Tuesday, November 20, 2012

1211.4257 (P. Popesso et al.)

The effect of the high-pass filter data reduction technique on the Herschel PACS Photometer PSF and noise    [PDF]

P. Popesso, B. Magnelli, S. Buttiglione, D. Lutz, A. Poglitsch, S. Berta, R. Nordon, B. Altieri, H. Aussel, N. Billot, R. Gastaud, B. Ali, Z. Balog, A. Cava, H. Feuchtgruber, B. Gonzalez Garcia, N. Geis, C. Kiss, U. Klaas, H. Linz, X. C. Liu, A. Moor, B. Morin, T. Muller, M. Nielbock, K. Okumura, S. Osterhage, R. Ottensamer, R. Paladini, S. Pezzuto, V. Dublier Pritchard, S. Regibo, G. Rodighiero, P. Royer, M. Sauvage, E. Sturm, M. Wetzstein, E. Wieprecht, E. Wiezorrek
We investigate the effect of the "high-pass filter" data reduction technique on the Herschel PACS PSF and noise of the PACS maps at the 70, 100 and 160 um bands and in medium and fast scan speeds. This branch of the PACS Photometer pipeline is the most used for cosmological observations and for point-source observations.The calibration of the flux loss due to the median removal applied by the PACS pipeline (high-pass filter) is done via dedicated simulations obtained by "polluting" real PACS timelines with fake sources at different flux levels. The effect of the data reduction parameter settings on the final map noise is done by using selected observations of blank fields with high data redundancy. We show that the running median removal can cause significant flux losses at any flux level. We analyse the advantages and disadvantages of several masking strategies and suggest that a mask based on putting circular patches on prior positions is the best solution to reduce the amount of flux loss. We provide a calibration of the point-source flux loss for several masking strategies in a large range of data reduction parameters, and as a function of the source flux. We also show that, for stacking analysis, the impact of the high-pass filtering effect is to reduce significantly the clustering effect. The analysis of the global noise and noise components of the PACS maps shows that the dominant parameter in determining the final noise is the high-pass filter width. We also provide simple fitting functions to build the error map from the coverage map and to estimate the cross-correlation correction factor in a representative portion of the data reduction parameter space.
View original: http://arxiv.org/abs/1211.4257

No comments:

Post a Comment