MADmap maps are currently disabled from the HSC piepline, until scan and crossed-scan maps
can be processed together.
- HPPNMAPB & HPPNMAPR stands for for "Herschel Pacs Photometer Naive Map Blue/Red"
Averaged signal map after pixel-to-pixel offset correction. This image is used by MADmap as its first
value for the sky map and is subsequently improved and optimized iteratively as described above, hence
the full optimized matrix inversion has not been performed on the data.
Again these products are not avaialble in recent HSC pipeline maps.