PACS Handbook Mar16

Herschel Explanatory Supplement
The Photodetector Array Camera and Spectrometer

(PACS)
HERSCHEL-DOC-xxxx, version 1.0, March 6, 2016
2
PACS Handbook
Version 1.0, March 6, 2016
This document is based on inputs from the PACS Consortium and the PACS Instrument
Control Centre. The name of this document during the active observing phase of Herschel
was PACS Observers Manual. The versioning IS OR IS NOT continued with the new
name.
Document editors:
Katrina Exter, Herschel Science Centre, European Space Astronomy Centre, European Space
Agency
Zoltan Balog, MPIA
Ulrich Klaas, MPIA
Dieter Lutz, MPE
Roberta Paladini, NHSC
custodians: Katrina Exter and Zoltan Balog
Contents
1 Introduction
11
1.1
The Observatory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
1.2
Purpose and Structure of Document . . . . . . . . . . . . . . . . . . . . . . .
12
1.3
Key PACS publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
1.4
Further sources of information
. . . . . . . . . . . . . . . . . . . . . . . . . .
12
1.5
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
1.6
Changes to this document . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
1.7
List of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
2 The PACS Instrument

2.1
Instrument design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
2.1.1
Warm electronics units . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
2.1.1.1
Digital Processing Unit (DPU) . . . . . . . . . . . . . . . . .
16
2.1.1.2
DEtector and MEchanisms Controler (DECMEC) . . . . . .
17
2.1.1.3
BOLometer Controler (BOLC) . . . . . . . . . . . . . . . . .
17
2.1.1.4
Signal Processing Unit (SPU)
. . . . . . . . . . . . . . . . .
17
Cold focal plane unit (FPU) . . . . . . . . . . . . . . . . . . . . . . . .
17
Common optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
2.2.1
Entrance optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
2.2.2
Calibration sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
2.2.3
Chopper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
Photometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
2.3.1
Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
2.3.1.1
Effective spectral response . . . . . . . . . . . . . . . . . . .
27
2.3.1.2
Filter wheel . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
2.3.2
Bolometer arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
2.3.3
Cooler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
2.1.2
2.2
2.3
15
CONTENTS
2.3.3.1
Procedure of the Cooler Recycling . . . . . . . . . . . . . . .
31
2.3.3.2
Establishment of the Cooler Hold Time Relation for Herschel

Mission Planning . . . . . . . . . . . . . . . . . . . . . . . . .
32
Mission Statistics of Cooler Recycling . . . . . . . . . . . . .
33
Spectrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.4.1
Instrument design . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.4.2
Image slicer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.4.3
Grating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.4.4
Order-sorting filters . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.4.5
Photoconductor arrays . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
On-board data handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.5.1
Data gathering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.5.2
On-board data reduction . . . . . . . . . . . . . . . . . . . . . . . . .
36
2.5.3
On-board data compression . . . . . . . . . . . . . . . . . . . . . . . .
36
2.3.3.3
2.4
2.5
3 Observing with PACS
37
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
37
3.2
Photometer observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
38
3.2.1
Scan map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
3.2.1.1
Point source photometry (mini-scan map) . . . . . . . . . . .
39
3.2.2
Point-source photometry (chop-nod) . . . . . . . . . . . . . . . . . . .
40
3.2.3
SPIREPACS parallel . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
Spectrometer observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
3.3.1
The background measurement method . . . . . . . . . . . . . . . . . .
42
3.3.1.1
Chop-nod mode . . . . . . . . . . . . . . . . . . . . . . . . .
43
3.3.1.2
The chop-nod AOT . . . . . . . . . . . . . . . . . . . . . . .
44
3.3.1.3
Unchopped mode . . . . . . . . . . . . . . . . . . . . . . . .
44
3.3.1.4
The unchopped AOT . . . . . . . . . . . . . . . . . . . . . .
45
The spectral coverage and sampling . . . . . . . . . . . . . . . . . . .
46
3.3.2.1
Line spectroscopy . . . . . . . . . . . . . . . . . . . . . . . .
47
3.3.2.2
The line spectroscopy AOT . . . . . . . . . . . . . . . . . . .
48
3.3.2.3
Range spectroscopy . . . . . . . . . . . . . . . . . . . . . . .
49
3.3.2.4
The range spectroscopy AOT . . . . . . . . . . . . . . . . . .
50
3.3.2.5
Wavelength switching . . . . . . . . . . . . . . . . . . . . . .
51
The pointing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
52
3.3.3.1
52
3.3
3.3.2
3.3.3
Pointed observation . . . . . . . . . . . . . . . . . . . . . . .
CONTENTS
5
3.3.3.2
Pointed with dither . . . . . . . . . . . . . . . . . . . . . . .
52
3.3.3.3
Rasters/mapping . . . . . . . . . . . . . . . . . . . . . . . . .
53
4 In-flight Performance
55
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
4.2
Photometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.2.1
Sensitivity and flux uncertainty . . . . . . . . . . . . . . . . . . . . . .
56
4.2.2
Noise properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.2.3
PSF and beam size . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.2.4
Anomalies and quality control . . . . . . . . . . . . . . . . . . . . . . .
56
Spectrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.3.1
Spectral range and resolution . . . . . . . . . . . . . . . . . . . . . . .
56
4.3.1.1
Line profile . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.3.1.2
Spectral bands . . . . . . . . . . . . . . . . . . . . . . . . . .
56
Spectral purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.3.2.1
Leakage regions . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.3.2.2
Second-pass ghosts . . . . . . . . . . . . . . . . . . . . . . . .
56
4.3.2.3
? 62 m dip ? . . . . . . . . . . . . . . . . . . . . . . . . . .
56
Wavelength calibration . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
4.3.3.1
Wavelength shifts with source position . . . . . . . . . . . . .
56
4.3.3.2
Flux calibration accuracy for the various modes . . . . . . .
56
4.3.3.3
Flux accuracy for point and semi-extended sources . . . . . .
56
4.3.3.4
Flux accuracy for extended sources . . . . . . . . . . . . . .
56
4.3.3.5
? The effect of transients ? . . . . . . . . . . . . . . . . . . .
56
Spatial accuracy and resolution . . . . . . . . . . . . . . . . . . . . . .
56
4.3.4.1
Pointed observations . . . . . . . . . . . . . . . . . . . . . . .
56
4.3.4.2
Mapping observations . . . . . . . . . . . . . . . . . . . . . .
56
Anomalies and quality control . . . . . . . . . . . . . . . . . . . . . . .
56
4.3
4.3.2
4.3.3
4.3.4
4.3.5
5 The Spectrometer Calibration
57
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
5.2
The spectro-photometric calibration . . . . . . . . . . . . . . . . . . . . . . .
59
5.2.1
The calibration sources and models . . . . . . . . . . . . . . . . . . . .
59
5.2.2
Chopped observations: calibrating the telescope background . . . . . .
59
5.2.2.1
Telescope spectrum . . . . . . . . . . . . . . . . . . . . . . .
59
5.2.2.2
Time variability . . . . . . . . . . . . . . . . . . . . . . . . .
59
CONTENTS
5.2.3
5.3
5.4
5.5
Unchopped observations: using the internal calibration sources . . . .
59
5.2.3.1
Calibration source fluxes . . . . . . . . . . . . . . . . . . . .
59
5.2.3.2
Absolute flux calibration . . . . . . . . . . . . . . . . . . . .
59
5.2.3.3
Relative spectral response function (RSRF) . . . . . . . . . .
59
5.2.4
Line flux calibration in order-leak regions . . . . . . . . . . . . . . . .
59
5.2.5
Saturation limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
5.2.6
Self-curing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
Spatial calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
5.3.1
Focal-plane geometry
. . . . . . . . . . . . . . . . . . . . . . . . . . .
59
5.3.2
Characterising the beam . . . . . . . . . . . . . . . . . . . . . . . . . .
59
5.3.2.1
Beam profiles . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
5.3.2.2
Point source calibration function . . . . . . . . . . . . . . . .
59
5.3.2.3
Semi-extended source correction . . . . . . . . . . . . . . . .
59
5.3.2.4
Extended sources and the spatial flatfield . . . . . . . . . . .
59
5.3.2.5
Intra-sapxel losses . . . . . . . . . . . . . . . . . . . . . . . .
59
Spectral calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
5.4.1
The line profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
5.4.2
Calibration the wavelengths . . . . . . . . . . . . . . . . . . . . . . . .
59
5.4.3
Spectral purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
Table of the calibration files . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
6 The Photometer Calibration
61
6.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
6.2
Calibration achievements during commissioning and the performance verification phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
Spatial calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
64
6.3.1
Focal plane mapping and field-of-view distortion . . . . . . . . . . . .
64
6.3.1.1
Non-uniform pixel size . . . . . . . . . . . . . . . . . . . . . .
67
PACS photometer point spread function . . . . . . . . . . . . . . . . .
67
6.3.2.1
PSF morphology . . . . . . . . . . . . . . . . . . . . . . . . .
67
6.3.2.2
Effects of source color and reduction schemes . . . . . . . . .
69
6.3.2.3
Effects of nonlinearity, saturation, crosstalk, ghosts, straylight

on observed PSF . . . . . . . . . . . . . . . . . . . . . . . . .
71
Encircled energy fraction . . . . . . . . . . . . . . . . . . . .
72
Flux calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
6.4.1
73
6.3
6.3.2
6.3.2.4
6.4
Flat-field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CONTENTS
6.4.2
7
Responsivity calibration . . . . . . . . . . . . . . . . . . . . . . . . . .
74
6.4.2.1
77
Internal calibration sources . . . . . . . . . . . . . . . . . . .

6.4.2.1.1
6.4.2.2
6.6
78
Celestial calibration sources and models . . . . . . . . . . . .
78
6.4.2.2.1
Fiducial standards . . . . . . . . . . . . . . . . . . .
78
6.4.2.2.2
Planets and moons . . . . . . . . . . . . . . . . . . .
79
6.4.2.2.3
Asteroids . . . . . . . . . . . . . . . . . . . . . . . .
84
6.4.2.2.4
Faint stars . . . . . . . . . . . . . . . . . . . . . . .
87
Evaporator temperature correction . . . . . . . . . . . . . . .
90
Photometry corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
6.5.1
Non-linearity corrections . . . . . . . . . . . . . . . . . . . . . . . . . .
90
6.5.1.1
Establishment of the non-linearity correction . . . . . . . . .
90
6.5.1.2
Verification with celestial standards . . . . . . . . . . . . . .
93
6.5.2
Aperture corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
6.5.3
Colour corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
6.5.3.1
Comparison with similar bands from other missions . . . . .
95
6.5.3.2
Colour correction factor tabulation . . . . . . . . . . . . . . .
96
Table of the calibration files . . . . . . . . . . . . . . . . . . . . . . . . . . . .
101
6.4.2.3
6.5
Calibration block observation data base . . . . . . .
7 The PACS Pipelines

7.1
Spectroscopy pipelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
103
7.1.1
SPG pipelines: Level 0 to 0.5 . . . . . . . . . . . . . . . . . . . . . . .
105
7.1.2
SPG pipelines: Level 0.5 to 2 . . . . . . . . . . . . . . . . . . . . . . .
106
7.1.2.1
Chop-nod AOTs . . . . . . . . . . . . . . . . . . . . . . . . .
107
7.1.2.2
Unchopped AOTs . . . . . . . . . . . . . . . . . . . . . . . .
110
SPG pipelines: final steps . . . . . . . . . . . . . . . . . . . . . . . . .
112
7.1.3.1
Pointed observations . . . . . . . . . . . . . . . . . . . . . . .
112
7.1.3.2
Mapping observations . . . . . . . . . . . . . . . . . . . . . .
113
7.1.4
The standalone browse products . . . . . . . . . . . . . . . . . . . . .
114
7.1.5
The interactive pipelines: leave for later . . . . . . . . . . . . . . . . .
114
Photometer pipelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
114
7.2.1
Scan map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
114
7.2.1.1
SPG to Level 1 . . . . . . . . . . . . . . . . . . . . . . . . . .
114
7.2.1.2
Level 1 to Level 2, 2.5, 3 . . . . . . . . . . . . . . . . . . . .
115
7.1.3
7.2
103
7.2.1.2.1
HighPass filter and PhotProject . . . . . . . . . . .
115
CONTENTS
7.2.1.2.2
JScanam . . . . . . . . . . . . . . . . . . . . . . . .
117
7.2.1.2.3
Unimap . . . . . . . . . . . . . . . . . . . . . . . . .
120
7.2.1.2.4
MADmap . . . . . . . . . . . . . . . . . . . . . . . .
124
Comparison of map-makers . . . . . . . . . . . . . . . . . . .
126
7.2.2
Chop-Nod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
127
7.2.3
PACS-SPIRE parallel mode . . . . . . . . . . . . . . . . . . . . . . . .
129
7.2.1.3
8 The PACS products
131
8.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
131
8.2
A PACS ObservationContext
. . . . . . . . . . . . . . . . . . . . . . . . . .
132
8.3
Spectroscopy observations: pipeline products . . . . . . . . . . . . . . . . . .
132
8.3.1
Viewing an observation in HIPE . . . . . . . . . . . . . . . . . . . . .
132
8.3.2
Viewing an observation on disk . . . . . . . . . . . . . . . . . . . . . .
134
8.3.3
The high-level spectroscopy pipeline products . . . . . . . . . . . . . .
134
8.3.3.1
The role of the observing mode . . . . . . . . . . . . . . . . .
134
8.3.3.2
The types of cubes produced . . . . . . . . . . . . . . . . . .
135
8.3.3.3
The types of spectrum tables produced . . . . . . . . . . . .
137
8.3.3.4
Cube and spectral table descriptions . . . . . . . . . . . . . .
137
Photometry observations: pipeline products . . . . . . . . . . . . . . . . . . .
138
8.4.1
Viewing an observation in HIPE . . . . . . . . . . . . . . . . . . . . .
138
8.4.2
Viewing an observation on disk . . . . . . . . . . . . . . . . . . . . . .
138
8.4.3
The high-level pipeline products . . . . . . . . . . . . . . . . . . . . .
138
8.4.3.1
. . . . . . . . . . . . .
139
8.4.3.1.1
highPass filtering + photProject . . . . . . . . . . .
139
8.4.3.1.2
JScanam . . . . . . . . . . . . . . . . . . . . . . . .
139
8.4.3.1.3
Unimap . . . . . . . . . . . . . . . . . . . . . . . . .
139
Spectroscopy Standalone Browse products . . . . . . . . . . . . . . . . . . . .
139
8.5.1
The SBP cubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
139
8.5.2
The SBP spectrum tables . . . . . . . . . . . . . . . . . . . . . . . . .
140
8.4
8.5
Science case related pipeline product
8.6
Photometry Standalone Browse products
. . . . . . . . . . . . . . . . . . . .
141
8.7
Spectroscopy Highly-processed data products . . . . . . . . . . . . . . . . . .
141
8.8
Photometry Highly-processed data products . . . . . . . . . . . . . . . . . . .
141
8.9
Spectroscopy Legacy data products . . . . . . . . . . . . . . . . . . . . . . . .
141
8.10 Photometry Legacuy data products . . . . . . . . . . . . . . . . . . . . . . . .
141
8.11 Calibration products and the calibration tree . . . . . . . . . . . . . . . . . .
141
8.11.1 Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
141
CONTENTS
9 Decision tables: how do I know which products I need for my science?
9
149
10
CONTENTS
Chapter 1
Introduction
1.1
The Observatory
The Herschel Space Observatory (pilbrat) was the fourth cornerstone mission of ESA science
programme, aimed at high spatial-resolution observations in the far-IR and sub-millimetre
regime; in fact it is the first space observatory to cover this part of the spectrum. It was
named after Sir William Herschel, the discoverer of infrared radiation.
Herschel was launched on the May 14th, 2009 aboard an Ariane 5 rocket together with Planck.
It was placed on a Lissajous 700 000 km diameter orbit 1.5 million kilometers away from Earth
at the second Lagrange point of the Earth-Sun system. The mission ended on April 29th,
2013 when its coolant ran out; the final command on June 17th was to put it into a safe
disposal-orbit around the Sun.
The Herschel mirror, which was radiatively cooled, had a diameter of 3.5m, and it was
the largest mirror launched at the time. The telescope performed photometry and spectroscopy in the 55670 mrange with three science instruments PACS, SPIRE, and
HIFI which were housed inside a superfluid helium cryostat. PACS (Photodetector Array Camera and Spectrometer) provided Herschel with the capabilities for spectroscopy and imaging/photometry in the 55210 mrange. A high-level description of
the Herschel Space Observatory is given in (pilbrat); more details are given in the
(http://herschel.esac.esa.int/Docs/Herschel/html/observatory.html to Herschel Observers
Manual). The first scientific results are presented in the special volume 518 of Astronomy
& Astrophysics journal. Information with latest news, documentation, data processing and
access to the Herschel Science Archive (HSA) is provided in the Herschel Science Centre web
portal (http://herschel.esac.esa.int).
PACS was designed and built by a consortium of institutes and university departments from
across Europe under the leadership of the Principal Investigator Albrecht Poglitsch at MaxPlanck-Institute for Extraterrestrial Physics, Garching, Germany. Consortium members came
from: Austria: UVIE; Belgium: IMEC, KUL, CSL; France: CEA, OAMP; Germany: MPE,
MPIA; Italy: IFSI, OAP/OAT, OAA/CAISMI, LENS, SISSA; Spain: IAC.
11
12
1.2
CHAPTER 1. INTRODUCTION
Purpose and Structure of Document
The purpose of this handbook is to provide relevant information to astronomers about the
PACS instrument, to help them understand and use the scientific observations performed with
it. In Chp. ?? is an overview of the PACS instrument, its design and capabilities. In Chp. 3
is an overview of how the observations were performed and how the observers could modify
their observations to suite their needs. In Chp. 4 is a summary of the in-flight performance of
PACS, with the various calibration uncertainties given. In Chps ?? and 6 are the details of how
PACS was calibrated (fluxes, wavelengths, etc), concentrating on the calibration necessary for
the respective data reduction pipelines. Chp. 7 is about these data reduction pipelines, and
in Chp. 8 is an explanation of the products provided for any observation downloaded from the
Herschel Science Archive. Finally, in Chp. ?? is a decision tree to help users of PACS data
to understand what they can do, need to do, and should pay attention to, when dealing with
PACS observations of different observing modes or source type Include in the handbook or
elsewhere?.
1.3
Key PACS publications
For a detailed description of the PACS instrument and its in-flight performances, we refer
you to Poglitch et al., (2010) LINK.
Links to the key refeered publications, from Poglitch et al 2010 to the latest. Compile before
releasing each completed version, since this will change with time.
1.4
Further sources of information
All sources of information about PACS organised in such a way as to place this Handbook
within that context. Link to the HELL and summarise what can be found there, and which
tier from there fed this Handbook.
1.5
Acknowledgements
This handbook is provided by the Herschel Science Centre, based on the PACS Observers
Manual, the extensive astronomer- and user-orientated documentation, and a wealth of technical notes written by the PACS Consortium and PACS Instrument Control Centre.
To cite this document, please use
PACS Handbook, 2016, herschel-hsc-doc-XXX vX.X.
1.6
Changes to this document
Need to get hold of HSO banner and PACS logo for cover page
1.7. LIST OF ACRONYMS
1.7
update!
List of Acronyms
13
14
AOR
AOT
DCU
DP
DPU
ESA
ESAC
FCU
FIR
FOV
FPU
FTS
HCSS
HIFI
HIPE
HSA
HSC
HSpot
IA
ILT
ISM
JFET
LSR
NEP
OD
OPD
PACS
PCAL
PFM
PLW
PMW
PSW
PTC
RMS,rms
RSRF
SCAL
SED
SLW
SMEC
SNR, S/N
SPG
SPIRE
SSW
ZPD
CHAPTER 1. INTRODUCTION
Astronomical Observation Request
Astronomical Observation Template
Detector Control Unit
Data Processing
Digital Processing Unit
European Space Agency
European Space Astronomy Centre, based near Madrid, Spain
FPU Control Unit
Far Infrared Radiation
Field of View
Focal Plane Unit
Fourier-Transform Spectrometer
Herschel Common Software System
Heterodyne Instrument for the Far Infrared
Herschel Interactive Processing Environment
The Herschel Science Archive
The Herschel Science Centre (based in ESAC)
Herschel Observation Planning Tool
Interactive Analysis
Instrument Level Test (i.e. ground tests of the instrument without the spacecraft)
Inter Stellar Medium
Junction Field-Effect Transistor
Local Standard of Rest
Noise-Equivalent Power
Operational Day
Optical Path Difference
Photodetector Array Camera and Spectrometer
Photometer Calibration Source
Proto-Flight Model of the instrument
SPIRE Photometer Long (500 m) Wavelength Array
SPIRE Photometer Medium (350 m) Wavelength Array
SPIRE Photometer Short (250 m) Wavelength Array
Photometer Thermal Control Unit
Root Mean Square
Relative Spectral Response Function
Spectrometer Calibration Source
Spectral Energy Distribution
SPIRE Long (316-672 m) Wavelength Spectrometer Array
Spectrometer Mechanism
Signal-to-Noise Ratio
Standard Product Generation
Spectral and Photometric Imaging REceiver
SPIRE Short (194-324 m) Wavelength Spectrometer Array
Zero Path Difference
Chapter 2
The PACS Instrument

2.1
Instrument design
PACS was designed as a general-purpose science instrument for the far-infrared wavelength
range 60 210 m. It incorporated two main astronomical measurement and analysis methods, namely
1) Imaging dual band photometry ( <0.5) over a field of view of 3.50 1.750 , with full
sampling of the telescope point spread function. Bandpass combinations were either 60
85 m and 125 210 m (blue/red filter) or 85 125 m and 125 210 m (green/red
filter).
2) Medium resolution (R 1500) integral-field spectroscopy between 51 and 220 m with
a velocity resolution of 75 - 300 km/s and an instantaneous coverage of 1500 km/s
over a field of view of 4700 47 arcsec
Scientific rationales for both photometer and spectrometer design are given in Poglitsch et al.
(2010, A&A 518, L2).
The instrument consisted of two units (see Fig. 2.1)
1) The cold focal plane unit (FPU) inside the Herschel cryostat on the payload module
2) The warm electronic unit on the Herschel service module
Both units were connected via a harness. The harness consisted of a cold part inside the
cryostat vessel and a warm part outside. Care was taken in the arrangement of the warm
harness to optimize electromagnetic compatibility (EMC), e.g. interference of the signals on
the harness by electrical switching of the solar panels.
2.1.1
Warm electronics units
The warm electronics unit of PACS had the following tasks

1) Control the instrument operations
15
16
CHAPTER 2. THE PACS INSTRUMENT
Figure 2.1: Main components of the PACS instrument. Left: Cold focal plane unit (T 4 K)
assembled at Kayser-Threde, Munich, containing the optical elements and the detectors (see
Section 2.1.2 for details). The location of some elements is indicated. Middle: The harness
connecting the cold focal plane unit inside the Herschel cryostat on the payload module with
the warm electronic unit on the satellite service module. Right: Warm electronic boxes (T
300 K) with the on-board CPUs controlling the instrument and receiving engineering and
science telemetry. Individual boxes are explained in Section 2.1.1.
2) Execute the data handling
3) Provide instrument autonomy for the about 20 h time interval between the Daily TeleCommunication Periods (DTCPs) with the Ground Stations and the Missions Operations Centre (MOC)
These tasks were too voluminous and complex to be handled by a single on-board computer.
Therefore, the various functionalities were associated with different sub-units.
2.1.1.1
Digital Processing Unit (DPU)
The DPU was the central PACS on-board computer. It provided the interface of PACS
to the satellite and was therefore responsible for receiving and decoding commands either
directly from the Missions Operations Centre (mainly for instrument commissioning and
contingency recovery) or routinely from the Mission Timeline uplinked in advance and stored
on-board. Decoded commands were forwarded to the appropriate subsystems for execution.
The DPU monitored the correct execution of all commands and raised error messages in case
of a failure. In case of hard limit violations of essential voltage or temperature values, an
autonomy function was triggered setting the instrument into a safe configuration. By daily
checking of the event log, the MOC staff could recognize quite early on in the DTCP, if there
was an instrument contingency and arrange for recovery.
In the opposite data flow direction, all housekeeping and compressed science data from the
various subsystems of PACS were collected, formatted into telemetry packets and sent to the
satellite mass memory.
2.1. INSTRUMENT DESIGN
17
For details, please refer to the DPU On-board Software User Manual (PACS-CR-UM-024,
issue 3.0, 10-Jun-2008).
2.1.1.2
DEtector and MEchanisms Controler (DECMEC)
The DECMEC received and handled all low level commands to all PACS systems, except
for the bolometers. Tasks that needed to be synchronous (e.g. detector readouts, chopper
motion, grating steps) were triggered from here.
The DEC part operated the photoconductor arrays and received their raw data which arrived
at 256 Hz. It also received the digitized bolometer data from BOLC. The detector raw data
were then sent on to the SPU for processing.
The MEC part contained drivers and control loop units for all mechanisms and regulated
temperatures in the FPU and generated most of the instrument housekeeping. For special
analysis of individual instrument components, diagnostic housekeeping with a rate of 1 kHz
was possible.
For more details, please refer to the DEC/MEC User Manual (PACS-CL-SR-002, issue 4.8,
11-Nov-2008).
2.1.1.3
BOLometer Controler (BOLC)
BOLC operated the bolometer arrays in the PACS photometer and provided a clock signal
to the MEC for the chopper synchronization in photometer mode. The digitized bolometer
signals were sent to the SPU via DECMEC for processing.
For details, please refer to the PACS FM Photometer Focal Plane Unit Users Manual (SApPACS-MS-0616-06, issue 1.0, 25-Feb-2009).
2.1.1.4
Signal Processing Unit (SPU)
The SPU reduced the raw data rate from the detector arrays, which exceeded the average
allowed telemetry rate of 130 kb/s by far. The SPU performed real-time reduction of the raw
data in the time domain, by averaging bolometer frames or fitting photoconductor integration
ramps, bit rounding and loss-less compression. The compressed science data were sent to the
DPU for formatting in the telemetry packets.
For a more detailed description of the algorithms employed see Section 2.5.
2.1.2
Cold focal plane unit (FPU)
The cold focal plane unit was divided into optically well separated compartments as shown
in Fig. 2.2. The front optics section, common to all optical paths through the instrument,
contained the internal calibration sources and the focal plane chopper. Thereafter followed
the separate photometer camera and spectrometer sections.
The FPU was mounted on an optical bench together with other two Herschel instruments
and was kept on the L1 temperature level of 3 to 5 K of the cryostat. The spectrometer
18
detectors had thermal straps connected to the L0 temperature level (1.65 K) of the liquid 4 He
bath inside the cryostat vessel. The bolometer detectors were operated at 285 mK provided by
a dedicated 3 He sorption cooler, which itself was connected to the L0 level during recycling.
All optical elements and detector units of the FPU are explained in more detail in the following
sections.
Figure 2.2: Left: PACS focal plane unit (FPU) functional block diagram. The arrows visualize the optical paths through the instrument. Imaging optics are highlighted by the light
brown boxes, filter components by the light beige boxes. Blocks containing active components (mechanisms, electronics) are outlined in bold. Right: FPU optical layout. After the
common entrance optics with internal calibrators and the chopper, the field was split into
the spectrometer train and the photometer trains. The two bolometer cameras (top left)
had partially separate re-imaging optics split by a dichroic beam splitter. Filter wheel I contained the filters for the blue bolometer channel. In the spectrometer train, the integral field
unit (IFU) image slicer (middle) converted the square field into an effective long slit for the
Littrow-mounted diffraction grating (top right). The dispersed light was distributed to the
two photoconductor (Ge:Ga) arrays, the first order to the red array (middle left), 2nd and
3rd order to the blue array (bottom right). The order sorting filters on filter wheel II were
used to choose between 2nd and 3rd order light.
2.2
2.2.1
Common optics
Entrance optics
The entrance or front optics had several instrument wide tasks: It provided for an intermediate
image of the telescope secondary mirror (the entrance pupil of the telescope) with the cold
Lyot stop and the first blocking filter (FL E, see e.g. Section 2.3.1), common to all instrument
channels. A further image of the pupil was reserved for the focal plane chopper. For more
details of this device see Section 2.2.3. The focal plane chopper allowed spatial chopping on
the sky and - through two field mirrors adjacent to the field stop in the telescope focal plane
- to switch between field of view on the sky and on the two internal calibration sources (see
Fig. 2.3).
2.2. COMMON OPTICS
19
In an intermediate focus after the chopper, a fixed field mirror split off the light for the
spectroscopy channel. The remaining part of the field of view passed into the photometry
channel. A footprint of the PACS focal plane is shown in Fig. 2.3. Proper centering of
a celestial target in the photometer or spectrometer field of view was therefore achieved
by offset pointing the telescope wrt. its optical axis. These offsets were determined during
the instrument focal plane geometry calibration and are listed in the Satellite/Instrument
Alignment Matrix (SIAM, see Herschel Satellite Handbook) as PACS instrument apertures
P01 0 (photometer; Y = 12.5000 , Z = 631.5100 ) and P02 0 (spectrometer; Y = 7.0500 ,
Z = 537.33). Additional angular offsets, that had to be taken into account, were the half
step sizes of chopper throws and telescope nodding steps.
Figure 2.3: PACS telescope focal plane lay-out. The axes give the angular offset in the
telescope coordinate system wrt. the telescope optical axis when looking towards the focal
plane (when looking towards the sky, signs are reversed). Photometer and spectrometer field
of view were separated by the fixed field mirror of the front optics. The smaller spectrometer
field of view was offset in the -Z direction by Z = 94.200 . Chopping was along the Y axis
direction. On both sides of the sky area in the focal plane, the fields of the internal calibration
sources were reachable only by the chopper. Typical chopper throws on the sky for photometry
and spectroscopy are given in Table 2.1.
2.2.2
Calibration sources
PACS contained two internal calibration sources (CS): Two grey body radiators with emissivity of about 4 8% (TBC), which were placed near the entrance of the instrument, outside
of the Lyot stop (see Fig. 2.2). They illuminated the full detector field of view, when the
chopper was deflected to its maximum angles in both directions, providing approximately the
same light path for observation and internal calibration.
The design of the internal calibration sources is shown in Fig. 2.4. The emitter was realized
by a platinum resistance PT500. The required temperature of the grey body was achieved
by a heater current flowing over the constant kept resistance. The goal was that the two
20
internal calibration sources should produce radiation loads slightly below (CS1, 48, 55 K)
and above (CS2, 58, 60 K) the telescope background at around 100m. Both CS loads were
below the telescope background for shorter wavelengths and above the telescope background
for longer wavelengths due to the difference in emissivity and hence also the to be selected
temperature. The resistance measurement of the heater allowed a direct determination of the
temperature by a four wire measurement technique. With a 15 bit resolution of the controller
the temperature stability of <5 m corresponding to 5 mK was guaranteed. The heat-up of
the sources was achieved by ballistic heating and needed about 40 min until full temperature
stabilization of the sources was achieved. This was incorporated into the general instrument
and detector set-up and took place before a PACS observation block. The sources were kept
illuminated until the end of the PACS observation block. The cool-down after switch-off
of the sources at the end of the PACS observation block was purely passive and lasted in
the order >12 h. A detailed determination of heating and cooling time constants is given in
Dannerbauer et al. (2007, PICC-MA-TR-033, Issue 1.1, Functional and Performance Test
of PACS Internal Calibration Sources during cold FM-ILT).
Figure 2.4: Design of the PACS internal calibration source (CS): The radiation emitter at
the lower left hand side was a platinum resistor. The four feet carrying the emitter was
a glass-fiber board to reduce the thermal heat loss due to low thermal conductivity. The
radiation was distributed via a scatter plate in front of the emitter into the integrating sphere
to increase the homogeneity at the pupil. The baffle cone at the right hand side including a
lens generated the illumination patch covering the field-of-view of the detector.
The utilization of the internal calibration sources for the photometer flux calibration is described in Sect. 6.4.2.1.
2.2. COMMON OPTICS
2.2.3
21
Chopper
The PACS focal plane chopper was a movable optical element in the PACS instrument light
path allowing to move the centre of the field-of-view in the focal plane. Such an element could
be used to
1) Modulate the incoming beam in order to subtract the background emitted by the 85 K
warm telescope, including any drifts and variations of this background slower than the
chopper frequency. This is usually done by chopping between a position centered on
the source and a nearby background reference position on the sky in square wave mode.
2) Mirror one of the two internal calibration sources into the detector field-of-view or
perform a square wave chopping between both sources.
3) Perform step scans with step sizes down to 1 to explore the homogeneity and borders
of the sky and internal calibration source field-of-views.
Design requirements for the chopper were:
High precision of the end position (0.75 on sky position and 1.5 on internal calibration source. This corresponds to 0.25 and 0.5 of a blue photometer pixel,
respectively.
Allow chopper frequencies up to 10 Hz
High duty cycle in square wave modulation (80% for sky chopping at 10 Hz, 70% for
chopping between the internal calibration sources at 5 Hz)
Low power dissipation (<4 mW)
Low micro-phonic noise
Very long lifetime under cryogenic conditions (execute around 630 million chopper cycles)
The design of the chopper and its essential elements is shown in Fig. 2.5. A detailed description
of the PACS chopper hardware is given in Krause et al. (2006, SPIE 6273, The Cold Focal
Plane Chopper of HERSCHELs PACS Instrument).
The angular calibration of the chopper (i.e. the relation field plate read-out to angular deflection) was established by the manufacturer (Carl Zeiss AG Oberkochen) with the help of a
laser measurement. The discrete measurements were interpolated with polynomial functions
and the Zeiss calibration was transferred into the PACS instrument system taking into account the different amplification factor (Nielbock et al. 2007, Angular calibration and zero
point offset determination of PACS FM1 chopper for cold HeII (T = 4.2 K) conditions, PICCMA-TR-021, issue 1.0). The angular calibration was verified on sky with the help of double
infrared sources with well known separation. Typical angular deflections and the related
position sensor read-outs used during the mission are given in Table 2.1
In order to achieve the fast transition times and high plateau position accuracy with only
small overshoots an active control loop of the drive had to be developed. The implementation comprised a PID loop, a velocity loop, a current feedback loop and a notch filter
22
Figure 2.5: Design of the PACS chopper showing its essential elements: The rotation axis
of the gold coated mirror was defined by two flexible pivots on both sides (rotation angle
< 10o ). Three drive coils provided redundancy for the rotational elongation of the mirror.
The rotation angle was monitored by field plates of the position sensor unit and fed back
to the chopper control electronics. Three mechanical interface points allowed to mount the
chopper accurately wrt. the PACS light path.
for suppression of axial resonances1 . The chopper control loop parameters were optimized
during a number of on-ground and in-flight tests. One performance example of the chopper
during the commissioning in-flight is shown in Fig. 2.6, details are given in Bouwman et al.
(2009, PACS Chopper Dynamic Behaviour Verification (HCOP PAC CHP2 01) during CoP
on June 1, 2009, PICC-MA-TR-083). The finally applied control loop parameters are stored
in calibration uplink file CONFChopper. The shortest chopper transition times for the best
performing set of control loop parameters are listed in Table 2.2. For a chopper frequency
of 5 Hz, the duty cycle was 80% on sky and 78% on the internal calibration sources. In the
photometric mode, the chopper was operated with a frequency of 1.25 Hz (95% duty cycle).
In-orbit performance checks revealed that the overall control loop gain was dependent on the
temperature of the DECMEC warm electronics (Nielbock et al. 2009, Analysis of Nominal HeII Open Loop Functional Tests of the PACS FM1 Chopper during the Commissioning Phase,
PICC-MA-TR-088, issue 1.1), resulting in less good performance for low temperatures. The
PACS warm electronics panel was therefore heated to 0o C during all PACS operations,
ensuring that the established control loop parameters gave the optimum performance.
A detailed description of the chopper control is given in (1) DECMEC User Manual (PACS-CL-SR-002)
and (2) PACS Calibration Document, req. 2.3.2 (PACS-MA-GS-001)
2.2. COMMON OPTICS
23
Table 2.1: Typical angular deflections used for photometer, spectrometer and internal calibration measurements.
deflection
angle on sky
(arcsec)
position sensor
read-out
(ADU)
664
-26.0
26.0
-29.8
29.8
-87.3
87.3
-176.2
176.2
-1146
2465
-1395
2710
-5483
6703
-11419
12296
-21350
21200
position reference
optical zero position

uplink table PHOT CHOP sky
chop/nod point source mode photometer
uplink table PHOT CHOP params
chopper throws for spectrometer science modes
corrected for distortion/rotation effects
uplink table SPEC CHOP params
CS1, uplink table CHOPPERCSparams

CS2, uplink table CHOPPERCSparams
Table 2.2: Final performance numbers of the PACS chopper for the transition times into the
accurate plateau position for the sky range (<15000 ADU) and for the internal calibration
source range (>15000 ADU).
deflection angle
(ADU)
3000
12000
18000
21000
chopper transition time

into negative plateau into positive plateau
(ms)
(ms)
17
19
21
22
17
20
21
22
24
Figure 2.6: Example of a PACS chopper performance test during in-flight commissioning.
The upper line shows the position read-out (in ADU), the lower line the drive current (in
mA). The first picture in each line displays the whole sequence, the other two the enlargement
of the critical swing-in phase with all chopper plateaux of the sequence overlaid. Due to the
diagnostic telemetry frequency of 1 kHz (corresponding to 1 ms temporal resolution), some
slight jitter between the different plateaux can be recognized. In the position read-out, blue
symbols represent the actual sensor read-out, while black symbols represent the stimulus. The
red horizontal lines display the corridor to achieve the required plateau position accuracy.
2.3. PHOTOMETER
2.3
2.3.1
25
Photometer
Filters
Far-infrared filters defined the photometric bandpasses of the PACS photometer Fig. 2.7 shows
the filter sets used in the two simultaneously operated blue and red branches by splitting short
and long wavelength radiation via a dichroic. The filter FL E in the top optics defined the
Figure 2.7: PACS photometer filter scheme for the red branch (top; >125 m with dichroic
in transmission) and the blue branch (bottom; <125 m with dichroic in reflection). The
blue branch contained a filter wheel which allowed to switch between the blue bandpass (60
85 m) and the green bandpass (85 125 m). This allowed the simultaneous observation in
one of these bandpasses and the red bandpass (125 210 m) with two bolometer cameras.
26
short wavelength edge of the overall PACS wavelength range (52 210 m). The blue branch
contained a filter wheel which allowed to switch between the blue bandpass (60 85 m) and
the green bandpass (85 125 m). The bandpass transmission is achieved by combination
of low-pass and high-pass edge-defining filters. In the blue branch, the filter combination
FBBP 1/FBBP 2 in front of the blue bolometer camera confined the whole blue branch range
(60 125 m), while the filter combinations FBBP2 1/FBBP2 2 and FBBP1 1/FBBP1 2
confined the proper blue and green passbands, respectively.
All filters were realized as metal mesh filters with inductive or capacitive grid geometry, either
as air-gap or hot-pressed types, and provided by Cardiff University. A review of these metal
mesh filters is given in Ade et al. (2006, SPIE 6275, 62750U-1). A few examples of PACS
flight model filters are shown in Fig. 2.8.
Figure 2.8: Examples of PACS flight model metallic mesh filters provided by Peter Ade and
colleagues from Cardiff University. Top left: Short-wave blocking filter FL E of the top optics
(diameter 36 mm). Top right: Dichroic beam splitter D P 1 (diameter 56 mm). Bottom left:
Low pass edge-defining filter FBBP2 1 for blue bandpass (diameter 20 mm). Bottom right:
High pass edge-defining filter FBBP2 2 for blue bandpass (diameter 20 mm); this filter is
very transparent in the visual. Image source: Geis (2005, PACS-ME-ISH-021, Incoming
Inspection of PACS FM Filters).
2.3. PHOTOMETER
27
The transmission of all PACS photometer filters was measured by transmission spectroscopy.
A complete overview is given in M
uller et al. (2011, PICC-ME-TN-038, v1.0, PACS Photometer Passbands and Colour Correction Factors for various Source SEDs). Here the example for the blue bandpass (60 85 m) is shown in Fig. 2.9.
Figure 2.9: Transmission measurements for the PACS flight model (PACS PFM) filter components of the blue bandpass (60 85 m). Note, that the x-axis is in wavenumber wn = 1 ;
wn = 100 cm1 corresponds to = 100m. From M
uller et al. (2011).
2.3.1.1
Effective spectral response
The effective spectral response of each bandpass is the product of the dichroic transmission/reflection, the transmission of all edge filters and the bolometer detector absorption. An
example of the product of the former two is shown in Fig. 2.9. The bolometer detector absorptions are based on reflection measurements (absorption = 1 ref lection). The detector
absorption was relatively flat and above 90% in the blue branch and dropping to around 80%
for the red branch (see M
uller et al. 2011).
Fig. 2.10 shows the resulting spectral response of the three bandpasses, the blue
or 70m band (60 85 m), the green or 100m band (85 125 m) and the
red or 100m band (125 210 m).
The filter transmission is stored in calibration file photometer.filterTransmission and the detector absorption in calibration file
photometer.absorption. The detailed shapes of the spectral response curves are used to
derive proper colour correction factors (see Section 6.5.3).
28
Figure 2.10: Effective spectral response (product of filter transmission and bolometer detector
absorption) for the three photometer bandpasses. The reference wavelength 0 (cf. Sections ??
and 6.5.3) is indicated for each bandpass.
2.3.1.2
Filter wheel
The filter wheel assembly is shown in Fig. 2.11. It comprised a disk with the two filter
positions, a magnetic motor drive, Hall effect sensors for position detection and 12 magnets
for filter wheel rotor positioning. Filter wheel mechanism and magnetic motor were mounted
to an adapter plate.
Design requirements for the filter wheel were:
Precision of end stop: 300
Transition time: 5 s
Power dissipation per position change: <50 mW
Warming of parts within view of detector: <0.1 K
Operational lifetime: 20 000 cycles
A description of the filter wheel movement is given in the DEC/MEC User Manual (PACSCL-SR-002).
The design of the filter wheel for the blue photometer branch was identical with the one of
the order sorting filter wheel for the blue spectrometer branch.
2.3. PHOTOMETER
29
Figure 2.11: PACS filter wheel assembly. Left: Design of the filter wheel, showing the disk
with the two filter positions and the magnetic motor drive underneath. Right: Image of a
flight model filter wheel coated with black paint.
2.3.2
Bolometer arrays
30
2.3.3
Cooler
A 3 He sorption cooler produced the operational temperature of 285 mK for the bolometer
arrays. The Herschel flight model cooler is described in detail in Duband et al. (2008, Cryogenics 48, 95, Herschel flight models sorption coolers). Fig. 2.12 gives a 3D view of the
device showing its main components. The evaporator contained a porous material, an alumina sponge (91 % Al2 O3 / 9 % SiO2 ), which trapped the liquid 3 He during the cold state.
This liquid 3 He evaporated providing the cooling to the detector focal plane. When all the
liquid 3 He had been evaporated into the gas phase, it needed to be recycled. The gaseous
3 He flew into the sorption pump which contained active charcoal for adsorption of the gas.
An important monitoring parameter both for the cooler recycling procedure and the detector
operation during the cold phase is the evaporator temperature TEMP EV.
Figure 2.12: Overall 3D view of the PACS cooler identifying its main components. The
bottom figure presents an exploded view identifying the elements represented schematically
in Fig. 2.13.
2.3. PHOTOMETER
2.3.3.1
31
Procedure of the Cooler Recycling
Fig. 2.13 right shows the evolution of the temperatures relevant for the cooler during the
different steps of the recycling process as monitored via Housekeeping (HK) parameters from
temperature sensors. At the beginning, the evaporator temperature (TEMP EV) was around
2 K indicating that the cooler had run out of liquid coolant. The recycling procedure was
completely controlled via heaters and heat switches integrated in the cooler (Fig. 2.13 left).
The heat switches were of the gas-gap type, where the presence or absence of gas between two
interlocked copper parts changed the heat flow between them. Gas handling was achieved by
means of a miniature cryogenic adsorption pump (Duband et al. 1995; A thermal switch for
use at liquid helium temperature in space-borne cryogenic systems. In: Proceedings of the
8th International Cryocooler Conference, p. 731. Plenum Press, NY).
Figure 2.13: Left: Schematic drawing of the PACS cooler elements and the thermal connections to the PACS bolometer detector focal plane unit (in green) and the liquid 4 He L0 bath
of the Herschel cryostat at 1.7 K. Right: Evolution of temperatures relevant for the PACS
cooler monitored by sensors and provided in the PACS instrument Housekeeping (HK) during
the cooler recycling process.
The first 15 minutes served to settle the thermal environment. The pump heat switch (HSP
in Fig. 2.13 left) had to be open, so that the pump did not dissipate heat into the instrument.
The evaporator heat switch (HSE in Fig. 2.13) was closed so that the evaporator thermalized
with the 2 K level and did not warm up too much when the pump was heated and thus still
trapped condensing gas. The heat switches HSP and HSE were closed by applying a current
and opened by setting the current to zero.
The recycling proper started with heating the sorption pump (SP) to desorb the gas that
had been trapped in the active charcoal. After 35 min the heater current was lowered to keep
the pump at the required temperature of about 40 K. At this time 3 He out-gased from the
32
pump and condensed. TEMP EV rose following the temperature rise of the pump due to the
enthalpy of the hot gas coming from the pump and not all being removed by the thermal shunt
and also due to the latent heat of 3 He. Since the evaporator and the shunt were connected
to the same thermal strap, all variations happening at the shunt were also registered by the
evaporator temperature sensor (TEMP EV). TEMP EV decreased when the pump was kept
at the required temperature. At around 80 min TEMP EV dropped below the 2 K level. The
finally achieved TEMP EV in this step characterized the efficiency of the recycling.
After 82 min the evaporator heat switch (HSE in Fig. 2.13) was opened (HSE = 0 mA) to
thermally isolate the evaporator. To establish the cooling functionality the pump heater was
switched off after 94 min and re-connected to the 2 K level by closing the pump switch (HSP
in Fig. 2.13). Once the pump was connected to the 2 K level the charcoal in the pump started
pumping and the 3 He pressure dropped. The thermally insulated liquid 3 He decreased in
temperature to below the 300 mK level which was reached after around 142 min.
The cooler recycling procedure with the tuned time steps as described above proceeded fully
automatically. The duration of the PACS-only cooler recycling block was 142.37 min, the duration of the parallel mode cooler recycling block 172.53 min. The procedure was implemented
by means of the Herschel Common Uplink System (CUS). It is documented in Sauvage et al.
(2014, Exp. Astronomy 37, 397, Operations and performance of the PACS instrument 3 He
sorption cooler on board of the Herschel space observatory).
2.3.3.2
Establishment of the Cooler Hold Time Relation for Herschel Mission

Planning
For an efficient Herschel Mission Planning it was necessary to have a reliable prediction of
the cooler hold time thold for PACS photometer operations. The cooler hold time was defined
as the period between TEMP EV going below 300 mK at the end of the recycling process (cf.
Sect. 2.3.3.1 and Fig. 2.13) and TEMP EV exceeding 320 mK when the cooler was exhausted.
The essential parameter determining the length of the cooler hold time was the time, when
the PACS bolometer detectors were biased for measurement. This period was set by the socalled orbit prologue and orbit epilogue, engineering type Astronomical Observations Requests
(AORs) setting and resetting the detector bias and bracketing the sequence of science AORs.
The so-called biased time tbias was defined as the period between the start of an orbit prologue
and the end of the subsequent orbit epilogue.
Such a dependence was monitored right from the beginning of the mission and after about
three quarters of a year there was enough statistics to derive the relation
thold (h) = 72.97 h 0.20 tbias (h)
(2.1)
established from all complete cooler periods up to OD 270, irrespective of the start conditions
or whether it was a PACS only or parallel mode cooler recycling, was reliably used in the
scientific mission scheduling until the end of the mission with a last PACS cooler recycling
on OD 1443 (cf. 2.15). The formula was applied for determination of the cooler hold time
following both a PACS only and a parallel mode (together with the SPIRE cooler) recycling.
In Eqn. 2.1 thold can be written as thold = tidle + tbias with tidle being the fraction of the hold
time period with the PACS bolometer detectors not biased, e.g. during spacecraft operational
2.3. PHOTOMETER
33
maintenance windows or SPIRE only operations in parallel mode.

In case tidle 0, then eqn.(1) can be re-written as
tbias,max (h) =
72.97
(h) = 60.8 h
1.2
(2.2)
This meant that a contiguous block of about 2.5 ODs of photometer observations could be
scheduled, thus minimizing the cooler recycling frequency. The remaining 0.5 OD was usually
filled with PACS spectrometer observations, not requiring any 3 He cooling, by switching
between the two PACS sub-instruments.
Since the cooler hold time relation in Eqn. 2.1 was determined for the point in time when
the evaporator temperature exceeded an out-of-limit value of 320 mK, a safety time buffer
was included when using this relation for mission planning, thus reducing tbias,max by tbuffer .
Since the evaporator temperature increased very steeply only at the end of the cooler hold
time (see Fig. 2.14), a buffer of 1.5 h was deemed sufficient initially. However, inspection of
photometer calibration observations, scheduled deliberately at the end of the cooler period for
cross-calibration with subsequently scheduled PACS spectrometer observations of the same
target, showed that there was already an increase in the evaporator temperature (cf. the
course of the temperature in Fig. 2.14 with an increase of up to 5 mK for the cycle labeled
A093) with an impact on the photometric calibration accuracy. This effect is described in
detail in Mo
or et al. 2014 (Exp. Astronomy 37, 225, PACS photometer calibration block
analysis) and characterized by appropriate correction functions (see Sect. 6.4.2.3) which put
all photometric measurements on a homogeneous temperature reference level. Despite the
final calibration and correction of this effect, it was decided to increase tbuffer to 3 h, which
meant that the evaporator temperature increased by at most 1 mK at the end of the cooler
period (see the cycle labeled A133 in Fig. 2.14. This meant a maximum contiguous PACS
photometer operation of 57.8 h.
2.3.3.3
Mission Statistics of Cooler Recycling
A total number of 239 PACS cooler recyclings was performed during the Herschel mission, the
first one on OD 26 for PACS commissioning and the last one on OD 1443. 139 cooler recyclings
were performed as PACS only, 100 recyclings in parallel with the SPIRE cooler recycling. It
should be noted that the parallel cooler recyclings were not only performed for parallel mode
observations of PACS and SPIRE, but also for separate PACS and SPIRE observations during
the subsequent hold time periods. The SPIRE cooler hold time was restricted to about 48 h,
the difference between the PACS and SPIRE cooler hold times was usually covered by PACS
photometer observations to achieve the highest possible efficiency of each cooler recycling
with regard to the foreseen and available photometer observing program.
Fig. 2.15 shows the relation of the cooler hold time versus the biased time for all complete
(with final exhaustion of the liquid 3 He) cooler cycles during the mission (192 out of 239, 115
out of 139 PACS only, 77 out of 100 parallel). The plot all complete hold times of the entire
mission shows some fine differences between the different modes and start conditions. The
PACS-only recyclings starting from a warm cooler cluster close to the relation established
from all complete cooler periods up to OD 270 (red line) with a dispersion of 0.5 h. The
parallel cooler recyclings starting from a warm cooler are slightly shifted to an about 1 h
34
A093 / w
A133 / w
0.295
0.295
3
TEV [K]
0.290
0.285
TEV [K]
TEV [K]
TEV [K]
2
0.280
10
20
30
t [h]
0
0
6
t [h]
40
50
0.290
0.285
2
0.280
60
10
20
30
t [h]
10
0
0
40
50
60
10
t [h]
Figure 2.14: Individual PACS-only cooler cycles (labeled A plus a sequence number): The
figure in the main panel shows the course of the evaporator temperature (T EV) during the
first 10 h following the start of the recycling, whereby the red part represents the proper
recycling process, the blue part the beginning of the subsequent operational period. The
inserts are a zoomed view with adapted dynamic range of T EV over the full operational
period. Left: Cycle A093 on OD 842 with a maximum contiguous bias period of nearly 59.2 h
(tbuffer = 1.5 h in hold time calculation). Right: Cycle A133 on OD 1354 with a reduced
(tbuffer = 3.0 h in hold time calculation) contiguous bias period of 57.8 h. For the latter one
the final steep temperature increase is less than 1 mK.
shorter cooler hold time with a similar dispersion. This one hour less efficiency was well
covered by the buffer time applied for mission planning aspects. Statistics for the cooler
recyclings starting from a still cold cooler are poorer, but there is some indication that the
hold times both for PACS-only and parallel cooler recycling were about 0.5 h longer than for
the warm start PACS-only cooler recyclings.
A more detailed statistics and description of the PACS cooler holdtime is given in Sauvage
et al. (2014, Exp. Astronomy 37, 397, Operations and performance of the PACS instrument
3 He sorption cooler on board of the Herschel space observatory). The PACS and SPIRE
cooler recyclings were also utilized to determine the lifetime of 4 He in the Herschel cryostat.
A description of this method is given in the Satellite Handbook in section Estimation of the
helium lifetime from cooler re-cyclings.
2.3. PHOTOMETER
35
Figure 2.15: Statistics of the cooler hold time versus bolometer biased time over the entire
Herschel mission. Hold time and biased time are defined in the text at the beginning of
Sect. 2.3.3.2. Different symbols and colors represent PACS only or parallel mode cooler
recyclings and the start conditions from a warm, i.e. exhausted liquid 3 He, or a cold, i.e. still
available liquid 3 He, cooler. The operational guideline established from the relation of all
cooler periods up to OD 270 (Eqn. 2.1) is shown as the red line.
36
2.4
Spectrometer
2.4.1
Instrument design
2.4.2
Image slicer
2.4.3
Grating
2.4.4
Order-sorting filters
2.4.5
Photoconductor arrays
2.5
On-board data handling
2.5.1
Data gathering
2.5.2
On-board data reduction
2.5.3
On-board data compression
Chapter 3
Observing with PACS

3.1
Introduction
Any observation with PACS (or any of the Herschel instruments) was performed following an
Astronomical Observation Request (AOR) made by the observer. The AOR was constructed
by the observer by filling in an Astronomical Observation Template (AOT) in the Herschel
Observation Planning Tool, HSpot. Each template contained options to be selected and
parameters to be filled in, such as target name and coordinates, observing mode etc. How to
do this is explained in details in the HSpot users manual link and the PACS AOT release
notes (link) and Observers Manual (link). In this chapter we do not explain how to fill in
an AOT (since that is no longer relevant), rather we explain the choices the observers had to
make when creating their AORs, and what the resulting observing modes created were.
If the observation request was accepted via the normal proposalevaluationtime allocation process then the AOR content was subsequently translated into instrument and telescope/spacecraft commands, which were up-linked to the observatory for the observation to
be executed. One special feature of PACS commanding was the use of On-board Control Procedures (OBCPs), command macros composed of a logical sequence of low level commands
needed to execute a certain type of observation, which were stored on-board. This reduced
the telecommand bandwidth considerably, because only the name of the OBCP and the related parameters, whose values depended on the AOT parameter selection of the observer,
had to be uplinked as part of the Mission Time-Line. The OBCPs were therefore essential elements of the AOT commanding design. A complete list of all PACS OBCPs and DEC/MEC
sequences is given in document PACS-ME-LI-005 (Issue 2.2, 14-Apr-2010).
The AOTs evolved in the first year of the mission check timeline of changes, as more efficient
ways to carry out observations were discovered, and observations carried out during the
Performance Verification (PV) and Science Demonstration (SD) Phases (during the first half
year of the mission) had slightly different observing sequences. Here we concentrate on the
final version of each AOT, which were those used for most of the Routine Science (RS) Phase.
Building Blocks: Observations were made up of logical operations, such as configuring the
instrument, initialisation and science data-taking operations. These logical operations are
referred to as building blocks (or just blocks). The science operations were usually repeated
several times to achieve the requested signal-to-noise ratio (SNR) and/or to map a given sky
37
38
CHAPTER 3. OBSERVING WITH PACS
area. Pipeline data reduction modules work on building blocks (see Chp. 7), and the data
during the pipeline processing is organised around building blocks.
3.2
Photometer observations
Since the noise of the bolometer/readout system had a strong 1/f component, photometer
observations had to be designed such that the signal was modulated in the frequency band
from 1 to 5 Hz in order to achieve optimum sensitivity. Signal modulation was either possible
by using the focal plane chopper or scanning with the whole telescope. In the pre-flight design
of the PACS photometer AOTs (see Sauvage & Okumura, Preparing AOTs for the PACS
photometer, SAp-PACS-MS-0186-03, Issue 2.1, 01-Dec-2005) utilization of the chopper was
considered to be more efficient. Beside a chop-nod point-source AOT, a small source AOT
and a chopped raster AOT were designed based on a combination of telescope raster execution
and chopping. These three modes were tested during the AOT validation in the Performance
Verification Phase together with the scan map technique with the chopper mirror staring in
the optical zero position. The AOT validation tests during the Performance Verification Phase
revealed quite soon, that the scan map technique was superior to the chop/nod technique.
Chopped small source and raster mode were therefore abandoned and made inaccessible in the
HSpot user interface. The chop-nod point source mode was maintained, because it provided
accurate astrometry for the Herschel Pointing Calibration program (performed throughout the
mission) and an independent reliable and low cost observing method to test the photometer
flux calibration consistency. With the invention of a special point-source mini-scan map
mode, utilization of the chop/nod point source AOT was no longer recommended to the
general observer, but it was exclusively used by the PACS ICC calibration team.
Three photometer AOTs were fully validated and released to the observer community
1) Point-source photometry in chopping/nodding technique with the restrictions outlined
above (see PACS Photometer - Point/Compact Source Observations: Mini-Scan Maps
& Chop-Nod release note, PICC-ME-TN-036, issue 2.0, 12-Nov-2010)
2) Scan map technique, both for large maps and dedicated mini maps for point source photometry (see PACS photometer - Prime and Parallel scan mode release note, PICCME-TN-035, issue 1.1, 23-Feb-2010 and PACS Photometer - Point/Compact Source
Observations: Mini-Scan Maps & Chop-Nod release note, PICC-ME-TN-036, issue
2.0, 12-Nov-2010)
3) Scan map technique within the SPIRE/PACS parallel mode (see PACS photometer Prime and Parallel scan mode release note, PICC-ME-TN-035, issue 1.1, 23-Feb-2010)
All photometer configurations performed dual band photometry with the possibility to select
either the blue (60 85m) or the green (85 125m) band for the blue channel, the red
band (125 210m) was always included. The two bolometer arrays provided full spatial
sampling in each band.
Source flux estimates given by the observer drove the selection of the photometer gain setting
for the ADC conversion. Default was the high gain, while low gain offered a larger
dynamic flux range. Above a certain flux threshold as given in Table 3.1, the AOT-to-ICP
3.2. PHOTOMETER OBSERVATIONS
39
(Instrument Command Parameter) logic switched to the low gain setting. The given flux
thresholds are quite conservative estimates, which should ensure that saturation occurred
only for a small percentage of pixels, even for the detector matrices with the highest signal
dispersion. More homogeneous behaving matrices allowed to measure considerably higher
fluxes. The low gain setting did not give the nominal factor of 4 in the ADC signal
conversion due to saturation occurring in the read-out circuit prior to reaching saturation
of the ADC range. The extension of the flux range was therefore limited to less than a factor
of 2. A description of how the saturation estimates were derived is given in Sauvage et al.
(2008, SAp-PACS-MS-0680-08, issue 1.0, 22-Jul-2008).
Table 3.1: Bolometer readout saturation levels for the high gain setting. These numbers are
conservative to ensure that saturation was avoided for most of the detector pixels.
filter
point source
[Jy]
extended source
[GJy/sr]
blue
green
red
220
510
1125
290
350
300
In the In-flight Performance Chapter 4 measured maximum fluxes depending on the gain
setting are given.
3.2.1
Scan map
The scan technique was the most frequently used Herschel observing mode. Scan maps were
performed by slewing the Herschel telescope at a constant speed along parallel lines, as shown
in Fig. 3.1. Initial available satellite speeds were 10, 20, 6000 /s, but during the mission the
1000 /s was made inaccessible to the general observer, because the 2000 /s achieved the same
performance within shorter time. The number of scan legs, the scan leg length, the scan leg
separation, and the orientation angles (either in array or sky reference frames) were freely
selectable by the observer to adjust the observation to required map size and sensitivity depth.
The sensitivity could also be increased via a repetition parameter repeatedly executing the
scan map n times. For optimizing the map cosmetics (destriping) it was recommended to
execute a cross scan map. As indicated in Fig. 3.1 an optimum orientation of the array for
large maps was 45o (and 135o cross scan) in the array reference frame to avoid artifacts by
the inter-matrix gaps. The performance for a given map configuration and repetition factor
could be evaluated and optimized during observation planning via sensitivity estimates and
coverage maps in HSpot.
3.2.1.1
Point source photometry (mini-scan map)
A special map design was invented for point source photometry with scan maps. The orientation of the array was here rather 70o in scan and 110o cross scan moving the photometer
arrays along their diagonals. Scan leg separations of 3400 provided a very high map coverage
40
Figure 3.1: Illustration of the photometer scan map scheme with an example of six scan legs.
After finishing the first leg, the telescope turned around and continued with the next scan line
in the opposite direction. The specified reference scan direction was the direction of the first
leg. An optimum orientation of the PACS blue photometer array wrt. the scan direction is
indicated. The rotation by about 45o in the array reference frame improved the map coverage
and avoided artifacts by the inter-matrix gaps.
and redundancy with many pixels seeing the source. The scan leg length was in the order of
30 and 10 scan legs provided a sufficient width of the map. Compared to the chop-nod point
source observation mode (see below), the advantage of the mini-scan map mode was a better
characterization of the source vicinity and larger scale structures in the background due to a
larger useful area, more homogeneous coverage inside the final map, the higher redundancy
wrt. the impact of noisy and dead pixels and the better sensitivity of the scan technique.
3.2.2
Point-source photometry (chop-nod)
The chop-nod point source mode used the PACS chopper to move the source on-array by
about 5000 , corresponding to the size of about 1 blue/green bolometer matrix (16 pixels) or
the size of about half a red matrix (8 pixels), with a chopper frequency of 1.25 Hz. The
nodding was performed by a satellite movement of the same amplitude but perpendicular to
the chopping direction. This chop-nod scheme is shown in Fig. 3.2. The offset of 5000 between
the source images on the array meant that this was approximately the side length of the
useful sky area.
On each nod-position the chopper executed 325 chopper cycles. The 3 sets of chopper
patterns were either on the same array position (no dithering) or on 3 different array positions
(dither option). In the latter case the chopper pattern was displaced parallel to the chopper
deflection by 8.500 (2 23 blue pixels or 1 31 red pixels). Each chopper plateau lasted for 0.4 s
(16 readouts on-board) producing 4 frames per plateau in the telemetry downlink. The full
325 chopper cycles per nod position were completed in less than 1 minute. In case of
repetition factors larger than 1, the nod-cycles were repeated in the following way (example
3.3. SPECTROMETER OBSERVATIONS
41
Figure 3.2: Illustration of the detector footprint of the blue detector array on the sky (left)
and the chop-nod source pattern (right) produced on the detector during the execution of the
chop-nod point source AOT. The detector array with a field of view of 3.50 1.750 consisted
of eight individual sub-matrices. The source was offset by about 5000 horizontally by chopping
(no dithering) and also by about 5000 vertically by the telescope nodding. The colours reflect
the four combinations of the nodding and chopping positions attained during the observing
sequence: black: nodA chop1, red: nodA chop2, blue: nodB chop1, cyan: nodB chop2.
for 4 repetitions): nodA-nodB-nodB-nodA-nodA-nodB-nodB-nodA to minimize satellite slew
times.
3.2.3
SPIREPACS parallel
The SPIREPACS parallel scan mapping mode allowed to obtain simultaneous five filter
photometry in two PACS bands (70 or 100m and 160m) and the three SPIRE bands
(250, 350, and 500m). Due to the fixed 210 separation of the PACS photometer and SPIRE
photometer footprints in the Herschel focal plane (cf. Herschel Satellite/Instrument Alignment
Matrix in the Satellite Handbook) only large maps were efficient, i.e. had sufficient common
overlap, in this observing mode. Offered scan speeds were 20 and 6000 /s. The orientation angle
wrt. the array was fixed to the optimum SPIRE orientation angle, which was 42.4o and 317.6o
(42.4o ) for scan and cross scan. Compared to the PACS prime mode with the SPU averaging
4 subsequent frames for all filters, in the SPIRE/PACS parallel mode 8 subsequent frames
were averaged for the blue/green filter, while for the red filter the averaging was identical with
the prime mode. This led to a noticeable smearing of the PACS point spread function, in
particular for the fast scan speed of 6000 /s. Also in the data compression step 1 bit more was
rounded for SPIRE/PACS high gain observations. For all other aspects the PACS instrument
set-up was identical with the prime mode.
3.3
Spectrometer observations
During a PACS spectrometer observation, data were collected continuously while various
parts of the instrument or telescope were moving: the grating, to cover the wavelength range
of the observation; the chopper (for chop-nod mode), to observe on-source and off-source sky
positions; and the telescope itself (for chop-nod and unchopped modes), nodding to observe
on-source and off-source sky positions or to move in a raster pattern. At any one point in
42
time data from only one grating position, i.e. one wavelength, were obtained, and successive
measurements where taken at a single grating position before the grating moved to its next
position, and so on until the entire requested wavelength range had been covered, twice. A
raw PACS spectrometer spectrum (at Level 0 of the pipeline) is in fact a time-line spectrum,
with units of counts vs. readout (which is a time counter).
The raw science data at Level 0 are called ramps. Each data-point is taken from an integration
ramp of 1/8 second length: the detector integrated the signal arising from the photon impacts
continuously for 1/8 of a second; these ramps were then fitted on-board PACS, and the signal
transmitted to the ground was the slope of the ramp in units of [ADU/reset interval].
Unchopped AOT release note of 20 sept 2010 it says Each integration lasts 1/8 s, resulting
in a grating scan which is faster than the one in chopping mode.. But in the OM it says The
extensive PV program aimed at optimising the AOT parameters led to ramps of 32 samples,
i.e. 1/8 second, for all all spectrometer AOTs.. So this implies all ramps are 1/8, and that
the unchopped ones are not faster. Which is the correct statement to make here and do I
need to modify what I say about the unchopped/chop-nod below? Double check that these
number apply to line and range/SED - as far as I can tell they do
All spectrometer observations began with a modulated chopped measurement on the two
internal calibration sources with the grating at a fixed position (at the key wavelength
nearest to the range requested in the science observation). The two sources were heated
to different temperatures, hence providing different signal levels (see Sec. ?? make sure list
key wavelengths there). The calibration block measurement started during the slew of the
spacecraft to the target in order to optimise the use of observing time. Data obtained in this
block was used for the flux calibration in one of the data-reduction pipelines (Sec. ??).
The choices that could be made by astronomers when designing their spectrometer observations were:
The method of sampling the background spectrum (a.k.a. telescope background): chopnod/chopping or unchopped
The spectral coverage and sampling: line spectroscopy (line-scan) or range spectroscopy
(range-scan)
The pointing: pointed or mapping
The two AOTs that were offered via HSpot for the PACS spectrometer were line and range
spectroscopy. (A wavelength switching mode was offered early in the mission, but was discontinued after the PV phase.) Within these two AOTs the observers could choose the
background measurement method, the pointing method, and a few other details. In this
section we will explain the choices the astronomers could make and what that meant for the
resulting observations.
3.3.1
The background measurement method
At the wavelength range Herschel operated at, and given the fact that the mirror was not
cooled, the primary source of emission in all PACS spectra is that of the telescope mirror.
For any observation of an astronomical source, the spectrum gathered includes the spectrum
43
of the source and that of the telescope background. To measure this background spectrum
to be able to subtract it out it was necessary to observe an off-source sky position free
of astronomical emission. Two observing methods were offered to allow this background to
be measured: chop-nod (a.k.a. chopping), and unchopped. The first was the main observing
mode: it used a rapid chopping between the on-source position and one of three possible
off-source positions, using an internal mirror (Sec. ??) to move between the two locations,
combined with a nodding movement of the telescope. The second mode was the choice for
observations of sources in crowded fields, where no suitable off-position within one of the
three standard chopping positions could be found, or if the source being observed was large;
instead, the telescope nodded at a much lower frequency between the on-source pointing and
a chosen off-source pointing.
3.3.1.1
Chop-nod mode
The chopper itself is shown in (Sec. ??). The chopper is an internal mirror that can flip at a
high frequency and so observe different light paths, and hence different parts of the skyp. The
mirror could flip between two positions (on-source and off-source), with a throw (separation)
of 1.500 (small), 300 (medium), or 600 (large). Now, the spectrum of the mirror varys with the
optical angle of the two chop positions, and hence the telescope background spectrum that is
measured in the two positions is not the same. To correct for this, an additional nodding of
the telescope was done, in such a way that the on-source chopper position in nod A was the
off-source chopper position in nod B, and vice versa: see Fig. ?? (OM Fig 6.3). The telescope
background can then be removed by combing the nod A and B data in such a way that the
two backgrounds are both subtracted:
F (end of pipeline) = ([N odAonoff ] + [N odBonoff ])/2
= ([(T 2 + Fsource T 1] + [(T 1 + Fsource T 2])/2
(3.1)
= F (source)
where F is a flux and T1 and T2 are the telescope background fluxes at the two chopper
positions. (In this term we assume the sky background is negligible compared to the telescope
background.) The nod sequence could be repeated within one observation to increase the
depth of the observation. Nod cycles were repeated in a way that A-B slew times were
minimised, for instance, in case of two repetitions the spacecraft follows the pattern: [A-BB-A].
The chopping was done at a high frequency so that the modulation could take place at each
grating step. Once the telescope had slewed to the first requested sky position (i.e. to the
source for pointed observations or the first raster position for mapping observations) and the
calibration block had been taken, a typical observing sequence was: the grating moved to the
first position of the requested wavelength range, the chopper was set to its on-source position
(chop+), and the telescope was in its first nod position (Nod A); two integration ramps were
recorded, and the chopper then flipped to its off-source position (chop-) to record another two
ramps this formed one On-Off sequence; two more ramps were taken in the chop- and then
another two in the chop+ position; the entire pattern (on-off-off-on) was repeated before the
grating moved to the next step.
44
The duration of such a grating plateau1 was [8 integrations] x [1/8 sec integration time] x
[2 on-off-off-on cycles] = 2 seconds check is for line and range and sed. After the grating
had moved over the entire wavelength range and back again, the telescope then moved to the
second nod position (Nod B) and the process was repeated, but now the chop+ position is
off source and chop- position is on source. This is shown in Fig. ?? (OM Fig 6.4) but need to
change the A and B there to chop+ and chop-.
We should mention here that an alternative calibration scheme, called the Telescope background normalisation method, was developed and that handled Eq. 3.1 differently: see Sec. ??.
3.3.1.2
The chop-nod AOT
The choices the observer had to make when designing the chop-nod part of their AOTs was:
the chopper throw, the number of times to repeat the nodding sequence.
The path of the chopper throw on the sky was curved. This resulted in a rotation of the
on-source sky footprint between the two nod positions, and this rotation become larger with
increasing chopper throw: see Fig. ??. The small chopper throw was hence the preferred
choice.
A chopper avoidance angle could also be set if a particular part of the sky was to be avoided
as the off position.
Increasing the number of nodding sequences could be done to improve the SNR of the final
spectra. However, it was also possible to increase the line repetition factor to improve the
SNR (this is discussed in Sec. 3.3.2), and this was more time-efficient: for a low number of
repetitions ( 6) it was recommended to repeat the line ather than the nodding sequence.
Include Fig 6.5 from OM?
3.3.1.3
Unchopped mode
For the unchopped mode there was no high-frequency chopping. Instead, the telescope background spectrum to be subtracted from the on-source spectra was taken from an off position
that the telescope moved to. The data-collection sequence in this mode depended on whether
the observer was using the line spectroscopy AOT or the range spectroscopy AOT. Once the
telescope was in position, the calibration block was taken and the the chopper was set to
its zero position. The grating then moved to its first position and integration ramps were
taken, four at each grating plateau rather than the two taken in the chop-nod mode. The
next grating position was moved to, and integration ramps taken again, etc., until the entire
requested wavelength range had been covered twice. say at what OD this switch to doing 2
by default occurred? also, I cannot find any mention speficially that 2 was the default but
the obsids in the ipipe scripts have this: check anyway.
For the range spectroscopy AOTs, once all the requested wavelength ranges and their repetitions had been executed, for pointed observations this was the end of the observation. For
mapping observations, the entire pattern was then repeated for the remaining raster pointings,
and then this was the end of the observation. To obtain off-source data for the unchopped
range scan mode, it was necessary to define another observation, ideally taken consecutively
with the on-source observation.
1
plateau refers to a single chopper, nod, or grating position where a sequence of data-points were collected
45
For pointed line spectroscopy AOTs, once all the requested lines and the repetitions had been
executed, the on-source block was followed by an off-source block. The telescope nodded to a
chosen off-source position, which had to fall within two degrees of the on-source position, and
the data for each requested line was again taken. The repetition factor for the off-source block
was always just 1, resulting in a shorter integration time and lower SNR. It was therefore
recommended that for such observations, the off-source spectra were smoothed before being
subtracted in the data-reduction pipeline. But we did not do this in the SPG, so should I take
this out?. The entire on-source-off-source block could be repeated. For mapping observations,
the observer could specify the number of raster positions to place between visits to the offsource block, e.g. for a 2 2 raster, asking for a repeat of 4 would result in the off-source
block taking place at the end of the raster.
One disadvantage of the unchopped mode is that it is less robust to instantaneous responsivity
changes caused by cosmic ray hits in the Ge:Ga detector pixels (see Sec. ??). To minimise
the effect of these response changes, the up/down grating scan was made faster in this mode
than in the chop-nod mode, and every line repetition requested by the observer was doubled
internally does this refer to the 4 rather than 2 integration ramps or doing a rep of 2 by
default - which all modes had by the end?. The duration of a single grating plateau was [4
integrations] x [1/8 sec integration time] = 1/2 seconds, i.e. 1/4 of that for the chop-nod
mode check: I looked at an unchopped obsid from the pipeline script and it has 5 ramps per
grating plateau, and looking at a chopnod in fact there are 17 in a plateau: so when was
1 extra added (the OM and release notes do not mention this?. Include? The number of
ramps and their duration per grating plateau is the same as for the line spectroscopy mode.
However, check: Can find no specific statement about the number of ramps in a plateau, but
in a chopnod range scan from pipipline there are 16 ramps per grating plateau so I assume
there is no difference here for chopnod or unchopped. Also, is the range doubled automatically
here also, or not? I can find no specific mention of this but looking at some data says that it
was.
Long-term responsivity changes can also occur during PACS spectroscopy observations (for
all observing modes, but it is primarily a problem for the unchopped mode), especially at
the beginning of an observation after the data-collection of the (fairly bright) calibration
source signals. Long-term response drifts mainly affect the level of the continuum, and for
the unchopped mode are the reason why the continuum uncertainty is higher than for the
chop-nod mode (Sec. ??). Long-term drifts basically mean that the response of the detector
is different during the observation of the off-source (telescope background) position compared
to that during observation of the on-source position. The effect of this drift can be minimised
by certain pipeline tasks (Sec. ??), but it is not corrected for perfectly. Imperfectly-corrected
response drifts result in the continuum level of the off-source data (the putative telescope
background spectrum) being wrong, and hence too much or too little background is subtracted
from the on-source data: resulting in a higher uncertainty in the recovered continuum level.
3.3.1.4
The unchopped AOT
When choosing the unchopped mode in their AORs, the choices to be made related to the
background sampling method were where the off position would be and how often to go there.
If using the range spectroscopy AOT, the off position had to be taken via a second observation.
This observation preferably had the same instrument settings as those of the on-source AOR,
46
indeed it was mandatory to request identical wavelength settings for the on-source and offsource observations but was this always followed?. Ideally, the off-source observation would
be as deep as the on-source observation, so as to not add noise when doing the background
subtraction and to minimise the effect of RSRF noise (Sec. ??). It was also recommended
to take the off-source observation within 60 minutes of the on-source observation, to minimise the effect of the long-term response drifts. In practise this was not always done. Do
some stats to see how often this was not met?, and in fact some observers did not request
an off-source pointing at all (presumably because the background continuum subtraction was
not important stats?). For mapping observations, often only one off-source observation was
requested check.
For SED AORs, with a Nyquist spectral sampling (Sec. 3.3.2), it was possible to split the
off-source observation into two observations, i.e. one before and one after the on-source observation, and both could then have half the on-source integration time. The on-source and
off-source AORs then had to be concatenated so that the HSC knew they were to be executed
as a single block. In the SPG, however, only one off-source observation is subtracted from an
on-source observation: Sec. ?? check if there are any actually with 2 offs, and if are: HPDP?
In later versions of HSPOT the observer could also set a switch to indicate whether the AOR
was the off-source or the on-source observation. This was for information purposes, and it
was not always set.
If using the line spectroscopy AOT, the observer had to define the off-position as either an
offset in arcseconds or the RA, Dec coordinates. For mapping observation, they also had to
specify the number of raster positions to fall between revisits to the off position. Since for
this mode the relative timing and duration of the off-source pointing was hard-wired, the only
other free parameter was the the number of nod cycles to request.
3.3.2
The spectral coverage and sampling
Three classes of spectral coverage could be chosen, of which ony two survived the entire
mission:
Line spectroscopy (a.k.a. line-scan). A limited number of relatively narrow emission/absorption lines could be observed for either a single spectroscopic FOV (4700 4700 )
or for a larger raster map. The fixed angular range scanned by the grating mechanism
was optimised for a given diffraction order to ensure the detection of the full profile
of an unresolved line with sufficient continuum coverage symmetric to the line centre.
Faint- and bright-line options were available (with the faint option being the default).
Range spectroscopy (a.k.a. range-scan). This is a more flexible and extended version
of the line spectroscopy mode, where a freely-defined wavelength range was scanned by
stepping through the relevant angles of the grating with selectable grating sampling
density (high or Nyquist). This AOT was used for full coverage of broad spectral lines,
several lines, or the entire SED.
A wavelength switching mode was offered during the first few months of the mission
but was then decomissioned in favour of the unchopped line spectroscopy mode.
47
Spectral ranges could only be defined in the AOTs for a set combination of spectral bands.
If the observer wanted to mix bands, they needed to create separate AORs and concatenated
them, so that they would be sure to be observed consecutively. As both detector arrays were
used at a same time, spectra were obtained from the prime (nominal) and the parallel ranges
together (i.e. from the red and blue camera at the same time).
As mentioned previously (Secs 3.3.1.3 and 3.3.1.1), the number of integration ramps per grating plateau was 16 in the chop-nod mode and 4 in the unchopped, and for both each ramp
lasted 1/8 s: a single plateau lasts 2 seconds in the chop-nod mode and 1/2 second in the
unchopped. From OD onwards, the grating runs were doubled internally by PACS, i.e. two
scans on the requested wavelength range were done for each requested repetition (values of
which ranged from 1 to 10). check
Recall that the PACS detector consists of 25 modules with 16 pixels each (Sec. ??). Each
module corresponds to a single spaxel, and the 16 pixels each gather data at the same time.
For each requested wavelength range, the spectrometer performed grating scans with an
amplitude such that a given wavelength was seen successively by all 16 spectral pixels of
the detector. The instantaneous spectral coverage in each pixel is thus a single wavelength.
However, the wavelength measured by each pixel was slightly offset from that measured by
the neighbouring pixels. This improved the spectral sampling obtained by the sum of all 16
pixels after the entire spectrum was combined. However, the spectral coverage of this summed
spectrum is slightly lower at the edges of the range, where not all 16 pixels contribute to the
spectral coverage, and hence the edges of the spectra have the lowest SNR. For line scans, the
region of highest sensitivity (the spectral range with uniform data coverage) is at least 4
times larger than the resolved line in each pixel at all wavelengths of an unresolved line. (The
dataset called coverage, found in the Level 2 cubes, is a measure of the number of data-points
in each spectral bin :link: PPE sec 3.2.1.5. For range scans, the region of uniform coverage is
larger, and for the full SEDs only a few wavelength bins at the spectral edges are so-affected.
3.3.2.1
Line spectroscopy
In line spectroscopy it was possible to chose a combination of 10 lines and/or line repetition
factors: anything from 10 separate lines with only 1 repetition (i.e. only one scan) to 1 line
with a 10 repetitions. The repetition factor refers to the number of consecutive grating scans
for the final AOT release (Jan 2010) the default was to do each grating scan twice (up and
then down), so for these a line repetition factor 2 meant that it was done four times at what
OD did this switch happen?. This could be combined in the AOR with any number of nod
repetitions (explained in Secs 3.3.1.2 and 3.3.1.4). The spectral coverage ranged from 0.35 m
in the blue order 3 to 1.8 m in the red.
The sampling density per resolution element is higher than 3 samples per FWHM of an
unresolved line at each of the 16 pixels of each module and at all wavelengths. The scanning
parameters, including grating steps, sizes, wavelength coverage, range, and spectral density,
are given in Table 3.2. Fig. ?? is the visualisation of a line scan AOT on an unresolved
spectral line. from OM 6.6. For the unchopped mode, more steps where taken irrespective of
the spectral band, resulting in a longer wavelength coverage with the same spectral sampling,
however as less time was spent at each grating position, the overall integration time was
similar for the two modes. All other grating and wavelength details are the same as for
48
the chop-nod mode check - there is no table for unchopped mode in OM but at least the
wavelength range must be different, so table will need updating.
Table 3.2: The grating, wavelength range, and sampling details for line spectroscopy AOTs
check, esp limits. The oversampling factor gives the number of times a given wavelength is
seen by multiple pixels in the homogeneously-sampled part of the spectrum. The column
Range of highest sensitivity refers to the range that is seen by each of the 16 spectral pixels
in a module. Bright and faint refer to bright-line or normal spectral mode.
Band
Mode
B3A
B3A
B2B
B2B
R1
R1
Band
CN
UN
CN
UN
CN
UN
Wavelength
m
55
72
72
105
105
158
175
210
B3A
B3A
B2B
B2B
R1
R1
R1
R1
Wavelength
range (m)
5173
5173
71105
71105
103220
103220
Full range
m
1880
799
2058
1039
5214
2869
2337
1314
Grating step size

faint
bright
168
384
168
384
188
368
188
368
240
344
240
344
Full range FWHM
km s1
m
0.345
0.021
0.192
0.013
0.038
0.039
0.364
0.038
1.825
0.111
1.511
0.126
1.363
0.124
0.92
0.098
Grating
faint
48
75
46
75
43
75
FWHM
km s1
115
55
165
80
315
240
210
140
steps
oversample factor
bright faint
bright
10
41.1
80
10
41.1
80
10
36.2
80
10
36.2
80
10
27.9
80
10
27.9
80
Range of highest
sensitivity (m)
0.095
0.053
0.221
0.126
0.875
0.724
0.654
0.441
A bright line mode was also offered for sources with bright lines, which offered a shorter
spectral coverage and lower SNR but also a shorter observing time. For the chop-nod bright
line mode, up and down grating scans were still performed in this mode, but with only 10
grating steps taken the on-sky time was about 3-4 times shorter than in the standard mode.
The minimum total observing time (one line, one cycle) was about half that of the standard
mode, and it was about twice less sensitive. In the bright line mode even for unresolved
lines the central 3 pixels in the 16-pixel modules do not see the continuum, as they scan
over the line profile only. This could make difficult a proper response correction as no flat
continuum level can be determined for all the pixels: taken from OM, but question is, did did
the Ffing task in the SPG take this into account? HPDP?.
For the unchopped bright line mode, fewer grating steps were also taken, 50 rather than
75, resulting in a reduced wavelength coverage and an integration time reduced by by 1/3.
Everything else remains the same.
3.3.2.2
The line spectroscopy AOT
The choices related to the spectral coverage and sampling that could be made in the line
spectroscopy AOT were: the band, the central wavelength, the number of line repetitions
(how many repeats on the grating scan), the redshift, and whether to opt for the bright-line
mode.
49
The bands that could be chosen in HSPOT were R1 (103220 m) + B2B (71105 m), or R1
+ B3A (5173 m). For observations in the 5155 m region, band B3A is affected by a red
leak (Sec. ??): it was recommended to instead use the range spectroscopy AOT from where
band B2B could be selected. linkto where we discuss what the real blue limit is, because
B3A has a red leak and B2A has an incorrect RSRF, and summarise that here. also, was
B2B always out for line spec, or did this happen at a certain OD? If the observer specified a
line in a blue band, then the free or parallel red band would be at a wavelength 2 (B2B)
or 3 (B3A) that of the blue wavelength, and similar applies if the spectral line chosen fell
in the red band. Note that it was possible to chose a wavelength range in one camera that
resulted in a wavelength range in the other camera that fell outside of the band limits: in the
pipeline any data that fall in this range are not propagated to the Level 2 cubes, and if the
entire range is out of band, the resulting cubes are missing entirely.
The central wavelength could be taken from a list provided in HSPOT (of common lines in
the PACS spectral range) or could be set by the user. If the observed specified a redshift, the
wavelength range would be shifted by the appropriate amount.
Observers were asked to input the expected line and continuum fluxes from their source.
This was used by the uplink logic to set the integrating capacitance (Sec. ??), particularly
to minimise the chance of saturation. The saturation limits are given in Table ?? in the
instrument chapter, see 4.12 of the OM.
Increasing the line repetition factor allowed for a higher SNR in the final spectra. However,
in one AOR it was only possible to ask for a combination of up 10 lines+their line repetition
factors: anything from 10 separate lines with no repetition, to 1 line with a 10 repetitions.
Finally, the bright line mode could be selected (from the part of HSPOT where the background
sampling method was specified). In this mode the spectra range covered is less, and so is the
observing time. Selecting this mode switched the integrating capacitance from the default
high to low, to minimise the chance of saturation. also to increase the dynamic range?
3.3.2.3
Range spectroscopy
Range spectroscopy AOTs allowed observers to chose their own wavelength range; this was
used to cover broad lines, several lines at once, or the entire chosen band in one go (SED
mode).
Chosing ranges: A combination of 10 ranges and/or range repetition factors was allowed:
anything from 10 separate ranges with repetition factor 1 to 1 range with a 10 repetitions.
The repetition factor refers to the number of consecutive grating scans for the final AOT
release (Jan 2010) the default was to do each grating scan twice (up and then down), so for
these a repetition factor 2 meant that it was done four times at what OD did this switch
happen?. This could be combined in the AOR with any number of nod repetitions (explained
in Secs 3.3.1.2 and 3.3.1.4).
Chosing the SED: This mode allowed the observer to observe the entire range of the band
chosen. Only one choice of band (one blue+red combination) could be chosen in a single
AOR, but users could ask for multiple range and nod repetitions factors. To observe the
entire spectral range of PACS (the entire SED), the red and blue camera ranges of two
separate observations was necessaryat what OD did we stop offering B2A and B2B together?.
50
Two spectral sampling densities were offered: high, which gave the same sampling as for line
spectroscopy (Sec. 3.3.2.1), or Nyquist, which was shallower by a factor 1013 I got this from
Table 6.6, since in the text it says the step size is 6.25 spectral pixels, but I dont know what
that means, esp comparing to the 75 quoted for the line scan mode above) and hence allowed
for faster grating scans. The Nyquist sampling was used for the SED mode, as otherwise the
observation would take too long. The scanning parameters, including wavelength coverage,
range, and spectral density, are given in Table 3.3. The instrument FWHM in this mode is
the same as is given in Table 3.2.
Table 3.3: The grating and spectral sampling details for range spectroscopy AOTs check, esp
limits, and that unchopped and chopnod are the same. The oversampling factor gives the
number of times a given wavelength is seen by multiple pixels in the homogeneously-sampled
part of the spectrum.
Band
B3A
B2A
B2B
R1
Wavelength
range (m)
5173
5173
71105
102220
Grating step size

Nyquist SED
168
2220
188
2300
188
2400
240
2500
oversample factor
Nyquist SED
41.1
3.1
36.7
3.0
36.2
2.8
27.9
2.7
Instrument FWHM
m
km s1
0.0210.013
11555
0.0390.039
227161
0.03980.364
16580
0.1110.1260.098 315240140
To improve data quality for deep Nyquist sampled observations, a spectral dithering scheme
was been implemented from what OD?: if the line repetition factor was > 1, then subsequent
scans were performed with a small spectral offset; this improved the spectral sampling of the
combined result.
In the line and range spectroscopy AORs, the observers could set the details of the spectral
ranges/sampling for one of the cameras only (i.e. a range in one camera or one of the SED
modes): the nominal/prime range. The grating step size was always determined for this
prime range. The step sizes set for the red and blue cameras are not the same: see Tables 3.3 and 3.2. In practice this means that if comparing separate observations (e.g. different
programmes) with the same wavelength range, the spectra from the observations could have
different step sizes depending on in which order the range was defined (i.e. whether as the
prime or parallel). For instance, the 8090 m range is observed as a prime range with step
size 188 in the 2nd order, but this range is also observed as the parallel for observations with
the prime range 160180 m in 1st order, for which the grating step size is 240. This second
observation will result sub-optimal wavelength sampling in the blue range.
3.3.2.4
The range spectroscopy AOT
The choices related to the spectral coverage and sampling that to be made in the range
spectroscopy AOT were: the band, the wavelength range or a choice of the complete band,
the sampling density, and the number of range repetitions (how many repeats on the grating
scan). There was no option to add the redshift for this mode.
For ranges: The bands that could be chosen in HSPOT were: R1 (102220 mum) + B2B
(71105 m), R1 (102220 mum) + B3A (5173 m), and R1 (102146 mum) + B2A (51
73 m). Up to 10 ranges+reptitions could be chosen in any AOR. The third option provides
a higher continuum sensitivity but lower spectral resolution than that provided by band B3A,
51
therefore its use was recommended for observing broad spectral features, very strong lines, or
when the primary scientific interest was the determination of the continuum level over long
spectral ranges. This band was also recommended for observations in the range 5155 m
(which was not provided for in Line spectroscopy). linkto where we discuss what the real blue
limit is, because B3A has a red leak and B2A has an incorrect RSRF, and summarise that
here. The default sampling density for ranges was High, but Nyquist could also be chosen.
If the observer specified a line in a blue band, then the free or parallel red band would be
at a wavelength 2 (B2B) or 3 (B3A) that of the blue wavelength, and similar applies if
the spectral line chosen fell in the red band. Note that it was possible to chose a wavelength
range in one camera that resulted in a wavelength range in the other camera that fell outside
of the band limits: in the pipeline any data that falls in this range are not propagated to the
Level 2 cubes, and if the entire range is out of band, the resulting cubes are missing entirely.
For SEDs: The bands that could be chosen in HSPOT were: SED B2B + long R1 (70
105 mum, 140220 m), SED B2A + short R1 (5173 m, 102146 mum), and SED B3A +
long R1 R1 (4773 mum, 140216 m). Only one SED (with to to 10 range repetitions) could
be chosen per AOR. need here to also give the real final ranges, definitely not 47 microns
and I though long R1 with B2A went to 210 only. For SEDs, only Nyquist spectral sampling
density could be chosen; if the observer wanted High they would instead define a range that
was the full stretch of the filter.
To increase the depth of high sampling density range scans, even for relatively short ranges,
observers were advised to increase the range repetition and the number of nodding cycles. As
the timescale of the drifts in detector sensitivities should be considered shorter than an hour,
they will be better corrected with shorter nodding cycle durations.
Observers were asked to input the expected line and continuum fluxes from their source.
This was used by the uplink logic to set the integrating capacitance (Sec. ??), particularly
to minimise the chance of saturation. The saturation limits are given in Table ?? in the
instrument chapter, see 4.12 of the OM.
3.3.2.5
Wavelength switching
The wavelength switching mode was superceeded by the unchopped line scan mode after the
PV phase on OD?, as it was aimed at the same sorts of observations but produced inferior
results or was less efficient?. The continuum flux level recovered by the wavelength switching
mode is not preserved, and line profiles are not reliable if: a noticeable gradient is present in
the continuum flux over the performed wavelength throw, or if blends of line forests disturb
the wavelength-switch interval.
In wavelength switching mode, the line was scanned with the same grating step as in chopped
line spectroscopy, i.e. every spectral pixel was sampled at least every 1/3 of a resolution
element. This step was called a dither step. At every dither step, the signal was modulated
by moving the spectral range over about half of the FWHM of an unresolved line. This
allowed one to measure a differential line profile, canceling out the background (link OM
Fig 6.12 or here?). The modulation on every scan step followed an AABBBBAA pattern,
where A is a detector integration at the initial wavelength, and B is a detector integration at
the wavelength-switched wavelength. This cycle was repeated 20 times in one direction, and
52
repeated in the reverse wavelength direction. The switching amplitude was fixed for every
spectral band.
To create the clean source spectrum, a clean off position was visited at the beginning and the
end of the observation (via a telescope movement, not a chopper throw), where the same scan
was was performed every two or more raster positions (OM unclear here, is this correct?).
3.3.3
The pointing
Pointed and mapping observations could be programmed into the AORs. Pointing consisted
of a single position. Mapping observations came in three flavours: tiling, Nyquist mapping, or
oversampled mapping. A mode called Pointed with dither was provided for a while ODs?).
3.3.3.1
Pointed observation
A pointed observation consisted of a single position on the sky. It was particularly recommended for observations of point and point-like sources, for which the final spectrum can be
recovered using the point-source and semi-extended source correction tasks provided in the
pipeline (Secs ?? and ??).
Many extended sources that were slightly smaller or slighlty larger than the field-of-view of
the PACS IFU were also observed in this mode. Archive users of such observations should
note that the spatial footprint of the resulting cubes is that of the native IFU: a 5x5 slightly
irregular grid with spaxels of 900 .4 on a side: see Sec. ?? for an image of this footprint (need
the footprint with the crosses and with the spaxel squares overlaid). The Level 2 rebinned
cubes (the science-grade products produced by the pipeline for this mode) do not have a
regular spatial grid, even if it looks like they do when opened in a cube viewer (the spatial
axes of the WCS of these cubes are defined in spaxel rather than sky coordinates). Hence, for
extended sources, the morphology will be compromised. It should also be noted that the 900 .4
spaxel severely undersamples the PACS beam, which has a FWHM of about 900 1300 (Sec. ??
check exact numbers). How to deal with these limitation is explained in Link. Need a good
explanation of the footprint, and of the rebinned cubes, how they will look in a viewer vs how
they really are, and the use of interpolated cubes instead. Some of this is in the products
chapter, but maybe a separate note should be written to explain this in more detail than the
handbook can include, and make that then a Tier 1 document? Include mosaic cubes in that
user note as well.
3.3.3.2
Pointed with dither
The dithering scheme was offered prior to launch to perform a small (1 3) raster with 200
step size. However, during the PV Phase it was found that for sources with a well-known and
confined photocenter, the pointing accuracy of Herschel allowed for a reproducibility of line
fluxes on the percent level and well-behaved line centres and shapes. This mode was therefore
discontinued in favour of the Pointed mode. Very few observations in the archive were taken
with this mode how many and check what types of cubes are produced for this mode - count
as oversampled mapping?.

3.3.3.3
53
Rasters/mapping
The PACS spectrometer spaxels of 900 .4900 .400 undersample the diffraction beam of the
Herschel telescope. However, with several observations (the number depending on the wavelength) offset by fractional spaxel-size steps, the full spatial information can be retrieved. The
raster pointing mode was used to generate a suitable set of pointings, and depending on the
purpose of the raster map, different raster step sizes were recommended. This mode was also
used for observations covering a larger FoV for which recovering the full spatial information
was not important.
When defining their mapping AORs, observers using the unchopped mode had to be careful
about where to place and with what frequency to observe their off positions: see Sec. 3.3.1.4.
For the chop-nod mapping modes, the only extra restriction was that only the large chopper
throw was possible.
For mapping modes, the final cubes of the pipeline are mosaic cubes (Sec. 8.3.3.2), where the
spatial sampling achieved is reflected in the spaxel sizes given to these cubes.
Three classes of mapping mode were offered, and these are summarised in Table 3.4. Fig. ??
shows the pointing pattern of the Nyquist and oversampled mapping modes.
Table 3.4: The recommendations for raster settings for the various mapping modes. Y and Z
direction are the in the instrumental plane (Y is in the PACS chopping direction).
Mode
Tiling
Nyquist
Oversample
Offsets (Y,Z)
blue
red
3400 4700
Nr. of steps
blue
red
any
Z:1600 , Y:14.500
Z:2400 , Y:2200
33
22
Z,Y: 300
Z,Y: 4.500
33
22
use-case
To create a larger FoV where the spatial
sampling was less important
To Nyquist sample the beam at the
wavelength of the observation
To map the source at full PACS spectral
resolution
For the tiling mode, the sampling of the beam is the same as with pointed observation. The
Nyquist mode obviously Nyquist-sampled the beam, and the spaxel sizes of the Level 2/2.5
cubes reflect this (34.500 ). The steps are in this mode are larger than a spaxel, but, as can be
seen in Fig. ??, the resulting sampling is finer than a spaxel. Since the PACS spectrometer
spaxels in the sky are not square really? but we quote them always as square, the best
recommended step is not the same in the red and blue camera. For the oversampled mode,
the very small step sizes could be the same in both directions. It was important to have the
correct number of raster steps, not just the recommended offset, to result in the aimed-for
the spatial sampling.
In mapping AORs, it was necessary to set the number and spacing of the raster steps. For
chopnod observations the map could only be defined in instrument coordinates (Y and Z),
and the map size was restricted to 60 60 so clean offset positions with the large chopper throw
could be obtained. In the unchopped mode the map could be defined in instrument or sky
coordinates, with a maximum size of 2 .
54
Chapter 4
In-flight Performance
4.1
Introduction
This chapter will contain a summary of the performance and calibration, but not the details of
how they were obtained. Qualitative information if quantitative information is not suitable.
The aim is for readers who just want to know how far to take PACS data accuracies and
what to be aware of, and who dont care how the calibrations were obtained. This chapter
comes from the Spectrometer Performance and Calibration document, Poglitsch 2010, the
calibration tree, and the various photometer publications and notes on the public webpage.
The section on anomalies and quality control comes from LC and KE as it is partly on-going
work. The section on wavelength shifts with source position requires a little extra work
from KE (gathering results from SR and KE done at KUL on this topic) but is otherwise
documented in TNs from before launch.
55
56
4.2
CHAPTER 4. IN-FLIGHT PERFORMANCE
Photometer
4.2.1
Sensitivity and flux uncertainty
4.2.2
Noise properties
4.2.3
PSF and beam size
4.2.4
Anomalies and quality control
4.3
4.3.1
Spectrometer
Spectral range and resolution
4.3.1.1
Line profile
4.3.1.2
Spectral bands
4.3.2
Spectral purity
4.3.2.1
Leakage regions
4.3.2.2
Second-pass ghosts
4.3.2.3
? 62 m dip ?
4.3.3
Wavelength calibration
4.3.3.1
Wavelength shifts with source position
4.3.3.2
Flux calibration accuracy for the various modes
4.3.3.3
Flux accuracy for point and semi-extended sources
4.3.3.4
Flux accuracy for extended sources
4.3.3.5
? The effect of transients ?
4.3.4
Spatial accuracy and resolution
4.3.4.1
Pointed observations
4.3.4.2
Mapping observations
4.3.5
Anomalies and quality control
Chapter 5
The Spectrometer Calibration
5.1
Introduction
This is the chapter that requires most help from the ICC
57
58
CHAPTER 5. THE SPECTROMETER CALIBRATION
5.2. THE SPECTRO-PHOTOMETRIC CALIBRATION
5.2
The spectro-photometric calibration
5.2.1
The calibration sources and models
5.2.2
Chopped observations: calibrating the telescope background
5.2.2.1
Telescope spectrum
5.2.2.2
Time variability
5.2.3
Unchopped observations: using the internal calibration sources
5.2.3.1
Calibration source fluxes
5.2.3.2
Absolute flux calibration
5.2.3.3
Relative spectral response function (RSRF)
5.2.4
Line flux calibration in order-leak regions
5.2.5
Saturation limits
5.2.6
Self-curing
5.3
Spatial calibration
5.3.1
Focal-plane geometry
5.3.2
Characterising the beam
5.3.2.1
Beam profiles
5.3.2.2
Point source calibration function
5.3.2.3
Semi-extended source correction
5.3.2.4
Extended sources and the spatial flatfield
5.3.2.5
Intra-sapxel losses
5.4
Spectral calibration
5.4.1
The line profile
5.4.2
Calibration the wavelengths
5.4.3
Spectral purity
5.5
Table of the calibration files
59
60
CHAPTER 5. THE SPECTROMETER CALIBRATION
Chapter 6
The Photometer Calibration

6.1
Introduction
The calibration of the PACS photometric observing modes was addressed centrally by the
Observatory. The PACS Instrument Control Centre Team in collaboration with the Herschel
Calibration Scientists established a smooth evolution of the calibration from module characterization over the detailed instrument and integrated system tests on ground to the final full
system verification and calibration in-flight, as outlined in the PACS Calibration Document
(Klaas et al., 2014, PACS-MA-GS-001, Issue 1.10, Nov. 2014). The PACS calibration team
planned, worked out, executed and analysed dedicated calibration observations on ground
blackbody sources, generating continuum emission over the full field-of-view via integrating spheres or simulating point sources with hole masks in front, and celestial standards to
consistently and thoroughly characterise all instrumental effects of the PACS photometer.
Once in-orbit, after the initial instrument check-out, there was a larger calibration block during the Performance Verification Phase verifying the in-orbit performance of the photometer
observing modes, comparing it against the predictions and ground performance, and doing
the tuning for optimal performance in the space environment as well as providing the baseline
in-flight calibration. 528.6 h of observing time were spent on this, as outlined in detail in the
PACS Performance Verification Phase Plan (Klaas & Nielbock 2014, PICC-MA-PL-001, Issue
2.0, May 2014). Verification of the calibration stability as well as refinement and extension of
the calibration, taking into account the growing knowledge in data processing, was then done
during the Routine Science Phase of the Herschel mission. Additional 293.2 h were allocated
to this goal, as described in detail in the PACS Routine Science Phase Plan (Klaas & Nielbock
2014, PICC-MA-PL-002, Issue 4.01, May 2014).
Analysis of the calibration observations led to the generation of the calibration files needed
for the Standard Product Generation, as listed in Table ??, as well as to the implementation
of appropriate algorithms in the various data reduction steps.
Since the PACS blue photometer camera offered the best spatial resolution of all Herschel
instruments it was used for verifying the absolute pointing accuracy of the Herschel telescope.
Also the characterization of specular stray-light features generated by Herschel telescope parts
was established in a PACS/SPIRE photometer parallel mode calibration program. Both
Herschel telescope related calibration aspects are described in the Herschel Mission & Satellite
61
62
CHAPTER 6. THE PHOTOMETER CALIBRATION
Overview Handbook (Kidger et al., 2015, HERSCHEL-HSC-DOC-2062). Also the issue of

Herschel instrument cross-calibration is addressed there.
In the following, the individual calibration items are shortly described and their expected
accuracies are given. This includes also an overview on the celestial calibration standards.
6.2
Calibration achievements during commissioning and the

performance verification phases
The Herschel Commissioning Phase (CP) lasted for the first two months of the mission and
was characterized by a not yet final and stable observing environment, the first part being
executed with the cryo cover still closed and the second part with the telescope still cooling
down. Nevertheless, the following results were achieved:
The photometer was found fully functional in its nominal operation mode.
The smooth operation of the chopper control loop was verified and even slightly improved. The warm electronics was temperature stabilized, thus avoiding control loop
gain drifts.
The so-called sneak preview on OD 32 with the PACS photometer executing scan
maps on M 51 provided the first light results shortly after successful opening of the
cryostat cover. This verified early on that the telescope was in focus and revealed the
triangular wing feature of the PSF. The measurements also gave the first feedback to a
global telescope pointing improvement (SIAM update).
A perfect tracking of the telescope on the fast moving asteroid Melpomene (67/h) was
verified.
The bolometer bias voltages were successfully tuned during the cool-down of the telescope with no saturation effects being encountered in low gain mode, thus providing
optimal start conditions for the bolometer detector set-up optimization in the first part
of the Performance Verification Phase.
A first already quite accurate prescription for the 3 He cooler hold-time calculation was
derived, which enabled efficient usage of the cold bolometer state between two cooler
recyclings.
Objectives and details of the PACS photometer operations are given during the Commissioning Phase in detail in the PACS Commissioning Phase Plan (Feuchtgruber 2009, PACS-MEPL-024, version 2.0) and the PACS Commissioning Phase Timeline (Klaas & Nielbock 2009,
PACS-ME-PL-024 Addendum, Version 1.0).
The Performance Verification Phase for the PACS photometer lasted from OD 64 until
OD 191, at the end overlapping with the Observatory Science Demonstration Phase. Fig. 6.1
shows the logical build-up of the initial in-flight calibration of the PACS photometer during
the Performance Verification Phase:
6.2. CALIBRATION ACHIEVEMENTS DURING COMMISSIONING AND THE PERFORMANCE VERIF

CPPhotFPG
PVSpecFPG
initial search, SIAM update
initial search
PVPhotFPG
PVPhotFPG
abs. pointing error
PVPhotBol
low gain optimization
PVPhotFPG
SAA dep., SSO track
PVPhotBol
PVPhotBol
high gain direct
PVPhotAOTVal
stand. small source & raster map
rel. pointing error, scans
PVPhotSpatial
DDCS, low freq. noise, chop freq.
PVPhotAOTVal
PVPhotSpatial
stand. point source & scan map
PVPhotFlux
FOV I
chop ang. cal, spec PSF sources
PVPhotSpatial
abs. flux cal., resp. monitor, nonlinearity, flatfield
PSF, distortion, ghosts, straylight
PVPhotFlux
PVPhotSpatial
PVPhotAOTVal
abs. flux cal., resp. monitor, linearity
FOV II
nonstandard params., upgrades
PVSpecSpatial
straylight
PVSpecSpatial
PSF
Figure 6.1: Logical flow of addressing the PACS photometer calibration items during the
Herschel Performance Verification Phase, taking into account mutual dependencies and prerequisites. Each box is represented by the calibration proposal name and the list of calibration
items addressed during this block.
After adjusting the dynamic range of the bolometers to the telescope background at
its lower end towards the end of the Commissioning Phase, one of the first photometer
calibration tasks was to find the bolometer bias for optimal detector responsivity and
noise equivalent power (NEP) in the high gain mode. The direct read-out mode was
preferred to the double differential correlated sampling mode. In the last step the low
frequency noise and the signal dependence on the chopper frequency was characterized.
Parallel to the optimization of the detector settings, the Herschel telescope pointing
behaviour could be investigated and characterized, because this did not need the final
sensitivity, since the pointing calibration sources were usually reasonably bright.
With the finalization of the detector set-up, the performance of the pre-flight designed
observing modes could proceed. It turned out soon, that the scan map mode was
more efficient and sensitive for mapping than the small source and rastering chopped
modes. This led to an early deprecation of the two latter modes. For point source
observations dedicated mini-map scans were newly designed and optimized. A scan
speed of 20 arcsec/s was found superior to the original baseline scan speed of 10 arcsec/s,
so that this speed was not offered any more for scientific observations. See the PACS
photometer AOT release notes for more details. The original chop-nod point source
mode was maintained throughout the mission for pointing calibration (delivering more
precise astrometry) and as an independent check of the overall photometric calibration
on a subset of flux standards (Nielbock et al. 2013, Experimental Astronomy 36, Issue
3, 631).
64

The photometric flux calibration program, including non-linearity assessment, started in
PV Phase, however some of the fiducial standard star and planet measurements had to
be deferred to after the PV Phase for visibility reasons and also additional measurements
had to be accumulated to address the aspect of reproducibility. A consistent in-flight
calibration update also needed the evaluation of the beam profiles to derive the encircled
energy fractions for the applied aperture. Both steps were achieved only some time after
the end of PV Phase.
The first basic measurement set for the spatial calibration of the PACS photometer
was acquired. This included the verification of the chopper angular scale on double
sources with well known separation, the first PSF characterization on point sources with
different SED shapes and maps on a bright source allowing its instantaneous detection
in each detector frame in support of determining the photometers field distortion.
Dedicated programs around very bright sources aimed for characterising ghost and nearfield stray-light features generated inside the instrument and for detecting predicted
specular out-of-field stray-light spots by the telescope structure. Also the spatial extent
and structure of some very bright sources selected for the spatial calibration of the
spectrometer was checked with the better spatial resolution of the photometer.
Objectives and details of the PACS photometer operations during the Performance Verification Phase are given in detail in the PACS Performance Verification Phase Plan (Klaas &
Nielbock 2014, PICC-MA-PL-001, Version 2.0).
6.3
Spatial calibration
The spatial calibration of the PACS photometer covers two main topics:
1. Mapping the focal plane to the sky. This is needed to assign sky coordinates RA, DEC
to each individual flux measurement, as defined by a bolometer pixel and a time. Such
sky coordinates are needed for the center of a photometer pixel and/or its corners.
2. Characterizing the point spread function (PSF) and related effects such a ghosts or
straylight. This also includes the determination of encircled energy fractions that are
needed to interpret photometry.
6.3.1
Focal plane mapping and field-of-view distortion
The photometer pointing and spatial calibration is defined in reference to a virtual aperture
at the center of the blue array. This virtual aperture actually corresponds to a gap between
the matrices constituting the bolometer. Chopped/nodded observations where hence set up
to provide a symmetric beam pattern around this position. They were done for a large set of
pointing sources with accurate positions, to ensure that Herschel placed this virtual aperture
at the intended position. The location of the virtual aperture is captured in the siam calfile.
The following coordinate systems and transformations are used to assign a sky position to
photometer pixels:
6.3. SPATIAL CALIBRATION
65
Subarray coordinates p,q: A coordinate system for the subarrays (= bolometer matrices), trivially assigning the center of each pixel with an integer coordinate e.g. p,q=(4,5).
This corresponds to the indices of each pixel in a PACS frame.
Array coordinates u,v: A cartesian coordinate system in the PACS focal plane, with
the long v axis aligned with the chop direction. The subArrayArray calfile captures the
location of each pixel p,q in these coordinates. Here, each matrix was assumed to be
an ideal grid but the relative spacings and orientations of matrices are allowed to be
imperfect. For each pixel, the positions of the center and of the four courners of the
nominally active region (640640m within the full 750750m pixel) are provided.
Instrument coordinates y,z: An orthogonal local coordinate system on the tangential
plane of the sky. The two axes are aligned with the spacecraft Y and Z axes as projected
onto the sky, and the zero point is the virtual aperture of the bolometer. It corresponds
to the zero point of the array coordinate system for the blue array and chopper angle
= 0. The arrayInstrument calfile contains polynomial relations that permit to compute
y and z given u,v, and the chopper angle.
Sky coordinates: The J2000 ICRS right ascension and declination of a measurement.
Standard methods can be used to compute this from y,z, plus RA, DEC of the virtual
aperture and the spacecraft position angle.
The subArrayArray and arrayInstrument calfiles where initially derived on the basis of instrument level tests in the laboratory. A punched hole far-infrared source was rastered over
a grid of positions, in front of a setup that was comprising a test optics and PACS. Such
measurements were obtained for dense spatial grids and for several chopper positions, and
transferred to first on-sky versions of the calfiles via the optical model of the test optics. For
that step, the optical model had to be assumed to be fully accurate.
Over much of the mission, the quality of the reconstructed Herschel pointing was insufficient
for a fully independent re-derivation of these quantities. Instead, only scale and rotation
of the ground-based calibration were adjusted to match actual observations, but distortions
were still traced back to the ground tests. When more accurate gyro reconstructed pointing
became available, detailed scans across bright (near) point sources were used to re-derive the
subArrayArray and arrayInstrument calfiles. These observations were obtained only for the
zero chopper position because they are very time consuming, hence chopper related terms
in the most recent arrayInstrument calfile still make use of the ground tests. This has no
practical consequences because in-orbit experience led to the overwhelming majority of PACS
observations being unchopped scan maps with the chopper at zero position.
Fig. 6.2 uses results of an R Dor observation to visualize the footprint of the blue and red
photometer pixels on sky. Fits to these and other observations were used to derive the final
spatial calfiles. Salient feature include:
A several degree tilt of the photometer arrays with respect to the spacecraft y direction.
In contrast, the chopper direction is close to the y direction.
Slight deviations from a regular pattern for the relative positions of the individual
matrices - most visible for the two left blue matrices in Fig. 6.2.
66
Figure 6.2: Location of PACS photometer pixels on sky, as measured from OD1308 observations of R Dor (black symbols). Top: blue array, bottom: red array. The coordinates are
defined by the photometer virtual aperture and the y and z direction of the Herschel spacecraft coordinate system. Bad pixels and few outliers from the preliminary fit that is shown
in red are not plotted.
Field-of-view distortions requiring nonlinear terms, most clearly visible for the red array.
More detailed information can be obtained from a number of documents in the explanatory
library. PACS spatial calibration files (Lutz & Contursi 2013, PICC-ME-TN-019, Version
1.0) describes the coordinate systems, the calibration files transforming between them, and
their version history. PACS spatial coordinates cheat-sheet (Contursi & Lutz 2008, PICC-METN-027, Version 1.0) provides an overview sketch of some coordinate orientations. In-orbit
rederivation of PACS photometer FOV distortion (Lutz & Feuchtgruber 2013, PICC-ME-TN044, Version 1.0) presents the most accurate evidence on mapping the pixel positions to sky,
and the FOV distortion.

6.3.1.1
67
Non-uniform pixel size
As described below ins Sect 6.4, the photometric calibration of the PACS photometer uses
units of Jy/pixel both for the individual detector timelines in the frames, and for the final
sky-projected photometric maps. It is important to note that at the frames level Jy/pixel
refers to the PACS physical pixels. Given the distortions just discussed, for different pixels
these physical pixels project to different solid angles on sky, despite same physical size in m.
Mappers creating final maps with regular sky pixels from the frames have to consider this, via
use of the information in the subArrayArray and arrayInstrument calfiles. For some mappers
this may need to be applied via the hipe convertToFixedPixelSize task.
6.3.2
PACS photometer point spread function
The PSF of the PACS photometer is mostly derived from observations of the bright point or
near-point sources Vesta, Ceres, Tau and Boo. The bright planets Neptune and Mars
were used to probe the faint PSF wings, at the expense of heavy saturation of the PSF core for
Mars. In the explanatory library, PACS photometer point spread function (Lutz 2015, PICCME-TN-033, version 2.2) and its associated tarball are the main references for the observed
PSF.
Modeled PSFs have been made available in Herschel/PACS modeled point-spread functions
(Geis & Lutz 2010, PICC-ME-TN-029, version 2.0) and its associated data tarball. While
these reproduce many salient features of the observed PSF, they are NOT an accurate representation of reality, as would be needed for PSF fitting to data, deconvolution, or derivation
of encircled energy fractions. This arises because they reflect only pre-launch knowledge of
the Herschel mirror shape, and do not include effects inside PACS.
6.3.2.1
PSF morphology
In all three PACS bands, the PSF core is surrounded by a tri-lobe pattern at the level of up
to 10% (in blue) of the PSF peak (Fig. 6.3). This basic morphology reflects the wavefront
errors caused by the Herschel main mirror surface imperfections. Among other factors, they
reflect the 120 degree symmetry of the secondary mirror suspensions. In all three bands, the
top-right of the three lobes is found weakest, qualitatively reproduced by models.
The next fainter level of PSF structures is again not a clear diffraction ring but a knotty
pattern at the % level and below. It is clearly seen in blue and with less detail in green and
red. The blue data clearly show a weak spike in roughly vertical (spacecraft Z) direction
that is also present in green and red. Roughly 7000 towards the -Z direction, a bright spot is
superposed on this spike, clearly seen at 70 m (Fig. 6.4) and weakly at 100 m. It is induced
in PACS rather than the telescope, and is likely caused by an optical ghost. Yet fainter levels
show a knotty PSF structure and finally the telescope secondary support diffraction spikes
(Fig. 6.4).
The PSFs at the medium scanspeed 20 00 /s that was employed for almost all prime mode
PACS scanmaps and for the rarely used slow 10 00 /s are virtually identical. They are roughly
round for 70 m and 100 m but vertically (Z direction) elongated in the red. This elongation
68
Figure 6.3: PACS photometer PSF at 70 m (left), 100 m (center) and 160 m (right), as
derived from Vesta observations in OD160. The images use a linear stretch to 100%, 10%,
and 1% of the PSF peak from top to bottom. The spacecraft Z direction is on top, as for
observations with telescope PA=0, and the scale bar indicates 6000 .
is seen in the PSF core and not just an effect of wings or lobes. These 20 00 /s prime mode
PSFs reflect the optical properties of Herschel+PACS.
For fast 60 00 /s prime mode scans, detector time constants lead to a significant PSF elongation
in scan direction. For the blue channel in PACS/SPIRE parallel mode, the smoothing into
5 Hz rather than 10 Hz frames leads to an equivalent elongation that is already noticeable
at 20 00 /s and strong for the widely adopted 60 00 /s parallel mode observations. Fig. 6.5
uses the example of the 70 m band to visualize these elongations. Note that for a typical
69
Figure 6.4: Faint structures in the PACS 70m PSF. Left: Ceres with a stretch from -0.0005
to 0.001 of the PSF peak, scale bar indicates 6000 . Right: Diffraction spikes from observations
of Mars, scale bar indicates 60000 .
parallel mode map composed from to observations with crossed scans, these elongated PSFs
will superpose to a cross-like structure. The fast scan 70 and 100 m PSFs show a region in
scan direction from the PSF peak where the flux is undershooting to below zero, to a level
of about -1% of the peak. This is the consequence of an undershoot in the detector timeline
after a bright signal, i.e. it is not a processing artefact.
6.3.2.2
Effects of source color and reduction schemes
Table 6.1 lists the full width half maximum of the photometer PSF as obtained from a
Rayleigh-Jeans source and masked highpass plus photProject mapping onto small 100 pixels.
Several influences can modify these values.
Because of the large spectral width of the PACS filters, noticeable effects of the SED slope on
the width of the PSF exist. Convolving the modelled PSFs of PICC-ME-TN-029 for different
spectral slopes F with the PACS pixel size, the ratio of PSF FWHM for = 4
(as for a star or asteroid) and the FWHM for = 1 is 0.979, 0.971, 0.956 for the 70,
100, 160m filters. Exactly tuning the convolution to the precise observed FHWM for the
Rayleigh-Jeans = 4 and comparing to a specific sources can refine such comparisons
further. All stellar or asteroid bright point sources used to calibrate the PACS PSF have
Rayleigh-Jeans like slopes in the far-infrared. A FWHM comparison to observations of the
comparatively faint and lower S/N Blazar 3C345 which has a much redder SED is consistent
with the model-based estimates for FWHM increase.
The PSF images obtained from any mapper depend on the map pixel size. For the drizzlelike algorithm implemented by the photProject task the drop size (controlled via the pixfrac
70
Figure 6.5: Effects of scan speed. The PACS photometer PSF at 70m is shown for prime
mode scan speeds of 10 00 /s (top left), 20 00 /s (top right), and 60 00 /s (bottom left). The
additional smoothing because of the 5 Hz averaging in the blue channel for PACS/SPIRE
parallel mode is indicated in the bottom right panel for speed 60 00 /s. The PSF images are
at 70 m, use a linear stretch to the PSF peak, and the scale bar indicates 6000 .
parameter) is also relevant. PICC-ME-TN-033 lists the effect for map pixel sizes ranging from
100 to the nominal 3.2 or 6.400 photometer pixels, and for pixfrac from 0.01 to 1. The largest
increase of PSF FWHM over that range is 18%, 10%, 17% for 70, 100, 160 m. For PSF
fitting or similar tasks, it is strongly recommended to extract PSFs from the map proper, or
from point source data that were reduced in a fully equivalent way.
Other effects on the PSF are more specific to the mapper. In the simple masked highpass

Mode
AMA
00 /s
Prime 10
Prime 20 00 /s
Prime 60 00 /s
Parallel 20 00 /s
Parallel 20 00 /s
Parallel 60 00 /s
Parallel 60 00 /s
Parallel 20 00 /s
Parallel 60 00 /s
71
blue 70m
FWHM
PA
00
+63
+63
+63
+42
42
+42
42
+42,42
+42,42
5.20 5.56
5.41 5.72
5.70 9.05
5.44 6.51
5.31 6.68
5.8512.58
5.6912.74
5.746.26
8.809.60
61.7
30.8
-26.5
43.7
-36.9
0.4
4.4
green 100m
FWHM
PA
00
6.54 6.78
6.66 6.89
6.84 9.81
6.62 7.44
6.53 7.56
6.9913.15
6.8713.41
6.98 7.42
9.7310.69
61.8
31.1
-27.0
43.9
-37.1
2.9
3.8
red 160m
FWHM
PA
00
10.3811.97
10.5512.08
11.3913.37
10.2912.20
10.3712.27
10.9014.09
11.0114.53
10.4612.27
11.5113.65
6.1
9.1
41.2
8.5
-3.4
27.7
-23.7
3.1
5.3
Table 6.1: FWHM of the PACS PSF for several important cases. 2-dimensional gaussian fits
were used to derive FWHM for the small and large axis. For noticeably non-round PSF cores,
the position angle east of the spacecraft Z direction is noted. The array to map angle of the
scan is also specified. The maps used to derive the FWHM have been created by photProject
with map pixel size 100 and pixfrac=1. Entries above the line refer to single direction scans,
showing the in-scan elongation for fast scan and for parallel mode. Entries below the line
refer to coadded parallel mode crossed scans, where a cross-like PSF emerges from coadding
the two elongated PSFs.
filter and photProject scheme, the PSF is sensitive to the mask that is applied to a given
source. In particular, unmasked sources that may e.g. later be retrieved via stacking will
show negative filter residues in scan direction. This is related to the slight fluxes losses in
such reductions that are discussed by Popesso et al. 2012 (arXiv 1211.4257). Other mappers
have occasionally produced enhanced flux in the PSF wings or artefacts near very bright point
sources. Some information at status 2014 can be taken from the PACS Map-making Tools:
Update on Analysis and Benchmarking, Paladini et al. 2014, Version 2.0, but development of
the mapmaking codes continues.
6.3.2.3
Effects of nonlinearity, saturation, crosstalk, ghosts, straylight on observed PSF
A number of effects can influence the PSF in particular for very bright sources and may need
to be considered when e.g. probing the the reality of faint companion structures. More detail
is provided in PICC-ME-TN-033, but a brief summary is:
Nonlinearity. The PACS bolometers respond in a nonlinear way to very bright sources.
For point sources in the 100 Jy regime and more, this would lead to a flattening of the
PSF peak. All recent reductions invoking the photNonLinearityCorrection task are
already corrected for this.
Saturation aftereffects. After passing very bright and saturated sources such as
Mars, the signal does not instantanously return to its stable pre-saturation value. This
can cause trails that are not an artefact of the mapmaker that was used.
72

Detector crosstalk. An electrical crosstalk is observed between columns 15 and 0
in the same row of each bolometer matrix. The photMaskCrosstalk task excludes the
affected column 0 from the final maps. If this task is not used and column 0 included,
a faint negative signal may be seen about 5000 (70 or 100 m) or 10000 (160 m) offset
in spacecraft Y direction from a very bright source. Its strength depends on processing
details, e.g. the method of deglitching.
Ghosts, reflections, straylight. A number of weak effects have been characterized
for a situation where a very bright source falls onto certain spots in the focal plane.
These can be safely ignored for all cases except for testing the reality of very faint
structures near very bright sources. More detail can be found in PICC-ME-TN-033 and
in Faint linear artefact in PACS photometer, Okumura 2010, SAp-PACS-KO-0716-10,
Version 1.0.
6.3.2.4
Encircled energy fraction
Figure 6.6: Fraction of energy from a point source contained within a circular aperture of
given radius. Left: As applicable for the standard case of prime mode and slow or medium
scan speed. The three PACS bands are shown in blue, green, red for 70, 100, 160 m. Right:
Repeat of the same curves, plus additional ones showing the smoothing effect of fast scan
and/or parallel mode.
The radial profile of the PSF or equivalently the fraction of energy from a point source
6.4. FLUX CALIBRATION
73
contained in a given aperture (encircled energy fraction EEF) are the link between point
spread function and flux calibration. To achieve sufficient dynamic range, the PACS EEFs
are derived from a combination of Vesta data (for the inner part of the profile) and Mars
data (for the faint wings, but heavily saturated in the center, and far from being a point
source). Fig. 6.6 left shows the EEF for the three bands and the default case of slow or
medium scanspeed in prime mode. This EEF is underlying the PACS flux calibration. As
discussed above, fast scan and/or PACS/SPIRE parallel mode lead to elongated PSFs. The
right panel of Fig. 6.6 illustrates the resulting EEF curves.
6.4
6.4.1
Flux calibration
Flat-field
The flat-field describes the pixel-to-pixel response variation of the bolometer cameras under
homogeneous illumination. Each pixel had a somewhat different gain from its neighbor (detector flat-field, the gain specification was 8% rms over a whole 16 16 pixel detector matrix)
and the transmission of the optical system was not spatially homogeneous either (optical
flat-field).
The measurement condition of homogeneous illumination was practically fulfilled during the
instrument ground level tests, when two blackbody sources were imaged via integrating
spheres onto the photometer detectors, thus illuminating the whole arrays. The flat-field
was constructed from the difference of pairs of measurements with small illumination power
differences and appropriate normalization over a selected pixel area. This established the
first version of a flat-field used in the PACS data reduction. However, one deficit of this
pre-flight generated flat-field was, that the operational bias voltages applied to the detectors
were different from the final ones in-flight. The flat-field in the green filter was not measured
in this way but assumed to be the same as in the blue filter, since the detector was the same.
The use of the PACS internal calibration sources, which were relatively homogeneous sources
illuminating the whole array, too, for construction of a flat-field could in the end not be
realized. The reason was, that the optical field distortion at a chopper deflection on the
internal calibration sources (at about 6 arcmin from the optical zero position) was different
from the field distortion inside the sky field-of-view and was not measurable independently.
This would have been a significant drawback for the quality of the flat-field.
A determination of a sky flat-field had the difficulty that there is no real flat extended celestial
source. This was overcome by performing scan maps on bright point or compact sources. The
scan maps were performed with the slow scan speed (10 arcsec/s) and with 90 orientation
with regard to the instrument reference system (along the S/C y-axis) to minimize the scan
area and to illuminate each pixel by the source. Very small scan leg separations were selected (2 arcsec or 3 arcsec with 2 maps shifted relatively by 1.5 arcsec) to achieve a very high
coverage.
Two different reduction methods were applied:
1) The signal time line for each pixel was used. After removing glitches, the time line
data were first filtered to remove low frequency noise. For each pixel the value of the
74

maximum minus the median of the time line signals was computed. The variation of
this value over all pixels is the flat-field.
2) For each pixel an independent map of the object was generated with the PhotProject
projection algorithm (cf. data reduction section on PhotProject). This included
also glitch removal and high pass filtering to reduce the low frequency noise. Then,
for each pixel the flux of the source was measured by aperture photometry on the map
generated from the single pixel. The variation of the fluxes over all pixels is the flat-field.
There is one significant difference in reduction philosophy between the pre-flight flat-field and
the sky flat-fields: The treatment of electrical cross-talk. While for the pre-flight flat-field,
which was generated by extended illumination, it was assumed that the cross-talk is part of
the flat-field, for the sky flat-field, which was generated by scanning a compact source over
the array, pixels affected by cross-talk were discarded.
The resulting sky flat-fields were compared with the ground level flat-field by constructing the
ratios. This allowed to identify spurious features introduced by the processing. Using a point
source resulted in interference patterns. This feature was caused by the interference of the
map pixel grid and the telescope scanning pattern, which moved in the horizontal direction
of the bolometer array. It was most visible for the blue filter with the sharpest point spread
function causing the flux falling onto one pixel to vary quickly with the position of the source
within the pixel. Using a slightly extended source reduced this flux variation.
The final sky flat-field was constructed with the mapping method from special calibration
measurements taken on OD 780 on the compact planetary nebula NGC 6543. This flat-field
is most up-to-date wrt. instrument settings in-flight and most consistent with the overall
data reduction strategy due to the removal of cross-talk and optical distortion. A more
detailed description of the reduction is given in Okumura (2014, PACS photometer flat field
measurement, SAp-PACS-KO-0735-14, v1.2, 05 Feb.)
Fig. 6.7 shows the photometer flat-field for each filter.
6.4.2
Responsivity calibration
The physical quantity which determines the bolometers absolute photometric calibration is
its responsivity R. R, in units [V/W], is the ratio of the measured output signal Usig produced
by the infalling far-infrared radiation power Fphotband , hence
Usig = R Fphotband
(6.1)
With a source SED, conventionally expressed as f,s () [Jy] in the FIR, and the relation
f = c2 f , the flux in a PACS photometer band can be expressed as
Z
Fphotband [W ] = T A
1
c
f,s () S() d
2
(6.2)
with T being the product of reflection losses by the optical mirrors, A being the effective telescope area and S() being the relative photometer system response of the PACS photometer
band.
75
Figure 6.7: Photometer flat-fields for the three filter bands: blue (top), green (middle) and red
(bottom). Individual detector matrices making up the camera field-of-view can be recognized.
The bars underneath each image indicate the amplitude of variation over the field-of-view.
76
For the PACS photometer, the convention is, that the flux density at the reference wavelength
f,s (0 ) is determined for the reference SED f,1 = 1 , i.e. f,1 () = 0 f,1 (0 ) = 0 f,1 (0 ),
hence
c
Fphotband [W ] = T A f,1 (0 )
0
f,1 (0 ) =
T A
1
S() d
Fphotband
= Fphotband Cconv
R 2 1
1 S() d
c
0
(6.3)
(6.4)
with
Cconv =
T A
c
0
1
R 2 1
1
=
S() d
1
T A 0
(6.5)
and effective bandwidth

0 =
c
0
1
S() d
(6.6)
The 0 used in the calibration processing of PACS standard data products are listed in
Table 6.2.
Table 6.2: Effective bandwidth 0 for a constant in the flux per logarithmic frequency
2
interval SED as defined in Eq. 6.6. 0 = c0 0 is the corresponding bandwidth in
wavelength.
0
(m)
70
100
160
0
(1011 Hz)
0
(m)
6.48
5.09
3.54
10.6
17.0
30.2
If the source SED f,s () is known, the true photometer source flux is determined as
f,s (0 ) =
f,1 (0 )
Kcc,s
(6.7)
with Kcc,s being the appropriate color correction factor as described in section 6.5.3
This means that
R
1
Cconv
1
Fphotband
f,1 (0 )
T A 0 f,s (0 ) Kcc
(6.8)
hence the responsivity R can be calibrated, too, by flux density measurements of celestial
standards.
77
The responsivity of the bolometer arrays depends on their operating temperature, voltage
bias, and optical loading (see Sec. 6). R was thoroughly characterized in the lab (cf. ?),
and the ground-based calibration was used to estimate the initial in-flight responsivity at the
beginning of the mission when the telescope foreground emission had stabilized. For in-flight
calibration only a relative update in the form of
f,s (0 )
Rnew
=
Rold
f,standardmodel (0 )
(6.9)
was necessary, by using the ratio of the measured flux to the celestial standard star model
flux, when calibrating f,s (0 ) with Rold .
In practice, the update of R did not rely on one single flux measurement alone, but on a set
of measurements on several standards to minimize systematic effects
n
f,s,i (0 )
Rnew
1X
=
Rold
n
f,standardmodel,i (0 )
(6.10)
i=1
so that
1
n
f,s,i (0 )
i=1 f,standardmodel,i (0 )
Pn
= 1 by applying Rnew .
This scheme applies, when more standard star measurements become available, when additional systematic instrument effects altering the signal level are corrected for, when changes
in the encircled energy fraction alter the resulting flux (see below) or when new standard star
models become available.
Note, that R is a global value for the whole detector array. Individual pixel-to-pixels variations
are accounted for by the flat-field ffpix .
R is not a constant. It depends on temperature and the total flux load, hence R = R(T,
Btotalf lux ), as will be shown later.
The standard evaluation procedure for flux calibration measurements is aperture photometry.
Hence, not the the total source flux is measured, but only the fraction inside the measurement
aperture. For the derivation of the total measured source flux a correction for the encircled
energy fraction (EEF1 , see Table ??) inside the aperture must be performed:
f,1 (0 ) =
6.4.2.1
aperture
f,1
(0 )
fEEF
(6.11)
Internal calibration sources
As described in Section 2.2.2, the PACS internal calibration sources were grey body radiators
with emissivity of about 4 8% (TBC). The two sources were operated at temperatures
of 55 K (CS1, corresponding to a resistance of 48) and 60 K (CS2, corresponding to 58).
The observed resistance values were 48.0040 0.0015 (13 m peak-to-peak) for CS1 and
58.0000 0.0019 (14 m peak-to-peak) for CS2. A 20 m resistance variation corresponded
1
http://herschel.esac.esa.int/twiki/pub/Public/PacsCalibrationWeb/bolopsf 20.pdf
78
roughly to a 10 mK temperature variation in the selected temperature range, which was about
0.05% flux or surface brightness variation, well below the bolometer measurement accuracy.
6.4.2.1.1
Calibration block observation data base
Each PACS photometric observation includes at least one calibration block (CB) measurement which consists of chopped observations between the two internal calibration sources. It
preceded the sky observation and was performed during the target acquisition slew. In the
end, the scientific measurements needed not to be flux calibrated with the help of the calibration block measurements thanks to the fantastic stability of the bolometer response over
the whole mission (0.5% standard deviation) allowing to use a fixed and time-independent
detector responsivity (see Section 6.4.2).
The Herschel archive contains in total 22000 calibration block observations from 20795 PACS
photometric observations, which include both scientific and celestial calibration observations.
All of them are made with the identical CS set-up, either in blue/red or green/red filter
combination. This homogeneous data base covering the photometer operational period with
the densest time grid allowed to reveal a number of subtle instrumental flux calibration effects:
1) A correlation of the differential CS signal, i.e. the detector responsivity, with the 3 He
cooler evaporator temperature, see Sect. 6.4.2.3 for details
2) A correlation of the differential CS signal, i.e. the detector responsivity, with the temperature of the focal plane unit. This effect is weaker than the trend with evaporator
temperature
3) A signal drift of few percent for the first half hour after the end of the orbit prologue
(photometer instrument set-up prior to a photometer observation block) following a
recycling of the 3 He cooler.
Details of the photometer calibration block trend analysis are given in Moor et al. 2014 (Exp.
Astronomy 37, 225, PACS photometer calibration block analysis).
6.4.2.2
6.4.2.2.1
Celestial calibration sources and models

Fiducial standards
As prime flux reference, models of the photospheric emission of late type giants are used ?.
This type of stars was already used as absolute calibrators for earlier IR space missions (e.g.
IRAS: ?, ISO: ?, Spitzer: ?). The flux regime they cover is inside the linear flux behaviour
of the PACS bolometers. Ideally, stars of different stellar types would be used to prevent
systematic uncertainties from the modeling. Unfortunately only the late type giants are
bright enough at far infrared wavelengths to be observed at high signal to noise. One star,
Sirius ( CMa), was the only A type exception, but unfortunately had to be dropped from our
set of primary flux calibrators as will be discussed in ?? Models of the atmospheres of the giant
planets Uranus and Neptune ? are equally accurate, however these sources are already in the
non-linear flux regime of PACS. Asteroids are about to be established as independent prime
FIR flux calibrators ( ?, this volume). The five fiducial stellar standards Boo, Cet, Tau,
79
And and Dra were observed repeatedly during the mission for absolute calibration and
to monitor the system stability (for this Dra, which was visible during the whole mission,
was observed on a monthly basis). We also monitored the photometric system using the
Calibration Blocks (CalBlocks). They track the response several times a day throughout the
mission (at least one Calblock per OBSID) and show the high stability of the photometric
system with high accuracy (see ?).
The theoretical spectra of candidate stellar calibrators in the far infrared were generated using
the MARCS stellar atmosphere code (?, ?) and the TURBOSPECTRUM synthetic spectrum
code (?) and are presented in ?. The stellar parameters and their uncertainties were derived
by ?. The line lists used in the spectrum calculations and the model uncertainties are discussed
in ?, where it is estimated that the uncertainty of the models in the PACS wavelength range
is 5%. The absolute flux calibration is based on Selbys (?) K-band photometry, the
zero-point is determined on the basis of an ideal Vega, i.e. the K-band photometry of Vega
is corrected for a flux excess of 1.29% (cf. ?). The determined Selby K-band zeropoint
is 4.0517 1010 W/m2 /m. This and more detailed information on the individual stars
can be found in the headers of the fits files containing the theoretical spectra used for the
PACS flux calibration. The files can be found at ftp.ster.kuleuven.be/dist/pacs/calsources/
or ftp.sciops.esa.int and will be provided in the Herschel archive.
As the stars can show excess flux in the far infrared wavelengths, due to either debris disks
for early type dwarfs or in the case of late type giants, by a chromosphere or ionised wind,
it is important to investigate whether we can rule out the existence of such excesses for our
candidate calibrators. ? present observations in the sub-mm up to cm wavelength range for
nine late type giant stars, of which seven were in our original list of candidate calibrators.
These observations were made in preparation of the Herschel mission and performed on several
ground based telescopes. Also Sirius, a A1V star, was studied, but not included in the ? paper
as this only considered K and M giants. Although in seven out of nine studied giants an excess
was detected, it was found that for the eight stars, selected as prime flux calibrators for the
PACS photometer, the excess only started beyond the PACS wavelength range.
Based on early analysis of Herschel observations of potential calibration stars we decided to
use only the following 5 fiducial stars for the final analysis: And, Cet, Tau, Boo,
Dra. The three other potential calibrators were discarded ( CMa, Peg, UMi). In
the case of these three stars the fluxes deviated from the models by more than 10% in at
least one PACS band. For UMi no Selby K-band photometry was available for the absolute
calibration and only a less accurate ( 10%) Johnson K band from ? could be used. ? showed
that Peg is variable in the mid-IR bands by about 10%. CMa shows about 20% excess at
160 m. The underlying cause of the discrepancy is still unclear and is under investigation.
The monochromatic flux densities based on at 70.0, 100.0 and 160.0 m for the five finally
selected stars are given in Table 6.3.
6.4.2.2.2
Planets and moons
The planets of the solar system and their satellites beyond the Earth have surface and brightness temperatures of few hundred Kelvin and are thereby bright infrared emitters. Thanks
to radiometers on space probes, like the infrared radiometer for Mariner (Chase 1969, Applied Optics 8, 639), IRIS on Voyager (Hanel et al. 1980, Applied Optics 19, 1391), the
80
Table 6.3: Information on the selected fiducial stars. Monochromatic flux densities at 70.0,
100.0 and 160.0 m are given. The stellar temperatures are taken from ? and the fluxes are
based on models published in ?.
HR
HD
HIP
337
911
1457
5340
6705
6860
18884
29139
124897
164058
5447
14135
21421
69673
87833
ID
RA (J2000)
Dec (J2000)
SpType
Temp
[K]
Model flux [mJy]

70 m 100 m 160 m
And
Cet
Tau
Boo
Dra
01:09:43.9236
03:02:16.8
04:35:55.2387
14:15:39.6720
17:56:36.3699
+35:37:14.008
+04:05:24.0
+16:30:33.485
+19:10:56.677
+51:29:20.022
M0III
M1.5IIIa
K5III
K1.5III
K5III
3880
3740
3850
4320
3960
5594
4889
14131
15434
3283
2737
2393
6909
7509
1604
Photopolarimeter-Radiometer (PPR) on Galileo (Russell et al. 1992, Space Science Reviews

60, 531), accurate information on temperatures, thermal properties, albedo, energy balance
and the infrared emission spectra have been collected, making the outer planets and their
satellites suitable bright calibration standards. The inner planets were not accessible to Herschel, since they were inside the Sun constraint. Radio occultation data acquired with Voyager
have been used to probe the vertical structure of the planetary atmospheres (Lindal 1992, AJ
103, 967).
In particular, Uranus and Neptune have been established as excellent flux standards for contemporary far-infrared space observatories, providing flux levels well adapted to the dynamic
range of their instrument detectors, including the PACS photometer. Mars, Jupiter, and Saturn are too bright for sensitive photometers like PACS. In the PACS photometer calibration
scheme Uranus and Neptune as well as the planet satellites Titan, Callisto and Ganymede
were used as complementary flux calibrators to the fiducial stars and the asteroid prime calibrators to cover an as large as possible flux range to address the issue of the non-linear
detector response and to allow consistency checks in the case of overlapping fluxes of different
calibrator types and reference models.
In Table 6.4 the utilized planet and satellite models are listed together with flux predictions
for certain epochs and appropriate colour correction factors for comparison with measured
PACS fluxes. The SEDs of the planets and satellites are shown in Fig. 6.8.
1062
928
2677
2891
621
81
Figure 6.8: Absolute disk-integrated model flux density predictions for Callisto, Ganymede,
Uranus, Neptune, and Titan in the Herschel-centric reference system: For the three planet
satellites the predictions are for the epoch of the corresponding PACS measurements. The
minimum-maximum model predictions for Uranus (ura esa2 2 i.dat) and Neptune are referring to all available PACS measurements during the entire Herschel mission. The PACS
band-passes are shown in arbitrary units.
model source
orton uranus esa5 (Orton)
ura esa2 2 i.dat (Moreno)
nep esa5 2 i.dat (Moreno)
call esa2 2 i.dat (Moreno)
gany esa2 2 i.dat (Moreno)
tit esa3 2 i.dat (Moreno)
Source
Uranus
Uranus
Neptune
Callisto
Ganymede
Titan
789
1121
789
1121
540
1444
981
981
1138
OD
969.0
917.3
886.1
835.2
374.2
357.9
1272.8
1168.1
89.0
f 70
(Jy)
0.984
0.984
0.984
0.984
0.984
0.984
0.998
0.996
0.985
cc 70
943.3
887.5
884.5
833.7
374.4
358.1
788.5
741.3
81.5
f 100
(Jy)
0.995
0.995
0.992
0.992
0.993
0.993
1.016
1.014
1.002
cc 100
673.0
631.7
636.1
599.5
268.2
256.5
371.2
355.9
52.0
f 160
(Jy)
1.018
1.018
1.019
1.019
1.020
1.020
1.055
1.053
1.028
cc 160
Table 6.4: Model flux predictions for Uranus, Neptune, Callisto, Ganymede, and Titan.
Models are taken from
ftp://ftp.sciops.esa.int/planets/originalData/esan/, where then n in sub-directory esan indicates the version number
as reflected in the model source name. In the case of Callisto, Ganymede and Titan the model fluxes are given for the epoch of
observation (OD). For Uranus and Neptune the flux range encountered during the entire mission is reflected by maximum and minimum values and the epoch of observation (OD). Appropriate colour-correction factors (cc) to be applied to measured PACS fluxes
for comparison with the model predictions are given, too. The estimated maximal uncertainty for these corrections is 2% for Titan
and about 1% for the rest of the targets.
82
83
The model predictions for Uranus are based on two different models: (i) the Uranus model provided by Glenn Orton (Orton et al. 2014, Icarus 243, 494) and (ii) the Uranus model provided
by Raphael Moreno (Moreno et al. 2016, in preparation). Both models of Uranus were used to
check the PACS calibration. They are connected to slightly different thermal structures: For
the Moreno model the thermal profiles go back to Voyager radio-occultation measurements
(Pearl et al. 1990, Icarus 84, 12; Lindal et al. 1992, AJ 103, 967), while for the Orton model
measurement constraints from mid- and far-IR (Spitzer-IRS and Herschel/SPIRE) were considered. The radiative transfer modeling includes continuum opacities from collision induced
absorption of H2 , He, and CH4 . The brightness temperature differences between both models
are below 5% in the PACS wavelength range. The model uncertainties are linked mainly
to the input parameters and therefore reflect the absolute calibration accuracy. The model
brightness temperatures were combined with Uranus angular diameter as seen from Herschel,
taking into account the equatorial and polar radii and the sub-observer latitude (i.e. Ob-lat
in JPL/Horizon). The correspondence of the Moreno model with measured PACS fluxes is
better than 4%. The Orton model predictions are too high by about 8% at 70 m and still
about 4-5% too high at 100 m. The overall impression is that the model seems to work well
from 160 m onwards. The reason for the offset might be related to the He/H2 ratio used in
the model (Orton, priv. comm.).
The radiative transfer modeling for Neptune includes continuum opacities from collision induced absorption of H2 , He, and CH4 . The thermal profiles go back to Voyager measurements
(Pearl & Conrath 1991, Geophys. Research Supp. 96, 18921, Lindal et al. 1992, AJ 103, 967)
and fit the IR measurements of Akari spectra of Neptunes stratosphere (Fletcher et al. 2010,
A&A 514, A17). The Neptune model also fits the line-to-continuum (i.e. relative measurements) of SPIRE and PACS spectra including CO and HD lines (Moreno et al. 2016, in
preparation). The model brightness temperatures were combined with Neptunes angular
diameter as seen from Herschel, taking into account the equatorial and polar radii and the
sub-observer latitude. The correspondence with measured PACS fluxes is better than 3%.
The model of Titan is computed using the radiative transfer model for a disk-averaged geometry, described in Courtin et al. (2011, A&A 536, L2) and Moreno et al. (2012, Icarus 221,
753). This model includes continuum opacities from collision induced absorption of N2 -CH4
pairs and uses the thermal profile from Huygens probe measurements (Fulchignoni et al. 2005,
Nature 438, 785) combined with CIRS measurements (Vinatier et al. 2010, Icarus 205, 559).
Molecular lines of CO, HCN, and CH4 and their isotopes are included, which fit the HIFI
and SPIRE observations. The final absolute accuracy is estimated to be better than 5%. The
correspondence with measured PACS fluxes is better than 2%.
The disk-averaged models of Callisto and Ganymede are based on thermal models of the
sub-surface computed upon the Spencer et al. (1989, Icarus 78, 337) algorithm. The models
solve the heat diffusion equation in the planetary surface material as a function of longitude,
latitude, and depth. The thermal inertia used was originally derived by Spencer (1987,
PhD, University of Arizona) from 2-layer models and based on 10-20 m data from Voyager.
These models also include the surface dielectric constant and roughness, which are fitted from
the ground based measurements at mm-wavelength performed with the IRAM-PdBI and the
SMA (Moreno et al. 2008, BAAS 40, 478). Model accuracies were estimated to be better than
7% for Ganymede and Callisto, however, effectively, these Callisto and Ganymede continuum
models turned out to be not very accurate in the PACS wavelength range. The Callisto model
shows a 5% agreement at 70 m and is about 10% too low at 160 m. The Ganymede model
84
underestimates the observed and calibrated PACS fluxes by 10 to 25%. This is probably
linked to the effective albedo and different thermal inertia layers in the sub-surface, which are
very difficult to constrain from disk-averaged observations.
A detailed comparison of PACS photometry of Uranus, Neptune, Titan, Callisto and
Ganymede with the model predictions and the discussion of the results is given in M
uller
et al. (2016, A&A accepted).
6.4.2.2.3
Asteroids
The idea of using asteroids for calibration purposes goes back to IRAS (Beichman et al.
1988, Infrared Astronomical Satellite /IRAS) catalogs and atlases, vol. 1: Explanatory Supplement). The IRAS 12, 25 and 60 m bands were calibrated via stellar models and in that
way connected to ground-based N- and Q-band measurements. But at 100 m neither stellar
model extrapolations nor planet models were considered reliable. Asteroids solved the problem. Models for a selected sample of large main-belt asteroids were used to transfer the
observed IRAS 60 m fluxes out to 100 m and calibrate in that way the IRAS 100 m band.
The Infrared Space Observatory (ISO) (Kessler et al. 1996, A&A 315, 27) was also lacking
reliable photometric standards at far-IR wavelength (50 - 250 m) in the flux regime between
the stars and the planetary calibrators Uranus and Neptune. M
uller & Lagerros (1998, A&A
338, 340) provided a set of 10 asteroids, based on a previously developed thermophysical
model code by Lagerros (1996, A&A 310,1011; 1997, A&A 325, 1226; 1998, A&A 332, 1123).
AKARI (Murakami et al. 2007, PASJ 59, 369) followed the same route to calibrate the FarInfrared Surveyor (FIS) (Kawada et al. 2007, PASJ 59, 389) via stars, asteroids and planets
in the wavelengths regime 50 - 200 m. The Spitzer mission (Werner et al. 2004) considered
in the beginning only stars for calibration purposes. But due to a near-IR filter leak of the
MIPS (Rieke et al. 2004, ApJS 154, 25) 160 m band, the calibration scientists were forced
to establish and verify calibration aspects by using cooler objects. The asteroids served as
reference for the flux calibration of the 160 m band as well as for testing the non-linear MIPS
detector behaviour (Stansberry et al. 2007, PASP 119, 1038).
In preparation for Herschel a dedicated asteroid programme was established (M
uller 2005,
The Asteroid Preparatory Programme for HERSCHEL, ASTRO-F & ALMA. In: Wilson, A.
(ed) Proceedings of the Dusty and Molecular Universe: A Prelude to Herschel and ALMA,
p. 471, ESA SP-577). This led to a sample of about 50 well-known and well-characterized
asteroids and their corresponding model predictions. The asteroids fill a gap of more than
two decades of flux between the sub-mm/mm calibrators Mars, Uranus and Neptune, and the
mid-IR bright calibration stars (Fig. 6.9)
PACS photometry was obtained for in total 16 asteroids:
1) The four asteroids (1) Ceres, (2) Pallas, (4) Vesta, and (21) Lutetia. With newly established thermophysical models and photometric consistency verification against PACS,
SPIRE and HIFI measurements, they can be considered as prime flux calibrators (M
uller
et al. 2014, Experimental Astronomy 37, 253).
2) The regularly observed asteroids (10) Hygiea, (3) Juno, (52) Europa, (88) Thisbe,
(704) Interamnia, (8) Flora, (6) Hebe, and (20) Massalia
85
3) The asteroids (47) Aglaja, (423) Diotima, (65) Cybele and (29) Amphitrite
In Table 6.5 the utilized asteroid models are listed together with flux predictions for certain
epochs and appropriate colour correction factors for comparison with measured PACS fluxes.
Figure 6.9: Overview with the flux densities of the different far-IR/sub-mm/mm calibrators.
The Uranus and Neptune SEDs represent the minimum and maximum fluxes during Herschel
visibility phases. Three fiducial stars are also shown, their flux range coverage is representative
for the brightest stellar calibrators. For Ceres the maximum flux and for Lutetia the minimum
flux during Herschel observations are shown.
(21) Lutetia
(4) Vesta
(2) Pallas
(1) Ceres
Source
model source
1244
726
1295
686
743
160
859
400
OD
394.2
176.8
119.1
43.7
259.3
88.8
11.1
2.7
f 70
(Jy)
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
cc 70
226.6
102.4
68.8
25.4
146.7
51.2
6.4
1.6
f 100
(Jy)
1.02
1.02
1.02
1.02
1.03
1.03
1.02
1.02
cc 100
96.8
43.9
29.3
11.1
61.3
21.4
2.7
0.67
f 160
(Jy)
1.07
1.07
1.07
1.07
1.07
1.07
1.07
1.07
cc 160
Table 6.5: Model flux predictions for (1) Ceres, (2) Pallas, (4) Vesta, and (21) Lutetia. The flux range encountered during the entire
mission is reflected by maximum and minimum values and the epoch of observation (OD). Appropriate colour-correction factors (cc)
to be applied to measured PACS fluxes for comparison with the model predictions are given, too.
86
87
The applied thermophysical model (TPM) is based on the work by Lagerros (1996, A&A
310,1011; 1997, A&A 325, 1226; 1998, A&A 332, 1123) It takes into account the true observing and illumination geometry for each observational data point, a crucial aspect for
the interpretation of the main-belt asteroid observations which cover a wide range of phase
angles and helio-/observer-centric distances, as well as different spin-axis obliquities. High
quality size and geometric albedo values are fundamental for reliable TPM predictions. The
TPM allows to specify simple or complex shape models and spin-vector properties. The
one-dimensional vertical heat conduction into the surface is controlled by the thermal inertia
= c, where is the thermal conductivity, the density, and c the heat capacity. The
observed mid-/far-IR/sub-mm fluxes are connected to the hottest regions on the asteroid
surface and dominated by the diurnal heat wave. The seasonal heat wave is less important
and therefore not considered. Infrared beaming effects (similar to opposition effects at optical wavelengths) are calculated via a surface roughness model, described by segments of
hemispherical craters. Mutual heating is included and the true crater illumination and the
visibility of shadows is considered. For further details of the thermophysical models and
the photometric verification against PACS, SPIRE and HIFI measurements, see M
uller et
al. (2014, Experimental Astronomy 37, 253). The example of the shape model for asteroid
(2) Pallas is shown in Fig. 6.10.
73
(2) Pallas temperature [K]
2 Pallas
200
150
100
50
100 micron [Jy]
72
without thermal inertia
71
70
69
with thermal inertia
68
6
10
12
14
16
29Nov2012 (UT)
Figure 6.10: Left: Shape model of Pallas with the TPM temperature coding on the surface,
calculated for the Herschel point-of-view on OD 1295, OBSID 1342256236, rotation axis is
along the vertical direction. Right: the corresponding thermal light-curve at 100 m with and
without thermal effects included.
The correspondence of the TPM models with measured PACS fluxes is better than 2% for
(1) Ceres and (2) Pallas, better than 4% for (4) Vesta and better than 5% for (21) Lutetia.
6.4.2.2.4
Faint stars
As outlined above, the photometric calibration of the PACS photometer is based on the
fiducial stars. They are still relatively bright in the FIR (in the range 1 - 10 Jy) to achieve
high S/N ratios within reasonable measurement times. Beside repeated measurements of
these standard stars a set of fainter secondary standard stars was repeatedly measured by
PACS as part of the calibration program. This included sources down to a few mJy. Including
fainter sources with well known flux predictions allows to address the following issues:
88

1) How does the sensitivity scale with flux and time?
2) What is the impact and consistency of the applied data reduction scheme on the resulting
source flux for fainter and fainter flux contributions to the telescope background level?
In preparation of the PACS in-flight photometric calibration secondary standard source lists
with stars described in Cohen (1996, Spectral irradiance calibration, Astronomical Journal
112, 2274), Hammersley (1998, Infrared standards for ISO, A&A Supp. 128, 207), and Gordon
(2007, MIPS absolute calibration, PASP 119, 1019) were prepared. Depending on the source
visibility during the Herschel mission, a subset of sources from these lists were observed to
cover the flux range from 0.5 - 2.5 Jy down to 2 - 10 mJy over the 3 photometer wavelengths
70, 100 and 160 m. The finally observed 17 sources are listed in Table 6.6.
62509
12929
32887
198542
148387
180711
139669
41047
170693
138265
159330
152222
39608
181597
15008
156729
168009
HD
HR 7341
Hyi
e Her
HR 6847
Gem
Ari
Lep
Cap
Dra
Dra
Umi
HR 2131
42 Dra
HR 5755
HR 6540
other name
f70
model flux prediction (mJy)

f100
f160
spectral type
2457 (5.73%) 1190 (5.73%) 455.9 (5.73%) K0IIIb

1707 (5.9%)
831.4 (5.9%)
321.0 (5.9%) K2III
1182 (5.9%)
576.2 (5.9%)
222.7 (5.9%) K4III
857.7 (6.03%) 418.0 (6.03%) 161.5 (6.03%) M0III
479.5 (3.38%) 232.6 (3.45%)
89.4 (3.51%) G8III
428.9 (5.7%)
207.7 (5.7%)
79.6 (5.7%) G9III
286.2 (5.67%) 139.5 (5.67%)
53.9 (5.67%) K5III
195.6 (5.96%)
95.4 (5.96%)
36.9 (5.96%) K5III
153.74.6
75.3 (3.0%)
29.4 (3.0%) K1.5III
115.94.0
56.8 (3.5%)
22.2 (3.5%) K5III
64.22.1
31.5 (3.3%)
12.3 (3.3%) K2III
39.41.9
19.3 (5.0%)
7.5 (5.0%) K2
30.91.2
15.1 (4.0%)
5.9 (4.0%) K5III
28.0 (3.29%)
13.6 (3.34%)
5.2 (3.42%) K1III
22.90.8
11.2 (3.5%)
4.4 (3.5%) A1/2V
12.0 (3.21%)
5.8 (3.25%)
2.2 (3.28%) A2V
10.0 (3.40%)
4.9 (3.45%)
1.9 (3.50%) G2V
Source flux models are from
http://iso.esac.esa.int/users/expl lib/ISO/wwwcal/isoprep/cohen/extraps/
b http://iso.esac.esa.int/users/expl lib/ISO/wwwcal/isoprep/gbpp/
c,s
s
c,s
s
c,s
c,s
c,s
c,s
s
c,s
c,s
c,s
s
c,s
c,s
c
c
PACS
obs mode
Cohen (1996)a
Cohen (1996)a
Cohen (1996)a
Cohen (1996)a
Hammersley (1998)b
Cohen (1996)a
Cohen (1996)a
Cohen (1996)a
Gordon (2007)
Gordon (2007)
Gordon (2007)
Gordon (2007)
Gordon (2007)
Hammersley (1998)b
Gordon (2007)
Hammersley (1998)b
Hammersley (1998)b
reference
Table 6.6: Faint secondary standards observed by Herschel-PACS in chop-nod (c) and/or scan map (s) mode. Source fluxes from
Gordon (2007) are for an effective wavelength of 71.42 m and have been color corrected to the PACS central wavelength of 70 m,
by dividing by the factor 0.961 (cf. Sect. 6.5.3) for a Rayleigh-Jeans tail type SED. 100 and 160 m fluxes for these sources are then
extrapolated values for this adopted SED.

89
90
6.4.2.3

Evaporator temperature correction
Each PACS photometric observation includes at least one calibration block measurement that
consists of chopped observations between the two PACS internal calibration sources. Calibration block observations on the contrary to science measurements are always performed
using a standard, identical manner. This homogeneous data set enables the monitoring the
possible short- and long-term evolution of bolometer response during the full mission lifetime
on the basis of a well-defined differential signal from the PACS calibration sources.
During the mission a 3 He cooler was used to ensure sub-Kelvin temperature for the operation
of PACS photometer. The evaporation of 3 He provides a very stable temperature environment
at 300mK. After each cooler recycling procedure, that takes about 2.5 h, there are about
2.5 ODs of PACS photometer observations possible. The evaporator temperature (TEV )
rises towards end of cooling cycles (?). By investigating the behaviour of calibration blocks
differential signals (Sdif f ) as a function of TEV (see Fig. ??) we found a clear correlation
between the two parameters for both the blue and the red arrays. The observed trend can be
well fitted by a linear relationship enabling a correction for this effect. It is important to note
that measured flux densities of standard stars extracted from PACS calibration observations
showed very similar trend with evaporator temperature and their photometry can be improved
by applying our pixel-based TEV correction for their observations (?).
The correlation with TEV is very clear, but the related corrections are typically well below
1%, however, they vary slightly from pixel-to-pixel. It is only 1.2% of all measurements for
which this correction is larger than 1%.
6.5
6.5.1
Photometry corrections
Non-linearity corrections
Bolometers are thermal detectors where a thermistor converts the infrared heat radiation into
an electrical signal. The impedance of the thermistor depends strongly on its temperature,
and hence the relation between the incoming flux and the output voltage is non-linear.
During the ground level tests of the PACS instrument this non-linear behaviour of the bolometers was characterized with the help of black body sources in the ground calibration facility
for a wide illumination range from zero to an illumination power of at least 7 pW per pixel
in each filter band (which corresponds approximately to 125, 160, and 230 Jy/pixel in the
blue, green and red filter). The voltage measured across the thermistor increases relatively
less the larger the illumination power is, which means that the responsivity of the bolometers
decreases with an increase of the infalling flux.
6.5.1.1
Establishment of the non-linearity correction
There is no physical model available for the PACS bolometer arrays. As an empirical model
a logarithmic behaviour proved to be appropriate based on the finding that the impedance of
the thermistors varies exponentially with the detector temperature.
6.5. PHOTOMETRY CORRECTIONS
91
The fitting of the calibration curves is done with a monotonic function of the form
Sig(F )[Jy/pix] = a ln(b F [W ] + 1)
(6.12)
with S being the output signal converted to Jy/pix as described below and F being an illumination power of the applied range during the ground level test.
Before the fits, the following steps have to be applied:
The originally measured voltages across the thermistors for the nominal in-flight bolometer biases have to be transformed into Jy/pixels by using the transfer function of the
bolometer read-out electronics and by applying the nominal responsivity and flat-field
calibration.
The non-linearity is derived relative to the permanent telescope background level, whose
absolute value becomes the zero point on the calibration curve. The values used were
1.8 pW (30 Jy/pix), 1.3 pW (30 Jy/pix) and 4 pW (130 Jy/pix) for the blue, green, and
red filter band, respectively.
Figure 6.11: Calibration scheme of the non-linearity correction: The measured signal difference above telescope background Sis related to a linearized signal difference via the
correction factor which is dependent on the corresponding true flux difference F .
92
The establishment of the calibration correction is sketched in Fig. 6.11. The measured signals
are fitted with the function of Eq. 6.12. The linearized reference signal is given by the tangent
of the logarithmic function at F = 0, which is given by the function
Lin(F )[Jy/pix] = a b F [W ]
(6.13)
For a difference flux F wrt. the zero point, the measured signal difference is
Sig( F ) = a ln(b F + 1) = S
(6.14)
The linearized signal difference is

Lin( F ) = a b F = S
(6.15)
The correction factor can be expressed as a function of the measured signal above zero S
S
by solving Eq. 6.14 for F = 1b (e a 1) and inserting into Eq. 6.15
S
(e a 1)
F
=a
=ab
S
S
(6.16)
The correction factor is determined individually for each pixel of a detector. Fig. ?? shows
typical values of the correction factor for each filter band depending on the difference signal
above the telescope background, which has always to be subtracted and hence set to zero
in the data reduction chain before applying the non-linearity correction. In practice, the
correction is only applied above a threshold of 1.5%, corresponding to a differential source flux
of about 10 Jy/pix (cf. 6.12). One reason for restricting the correction to above the threshold
is that the correction task would be quite memory intensive, if a correction for all pixels in
all frames were performed, which is negligible for a large fraction of the data. Furthermore,
the correction is quite sensitive to the background subtraction and the correction of negative
differential fluxes is not supported (due to the logarithmic function). The threshold avoids
processing complications. Another aspect is that the correction introduces extra noise into
the flux time-lines. This is unproblematic for bright sources with a significant correction, but
it would raise the detection limit for faint sources with in principle negligible correction.
Note that this correction factor has to be applied to individual pixel signals during the level
0.5 to level 1 processing and is not directly applicable to correct the total flux of a source. The
source flux is distributed over several pixels with different signal levels above the background
involving quite different values of the correction. For a rough estimate what the correction
means in terms of absolute source flux we assume the following: The FWHM of a source, which
includes about 40% of the flux is typically distributed over 22 pixels. For a 400 Jy source,
this means a non-linearity correction of about 5% and for a 1000 Jy source a non-linearity
correction of about 15 20%.
Uncertainties of this calibration arise from the following issues:
The applied absolute telescope background levels are an approximation. The telescope
background was not constant, but varied over the mission due to the seasonal solar
illumination variation and a slight increase of the mirror emissivity. The variation of
the telescope background was about 8% at 70 m, 4% at 100 m, and 6% at 160 m,
93
Figure 6.12: Typical curves for the non-linearity correction factor per filter band as a function of the signal variation above the telescope background. The signal is already converted
to the flux unit Jy/pixel.
as derived from the PACS telescope background model in support of the spectrometer
calibration (ref. TBD). It was verified that a change of 10% in the telescope background
has only minor impact on the non-linearity coefficients.
Part of the calibrated range is covered by extrapolating the fitting function. Since for
the red channel the fitting could be verified up to 32 pW, there is confidence that the
extrapolations for the blue and green channel work reliably.
A detailed description of the original calibration analysis, including the applied reduction
scripts is given in Billot (2011; Non-linearity correction module for the PACS photometer,
PICC-NHSC-TR-021, v0.2, February 2011).
6.5.1.2
Verification with celestial standards
The non-linearity correction was verified in-flight with the help of flux standards covering
nearly the whole accessible flux range of PACS (cf. section 6.4.2.2. This is shown in Fig. 6.13
94
1.2
1.2
1.1
1.1
1.1
0.9
flux/model flux
1.2
flux/model flux
flux/model flux
for the best characterized flux standards both in measurement and model accuracy. From the
statistics of this sample no deviation of >2% TBC on average from linearity over more than 5
orders in flux can be found. There are also no offsets between low and high gain observations,
fixed or solar system tracked targets and measurements taken in chop-nod or in scan map
mode. In particular, the results for Neptune, Uranus, and Callisto show the reliability of PACS
fluxes in the high flux regime, where the bolometer responsivity is noticeably in a non-linear
regime. The rare cases of even brighter sources up to the saturation limit, where non-linearity
corrections of >20% are needed are formally not verified. The residual uncertainty for these
sources is estimated to be 10% (M
uller et al., 2016, A&A accepted).
0.9
0.8
0.9
0.8
0.1
10
100
1000
model flux [Jy]
0.01
0.8
0.1
10
model flux [Jy]
100
1000
0.01
0.1
10
100
1000
model flux [Jy]
Figure 6.13: In-flight verification of the non-linearity correction with flux standards over
nearly the whole accessible flux range of PACS. For the blue (left), green (middle) and red
(right) band the measured flux to model flux ratio for the most reliable flux standards is
shown. Triangles represent faint secondary standard stars, circles the fiducial prime standard
stars, boxes prime asteroid calibration standards and reverse triangles planets (Neptune &
Uranus) and planet satellites Callisto & Titan.
6.5.2
Aperture corrections
6.5.3
Colour corrections
In the description of the basic photometric flux calibration strategy in Sect. 6.4.2 the concept
of colour correction was introduced in Eq. 6.7. This is necessary, because the contribution
of the monochromatic flux density f,s (0 ) at the reference wavelength 0 to the total flux
inside the broad photometric bands ( = 6.6, 5.9 and 5.3 for the blue, green, and red
band, respectively) depends on the shape of the source SED within the bandpass. In other
words, the effective bandwidth depends on the source SED. The PACS photometer standard
data products are calibrated in [Jy/pixel] and quoted for the reference wavelengths and an
input energy distribution which is constant in the flux per logarithmic frequency interval (flux
per octave): The flux density with frequency goes as f,1 1 or the flux density with
wavelength goes with f,1 1 . This is in accordance with the flux calibration schemes
used for IRAS, ISO and Akari flux densities.
s
The colour correction factor Kcc,s is therefore the ratio
0 with 0 being the effective
bandwidth for the constant in the flux per logarithmic frequency interval SED used in Eq. 6.5
95
and as defined in Eq. 6.6 and tabulated in Table ??:

R 2
Kcc,s
SED() S() d
= 1 R 2
=
1
1 S() d
R 2
1
SED() S() d
0
(6.17)
SED() is the function of the source SED and S() is the photometer spectral response
S(), i.e. the convolution of the filter transmission curves and the bolometer responses (see
Section 2.3.1.1).
In Tables 6.7 to 6.9 the colour correction factors Kcc,s for representative source SED models
like black bodies (SED() B(, T )), modified blackbodies (SED() B(, T )) or
power laws (SED() ) are tabulated for the PACS reference wavelengths 70, 100, 160 m
(Note: Due to the usage of the unit [Jy] for the PACS fluxes, i.e. per unit frequency
fluxes, one has to consider the correct conversion between f and f , which is given by
the relation f = f )). The correction factors are usually small, only for sources with
temperatures below 20 K they can become quite significant, in particular at 70 m. The
PACS bandpasses have been measured at cold (operational) temperatures and are accurately
known. Nevertheless, large colour correction factors above 1.5 or below 0.5 (mainly for very
cold sources) have to be taken with care, and they might even dominate the final flux accuracy.
The fluxes derived from the PACS photometer standard data products f,1 (0 ) have to be
divided by the appropriate colour correction factor2 to derive the source flux for the proper
source SED:
f,s (0 ) =
6.5.3.1
f,1 (0 )
Kcc,s
Comparison with similar bands from other missions
In addition to the colour correction terms, we provide corrections factors Klc,s which allow
to perform a direct comparison with published monochromatic colour-corrected flux densities
derived from other projects.
For a comparison of a PACS photometer blue band measurement with monochromatic
flux densities derived from the IRAS 60-band or the ISOPHOT P3 60-band: Use the
values in column IRAS 60.0 m
For a comparison of a PACS photometer blue band measurement with monochromatic
flux densities derived from the Spitzer/MIPS 70 m band: Use the values in column
MIPS 71.42 m
For a comparison of a PACS photometer green band measurement with monochromatic
flux densities derived from Akari/FIS Wide-S or the ISOPHOT C100 C90 band: Use
the values in column ISOPHOT 90.0 m
For a comparison of a PACS photometer green band measurement with monochromatic flux densities derived from ISOPHOT C100 C105 band: Use the values in column
ISOPHOT 105.0 m
2
In the technical literature, this term is traditionally called a factor. To avoid confusion, we emphasise
here again, that the raw flux density values have to be divided by, and not multiplied with, the correction
term.
96

For a comparison of a PACS photometer red band measurement with monochromatic
flux densities derived from the Akari/FIS Wide-L band: Use the values in column FIS
140.0 m
flux densities derived from Spitzer/MIPS 160 m band: Use the values in column MIPS
155.9 m
flux densities derived from the ISOPHOT C200 C160 band: Use the values in column
ISOPHOT 170.0 m
These correction factors Klc,s (wavelength corrections lc) have been determined for the
assumed object SED by simply moving along the SED curve from the PACS key wavelength
to the new wavelength and determining the ratio of the flux densities. They have to be applied
in the following way:
1. Determine the monochromatic colour-corrected flux density at the PACS reference wavelength:
f (0 )
f,s (0 ) = ,1
Kcc,s
2. Transport the monochromatic colour-corrected flux density at the PACS reference
wavelength to a neighbouring wavelength along the given object SED:
f,s (lc ) = f,s (0 ) Klc,s
6.5.3.2
Colour correction factor tabulation
The correction factors listed in Tables 6.7 to 6.9 are taken from M
uller et al. (2011, PICCME-TN-038, v1.0, PACS Photometer Passbands and Colour Correction Factors for various
Source SEDs)
70.0
1.016
1.016
1.013
1.011
1.005
0.989
0.982
0.992
1.034
1.224
1.269
1.325
1.396
1.488
1.607
1.768
1.992
2.317
2.816
3.645
5.175
8.497
17.815
58.391
456.837
10 000 K
5 000 K
1 000 K
500 K
250 K
100 K
50 K
40 K
30 K
20 K
19 K
18 K
17 K
16 K
15 K
14 K
13 K
12 K
11 K
10 K
9K
8K
7K
6K
5K
blue
CC 70
ref [m]
PACS
1.034
1.033
1.031
1.029
1.023
1.007
0.985
0.980
0.982
1.036
1.051
1.069
1.093
1.123
1.162
1.213
1.282
1.377
1.512
1.711
2.024
2.554
3.552
5.774
12.259
100.0
green
CC 100
1.074
1.074
1.072
1.068
1.062
1.042
1.010
0.995
0.976
0.963
0.964
0.967
0.972
0.979
0.990
1.005
1.028
1.061
1.110
1.184
1.300
1.491
1.833
2.528
4.278
160.0
red
CC 160
1.359
1.356
1.337
1.312
1.258
1.081
0.794
0.672
0.507
0.286
0.262
0.237
0.212
0.187
0.162
0.137
0.114
0.091
0.071
0.052
0.035
0.022
0.012
0.005
0.002
0.961
0.962
0.963
0.965
0.970
0.987
1.023
1.043
1.078
1.153
1.165
1.179
1.194
1.212
1.233
1.257
1.285
1.318
1.359
1.410
1.475
1.559
1.676
1.845
2.110
PACS blue band

from 70.0 m to
IRAS
MIPS
60.0
71.42
1.234
1.233
1.224
1.214
1.192
1.118
0.980
0.911
0.802
0.617
0.591
0.564
0.536
0.505
0.473
0.438
0.401
0.362
0.321
0.277
0.232
0.186
0.140
0.096
0.056
0.907
0.908
0.910
0.914
0.921
0.946
0.999
1.031
1.088
1.217
1.239
1.264
1.293
1.326
1.364
1.409
1.463
1.529
1.610
1.714
1.849
2.034
2.299
2.706
3.400
PACS green band

from 100.0 m to
ISOPHOT
90.0
105.0
1.305
1.304
1.298
1.289
1.271
1.212
1.105
1.049
0.955
0.781
0.756
0.729
0.699
0.667
0.633
0.596
0.555
0.512
0.464
0.413
0.358
0.300
0.238
0.175
0.114
1.053
1.053
1.052
1.051
1.048
1.039
1.022
1.012
0.995
0.959
0.953
0.947
0.940
0.932
0.923
0.913
0.901
0.888
0.872
0.853
0.831
0.804
0.771
0.729
0.674
0.886
0.886
0.888
0.891
0.896
0.913
0.948
0.968
1.005
1.090
1.104
1.121
1.140
1.162
1.187
1.217
1.253
1.296
1.349
1.415
1.501
1.615
1.775
2.013
2.401
PACS red band

from 160.0 m to
FIS MIPS ISOPHOT
140.0 155.9
170.0
Table 6.7: Photometric colour corrections for a range of different blackbody temperatures from 5 K to 10 000 K. Bold values in columns
2 to 4 are the Kcc,BB factors for the PACS bands. Columns 5 to 11 contain the Klc,BB factors needed to obtain a monochromatic flux
density at neighbouring key-wavelengths used by other missions.

97
70.0
5.150
2.013
1.442
1.141
1.058
1.026
1.001
0.996
4.307
1.790
1.322
1.081
1.019
0.998
0.987
0.987
3.116
1.456
1.144
1.000
0.976
0.975
0.989
1.001
2 B (10 K)
2 B (15 K)
2 B (20 K)
2 B (30 K)
2 B (40 K)
2 B (50 K)
2 B (75 K)
2 B (100 K)
1 B (10 K)
1 B (15 K)
1 B (20 K)
1 B (30 K)
1 B (40 K)
1 B (50 K)
1 B (75 K)
1 B (100 K)
+1 B (10 K)
+1 B (15 K)
+1 B (20 K)
+1 B (30 K)
+1 B (40 K)
+1 B (50 K)
+1 B (75 K)
+1 B (100 K)
blue
CC 70
ref [m]
PACS
1.541
1.092
0.999
0.973
0.983
0.995
1.018
1.031
1.916
1.249
1.086
1.003
0.988
0.986
0.990
0.994
2.162
1.356
1.152
1.037
1.008
0.998
0.993
0.993
100.0
green
CC 100
1.083
0.956
0.957
0.997
1.029
1.051
1.082
1.098
1.319
1.045
0.988
0.974
0.981
0.988
1.001
1.008
1.495
1.126
1.035
0.993
0.987
0.987
0.991
0.994
160.0
red
CC 160
0.060
0.189
0.334
0.591
0.784
0.926
1.144
1.262
0.044
0.139
0.245
0.434
0.576
0.680
0.841
0.927
0.038
0.119
0.210
0.372
0.494
0.583
0.721
0.794
1.382
1.208
1.130
1.057
1.022
1.003
0.979
0.968
1.438
1.257
1.175
1.099
1.064
1.043
1.018
1.007
1.466
1.282
1.199
1.121
1.085
1.064
1.038
1.027
PACS blue band

from 70.0 m to
IRAS
MIPS
60.0
71.42
0.308
0.525
0.685
0.891
1.013
1.089
1.192
1.242
0.250
0.425
0.555
0.722
0.820
0.882
0.966
1.006
0.225
0.383
0.499
0.650
0.738
0.794
0.869
0.905
1.632
1.299
1.159
1.036
0.982
0.952
0.916
0.901
1.800
1.432
1.278
1.142
1.082
1.049
1.010
0.993
1.890
1.504
1.342
1.199
1.136
1.102
1.061
1.043
PACS green band

from 100.0 m to
ISOPHOT
90.0
105.0
0.472
0.723
0.893
1.092
1.198
1.263
1.346
1.386
0.361
0.554
0.683
0.836
0.918
0.967
1.030
1.061
0.316
0.485
0.598
0.731
0.803
0.846
0.902
0.928
0.876
0.947
0.984
1.021
1.039
1.049
1.061
1.066
0.831
0.899
0.935
0.970
0.986
0.996
1.007
1.012
0.810
0.876
0.911
0.945
0.961
0.970
0.981
0.986
1.332
1.118
1.026
0.946
0.911
0.892
0.869
0.859
1.503
1.262
1.158
1.067
1.028
1.007
0.982
0.970
1.597
1.341
1.230
1.134
1.092
1.070
1.043
1.031
PACS red band

from 160.0 m to
FIS MIPS ISOPHOT
140.0 155.9
170.0
Table 6.8: Photometric colour corrections for different modified black-bodies (Fmodif ied = B (, T )). Bold values in columns 2 to
4 are the Kcc,mBB factors for the PACS bands. Columns 5 to 11 contain the Klc,mBB factors needed to obtain a monochromatic flux
density at neighbouring key-wavelengths used by other missions.
98
70.0
2.691
1.333
1.080
0.977
0.970
0.979
1.005
1.023
+2 B (10 K)
+2 B (15 K)
+2 B (20 K)
+2 B (30 K)
+2 B (40 K)
+2 B (50 K)
+2 B (75 K)
+2 B (100 K)
blue
CC 70
ref [m]
PACS
1.399
1.038
0.974
0.974
0.997
1.017
1.049
1.067
100.0
green
CC 100
1.009
0.943
0.971
1.037
1.083
1.113
1.155
1.175
160.0
red
CC 160
0.070
0.220
0.390
0.689
0.915
1.080
1.335
1.472
1.355
1.185
1.108
1.036
1.002
0.983
0.959
0.949
PACS blue band

from 70.0 m to
IRAS
MIPS
60.0
71.42
0.342
0.583
0.761
0.991
1.125
1.210
1.325
1.380
1.555
1.237
1.104
0.987
0.935
0.907
0.873
0.858
PACS green band

from 100.0 m to
ISOPHOT
90.0
105.0
0.540
0.827
1.020
1.248
1.370
1.443
1.538
1.584
0.899
0.972
1.010
1.048
1.066
1.076
1.089
1.094
1.253
1.052
0.965
0.890
0.857
0.840
0.818
0.809
PACS red band

from 160.0 m to
FIS MIPS ISOPHOT
140.0 155.9
170.0
Table 6.8: Photometric colour corrections for different modified black-bodies (Fmodif ied = B (, T )) continued. Bold values in
columns 2 to 4 are the Kcc,mBB factors for the PACS bands. Columns 5 to 11 contain the Klc,mBB factors needed to obtain a
monochromatic flux density at neighbouring key-wavelengths used by other missions.

99
70.0
1.043
1.016
1.000
0.995
1.000
1.016
1.043
=-3.0
=-2.0
=-1.0
= 0.0
= 1.0
= 2.0
= 3.0
blue
CC 70
ref [m]
PACS
1.037
1.012
1.000
1.000
1.011
1.034
1.069
100.0
green
CC 100
1.056
1.017
1.000
1.004
1.029
1.075
1.142
160.0
red
CC 160
0.630
0.735
0.857
1.000
1.167
1.361
1.588
1.061
1.040
1.020
1.000
0.980
0.961
0.942
PACS blue band

from 70.0 m to
IRAS
MIPS
60.0
71.42
0.729
0.810
0.900
1.000
1.111
1.235
1.372
1.158
1.103
1.050
1.000
0.952
0.907
0.864
PACS green band

from 100.0 m to
ISOPHOT
90.0
105.0
0.670
0.766
0.875
1.000
1.143
1.306
1.493
0.925
0.949
0.974
1.000
1.026
1.053
1.081
1.199
1.129
1.062
1.000
0.941
0.886
0.833
PACS red band

from 160.0 m to
FIS MIPS ISOPHOT
140.0 155.9
170.0
Table 6.9: Photometric colour corrections for a range of different power-law spectra (F ). Bold values in columns 2 to 4 are
the Kcc, factors for the PACS bands. Columns 5 to 11 contain the Klc, factors needed to obtain a monochromatic flux density at
neighbouring key-wavelengths used by other missions.
100
6.6. TABLE OF THE CALIBRATION FILES
6.6
Table of the calibration files
101
102
Chapter 7
The PACS Pipelines

There is a Herschel-wide convention on the processing levels of the different instruments. All
observations are processed automatically by SPG (systematic product generator) pipelines at
the HSC, and in the ObservationContexts (Chp. 8) from PACS are data at Levels 0, 0.5, 1,
2, 2.5, and 3. A more detailed product description is given in the PACS Products Explained
(link PPE) and the PACS Data Reduction Guide (link PDRG). A product description is also
provided in Sec. ??.
In this chapter we will explain the SPG and the interactive pipelines that produce the ObservationContexts provided via the HSA. The SPG pipelines are AOT-dependent, and are
based on one of the interactive pipelines provided for each AOT. Other pipeline scripts are
provided via HIPE for interactive processing and/or to provide an alternative processing for
some AOTs. For all pipeline scripts, the red and blue camera data must be processed separately, and most products produced have a red and a blue version; these are identified by an
R or a B at the end of the product name. All product names start with HPS (Herschel PACS
Spectroscopy), or HPP (Herschel PACS Photometry).
Link to somewhere where the SPG and ipipe scripts are provided outside of HIPE. These
should be suitable commented-in, so that those who will never run HIPE will be able to
understand the important things: the tasks and parameters that are important (e.g. up and
oversample), and in the ipipe where the differences are compared to the SPG
7.1
Spectroscopy pipelines
There are four groups of spectroscopy pipeline to deal with: chop-nod range spectroscopy,
chop-nod line spectroscopy, unchopped range spectroscopy, and unchopped line spectroscopy
(which also deals with the few wavelength switching AOTs in the archive: Sec. 3.3.2.5). Within
each group there are several pipeline scripts, which differ in the calibration scheme used or
can be used to process particular sources, produce combined observations, or to inspect
intermediate pipeline products. These scripts can be found in the HIPE Pipelines menu (and
included as tarball somewhere?). Every pipeline script can deal with observations that include
multiple wavelength ranges and with pointed and mapping AORs (Sec. 3.3.3).
In addition, useful scripts are provided in the HIPE Scripts menu that show how to fit spectra,
combine observations, or to produce the so-called standalone browse products (Sec. ??). The
103
104
CHAPTER 7. THE PACS PIPELINES
scripts in these two HIPE menus are explained in the link PDRG and PPE. In this section we
will explain the spectroscopy SPG pipeline scripts that are used to produce the products you
obtain from the HSA. After that, the interactive pipeline scripts are explained, emphasising
the differences they provide with respect to the SPG scripts.
An overview of the data products found in the different levels of an observation:
Level 0 products are the starting point for scientific data reduction. For all standard
observing modes, the scientific information is contained in objects called HPSFITB/R
(which are of the Frames product class). These constitute the starting point for the
pipeline, and are contained in a layer of the ObservationContext called level0 (Table 8.3.1). The Frames contain all measured ramps, i.e. the fitted slopes of the detector,
and have the dimensions 18 25no. readouts link to where explain ramps. Correct
interpretation of the signal requires knowledge of the instrument status at every time.
That can be found in sub-structures of these Frames. Other Level 0 products contain
other useful information, such as the spacecraft pointing, time correlation, and selected
spacecraft housekeeping information, raw data (i.e. un-fitted ramps) for some selected
pixels, etc. Finally, the calibration data needed for data reduction are also Level 0
products.
Level 0.5. The pipeline from Level 0 to 0.5 is AOT-independent. The first steps
of the pipeline processing do the following: basic unit conversions (digital readouts
to Volts/s); processing flags and masks (saturation, damaged pixels, signals affected by
chopper and grating transitions); generate block selections and organise the data around
logical blocks; calibrate the wavelengths (including the Herschel velocity correction); and
the centre-of-field coordinates are computed for every frame and sky coordinates are
assigned to every pixel. The output products here are still HPSFITB/R, of dimensions
18 25no. of readouts, and there is one (red and blue) Frame per wavelength range
requested by the observer, and per pointing for raster observations.
Level 1. The generation of Level 1 products is partly AOT-dependent. Level 0.51
processing adds further status information to the product (e.g. chopper angle) and
more masking is done (a de-glitching). For chop-nod AOTs the differential signals are
created (Sec. ??) during which the dark is subtracted, and for unchopped AOTs the dark
is subtracted and a transients correction may be done. The spectral flatfielding is also
done at this level (Sec. ??). Finally, detector readouts are flux calibrated and converted
to physical units (Jy or in units of the telescope flux: Sec. ??). The spectroscopy Level
1 final product is a cube, of class PacsCube, HPS3DB/R, of size 5 5no. of readouts.
Level 2: Further-processed Level 1 data to such a level that scientific analysis can
be performed. The data are spectrally regridded (creating PacsRebinnedCubes and
masked for outliers. For the line spectroscopy unchopped observations, the off-source
(telescope background) spectra are subtracted. For mapping observations, mosaic cubes
are created. Various spectrum tables are also created.
Level 2.5. For unchopped range spectroscopy observations for which the observer
requested an off-source observation, the products at this level are created from the
Level 2 rebinned cubes taken from the on-source observation, from which the Level 2
7.1. SPECTROSCOPY PIPELINES
105
rebinned cubes of the off-source observation have been subtracted. All subsequent cubes
and spectral tables are then produced from these Level 2.5 PacsRebinnedCubes.
Level 3. At this level spectral tables for chop-nod pointed SED observations can be
found link.
7.1.1
SPG pipelines: Level 0 to 0.5
All pipelines start by grabbing the observation and the calibration tree. They process one
camera at a time: the blue and red data are reduced separately, and then pushed into one
ObservationContext, to then be provided to the users via the HSA.
The steps in the first part of the pipeline, being the same for all four SPG pipeline scripts,
are:
Extract the pointing product and the orbit ephemeris (to later add pointing information), the horizons product (used for Solar system object coordinate setting), and time
correlation (used to set the UTC in the data) from the ObservationContext.
Extract two science data products the slicedFrames containing all the science data
and the slicedRawRamps which contain slopes for a single pixel and will be used in the
saturation-masking task and the DecMec data product, which contains the detector
and mechanism information, and will be used to organise the data into logical blocks.1
Mask out saturated data in the slicedFrames, adding the masks SATURATION and
RAWSATURATION.
Convert the units from ADU to V/s (calibration file Readouts2Volts), detect where the
calibration block data lie, add the UTC to each readout.
Add the pointing information of the central spaxel (uses the Siam calibration file), and
for Solar system objects move the target to a fixed position on the sky.
Add grating and chopper information (using calibration file ChopperThrowDescription)
and convert the chopper readouts to an angle wrt. the focal plane unit and the sky (using
calibration files ChopperAngle and ChopperSkyAngle).
Add the sky positions to each pixel (the Ra, Dec datasets) using the calibration files
ArrayInstrument and ModuleArray.
Add wavelengths to each pixel (the Wave dataset), using the calibration file WavePolynomes. Correct the wavelengths for the spacecraft velocity, using the pointing,
orbitEphemeris and TimeCorrelation product. This task also sets a mask, OUTOFBAND, to flag data-points in spectral regions that lie outside of the wavelength range
of the band.
Find the logical blocks of this observation (calibration file ObcpDescription).
1
The word sliced in slicedFrames (etc.) refers to the fact that while all the science and calibration data
of the observation are held in a single products, within that product the data are organised into blocks, aka
slices.
106

Mask out bad and noisy pixels in the masks BADPIXELS and NOISYPIXELS (using
the calibration files BadPixelMask and NoisyPixelMask). Mask the data take while the
chopper and gratings were moving (Sec. ??), placing them in the masks UNCLEANCHOP and GRATMOVE. None of these masked data will be used any more in the
pipeline.
Finally, slice explain/link to slicing explanation the data by Line/Range, Raster Point,
nod position, nod cycle, on/off position and per band. In this way, the data can be
reduced with the pipeline while still honouring differences in pointing on the sky, wavelength range (as requested by the observer), nod position (chop-nod AOTs) and on or
off pointing (unchopped AOTs).
The task that masks for saturation requires a little more explanation. In the main mask,
SATURATION, a flag (with a value 1) is set for all datapoints above or below? the limit
found in the calibration file SignalSatLimits. A warning mask, RAWSATURATION, is also
added, to indicate potential saturation. This warning is based on an inspection of the raw
slopes that are found in slicedRawRamps2 . These raw slopes are a more secure way to detect
saturation, however it was not possible to downlink from Herschel the raw slopes for all 1625
pixels for both detectors: we only got the slopes from one pixel of one module for the red and
the blue detector each. Therefore, these slopes are compared to the raw saturation limits in
the calibration file RampSatLimits, and if saturation is detected then all the readouts taken
at that same time for all the pixels of all the modules are also flagged. But this mask is only
a warning: this single pixel is not only the most responsive (and hence will saturate earlier),
but being in the centre of the FoV is more likely to have the brightest part of the source in
it (and hence more likely to have high fluxes). Therefore, saturation in this pixel does not
necessarily mean saturation in all pixels of all modules. It was intended that the users inspect
their data to decide whether to use this warning mask, or not. In the SPG pipelines, this
mask is created but it is not activated when the final products are created.
7.1.2
SPG pipelines: Level 0.5 to 2
At the second (0.51) and third (12) levels of the pipeline there are some tasks that the
user can interact with, and hence can change the spectra in the final products. Naturally,
in the SPG pipeline there is no possibilty of interaction, and for these tasks the parameters
used are those found to work well for most observations. Those who wish to know what these
parameter settings were can inspect the pipeline scripts link!.
The pipeline steps that are followed by all AOTs:
Mask out glitches (adding the mask GLITCH), going over the spectrum of each of the
16 pixels of each of the 25 modules, following the time-line spectrum rather than vs.
wavelength.
Convert the signals the value they would be if the observation was taken at the standard
capacitance (see Sec. ??). This uses the capacitanceRatios calibration file.
2
The PACS detector measured integration ramps, i.e. 32 check separate non-destructive, integrated readouts
were taken before a destructive readout. The slope of the 32 readouts was fitted on-board Herschel by a
polynomial. It was the the fitted slope, aka ramp, that was downlinked. With only one value of a slope rather
than 32 check, the bandwidth necessary was much reduced.
107
Here the unchopped and chop-nod pipelines diverge.

7.1.2.1
Chop-nod AOTs
Following on from Sec. 7.1.2, the pipeline steps are:

Calculate the differentialratio signal by subtracting the off-source readouts from the
on-source readouts and dividing by the sum (see below). At this point the data are
background subtracted and are in units of telescopes.
The order of the next steps depends on the specific AOT:
1. For line spectroscopy, the next step is to convert the format of the data from
Frames (with dimensions 18 25no. of readouts) to the first cube of the pipeline,
PacsCubes, with dimensions 5 5(16no. readouts). In each of the 25 spaxels of
this cube, you will find the spectra in each of the 16 pixels (these spectra themselves
being repeated if multipled scans on the wavelength range were requested).
2. Then the spectral flatfielding is done (see below).
3. For range spectroscopy, the flatfielding task is done first, and the PacsCubes
created second.
4. For line spectroscopy for Nyquist and oversampled mapping observations (see Sec. 3.3.3.3) for which a drizzled cube is created, one extra step is done
before the PacsCube is created and the flatfielding done. This is to convert the flux
units of the Frames from telescopes to Jansky, using the conversion in the calibration file offRatio (the telescope spectrum for the so-called asymmetric chopping
case: need to link to a TN to explain this). The calibration is scaled by the OD of
the observation (specifically, the temperature of the mirror, which aged over the
mission).
Formally, the end of SPG Level 1 is the creation of the PacsCube, with no flatfielding
applied. The flatfielding is the first step of the SPG Level 12 pipeline. The Level 1
PacsCubes (Table 8.3) in any ObservationContext downloaded from the HSA are not
flatfieled, while the Level 2 PacsCubes are.
Some of these tasks require a little more explanation.
The differentialratio signal. The standard observing pattern for chop-nod observations had two consecutive on-on-off-off-off-off-on-on chopping patterns per grating plateau3
(Sec. 3.3.1.1). The task breaks this pattern into four block (on-on-off-off and off-off-on-on),
and computes the pairwise differenceratios for each: (on(a)-off(b))/(on(a)+off(b)), where
a and b are 1st and 2nd and all four combinations of the ratio computed. This subtracts
the telescope background signal (the off-source signals) and converts the signals into units of
telescopes link to where this is explained better - AP wrote a note on this which I have, and
it should go into HELL but need to know if it will go as-is or if a formal note will be written.
Conversion to Jy by multiplying by a smoothed telescope background spectrum is done later
in the pipeline.
3
plateau refers to a single chopper, nod, or grating position where a sequence of data-points were collected
108
The spectral flatfielding. This task is explained in detail in link to best place to explain
the FF - PRRG?. In each spaxel of the PacsCube there are 16no. repeats spectra. These
spectra are all slightly offset wrt. each other, because of response drifts during the observing
and because of the slight response differences of the 16 detector pixels. If you were to combine
(average) the spectra together, the scatter in the resulting spectrum would not just reflect the
noise of the spectra, but the offset of each spectrum wrt. the rest. Therefore, the individual
spectra of each spaxel are normalised to the flux level of the mean spectrum of the spaxel:
this brings those spectra with fluxes too high, down, and those spectra with fluxes too low,
up. The mean spectrum of the spaxel is again calculated, and now its scatter/noise will not
include the detector offsets.
The essence of this task is the same for the line and range spectroscopy AOTs, although the
actual task doing the work is different. For line spectroscopy, a few steps are performed before the actual flatfielding task, to identify and then mask out spectral lines in each spectrum
of each spaxel of the cube so that the normalisation is done on continuum signals only. For
range spectroscopy, rather than identifying the lines to mask them out, the spectra are heavily
smoothed and then fitted with a spline, and these fits are used to compute the normalisation
factors.
The flatfielding tasks allow users to limit the flatfielding to certain wavelength regions, e.g.
to concentrate on a certain part of the spectrum, but of course this is not done in the SPG
pipeline. However, what is done in the SPG pipeline is to not flatfield the regions that lie
in the red leak (Sec. ??). These data are flagged in the mask NOTFFED. link to where
describe either red leak HPDP or red leak spg script and check if any blue regions are also
excluded.
The flatfielding tasks are the first place where users can change the final spectra by changing
the parameters in the flatfielding task(s): e.g. by manually identifying lines for the line spectroscopy task (e.g. for wide or absorption lines), or by changing the spline knots or using a
polynomial fitting instead for the range spectroscopy task. The parameters used in the SPG
pipeline, however, work well for the majority of observations.
Line spectroscopy drizzled cubes. To be able to create drizzled cubes for line spectroscopy
mapping observations (Sec. 8.3.3.2), the step to convert the data from telescopes to Jy is
taken at this stage of the pipeline, while the same conversion occurs near the end of the
pipeline for all other AOTs. The reason is a formality: the drizzle task to create the sciencegrade (Level 2) drizzled cubes works on PacsCubes, and these need to be fully calibrated
before the drizzle task runs. The tasks that create all other science-grade (Level 2) cubes
work on the so-called PacsRebinnedCubes (see below), and for these the calibration to Jy
comes later. This is the only difference in the pipeline between these particular mapping
line spectroscopy chop-nod observations, and the pointed and all other mapping chop-nod
observations.
There is, however, a slight difference in the calibration applied to the PacsCubes at this point
in the pipeline, compared to that applied to the PacsRebinnedCubes later. The means that
the calibration used for all drizzled cubes will be slightly different to the calibration used for
all other cubes. That applied to the PacsRebinnedCubes was the standard calibration used in
the pipeline before the drizzle task was introduced. The calibration applied to the PacsCubes
is one that was originally created for the pointing offset correction pipeline script (Sec. ??)
which needs to also calibrate PacsCubes. The two calibrations differ very slightly: Say how
much!.
109
The next pipeline steps, taking us close to the end of the pipeline, are:
Create a single wavelength grid over which to spectrally combine the individual spectra
in each spaxel of the PacsCube.
Detect glitches (creating the mask OUTLIERS), working this time with the spectra in
wavelength space. (There is nothing that can usefully be changed with this task.)
Activate all the masks for which the data flagged as bad are not wanted in the any of
the final pipeline cubes. In the SPG the masks that are activated are: OUTOFBAND,
GLITCH, UNCLEANCHOP, SATURATION, GRATMOVE, BADFITPIX (a useless
mask), OUTLIERS, BADPIXELS, and NOTFFED check.
The second cube of the pipeline is now created, the PacsRebinnedCubes. The multiple
spectra from the PacsCube are combined along the wavelength grid into a single spectrum, separately for each spaxel. (The wavelength grid of each spaxel and for each cube
of the same wavelength range is exactly the same.)
By this point in the pipeline, all the data from each slice have been fully reduced: cleaned
of bad data, corrected for the instrument response, flux calibrated (either in telescopes
or Jy), wavelength calibrated (microns), flatfielded, and spectrally combined. Multiple
PacsRebinnedCube slices will exist if more than one wavelength range was requested
for the observation and for mapping observations (multiple raster positions). There are
always separate nod (A and B) and nod cycle (for a nod repetition factor > 1) slices.
The next stage is then to combine nod A and nod B slices, to create one cube for each
wavelength range the observer requested, and for each separate raster position.
Then the conversion of the flux units from telescopes to Jy is done, for the line and
range spectroscopy SPG scripts but not for the line spectroscopy drizzle script (see
above), as that conversion was done earlier.
Finally, the RA, Dec coordinates of the central spaxel of the central cube is pushed into
the ra and dec meta data.
Some extra detail on the wavelength grid created in the first step. This grid is created
to allow the individual spectra contained in each spaxel of the PacsCube to be combined
into a single spectrum per spaxel. This is done in the following way: for each bin of the
wavelength grid, the data of the same wavelength range in each spaxel of the PacsCubes are
averaged, and these 25 average spectra are placed in an image dataset of the new cube
a PacsRebinnedCube. The standard deviation is placed in a stddev dataset, which later
becomes the weights and error datasets of all subsequent products. The wavelength grid
chosen will obviously dictate the spectral sampling achieved in the spectra of the final cubes
(not the resolution, only the sampling).
This wavelength grid scales with the spectral resolution, which scales with wavelength. Bins
are generally smaller at shorter wavelengths and larger at longer wavelengths. This is done so
that the spectral sampling is at least Nyquist at all wavelengths, i.e. the 50 m to 220 m
stretch of PACS. The task that creates the wavelength grid choses one appropriate for the
type of observation the detail may change in 14.2, e.g. we do not distinguish between sed
with rep factor 1 and greater and will we go for up=1 or not? Explain here when final SPG
110
has been written. It uses the calibration file wavelengthGrid or littrowPolynomes? to get the
details of the spectral resolution for the band of the cube it is working on. To learn more
about how changing the wavelength grid can change the appearance of the final spectra, see
link PDRG.
Errors: the information about stddev and errors in the PDRG, where do I put that? Here in
the pipeline chapter, or in the calibration/in-flight performance chapter?
7.1.2.2
Unchopped AOTs
Most of the pipeline tasks for the unchopped mode are the same as for the chop-nod mode,
so this section repeats much of what is written above. The differences are related to the
fact that for unchopped observations, the off-source data come from a separate pointing (line
spectroscopy) or a separate obsid (range spectroscopy).
Following on from Sec. 7.1.2, the next pipeline steps are:
Derive the response of the detector (red or blue) from the calibration block data of the
observation (using calibration file observedResponse and calSourceFlux). Subtract the
nominal dark current, taken from the calibraiton file darkCurrent.
Divide by the Relative Spectral Response Function (RSRF) taken from the calibration
file rsrf[R1,B2B,B3A,B2A], and then divide by the response computed previously.
The order of the next steps depends on the specific AOT:
1. For line spectroscopy, the next step is to convert the format of the data from
Frames (with dimensions 18 25no. of readouts) to the first cube of the pipeline
(PacsCube), with dimensions 55(16no. readouts). In each of the 25 spaxels of
this cube, you will find the spectra in each of the 16 pixels (these spectra themselves
being repeated if multipled scans on the wavelength range were requested).
2. Then the spectral flatfielding is done (see below).
3. For range spectroscopy, the flatfielding task is done first and the PacsCubes
created second.
Formally, the end of SPG Level 1 is the creation of the PacsCube, with no flatfielding
applied. The flatfielding is the first step of the SPG Level 12 pipeline. The Level 1
PacsCubes (Table 8.3) in any ObservationContext downloaded from the HSA are not
flatfielded, while the Level 2 PacsCubes are.
Some extract detail on the the spectral flatfielding. This task is explained in detail in link
to best place to explain the FF - PRRG?. In each spaxel of the PacsCube there are 16no.
repeats spectra. These spectra are all slightly offset wrt. each other, because of response
drifts during the observing and because of the slight response differences of the 16 detector
pixels. If you were to combine (average) the spectra together, the scatter in the resulting
spectrum would not just reflect the noise of the spectra, but the offset of each spectrum wrt.
the rest. Therefore, the individual spectra of each spaxel are normalised to the flux level of
the mean spectrum of the spaxel: this brings those spectra with fluxes too high, down, and
111
those spectra with fluxes too low, up. The mean spectrum of the spaxel is again calculated,
and now its scatter/noise will not include the detector offsets.
The essence of this task is the same for the line and range spectroscopy AOTs, although the
actual task doing the work is different. For line spectroscopy, a few steps are performed before the actual flatfielding task, to identify and then mask out spectral lines in each spectrum
of each spaxel of the cube so that the normalisation is done on continuum signals only. For
range spectroscopy, rather than identifying the lines to mask them out, the spectra are heavily
smoothed and then fitted with a spline, and these fits are used to compute the normalisation
factors.
The flatfielding tasks allow users to limit the flatfielding to certain wavelength regions, e.g.
to concentrate on a certain part of the spectrum, but of course this is not done in the SPG
pipeline. However, what is done in the SPG pipeline is to not flatfield the regions that lie
in the red leak (Sec. ??). These data are flagged in the mask NOTFFED. link to where
describe either red leak HPDP or red leak spg script and check if any blue regions are also
excluded.
The flatfielding tasks are the first place where users can change the final spectra by changing
the parameters in the flatfielding task(s): e.g. by manually identifying lines for the line spectroscopy task (e.g. for wide or absorption lines), or by changing the spline knots or using a
polynomial fitting instead for the range spectroscopy task. The parameters used in the SPG
pipeline, however, work well for the majority of observations.
The next pipeline steps, taking us close to the end of the pipeline, are:
Create a single wavelength grid over which to spectrally combine the individual spectra
in each spaxel of the PacsCube.
Detect glitches (creating the mask OUTLIERS), working this time with the spectra in
wavelength space. (There is nothing that can usefully be changed with this task.)
Activate all the masks for which the data they flag as bad are not wanted in the any of
the final pipeline cubes. In the SPG the masks that are activated are: OUTOFBAND,
GLITCH, UNCLEANCHOP, SATURATION, GRATMOVE, BADFITPIX (a useless
mask), OUTLIERS, BADPIXELS, and NOTFFED check.
The second cube of the pipeline is now created, the PacsRebinnedCubes. The multiple
spectra from the PacsCube are combined along the wavelength grid into a single spectrum, separately for each spaxel. (The wavelength grid of each spaxel and for each cube
of the same wavelength range is exactly the same.)
By this point in the pipeline, all the data from each slice have been fully reduced:
cleaned of bad data, corrected for the instrument response, flux calibrated (either in
telescopes or Jy), wavelength calibrated (microns), flatfielded, and spectrally combined.
Multiple PacsRebinnedCube slices will exist if more than one wavelength range was
requested for the observation and for mapping observations (multiple raster positions).
There are always slices for the on- and off-positions for line spectroscopy AOTs (for
range spectroscopy there is only either on-source or off-source data in the observation),
and for repetitions of the on-off cycle or the grating scan.
For line spectroscopy AOTs, the data of the off-source cube(s) are subtracted from those
of the on-source cubes(s).
112

For line spectroscopy AOTs, the repeated on- and off-source pointings are combined,
and for range spectroscopy AOTS repeates on the grating scan are combined check that
repeats on grating are not always also separate slices, to create one cube for each wavelength range the observer requested, and for each separate raster pointing for mapping
observations.
Finally, the RA, Dec coordinates of the central spaxel of the central cube is pushed into
the ra and dec meta data.
For unchopped range spectroscopy AOTs, one final step is necessary. The on-source and
off-source data come from separate observations (if the observer requested an off-source
observation). These observations are reduced separately to this stage of the pipeline.
Then the off-source subtraction is done, using the data from the off-source observation.
This forms the Level 2.5 of an unchopped range spectroscopy on-source observation.
Some extra detail on the wavelength grid created in the first step. This grid is created
to allow the individual spectra contained in each spaxel of the PacsCube to be combined
into a single spectrum per spaxel. This is done in the following way: for each bin of the
wavelength grid, the data of the same wavelength range in each spaxel of the PacsCubes are
averaged, and these 25 average spectra are placed in an image dataset of the new cube
a PacsRebinnedCube. The standard deviation is placed in a stddev dataset, which later
becomes the weights and error datasets of all subsequent products. The wavelength grid
chosen will obviously dictate the spectral sampling achieved in the spectra of the final cubes
(not the resolution, only the sampling).
This wavelength grid scales with the spectral resolution, which scales with wavelength. Bins
are generally smaller at shorter wavelengths and larger at longer wavelengths. This is done so
that the spectral sampling is at least Nyquist at all wavelengths, i.e. the 50 m to 220 m
stretch of PACS. The task that creates the wavelength grid choses one appropriate for the
type of observation the detail may change in 14.2, e.g. we do not distinguish between sed
with rep factor 1 and greater and will we go for up=1 or not? Explain here when final SPG
has been written. It uses the calibration file wavelengthGrid or littrowPolynomes? to get the
details of the spectral resolution for the band of the cube it is working on. To learn more
about how changing the wavelength grid can change the appearance of the final spectra, see
link PDRG.
Errors: the information about stddev and errors in the PDRG, where do I put that? Here in
the pipeline chapter, or in the calibration/in-flight performance chapter?
7.1.3
7.1.3.1
SPG pipelines: final steps

Pointed observations
For pointed observations, the SPG creates two cubes (Sec. 8.3.3.2), and two or three spectral
tables (Sec. 8.3.3.3).
The first of the cubes is the PacsRebinnedCubes, which are the end product of the previous
section (for all AOTs). For observations of point or semi-extended sources, it is from these
cubes that the final spectra will be extracted. The second cube is an interpolated cube
113
(Sec. 8.3.3.2). For observations of extended sources, science can be done with the rebinned
and the interpolated cubes is that right, or do we still only recommend this as a browse
product?, depending on the science-case link to doc to guide users thru this morass. The
interpolated cubes have spaxels of 300 size (compared to the 900 .4 native spaxel size of the
rebinned cubes).
The Level 2 spectral tables created by the SPG pipeline are: the data of the rebinned cubes,
but organised as a large table instead of as a cube (Sec. 8.3.3.3); a table with the centralspaxel spectrum and one or two point-source corrected spectra (Sec. 8.3.3.3), both taken from
the PacsRebinnedCubes. If the observation is of a point source that is located in the central
spaxel of the cube, then the point-source spectra of this table can be used for science: see
link to PDRG or a cheat-sheet for more information about using these point-source corrected
spectra. if create POC results that can also be used and also for off-centred point sources,
then add that information here.
One extra Level 3 table is provided (Sec. 8.3.3.3), but only for the chop-nod pointed observations taken in SED range spectroscopy mode: the table of point-source corrected spectra,
with the red and blue camera results combined into a single table (for all other products, the
red and blue products are separate), and with the data from all observations that were taken
of that source in SED mode, combined. From this table you can extract the combined SED
of a source. If the source is moreoever a point source located in the central spaxel, then this
spectrum can be used for science.
7.1.3.2
Mapping observations
For mapping observations, the SPG creates two cubes (Sec. 8.3.3.2), and one spectral table
(Sec. 8.3.3.3).
The first of the cubes produced, for all raster types, is the PacsRebinnedCubes, which are the
end product of the previous section (for all AOTs). The data of this cube is also provided in
the form of a rebinned cube spectrum table (Sec. 8.3.3.3).
Which cube is produced as the second cube depends on the details of the raster. See
Sec. 8.3.3.1, and in particular Table 8.3.3.1, for information about the different mapping
modes.
For Tiling observations (i.e. with a large offset, of order a spaxel in size) interpolated cubes (spaxel size 300 ) and projected cubes (spaxel size 1.500 ) are produced (see
Sec. 8.3.3.2).
For Nyquist-sampled range spectroscopy AOTs (with offsets of about 1/31/2 a cube)
these same two cubes are also produced, but the projected cubes have 300 -sized spaxels.
For oversampled range spectroscopy AOTs (offsets much less than a spaxel), only the
projected cubes (spaxel size 300 ) are produced.
For Nyquist and oversampled line spectroscopy AOTs, drizzled and projected cubes
are created. The spaxel size is optimised to the central wavelength and the requested
oversampling and upsampling factors (values 3 and 2 in the SPG pipeline) (link to
PDRG?).
114
In all cases there will be one red and one blue cube for each wavelength range in the observation.
7.1.4
The standalone browse products
The standalone browse products (SBPs) are explained in Sec. 8.5. These are produced as part
of the SPG pipeline: the rebinned cube spectrum tables for all AOTs, the tables of central and
point source spectra for all pointed AOTs, one of the mapping cubes but spectrally resampled
onto an equisitant wavelength grid for all mapping AOTs, and an equidistant interpolated
cube for pointed AOTs: these are taken from Level 2.5 for the on-source observation of unchopped range scans, and Level 2 from all others. For all pointed chop-nod range spectroscopy
AOTS, where the full SED of PACS was requested (in two or more observations on the same
source), this is also provided as an SBP, taken from Level 3.
7.1.5
7.2
7.2.1
7.2.1.1
The interactive pipelines: leave for later
Photometer pipelines
Scan map
SPG to Level 1
Two scripts take the raw data to a fully calibrated Frames product (calibrated data cubes in
Jy/detector pixel). The first one starts from Level 0 and prepares the data for the second
one. This script creates a so-called Level 0.5 product. It makes sure that all the needed meta
data entries are present, detects and removes the calibration blocks from the beginning of the
observation. It also adds the basic astrometric information to each frame. The second script
performs the basic calibration steps. These are:
flagging bad and saturated pixels
masking the columns affected by cross-talk
converting the digital units to volts (needed for the responsivity calibration)
applying the flat field and responsivity correction
applying the non-linearity correction (if neccessary)
applying the correction needed to alleviate the response changes of the bolometers due
to the changes in the evaporator temperature. Moor et al. (2015)
if the the target is a Solar System object performs a correction which converts the
observation into the reference frame of the moving object
7.2. PHOTOMETER PIPELINES

7.2.1.2
115
Level 1 to Level 2, 2.5, 3
The handling of data obtained in scan map mode depends strongly on the scientific goal.
There are three distinguishable cases:
High pass filtering and photProject: suitable for point (and slightly extended) sources.
Generalized Least Square (GLS) mapmakers: we provided two mapmakers - Unimap
and MadMad - that exploit the GLS method. Starting from SPG13, the maps from
Unimap replace those from MadMap in the Herschel Science Archive.
Destriper mapmaker: JScanam is the HIPE implementation of the IDL mapmaker called
Scanmorphos.
Unimap (MadMap) and JScanam are well suited for point and extended sources: they give
similar results so it is up to the user to decide which one to use. They exploit the redundancy
provided by the observations (mainly scan and cross-scan) and they generate Level2.5 products. The high pass filter method is different, it is applied to a single observation (Level2)
and then the individual maps (scan, cross-scan) are combined with the mosaic task to create
Level2.5 products. In the sections that follow we explain these pipelines. We give only a
short overview of the scripts. A step by step walkthrough of the processing can be found of
the PACS phot PDRG.
7.2.1.2.1
HighPass filter and PhotProject
This script processes scan map and mini-scan map observations containing mostly point-like
or relatively small extended sources. It starts from Level 1. The final map is in units of
Jy/map pixel. This script uses the high-pass filter (HPF) method to remove the 1/f noise.
This processing is not adequate for extended sources as it removes large-scale structures,
which cannot be properly protected (masked) from the filtering.
The script is able to combine several obsids but it is also useful for processing a single obsid.
It performs iterative masked highpass filtering of the timelines and projects the final map
using photProject: a first pass is done using a S/N filtering of the timeline, then a second
pass is done using a full circular patch masking of the source(s). The placement of the circular
patch is very important to ensure the correct positioning, therefore three options is proposed:
Source fitting: if the source is the brightest object in the field source fitting will find
the source and set the center of the patch to its fitted coordinates.
Target coordinate list: the script reads the coordinates given in a text file and sets the
coordinates to the center of the patch.
Target source list and source fitting: if the coordinates are only approximate a source
fitting is done on a sub-image centered on the given coordinates. The sub-image small
size ensures that the fitting does not diverge seeking the brightest object in the image.
116
Beside the appropriate placement of the mask the correct filtering of the data also relies on the
correct setup of the filter width. An optimum value is given as default but it can be modified
as desired and/or necessary for specific science cases. If the HPF width (the half-width of
the highpass filter) is too tight some extended features of the compact sources will be filtered
out, on the other hand a wider HPF width will increase the left over 1/f noise.
An important parameter for the combination HPF and PhotProject is the ratio between
the original pixel size, the re-gridding onto an output pixel (ouputpix) and the active pixel
fraction (pixfrac).
Both the output pixel size and the drop pixel size influence strongly the noise in the resulting
map. A smaller drop size or output pixel size will allow a better PSF sampling.
The processing itself consists of three major steps.
a first pass is done per observation masking the source(s) by signal-to-noise
a second pass is done by building a better mask from the combined map of the first
pass, again with a signal-to-noise thresholding
a third pass is performed by masking the central source or a list of sources with circular
patching
First we create a primary map for each individual obsids without highpass filtering then using
this map we construct a signal-to-noise ratio mask. This mask is then used to create a first
set of maps using the HPF with the new mask. then we cycle through the obsids and create
the first map and the first mask based on the S/N of the map.
Then we can do our first pass with the highpass filter and construct our first source mask
based on S/N. and perform a second pass using our mask to provide a better highpass filtering
on the original frames saved before the first pass. After creating our initial mask, and highpass
filtered data we now combine all of our maps to get a better S/N for the sources to be masked
out: Then we create a new sourcemask using the combined map. The mask is then applied
during the HPF process of the original restored frames, and the new frames are deglitched
according to the chosen method. Then we use the mosaic map to create a new multiple
source mask based on S/N. This multiple source mask will be combined during the third and
ultimate pass to a circular patch centered on the main source. If a file with the coordinate of
the expected source(s) on the maps are provided then we can create mask using this list and
a radius of 20 arcsec. But first we need to be sure that our coordinates are accurate so we
fit a gaussian on each source and calculate the exact Ra and Dec. If the source is too faint
for the fitter we just use the supplied coordinate. Of course if we are sure that our supplied
coordinates are perfect we can skip the fitting part by setting the doSourceFit variable to
False at the beginning of the script. If there is no target list given the script will use the
source coordinates given in HSPOT as a center for the circular patch. Then both the multiple
sources mask and a circular patch mask centered on the main source(s) are combined and
used during a final HP filtering on restored original frames. The final step is the actual map
making from Level 1. So the frames are:
Highpass filtered using the final combined mask
The turnovers are removed from the frames
117
The frames are finally projected onto the final individual maps taking into account the
chosen pixfrac and output pixel size. An extra set of map is created for comparison
with a pixfrac = 1
The final mosaics are combined for each given pixfrac.
If chosen, aperture photometry is performed on the source and several aperture on the sky
around the source to determine the uncertainties and S/N of the source. Results are given for
the given pixfrac and each the apertures: centered on source and on the sky are displayed.
The aperture on sky are chosen such that they overlap as little as possible with the source.
The sky aperture are displayed: blue all, red selected for aperture photometry of the sky
background. The photometry is given: un-corrected and corrected for the used aperture.
The on-source apertures can be modified according to the source size.
By definition, HPF removes the sky contribution, i.e. the skybackground values of the mosaics
should be distributed around a zero value. However, this is in theory, and the sky aperture
photometry gives a good indication of the residual noise. The sky photometry is also corrected
for the apertures used.
Each intermediate mosaics, masks and the original frames are saved to disk during the processing. If the user chose to not save the intermediate steps, then all but the final mosaic are
removed from disk.
7.2.1.2.2
JScanam
JScanam is the HIPE implementation of the IDL mapmaker called Scanmorphos.

Please look at the following link (http://www2.iap.fr/users/roussel/herschel/) for more
information and a detailed description of this map making tool.
The ipipe script
(scanmap Extended emission JScanam.py) starts from level1 frames and creates a map from
the data. It always combines two (and only two) obsids. If you have more than one pair use
the scanmap Extended emission JScanam multiplePairs.py script which is basically the
same script embedded into a for cycle and can combine any number of scan and cross-scan
obsid pairs. The script is designed in a way that you can run it line by line or in one
go. It is also possible to check all the intermediate products created during the processing.
5makeFinalMap = True/False We remove the unnecessary frames that are taken during the
turnaround of the spacecraft
and mask long the term glitches. A mask called Scanam LongTermGlitchMask is produced.
You should check this mask and if the results are not optimal (some glitches are not detected
or some sources are flagged as glitches), you can try to modify the parameters in order to get
better results.
we save the scan and cross-scan for later use, then we subtract the baseline of every scanleg
in every pixel using a task that subtracts a linear baseline to the signal of every pixel and
every scanleg in the frames. The objective of this task is to remove signal drifts with time
scales larger than a scanleg, preserving real sources as much as possible, with the intention
of creating a first order map that can be used for the source mask calculation. Extended
emission is generally not preserved by this task.
These are the main algorithm steps:
118

Select the unmasked timeline data for each instrument pixel and every scanleg. If a
source mask is provided via the sourceMask optional parameter, the points in this mask
will not be used.
If there is enough points, fit them with a linear baseline. If this is not the case, calculate
their median value.
The linear baseline fit is performed on a iterative way. The unmasked data is fitted
first. Fit positive outliers are then removed, and the fit is performed again with the
clean data. The same steps are repeated until the iteration converges. This happens
when either there is no new outliers found, the number of iterations is larger than 50,
or there is too few data points to continue with the fitting process. In this last case,
the median of the unmasked data is calculated.
Subtract the fit (or the median) to the whole scanleg signal, including the masked values.
The next step is then to create the source mask. But before we do that we need to join the
scan and cross-scan to increase the signal to noise for detecting the sources.
Then we replace the scan and cross-scan with the saved ones. and add the mask to the scans.
After adding the source mask to the data we can start the real processing by removing the
general offset using a simple baseline removal. In this step we calculate the signal median
value in every pixel array and every scan or scanleg , and subtracts it from the frames. Only
the unmasked data is considered in the median offset calculation.
Before further processing we need to identify and mask the signal drifts produced by
the calibration block observation. The PACS photometer signal is sometimes affected after the calibration block observation by a strong signal drift with a typical time length
of 1000 time indices. The task scanamorphosBaselinePreprocessing identifies those
drifts and masks them.
The derived mask is saved in the frames with the name
Scanamorphos CalibrationBlockDrift. This mask will be active after the execution of
the task. If a mask with the same name is already present in the frames, it will be overwritten with the new mask values. After we got rid of the effect of the calibration blocks we can
remove the real baseline from the scans (not only a simple offset as we did before) by running
the basline subtraction again: After the final baseline removal we can remove signal drifts
with time scales longer than a scan-leg. This is called de-striping. It assumes that the scans
and the cross scans have perpendicular scan directions, and as a result the projected maps
have different drift gradients. The drift removal technique is based on the following points:
The drift power increases with the length of the considered time scale (1/f noise). For
that reason the task starts removing the strongest drift component on a time scale equal
to the scans length, decreasing this scale in the next step to 4 scalegs, and finally to 1
scanleg. In each step the remaining drifts are weaker.
Back projecting a map obtained in a given scan direction, to the perpendicular direction,
will transform the generally increasing or decreasing signal drifts into oscillatory drifts
that cancel out on large time scales.
These are the main algorithm steps:
119
Create a map of the cross scans and back project it to the scan direction of the
scans. By default, the brightest sources will be removed using the mask created by
scanamorphosCreateSourceMask (Scanamorphos SourceMask). This can be changed
with the task optional parameter sourceMask.
Subtract the back projected timeline from the scans timeline. This will remove the
remaining sources (common in the two timelines) a will leave the combination of the
two drifts.
Fit a line to the timeline difference. The fit will contain mostly the effect of the scans
drift, because the contribution from the cross scans drift cancels due to its oscillatory
nature.
Subtract the fit (i.e. drift) to the scans timeline.
Repeat the previous steps on the cross scans, using the back projection of the scans. Repeat all steps, decreasing the considered time scale from the scans length, to 4 scanlegs,
and finally 1 scanleg.
Iterate until convergence for timescales equal to one scanleg. Convergence is reached
when the fit of the differences is considerably smaller than the fit residuals. The maximum number of iterations can be controlled with the task optional parameter iterations.
After destriping we can finish working with the scans and cross-scans separately and can join
all the data together and remove signal drifts with time scales shorter than a scan-leg using
the following method
Measure the scanspeed and calculate the size of a mappixel, that can hold 6 subsequent
samples of one crossing of a detector pixel.
Make a mapindex with the specified gridsize. Collect only pixfrac = 0.001 (only pixelcenters)
Count the crossings of all pixels into every map pixel
Calculate the average flux value for each crossing and its standard deviation: bi and si
Now use the threshold noise. This is the blue line in Fig. 1 called HF noise. Check, how
many percent of the stddev of every mappixel are smaller than this threshold noise. -
80%; do not use that mappixel for the drift correction - = 80%: use the mappixel
but dont use the values with stddev HF noise
Calculate the average flux fi in every mappixels. Use the white noise of the detectors
as a weight.
for each crossing calculate: average time ta, drift of the time ti as a difference of the
average flux of the crossing and overall average di = bi - fi.
build the drift timeline from the di at the times ti.
interpolate the missing values of the real timeline (linear interpolation) and subtract
the drift from the real timeline.
120

new map for pointing corrections
calc the rms (stddev) of the drift timeline sdi. If sdi HFnoise, this iteration is converged
if sdi HFnoise, build the mapindex from the new corrected timelines and iterate.
empirical improvement: average the drifts to time-bins of 27, 9, 3 and 1 (in subsequent
iterations) and build the drift timeline by interpolating these much larger bins.
Then we deglitch the merged scans and remove the individual drifts If the deglitch parameter
is set to True here comes a new deglitching with the user defined nSigma.
Finally we can create the final map using photProject using the desired pixel size and pixfrac:
If we process multiple obsid pairs the end of the processing is a little bit different. First we
need to save all of our pairs after the final deglitching using a product sink (we also examine
here if the observation are done in parallel mode): then we merge all the observations by filter
and finally we project the map using all of our obsid pairs:
7.2.1.2.3
Unimap
Unimap is a Generalized Least Squares (GLS) mapmaker developed in MATLAB and released
by the DIET department of the University of Rome La Sapienza. Please, look at the Unimap
Documetation web page for more information, and in particular for a detailed description of
this mapmaker, look at the Unimap Users Manual.
Unimap performs an accurate pre-processing to clean the dataset of systematic effects (offset, drift, glitches), and it uses the GLS algorithm to remove the correlated 1/f noise while
preserving the sky signal over large spatial scales.
The GLS estimator provided by Unimap can introduce distortions at the positions of bright
sky emissions, especially if point like. The distortions, that generally appear as cross-like
artifacts, are due to the approximations of the signal model and to a not perfect compensation
of disturbances at the pre-processing stage.
A high-pass filtering approach is implemented within Unimap to remove these distortions
within a specified spatial scale, by generating the Post-GLS map (PGLS).
The application of the high-pass filter method has the drawback to inject back correlated
noise over the same spatial scale used for the removal of the distortions. To minimize this
noise injection, the filtering can be applied only over the bright emissions that are selected
by using a threshold approach, having at the latest stage the Weighted-GLS map (WGLS).
The values of parameters that contribute for generating the GLS, PGLS and WGLS maps
depend on the characteristics of the sky emission and on the level of accuracy to be achieved
by the different estimators. The best choice for the parameters values is automatically set
by Unimap (from the track 6 and beyond) by performing a statistical analysis of the sky
emission. These parameters can be always fine tuned by the users according their purposes.
The Unimap interactiv script starts from Level1 frames and invokes the runUnimap task for
generating final maps.
Unimap makes use of a large number of parameters that are defined and managed within
the unimap par.txt file. In the interactive script a limited number of parameters (6) are
121
included and propagated. The default values adopted in the script allow the pipeline to
calculate their optimal values, while these parameters can be always fine-tuned by the user.
The other Unimap parameters are adopted as they are into the unimap par.txt and in the
section Advanced use of Unimap is explained how to modify all of the Unimap parameters.
In the following, the parameters that the user can edit within the ipipe script are described:
filterSizeInArcsec: sets the dimension (in arcseconds) of the high-pass filter adopted
by the PGLS method for removing possible distortions introduced by the GLS. If it is
set 0, Unimap calculates the best dimension of the PGLS filter by using an iterative
approach, where the convergence is controlled in the similar way as it is for the GLS
algorithm. If the automatic setting doesnt provide a satisfactory result, a fine-tuning of
this parameter can enhance the quality of PGLS map (see the section Quality inspection
of final maps).
startImage: is the initial guess adopted by the GLS for starting iterations. It can be:
0 for a zero (flat) image, 1 for a nave map, 2 for a mixture map. The mixture map is
composed by portions of a flat map at the positions recognized as background regions,
while for the signal regions, the mixture map is formed by portions of the nave map.
Background and signal regions are identified by adopting a threshold approach over
the computation of the noise level of the observation. If the parameter is equal to 3,
Unimap automatically select the starting image. The number of iterations required by
the GLS to converge depends on the adopted starting image, thus the selection of proper
initial guess can affect the quality of the final maps and it can significantly reduce the
process running time. If the observation is signal dominated, the best initial guess is
the Nave map, while the mixture (or flat) map is suited for observations dominated by
the background. This is the rational adopted by setting startImage = 3.
outputPixelSize: pixel size for the final map in arcsec. If 0, the pixel size is 3.2, except
for non-parallel blue observations, where it is 1.6
wglsDThreshold: threshold for the detection of distortions introduced by the GLS algorithm, for the building of the WGLS mask (mask of artifacts). The greater the value,
the smaller the number of detected artifacts. The optimal value (wglsDThreshold =
0) is computed by varying the value of the threshold and by performing a statistical
analysis on the map of the distortions.
wglsGThreshold: threshold for the building of the WGLS mask (mask of artifacts),
starting from the anchor points provided by the wglsDThreshold parameter. The greater
the value, smaller the extent of the masked artifacts. The optimal value (wglsDThreshold = 0) is computed by varying the value of the threshold and by performing a statistical
analysis on the map of the distortions.
minimumChange: expressed in dB, it is the minimum change in the GLS map to keep
on iterate. If it is set to a positive number, the stop level is automatically selected by
taking into account the morphology of the map. The rational is that a higher precision
is requested for images background dominated and for intermediate cases where bright
and extended emissions are over a flat background (e.g. nearby galaxies, nebulae).
122

maskToUse: the masks that should be considered by Unimap. The Scanamorphos CalibrationBlockDrift mask is added to the default masks because the JScanam
module scanamorphosBaselinePreprocessing is used within the Unimap preprocessing
for removing drift due to calibration blocks.
rewriteFramesFitsFiles: set to True if you want to generate the FITS files used by
Unimap as the input dataset and to save them in the data directory (see the section
Advanced Use of the script).
rewriteParametersFile: set to True if you want to rewrite the parameter file in the data
directory (see the section Advanced Use of the script).
cleanDataDir: set to True if you want to remove input and output files from the data
directory (see the section Advanced Use of the script).
After setting the initial parameters, the first part of the script creates a list with the Level1
frames (frameList) of the observations that you want to reduce. Once the frameList is complete, it can be passed to the runUnimap task, together with the parameter previously defined.
RunUnimap works in two steps: the first converts the Level1 frames into FITS files suitable
for Unimap and writes the Unimap parameter file, the second spawns Unimap from HIPE and
creates the final maps, together with intermediate evaluation products. All files generated by
the task are saved in the directory dataDir/firstObsid lastObsid/camera, and this directory
is removed if cleanDataDir=True.
RunUnimap creates a FITS file for every obsids by adopting the naming convention
unimap obsid camera obsid.fits, and writes the Unimap parameter file unimap par.txt. Then
it launches the Unimap pipeline. While the pipeline is running, the log is displayed into the
HIPE Log and a unimap log.txt file is generated at the end of the process. For a detailed
description of the pipeline, read the Unimap Users Manual, while here the main steps of the
pipeline are summarised:
Figure 7.1: The UNIMAP pipeline
TOP: Time Ordered Pixels. In the first step the sky grid is built and every readout is
assigned to a sky pixel according the specified coordinate system (equatorial or galactic).
In addition, offsets are removed by subtracting the median value from each timeline,
and the coverage map is generated (cove full.fits).
Pre: This module works on the detector timelines. Signal jumps are identified and
flagged, and onsets are corrected by subtracting an exponential fit.
Glitch: this module detects and flags glitches by adopting a sigma-clipping approach
that exploits the spatial redundancy. To simplify the glitch detection, a high pass filter
is first applied over the timelines, to remove long-scale drift and 1/f noise. A glitch
mask (flag glitch.fits) is created by this module.
123
Drift: this module removes the drift affecting the timelines by performing a polynomial
fitting. The fit procedure is based on an iterative approach that uses the Subspace Least
Square (SLS) drift estimation [2]
Noise: this module estimates the noise spectrum and constructs the noise filters to
be used by the GLS algorithm. The GLS implementation used by Unimap exploits
time-domain filtering. With this step the pre-processing is completed and the dataset,
together with the flags, are formatted for the GLS module.
GLS: this module computes the GLS estimator, adopting an iterative approach. GLS is
used for removing 1/f noise and, for an accurate result, all other sources of noise (glitch,
drift, offset) must be removed within the previous tasks. The module is the most timeconsuming step of the whole pipeline and it also generates a nave map (img rebin.fits)
and the associated standard deviation map (img noise.fits). The main parameters used
by the GLS module are the initial guess for starting iteration and the level of precision
that you want to achieve between two iterations. Distortions can be introduced by
the GLS method at the positions of bright sources and the Unimap pipeline includes
post-processing modules to estimate and remove these distortions.
PGLS: this module estimates the GLS distortions, by using a non-linear, median highpass filtering, and subtracts them from the GLS image, to produce a clean Post-GLS
map [1]. The critical parameter is the dimension of the high-pass filter, which dimension should be enough large to include the distortions but not too much large for not
amplifying the background noise of the image. Generally the increase of the noise level
introduced by the PGLS can be negligible for images with high signal-to-noise ratio,
which are also the images for which distortions are generally present. Nevertheless, it is
not always obvious that the PGLS image is better than the GLS one, and that is why
the further WGLS module was introduced. The PGLS module also saves the difference
map between GLS and PGLS (delta gls pgls.fits), that is a useful image to evaluate the
distortion estimate.
WGLS: this module applies the PGLS method only in the map sky regions where the
distortions are relevant (Weighted-GLS). Distortions are detected and masked by using
the wglsDThreshold and wglsGThreshold parameters. In the WGLS map, the PGLS
method is applied only where it is necessary, by minimizing the injection of background
noise generated by the PGLS. The mask of distortion is saved by the WGLS module
for evaluation purposes (flag wgls.fits).
The output of runUnimap task is a single SimpleImage, called unimapMap. It is a context
that contains all the relevant maps generated by Unimap. The products that are present
within unimapMap are listed below. For every product, it is specified the name of the file
generated by Unimap, which is also saved into the working directory if clearDataDir = False:
image is the WGLS map (img wgls.fits). In most cases this is the map you can take;
pgls is the Post-GLS map (img pgls.fits);
gls is the GLS map (img gls.fits); naive is the Nave map (img rebin.fits);
124

stddev is that is the standard deviation map (img noise.fits). It can be properly used as
an error map only if associated with the Nave map. Actually, errors are not propagated
through the pipeline and final error maps can not be provided.
coverage is the coverage map obtained with no flagged data (cove full.fits).
7.2.1.2.4
MADmap
The Microwave Anisotropy Dataset Mapper (MADmap), loosely based on the CMB code
bearing the same name 4 , is - like Unimap - a Generalized Least Square (GLS) map-making
technique that allows a user to generate maximum likelihood sky maps from the ToDs (Timeordered-Data). The point of using MADmap is to correct for signal drifts due to 1/f noise
while preserving emission at all spatial scales. One of the main differences between MADmap
and Unimap is that in MADmap the (uncorrelated) noise properties of the detectors are not
estimated directly from the data, rather taken from a fixed model, passed to MADmap by
means of specific calibration files, typically referred to as InvNtt (Inverse Time-Time Inverse
Noise) files. The MADmap algorithm implemented in HIPE assumes that the noise in the
time streams is purely uncorrelated. In reality the PACS bolometers are affected also by
correlated noise, as well as by pixel-to-pixel electronic offsets. Both the effect of uncorrelated
noise and the pixel-to-pixel offsets have to be corrected for before applying the MADmap
task. The combined removal of correlated noise and pixel-to-pixel offsets from the time lines
takes the name of MADmap pre-processing.
The MADmap ipipe script (L25 scanMapMadMap.py) starts from Level 1 frames and ends
with the creation of an artifact-free 2-d map. The ipipe scripts is organized in four parts: 1)
variables setting; 2) MADmap pre-processing; 3) MADmap run; 4) MADmap post-processing.
Note that the script only works for one or more pairs (scan and cross-scan) of PACS obsids.
It is not designed to generate Level 2 products from a single scan.
At the beginning of part 1), the user sets the values of: obsid numbers; camera (red or blue);
if the observations are of Solar System objects. Then, the user has the choice to tweak the
values of the pre-processing parameters. The meaning and recommended values of these
parameters is shown in Fig. 7.2.
Figure 7.2: Meaning and recommended values for the MADmap pre-processing.
4
http://crd-legacy.lbl.gov/ cmc/MADmap/doc/
125
Finally, the user can also tweak the values of the parameters that enter in the call to the
MADMAP task. See Fig. 7.3 for their meaning and recommended values.
Figure 7.3: Meaning and recommended values for the call to the MADmap task.
Part 2), i.e., the MADmap pre-processing, initially consists in:
the Level 1 refinement: build pointing cube, remove glitches, apply optical distortion
correction
removal of the pixel-to-pixel offsets
the exponential drift correction
The exponential drift correction starts with an initial guess of the average drift, which is
removed from the time streams; then an approximate map is generated, back-projected and
subtracted from the time lines; finally an exponential model is fit to each pixel time line.
In the procedure now described, the user can decide whether to perform the initial guess or
ignore it. In the former case (useMinMedFirstGuess set to True) - the most suitable for bright
sources - the median array, which is created by taking the median of all detector readouts at
each given time, is divided into N bins (N is typically 1000 readouts); then the minimum in
each bin is computed and a polynomial is fit to the resulting distribution of minimum points.
If the time lines contain only weak sources (useMinMedFirstGuess set to False), it is best to
fit directly the median values of the N readout bins with a polynomial ( perPixelExpPolyFit
set to True), and then repeat the steps described above, that is: the fit is removed from the
time streams; an approximate map is generated; this is back-projected and subtracted from
the time lines; an exponential model is fit to each pixel time line. All these processing steps
are applied to each input scan (or obsid). Therefore, before moving further, all the scans
and cross-scans are merged into a single structure. At this point, the time streams are still
affected by some residual drift. In order to fully remove this remaining correlated noise, the
following iterative loop is executed: at the start of the iterations, the current best-estimated
map is generated, back-projected and subtracted from the time lines; then the residual drift
is fit with a baseline (i.e. a first order polynomial) and subtracted from the time lines; a new
map is created. The procedure is repeated for a user-selected number of iterations (by default
set to 5).
With the removal of the global drift (i.e., the correlated noise), the MADmap pre-processing
is complete. However, before moving to part 3), deglitching is performed. The reason why
glitches are removed only at this stage is that it is much easier to perform this operation
once the global correlated drift is corrected for. Finally, part 3), that is the actual MADmap
processing, starts. The call to the MADmap routine (runMadMap) requires that the PACS
126
time lines are re-arranged in a specific format. This is done by the MakeTodArray task,
which will create a binary file, containing the newly generated ToDs, in the /tmp area5 of
the local machine where HIPE is running. The script has two calls to the runMadMap task:
the first call (runNaiveMapper set to True) generates a 2-d naive map, that is a map which is
obtained by a straightforward (or naive) combination of scans and cross-scans and which is
not corrected for uncorrelated 1/f noise. This map allows the user to highlight the presence
of residual correlated drift, hence to track down potential problems with the pre-processing
phase; the second call (runNaiveMapper set to False) generates instead the 2-d GLS map
corrected for 1/f noise, as well as a corresponding coverage map. An indicative error map is
also created (by photMadmapErrors). This map is based on the coverage map, by using the
same algorithm that generates the sensitivity estimates in HSPOT.
Part 4) in the script is the so-called MADmap post-processing. This part takes care of
correcting for artifacts - in the form of crosses in correspondence of bright sources - that are
typically introduced by GLS mappers, such as MADmap and Unimap. Indeed, the algorithm
used in the task (photCorrMadmapArtifacts) - which estimates the artifacts in the GLS map
and subtracts these from the same map, therefore producing a clean GLS map - was designed
within the Unimap framework, and later adopted by MADmap.
7.2.1.3
Comparison of map-makers
During the PACS Post Operation Phase, two map-making challenges were organized, with
the ultimate goal of selecting a map-making code for the generation of the PACS photometer
legacy archival data. For the implementation in the SPG (Standard Product Generation)
pipeline, the selected map-making code - which would add to the High Pass Filter code - had
to satisfy two main requirements: 1) preservation of the emission on all angular scales; 2)
high photometric accuracy.
The first challenge took place in 2012 and culminated in a workshop held at ESAC (Spain)
on January 28 - 31 20136 . A report summarizing the outcome of the challenge was released to
the astronomical community on November 1 20137 . The report contained a preliminary assessment of the performance of six map-making codes: Jscanam, MADmap, SANEPIC, IDL
Scanamorphos, Unimap, Tamasis. Following the document release, the PACS ICC, with the
endorsement of the Herschel Science Team (HST) and Herschel User Group (HUG), decided
to implement Unimap in the SPG pipeline for a target HSCC 13 bulk reprocessing. Given
that the Unimap implementation in SPG encountered some delay, and that other codes (e.g.
MADmap, JScanam) greatly improved between November 2013 and the first months of 2014,
the ICC requested a re-assesment of the status of the various softwares, this time limiting
the challenge to the codes either already present in the pipeline (MADmap, JScanam), or
under consideration for implementation, i.e. Unimap. A second report8 was, therefore, made
publicly available on March 30 2014.
The two challenges were carried out following a common philosophy, namely: first a selection
of representative data sets was done, assuring that the chosen fields contained both faint and
bright sources, and/or flat/highly structured background; then the selected data sets were
5
Or in whatever directory is specified in var.hcss.workdir.

http://www.cosmos.esa.int/web/herschel/pacs-and-spire-map-making-workshop
7
http://herschel.esac.esa.int/twiki/pub/Public/PacsCalibrationWeb/pacs mapmaking report ex sum v3.pdf
8
http://herschel.esac.esa.int/twiki/pub/Public/PacsCalibrationWeb/pacs mapmaking report14 v2.pdf
6
127
processed with the map-making codes participating in the challenge; finally several metrics
were used to evaluate the performance of the codes. Among the metrics, the main ones
were: 1) power spectrum analysis; 2) difference matrix analysis; 3) point source photometry
analysis. Metric 1) consisted in computing a 2-d power spectrum from the final maps, and
allowed the estimation of the amount of flux loss on different angular scales. Metric 2) was
applied by taking the difference between the maps generated by the different codes and a
reference map assumed to be artifact-free. This metric was a powerful tool for revealing potential residual artifacts in the maps. Finally, metric 3) consisting in performing photometric
measurements of both faint and bright sources in the maps generated by the various mappers,
while comparing the measured fluxes to reference values. This metric made it possible to test
the photometric accuracy of each method.
As a result of the first challenge, JScanam, Unimap and MADmap turned out to have the
best performance among the participating map-making codes, while the second challenge
highlighted the slight superiority of JScanam and Unimap compared to MADmap. As a consequence, both methods (JScanam and Unimap) were implemented in the SPG environment
and used for the generation of the legacy archival products, while MADmap was provided to
users, in the form of a HIPE ipipe script, as an alternative mapping method.
7.2.2
Chop-Nod
There is a number of differences in the data structures of chopped and scanned observations
that are also reflected in the various processing steps and the tasks that are applied. In short,
chopped data do not need highpass filtering nor deglitching. The latter is done automatically
when coadding the individual frames of a chopper position by applying a sigma clipping
algorithm. In addition, the final map is created with a shift-and-add procedure that combines
the four images of two chopper and the two nod positions to a single combined signal. A
drizzling algorithm as provided for scan map data is not implemented.
The chop-nod processing pipeline is working on sliced data (slicedFrames) and starts the
processing from level 0. The default slicing rule structures the data according to nod cycles.
This is done in the background, so the user does not even notice the difference (except for the
final step, when combining the individual maps). Each slice appears as an individual index
number in the frames. The tasks internally loop over these items to perform the processing
steps.
Because the pipeline starts from level 0 one needs to perform the basic calibration steps
which are done by the Level 1 SPG script like: saturation and bad pixel flagging, adding
basic pointing information, applying flat-field and responsivity correction.
Then we need to apply some chop-nod specific calibration. First the calibration of the chopping mirror. Here, the digital position sensor readout is converted into chopping angles, so
that we can flag data that were taken during the transition between two chopper plateaus,
i.e. the chopper was neither on the off nor the on position. Then we have to deal with the
chop/nod pattern. During the planning of the observation the observer can specify, if the
chopper should perform a dithering pattern. This displaces the on and off positions by a few
fractions of a pixel for each chopping cycle. The task photMakeDithPos identifies such a pattern and corrects the pointing information accordingly. Then, the global chop/nod pointing
pattern is identified by photMakeRasPosCount as well.
128
There are four readouts per chopper on and off position (plateau). The corresponding flux
values are averaged, so that photAvgPlateau produces a single value for each. The option
skipFirst=True discards the first of each of the four readouts, because it was found that it
is still affected by the chopper movement and contaminates the measurement. Therefore, we
recommend to always use this option. Now, we apply the pointing information to each detector
pixel. The, photAssignRaDec applies the pointing information (astrometry) to each detector
pixel, while photAddPointings4PointSource corrects the coordinates for the displacements
produced by the chop/nod pointing pattern. It relies on the information extracted previously
by photMakeRasPosCount.
Then we need to produce the differential signal of the chopping cycles per dither position and
subtract a constant signal component that is dominated by the telescope mirror emission.
At this stage, we apply a few flux corrections that are based on a thorough investigation of
detector response effects. Note that the order of applying these corrections matters. The
individual corrections are:
Detector non-linearity: For very high source fluxes, the bolometers exhibit a non-linear
relationship between the incident signal and the detector response. Applying the task
photNonLinearityCorrection corrects for this effect.
Evaporator temperature: The temperature of the evaporator of the 3He cooler has a
small but measurable effect on the detector response. Applying the task photTevCorrection provides a correction for this effect at a level of a few percent. For details see:
Balog et al. (2013)
Scan map based flux calibration scheme: The flux calibration of the PACS photometer
is based on scan map measurements, which introduces a time dependent flux scaling
difference for the chopped observations. The correction is done by applying the task
photFluxCalPsToScan. For details see: Nielbock et al. (2013)
Then the three dithering positions are averaged. At the same time, a sigma clipping algorithm
is applied to account for and mask glitches. Note that we do not perform MMT nor 2nd order
deglitching on the chopped/nodded data.
Now that we have the full set of differential signals of the chopper cycles, we also subtract the
nod positions from each other. In this way, small scale differences in the telescope background
due to the varying line of sight caused by the chopper positions cancel out. Finally, we combine
all data to a coherent signal timeline. If the object is bright enough, one should be able to see
the characteristic 2x2 chop/nod pattern on each of the individual frames. In order to improve
the S/N, we combine the flux of the four object images by a shift-and-add algorithm.
This is done per slice, so it produces one map each. In this way, only the centre of the map
contains the valid co-added image of the source. The eight images around the central one are
to be disregarded, as they are just the byproduct of the shift-and-add algorithm, and do not
contain any additional information. In particular, they must not be used to try to improve
the S/N of the target. The central image is already the final result. This image cube is finally
merged to a single map product by applying the mosaic task.
The result is a single Level 2 map that contains both the final image and a coverage map.
Finally, we produce a noise map from the coverage map and add it to the Level 2 product.
7.2.3
129
PACS-SPIRE parallel mode
The processing of the PACS part of the PACS-SPIRE parallel mode observations is identical
to that of the standalone PACS observations. The PACS part of the data can be downloaded
and processed independently from the SPIRE part.
130
Chapter 8
The PACS products

8.1
Introduction
In this chapter we explain the PACS SPG (systematic product generator) products which
were produced by the automatic pipeline run at the HSC, the Standalone Browse products
(SBPs) which are specific SPG products provided for a first look, the Highly-processed Data
Products (HPDPs) which were created by the calibration scientists at the HSC, and the
Legacy products which are a mix of the SBPs and the HPDPs.
PACS products are provided via the HSA in two main forms: an ObservationContext, which
is a wrapper (context) containing the raw data and the different types of products created by
the various levels of pipeline processing; or as individual FITS files wrapped up in a tarball.
An ObservationContext includes raw science data, instrument and satellite housekeeping
data, pointing data, and SPG automatic pipeline-reduced data (cubes, tables, images).
The SBPs are final-level products taken from the ObservationContext, provided as FITS
files in a tarball. They are intended to be easy to read into any software (this being
mainly a problem with the spectroscopy cubes) and also suitable for science work (with
some provisos).
HPDPs are products that have been manually processed at the HSC, and are intended
to replace the equivalent product in an ObservationContext. These products are created
for observations for which the automatic processing cannot produce the most scienceready result, usually because of a problem (technical or otherwise) with the observation.
The Legacy products can be a mix of SBPs and Legacy products. They are provided as
individual FITS files of the cubes, spectral tables, or images that the PACS team have
decided are the most science-ready and user-friendly products for astronomers to work
on, and are the legacy of PACS. Update text when official legacy products exist; will
Legacy be provided as a separate tab in the HSA?
PACS ObservationContexts are explained in detail in the PACS Products Explained PPE:
link, and the pipeline tasks that create all the products are explained in the PACS Data
Reduction Guide PDRG: link. In this chapter the details provided in those documents is
131
132
CHAPTER 8. THE PACS PRODUCTS
summarised. For Standalone and Lecagy products and HPDPs downloaded from the HSA,
specific information about those products is written in a README, and that information is
mostly not repeated here. Should make sure that there READMEs do actually contain all
this information.
8.2
A PACS ObservationContext
For every new user release of HIPE, the entire Herschel archive was re-reduced automatically
with a chosen set of pipelines (per instrument and per observing mode). The resulting SPG
results were then made available through the HSA. The final SPG processing for PACS ended
with HIPE v 14.2. Confirm
All PACS data products are grouped into contexts. A context is a special kind of product
linking other products together in a coherent description and can be thought of as an inventory
or catalogue. The PACS processed observations consist of many such contexts enclosed within
one grand ObservationContext. Each processing level in the PACS pipeline has a context and
the entire observation has a context. There are up to six processing levels for any PACS
observation: 0 (raw data), 0.5 (partially reduced), 1 (instrumental signature removed and
astronomical units added), 2 (final cubes, images, or tables), 2.5 (combined Level 2 data from
two related observations), and 3 (provided for some observations, and are a combination of
observations of that field of view [phot] or of that source [spec]). In addition to the context
containing these levels, an ObservationContext contains contexts for Calibration Products,
contexts for Auxiliary Products (e.g. telescope pointing, mirror temperatures etc.), and a
context for a high-level Quality Control. The entire PACS ObservationContext is shown in
Fig. X: based on SPIRE 6.1.
8.3
Spectroscopy observations: pipeline products
For PACS spectroscopy, a separate pipeline script was run for the following observing modes:
chop-nod line scan, chop-nod range scan, unchopped line-scan and wavelength switching, and
unchopped range-scan (see Sec. 8.3.3.1 for a definition of these modes). The products created
by these four pipelines are exactly the same up to Level 2, and from Level 2 upwards the
differences are minor.
Fig. X, based on SPIRE 6.2 is an an overview of the contents on an ObservationContext, giving a general idea of how it is organised and how its contents relate to the pipeline processing.
The levels of most interest are from 2 upwards and this is explained in this handbook. For
those wishing to re-run the pipeline or wishing to use lower-level products, the PDRG and
the PPE should be consulted.
8.3.1
Viewing an observation in HIPE
Make screenshots from 14.2 (on a big screen!)

An ObservationContext is best viewed in HIPE with the Observation viewer: Fig. ??. At the
top part of this viewer is a summary of the observation and a clickonable postcard (which
8.3. SPECTROSCOPY OBSERVATIONS: PIPELINE PRODUCTS
133
is also available from the HSA directly). In the panel just below the Summary are the Meta
data for the context/product that is currently highlighted in the Data panel. At the bottom
of the viewer is the Data panel which lists the layers of the ObservationContext. Hovering
the mouse over the entries in this directory will bring up a tooltip with information about
the contents.
The layers in an ObservationContext are summarised in Table 8.3.1. The contents of Level 2.
2.5, and 3 are explained in Sec. 8.3.3, and the SBPs, which are taken from these levels and
are also provided in the browseProduct context, are explained in in Sec. 8.5.
Context
PacsObsSummary
History
Auxiliary
BrowseImageProduct
BrowseProduct
Calibration
Level 0
Level 0.5
Level 1
Level 2
Level 2.5
Level 3
Quality
QualitySummary
TrendAnalysis
Description
An observation summary (observer, date, observing mode, wavelength
ranges, and comments)
A history of the SPG processing done to the observation
Satellite and instrument housekeeping, timing, pointing information, etc,
required to run the pipeline
Is the postcard
Contains links to the SBPs, which are otherwise located in Level 2, 2.5,
or 3 (see Sec. 8.5)
Contains the calibation tree that the SPG pipeline used to reduce the
data
The raw science data
The science data partially-processed (engineering conversions, some flagging tasks, data organisation)
More fully processed data, spectrally flat-fielded and organised into the
first of the PACS cube formats
First level with science-grade cubes and spectrum tables, and containing
the final science-grade cubes for most observing modes
Provided only for the unchopped range-scan observations. The off-source
(background) data come from a separate observation to the on-source
data. The two observations are first independently processed, the offsource are subtracted from the on-source, and the full set of high-level
cubes and spectrum tables are created. If the observer did not request an
off-source pointing, the final pipeline products for this mode are found
instead in Level 2
Contains a spectral table for chop-nod pointed SED mode observations
only (see Sec. 8.3.3)
Automatically-added quality information including flags set during the
observing or SPG processing Calibration observations also have this?
Manually-added quality information. Flags raised during the observing
or processing and comments about the quality of the observation can
be found here. A copy of the QualitySummary is also provided directly
from the HSA Calibration observations also have this?
contains something from the calibration block, probably created by
specDiffCs, but not sure if this is at all useful. Ask.
In Fig. ?? (a screenshot from 14.2 SPG) the contents of all of the pipeline levels is shown. The
names HPSXXX (Herschel PACS Spectroscopy XXX) are the names of contexts containing
the products created by the pipeline, mainly cubes and spectral tables. Definitions of these
contexts is given in Table 8.3 and 8.4.
134
8.3.2
Viewing an observation on disk
Downloading an observation from the HSA produces a tarball, which when unpacked has a
layout similar to the layout of the ObservationContext in HIPE: see Fig. ??. All the products
are held as FITS files organised in directories with the names of the layers/contexts they
belong to (these contexts are defined in Tables 8.3 and 8.4). The only product missing on
disk is the PacsObsSummary (Table 8.3.1), but this available from the HSA itself check that
this is true. A very few of the products the History, the quality and qualitySummary are
awkward to read without the handy HIPE viewers.
The data that most astronomers will be interested in can be found in Levels 2 and higher,
being cubes and spectral tables, organised in directories that carry the name of the context to
which they belong. The actual FITS files that contain the cubes or spectral tables are found
the very end of the directory tree: see Fig. ??. The names of these FITS files are built from:
the hpacs + observation id + level (e.g. 20 = 2.0) + the name of the context it comes from
+ a unique identifier. Naming changed in 14.0? find out what the unique bit is, and if the
SBP/level 3 names will be different.
Will need to write a TN explaining how to navigate the actual fits files - including the
important meta data, the layers, and the type of dataset in each extension. Can be provided
as a separate document.
8.3.3
The high-level spectroscopy pipeline products
In this section we explain the cubes and spectrum tables that are the high level pipeline
products on which science can be carried out.
8.3.3.1
The role of the observing mode
The types of cubes and tables provided depend on the observing mode. The observing modes
(a.k.a AOTs) that were offered for PACS spectroscopy were:
Chop-nod and unchopped: refer to the two different ways of gathering data from the
sky background (aka telescope background) during an observation. Chop-nod was the
most common mode, and unchopped was used for crowded fields. See link for more
detail.
Pointed and mapping: observations were either taken as a single pointing or as a raster
used to create a mapping observation.
Range scan and line scan: refer to the wavelength coverage. A range scan could cover
any requested range up to the entire range of the filter used (aka SED mode), and a line
scan would cover one unresolved line and the necessary continuum to study that line.
In Table 8.3.3.1, the Meta data and FITS header keywords (found in the first extension of
the FITS file) that allow the observing mode for any PACS spectroscopy observation to be
determined, are given.
135
Table 8.1: The Meta data/FITS keywords to indicate the observing mode for a PACS spectroscopy observation
Observing mode
chop-nod
unchopped, wavelength switching
line scan
range scan
pointed
mapping
Meta datum
ChopNod
ChopNod
instMode
instMode
obsMode
obsMode
FITS keyword
CHOPNOD
CHOPNOD
INSTMODE
INSTMODE
OBS MODE
OBS MODE
Value
True
False
PacsLineSpec
PacsRangeSpec
Pointed
Mapping
For the mapping modes, it is also necessary to distinguish between the spatial sampling factors. Three spectral sampling modes were offered, each of which had a recommended number
and size of steps between the pointings of the raster sequence. The step sizes and number of
steps that were recommended for the three mapping modes are listed in Table 8.3.3.1.
Nyquist-sampled mapping observations were taken to allow for the eventual mosaic
cubes (the combination of all the cubes of the raster) to have a spatial sampling that
Nyquist-sampled the beam.
Oversampled observations had smaller step sizes (but the same number of steps) and
were taken to allow for even finer sampling of the PACS spatial resolution.
Undersampled mapping observations, aka tiling, were taken to cover a large field-of-view
and were essentially a set of pointed observations with a very small overlap (one spaxel
or less).
In addition, any observation which had either too few or too large steps is officially counted
as being undersampled. Noting that the recommended number of steps in the blue (2x2)
is less than that in the red (3x3), it is possible that e.g. an observation was taken to be
Nyquist sampled in the blue but as a consequence is undersampled in the red: as a result the
blue cubes provided will be that appropriate for Nyquist observations but the red cubes that
appropriate for undersampled observations.
8.3.3.2
The types of cubes produced
The products of interest to the astronomer are found in Level 2, Level 2.5, Level 3, and in
the browseProduct layer...will the SBPs still be in here in 14.2?. The types of cubes placed in
Levels 2 and 2.5 and the spectrum tables placed in Level 2, 2.5, and 3 by the SPG processing
depends on the type of observation. In the PPE and the PDRG the method of determining
which observation got which type of cube/table is explained, together with details of the
tasks that created the products. The interested reader should consult these references. In
this section a summary of that information is provided.
In total there are four types of cubes created by the pipeline:
Rebinned cubes have the footprint of the PACS IFU a 5x5 irregular spatial grid
(see Fig. ??) with spaxels of 900 .4 on a side and a regular, but non-equidistant spectral
grid (the bin sizes scale with the spectral resolution: see link). The wavelength grid,
136
Table 8.2: The Meta data/FITS keywords that determine the type of mapping mode of any
observation. All conditions must be met for a cube to be considered to be either oversampled
or Nyquist, otherwise it is considered to be undersampled/tiling. The blue and red camera
cubes are considered separately when this determination is made.
Meta datum
pointStep
FITS keyword
RASSTEPC
lineStep
RASSTEPL
numRasterLine
NRASTLIN
numRasterCol
NRASTCOL
Value
3 (blue), 2 (red)
< 3 (blue), < 2 (red)
any
3 (blue), 2 (red)
< 3 (blue), < 2 (red)
any
300 (blue), 400 .5 (red )
1600 (blue), 2400 (red )
> 1600 (blue), > 2400 (red)
300 (blue), 400 .5 (red )
1600 (blue), 2400 (red )
> 1600 (blue), > 2400 (red)
Spatial sampling
Nyquist, overampled
undersampled/tiling
undersampled/Tiling
Nyquist, overampled
undersampled/tiling
undersampled/Tiling
oversampled
Nyquist
oversampled/tiling
oversampled
Nyquist
oversampled/tiling
spatial coordinates (RA, Dec), fluxes, and errors are all contained in datasets within
the cube, and extensions within the FITS file.
Projected cubes are mosaic cubes suitable for mapping observations. They are created
with the pipeline task specProject (link to URM?). This mosaic cube is created by
projecting, for each wavelength, the fluxes in the spaxels of the input rebinned cubes
onto a spatial grid that covers the entire field-of-view of the raster. The spatial grid of
these cubes is regular and is defined in the WCS, in units of degrees. The spaxel sizes
are 000 .5 for pointed observations (although use of this cube for pointed observations
is not recommended ??), 1.500 for tiling (undersampled) observations, 300 for range-scan
mapping observations, and for line-scan mapping observations the spaxel size is set by
those in the drizzled cubes of the same observation.
Drizzled cubes are those created by the pipeline task drizzle (link to URM?). This
task also creates mosaic cubes from the individual cubes of a mapping observation, but
uses a more sophisticated algorithm based on the drizzle work of Fruchter & Hook
(reference). The spatial grid is also regular and is defined in the WCS, in units of
degrees. The task optimises the size of the spaxel to be Nyquist at the wavelength of
the observation. These cubes are only provided for line-scan observations, since the
optimal spaxel size is suitable only for short spectral stretches.
Interpolated cubes are provided for the undersampled mapping/tiling and for the
pointed observations, and are created with the task specInterpolate (link to URM?).
For mapping observations these cubes are a mosaic of the raster, and the algorithm used
is a very simple interpolation of the fluxes, at each wavelength, from the input spatial
grid onto an output grid, with spaxels of 300 . For pointed observations, these cubes are
also an interpolation, this time from the input 5x5 irregular grid of 900 .4 spaxels (that
is the IFU footprint of a single pointing) onto an output grid of 300 spaxels. The spatial
grid is therefore regular and is defined in the WCS, in units of degrees.
Equidistant cubes are copies of the other cubes, but they have been spectrally resam-
137
pled to have an equidistant spectral grid (all bins have the same size). These cubes are
provided as SBPs or SBP? and are explained in Sec. 8.5.
See link for a comparison of these cubes to each other.
8.3.3.3
The types of spectrum tables produced
There are three types of spectral tables provided by the pipeline:

For all observations: the rebinned cube spectral tables contain the data of the
rebinned cubes organised in a tabular format. This table is provided in preference to
the rebinned cubes as actual cubes, since these cubes have neither an equidistant spectral
nor spatial grid, i.e. none of the WCS axes can be properly filled. This makes it difficult
to read them into other software. The rows of each table contains the wavelength, flux,
error, spaxel number and coordinate, spaxel RA and Dec, spectral band, and raster
coordinate. The spectra iterate over spaxel number (the entire spectrum of spaxel 0 is
given first, then spaxel 1, etc) and then raster position. One table is provided for each
wavelength range the observer requested. So, for a mapping observation with a 2x2
raster pattern and two requested spectral regions, there will be 4 tables (two red, two
blue) and in each table will be the data of 25 spaxels times 4 pointings.
For all pointed observations: a central and point-source spectrum table containing
the spectrum of the central spaxel (wavelength, flux, error, in Jy/spaxel and m); the
spectrum of the central spaxel point-source corrected (flux, error, in Jy and m); and
the central spaxel-spectrum again point-source corrected but also scaled to the central
3x3 spaxel point-source corrected flux levels (flux, error, in Jy and m). The second
point-source spectrum is recommended for all except the very faintest observations: see
link for more information on these point-source corrections. The scaled spectrum is
not provided for unchopped observation as the scaling does not work for this type of
observation.
For some chop-nod pointed observations with full SED coverage: a single SED spectrum table, that is the combination of all the spectrum tables mentioned above, for
the red and blue camera together and for all observations of that source that were
taken to cover the full PACS spectral range (53120 m and in most cases covered by
two separate observations). This table is provided for the subset of observations which
requested a full SED coverage.
Note that for the point-source corrected spectrum tables, it has not been determined that the
target actually is a point source: that evaluation needs to be made by the user of the data.
8.3.3.4
Cube and spectral table descriptions
Tables 8.3 and 8.4 lists which cubes and tables can be found at which pipeline Level and for
which observing mode. In Table 8.3.3.1, the Meta data and FITS header keywords (found in
the primary extension of the FITS files) that allow one to determine the observing mode for
138
any observation are listed; taken together these tables will allow a user to know what type of
product should be found in any observation.
Tidy the tables up: too busy. Get latex to put them in the text
May put this somewhere else, but it is necessary to put it somewhere. In Table 8.5, information
is given to help the user determine which cube or table can be used depdending on what type
of science they want to do with the PACS observation.
8.4
Photometry observations: pipeline products
For PACS photometry a pipeline was run for all the observation. The most important part
of the photometry pipeline products are Frames and Simpleimages, but they contain several
other datasets that have been produced by the pipelin in a multi-layered organisation. These
variously including: a BlockTable, Meta data, a Status table, Masks etc.
8.4.1
Viewing an observation in HIPE
After downloading your observations in HIPE you can take a look at your fully pipeline
processed data using the ObservationViewer. The outline of a photometry observation is
very similar to that of a spectroscopy observation. The upper two panels show the Summary
and the MetaData. directly below on the left you see the product Tree while on the right
the product is shown which is currently highlited in the left panel. See Table 8.4.1 for the
products descriptions.
8.4.2
Viewing an observation on disk
The observations can be downloaded from the HSA also as a tarball. When unpacked it
produces a directory tree which has a similar layout to that of the ObservationContext in
HIPE. Here the datasets most astronomer intrested in are the level2, 2.5 and 3. The Level
2 maps have file names built as hpacs + the obsid + the level + product name + slice
number (which is always 00 because all photometer observation contain only one slice). E.g.:
hpacs1342223608 20hpppmapb 00.fits for obsid 1342223608 from level 20 (=2.0) with
product name hpppmapb, which in HIPE you see as HPPPMAPB.The Level 2.5 and 3
maps have file names built as hpacs + the level ( 30 25) + product name in capitals
(e.g. HPJSMAPB or HPUNIMAPB) + ra+dec (e.g. 1318 m1819) +00 + version
number (e.g. v1.0).
8.4.3
The high-level pipeline products
The photometry pipeline product from Level 2 are Maps. These maps contain the image in
Jy/pixel, the coverage, some measure of error plus several mapmaker-related auxiliary images.
If your observation contain only point sources (e.g. in mini-map mode) the Level 2 maps are
of science quality, however it is recommended to use the Level 2.5 or Level 3 if you can, since
these maps have a better S/N (being a combination). For extended sources or large maps,
the Level 2 products should not be used for science.
8.5. SPECTROSCOPY STANDALONE BROWSE PRODUCTS

8.4.3.1
139
Science case related pipeline product
As it was mentioned before, the product one should use for analysis depends strongly on the
scientific goal. In the PACS photometer ObservationContext there are three different kind of
products.
8.4.3.1.1
highPass filtering + photProject
In Level 2 and 2.5 of the ObservationContext one can find maps processed using the highPass
filtering method. These maps (called HPPPMAPBR in Level 2 and HPPHPFMAPBR
in Level 2.5) contain several layers not exctly in the same order. The first one however, is
in both cases, the image itself calibrated in Jy/pixel. The other layers are: the coverage, a
measure of the number of readouts that were used to assign flux to a spatial map pixel. In
mappers using photProject, the general level of the coveage values strongly depends on the
parameter pixfrac in the task, the error dataset described in the Chp 4.10 in the PDRG and
also in Popesso et al. (2012), the stddev which is the standard deviation of the projected
pixels and the HPFMask which is the mask used for high-pass filtering. We emphasize once
again that these product are suitable for science only in the case of point sources and if there
is no extended emission is present in the maps.
8.4.3.1.2
JScanam
The maps created with JScanam are present at Level 2.5 and 3. At both levels they are called
HPPJSMAPBR. The layeers are identical to that of the highPass filtered products except
the errors and the HPFMask layers which are missing.
8.4.3.1.3
Unimap
The Unimap maps can also be found at Level 2.5 and 3. however they structure is quite
different from the maps created by the previous two map-makers.
8.5
Spectroscopy Standalone Browse products
PACS spectroscopy SBPs are cubes and spectral tables taken from Levels 2, 2.5, or 3. These
cubes and tables are explained in Sec. 8.3.3: which products you get for which observing
modes, and how to determine the observing mode for an observation (Tables 8.3.3.1 and
8.3.3.1). Exactly which of the cubes and tables produced by the SPG pipeline are then
provided as SBPs also depends on the observing mode, since the most science-ready and
useful products depend also on the observing mode.
8.5.1
The SBP cubes
The cubes provided as SBPs are:
140

For Nyquist and oversampled mapping line scan observations: equidistant drizzled
cubes.
In HIPE these cubes are contained in a context called HPS3DEQDR (Herschel PACS
Spectroscopy 3d [cube] equidistant drizzled red) and HPS3DEQDB (...blue).
For Nyquist and oversampled mapping range scan observations: equidistant projected
cubes.
In HIPE these cubes are contained in a context called HPS3DEQPR (... projected red)
and HPS3DEQPB (...blue).
For all other observations: equidistant interpolated cubes.
In HIPE these cubes are contained in a context called HPS3DEQIR (... interpolated
red) and HPS3DEQIB (...blue).
In Tables 8.3.3.1 and 8.3.3.1 are given the Meta data/FITS keywords necessary to decifer the
specific observing mode of any observation.
For each wavelength range that was requested by the PI of an observation, there will be one
red and one blue cube: so e.g. an observation with two requested line scans will have two
cubes of HPS3DEQDR two of and HPS3DEQDB type present. The names of these FITS
files are built from: the hpacs + observation id + level (e.g. 20 = 2.0) + the name of the
context it comes from + a unique identifier. Naming changed in 14.0? find out what the
unique bit is, and if the SBP/level 3 names will be different.
These equidistant cubes are copies of the Level 2 or 2.5 cubes, but with one difference: they
have an equidistant spectral grid. The default wavelength grid for all PACS spectral products
is not equidistant the bin sizes scale with the spectral resolution (see link for plots). This
was done because PACS spectra cover a large wavelength range and the spectral sampling
was chosen to reflect the spectral resolution at each wavelength. However, non-equidistant
spectral products can be difficult to read into software other than HIPE. For cubes this is
particularly the case, since most software assumes that the wavelength grid can be defined
with the WCS keywords CRPIX3, CDELT3, and CRVAL3, and naturally this is impossible
if the wavelength grid is non-equidistant but wait to see if HCSS-20058 is implemented.
Therefore, the SBP cube for each observation is created by a task that spectrally resamples
the chosen Level 2.5 (unchopped range) or Level 2.0 (all other) cubes on to an equidistant
grid. The equidistant wavelength grid chosen is finer than the original grid, so that the spectra
are practically idential in appearance to the original spectra. The only two things the user
must be aware of when using these cubes are (i) the finer spectral sampling results in spectra
with very many more datapoints and so they are larger on disk, and (ii) the spectral sampling
is no longer a reflection of the spectral resolution, being much finer necessary.
8.5.2
The SBP spectrum tables
Three types of spectrum tables are also provided as SBPs. These are the same tables that
are described in Sec. 8.3.3.3. The rebinned cube tables provided as SBPs are taken from
Level 2.5 for unchopped range observations and Level 2 for the others, as are the central
and point-source spectrum tables for all pointed observations, and the SED point-source
8.6. PHOTOMETRY STANDALONE BROWSE PRODUCTS
141
spectrum tables are taken from Level 3 for all pointed chop-nod full SED spectral coverage
observations.
8.6
Photometry Standalone Browse products
8.7
Spectroscopy Highly-processed data products
To Be Written when the HPDPs exist
8.8
Photometry Highly-processed data products
To Be Written when the HPDPs exist
8.9
Spectroscopy Legacy data products
To Be Written when the they exist
8.10
Photometry Legacuy data products
To Be Written when the they exist
8.11
Calibration products and the calibration tree
8.11.1
Concept
Each PACS ObservationContext downloaded from the HSA contains also the calibration tree,
which is shown in Fig. ??. The calibration tree contains all the calibration products necessary
to reduce any spectroscopy or photometry observation, and is also the same calibration
tree that was used to reduce the data in the ObservationContext. A standalone calibration
tree can also be downloaded from ...how this is done may change, so give details about getting
it via HIPE or independently once they have been settled.
The calibration tree in HIPE is organised into a common, a photometry, and a spectroscopy
section. Inside each are the different calibration products which are used by the pipeline tasks.
On disk these products are held as FITS files. The name of the FITS files are the same as the
names of the calibration products in HIPE: PCalPhotometer TevCorrection FM v1.fits on
disk is called calibration.photometer.tevCorrection in HIPE. The datasets that contain the
actual tev calibration are inside the tevCorrection product. (FM simply refers to flight
model, i.e. the PACS instrument that actually flew on Herschel.)
Obtaining and working with the calibration tree while pipeline processing data in HIPE
is explained in the PDRG. A new version of the tree was produced whenever new calibration products were added or updated, and users who next opened HIPE were offered
142
the chance to update their calibration tree on disk (in the directory .hcss/data/pcalcommunity location change?). A history of the calibration tree versions can be found in
http://herschel.esac.esa.int/twiki/bin/view/Public/PacsCalTreeHistory?template=viewprint
and linked from there are the release notes for each version. These release notes are also
included in the calibration tree directory on disk, as html files. Most of the calibration
product FITS files also have an html info file on disk: Fig. ??.
If working with the calibration tree that comes with a downloaded observation, the organisation of the data on disk is very similar to the organisation in HIPE. Fig. ?? is a screenshot
of the calibration tree from a downloaded observation. Together with the release notes, this
makes it easy to identify which files contain which calibration information. In Tables ??, ??,
and 8.11.1, the products in the calibration trees for common, photometry, and spectroscopy
are described. The calibration products that could be useful to use outside of HIPE are highlighted make sure to do that, and more information about the data in these products can be
found in link I think this guide still need to be written.
8.11. CALIBRATION PRODUCTS AND THE CALIBRATION TREE
143
Table 8.3: The science-grade products provided for PACS pointed observations of different observing modes. See Sec. 8.3.3.2 for more information on the names of the cubes and
Table 8.3.3.1 for the Meta data/FITS keywords that define the observing modes.
Chop-nod and unchopped line scans and wavelength switching
HPS3D[B|R]a
PacsCubes. Spectral units are Jy (per spaxel) and m. The footprint
is irregular and spaxels are 900 .4 on a side. One cube is provided for
each wavelength range observed. These cubes are not to be used for
science, as they are an intermediate pipeline product.
Level 2
HPS3DR[B|R]a Rebinned cubes. Spectral units are Jy (per spaxel) and m. The
footprint is irregular and spaxels are 900 .4 on a side. One cube is
provided for each wavelength range observed
Level 2
HPSTBR[B|R]
Rebinned cubes in table format. One cube is provided for each wavelength range observed
Level 2
HPSTBP[B|R]
Projected cubes. Spectral units are Jy (per spaxel) and m. One
cube is provided for each wavelength range observed
Level 2
HPS3DI[B|R]
Interpolated cubes. Spectral units are Jy (per spaxel) and m. One
Level 2
HPSSPEC[B|R] Table of central and point-source corrected spectra. Note that no
determination is made as to whether the source is actually a pointsource. One table is provided for each wavelength range observed
Chop-nod range scans
Level 2
HPS3D[B|R]a
PacsCubes. See previous
Level 2
HPS3DR[B|R]
Rebinned cubes. See previous
Level 2
HPSTBR[B|R]
Rebinned cube tables. See previous
Level 2
HPS3DI[B|R]
Interpolated cubes. See previous
Level 2
HPSTBP[B|R]
Projected cubes. See previous
Level 2
HPSSPEC[B|R] Spectrum tables. See previous
Level 3
HPSSPEC
Table of SED spectrum taken from the HPSSPECB and HPSSPECR
tables, moreoever for all observations of the source that were taken
to create one complete SED of the source. Both cameras and all
observations are included in the single table
Unchopped range scans
Level 2
HPS3D[B|R]a
Level 2
HPS3DR[B|R]
Rebinned cubes. See previous. These observations often came in
pairs: an off-source and an on-source observation
Level 2
HPSTBR[B|R]
Rebinned cube tables. See previous
Level 2
HPS3DI[B|B]
Level 2
HPSTBP[B|R]
Projected cubes. See previous
Level 2.5 HPS3DR[B|B]
Rebinned cubes. See previous. If the observer requested both an
on-source and an off-source observation, this level contains the data
of the rebinned cubes of the off-source observation subtracted from
that of the on-source observation
Level 2.5 HPSTBR[B|R]
Rebinned cube tables. See previous. This product is created from
Level 2.5 rebinned cubes
Level 2.5 HPS3DI[B|B]
Interpolated cubes. See previous. Created from Level 2.5 rebinned
cubes
Projected cubes. See previous. Created from Level 2.5 rebinned
cubes
Level 2.5 HPSSPEC[B|R] Spectrum tables. See previous. Created from Level 2.5 rebinned
cubes
a: B|R means blue and red camera
Level 2
144
Table 8.4: The science-grade products provided for PACS mapping observations of different observing modes. See Sec. 8.3.3.2 for more information on the names of the cubes and
Table 8.3.3.1 for the Meta data/FITS keywords that define the observing modes.
Undersampled mapping observations
Chop-nod line and range scan, unchopped line scan, wavelength switching
Level 2
HPS3D[B|R]a
PacsCubes. Spectral units are Jy (per spaxel) and m. The footprint
is irregular and spaxels are 900 .4 on a side. One cube is provided for
each wavelength range observed. These cubes are not to be used for
science, as they are an intermediate pipeline product
Level 2
HPS3DR[B|R] Rebinned cubes. Spectral units are Jy (per spaxel) and m. The footprint is irregular and spaxels are 900 .4 on a side. One cube is provided
for each wavelength range observed and for each raster pointing
Level 2
HPSTBR[B|R] Rebinned cubes in table format. See previous.
Level 2
HPS3DI[B|R]
Interpolated cubes. Spectral units are Jy (per spaxel) and m. One
Level 2
HPS3DP[B|B] Projected cubes. Spectral units are Jy (per spaxel) and m. One
Level 2
HPS3D[B|R]a
Level 2
HPS3DR[B|B] Rebinned cubes. See previous. These observations often came in
Level 2
HPSTBR[B|R] Rebinned cube tables. See previous
Level 2
HPS3DP[B|B] Projected cubes. See previous
Level 2
HPS3DI[B|B]
Level 2.5 HPS3DR[B|B] Rebinned cubes. See previous. If the observer requested both an
Level 2.5 HPSTBR[B|R] Rebinned cube tables. See previous. This product is created from
Level 2.5 rebinned cubes
Level 2.5 HPS3DP[B|B] Projected cube. See previous. This product is created from Level 2.5
rebinned cubes
Interpolated cube. See previous. This product is created from Level
2.5 rebinned cubes
Nyquist and oversampled mapping observations
Chop-nod and unchopped line scans and wavelength switching
Level 2
HPS3DR[B|R] Rebinned cubes. See previous
Level 2
Level 2
HPS3DP[B|R] Projected cubes. See previous, however the size of the spaxels is set
by those in the drizzled cubes
Level 2
HPS3DD[B|B] Drizzled cubes. Spectral units are Jy (per spaxel) and m. One cube
is provided for each wavelength range the observer requested
Chop-nod range scans
Level 2
HPS3DR[B|R] Rebinned cubes. See previous
Level 2
Level 2
HPS3DP[B|R] Projected cubes. See previous
Level 2
HPS3DI[B|B]
Interpolated cubes. See previous. Not provided for oversampled mapping observations
Level 2
HPS3DR[B|B] Rebinned cubes. See previous. These observations often came in
Level 2
Level 2
HPS3DP[B|B] Projected cubes. See previous
Level 2
HPS3DI[B|B]
Level 2.5 HPS3DR[B|B] Rebinned cubes. See previous. If the observer requested both an
Level 2.5 HPSTBR[B|R] Rebinned cube tables. See previous. Created from Level 2.5 rebinned
145
Table 8.5: TBW

Observing mode
Use case
Level & Product
Table 8.6: Photometer pipeline products

Context
PacsObsSummary
History
Auxiliary
BrowseImageProduct
BrowseProduct
Calibration
Level 0
Level 0.5
Level 1
Level 2
Level 2.5
Level 3
quality
Description
An observation summary (observer, date, observing mode, wavelength
ranges, and comments)
A history of the SPG processing done to the observation
Satellite and instrument housekeeping, timing, pointing information, etc,
required to run the pipeline
Is the postcard
Contains links to the SBPs, which are otherwise located in Level 2,
2.5, or 3 depending of the highest level of processing (generally leve2.5
processed with JScanam)
Contains the calibation tree that the SPG pipeline used to reduce the
data
The raw science data
The science data partially-processed (engineering conversions, some flagging tasks, data organisation)
Fully calibrated data,
High Pass Filter + PhotProject processed maps of the individual observations. Should not be used for science in case of extended objects
Combined maps of scan and cross-scan (if both exist) processed with
high pass filter+photProject, JScanam, and Unimap
The combination of all observations taken with the PACS photometer
during the Herschel mission of the same object (the same coordinates)
from within the same observing programme.
Automatically-added quality information including flags set during the
observing or SPG processing and quality reports
146
Table 8.7: The contents of the PACS Spectrometer calibration tree version XXX update!.
Each product name corresponds to a context in the calibration tree in HIPE. These products
are also available in the HSA tar file, under a folder with a name (or part of the name)
matching the first column. Products with relevance to particular sections in this Handbook
are provided with links PLEASE CHECK all of the descriptions, especially those with an
opaque meaning. Do we want to include all of them?.
Product
absoluteCapacitance
Description
The measured capacitances for the red and blue detector
arrays
arrayInstrument
Array to Instrument coordinate conversion
badPixelMask
Bad pixels mask
beamSize
Measured FWHM of the beam vs. wavelength
beams[B2AkB3AkB2BkR1]
Beam profiles used in the Pointing Offset Correction
pipeline (for the named bands)
beamsPerSpaxel[B2AkB3AkB2BkR1] Beam profiles for each spaxel (for the named bands)
calSourceFlux
The fluxes (Jy) of the internal calibration sources at the
primary key wavelengths
calSourceFlux3x3
An update of calSourceFlux computed from an improved
calibration scheme
capacitanceRatios
The capacitance ratios for the red and blue detector arrays
chopperThrowDescription
The CPR (chopper position readouts) with verbal descriptions
crosstalkMatrix
Crosstalk matrices for the red and blue detector arrays
darkCurrent
Dark current (V/s) for the red and blue detector arrays
detectorSortMatrix
Detector sorting matrices for the red and blue detector arrays
discardRampHooks
Number of discarded readouts at the hook-shaped start of
the ramps
effectiveCapacitance
Effective measured capacitances of the four commandable
capacitances of the spectrometer
extendedSourceLoss
Fraction of the signal of an extended source seen in a single
spatial pixel
filterBandConversion
The wheel position (WPR) readout-to-band conversion
gprHall
The GPR (DM GRAT CUR POS) versus Hall sensor readback calibration object
gprHallRedundant
The redundant GPR (DM GRAT CUR POS) versus Hall
sensor readback calibration object
gratingJitterThreshold
Value for the jitter threshold of the final grating positions
in readout units
keyWavelengths
The primary and secondary key wavelengths
labelDescription
The bit-coded labels with verbal descriptions
littrowParameters
Littrow parameters for wavelength calibration
littrowPolynomes
Grating wavelength calibration: Littrow equation parameters / polynomial approximation for alpha per pixel
147
Table 8.7: cont

Product
moduleArray
noisyPixelMask
nominalResponse
nonLinearity
observedResponse
observedResponse3x3
offRatio[B2AkB3AkB2BkR1]
pointSourceLoss
psf
rampSatLimits
readouts2Volts
relCalSourceFlux
rsrf[B2AkB3AkB2BkR1]
sensitivity
signalSatLimits
specProperties
telBackCor[B2AkB3AkB2BkR1]
telescopeBackground
timedep
wavePolynomes
wavelengthGrid
Description
Module to Array coordinate conversion
Noisy pixels for the red and blue detector arrays
The nominal responses in V/s/Jy at the prime key wavelengths
Coefficents of a second order polynomial fit to linearise the
signals for the red and blue detector arrays
The responses (V/sJy) for red and blue detector arrays,
computed via observations of astronomical standards
An update of the observedResponse based on an improved
calibration scheme
Signal ratio of the off positions used in the Pointing Offset
Correction pipeline (for the named bands)
Fraction of the signal of a point source seen in a single
spatial pixel, in the central 3x3 and in the 5x5 spatial pixels
Point spread functions for the red and blue spectrometer
Ramp saturation limits (digits) for the red and blue detector arrays
The ramp readout to volt conversion
The flux ratios of the calibration sources at the key wavelengths to those at the prime key wavelengths
The relative spectral response function at the key wavelengths (for the named bands)
The line and continuum RMS noise fluctuations for a 1 sec
integration time
The signal saturation limits (digits/second) for the red and
blue detector arrays scaled to 1 second reset interval
Constants for calculating the spectral resolution vs. wavelength
The wavelength, position and time dependent correction
factors for the telescope background (for the named bands)
Telescope background SED
The time dependency for the calibration products, and
which version is current
Grating wavelength calibration: grating step parameters /
polynomial coefficients
The wavelength grid for the three grating orders and for
different upsample values
148
Chapter 9
Decision tables: how do I know

which products I need for my
science?
A mix of decision trees and decision flowcharts, depending. Will be focussed on the (i) legacy
products, (ii) ObservationContext products, (iii) HPDPs, and then (iv) do I need to reprocess
my data in some way (for the cases where it is really necessary to do something to the data
before doing science with it)?
149
150CHAPTER 9. DECISION TABLES: HOW DO I KNOW WHICH PRODUCTS I NEED FOR MY SCIEN
List of Figures
2.1
2.2
2.3
Main components of the PACS instrument. Left: Cold focal plane unit (T
4 K) assembled at Kayser-Threde, Munich, containing the optical elements
and the detectors (see Section 2.1.2 for details). The location of some elements
is indicated. Middle: The harness connecting the cold focal plane unit inside
the Herschel cryostat on the payload module with the warm electronic unit on
the satellite service module. Right: Warm electronic boxes (T 300 K) with
the on-board CPUs controlling the instrument and receiving engineering and
science telemetry. Individual boxes are explained in Section 2.1.1. . . . . . .
16
Left: PACS focal plane unit (FPU) functional block diagram. The arrows visualize the optical paths through the instrument. Imaging optics are highlighted
by the light brown boxes, filter components by the light beige boxes. Blocks
containing active components (mechanisms, electronics) are outlined in bold.
Right: FPU optical layout. After the common entrance optics with internal
calibrators and the chopper, the field was split into the spectrometer train and
the photometer trains. The two bolometer cameras (top left) had partially
separate re-imaging optics split by a dichroic beam splitter. Filter wheel I contained the filters for the blue bolometer channel. In the spectrometer train, the
integral field unit (IFU) image slicer (middle) converted the square field into
an effective long slit for the Littrow-mounted diffraction grating (top right).
The dispersed light was distributed to the two photoconductor (Ge:Ga) arrays,
the first order to the red array (middle left), 2nd and 3rd order to the blue
array (bottom right). The order sorting filters on filter wheel II were used to
choose between 2nd and 3rd order light. . . . . . . . . . . . . . . . . . . . .
18
PACS telescope focal plane lay-out. The axes give the angular offset in the telescope coordinate system wrt. the telescope optical axis when looking towards
the focal plane (when looking towards the sky, signs are reversed). Photometer
and spectrometer field of view were separated by the fixed field mirror of the
front optics. The smaller spectrometer field of view was offset in the -Z direction by Z = 94.200 . Chopping was along the Y axis direction. On both sides
of the sky area in the focal plane, the fields of the internal calibration sources
were reachable only by the chopper. Typical chopper throws on the sky for
photometry and spectroscopy are given in Table 2.1. . . . . . . . . . . . . .
19
151
152
LIST OF FIGURES
2.4
Design of the PACS internal calibration source (CS): The radiation emitter
at the lower left hand side was a platinum resistor. The four feet carrying
the emitter was a glass-fiber board to reduce the thermal heat loss due to low
thermal conductivity. The radiation was distributed via a scatter plate in front
of the emitter into the integrating sphere to increase the homogeneity at the
pupil. The baffle cone at the right hand side including a lens generated the
illumination patch covering the field-of-view of the detector. . . . . . . . . .
20
Design of the PACS chopper showing its essential elements: The rotation axis
of the gold coated mirror was defined by two flexible pivots on both sides (rotation angle < 10o ). Three drive coils provided redundancy for the rotational
elongation of the mirror. The rotation angle was monitored by field plates of
the position sensor unit and fed back to the chopper control electronics. Three
mechanical interface points allowed to mount the chopper accurately wrt. the
PACS light path. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
Example of a PACS chopper performance test during in-flight commissioning.

The upper line shows the position read-out (in ADU), the lower line the drive
current (in mA). The first picture in each line displays the whole sequence,
the other two the enlargement of the critical swing-in phase with all chopper
plateaux of the sequence overlaid. Due to the diagnostic telemetry frequency of
1 kHz (corresponding to 1 ms temporal resolution), some slight jitter between
the different plateaux can be recognized. In the position read-out, blue symbols represent the actual sensor read-out, while black symbols represent the
stimulus. The red horizontal lines display the corridor to achieve the required
plateau position accuracy. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
PACS photometer filter scheme for the red branch (top; >125 m with dichroic
in transmission) and the blue branch (bottom; <125 m with dichroic in reflection). The blue branch contained a filter wheel which allowed to switch between
the blue bandpass (60 85 m) and the green bandpass (85 125 m). This
allowed the simultaneous observation in one of these bandpasses and the red
bandpass (125 210 m) with two bolometer cameras. . . . . . . . . . . . .
25
Examples of PACS flight model metallic mesh filters provided by Peter Ade
and colleagues from Cardiff University. Top left: Short-wave blocking filter
FL E of the top optics (diameter 36 mm). Top right: Dichroic beam splitter
D P 1 (diameter 56 mm). Bottom left: Low pass edge-defining filter FBBP2 1
for blue bandpass (diameter 20 mm). Bottom right: High pass edge-defining
filter FBBP2 2 for blue bandpass (diameter 20 mm); this filter is very transparent in the visual. Image source: Geis (2005, PACS-ME-ISH-021, Incoming
Inspection of PACS FM Filters). . . . . . . . . . . . . . . . . . . . . . . . .
26
Transmission measurements for the PACS flight model (PACS PFM) filter
components of the blue bandpass (60 85 m). Note, that the x-axis is in
wavenumber wn = 1 ; wn = 100 cm1 corresponds to = 100m. From M
uller
et al. (2011). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
2.10 Effective spectral response (product of filter transmission and bolometer detector absorption) for the three photometer bandpasses. The reference wavelength
0 (cf. Sections ?? and 6.5.3) is indicated for each bandpass. . . . . . . . . .
28
2.5
2.6
2.7
2.8
2.9
LIST OF FIGURES
153
2.11 PACS filter wheel assembly. Left: Design of the filter wheel, showing the disk
with the two filter positions and the magnetic motor drive underneath. Right:
Image of a flight model filter wheel coated with black paint. . . . . . . . . .
29
2.12 Overall 3D view of the PACS cooler identifying its main components. The
bottom figure presents an exploded view identifying the elements represented
schematically in Fig. 2.13. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
2.13 Left: Schematic drawing of the PACS cooler elements and the thermal connections to the PACS bolometer detector focal plane unit (in green) and the liquid
4 He L0 bath of the Herschel cryostat at 1.7 K. Right: Evolution of temperatures relevant for the PACS cooler monitored by sensors and provided in the
PACS instrument Housekeeping (HK) during the cooler recycling process. .
31
2.14 Individual PACS-only cooler cycles (labeled A plus a sequence number): The
figure in the main panel shows the course of the evaporator temperature (T EV)
during the first 10 h following the start of the recycling, whereby the red part
represents the proper recycling process, the blue part the beginning of the
subsequent operational period. The inserts are a zoomed view with adapted
dynamic range of T EV over the full operational period. Left: Cycle A093
on OD 842 with a maximum contiguous bias period of nearly 59.2 h (tbuffer =
1.5 h in hold time calculation). Right: Cycle A133 on OD 1354 with a reduced
(tbuffer = 3.0 h in hold time calculation) contiguous bias period of 57.8 h. For
the latter one the final steep temperature increase is less than 1 mK. . . . . .
34
2.15 Statistics of the cooler hold time versus bolometer biased time over the entire
Herschel mission. Hold time and biased time are defined in the text at the
beginning of Sect. 2.3.3.2. Different symbols and colors represent PACS only
or parallel mode cooler recyclings and the start conditions from a warm, i.e.
exhausted liquid 3 He, or a cold, i.e. still available liquid 3 He, cooler. The
operational guideline established from the relation of all cooler periods up to
OD 270 (Eqn. 2.1) is shown as the red line. . . . . . . . . . . . . . . . . . . .
35
3.1
3.2
Illustration of the photometer scan map scheme with an example of six scan
legs. After finishing the first leg, the telescope turned around and continued
with the next scan line in the opposite direction. The specified reference scan
direction was the direction of the first leg. An optimum orientation of the
PACS blue photometer array wrt. the scan direction is indicated. The rotation
by about 45o in the array reference frame improved the map coverage and
avoided artifacts by the inter-matrix gaps. . . . . . . . . . . . . . . . . . . .
40
Illustration of the detector footprint of the blue detector array on the sky
(left) and the chop-nod source pattern (right) produced on the detector during
the execution of the chop-nod point source AOT. The detector array with
a field of view of 3.50 1.750 consisted of eight individual sub-matrices. The
source was offset by about 5000 horizontally by chopping (no dithering) and
also by about 5000 vertically by the telescope nodding. The colours reflect the
four combinations of the nodding and chopping positions attained during the
observing sequence: black: nodA chop1, red: nodA chop2, blue: nodB chop1,
cyan: nodB chop2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
154
LIST OF FIGURES
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
Logical flow of addressing the PACS photometer calibration items during the
Herschel Performance Verification Phase, taking into account mutual dependencies and prerequisites. Each box is represented by the calibration proposal
name and the list of calibration items addressed during this block. . . . . . .
63
Location of PACS photometer pixels on sky, as measured from OD1308 observations of R Dor (black symbols). Top: blue array, bottom: red array. The
coordinates are defined by the photometer virtual aperture and the y and z
direction of the Herschel spacecraft coordinate system. Bad pixels and few
outliers from the preliminary fit that is shown in red are not plotted. . . . . .
66
PACS photometer PSF at 70 m (left), 100 m (center) and 160 m (right),

as derived from Vesta observations in OD160. The images use a linear stretch
to 100%, 10%, and 1% of the PSF peak from top to bottom. The spacecraft Z
direction is on top, as for observations with telescope PA=0, and the scale bar
indicates 6000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
Faint structures in the PACS 70m PSF. Left: Ceres with a stretch from 0.0005 to 0.001 of the PSF peak, scale bar indicates 6000 . Right: Diffraction
spikes from observations of Mars, scale bar indicates 60000 . . . . . . . . . . . .
69
Effects of scan speed. The PACS photometer PSF at 70m is shown for prime
mode scan speeds of 10 00 /s (top left), 20 00 /s (top right), and 60 00 /s (bottom
left). The additional smoothing because of the 5 Hz averaging in the blue
channel for PACS/SPIRE parallel mode is indicated in the bottom right panel
for speed 60 00 /s. The PSF images are at 70 m, use a linear stretch to the
PSF peak, and the scale bar indicates 6000 . . . . . . . . . . . . . . . . . . . . .
70
Fraction of energy from a point source contained within a circular aperture of

given radius. Left: As applicable for the standard case of prime mode and slow
or medium scan speed. The three PACS bands are shown in blue, green, red
for 70, 100, 160 m. Right: Repeat of the same curves, plus additional ones
showing the smoothing effect of fast scan and/or parallel mode. . . . . . . . .
72
Photometer flat-fields for the three filter bands: blue (top), green (middle) and
red (bottom). Individual detector matrices making up the camera field-of-view
can be recognized. The bars underneath each image indicate the amplitude of
variation over the field-of-view. . . . . . . . . . . . . . . . . . . . . . . . . . .
75
Absolute disk-integrated model flux density predictions for Callisto, Ganymede,

Uranus, Neptune, and Titan in the Herschel-centric reference system: For the
three planet satellites the predictions are for the epoch of the corresponding
PACS measurements. The minimum-maximum model predictions for Uranus
(ura esa2 2 i.dat) and Neptune are referring to all available PACS measurements during the entire Herschel mission. The PACS band-passes are shown
in arbitrary units. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
81
LIST OF FIGURES
6.9
155
Overview with the flux densities of the different far-IR/sub-mm/mm calibrators. The Uranus and Neptune SEDs represent the minimum and maximum
fluxes during Herschel visibility phases. Three fiducial stars are also shown,
their flux range coverage is representative for the brightest stellar calibrators.
For Ceres the maximum flux and for Lutetia the minimum flux during Herschel
observations are shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
6.10 Left: Shape model of Pallas with the TPM temperature coding on the surface,
calculated for the Herschel point-of-view on OD 1295, OBSID 1342256236,
rotation axis is along the vertical direction. Right: the corresponding thermal
light-curve at 100 m with and without thermal effects included. . . . . . . .
87
6.11 Calibration scheme of the non-linearity correction: The measured signal difference above telescope background Sis related to a linearized signal difference
via the correction factor which is dependent on the corresponding true flux
difference F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
6.12 Typical curves for the non-linearity correction factor per filter band as a
function of the signal variation above the telescope background. The signal is
already converted to the flux unit Jy/pixel. . . . . . . . . . . . . . . . . . . .
93
6.13 In-flight verification of the non-linearity correction with flux standards over
nearly the whole accessible flux range of PACS. For the blue (left), green (middle) and red (right) band the measured flux to model flux ratio for the most
reliable flux standards is shown. Triangles represent faint secondary standard
stars, circles the fiducial prime standard stars, boxes prime asteroid calibration standards and reverse triangles planets (Neptune & Uranus) and planet
satellites Callisto & Titan. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
7.1
The UNIMAP pipeline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
122
7.2
Meaning and recommended values for the MADmap pre-processing. . . . . .
124
7.3
Meaning and recommended values for the call to the MADmap task. . . . . .
125
156
LIST OF FIGURES
List of Tables
2.1
2.2
3.1
3.2
3.3
3.4
6.1
Typical angular deflections used for photometer, spectrometer and internal

calibration measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23
Final performance numbers of the PACS chopper for the transition times into
the accurate plateau position for the sky range (<15000 ADU) and for the
internal calibration source range (>15000 ADU). . . . . . . . . . . . . . . . .
23
Bolometer readout saturation levels for the high gain setting. These numbers
are conservative to ensure that saturation was avoided for most of the detector
pixels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
The grating, wavelength range, and sampling details for line spectroscopy
AOTs check, esp limits. The oversampling factor gives the number of times a
given wavelength is seen by multiple pixels in the homogeneously-sampled part
of the spectrum. The column Range of highest sensitivity refers to the range
that is seen by each of the 16 spectral pixels in a module. Bright and faint
refer to bright-line or normal spectral mode. . . . . . . . . . . . . . . . . . . .
48
The grating and spectral sampling details for range spectroscopy AOTs check,
esp limits, and that unchopped and chopnod are the same. The oversampling
factor gives the number of times a given wavelength is seen by multiple pixels
in the homogeneously-sampled part of the spectrum. . . . . . . . . . . . . .
50
The recommendations for raster settings for the various mapping modes. Y
and Z direction are the in the instrumental plane (Y is in the PACS chopping
direction). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
53
FWHM of the PACS PSF for several important cases. 2-dimensional gaussian
fits were used to derive FWHM for the small and large axis. For noticeably
non-round PSF cores, the position angle east of the spacecraft Z direction is
noted. The array to map angle of the scan is also specified. The maps used to
derive the FWHM have been created by photProject with map pixel size 100 and
pixfrac=1. Entries above the line refer to single direction scans, showing the
in-scan elongation for fast scan and for parallel mode. Entries below the line
refer to coadded parallel mode crossed scans, where a cross-like PSF emerges
from coadding the two elongated PSFs. . . . . . . . . . . . . . . . . . . . . .
71
157
158
LIST OF TABLES
6.2
6.3
6.4
6.5
6.6
6.7
6.8
6.8
Effective bandwidth 0 for a constant in the flux per logarithmic frequency

2
interval SED as defined in Eq. 6.6. 0 = c0 0 is the corresponding
bandwidth in wavelength. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
76
Information on the selected fiducial stars. Monochromatic flux densities at

70.0, 100.0 and 160.0 m are given. The stellar temperatures are taken from ?
and the fluxes are based on models published in ?. . . . . . . . . . . . . . . .
80
Model flux predictions for Uranus, Neptune, Callisto, Ganymede, and

Titan.
Models are taken from ftp://ftp.sciops.esa.int/planets/originalData/esan/, where then n in sub-directory esan indicates the version number as reflected in the model source name. In the case of Callisto,
Ganymede and Titan the model fluxes are given for the epoch of observation
(OD). For Uranus and Neptune the flux range encountered during the entire
mission is reflected by maximum and minimum values and the epoch of observation (OD). Appropriate colour-correction factors (cc) to be applied to
measured PACS fluxes for comparison with the model predictions are given,
too. The estimated maximal uncertainty for these corrections is 2% for Titan
and about 1% for the rest of the targets. . . . . . . . . . . . . . . . . . . . . .
82
Model flux predictions for (1) Ceres, (2) Pallas, (4) Vesta, and (21) Lutetia.
The flux range encountered during the entire mission is reflected by maximum
and minimum values and the epoch of observation (OD). Appropriate colourcorrection factors (cc) to be applied to measured PACS fluxes for comparison
with the model predictions are given, too. . . . . . . . . . . . . . . . . . . .
86
Faint secondary standards observed by Herschel-PACS in chop-nod (c) and/or

scan map (s) mode. Source fluxes from Gordon (2007) are for an effective
wavelength of 71.42 m and have been color corrected to the PACS central
wavelength of 70 m, by dividing by the factor 0.961 (cf. Sect. 6.5.3) for a
Rayleigh-Jeans tail type SED. 100 and 160 m fluxes for these sources are then
extrapolated values for this adopted SED. . . . . . . . . . . . . . . . . . . . .
89
Photometric colour corrections for a range of different blackbody temperatures

from 5 K to 10 000 K. Bold values in columns 2 to 4 are the Kcc,BB factors for
the PACS bands. Columns 5 to 11 contain the Klc,BB factors needed to obtain
a monochromatic flux density at neighbouring key-wavelengths used by other
missions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
97
Photometric colour corrections for different modified black-bodies (Fmodif ied =

B (, T )). Bold values in columns 2 to 4 are the Kcc,mBB factors for the
PACS bands. Columns 5 to 11 contain the Klc,mBB factors needed to obtain
a monochromatic flux density at neighbouring key-wavelengths used by other
missions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
98
Photometric colour corrections for different modified black-bodies (Fmodif ied =

B (, T )) continued. Bold values in columns 2 to 4 are the Kcc,mBB factors
for the PACS bands. Columns 5 to 11 contain the Klc,mBB factors needed to
obtain a monochromatic flux density at neighbouring key-wavelengths used by
other missions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
99
LIST OF TABLES
159
Photometric colour corrections for a range of different power-law spectra (F

). Bold values in columns 2 to 4 are the Kcc, factors for the PACS bands.
Columns 5 to 11 contain the Klc, factors needed to obtain a monochromatic
flux density at neighbouring key-wavelengths used by other missions. . . . .
100
The Meta data/FITS keywords to indicate the observing mode for a PACS
spectroscopy observation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
135
The Meta data/FITS keywords that determine the type of mapping mode of
any observation. All conditions must be met for a cube to be considered to
be either oversampled or Nyquist, otherwise it is considered to be undersampled/tiling. The blue and red camera cubes are considered separately when
this determination is made. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
136
The science-grade products provided for PACS pointed observations of different observing modes. See Sec. 8.3.3.2 for more information on the names of
the cubes and Table 8.3.3.1 for the Meta data/FITS keywords that define the
observing modes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
143
The science-grade products provided for PACS mapping observations of different observing modes. See Sec. 8.3.3.2 for more information on the names of
the cubes and Table 8.3.3.1 for the Meta data/FITS keywords that define the
observing modes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
144
8.5
TBW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
145
8.6
Photometer pipeline products . . . . . . . . . . . . . . . . . . . . . . . . . . .
145
8.7
The contents of the PACS Spectrometer calibration tree version XXX update!.
Each product name corresponds to a context in the calibration tree in HIPE.
These products are also available in the HSA tar file, under a folder with a name
(or part of the name) matching the first column. Products with relevance to
particular sections in this Handbook are provided with links PLEASE CHECK
all of the descriptions, especially those with an opaque meaning. Do we want
to include all of them?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
146
cont . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
147
6.9
8.1
8.2
8.3
8.4
8.7

PACS Handbook Mar16

Cargado por

Información del documento

Derechos de autor

Formatos disponibles

Compartir este documento

Compartir o incrustar documentos

Opciones para compartir

¿Le pareció útil este documento?

¿Este contenido es inapropiado?

Copyright:

Formatos disponibles

PACS Handbook Mar16

Cargado por

Copyright:

Formatos disponibles

Herschel Explanatory Supplement

The Photodetector Array Camera and Spectrometer

HERSCHEL-DOC-xxxx, version 1.0, March 6, 2016

Purpose and Structure of Document . . . . . . . . . . . . . . . . . . . . . . .

Key PACS publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Further sources of information

Changes to this document . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 The PACS Instrument

Warm electronics units . . . . . . . . . . . . . . . . . . . . . . . . . . .

Digital Processing Unit (DPU) . . . . . . . . . . . . . . . . .

DEtector and MEchanisms Controler (DECMEC) . . . . . .

BOLometer Controler (BOLC) . . . . . . . . . . . . . . . . .

Signal Processing Unit (SPU)

Cold focal plane unit (FPU) . . . . . . . . . . . . . . . . . . . . . . . .

Effective spectral response . . . . . . . . . . . . . . . . . . .

Procedure of the Cooler Recycling . . . . . . . . . . . . . . .

Establishment of the Cooler Hold Time Relation for Herschel

Mission Statistics of Cooler Recycling . . . . . . . . . . . . .

On-board data handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

On-board data reduction . . . . . . . . . . . . . . . . . . . . . . . . .

On-board data compression . . . . . . . . . . . . . . . . . . . . . . . .

3 Observing with PACS

Point source photometry (mini-scan map) . . . . . . . . . . .

Point-source photometry (chop-nod) . . . . . . . . . . . . . . . . . . .

The background measurement method . . . . . . . . . . . . . . . . . .

The chop-nod AOT . . . . . . . . . . . . . . . . . . . . . . .

The unchopped AOT . . . . . . . . . . . . . . . . . . . . . .

The spectral coverage and sampling . . . . . . . . . . . . . . . . . . .

The line spectroscopy AOT . . . . . . . . . . . . . . . . . . .

The range spectroscopy AOT . . . . . . . . . . . . . . . . . .

Pointed with dither . . . . . . . . . . . . . . . . . . . . . . .

Sensitivity and flux uncertainty . . . . . . . . . . . . . . . . . . . . . .

PSF and beam size . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Anomalies and quality control . . . . . . . . . . . . . . . . . . . . . . .

Spectral range and resolution . . . . . . . . . . . . . . . . . . . . . . .

Wavelength shifts with source position . . . . . . . . . . . . .

Flux calibration accuracy for the various modes . . . . . . .

Flux accuracy for point and semi-extended sources . . . . . .

Flux accuracy for extended sources . . . . . . . . . . . . . .

? The effect of transients ? . . . . . . . . . . . . . . . . . . .

Spatial accuracy and resolution . . . . . . . . . . . . . . . . . . . . . .

Anomalies and quality control . . . . . . . . . . . . . . . . . . . . . . .

5 The Spectrometer Calibration

The spectro-photometric calibration . . . . . . . . . . . . . . . . . . . . . . .

The calibration sources and models . . . . . . . . . . . . . . . . . . . .

Chopped observations: calibrating the telescope background . . . . . .

Unchopped observations: using the internal calibration sources . . . .

Calibration source fluxes . . . . . . . . . . . . . . . . . . . .

Absolute flux calibration . . . . . . . . . . . . . . . . . . . .

Relative spectral response function (RSRF) . . . . . . . . . .

Line flux calibration in order-leak regions . . . . . . . . . . . . . . . .

Characterising the beam . . . . . . . . . . . . . . . . . . . . . . . . . .

Point source calibration function . . . . . . . . . . . . . . . .

Semi-extended source correction . . . . . . . . . . . . . . . .

Extended sources and the spatial flatfield . . . . . . . . . . .

The line profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Calibration the wavelengths . . . . . . . . . . . . . . . . . . . . . . . .

Table of the calibration files . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6 The Photometer Calibration

Calibration achievements during commissioning and the performance verification phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Focal plane mapping and field-of-view distortion . . . . . . . . . . . .

Non-uniform pixel size . . . . . . . . . . . . . . . . . . . . . .

PACS photometer point spread function . . . . . . . . . . . . . . . . .

Effects of source color and reduction schemes . . . . . . . . .