UNIVERSITY OF CANTERBURY

Department of Physics and Astronomy

CHRISTCHURCH NEW ZEALAND

UC ·
U IVER ITYOF
C T RBU Y
Tr ~Vlla"''' l1'LlruHlgu (I Wall(lhCl
IIRISle", ReI! !IollI ZIAI ,\'O

The design and performance of high

resolution echelle spectrographs in
astronomy

A thesis submit ted in

partial fulfilment of the
requirements for the degree of
Doctor of Philosophy in Astronomy

by

Stuart Barnes

University of Canterbury
2004
'HYSICAl.
SCIENCES
LI6RARY

~
z
,D1bl
;004

The design and performance of high

resolution echelle spectrographs in
astronomy

Stuart Barnes

17 JUN 2005
 Abstract

The design and performance of several high resolution spectrographs for use in as-
tronomy will be described. After a basic outline of the required theory, the design and
performance of HERCULES will be presented. HERCULES is an R2 spectrograph fibre-fed
from the MJUO 1-m telescope. The echelle grating has 31.6 grooves/mm and it uses a
BK7 prism with a 50° apex angle in double-pass for cross-dispersion. A folded Schmidt
camera is used for imaging. With a detector having an area 50 x 50 mm, and pixels less
than 25 !-Lm, HERCULES is capable of resolving powers of 40000 to 80000 and wavelength
coverage from 380 to 880 nm. The total throughput (from the fibre entrance to the CCD) is
expected to be nearly 20% (in I" seeing). Measured efficiencies are only slightly less than
this. HERCULES is also shown to be capable of excellent radial velocity precision with
no apparent difference between long-term and short-term stability. Several significant
upgrade options are also described.
As part of the evolution of the design of a high resolution spectrograph for SALT,
several instruments were developed for 10-metre class telescopes. Early designs, based in
part on the successful HERCULES design, did not meet the requirements of a number of
potential users, due in particular to the limited ability to inter-leave object and sky orders.
This resulted in the design of SALT HRS R2 which uses a mosaic of two 308 x 413 mm R2
echelle gratings with 87 grooves/mm. Cross-dispersion is achieved with a pair of large
40° apex angle BK7 prisms used in double-pass. The echelle grating accepts a 365-mm
collimated beam. The camera is a catadioptric system having a 1.2-m primary mirror
and three lenses made of BK7 each around 850 mm in diameter. Complete unvignetted
(except by the CCD obstruction) wavelength coverage from 370nm to 890nm is possible
on a mosaic of three 2k by 4k CCDS with 15!-Lm pixels. A maximum resolving power of
R ~ 80000 is possible. For immunity to atmospheric pressure and temperature changes
the entire spectrograph is designed to be housed inside either a helium atmosphere or a
light vacuum. The spectrograph chamber is nearly seven metres long.
An alternative to the R2 SALT HRS is also described. This instrument is an R4 dual
beam spectrograph based on a white pupil layout. The design is based on suggestions
by B. Delabre and follows closely this authors SOAR HRS instrument. SALT HRS R4 uses
volume-phased holographic gratings for cross-dispersion and a 836 x 204 mm echelle grating
with 41.6 grooves/mm. The grating will be replicated from two smaller gratings onto a
single Zerodur blank. The spectrograph is split into blue and red arms by a dichroic
located near the white pupil relay intermediate focus. Wavelengths from 370 nm to 890 nm
are covered by two fixed format blue and red dedicated dioptric cameras. The detectors
will be a single 2k by 4k CCD with 15!-Lm pixels for the blue camera and a 4k by 4k
CCD with 15!-Lm pixels for the red. The size of the cameras is reduced significantly by
white pupil demagnification from an initial 200-mm diameter collimated beam incident
on the echelle grating to around 100 mm (in undispersed light) on the VPH gratings. The
final SALT HRS R4 instrument is also designed to be immersed in a vacuum vessel which is
considerably smaller than that proposed for the R2 spectrograph. SALT HRS R4 is currently
being developed in detail and will be presented for a critical design review in 2005 April.
Contents

Figures x
Tables xv

Acknowledgments xvii

1 Echelle spectrograph theory 1

1.1 Introduction 1
1.1.1 The development of astronomical spectroscopy 1
1.2 Properties of echelle gratings 4
1.2.1 Grating equation 4
1.2.2 Blazed gratings 4
1.2.3 Angular and linear dispersion 7
1.2.4 Free spectral range 7
1.2.5 Anamorphic magnification 8
1.2.6 Direct and fibre spectrographs 9
1.2.7 Slit width and height 10
1.2.8 Line tilt 11
1.2.9 Cross dispersion 15
1.2.10 Resolving power 17
1.2.11 Efficiency 26
1.2.12 Overfilling 30
1.3 Design of echelle spectrographs 32
1.3.1 Choice of echelle 32
1.3.2 Cross dispersion 34
1.3.3 Collimator, camera and detector properties 36
1.3.4 Fibres 37
1.3.5 Merit functions 38
1.4 Summary 39

2 Design and performance of HERCULES 41

2.1 Design 41
2.1.1 Introduction 41
2.1.2 Optical design 44
2.1.3 Fibre feed 55
2.2 Performance 64
2.2.1 Efficiency predictions 64
2.2.2 Signal to noise predictions 74
2.2.3 Efficiency measurements 76
2.2.4 Environmental stability 80
2.3 HERCULES in the future 84
2.3.1 CCD 84
2.3.2 Collimator and fibre feed 88

vii
viii Contents

2.3.3 Guiding 89
2.3.4 Mechanical stability 89
2.3.5 Efficiency 89
2.4 Summary 90

3 The design of SALT HRS 91

3.1 Introduction 91
3.1.1 SALT 91
3.1.2 Fibre Instrument Feed 96
3.1.3 SAC calibration optics 98
3.1.4 HRS science requirements 98
3.2 SALT HRS fibre feed 101
3.2.1 The fibre modes 101
3.2.2 Fibre slicing options 103
3.3 R2 and R3 designs 108
3.3.1 CELESTIA optical design 108
3.3.2 Alternative designs 116
3.3.3 SALT HRS R2 123
3.4 R4 designs 127
3.4.1 Conceptual design 127
3.4.2 Comparison of efficiencies: R2 vs. R4 132
3.4.3 SALT HRS R4 136
3.5 Summary 138

4 Conclusion 139

A ECHMOD - a Matlab tool for echelle spectrograph modelling 141

A.l Basic outline 141
A.2 Example input/output files 141
A.2.1 HERCULES 142
A.2.2 CELESTIA 144
A.2.3 SALT HRS R2 146
A.2.4 SALT HRS R4 148

B Optical prescriptions 151

B.l HERCULES 152
B.2 CELESTIA 154
B.3 SALT HRS R2 156
B.4 SALT HRS R4 158
B.4.1 SALT HRS R4 - Blue arm 158
B.4.2 SALT HRSR4 - Red arm 160

C HERCULES observing manual 163

C.1 Initializing HERCULES 163
C.1.1 Before observing begins 163
C.1.2 Initializing the HERCULES fibre-feed control 163
C.1.3 Using MoJo 166
C.2 0 bserving 169
Contents ix

C.2.1 CCD position 169

C.2.2 CCD focus 169
C.2.3 Fibre choice 169
C.2.4 Calibration spectra 171
C.2.5 Stellar spectra 171
C.2.6 Guiding 173
C.3 Miscellaneous additional information 174
C.3.1 Computing atmospheric seeing 174
C.3.2 HERCULES log files 175
C.4 Trouble shooting 178
C.4.1 The CCD dark readout is not what was expected 178
C.4.2 CCD is contaminated with thorium or white light 178
C.4.3 Stellar CCD signal is not what expected from exposure meter counts178
C.4.4 Exposure times much longer than expected 179
C.4.5 Hercules screen locks 179
C.4.6 Filter wheel or turn-table out of alignment. 179
C.4.7 The fibre-feed control is acting "strangely"... 179
C.4.8 Auto-guide fails 179
C.4.9 Exposure meter dark count high 180
C.5 Focusing HERCULES 181
C.5.1 Introduction 181
C.5.2 The focuser 181
C.5.3 Collecting the images 181
C.5.4 Running focus_hercules 182
C.5.5 Some useful tips 183

D SALT HRS R2 optical design 185

D.1 Scope 185
D.2 SALT HRS optical design 185
D.2.1 Overview 185
D.2.2 Fibre input 185
D.2.3 Fibre output, collimator and vacuum window 186
D.2.4 Dispersive system 187
D.2.5 Camera 192
D.2.6 Detector 195
D.3 Instrument performance 197
D.3.1 Spectral format 197
D.3.2 Image quality and vignetting 200
D.3.3 Throughput 204
D.3.4 Stray light and ghosts 206
D.4 Exposure meter 208
D.5 Opto-mechanical tolerances 209
D.6 Procurement 209
D.6.1 Optical components 209
D. 6.2 Figuring 209
D.6.3 Coatings 209
x Contents

E SALT HRS R4 optical design 211

E.1 Scope 211
E.2 SALT HRS R4 optical design 212
E.2.1 Overview 212
E.2.2 Fibre injection design 213
E.2.3 Collimator and blue pupil mirror 215
E.2.4 Red pupil mirror 216
E.2.5 Echelle grating 216
E.2.6 Dichroic 217
E.2.7 The cross-dispersers 217
E.2.8 Cameras 218
E.2.9 CCDs 219
E.2.10 Exposure meter 220
E.3 Performance 221
E.3.1 Spectral formats 221
E.3.2 Image quality 223
E.3.3 The white pupil optics 223
E.3.4 Efficiency 228
E.3.5 Signal to noise predictions 234
E.4 Discussion 236
E.4.1 The white pupil optics 236
E.4.2 VPH gratings 236
E.4.3 Cameras 236

Bibliography 243
Figures

1.1 Schematic diagram of a (reflection) diffraction grating 4

1.2 Schematic diagram of a grating where 'Y ::j: 0 5
1.3 Schematic diagram of an echelle grating 6
1.4 Effect of anamorphic magnification on beam size 8
1.5 Schematic digram of a slit-limited spectrograph 9
1.6 Schematic diagram of a fibre-fed spectrograph 9
1. 7 Schematic of a tilted slit image 11
1.8 A fibre image sheared by line tilt 14
1.9 Schematic of a tilted fibre 14
1.10 Schematic of echelle cross-dispersion 15
1.11 Flux weighted fibre width 19
1.12 FWHM of synthetic fibre profiles 21
1.13 Synthetic fibre images and profiles 23
1.14 FWHM of extracted and tilted fibre profiles 24
1.15 Equivalent width of extracted and tilted fibre profiles 24
1.16 Relative FWHM of extracted and tilted fibre profiles 24
1.17 Diffracted intensity of a single wavelength 26
1.18 Effective facet size of a blazed grating 27
1.19 Diffracted intensity of a single wavelength using blazed grating 27
1.20 A method for computing the efficiency of an echelle grating 28
1.21 Relative efficiency of an echelle grating which is blazed at BB = 63° 29
1.22 Blaze function for a range of Littrow angles 29
1.23 Overfilling of an echelle grating 30
1.24 Computing grating overfilling 31
1.25 Absolute and relative efficiency for an R2 echelle grating 32
1.26 Effect of changing the echelle groove ruling density on order separation 33
1.27 A prism used at minimum deviation 34
1.28 Prism vs gratings 36

2.1 Mt John University Observatory Cassegrain echelle spectrograph 42

2.2 A photographic spectrum taken with the MJUO Cassegrain echelle 43
2.3 Optical design of HERCULES 45
2.4 HERCULES inside the vacuum tank 46
2.5 The HERCULES camera 49
2.6 HERCULES spot diagrams 50
2.7 A small region of the HERCULES spectrum. 51
2.8 HERCULES spectral format with a single 2k x 2k CCD 52
2.9 HERCULES spectral format showing the nominal positions of the 1k x 1k CCD 53
2.10 HERCULES spectral format showing the actual positions of the 1k x 1k CCD 54
2.11 Transmission of the HERCULES CeramOptec fibres 55
2.12 The FRD test setup. 57
2.13 Examples of the far-field and near-field images of the fibre output 58
2.14 MeasuredFRD curves for four different fibres 59
2.15 The HERCULES micro-lens 59
2.16 The McLellan 1 metre telescope and fibre feed guide camera 60
2.17 The HERCULES fibre feed guide and acquisition camera 61
2.18 The spectral response of the Thorn EMI 9924 photomultiplier tube. 62
2.19 Exposure meter photocathode current as a function of wavelength for a mv = 0 star 63
2.20 Exposure meter photocathode current as a function of stellar magnitude 63
2.21 Examples of exposure meter log files 64

xi
xii Figures

2.22 Transmission of the seeing disk through the HERCULES fibres 65

2.23 Transmission of the single layer MgF 2 anti-reflection coating on the microlens 65
2.24 Total transmission of the HERCULES fibre 66
2.25 Reflectivity of Laserdyne's uv-enhanced silver mirror coating 66
2.26 Total throughput of the HERCULES fibre feed and collimator 67
2.27 Overfilling of the HERCULES echelle grating 67
2.28 Diffractive efficiency of the HERCULES echelle grating 68
2.29 Efficiency of the HERCULES echelle grating 68
2.30 Transmission of the prism anti-reflection coatings 69
2.31 Total efficiency of the HERcuLEsprism 69
2.32 Total efficiency of the HERCULES dispersive elements 70
2.33 Transmission of the Laserdyne single layer MgF 2 anti-reflection overcoat 70
2.34 HERCULES camera vignetting function across an order 71
2.35 HERCULES camera vignetting function at all wavelengths 71
2.36 Total transmission of the HERCULES camera 72
2.37 Quantum efficiency of the SITe SI003AB CCD 72
2.38 Total efficiency of the HERCULES spectrograph 73
2.39 Atmospheric extinction over MJUO 75
2.40 HERCULES signal-to-noise predictions 76
2.41 Measured efficiency of HERCULES 78
2.42 "Guide corrected" efficiency of HERCULES 78
2.43 Relative efficiency of HERCULES 79
2.44 Pressure of the HERCULES vacuum tank 80
2.45 Temperatures inside HERCULES 81
2.46 Temperatures inside HERCULES during 2002 Feb-Mar 81
2.47 Thorium lines x-shift 82
2.48 Thorium lines y-shift 82
2.49 HERCULES spectral format showing the possible locations of a single 2k x 4k CCD 85
2.50 HERCULES spectral format with a 4k x 4k CCD 86
2.51 Efficiency of HERCULES using a Fairchild CCD with a broadband overcoat 87
2.52 An upgraded collimator for HERCULES 88

3.1 SALT telescope 92

3.2 SALT telescope and detail of the spherical aberration corrector 93
3.3 Variable illumination of the SALT entrance pupil 93
3.4 SALT optical error budget 94
3.5 Fibre entrance aperture efficiencies for SALT 95
3.6 SALT transmission 95
3.7 SALT prime focus payload 96
3.8 Model of the SALT fibre instrument feed 97
3.9 Telecentric angle at the SALT focal plane as a function of field angle 97
3.10 A possible SALT prime focus calibration system 98
3.11 Nod and shuffle concept for fibre-fed spectrographs 102
3.12 Fibre bundle formats in fixed object mode 105
3.13 Throughput of a Bowen-Walraven type image slicer 106
3.14 Slice geometry for fixed object plus sky mode 106
3.15 Image slicer concept for SALT HRS 107
3.16 CELESTIA optical layout 108
3.17 Efficiency of prisms and gratings 110
3.18 CELESTIA spectral format. 112
3.19 CELESTIA camera. 113
3.20 Spot diagrams for CELESTIA 114
3.21 CELESTIA geometric encircled energies 114
3.22 Mechanical design of CELESTIA 115
3.23 Spectral format for R2.8 with 57lines/mm and 47.7° prisms 118
3.24 Spectral format for an R2.8 grating with 57lines/mm and 57.4° prisms 118
3.25 Spectral format for an R2 grating with 110lines/mm and 32.8° prisms 119
Figures xiii

3.26 Spectral format for an R2 grating with 110 lines/mm and 40.0° prisms 119
3.27 Spectral format for an R2 grating with 87lines/mm and 38.8° prisms 121
3.28 Spectral format for an R2 grating with 87lines/mm and 38.8° prisms with tilt able grating 121
3.29 SALT HRS camera design concept 122
3.30 SALT HRS concept camera image quality 122
3.31 Plan and elevation views of SALT HRS 123
3.32 SALT HRS R2 camera 124
3.33 SALTHRS R2 spectral format 125
3.34 Conceptual design for SALT HRS R4 127
3.35 SALT HRS R4 blue camera spectral format 130
3.36 SALT HRS R4 red camera spectral format 131
3.37 Theoretical efficiency of a 900-line/mm blue VPH grating 132
3.38 Theoretical efficiency of a 900-line/mm red VPH grating 133
3.39 Dichroic efficiency 133
3.40 Efficiencies of the R2 and the R4 SALT HRS designs 134
3.41 Relative efficiencies of the R2 and R4 SALT HRS designs 135
3.42 SALT HRS R4 2004 July design 136
3.43 Revised SALT HRS R4 design 137

C.1 HERCULES fibre-feed control graphical user interface. 165

C.2 MoJo control panel. 166
C.3 MoJo image display. 168
C.4 CCD positioning template 169
C.5 Fibre choices. 170
C.6 HERCULES fibre throughputs. 170
C.7 Choosing an isolated spectral line for focus determination. 183
C.8 Determining the best focus position 184

D.1 Plan and elevation views of SALT HRS. 186

D.2 Schematic of the collimator fold prism and focal modification optics. 187
D.3 A schematic of the footprint of the collimated beam on the echelle gratings. 189
D.4 The effect of prism inhomogeneity on image quality. 190
D.5 The SALT HRS camera. 193
D.6 Footprint on the camera primary mirror. 194
D.7 The footprint diagram on the field-flattening lens 196
D.8 The format of the SALT HRS CCDs. 196
D.9 The SALT HRS spectral format. 198
D.10 The position of thorium-argon calibration lines. 199
D.11 Spot sizes of representative wavelengths. 200
D.12 Ensquared energies of representative wavelengths. 201
D.13 The ensquared energy (%) within one pixel at all wavelengths. 202
D.14 The vignetting function of the spectrograph. 203
D.15 The exposure meter. 208

E.1 The ray diagram of SALT HRS R4. 212

E.2 A solid model view of the SALT HRS R4 optics. 213
E.3 The slit fore-optics convert from f /3.8 to f /20. 214
E.4 The focal conversion optics provide the conversion from f /20 to f /10. 215
E.5 Collimator (M 1 ) and blue pupil mirror (M 2 ) dimensions. 215
E.6 Red pupil mirror (M3) dimensions. 216
E.7 The SALT HRS R4 blue arm camera. 218
E.8 The SALT HRS R4 red arm camera. 218
E.9 Blue camera field-flattening lens. 219
E.10 Red camera field-flattening lens. 219
E.11 SALT HRS R4 blue camera spectral format. 221
E.12 SALT HRS R4 red camera spectral format. 222
E.13 Spot diagrams of the slit fore-optics. 223
xiv Figures

E.14 The image quality of the focal conversion optics. 223

E.15 The image quality of the SALT HRS R4 blue arm white pupil relay. 224
E.16 The image quality of the SALT HRS R4 red arm white pupil relay. 224
E.17 Spot diagram for the SALT HRS R4 blue camera. 225
E.18 Spot diagrams for the SALT HRS R4 red camera. 225
E.19 The total image quality of the SALT HRS R4 blue arm. 227
E.20 The total image quality of the SALT HRS R4 red arm. 227
E.21 The reflectivities of various coatings by Laserdyne. 229
E.22 The UV close-up of the Laserdyne mirror reflectivities. 229
E.23 The dichroic efficiency. 230
E.24 Theoretical efficiencies of a 1050line/mm VPH grating from Wasatch Photonics. 230
E.25 Theoretical efficiencies of a 650line/mm VPH grating from Wasatch Photonics. 231
E.26 The measured efficiencies of VPH gratings supplied by three different vendors. 231
E.27 The reflectivity of the multi-layer coatings from Laserdyne. 232
E.28 The reflectivity of the single-large MgF 2 coatings from Laserdyne. 232
E.29 The predicted signal to noise ratio (S / N) of SALT HRS R4 at A = 650 nm. 235
E.30 PEPSI catadioptric white pupil relay 236
E.31 The FEA of the red camera vacuum window. 237
E.32 The deformation of the red camera vacuum window with respect to the original sphere. 238
E.33 Spot diagrams showing the effect on image quality of the deformation of the camera
vacuum window. 238
E.34 The residuals of the red camera vacuum window with respect to the best fit sphere. 239
E.35 The vignetting of the blue camera with reduced apertures. 240
E.36 The vignetting of the red camera with reduced apertures. 241
E.37 The current camera field-flattening lens (a) and an alternative design (b) which would
increase the spacing between the CCD and this lens. 242
Tables

2.1 Refractive index melt data for HERCULES BK7 prism 48

2.2 HERCULES fibres and resolving powers 55
2.3 Description of fibres tested for FRD 56
2.4 Selected spectrophotometric standards 77

3.1 SALT parameters. 94

3.2 High resolution spectrographs on other large telescopes 100
3.3 Optimal configuration of fibres and micro-slits and their throughputs in median seeing 103
3.4 Fibre bundle efficiencies for fixed-object mode 104
3.5 Image slicer parameters for fixed object plus sky mode 104
3.6 Properties of large echelle gratings 109
3.7 Optimal configuration of fibres and micro-slits 110
3.8 CELESTIA prism parameters 111
3.9 Parameters of the CELESTIA camera 113
3.10 Minimum prism apex angles for various echelle gratings 117
3.11 Echelle grating parameters for SALT HRS R4 128
3.12 Fibre and image slicer properties for the SALT HRS R4 132

C.1 Fibre type and resolving power 169

D.1 SALT HRS grating parameters. 187
D.2 Fibre diameters, resolving powers, and entrance aperture transmissions. 189
D.3 Effect of prism homogeneity on image quality. 191
D.4 The physical properties of BK7 and fused silica. 195
D.5 Order numbers and wavelengths for SALT HRS. 197
D.6 Geometrical throughput of the fibre feed and image slicer. 204
D.7 Fibre feed and image slicer throughput. 204
D.8 Fold mirror/focal modifier and collimator throughput. 205
D.9 Prisms and echelle throughput. 205
D.10 Camera throughput. 206
D.11 Total SALT HRS throughput. 207
D.12 SALT HRS and telescope detective quantum efficiency. 207
D.13 SALT HRS and telescope detective quantum efficiency assuming Solgel coatings. 207

E.1 Summary of the SALT HRS R4 fibre modes. 214

E.2 The SALT HRS R4 grating parameters. 216
E.3 Parameters of the VPH gratings for SALT HRS R4. 217
E.4 Detailed efficiciencies of the SALT HRS R4 blue arm in "Fixed Object" mode at the lowest
resolving power. 233
E.5 Detailed efficiciencies of the SALT HRS R4 red arm in "Fixed Object" mode at the lowest
resolving power. 233
E.6 Summary of efficiencies of the SALT HRS R4 blue arm. 234
E.7 Summary of efficiencies of the SALT HRS R4 red arm. 234

xv
Acknowledgments

The design and construction of HERCULES would not have been possible without the
support of many people. I thank my supervisor, John Hearnshaw, for leading this project
in which I have taken great pleasure being involved. I would like to thank all the members
of the HERCULES design and construction team. In particular, the support of Graeme
Kershaw, Nigel Frost, Ross Ritchie, and Geoff Graham from the Department of Physics
and Astronomy, and optical fabricators Gary Nankivell and Dave Cochran has been of
great value. The excellent performance of HERCULES would never have been demonstrated
without the considerable efforts of Jovan Skuljan and David Ramm. Thanks also to Jovan
(and Ljiljana) , and to David for many interesting conversations and for tolerating my
sometimes wild ideas.
The design of SALT HRS has involved a large number of people. The principle investi-
gator Peter Cottrell, and project scientist Michael Albrow have both given considerable
support to my design work. Without their continued enthusiasm, this project would not
have continued to progress Peters encouragement in particular has allowed us all to per-
severe through the sometimes difficult times we have faced. Thanks to Andrew Rakich
and Damien Jones for casting their expert eyes over my optical design work. I would
also like to acknowledge the support of SALT project scientist David Buckley. Along with
the combined SALT science working group, David has ensured that the SALT HRS design
has matured into what will become a very capable instrument. The especially thorough
examination by the SALT HRS external reviewers has been greatly appreciated. Reviewers
have included Richard Bingham, Bernard Delabre, Hans Dekker, Steve Shectman and
David Walker.
I would like to acknowledge being in receipt of the Michael Kidger Memorial Scholar-
ship and the William PriCe Scholarship for Optical Design. I also received a University
of Canterbury Doctoral Scholarship for which I am grateful. I acknowledge support from
the Moore Fund, the Royal Society of New Zealand (Canterbury Branch), SPIE, for funds
to attend overseas conferences. The financial and logistical support of the Department
of Physics and Astronomy has been considerable for both the HERCULES and SALT HRS
projects.
Of course, I thank all of my family, without whom none of this would have been
possible. Special thanks to my good friends Mike and Teina, and to J. Bedford and the
rest of the derelicts on Tilford Street for providing many welcome distractions. Thanks
also to Katja for her love and support during recent months. Finally, thanks also to all
the beautiful freaks I have come to know who make life so interesting.
This research has made use of NASA's Astrophysics Data System.

l>.'vii
Chapter 1

Echelle spectrograph theory

1.1 Introduction
Even at the time the French philosopher August Comte (1835) wrote despairingly of the
hope of how to "study by any means" anything other than the "geometrical or mechanical
phenomena" of stars, Fraunhofer, using both prism and grating spectroscopes, had already
observed absorption lines in the spectra of the sun (see Hearnshaw, 1986 pp. 24-29
and references therein). Later, when Kirchhoff and Bunsen (1860) made the connection
between these lines and the chemical composition of the Sun, it became possible to extend
the reach of spectrographic analysis to what were then the most distant known objects in
the universe: the stars. This ability was described by Sir William Huggins as being "like
the coming upon a spring of water in a dry and thirsty land" (Huggins, 1897). The science
of spectroscopy has since become one of the most fundamental tools used in astronomy.

l.J.1 The development of astronomical spectroscopy

The early experiments by Issac Newton with prismatic dispersion of sunlight in 1666 mark
the beginnings of astronomical spectroscopy. Newton is acknowledged to have made an
observation of the spectrum of Venus although the resolving power of his prismatic instru-
ments was insufficient to recognize anything other than a continuous spectrum. It took
another 135 years before Thomas Young, using the results of interference experiments,
demonstrated that colour and wavelength are the same thing and interest in astronomical
objects did not develop until the 1820s when Joseph Fraunhofer made numerous obser-
vations of the spectra of stars and planets as part of his examination of optical glasses.
Later, laboratory flame spectra were used in an attempt to explain the "Fraunhofer lines",
and progress was made after the observational and theoretical work of Gustav Kirchhoff
and others. This ultimately led to the formulation of Kirchhoff's laws of emission and
absorption. 1

1 Kirchhoff's law states that the ratio between the degree of emission and the degree of absorption for
rays of the same wavelength is constant for all bodies at the same temperature. This can be written as:

E>.(T)
K,>. (T) = constant

where E and K, are the coefficients of emission and absorption at a wavelength)" and temperature T.
(After Kitchin, 1995.) To this law should be added the three corollaries:
1. The wavelengths emitted by a substance depend upon that substance and the temperature
2. The absorption of a substance is a maximum at those wavelengths which it also emits.
3. A luminous solid, liquid, or compressed gas emits a continuous spectrum whereas a rarefied
gas produces a discontinuous spectrum of bright lines.
2 Chapter 1. Echelle spectrograph theory

By the late 19th and early 20th centuries prismatic spectrographs were commonplace
on telescopes as large as 15-30 inches and some objective prisms as large as 4-8 inches
were also being used. In 1862 Huggins built his first (prismatic) stellar spectroscope
and began spectroscopic observations. He gave life to modern astrophysics by making
the fundamental observation that laboratory flames, our Sun, the planets and the stars
share a common chemistry. Huggins was also the first person to attempt to measure the
radial velocity of a star. Over the next three or four decades prism spectrographs became
commonplace at numerous astronomical observatories.

The development of grating spectrographs

It was Lord Rayleigh who showed that the ideal diffraction grating would be better suited
than prisms for achieving high resolution. However, it was extremely difficult to produce
gratings of the required quality. The efficiency of diffraction gratings was also quite low
as the light is dispersed into several orders. Gratings which consisted of many finely
ruled apertures on glass were used by Fraunhofer, and in the 1870's Lewis Rutherfurd
ruled a small number of gratings in speculum metal (see Palmer, 2000, pp.9-10). In
1882 Henry A. Rowland of Johns Hopkins University perfected his "ruling engine" and
was subsequently able to produce gratings which approached the necessary tolerances
(see Palmer and Verrill, 1968) and in 1912, J.A. Anderson, who succeeded Rowland in
the manufacture of gratings at John Hopkins University, was able to produce "blazed"
gratings (see Hearnshaw, 1986, p.ll). It was claimed that these blazed gratings were able
to diffract up to 50% of the light into the first order. Large gratings with near theoretical
resolving power became possible after the development by G. Harrison and G.W. Stoke in
the 1950's of interferometrically controlled ruling engines (see Palmer and Verrill, 1968).
The earliest grating instrument to be used for stellar spectroscopy was likely to have
been used by H. C. Vogel in 1881 and J. Keller from 1890 to 1891 used the Lick refrac-
tor with a spectrograph which had interchangeable prisms and gratings (see Hearnshaw,
1986, p.10). The 1929 Cassegrain spectrograph constructed by P. W. Merrill (1931) was
a significant advancement and incorporated several techniques for the control of flexure.
The development of the coude telescope allowed flexure to be eliminated. Coude spec-
trographs were first used with prisms, but quickly took advantage of first blazed gratings
and later the revolutionary Schmidt camera. In the coude configuration large gratings
(needed for the highest resolving powers) could readily be used without the limitations of
space at the Cassegrain focus and long focal length cameras could be used in conjunction
with large photographic plates in order to achieve high dispersion. These advantages were
first demonstrated on the Mt Wilson spectrographs in the mid 1930s, and were thereafter
copied by many observatories around the world (see Hearnshaw, 1986, pp.14-17).

From blazed to echelle grating spectrographs

The resolving power of a grating is proportional to the product of N, the total number of
grating rulings, and m, the order of diffraction. Given this, the quest for higher resolving
power can be achieved in one of two ways: (i) by increasing the number of rulings on a
grating, or (ii) by increasing the order of diffraction. During the first half of the twentieth
century much work had been done on achieving the former, with the result that larger
more finely ruled gratings were being used; particularly in coude spectrographs. Some
1.1. Introduction 3

progress toward the latter was made by the ability precisely to shape the tools of ruling
instruments so that gratings blazed in the second or third order could be used efficiently.
However, as noted by Michelson, little progress was being made to produce gratings which
efficiently disperse light into orders as high as one hundred (Michelson, 1898). Clearly, if
such a feat were possible, gratings having a fraction of the number of rulings as before
would achieve comparable resolving powers.
Michelson (1898) experimented with producing high order "echelon" gratings which
were comprised of a small number of parallel plates of glass and were used in transmission.
This method was later used by Williams in 1933 (see Harrison, 1949b) to produce a grating
having a resolving power of the order of one million, but the difficulties of producing the
glass plates limited the size of such gratings. R. W. Wood, in 1910, proposed and then
constructed a reflecting echelon grating for use in the infra-red (Wood, 1910). This
grating, which was ruled on metal, he termed the "echelette" and it threw light into only
a few orders. The advantages of the reflecting "echelle ,,2 grating, which, like the echelon
grating, works in high order numbers, but is more coarsely ruled than an echelette, were
described in detail by Harrison in 1949 (op. cit.). However, the echelle grating demands
groove profiles where the reflecting facets are accurate to ,\/10 and where the relative
position of all grooves is maintained to a similar accuracy. By the early 1950s echelle
gratings of up to 126 x 254-mm in area had been constructed which had resolutions close
to the theoretical (see Harrison et al., 1976 and references therein) and by 1970 even better
gratings which were up to 300 x 400-mm in area were possible (op. cit.).
The theoretical properties of the echelle grating will be discussed in the following
section (Section 1.2). In Section 1.3 the requirements for efficient use of echelle gratings
on astronomy will be outlined.

2The term appears to have been coined by Harrison (op. cit) and derives from the French for a "ladder,
scale, or pair of steps"
4 Chapter 1. Echelle spectrograph theory

1.2 Properties of echelle gratings

1.2.1 Grating equation
A schematic of a diffraction grating is shown in Figure 1.1. A reflective surface (having
a normal N) has been ruled with grooves which have spacing a. These grooves cause the
light incident at an angle a to be diffracted through an angle 13. According to Huygen's
principle each groove facet, which has a width as, acts as a source for (plane) diffracted
wavefronts. A given wavelength A will interfere constructively only if the following con-
dition applies:
rnA = a(sin a ± sin 13) (1.1)
This is the classical form of the diffraction grating equation which assumes that the
incident and diffracted rays are all perpendicular to the grooves. It is also possible to
illuminate the grating at angle 'Y with respect to the facet normal (in the x-z plane, see
Figure 1.2) in which case the grating equation becomes

rnA = a(sin a ± sin 13) cos 'Y (1.2)

Figure 1.1: Schematic diagram of a

(reflection) diffraction grating.

1.2.2 Blazed gratings

The grating can be made to diffract a high proportion of the energy into a single diffraction
direction by orientating the grating facets such that a chosen wavelength (in a given
diffraction order) is incident and diffracted at very nearly the same angle. This effect is
termed blazing, and is achieved by orientating the grating facets so that the diffraction
angle is very nearly same as the angle of specular reflection. As shown in Figure 1.3a, the
grating facet angle with respect to the grating normal is called the blaze angle BB. An
echelle grating is simply a standard blazed grating which has a large blaze angle. Such
gratings are often referred to in terms of an "R-number" which is the tangent of the blaze
angle. For instance an R2 grating has a blaze angle BB = 63.4 while an R4 grating has a
0
1.2. Properties of echelle gratings 5

I
IN
I ,
I
I /',
f-------------
//
\ 1"/ J' 1,;/

1)\-- - - - - - - - - /-/I!-"'- ~ - - - -- - - - - ;,"

"/,,
' ____ L _____
\ ,,/
"
1 r J, "
I

Figure 1.2: Schematic diagram of a

I~ ~I ~
grating where,,! =1= o.
cr s

blaze angle eB = 76.0°. From Figure 1.3a it can be seen that the angles of incidence and
dispersion a and fJ are related to the blaze angle ()B of the grating by:

a eB + e and
fJ eB - () (1.3)

where e is the facet illumination angle with respect to the facet normal. That is fJ is the
angle of diffraction for a wavelength AB (the blaze wavelength) in the centre of order m.
Echelle gratings can also be illuminated out of the normal plane (see Figure 1.3b) and it
follows that the blaze wavelength AB is defined in terms of the grating equation (equation
1.2) as

mAB CT(sin a + sin fJ) cos ')I

2CT sin BB cos ecos "I (1.4)

For reasons of efficiency the only viable modes in which an echelle grating can be operated
are where a > fJ or that a ~ fJ (see Schroeder and Hilliard, 1980 and Section 1.2.11).
The situation where e = 0 (i.e., a = fJ) is termed the Littrow condition and if "I =J. 0
the condition becomes quasi-Littrow. Under Littrow illumination, the optical depth of a
grating CTt is given by
(1.5 )
and the facet width is
(1.6)
This determines the order of interference for diffracted light. That is,

2CTt
m=- (1. 7)
A
6 Chapter 1. Echelle spectrograph theory

8
8 z

(a) Profile of echelle grating.

(b) Isometric view of echelle grating.

Figure 1.3: Schematic diagram of an echelle grating. The definitions of the blaze angle eB) angle of .
incidence a and the angle of diffraction f3 are shown in (a). The angle 13 is the angle of diffraction in
the centre of each order m. The facet illumination angle e is defined with respect to the facet normal
O-z. All these angles are defined in the y-z plane. The definition of'Y is shown in (b). It is the angle Of
incidence with respect to the facet normal as measured in the x-z plane.
 1.2. Properties of eehelle gratings 7

1.2.3 Angular and linear dispersion

The angular dispersion of a grating is found by differentiating equation 1.2 with respect
to A for a given a. This gives

dfJ m
(1.8)
dA (J' cos fJ cos r
or
dfJ sin a + sinfJ
- (1.9)
dA A cos fJ

which in the centre of an order, at the blaze wavelength, becomes

dfJ 2 sin BB cos B

(1.10)
dA AB cos fJ

From these equations it can be seen that for a given wavelength high angular dispersion
can be obtained either by making a (and fJ) large or by increasing the grating groove
density (i.e., small 0"). Echelle gratings make use of this fact by having large blaze angles.
Typical echelle gratings have from 30 to 300 grooves mm- I ) and they therefore operate
with large values of m (i.e., m = 10 to > 100).
The angular dispersion is independent of the optical system of which the grating is
part. The linear dispersion determines the extent 6l of a spectral region 6,\ on a given
detector and is given by
(1.11)

where fearn is the focal length of the camera used to image the spectrum. The plate factor
P is the reciprocal linear dispersion and is therefore

dfJ)-I
P = ( fearndA (1.12)

1.2.4 Free spectral range

The free spectral range 6.AFSR is defined as the change in wavelength from an order m
to the next (m ± 1). Any wavelength that appears in an order m will also appear in
. orders m - 1 and m + 1; however the angle of diffraction will be quite different as will the
diffracted intensity. The free spectral range is given by

(1.13)

which in terms of the blaze wavelength, AB, becomes

2(J' sin BB cos Bcos r

A~
(1.14)
2t
8 Chapter 1. Echelle spectrograph theory

The angular extent of one free spectral range is determined by multiplying the free spectral
range (equation 1.13) with the angular dispersion (equation 1.8). That is,

(1.15)
0' cos f3 cos,
which, if e = 0, becomes llf3FSR = AB/(O'scos,). This is simply diffraction from a
rectangular slit of width O's· The diffraction pattern has an angular width AIO's. From
the above equations it can be seen that for a given diffraction angle f3 and order number
m the angular extent of an echelle spectrum depends largely on the density of the echelle
rulings. A coarsely ruled grating (large 0') will produce a spectrum with a smaller angular
extent (per free spectral range) than a more finely ruled grating.

1.2.5 Anamorphic magnification

If a source is of angular distance 6a as viewed from the grating then after dispersion it
will have an angular separation 6f3, where

(1.16)

Now, from equation 1.1, it is straight-forward to show that

r- df31 - cos a (1.17)

- I da - cosf3

The quantity r = cos al cos f3 is called the anamorphic magnification. The effect of
anamorphic magnification on the dispersed light from an echelle grating is illustrated in
Figure 1.4. It can be shown that a beam with a diameter B which is incident on a grating
at angle a will after diffraction through an angle f3 have a diameter B' given by

B'= B (1.18)
r

It should be noted that the anamorphic magnification can vary considerably across a
single free spectra range. This is particularly significant for high R-number gratings,
which generally have a larger angular free spectral range.

I
'N

B Figure 1.4: The effect of

anamorphic magnification on
beam size.
1.2. Properties of eehelle gratings 9

/-Grating/prism

w'

I- ftel feol ·1 feam

Figure 1.5: Schematic diagram of a slit limited spectrograph (after Schroeder, 2000).

I- Grating/prism
w'
B

FRD=p

I- ftel
.1 fcol

Figure 1.6: Schematic diagram of a fibre-fed spectrograph. The cone of light which exits the fibre is
slightly larger than would be expected in the absence of a fibre (dashed line).

1.2.6 Direct and fibre spectrographs

A schematic slit-limited spectrograph is shown in Figure 1.5. A telescope of diameter D
and focal length ftel feeds a spectrograph which has an entrance slit width w. The fibre-
fed spectrograph (Figure 1.6) is identical to the directly fed spectrograph except that a
fibre of diameter d replaces the slit. In both cases the angle subtended by the slit or fibre
on the sky is
w
es - or
ftel
d
es - (1.19)
ftel

If the spectrograph is directly coupled to the telescope (for instance, m coude,

Cassegrain or N asmyth configurations) then the following equality will apply:

ftel feol
(1.20)
D B
where feol is the focal length of the collimator. If instead the spectrograph is coupled to
a telescope via an optical fibre then after the light has passed through the fibre it will
10 Chapter 1. Echelle spectrograph theory

emerge with an output focal ratio Fout which is faster than the input focal ratio Fin, where

R - ftel D fcol
m- D an d rout = B (1.21 )

This effect is known as focal ratio degradation (FRD, see for example Angel, 1977 or
Ramsey, 1988) and can be described in terms of a FRD parameter p:

p=-
Fin (1.22)
F out
Although FRD always has the effect of decreasing the focal ratio, the amount by which
it is decreased depends upon the focal ratio at which the fibre is fed. A typical fibre fed
at an optimal focal ratio will decrease the focal ratio by about 10% to 20% (i.e., p = 1.1
to 1.2). Now, because of FRD, the equality given in equation 1.20 becomes for fibre-fed
instruments
ftel fcol
-=p- (1.23)
D B
That is, in order for the beam size to remain constant on the same spectrograph which
is first directly fed and then later fibre-fed, the focal length of the collimator must be
reduced. In order to preserve throughput, the effective resolving power will thereby be
reduced (see Section 1.2.10) which justifies this effect being termed a degradation. The
use of fibres for spectroscopy will also be briefly discussed in Section 1.3.4.

1.2.7 Slit width and height

As viewed from the grating, the angular size of the slit is 6a = w / fcol or 6a = d/ fcol. It
was shown above that this slit will undergo anamorphic magnification and therefore the
image of this slit will have a width w' given by

fcarn
w' w--r or
fcol
fcarn
d --r (1.24)
fcol

The slit height h' will be

h' 1?'--r
fcarn I
or
fcol
dfcarn
--r I (1.25)
fcol

where r' is the anamorphic magnification introduced by the cross-disperser. This is gen-
erally (but not always) negligible.
1.2. Properties of echelle gratings 11

1.2.8 Line tilt

Although the quasi-Littrow mode of grating illumination offers advantages in terms of
efficiency, a non-zero, has the effect of tilting the slit image with respect to the direction
of dispersion. Due to the finite height of the slit, there is a small change in the angle
of incidence with respect to the facet normal (in the x-z plane) from the bottom of the
slit to the top. As shown in Figure 1.7, if the change in , is 6" then there will be a
corresponding change in the angle of diffraction, 6(3, which will result in a line tilt ¢ given
by
6(3 d(3
tan¢;= - = - (1.26)
0, d,
where it should be noted that 0(3/0, is not necessarily a constant, and hence the tilt angle
¢ will vary across the slit image height. This line curvature will only be noticeable for
very long slit heights. It follows that
d(3 d(3 dA
-- (1.27)
d, dA d,

where d(3 / dA is the echelle angular dispersion (equation 1.8 or 1.9) and from the grating
equation
-dA = - -
(J' ( •
sma+sm • (3) sm,
. (1.28)
d, m
Therefore
sin a + sin (3 sin,
tan¢
cos (3 cos,
d(3
AdA tan, (1.29)

which at the blaze wavelength AB the line tilt becomes

tan ¢; = 2 tan BB tan, (1.30)

Note that from equation 1.30 it can be seen that high R-number gratings are more sus-
ceptible to line tilt. It is also significant to note that if some of the cross-dispersive power
occurs before the echelle grating then the line tilt will have a wavelength dependence.

Figure 1. 7: Schematic of a tilted slit image.

This is due to the finite height of the slit which
slightly changes 'Y.
12 Chapter 1. Echelle spectrograph theory

Fibre tilt
The effect of line tilt on a fibre requires more detailed consideration. As before, a non-zero
r will tilt the dispersed fibre image by an amount ¢ given by equation 1.29. This tilt will
however simply shear the fibre in the direction of echelle dispersion (see Figure 1.8). A
detailed schematic of this sheared fibre is shown in Figure 1.9. The unsheared image of
the fibre is an ellipse (due to anamorphic magnification of the circular fibre) which has a
height h and width w. This ellipse has an equation

(1.31)

The ellipse is then sheared through an angle ¢ giving

x' x + y tan ¢ and

y' y , (1.32)

which if substituted into equation 1.31 gives

(1.33)

Equation 1.33 can be recognized as a quadratic equation of the form

A'x,2 + B'x'y' + G'y,2 + FI = 0 (1.34)

where the coefficients are

A' 1
B' -2tan¢
2 w2
G' tan ¢ + h2 and
w2
F' (1.35)
4

Because the discriminant B,2 - 4A' G' = -4 ~: < 0 this sheared ellipse is also an ellipse.
However, the major axis of this ellipse does not form an angle ¢ to the major axis of the
unsheared ellipse. In fact, it can be shown that the sheared ellipse is equivalent to an
ellipse of the form
A" x"2 + G" y,,2 + F" = 0 , (1.36)
which has been rotated through an angle ¢e given by

A'-O'
cot 2¢e
B'
2 t~n ¢ ( tan 2
¢+ '~: - 1)
1 w2
"2 h2 cot ¢ - cot 2¢ (1.37)
1.2. Properties of echelle gratings 13

The coefficients of the unrotated ellipse are

A" A' cos 2 cPe + B' cos cPe sin cPe + C' sin 2 cPe
C" A' sin2 cPe - B' sin cPe cos cPe + C' cos 2 cPe and
F" F' (1.38)

Now, if equation 1.36 is rewritten in the form

(1.39)

we find that the major and minor axis lengths a and b are given by

2 -F"
a =)liI and (1.40)

The full width We of the sheared ellipse is given by

We
2 = Xe + Ye tan cP , (1.41)

and can be derived by noting that at (x, y) = (xe, Ye)

dy
-dx = -cotcP (1.42)

Given that in polar coordinates

x= w
"2 cose d
an Y ~ 2"h sm
. e (1.43)

it is straightforward to show that

We = wseccPe (1.44)

where
tan cPe = tan ( ~ tan cP ) (1.45)

The relevance of the above derivation will become apparent when the resolving power of
fibre-fed echelle spectrographs is considered below (Section 1.2.10).
14 Chapter 1. Echelle spectrograph theory

Figure 1.8: A fibre image sheared by line

tilt.

I
I
I

Y x+ y tan~ ,/
.... ....,'
I y"

h x

Figure 1.9: Schematic of a

tilted fibre. The fibre im-
age has been sheared by line
tilt through an angle cp. The
sheared image is an ellipse with
axes which have been rotated
w through an angle ¢e. See text
for details.
1.2. Properties of echelle gratings 15

1.2.9 Cross dispersion

Because echelle gratings generally work at relatively high order numbers (i.e, m » 1)
there are many combinations of m and A that satisfy the grating equation. Therefore, an
echelle grating will usually be used in conjunction with a second dispersive element which
will disperse the spectra in a direction that is orthogonal to the main echelle dispersion.
This element could be either a grating or a prism (or a combination of the two; for
instance, a grism). This is shown schematically in Figure 1.10. It would also be possible
to separate the orders by using a filter which is tuned to allow transmittance of only one
free spectral range centred on the wavelength of interest. However, this would negate one
of the most attractive features of an echelle spectrograph. That is, if the order separation
is done by an element with low dispersive power it is possible to arrange many orders
into a 2-dimensional format which can be simultaneously imaged by a single camera. The
choice of cross-dispersers will be discussed further in Section 1.3.2.

Order separation

If the spectrograph camera has a focal length fearn, then the separation between orders
will be
(1.46)

where d,B / dAxD is the angular dispersion of the cross-disperser. If we express the free
spectral range in terms of the blaze wavelength AB then equation 1.46 becomes

d,B A~
6.y = fearn \" . e e
dA XD 20' sm B cos cos 'Y
(1.47)

~-~--- .... --
y

m+l

nl-1-r----------------------------------~
. . ----=- Figure 1.10: Schematic of echelle
cross-dispersion.
x

Order tilt and curvature

As shown in Figure 1.10 the orders will be tilted by an amount 'IjJ. The angle '1/) is given
by
d,B/dAxD
tan'IjJ = d,B/dAEcH (1.48)
16 Chapter 1. Echelle spectrograph theory

where d,B / dAEcH is the angular dispersion of the echelle grating and d,B / dAxD is the
angular cross dispersion. In the order centre the tilt is

tan'l/JB = d,B A cos ~ (1.49)

dA XD 2 sin ()B cos ()
However, because the echelle angular dispersion is not completely uniform throughout an
individual order (i.e, equations 1.8 and 1.9) the orders will be slightly curved.
1.2. Properties of echelle gratings 17

1.2.10 Resolving power

If a spectrograph has marginally sufficient resolution to distinguish between two wave-
lengths Al and A2 = Al + oA then the resolving power is defined as

A
R= oA' (1.50)

where A ~ Al ~ A2. The angular width between the two wavelengths Al and A2 in the
dispersed beam will be 0{3, so in terms of the angular dispersion (d{3 / dA), equation 1. 50
may be written as
A d{3
R = 0{3 dA (1.51)

Now, from equation 1.16 the above becomes

R=~do; (1.52)
oo;dA
The resolving power may now be written in a more useful form by noting that

do; sin 0; + sin {3

(1.53)
dA A coso;
which gives

1 sin 0; + sin {3
R
00; cos 0;
1 2tanBB
(1.54)
00; (1 - tan BB tan B)

The term cos I is ignored here as I is always small and therefore cos I ~ 1.

Diffraction limit
The diffraction limited resolving power can be derived from equation 1.54 by noting that,

mA = t(sino; + sin(3) (1.55)

where N is the number of grooves across a grating which has a length L. If the collimated
beam size is B, then it follows that B = L cos 0; and that the diffraction limited angular
slit size 00; is approximately A/B (or, equivalently Bs ~ A/D). Therefore, in the diffraction
limit,
R=mN . (1.56)

Directly fed spectrographs

In the case of a directly fed spectrograph equation 1.54 becomes

R = icol 2 tan BB
(1.57)
w (1 - tan BB tan B)
18 Chapter 1. Echelle spectrograph theory

Combining equations 1.57, 1.19 and 1.20 gives

R = 2Btane B
(1.58)
esD(l - tan eB tan e)

This provides a very useful way of determining the resolving power of a spectrograph
in terms of the diameter and focal length of the telescope, the slit width (expressed in
terms of the angle the slit subtends on the sky), and the size of the collimated beam
which is incident on the spectrograph's grating. If the collimated beam can be matched
to the projected length of the echelle grating (i.e., B = L cos a) then equation 1.58 can
be rewritten as
R = 2L sin eB cos e
(1.59)
BsD
This equation was first given by Bingham (1979).
What these equations (1.58 and 1.59) show is that in order to obtain a large resolving
power with a given slit size it is necessary either to have a large grating size or a large
collimated beam (i.e., large L or B). This was the solution for the large coude spectro-
graphs used from 1910 to 1980. Equation 1.58 shows the merit of the echelle solution;
that is, to use large eB . However, as shown by equation 1.59, the usefulness of increasing
the blaze angle is not without limits; that is, for R2 gratings, sin eB = 0.89 while for R4
sin BB = 0.97. Also, if the dimensions of the collimated beam are such that B > L cos a
then equation 1.58 is more appropriate. As will be shown (Section 1.2.12) although the
overfilled grating will result in the loss of light, it is still possible to improve the overall
throughput of the spectrograph for a given product of Rand es.

Fibre-fed spectrographs
The effect of FRD has been described in Section 1.2.6. The FRD term p modifies the
resolving power of a fibre-fed spectrograph to
2Btan eB
R and (1.60)
pBsD(l - tan eB tan e)
2L sin BB cos e
R = (1.61 )
pBsD
It is relevant to note that because the FRD of a given fibre depends only on the input
focal ratio, this is the only means by which the focal ratio of the telescope can effect the
resolving power of a fibre-fed spectrograph.

Effective fibre resolving power

The resolving power of a slit limited spectrograph will be given by equations 1.58 and
1.59 only if the seeing disk is considerably larger than the slit width, thereby providing a
uniformly illuminated rectangle. If the seeing disk only partially fills the spectrograph's
entrance slit, or if the entrance slit is entirely absent, then the resolving power equations
must be somewhat modified. This will not be discussed here. However, for a discussion
see Schroeder (2000, pp 318-320).
The entrance slit of a fibre is always circular and essentially uniform in surface bright-
ness. The effective resolving power R' of a circular fibre of diameter d can be calculated
1.2. Properties of echelle gratings 19

Wi-I
r,.L'=====~WPi~====='~
~ b..Yi
r-I
Wi+l \ I

Figure 1.11: The flux weighted fibre

width is calculated by weighting each
d chord by the area it encompasses.

by subdividing the fibre into many narrow slits which have widths Wi equal to the chord
which is parallel to the direction of dispersion. This is shown in Figure 1.11. Each slit will
then have a weighting which equals the fraction of the total flux which the slit encloses.
This fraction is proportional to the area of each slit, where the normalized area Ai of each
slit is
. _ WillYi
A-1 - d2 (1.62)
7f-4

Therefore, the flux weighted fibre width is given by

n

iIJ = LWiAi (1.63)

i=l

In the limit where n -t 00 llYi -t dy and Ai -t wdy equation 1.63 can be solved to give

8
iIJ=-d (1.64)
37f

This factor was first derived by Vaughnn (1994), although he gave an expression for the
flux-weighted slit width of a fibre which has been reimaged onto a slit which is smaller
than the fibre diameter (or alternatively, the slit could be imprinted directly on the fibre
exit face). If the slit width is Ws the flux-weighted slit width becomes

_
W
8 (
= 37f d 1 -
( Ws
1- (d )
2) 3/2) T1 (1.65 )

where the relative transmission T of the clipped fibre is given by

(1.66)
20 Chapter 1. Echelle spectrograph theory

Obviously, if Ws = d, T = I, and w = 8/ 37r d. Hence, the effective resolving power of a

(fully illuminated) fibre is
d
R' R
W
37r R
8
R
'"
'" (1.67)
0.849
While the above shows that a fibre will deliver a resolving power that is considerably
better than the resolving power that can be achieved with a uniformly illuminated slit
with a width equal to the diameter of the fibre, the convention is to measure the full-
width at half-maximum (FWHM) of a monochromatic light source which can be either a
single laser line or the emission lines formed by an appropriate calibration light source.
This method makes the assumption that the profile of a single line, after extraction to
one-dimension (hereinafter called the line-spread function, or LSF) can be approximated
by a gaussian. In fact, as will be shown below, this will not be the case, and the resolving
power measured by this method will be quite different from that derived above.
The extracted profile (or LSF) of a fibre image can be determined by noting that the
extraction in one dimension of an elliptical fibre image produces another ellipse which will
have a normalized height of one and a minor-axis equal to the fibre image width w. That
is, the equation of an extracted fibre will be:

hrb =
n1--
w2
X2

'
-W
-<X<-
2
W
2
In order to approximate the observed fibre profile Jobs the extracted profile is then con-
(1.68)

volved with a one-dimensional point-spread function (PSF); i.e.,

(1.69)

The effect of the PSF will depend on the image quality that the spectrograph produces.
To show how image quality affects the final line profile, gaussians with FWHMS, which
varied in proportion to the fibre image width, were used. The FWHM of the PSF varied
from Wpsf = O.lw to Wpsf = 0.5w. The changing ratio can be used to represent either a
change in image quality or a change in the size of the fibre image.
The FWHM of the fibre profiles can now be determined by fitting a gaussian to the
extracted and convolved fibre profile. It appears reasonable to insist that the fitted
gaussian is normalized to have the same equivalent width as the fibre profile, although
in practice this makes little difference to the parameters of the fitted gaussian (assuming
both width and height are variables). The results are shown in Figure 1.12. The fibre
profiles determined using this method are shown in Figure 1.13. It is noted that the fit
to a gaussian is very poor when the effect of the PSF is small, although as the relative
effect of the PSF increases the approximation by a gaussian becomes more appropriate.
The limit of the FWHM as Wpsf -+ 0 is Wfwhm = 0.682w. Hence, if the spectrograph has
perfect optics, the resolving power would be measured as
R
R(Wpsf = 0) ::::; 0.682 (1.70)
1.2. Properties of echelle gratings 21

- FWHM
0.9 -- EW

$0.85
l.LI

?e 0.8 ---
$ Figure 1.12: The FWHM
lL.
~ 0.75 of synthetic fibre profiles.
.~
Normalized gaussians
Qi
were fitted to the profiles
0::: 0.7
shown in Figure 1.13 in
order to obtain the FWHM.
0.65 Equivalent widths of the
fibre profiles are shown for
~--'-_--'-_---L_---L_--'--_--'--_--'--_--'--_--'--_-'-_---'--' comparison.
o 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Relative PSF width (wpSF)

and to a good approximation, the resolving power as a function of Wpsf is given by

= 0)
"-'
R( Wpsf ) "-' -;========
R(Wpsf

VI + 1.6 W;sf
(1.71)

In practice, the optics of the spectrograph (as well as the properties of the CCD)
will tend to degrade the resolving power. If a degradation in resolving power (which is
measured using the above method) due to optical performance of 10% is acceptable, then
the FWHM of the PSF should be no more than 0.35 -+ 0.40 x w.

Effect of line tilt

As described in Section 1.2.8 the effect of a non-zero "I will be to tilt the dispersed slit
image by an amount ¢ given by equation 1.27. This will have the effect of decreasing the
resolving power. In the case of a slit spectrograph it may be possible to counter-rotate
the slit in order to minimize this tilt, although as pointed out by Schroeder and Hilliard
(Schroeder and Hilliard, 1980) the throughput-resolution product remains constant.
The entrance slit of a fibre-fed spectrograph cannot however be rotated. It will be
stated without proof that the extracted profile of a tilted fibre is simply equation 1.68
where the fibre image width W is replaced by the full-width We of the tilted fibre (equation
1.44). Hence, the extracted profile of a tilted fibre will be given by

I' Jl-
= 4x'2
We '
-We
--<x<-
2
We
2
(1.72)

The observed fibre profile can now be obtained by convolving equation 1.72 with a one-
dimensional PSF. As above, the FWHM of this profile can be measured. This is shown in
Figure 1.14 and the measured equivalent widths are shown in Figure 1.15.
The relative change in FWHM is shown in Figure 1.16. It can be seen that as image
22 Chapter 1. Echelle spectrograph theory

quality becomes worse the relative effect of line tilt decreases. If, for a given image quality
the resolving power at zero line tilt is R'(¢ = 0) (see equation 1.71) then the resolving
power as a function of line tilt is given by

R'(¢) = R'(¢ = O)~ (1.73)

We

In the small angle approximation this reduces to

R'(¢) = R'(¢ = 0) (1.74)

/1 + ~~¢2
where it should be noted that the fibre image height h must also be considered. Hence,
in order for line tilt to degrade the resolving power by less than 10% it can be seen that
a line tilt of up to ¢ ~ 20° can be tolerated.
1.2. Properties of echelle gratings 23

1.5r----_-_-_-~-__,

0.5

Wpsf = O.4W:

Wpsf = O.5w:

Figure 1.13: Synthetic fibre images and profiles. The output from a circular fibre has been convolved
with a gaussian PSF (left). The PSF's have a FWHM ranging from 0.1 to 0.5 times the fibre image width
w. The extracted profile (bold) is then normalized and fitted by a gaussian (dot-dashed). Note that the
extracted fibre profile is well approximated by a gaussian only when the influence of the PSF is large. The
dashed line shows the extracted profile of an unconvolved fibre
24 Chapter 1. Echelle spectrograph theory

0.85,.----,--,---r--,--,---,----,--,-------,-----,----,--, 0.95,----,---,------,-----,---,------,-----,----,-------,-----,---,....,

0.8 0.9

~;:
:;;
§0.75 lo.85
u.
:J
cr
W

0.7 --- - 0° 0.8 - 0°

._.- 5° '-'-'-'-'- .- - 5°
- - 10° - - 10°
.. " .. 15° . 15°
0.65 L-'-_"----'-_--'-_"----'-_-'-_-'----'-_-'====_ 0.75 L-:---:-:-:--::'-:---::-7::---:-,:::-::-:-::-:,,::-::-:::-:-,-:--:'=c=:=:o==:::J
o 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 o 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Relative PSF width (wpSF) Relative PSF width (wpSF)

Figure 1.14: The FWHM of extracted and tilted Figure 1.15: The equivalent width of extracted
fibre profiles. and tilted fibre profiles.

o
1.16 0.25
1 .14 u____0:...:..5~

1.12 ,.. ,..

:::2: ,.. ,..' /
,..
~ 1.1 ,.. ,.. ,..
/

LL
OJ 1.08 ,..
/
> /

~ ,/ /'

ill 1.06
0:
1.04

1.02
Figure 1.16: The rela-
tive FWHM of extracted
and tilted fibre profiles.
o 5 10 15 20 25
Tilt angle <I> (degrees)
1.2. Properties of echelle gratings 25

Total resolving power

The above equations (equations 1.58, 1.59, 1.60 and 1.61) give only the slit or fibre limited
resolving power of a given spectrograph. The total effective resolving power will however
be degraded by several additional factors; for example by imperfect optics and detectors.
If these degrading factors can each be assigned to resolving power influences Ri , then the
total resolving power of the spectrograph R tot will be given by,

(1.75)

where it is assumed that each factor can be modeled by a gaussian function with a FWHM
given by R i . Some of the contributors to the total resolving power include the following:

• the diffraction limited resolving power given by Rdif = mN, where N is the to-
tal number of grooves being illuminated. The diffraction limit will invariably be
somewhat degraded by imperfections in the grating surface; for instance, surface
irregularities and groove ruling errors.

• optical aberrations. The use of imperfect optics is unavoidable.

• detector properties. These include the effects of the depletion layer and charge
migration in silicon detectors. The effects of finite pixel sampling must also be
considered.

The optical quality and detector properties must therefore be chosen in order to ensure
that the degradation is acceptable. While all of the above influences are unavoidable,
there may be other transient effects such as focus· errors or image motion which will
further degrade the image quality of the spectrograph. The design of the spectrograph
should attempt to mitigate all such effects.
26 Chapter 1. Echelle spectrograph theory

1.2.11 Efficiency

Upon striking a grating at an angle a a collimated beam will be diffracted through an

angle (3, which depends on the wavelength A and the grating groove spacing 0" according
to the grating equation (equation 1.2.1). The intensity I of this diffracted beam results
from the combination of an interference function (I F) and a blaze function (B F); that
is,
I=IFxBF (1. 76)
The interference function that results from a grating which has a total of N equally spaced
grooves is given by
IF = (sinNv')2 (1. 77)
Nsinv'
where 21/ is the phase difference between rays diffracted off the centres of adjacent grooves.
The blaze function is given by
BF = (SiI:V)' (178)

where v is the phase difference between the centre and edge of an individual groove. These
phase differences are given by

. (3)
2v' T
21W ( .
sma+sm and (1.79)

v 7rO"s ( . . (3) (1.80)

T\sma+sm

Each individual grating facet has a width o"s which may be smaller than the groove spacing
0" in which case the blaze function will be broadened.
The diffracted intensity pattern for a single wavelength is shown in Figure 1.17. It

/1"', Figure 1.17: The

/ \ diffracted intensity of a
/ \
/ \ single wavelength (solid
0.8 I \ line) . The blaze function
/ \
c / \ (dashed line) modulates the
'(jj
c: / \ interference function, which
.2) 0.6 / \ is maximum when the order
.~
OJ
/ \ number m is an integer.
> / \
~ / \ The intensity in diffracted
(jj 0.4 orders (m i= 0) is low.
0: / \
I \
/ \
I \
0.2 /

//(
o~~-_~~_-~~·"~'~,~~~~·__~·J~____~·I~l~~~~~_~~I-_-_"_'
\,\ __~
-3 -2 -1 0 1 2 3
Order number (absolute)

can be seen that the majority of the energy incident on the grating is returned in the
zeroth order (m = 0) where it is simply reflected. Only a small portion of the energy is
diffracted into other orders.
1.2. Properties of echelle gratings 27

Blaze function
The purpose of blazing a grating is to shift the blaze function so that the maximum
diffracted intensity of a given wavelength coincides with the chosen diffraction order. The
phase difference between successive grooves (equation 1.80) is now given by
I

V = 1f~s (sin( a - eB ) + sin(,8 - eB )) (1.81 )

where (J~ is the effective size of each facet (see Figure 1.18). The effective facet size (when
a > ,8 is

(J~ (J( cos eB - sin eB tan e)

(J cos a
(1.82)
cos e
which should be compared to the size of a clear facet (Js, which is (Js = (Jcose B .
I
I

I
I
I
I
I
I
I
I
I
I
I
I
I
I
I

e
Figure 1.18: The effective facet size
of a blazed grating is reduced be-
cause of shadowing (after Schroeder,
2000).

The normalized intensity of a wavelength diffracted by a grating blazed at eB = 63.5°

is shown in Figure 1.19. The wavelength has been chosen so that it coincides with the
maximum of the blaze function which occurs in order m = 40.

/ " Figure 1.19: The diffracted

/ \ intensity of a single wave-
0.9 I \
I \
I \
length using blazed grating.
0.8 I \ The blaze function is now
~0.7 I
/ \
\
centred on an order m f O.
I \ Compare with Figure 1.17.
5i 0.6 I \
.5 I '.
~ 0.5 I \
~ I '.
~ 0.4 I
I '.
\
I '.
0.3 I \

0.2
/ \
\
\
/
0.1
/
/
I
/
\~\ .,"""" .... --.... Jk...,
~7 38 39 40 41 42 43
Order number (absolute)
28 Chapter 1. Echelle spectrograph theory

Absolute efficiency
In order to calculate the efficiency of an echelle grating it is necessary to determine the
distribution of light of a given wavelength across all possible orders. A wavelength that
is not at the centre of the blaze function will have a significant fraction of its energy
diffracted into other orders. This is shown in Figure 1.20. The method prescribed by

1-' Figure 1.20: A method for

\
\ computing the efficiency of an
\ echelle grating. See text for
/
0.8 I \ details.
\
\
I \
~ I \
'iii I
\
c I
\
2 0 .6 I
I
\
.f: I \
I
~
\
I \
~ I \
Q5 0.4 I \
a: /
\
\

I \
I \
0.2 I \
/ \
/ \
-------,---- /(
O~-w~--~--~~~----~----~~~-L~-'~~
\ """-[," -- ..
-3 -2 -1 0 1 2 3
Order number (relative)

Schroeder and Hilliard (1980) is simply to sum the intensities across all possible orders
and then derive the fraction that remains in the order of interest. However, as commented
by Bottema (1981), this definition of efficiency is not quite correct, although it is conceded
that in most cases of interest the results will be correct (Schroeder, 1981). Therefore, the
absolute diffractive efficiency of an echelle grating for a wavelength in order m is

_ It::.m=o
T.ech - (1.83)
It::.m=o + '6t::.m#O
"
It::.m

where Clm and I are the relative order number and intensity respectively.
The relative efficiency of an echelle grating which is blazed at OB = 63 is shown in 0

Figure 1.21. This grating is illuminated at 0 = 0 which means that the wavelength free 0
,

spectral range is equal to the FWHM of the blaze function. If the grating is illuminated
in a non-Littrow mode (0 =I- 0) then the fraction of the blaze function that is covered by
one free spectral range is increased by a factor cos fJ I cos a; that is

;\ \ _ Cl.AFSR
U/\FWHM - (1.84)
r
where r is the anamorphic magnification. The blaze function for a range of Littrow angles
(0), such that 1.0 < 1/r < 1.5 is shown in Figure 1.22. Note that the values of r refer to
the order centre only.
1.2. Properties of echelle gratings 29

m=41 m=40 m =39 Figure 1.21: Relative effi-

, .......... , /"'" - ...
/ \ /
/ \
\
ciency of an echelle grating
0.9
I
/ \
\
\
\ which is blazed at (}B = 63°.
\
l \
\
0.8 / \ \

/ \
()' 0.7 / \
C \
Q)
\
~ 0.6 I

Q) I

.~ 0.5 ,
1il I
\
\
£ 0.4 I
\

I
0.3 I

I \
0.2 I \
I \
\
I
0.1 /
\

/
/
~.
.... -.-.- .... \
\

0 '-'-
490 495 500 505 510 515 520 525 530 535
Wavelength (A.)

,.=cosa/cos~ : Figure 1.22: Blaze function

I
0.9 V1.0 -----r------ for a range of Littrow angles
1/1.1 - - - - - ... - - - - - - - - -, - - ,
In2 _____ L ________ / , _ (e).
I / /'
0.8 111.3 - - - - - L - - _ _ _ .L .i 7, _ . ,

()' 0.7
c
<D
~ 0.6
:;:: ::::r:,/;:--,~'"
'>" 1/:' r ... ,\ \
<D I Ij" ~ / ... '\ "\ \
I I " .\\

~ 0.5 I " ':,~

~
I

:!:: 0.4
is
0.3
,
0.2 ,, ·'11
i /

-0.5 o 0.5
Relative order number
30 Chapter 1. Echelle spectrograph theory

1.2.12 Overfilling

Although it has been assumed thus far that the collimated beam is matched to the pro-
jected size of the echelle grating, it is not always possible to do this. As shown in Figure
1.23, the amount by which a grating is overfilled is a function of the size of the grating
(Wand L) and the angle of illumination (a). Depending on the size of the grating, the

,'r:::.. .. .. -- -------- A_ --- ----- -- --- - - -- - ----- ------------ - -- -- ---- - -- -.- --- ---- --- --- - - - ------ - - - -- --::..-..-..-.-.....
~:: ~ ~
...............
,,
,, ,
,, ·
L .
~~------------~----~1-------+------4+B
. .
·

Figure 1.23: The overfilling of an echelle grating. The collimated beam, which has a diameter B,
projects to an ellipse on the echelle grating (width W, length L). This projects to a height L' = L cos a
in the collimated beam.

elliptical footprint of the collimated beam may overfill the grating either perpendicular or
parallel to the direction of the rulings (or both, as shown in Figure 1. 23). If, as is shown
in Figures 1.24a and 1.24b, we consider the overfilling in each of the directions separately
then the fraction F of a collimated beam that is incident on a grating is given by

F = Fw +FL -1 (1.85)

where Fw and FL are the fractions of the beam captured when the overfilling in the
parallel and perpendicular directions respectively. Once the grating has been projected
into the collimated beam these fractions may be calculated by integrating the equation
of a circle, with appropriate limits. That is,

Fw -16-
nB2
l 0
W2
/ ~
-- -xdx
4
(1.86)

FL -16-
n B2
['/2 ~
0
. --xdx
4
(1.87)
1.2. Properties of echelle gratings 31

Figure 1.24: The grating may be

overfilled parallel and perpendicular
to the rulings. The fraction F of the
collimated beam that is incident on the
echelle grating may be computed by
considering the amount the grating is
overfilled parallel (Fw) and perpendic-
ular (FL) to the ruling separately. See
text for details.

(a) W (b) L

In polar coordinates, these equations become

(1.88)

(1.89)

where the polar angle limits ()w and ()L are given by

w D
cosew = 13 and cos(h = -
B
(1.90)

Evaluating equations 1.88 and 1.89 gives

Fw -4 [7r
- - -1 ()w + -1.sm 2()w1 (1.91)
7r 4 2 4

7r
[7r
-4 - - -()L
4
1
2
+ -sm2()L
1.
4
1 (1.92)

If TV > B, and/or L' > B, then the grating is not overfilled (or overfilled in one direction
only) and consequently either ()w = 0 or ()L = O. Hence, either Fw = 1 or FL = 1
depending on the direction of overfilling.
32 Chapter 1. Echelle spectrograph theory

1.3 Design of echelle spectrographs

1.3.1 Choice of echelle
The choice of grating is one of the most fundamental choices in the design of an echelle
spectrograph. In the following sections some factors which influence this choice will be
discussed.

Beam size
For a given echelle grating with a blaze angle ()B which is used to obtain a given resolving
power R it is not necessarily the case that the ideal beam width will be the width of
the echelle grating W. It has been pointed out by Diego and Walker (1985) (see also
Walker and Diego, 1985) that the echelle grating may be considerably overfilled without
compromising throughput. This is because while the grating becomes less efficient as the
beam size increases (due to overfilled light being lost) the angular size of the slit on the
sky can be increased in order to maintain a constant resolving power.
The effect is illustrated in Figures 1.25a and 1.25b. Here an R2 echelle with W x L =
300 x 840 mm is illuminated (in Littrow configuration) by a beam which can vary in
diameter. This is done in practice by varying the telescope focal ratio. The angular slit
width is varied so that a constant resolving power of R = 25 000 is maintained at all beam
sizes. The efficiency of the grating is therefore a function of both beam size and seeing.
Figure 1.25b shows the throughput relative to a beam size of 300 mm. For small seeing
values it can be seen that increasing the beam size leads to rapidly decreasing throughput
as the slit throughput always remains high. However, at larger values of seeing, the
throughput of the spectrograph actually increases as the beam size is increased. This is
because the overfilling of the echelle becomes increasingly mitigated by the larger angular
slit width.

1oo,---,-----,--------,-------r-----, 140~------r----~----~~

80 -----=:::::::::::::::=::;::::====- 3.0"
2.5"
?120
2.0"
;€' '5
e- 60 0.5" E-
'5 O>
Q.
::J 1.5"
.r:
0> 1.0" ~100
::J I-
e 40 e .~
~ -1.5" iii
(j)
1.0"

20
-----
-
_-------e----- 2.0"
_-----------<3----- 2.5"
3.0"
0: 80

0.5"
OL3~0-0----3~5-0----4~0-0----4~50~ 60 300 350 400 450
Beam size (mm) Beam size (mm)

(a) Absolute efficiency (b) Relative efficiency

Figure 1.25: The absolute (a) and relative (b) efficiency for an R2 echelle grating with Hi xL = 300 x 840
for a resolving power of R = 25000 as a function of beam size and atmospheric seeing. Open circles
indicate the most efficient beam size as a function of seeing.
1.3. Design of echelle spectrographs 33

Using such an analysis, for a given telescope and echelle grating combination, it is
possible to choose an optimum beam size, where the weighting function would depend on
the expected seeing conditions. A more detailed analysis would require that the effects of
the secondary obstruction be considered (for directly fed spectrographs) and/or the effects
of non-uniform illumination of the echelle grating (due, for instance, to the incomplete
radial scrambling of the fibre far-field).

Blaze angle
As is shown in the following section, the choice of blaze angle will have little direct impact
on the cross-dispersion. However, the blaze angle has a significant effect on the collimator
and camera properties. These will be discussed in Section 1.3.3. For a further discussion of
the choice of echelle grating blaze angle, and its implications on the spectrograph design,
the reader is referred to Section 2.1.2 and to Hearnshaw et al. (1999).

Effect on cross-dispersion
The amount of inter-order space can be tuned by altering the properties of the echelle
grating. As shown above (equations 1.46 and 1.47) the inter-order spacing depends on
the free spectral range of the echelle grating. That is, if the wavelength extent from one
order to the next is increased, while the cross-dispersion remains constant, then the inter-
order spacing will increase accordingly. Given that the free spectral range depends most
sensitively on the grating groove spacing (equation 1.14), simply changing () will change
the inter-order spacing. If the echelle grating is more densely ruled (() decreased) then
the free spectral range will increase, and therefore the total number of orders over a given
wavelength range will decrease. The effect this will have on the spectral format is- shown
in Figure 1.26. One consequence of changing the echelle ruling density simply to increase
the inter-order spacing is that the angular width of the orders also increases. This might
be a problem if the angular field of view of the camera and detector is limited.

T = 75 grooves/mm T = 100 grooves/mm

Figure 1.26: The effect of changing the echelle groove ruling density on order separation. The same
(prism) cross-disperser and camera is used for the two examples however the echelle groove spacing has
changed as indicated.

Changing the blaze angle of the echelle grating has relatively little effect on the spectral
format. That is, as discussed above in Section 1.2.9, the order separation 6y, for a given
echelle and cross-disperser combination, is given by

1
6y = Const. x -'-e-
sm B
(1.93)
34 Chapter 1. Echelle spectrograph theory

(i.e., equations 1.47). Therefore, changing from an R2 grating to an R4 grating will

decrease the order spacing by less than 10%. A description of the methods of cross-
dispersion follows.

1.3.2 Cross dispersion

As has already being discussed, any combination of order number m and wavelength A
that satisfies the grating equation (equation 1.1) will have equal diffraction angles. If
the wavelength coverage of interest spans more than a single free spectral range it will
therefore be necessary to introduce dispersion in a direction that is orthogonal to the
main echelle dispersion. Some possibilities for such cross-dispersion are discussed below.

Grating cross dispersion

The angular dispersion of a grating cross-disperser is

dt1 mg
(1.94)
dAxD a g cos t1g

where the grating order number mg is generally low and the grating ruling density a g
is high. Because the overall angular dispersion is quite low, the cross-disperser will be
blazed at quite a shallow angle (i.e, t1g is small). The physical separation between orders
is given by combining equation 1.94 with equation 1.47. That is,

f mg A~
cam a g cos t1 g 2 sin BB cos B

Const. x A~ (1.95)

Prism cross dispersion

When using a prism for cross-dispersion the angle of incidence is usually such that the
dispersed rays are very nearly parallel to the base of the prism (see Figure 1.27). While

Figure 1.27: A prism used at minimum

b deviation.

this is close to the situation where a prism with a given apex angle has the least overall
dispersion, this arrangement minimizes the total path length (of a wavelength which
travels parallel to the base) and lessens the effects of polarization and refiection losses at
1.3. Design of echelle spectrographs 35

each face. The size of the prism is also minimized. The angular dispersion of a prism
used near minimum deviation is given by
d,B b dn
(1.96)
dAXD B dA
where b is the length of the prism's base and B is the diameter of the incident beam. The
ratio b/ B effectively determines the prism apex angle ap. That is,
b tanBi ap
-=--cot- (1.97)
B n).. 2

where Bi is the angle of incidence of a wavelength (for which the prism refractive index is
n)..) such that
sin Bi = sin n).. a; (1.98)

This is the angle of incidence of a wavelength which has a minimum path length through
the prism. Now, the refractive index of a prism can be approximated using the Conrady
formula by
k2
n(A) = k1 + A2 (1.99)
where kl and k2 are constants, and hence the angular dispersion of a prism is
d,B b k2
dAXD = -2 B A3 (1.100)

The separation between orders produced by a prism can be found by substituting Equation
1.100 into Equation 1.47 which gives

fly = -2f ~~ A~
cam B A~ 2 sin BB cos B
1
Const. x AB (1.101)

Gratings or prisms?
The order separation for both prisms and gratings was derived above (equations 1.95 and
1.101). It was shown that the order separation was

Gratings: fly = Const. x A~ and

1
Prisms: fly = Const. x -
AB
This shows that the order separation given by a grating increases rapidly as the wavelength
increases (i.e, as the square of the wavelength), while the order separation decreases (at
a lesser rate) for a prism. This fact makes prisms particularly attractive in situations
where it is desirable to capture a large wavelength range on a single detector. That is, if
a prism and grating is chosen such that the total cross dispersion is the same (see Figure
1.28), the range of inter-order spacing of a grating will vary considerably. Generally this
forces the design of an echelle spectrograph with grating cross-dispersion to incorporate
several grating cross-dispersers so that the inter-order spacing at a chosen wavelength can
 36
Chapter 1. Echelle spectrograph theory

(a) Grating cross dispersion (b) Prism cross dispersion

Figure 1.28: The relative order separation of gratings (right) and prisms (left).

be varied. However, a prism has relatively uniform inter-order spacing, and one prism (or
prismatic system) is sufficient for all wavelength regions. Alternatively, a combination of
gratings and prisms (or a grism) could be considered.
Another aspect to consider is the relative efficiency of grating and prism cross-dispersers.
It is generally the case that a high quaiity prism will have significantly higher efficiency
over a broader wavelength range than any grating. This is because gratings are subject
to the effects of the blaze function. A typical high efficiency surface relief grating will
have a FWHM which is about equal to the blaze wavelength, and hence may only be con-
sidered useful over a small wavelength range. Recently however high efficiency gratings
have been developed that have a periodic grating structure which arises from modulation
of the index of refraction of a thin layer of light sensitive material. Such gratings are
termed volume-phase holographic (VPH) gratings and are discussed further in Chapter 3.
These gratings cannot however be used over more than a single octave of spectral coverage
and two or more gratings would still be required to cover a wavelength band spanning the
near-uv to the near-IR (i.e., the approximate pass-band of a high efficiency CCD detector).

1.3.3 Collimator, camera and detector properties

Collimator

The required focal length of the collimator (fcol) follows from the equality given by equa-
tion 1.23. That is,

f - ~ftel B
. col - p D (1.102)

where the focal ration degradation factor p = 1 if the spectrograph is directly fed. The
actual collimator focal length is a completely free parameter as long as the equality given
by equation 1.23 is maintained.
1.3. Design of echelle spectrographs 37

Camera
The focal length of the camera lcarn is determined by noting that in order for the maximum
resolving power Rrnax to be achieved the OOD must sample at least two resolution elements.
It therefore follows that

lcarn = ns;rnp RrnaxSpix cot BB (1 + tan BB tan B) (1.103)

where n sarnp is the number of OOD pixels per resolution element. Typically n sarnp = 2 for
critical Nyquist sampling with pixels each having a size Spix, giving

(1.104)

for small B. This shows that large blaze angle gratings require short focal length cameras.
However, because R = Const. x BtanBB, (equation 1.58) the monochromatic focal ratio
of the camera will be given by

lcarn C
B = onst. x Spix (1.105)

That is, for a given maximum attainable resolving power, the focal ratio of the spectro-
graph's camera will depend only on the COD pixel size. The effective focal ratio of a
spectrograph camera, which determines the camera's actual size, depends rather more on
the location of the entrance pupil.

Detector
That the chosen pixel size influences the camera's focal length was pointed out in the
previous section. The number of pixels npix required by a detector to completely sample
an order is given by
(1.106)
Spix
which, given equations 1.15 and 1.104, can be approximated to give

RrnaxAB
npix ~ --- (1.107)
a sin BB
Given that mAB ~ 2a sin eB , equation 1.107 can also be written as

(1.108)

which shows that all high resolution spectrographs require large detectors if wavelength
coverage is complete. Often, for the sake of economy, compromises are made either in the
maximum resolving power and/or wavelength coverage.

1.3,4 Fibres
The use of fibres in astronomy was first suggested by Angel et al. (1977). Their idea,
which was made possible by the recent development of high quality fused silica fibres,
was to link numerous small aperture telescopes to a single instrument. Subsequently
38 Chapter 1. Echelle spectrograph theory

fibres were used in multi-fibre applications such as the simultaneous observation of many
objects (for example, the Medusa spectrograph (Hill et al., 1980)), or to obtain spectra
over a two-dimensional area (for example, the DensePax fibre optic array (Barden and
Wade, 1988)). Both of these applications demonstrate that fibres contribute towards
considerable improvements in the efficiency of spectroscopic observations.
Another practical benefit of the use of fibres is that the instrument is removed from
the telescope. Hubbard et al. demonstrated the feasibility of this in 1979 (Hubbard et al.,
1979). This removes the constraints of size and weight of any fibre-coupled instrument,
while also allowing such an instrument to be placed in a potentially more stable envi-
ronment, where the effects of flexure, temperature, and pressure changes may be absent.
Hence fibres are of particular value for the high-precision measurement of radial velocities.
A further advantage of the use of fibres in precision spectroscopy is the ability of a
fibre to scramble the input image structure. This means that regardless of the distribution
of light on the input face of the fibre, the output face will appear more uniform. Hence,
systematic errors due to slit illumination may be reduced. This type of image scrambling
is referred to as "near-field" scrambling. It was also realized that the optics of a fibre-fed
instrument may be illuminated more uniformly due to the scrambling properties of a fibre.
That is, the angular distribution of rays exiting a fibre will not betray the distribution
that entered the fibre. This type of "far-field" scrambling also has the potential to increase
the stability of the spectrograph. However, as observed by Hunter and Ramsey (1992),
and predicted by Heacox (1987), while the azimuthal scrambling of rays in the far field
is nearly complete, the radial scrambling is not quite as good. These effects also impinge
subtly on the illumination of the slit exit (or the "near-field" image) and hence may cause
significant drifts in line profiles or positions. A method for increasing the scrambling via
means of a "double-scrambler" has been proposed by Brown (1990). The double-scrambler
is inserted in a break in the fibre and its purpose is to invert the positional and angular
dependence of the rays crossing the junction between the two fibre halves.
The implications of coupling a spectrograph to a telescope via optical fibres was dis-
cussed above in Sections 1.2.6, 1.2.8, 1.2.10.

1.3.5 Merit functions

A common merit function used for comparing spectrographs is the slit-resolving power
(Res) product which follows from equation 1.61 (or equation 1.59). That is,

Res = 2L sin eB cos e (1.109)

pD

As stated in Section 1.2.10, this equation shows that for a given resolving power and
angular slit size, a large telescope requires a large grating. A more complete merit function
would also take into account the throughput of the spectrograph, T; i.e., T Res. This was
introduced by Jacquinot (1954) in a different form. However, as pointed out by Vaughnn
(1994), a more appropriate merit function would maximize the product of the signal-to-
noise ratio (for each spectral element) and the total number of resolution elements. This
is applicable because the performance of the spectrogragh is then intimately linked to the
telescope and its environment.
1.4. Summary 39

1.4 Summary
A theoretical basis for the design of high resolution echelle spectrographs in astronomy
has been outlined. A particular emphasis has been placed on the implication of coupling
the spectrograph to a telescope via an optical fibre. In the following chapter (Chapter
2) the design and performance of a fibre-fed spectrograph designed for small to medium-
sized telescopes will be described. In Chapter 3 the design evolution of a high resolution
spectrograph for an ll-metre telescope will be presented.
Chapter 2

Design and performance of HERCULES

The optical design of HERCULES is described in detail in the following section (Section 2.1).
The detailed design of HERCULES was not part of this thesis, however some justification
for the design is given here. In Section 2.2 a summary of the predicted and measured
performance is given. Finally, Section 2.3 describes a few upgrade options which could
improve the performance of HERCULES. The HERCULES observing manual is included as
an appendix (Appendix C).

2.1 Design

2.1.1 Introduction

Since 1977 a Cassegrain echelle spectrograph has been in operation at Mt John University
Observatory (MJUO). Initially this instrument was designed to operate with the Boller and
Chivens 0.60m telescope. The spectrograph design is based on the Harvard-Smithsonian
design (see Hearnshaw 1977, 1978). It uses a 79 grooves per millimeter grating, which has
an area of 102 x 206mm. The grating has a blaze angle of BE >=::::: 63° (i.e., tanBB = 2).
In 1987 the 1 m McLellan telescope was built, and the echelle spectrograph was used
with this telescope until 2001 when it was decommissioned. The McLellan telescope is a
Dall-Kirkham design which uses an ellipsoidal primary mirror and a spherical secondary
(Nankivell and Rumsey, 1986). Although the telescope delivers an f/13.5 beam to the
spectrograph a focal reducing lens immediately in front of the slit reduces this to f/10. A
100 Mm slit therefore subtends an angle of 2.1/1 on the sky and delivers a resolving power
of R >=::::: 35000 with a collimated beam size of 54mm. The camera, which is a spherical
mirror, was designed for use with photographic plates approximately 50 mm square and
the collimator, which is an off-axis paraboloid, produces astigmatism in the direction
of cross-dispersion. This was deliberate in order to avoid saturating the photographic
emulsion. A schematic of this instrument is shown in Figure 2.1, and Figure 2.2 shows
an early photographic spectrum.
For increased sensitivity the spectrograph was used with a cryogenically cooled linear
diode array (MacQueen, 1986). The Reticon RL1872F chip was capable of observing a
25 mm length of a single order. A Photometries CCD with a Thomson TH7882 CDA detector
(384 x 576 pixels each 24Mm square) was acquired in 1988 (Tobin, 1992). In April 1996
this CCD was replaced with a larger format CCD (Barnes et al., 2000); a thinned and
back-illuminated SITe SI-003AB which has 1024 x 102424 Mm square pixels. However, this
ceD was still incapable of observing more than a small fraction of the available spectral
format. The spectral coverage was considerably improved in 1998 when a focal reducer
was installed (Tobin et al., 1998).

41
42 Chapter 2. Design and performance of HERCULES

Figure 2.1: The Mt John University Observatory Cassegrain echelle spectrograph. The design uses
an R2 grating, with 79 grooves per millimeter. The grating has an area of 102 x 206 mm, and three
interchangeable gratings are available for cross-dispersion.

The fibre-fed adaptation

The MJUO echelle spectrograph was adapted for fibre fed work in 1989 (Kershaw and
Hearnshaw, 1989). By removing the spectrograph from the Cassegrain mount, where
flexure is a major problem, and enclosing the spectrograph in a thermally isolated room
the spectrograph could be used for obtaining precise radial velocities. Initially the detector
used was the Reticon linear diode array. Later CCD s were used, and radial velocities of
bright stars (V :::; 7) could be obtained to a precision of about 50 ms- 1 (Murdoch et al.,
1993; Skuljan et al., 1999; Cummings et al., 1999).
Although this system worked well, and gave some excellent and precise radial velocity
measurements, it was inherently inefficient because of the low throughput into the fibre,
the need to reimage the fibre onto a slit to improve resolving power, the small beam size
of the spectrograph, the use of grating rather than prism cross-dispersion, the absence
of anti-reflection coatings or high-efficiency reflecting surfaces and the large cross-order
profile (due to deliberate collimator astigmatism) and the consequently limited number
of orders, and hence overall wavelength coverage, that were available on the CCD. The
need to allow this instrument to be used in both Cassegrain and fibre-fed modes, and the
variety of manual adjustments and interchangeable gratings undoubtedly contributed to
its instability.
It was realized that a much more efficient and stable spectrograph was possible in
principle if care were taken in matching all parameters to the image size, pixel size and
OOD dimensions, if very efficient optical coatings were applied to relevant surfaces and if
great care was taken to stabilize the environment in which the spectrograph operated.
This new instrument, which meets all these design criteria, is known as HERCULES
(High Efficiency and Resolution Canterbury University Large Echelle Spectrograph) and
came into operation at MJUO in April 2001.
44 Chapter 2. Design and performance of HERCULES

2.1.2 Optical design

The optical layout of HERCULES is shown in Figure 2.3 and a schematic of the mechanical
design is shown in Figure 2.4. The design of HERCULES was done by J. Hearnshaw.
G. Nankivell and N. Rumsey advised on the prism glass and developed the Zemax optical
model and G. Nankivell produced the Schmidt plate parameters. The present author
designed the fibre feed optics, including the guide and acquisition cameras, as well as
the exposure meter. A summary of the HERCULES parameters can be found in Appendix
A.2.1 and the design is summarized by Hearnshaw et al. (2002).
The nominal wavelength coverage of HERCULES is from 380 nm to 880 nm and with a
fibre entrance aperture of 4.5/1 (on a I-metre telescope) a resolving power of R = 40000
is possible.

Design evolution
Design work on HERCULES commenced in 1995 and from 1995-1996 two design options
were explored. One was an R2 spectrograph with a folded Schmidt camera. The other was
an R4 instrument with a white pupil and employing a multi-element refracting camera.
The white pupil reimages the undispersed pupil incident on the echelle onto the entrance
pupil of the camera, thereby ensuring smaller optical components and better control of
camera aberrations. These two designs were developed further in 1997-98. The R4 in-
strument was considerably more compact and used a 10-cm collimator beam size falling
on an area of 400 x 100 mm (ruled surface area) for the R4 echelle. The cost advantage
of this compactness was offset by the complexity of the camera which was necessary for
such a wide wavelength coverage; three different all-refracting cameras were designed, all
with 10 optical elements, some of which entailed use of rather exotic (hence expensive)
glasses (the best performance came from the camera using several fluorite lens compo-
nents). The R2 spectrograph was relatively large, thereby adding to the cost of some
components, including the echelle (about 400 x 200mm), but this instrument employed a
folded Schmidt camera, giving superior near achromatic performance over a wide field of
view and wavelength range, together with relative simplicity of camera fabrication. The
light loss inherent in the folded Schmidt design due to the hole in the fold flat is moder-
ate (::; 23%) for the R2 design, but would be prohibitive for an R4 instrument, which is
why an all-refracting camera is unavoidable for that spectrograph. The overall size of the
instrument was not considered an issue given that it would be fibre-feed. In mid-1998,
after carefully evaluating both R2 and R4 designs, a decision was made to proceed with
the former (Hearnshaw et al., 1999).
Construction of HERCULES proceeded from July 1998. The prism was optically figured
and polished by D. Cochrane at Industrial Research Limited 1 . All other optical compo-
nents of HERCULES were fabricated by G. Nankivell, also in Lower Hutt, at his private
optical workshop. The mechanical design and construction took place in the workshops
of the Department of Physics and Astronomy under the direction of G. Kershaw. N. Frost
designed and built the fibre-feed module with electronics provided by R. Ritchie. The
control software was developed by G. Graham. HERCULES was commissioned on 2001
April 3.

1 Industrial Research Ltd, Lower Hutt, New Zealand

46 Chapter 2. Design and performance of HERCULES

Cmnera primary

Collimator

Figure 2.4: The HERCULES spectrograph inside the vacuum tank. The tank is formed in three sections.
The first section, which encompasses the camera optics, is rigidly connected to the spectrograph bench.
The lid (on which the camera is mounted) and the other two sections are free to roll away on rails.
All optical mounts are fabricated from cast aluminium. The mirrors are all supported by thin stainless
steel bands. There is minimal provision for optical alignment as a single alignment is made during final
assembly.
2.1. Design 47

Vacuum tank
From the outset it was intended to put HERCULES in an environment which would be
immune from the effects of changes of either atmospheric pressure or the temperature of
the spectrograph room. The importance of such immunity to the precision of the spectro-
graph is emphasized by considering that a 1 mbar (hPa) increase in air pressure (at sea
level) will give rise to a spurious Doppler shift of -80 mis, while a temperature change of
l°(at 25°) will result in a shift of 270m/s (Murdoch et al., 1993). These effects, which
are caused by the pressure and temperature dependence of the refractive index of air, can
be completely eliminated if the spectrograph were enclosed inside a vacuum. A vacuum
will also eliminate both convection currents and temperature stratification; both of which
would impact on the spectrograph's stability. The use of helium was also considered.
This monotonic gas has a refractive index very close to that of a pure vacuum with a con-
sequently small dependence on temperature or pressure. The high thermal conductivity
of the gas would aid the thermal equalization of the spectrograph structure. Concerns
about the permeability of helium versus the relative ease of constructing and maintaining
a light vacuum tank were behind the decision to use a vacuum. The vacuum is supplied
by an aluminizing chamber located around 20 m from the HERCULES room.

Echelle grating
HERCULES uses a large R2 31.6 grimm echelle grating from the master ruling MR152 from
the Richardson Grating Laboratory2. for which the BB = 64.33°; tan BB = 2.08. The
ruled area is 204 x 408 mm. The relatively coarse ruling gives a small angular width of the
orders and hence an optimum match to a 50 mm square CCD detector. The Littrow angle
B is made as small as practicable (small B requires a much longer spectrograph, in order
to separate the incident and diffracted echelle beams) so as to give high peak efficiency
to the orders within a free spectral range (see Section 1.2.11 and references therein). In
the case of HERCULES we chose B = 3.0°, which with a 200 mm collimated beam allows
the entire spectrograph to be housed inside a 4.50-m long tank.

Cross-dispersion
The choice of prism cross-dispersion gives high efficiency at all wavelengths (a maximum
of 90% of the photons are preserved after two passes) while ruled grating cross-dispersion
peaks at about 70% and suffers from a limited wavelength range each side of the blaze. The
cross-dispersion arrangement adopted was influenced by the similar design of Libbrecht
and Peri (1995) for the Hale 5-m telescope fibre-fed echelle at Palomar, but HERCULES
differs from the Hale instrument in most other aspects of its design. Cross-dispersion
is provided by a single BK7 prism (PH3 quality glass which has a maximum deviation
of the refractive index of no more than ±2 x 10- 6 ) in a double pass-mode. The prism
apex angle is elp = 49.50° and dimensions are a height of 276 mm, a base of 258 mm and
a length (between triangular faces) of 255 mm. The mass of the prism is 22.7 kg. The
cross-dispersion is sufficient for a minimum of 17/1 separation between adjacent orders.

2Richardson Grating Laboratory, Rochester, New York, U.S.A.

48 Chapter 2. Design and performance of HERCULES

The refractive indices at standard wavelengths were measured by Glass Fab 3 and the
data are given in Table 2.1. The measured dispersion (at 587.6 nm) is

Vd = 64.25

and both this measurement and the refractive indices are within the standard tolerances
for this glass; that is,

nd = 1.51680 ± 0.001 and Vd = 64.17 ± 0.8%

The refractive indices are reliably modeled by the Conrady formula

(2.1)

where the coefficients are:

no 1.500301,
A 8.24268 X 10- 3 and
B 4.46946 x 10-4

Line Wavelength (nm) Refractive index

nF' 480.0 1.52330
nF 486.1 1.52285
ne 546.1 1.51909
nd 587.6 1.51722 Table 2.1: Refractive index melt data for
nc, 643.8 1.51520 HERCULES BK7 prism. The measurements

nc 656.3 1.51480 were provided by Glass Fab.

As shown in Section 1.2.8 every spectral line will be tilted by an amount ¢ given by

tan ¢ ~ 2 tan BE tan "I . (2.2)

Because half the cross-dispersive power is located before the echelle grating and the to-
tal variation in the angle "I will be approximately equal to half the total angular cross
dispersion b..,8xD.Therefore, the variation in line tilt tan(b..¢) will be

tan(b..¢) 2 tan "I tan BE

~ 4tan (~b..,8XD)
and given that b..,8XD ~ ±2.5°, the variation in "I is b.., ~ ±1.25°, and hence the variation
in line tilt across the HERCULES spectrum will be b..¢ ~ ±5°.
3Glass Fab Inc, Rochester, N.Y.
2.1. Design 49

Collimator
The collimator mirror of HERCULES is an on-axis paraboloid with a 783 mm focal length.
The diameter of 210mm is designed to be used at f/3.75 where the fibres are placed at
the mirror's focal point. This is possible because the number of fibres used is small, and
there are no pre-slit optics, and therefore the resulting obstruction is small « 5 mm in
diameter). The elliptical illumination on the echelle overfills the ruled area by 14.5% (i.e.,
the semi-major axis of the beam is 272mm). However, a large collimated beam size B
allows a larger angular slit size es for a given resolving power, and hence a net efficiency
gain (see Section 1.2.12 and references therein). The collimator mirror was made from
Zerodur with a 35-mm edge thickness.

Camera
The HERCULES camera is a folded Schmidt which gives outstanding performance with
respect to aberrations over a very large wavelength range (380-880 nm) and high efficiency.
The arrangement with a perforated fold flat (perforation diameter 100 mm at an angle of
19.0° to the mirror's normal) gives rise to a light loss of up to about 23% at some central
wavelengths. On the other hand, placing the CCD dewar in the beam would have led to
even more obscuration and lack of access to the detector. The fold flat is 55 cm in diameter
and tilted at 19.0° to the optical axis from the Schmidt plate. In our folded design the
CCD dewar is located outside of the HERCULES vacuum tank, and the field-flattening lens
immediately in front of the detector acts as a window for the tank. The BK7 Schmidt
corrector plate has a 525-mm trim diameter and 15-mm thickness. The camera mirror of
HERCULES is 500 mm in diameter with an edge thickness of 66 mm and a centre thickness
of 50 mm. The optical surface is spherical and the concave radius of curvature is 1946 mm,
giving a focal length of 973 mm. The material is Zerodur. The camera is shown in Figure
2.5.

minor

Field-flattening
/ lens

Figure 2.5: The HER-

CCD CULES camera. The de-
Conector plate sign is a folded Schmidt.
50 Chapter 2. Design and performance of HERCULES

Spot diagrams for a variety of wavelengths and spectral orders are shown in Figure
2.6. The rms spot size ranges from 4.5/Jm in the uv to below 1 /Jm near H;3, 3 /Jm at
Ha, rising to 6 /Jm near the ends of the red orders.

Order 65:

>.. = 870Anm >.. = 875.1nm >.. = 879.8nm

Order 90:

>.. = 628.6nm >.. = 635.6nm

Order 120:

>.. = 474.1nm
Order 140:

>.. = 405.0nm >.. = 407.9nm

Order 150:

>.. = 378.1nm >.. = 379.3nm >.. = 380.6nm

Figure 2.6: HERCULES spot diagrams as predicted by ZEMAX. All boxes are 15 Mm square.
2.1. Design 51

Spectral format
The spectral format of HERCULES is shown in Figures 2.8 to 2.10. Examples of small
regions of spectra are shown in Figure 2.7. As was noted above, HERCULES was origi-
nally intended to be used with a 2k by 2k CCD having 25 /km pixels which would have
enabled complete spectral coverage from 380 nm to 880 nm (see Figure 2.8). At the time
of construction only the Series 200 CCD with 1024 x 1024 pixels each 24/km square was
available to be used with HERCULES. It is therefore not possible to observe the entire
spectral range simultaneously. In order to observe all of the spectral format a detector
cradle was constructed which has four discrete CCD positions. The design was intended
to cover the spectral regions shown in Figure 2.9. However, during assembly the spectral
format was slightly altered to give a more centrally located CCD position which is better
suited to precision radial velocities. The CCD positions currently available with HERCULES
are shown in Figure 2.10. Note that it is currently impossible to observe spectra above
720 nm and the lower limit of 370 nm is a result of the instrument efficiency (i.e., fibres,
mirrors and CCD). It is likely that a new CCD will be acquired which will capture the
entire visible region from 360nm to l/km (see Section 2.3.1).

III
" It "'.
'" ill'
• lit

•
lot
oiII

It
•
ill .. .

(a) A Th-Ar lamp spectrum (b) A stellar spectrum (31 Aql, G8IV)

Figure 2.7: A small region of the HERCULES spectrum.

 CJ1
t-V

Figure 2.8: The HERCULES

Order AB l1y spectral format with a single 2k
30 . x 2k CCD with 25J-Lm pixels.
59 ,
63
, tt Py

~
~ '965
904
17.~ #
" .3"
Each order extends over two free
spectral ranges (FSR) except in
~
67 ~ 85 ~ 16.9"
the far red. At these wavelengths
71 1102 16.8"
~ 759 16.9"
(which are beyond the nominal
20 ---- ~ design limits) the vignetting due
83
87
Hel All
Nil
°2(8)
Lil
-
811
,7 721
686
17.1"
17.4" to the hole in the fold-flat mirror
and the field-flattening lens lim-
- - - l:Ia 655
626
17.8"
18.2" its spectral coverage.
600 18.8"

10 99 75 19.3"
103 53 20"
--.
107
E 32 20.6"

E
---
111 513 21.3"

C 115 495 22.1"

0
:.t=' 0 119
·w0 123
479

463
22.8"

23.6"
Q. 8a
I 127 Mgll 449 24.4"
>-
131 FFMI(d) 435 25.3"
-10 135
H(G)
422 26.2"
Q
..,::r
C '0
c+
139 O,1)~ 410 27.1" (0
'"1

143
He ~
398 28"
tj
Call(l<)
-20 147 CN(1, CN(O,~?8 29"
(0

oq.
Ul

Hel ::;
151 H1 3n 30" ..,
, ::;

155 I
! , Oil
368 31 "
0..
'0
(0
....,

~
I 0'
....,
-30 159 358 32.1 "
S
..,
FSR ::;
(")
(0

FWHM 0
>-+,
::z::
tyj

-30 -20 -10 0 10 20 30 ~

Q
c:
X-position (mm) t-<
tyj
Ul
 ~
~

Figure 2.9 : The HERCULES

t:J
Order AB /1y spectral format showing the four ('0
til
Qq'
30 ., " Py
~ '965
nominal posit ions of t he Series
200 1k x 1k CCD wit h 24 J.Lm
~

59 ,
63
v
, 904
17.~ ~
" .3" pixels, The actual positions are
67
71
.
~ 850~
1KJ2
~
16.9"
16.8"
shown in Figure 2.10,
. ~

, ~ 759 16.9"
20 "71J:
83
- - ,
" '\ 1, ,
\
_-_ I
I All
° 2(B)
- CTI ~ I:

I ,~-,. , -, 721
686
17.1 "
17.4"
87 --~\ - :~ Hel
f-
--- 1-- - Nil
Lil SII
Ha
SII
I~ 655 17.8"
'"
626 18.2"
600 18.8"
99 75 19.3 "
10
103 53 20"
....-..
107 32 20.6"
E
E
'-..-"
111 513 21 .3"

C 11 5 495 22.1"
0
:;::::; 0 119 479 22.8"
'00
0 123 463 23.6 "
C. Barr
I 127 MgII 449 24.4"
>- 131 F~MI(d) 435 25.3"
e ~H(G)
- 10 135 422 26.2"

C
139 (O , 1 )~ 410 27.1 "

He
143 398 28"

-20 147 CN(1 ,1T .'

,,..
n '
,v ,
I.
.•
Call(K)
CN(O,~~8 29"

151
Hel
ru
, H10 3n 30"

155
:- 011
I
011

! ,
I 011
368 31 "
I
-30 159
~ 358 32.1"

FSR
FWHM

- 30 -20 -10 0 10 20 30
X-position (mm) Ql
w
 Figure 2.10 : The HERCULES
Order spectral format showing the ap-
30 ., , Py
~ "96 5
proximate location of the four
positions of the Series 200 lk x
,,
59 , 17 .~ #
63 . P9
~: 11
Po
904
.,
" ,3" lk CCD with 24 fJ,m pixels.

'. [I i,'
Cjj
67 , 9! ! e ~ ~e 85o ~ 16.9"
71

;:
,
\
'I: [!

-
, , ~ I/O2
~ 75 9
16.8"
16.9"
20
83
- .~
I _ I _ "" ,

\
_ I _ I _ I

Hel
.
I
-
All
U 1tl)
2
lil
- ---
~I~
ClT_ ;;;-;.
811 '"
- I ~i1
, ,-,
~(a~ 72
6 86
17.1"
17.4"
87 -- -\ ~- ~, - - Nil ..
- -~
-
65 5 17.8"
18.2"
- 62 6
60 o 18.8"
99
- 7 19.3"
10
103 53 20"
.....-... - -
107 3 20.6"
E
g 111 ) 51 3 21 .3"

c 115 49 22.1"
0
.~ 0 119 479 22.8"
·w
0 123 463 23.6"
Q. Sa
I I
127 449 24.4"
>-
MglI - I
~

131 FFMI(d) 43 5 25.3"

rT('
H(G)
-10 0
135 42 2 26.2"
..,::r-
Crvr ~
139
"
8: (O.l)~ 41 27.1"
.,
([)

He . ~
143 .~
398 28"
- . " ~
,~,

t:I
- 20 147 CN(l . ~
I"
- • ... '"
~,
Call (1<)
CN(o.n~ 8
29"
([)
en
oq.
Hel
-rrv
·• · .~.
~
..,
·
-
•
151 Hl0 377 30"
~
~,
I 0..
• -• 011
"0
155
- 368 31 "
.,
·
([)
•
I .,
0'
- 30 159 • 35 32.1"
..,S
FSR ~
(")
([)

FWHM 0
......,
::r:
M
-30 -20 -10 0 10 20 30 ;J:!
0
C
X-position (mm) t<
M
en
2.1. Design 55

2.1.3 Fibre feed

Fibres
The three fibres used with HERCULES are all mounted in close proximity around the focal
point of the collimator. Table 2.2 gives details of the fibres available and the expected
resolving powers. Fibres 1 and 2 are used as bare fibres and fibre 3 (a 100-jLm core
diameter fibre) has a 50-jLm micro-slit on the exit face.

Fibre # core diam. (jLm) micro-slit (jLm) R

1 100 41000
2 50 82000
100 70000 Table 2.2: HERCULES fibres and re-
3 50
solving powers.

The fibres selected are Ceramoptec step index fibres with enhanced uv transmission
(see Figure 2.11). The claddingjjacket diameters are 60/70 Mm or 140/155 Mm for the
50-Mm fibre and 100-jLm fibres respectively. Each of the three fibres was mounted inside
a hypodermic needle. The 100-jLm fibres were then placed inside a 150-Mm needle, while
the 50-Mm fibre was placed inside a 200-jLm needle along with several short pieces of fibre
for packing. The fibres were fixed in place with an epoxy resin, and the three needles were
in turn mounted inside a brass ferrule. The exit faces of these fibres were then cleaved
and polished. A length of 22-m is needed to span the distance between the telescope's
focal plane and the spectrograph room. All fibres were prepared by G. Kershaw.

(iJ
.0
«J
:c
I-t;::
c
0
'iii
(f)

'E
(f)
c

-
(Ii
....
....
( ])

@
Figure 2.11: The trans-
mission of the HERCULES
fibres, Tfib(abs)' The fibres
-
()
(])
0-
0
E
0.8 are CeramOptec and have
high-OH content (for enhanced
(Ii
.... 0.75 uv transparency). Each fibre
(]) is 22 m long. The absorption
0
peaks are due to OH.
400 450 500 550 600 650 700 750 800 850
Wavelength 'A

Focal ratio degradation and focal reducer

The fibres listed in Table 2.3 were tested for their focal ratio degradation (FRD) properties
using the apparatus which is shown schematically in Figure 2.12. All the lenses were
achromatic camera lenses. The input source was a 400 Mm pinhole which was illuminated
with diffuse white light. The lens Ll created a collimated beam, the diameter of which
was controlled by an adjustable iris diaphragm. By this means a beam of known focal
56 Chapter 2. Design and performance of HEROULES

Company Fibre type Corel cladding/buffer length

dimensions (/-Lm ) (m)
CeramOptec Optran UV 050/060/070 20
CeramOptec Optran UV 100/140/155 20
Polymicro FH-type 055/077/220 21
Polymicro FH-type 105/147/210 21

Table 2.3: Description of fibres tested for FRD. All the fibres are uv-enhanced
high-OR fibres. The 55 and 100/-tm Polymicro fibres have additional buffers
which extend to 500/-tm made of nylon and acrylate respectively. These fibres
were kindly provided by Prof. Fred Watson of the Anglo-Australian Observa-
tory, Siding Spring, Australia and were manufactured for the FLAIR multi-
object spectrograph.

ratio can be produced at the focus of the lens L2. That is,

(2.3)

where 12 is the focal length of lens L2 and Diris is the diameter of the iris. The purpose
of creating a collimated beam from the output of the fibre using the lens L3 was to allow
access to both the near-field (PI) and far-field (P 2) positions in the emergent beam with
the repositioning of a single element. In practice, this was done by mounting the lens
L4 on a micrometer translation stage, while the detector remained in a fixed location.
The detector was a Lynxx CCD camera which has a 2.64 x 2.64mm detector area with
192 x 164 pixels which are 13.75 x 16/-Lm in size. The relationship between the distance
d from the near-field focus and the diameter of the far-field image gives the focal ratio;
i.e,
(2.4)
where D95 denotes the diameter within which 95% of the flux is enclosed.
Image reduction was done using ESO-MIDAS4. This involved subtracting the back-
ground, centring and normalizing the image, then computing the enclosed flux as a func-
tion of image diameter. Some examples of the far-field and near-field images are shown
in Figure 2.13.
The resulting focal ratio degradation curves are given in Figure 2.14. As a check on
the effects of end-face preparation and on the repeatability of results, each of the fibres
(except for the 55/-Lm Polymicro fibre) was tested using both ends as the input. No
significance difference was noted. A puzzling aspect of these results is the apparently
poor FRD properties of the two Polymicro fibres (especially the 55/-Lm fibre). These
results compare badly with other published test results for similar Polymicro fibres (for
example, Ramsey, 1988). However, similar results as for the 105/-Lm fibre were obtained
by Carrasco and Parry (1994) for a Polymicro 200/-Lm fibre. Whether the results obtained
for both Polymicro fibres are indeed consistent with the expected behaviour of these fibres
is unknown - especially given the wide variation observed in the FRD behaviour of optical
fibres. It is possible that the Polymicro 50/-Lm fibres were damaged in transit to New
Zealand, although this has not been confirmed.
4European Southern Observatory - Munich Image Data Analysis System
2.1. Design 57

Inputi-------------------------------------------------------:
I I

I
Fibre I
input :

I
I
I
Diffuse light source :
and pmhole I
I

Adjustable iris LI :
________ _ ::. ________ ~~~l~:g~_____________________________ __ :

Test fibre

r----------------------------------------------------- --~

Fibre
I
I output
P2 PI I
I
I
CCD (far-field I

and near-field)
I
I
I
I
Figure 2.12: The FRD
: __________________ ~4__________________________ ~~ _________ : Output test setup.

On the basis of the above results, in order to obtain the output focal ratio of 1/3.75
that HERCULES requires (see Section 2.1.2) the following input focal ratios are required
(assuming CeramOptec or equivalent fibres are used):

CeramOptec 100 pm : 1/Dg5 ='1/4.6

CeramOptec 50 pm : 1/D95 = 1/4.5,
where in both case 95% enclosed flux is assumed. The silica micro-lens shown in Figure
2.15 is used to provide the focal conversion from the McLellan telescope 1/13.5 beam to
1/4.5 (or 1/6.8 in glass). The size ofthe lens, which is arbritary, is a compromise between
manufacturability, sensitivity to misalignment, and optical performance.
58 Chapter 2. Design and performance of HERCULES

(a) Far-field image, f / Din = 2.78 (b) Far-field image, f / Din = 4.17

Figure 2.13: Examples of (raw) far-field

and near-field ~mages of the fibre output
(lOOl1m CeramOptec fibre). The far-field
images (a) and (b) are taken with the COD
positioned 4.50mm from the near-field fo-
cus and are both 2.64mm square. Note that
various blemishes and dark patches are vis-
ible in each of the images. These are due
to dust on the OOD window, but have only
a small' effect on the measured intensities.
The near-field image is of the exit face of
the fibre. The size of the frame is approxi-
mately 1.3mm square, where there is a 1.1:1
scale factor between the size of the object
and its image.

(c) Near-field image

2.1. Design 59

6 6

0.93 0.93
0.95 0.95

0.97 5 0.97

• 0 • 0

o •
:' 0 • 0
• 0
00
:'0·
3 •• 0 3
~.~~O
.'...
• 0

3 4 5 6 7 B 9 10 11 3 4 5 6 7 B 9 10 11
Input liD Input liD

(a) CeramOptec 100 f..tm (b) CeramOptec 50 f..tm

0.93
4 0.95 4
0.97

o • o
O' 0
:::-
.. o· [) ::; 3
.:
:' o. D
0·0
.3-
:' caD
: [lac
. '0
"
o
o

2 2

2 3 4 5 6 7 B 9 10 11 3 4 5 6 7 B 91011
Input l/D Input liD

(c) Polymicro 105f..tm (d) Polymicro 55 f..tm

Figure 2.14: Measured FRD curves for four different fibres. The focal ratios have been measured at 93,
95 and 97% enclosed flux. Note the change of scale for graphs (c) and (d) and the apparently very poor
performance of the 55 f..tm Polymicro fibre.

• v;;;: 1746mm
•I
I
~L-....-.
_ _ r_=9.031_nm _ _ _ _ I]
_fll3.5_ -C~f~/6'8~~
-= Fibre entnmce

Figure 2.15: The HERCULES micro-lens.

60 Chapter 2. Design and performance of HEROULES

Guide camera
An intensified CCD camera from DEp5 is used for both acquisition and guiding. The
intensifier has a useful input diameter of 18 millimeters. This is demagnified 4.5 times in
order to match it with the image area (6.0 mm by 4.5 mm) of the NXAI011 CCD. Camera
optics have been designed to reimage the telescope's focal plane at either f /8.4 or f /21.6
in order to provide the option of 5.3' x3.7' or 2.1' x 1.4' fields of view. The optics are
based on a design of G. Nankivell which was reoptimized to allow the use of off-the-shelf
components. The two camera modes were intended for the "acquistion" (at f /8.4) or
"guiding" (f /21.6) of an object. The object can be viewed by inserting either a fully
reflective diagonal mirror or a 92/8% beam-splitter in the optical path (see Figure 2.16).

Secondary
lllllTor

II
t \
/
I \
I \

\ II
Primary
\ \
/milTor

Guide
F?ld _______ ~camera
IllllTOr ~~__~~~

:I .......l - - - - - - Fibre
"
Figure 2.16: The McLellan 1 metre tele-
entrance scope and fibre feed guide camera.

5Delft E1Elctronische Producten, Holland

2.1. Design 61

The guide and acquisition modes are selected by rotating a single set of lenses about a
central point (see Figure 2.17). In practice the "acquisition" mode is generally also used
for guiding because the faster focal ratio mitigates to some extent the poor sensitivity of
the DEP camera.

1\ I f From telescope
I I
! 1

"Aguisition" mode

Fold min'or /
or
b"m-'PHtt,,\ T,IM",'
plane
foc>l(
~ "Guide" mode

To fibre

Figure 2.17: The HERCULES fibre feed guide and acquisition camera.

The calibration modes

Two calibration lamps are available; a thorium-argon hollow cathode lamp for wavelength
calibration and a quartz halogen white lamp for order definition and/or flat-fielding.
Simple reimaging optics have been designed for each lamp which delivers an f /13.5 beam
to the optical axis of the fibre micro-lens. In the case of the thorium-argon lamp the
cathode is used as the object whereas in the case of the white lamp a small diffuser is
used instead. It is also possible to reimage the fibre entrance onto the guide camera. A set
of LEDS mounted around the collimator mirror provide back illumination. This provides
a means of locating the approximate position of the fibre input on the sky.
62 Chapter 2. Design and performance of HERCULES

The exposure meter

Light that would otherwise be lost due to the hole in the fiat mirror of the folded Schmidt
camera is used by the exposure meter. A diagonal mirror and relay optics are placed
immediately behind the Schmidt corrector plate which directs light to an exposure meter
located outside the evacuated tank. The HERCULES exposure meter is a Thorn EMI9924
photomultiplier tube (PMT). This type of high gain PMT has a rubidium bialleali pho-
tocathode which gives a high quantum efficiency in the blue (see Figure 2.18) although
it has no response beyond 700 nm. At an operating voltage of approximately 1 leV, this
photomultiplier has nominal gain of G = 2.1 X 10 6 .

3o,.--,.--,----,--,--,--..,--..,-----.
_0,1

'I
E 0.09
25 c:
I" 0.08
$:
20 ::s 0.07
>-
:s: 0.06
E
~ 0.05
0
C.
en 0.04
10 l!:
.g; 0.03
0
1ij 0.02
t.l
0
(50.01
.c:
g5~0-~400~~45~0-~6~00-~5~50-~6~00-~6~~~~7~00--~
a. 0
750 350 400 450 500 550 BOO 650 700 750
Wavelength (nm) Wavelength (nm)

(a) Quantum efficiency (b) Responsitivity

Figure 2.18: The spectral response of the Thorn EMI 9924 photomultiplier tube.

The exposure meter uses around 1 to 1.5% of the light exiting the fibre feed. After
accounting for atmospheric transmission, telescope efficiency, telescope to spectrograph
coupling efficiency and PMT responsitivity, the exposure meter photocathode current will
be as shown in Figure 2.19. The integral of these curves gives the total photocathode
current. The maximum and minimum photocathode currents expected are shown in
Figure 2.20. This calculation shows that the photocathode current can be expected to
vary anywhere from 5 x 10- 11 A to 1 X 10- 16 A depending on the stellar magnitude, spectral
type and atmospheric seeing. Therefore, with a gain of G = 2.1 X 10 6 , the anode current
will range from 100 mA to 0.2 nA. These values are expected to enable the exposure meter
to perform adequately on stellar exposures down to mv = 10 and perhaps slightly fainter
in ideal conditions.
2.1. Design 63

X 10- 14
4~----,-----,-----,-----,-----,-----,---~c===~
- 80
- AO
- FO
--. - GO
T"""
- KO
I MO
E3
c

--«
+-'
C
CD
~
~

~2
CD
"0
o
£.
ro
()
01
(5
£. Figure 2.19: The exposure
a.. meter photocathode current as
a function of wavelength for an
OL-----~-----L----~------~-----L-- __~_=~_ _~_ _ _ _~ mv = 0 star.
350 400 450 500 550 600 650 700 750
Wavelength (nm)

10- 10 .-----------r-------,-------,--------,---------,-------r----------=l

Figure 2.20: The exposure

meter photocathode current as
a function of stellar magnitude.
The upper line is for an BO
star while the lower line is for
10- 17 L - -_ _ _ _- - ' -_ _ _ _ _ _- ' - -_ _ _ _ _ _-'---_ _ _ _----'-_ _ _ _ _ _....L-_ _ _ _ _ _-'----_ _ _ _- - - ' an MO star.
-2 o 2 4 6 8 10 12
Stellar magnitude
64 Chapter 2. Design and performance of HERCULES

2.2 Performance
2.2.1 Efficiency predictions
When computing the efficiency of HERCULES, everything from the fibre feed (including
the effects of seeing and guiding), to the CCD detector will be considered.

Fibre feed and collimator

Apart from the fibre itself, the fibre feed also includes the losses due to the finite size, and
perhaps variable location, of the stellar seeing disk.

Guiding
Although the guiding is generally good and the exposure meter records essentially constant
flux during an exposure, it is occasionally observed that the flux varies dramatically (see
Figure 2.21). This could be due to either poor tracking of the telescope or to varying
seeing conditions (including the passage of clouds). To correct for these fluctuations it
is assumed that the maximum flux observed during an exposure represents the flux that
would be observed if the guiding were perfect. The correction (Tgui) is simply the ratio
between the integral of the observed flux and the maximum flux multiplied by the time
of the exposure. That is,
fct exp
j (t ) dt
T - ---'0"---:----:--:--:--_ (2.5)
gui - max(f(t))texp
where j(t) is the exposure meter flux and t exp is the length of the exposure.

2.5

(i) 2
C
::J 00
o
~1.5 0
~
u:

0.5
0.5

50 100 150 200 250 20 60 80 100 120 140

Exposure time (sees) Exposure time (sees)

Figure 2.21: Examples of exposure meter log files. On the left is a typical log file where the star was
keep well centred throughout an exposure. The example on the right might have been taken during the
passage of light to heavy cloud or the variation could be due to extremely poor guiding.
2.2. Performance 65

Seeing
The seeing disk of a star is assumed to be of the same form as that described by Mof-
fat (1969) to model the point spread function (PSF) of a star. Figure 2.22 shows the
throughput of the three HERCULES fibres in various seeing conditions.

- 100llm Figure 2.22: The transmission

. _. - 100llm with 50llm slit of the seeing disk through the
Ql
Ql
- _. 50llm HERCULES fibres. The entrance
I- en 0.8 of the 100/-lm fibres is 4.50" and
c the entrance of the 50/-lm fibre is
o
'00 2.25". One of the 100 p,m fibres
C/)

'E
C/)
06
•
has a 50 p,m slit on the exit face.
c It is apparent that the 100/-lm
.....~ fibre with micro-slit has superior
~
C/) 0.4 throughput to that of the 50 p,m
'5 fibre only when the seeing is worse
0)
c than 1.8/1. If the seeing is better
'(i)
Q)
0.2 than this then the 50 p,m fibre
(J)
will allow substantially more light
through to the spectrograph for
o~--~----~----~----~----~~----~ high-resolution observations.
23456
Seeing fwhm (arcsecs)

Focal reducer
A (micro-)lens made from BK7 glass is used as a focal reducer. The front surface of the
lens has an extremely efficient multilayer anti-reflection coating applied by Fisba Optik
(see Figure 2.23), and the rear surface has been cemented directly to the fibre entrance
with an index matched cement. A coupling efficiency of 99% is assumed. The average
overall transmission of the focal reducer is therefore greater than 98%.

E Ql
U
o
1-.....
c
o
'00
en
'E
:g 0.98
.....~
cr;
a 0.97 Figure 2.23: The transmis-
2 sion of the single layer MgF 2
~ coating applied by Fisba Optik
a
en 0.96
cQ)
---- --------------------------------- on the microlens. For compari-
son the Fresnel reflection losses
"2Q 0.95 for the uncoated surface are
~ also shown (dashed line).
400 450 500 550 600 650 700 750 800 850
Wavelength A
66 Chapter 2. Design and performance of HERCULES

Fibres
The fibres used in HERCULES are Ceramoptec step index fibres, with high OR content
for enhanced ultra-violet transmission (see Section 2.1.3). Each fibre has a length of
22m which has an absorption shown in Figure 2.11. The transmission through the fibre
output glass/air interface can be calculated using the Fresnel laws of reflection. The total
transmission of the HERCULES fibre is shown in Figure 2.24 .

.n
1--:;:: 0.95
c
o
'w
CI)

'E
CI)
c
-to
.....
~
.0 0.8
:;:::
:§ Figure 2.24: The total trans-
o 0.75 mission of the HERCULES fibre.
I-
The mean transmission of the
0.7 fibre is 90% and the maximum
is 94% at ,\ = 770 nm.
400 450 500 550 600 650 700 750 800 850
Wavelength A

Collimator
The focal ratio of the collimator was chosen so that 95% of the beam emerging from the
fibre is captured. The mirror was coated with uv-enhanced overcoated silver by Laser-
dyne 6 (see Figure 2.25). The obstruction of the fibre-feed assembly has been estimated
at approximately 1%.

0)
c
~ 0.95
o
c.:>
0)
<C
"'0 0.9
Q)
c.:>
C
to
-§ 0.85
Q)
Q)
c
>. Figure 2.25: The reflectivity
"E 0.8
Q)
CI) of Laserdyne's uv-enhanced
to
....J silver mirror coating .
400 450 500 550 600 650 700 750 800 850
Wavelength A

6Laserdyne Pty Ltd, Queensland, Australia.

2.2. Performance 67

Fibre feed and collimator summary

The total throughput of the fibre feed and collimator can now be calculated. This is
shown in Figure 2.26.

1-$ 0.7
c:
'wo
(J)

'E
(J)
c:
~
......
....
o
1ti
.~
(5
()
Figure 2.26: The total
"0 throughput of the HERCULES
c:
co fibre feed and collimator.
~ The calculations assume good
..D
u:: 0.45 guiding and median (2.5//)
seeing.
400 450 500 550 600 650 700 750 800 850
Wavelength A

Echelle grating and prism

Echelle grating
The B = 210 mm collimated beam forms an elliptical footprint on the echelle grating
which has a major axis of B / cos a = 545 mm. This extends beyond the length of the
grating (408 mm) giving an overfilling in this direction of 14.5%. The grating is also
overfilled across its width (204 mm) by 0.6%. Also, because of the dispersion by the
first pass through the prism, the elliptical footprint is not always centred on the echelle
grating. However this effect is small and can be ignored. The echelle grating and elliptical
footprint are shown in Figure 2.27. The total echelle overfilling is 15.1%.

Figure 2.27: The

100 ........... - -~~ overfilling of the HER-
~~~ -~~
. .
. .- .
~~ ~
CULES echelle grating.
... ...
E 50 ,, ...
...
The collimated beam
g I
I ... forms an elliptical
\
C
0
'';:::;
0
I
I ,
I footprint on the echelle
'00 \ I grating which has a
0
c..
... , ... , I
I
major axis .of 545 mm.
>- -50
... ... ,,
.. .. .
"' ..... The overfilled portion
.... ......
~
~ .. ~
~~
of the beam is 15.1%
-100 of the total beam's
area.
-250 -200 -150 -100 -50 0 50 100 150 200 250
X position (mm)
68 Chapter 2. Design and performance of HERCULES

The diffractive efficency of the HERCULES echelle grating has been calculated using
the method outlined in Section 1.2.11. The energy distribution across one order is shown
in Figure 2.28. It can be seen that the peak theoretical efficiency at the blaze wavelength
for this order is 80.8%.

0.9
~ Figure 2.28: The diffractive
1-'5 0.8
efficiency of the HERCULES
>. 0.7 echelle grating. The blaze func-
u
c tion is plotted for a single order
Q)
0.6
'(3
m = 87, centred on the blaze
-
'+=
Q)
Q)
>
0.5 wavelength AB = 654.8nm.
The free spectral range for this
:;::; 0.4
u order spans the wavelengths
co
....
tI= 0.3 within the dashed lines. The
i:S efficiency at the blaze peak
0.2
is 80.8%. Compare this fig-
0.1 ure with Figure 1.22, with
0 cos (3/ cos 0: = 1.24.
648 650 652 654 656 658 660 662
Wavelength A

The refiective coating applied to the echelle grating is standard aluminium. The
combination of the refiectivity of aluminium, the overfilling by the incident beam, and
the diffractive efficiency gives the echelle efficiency. This is shown in Figure 2.29.

0.65
.c
()
I-Q)
>.
u
cQ) Figure 2.29: The efficiency

-
'(3
'+=

~
Q)

ill
at the blaze centre of the
HERCULES echelle grating. The
broad variation in efficiency as
..c a function of wavelength is due
u
W
primarily to the reflectivity
0.55 of aluminium. The high fre-
quency variation is the result
of numerical errors.
400 450 500 550 600 650 700 750 800 850
Wavelength A
2.2. Performance 69

Prisms
Both refractive surfaces of the prism have been anti-reflection coated with a single layer of
MgF 2 . The layers are optimized for an angle of incidence of (Jj = 40.0° (see Figure 2.30),
although the incident angles deviate slightly from this after several refractions. The mean
path length through the prism (in one direction) is 128 mm. The total throughput of the
double-pass prism is shown in Figure 2.31.

.~ 0.99
'iii
(f)

'E(f)
c
~
.....
:? 0.97
1ao Figure 2.30: The trans-
(.) mission of the anti-reflection
(]) 0.96 coating applied by Laserdyne
c
>-
---- -- -- - - - - - - -- - - - - - - -- - - - - - - -- - - - - - - -
'E to the prism. The coating has
(])
(f) 0.95 been optimized for an angle of
C1l
-l incidence OJ = 40°.
400 450 500 550 600 650 700 750 800 850
Wavelength Iv (nm)

]
E
11
I- 0.95
c
o
'iii
(/)

'E
(f)
c
.....~ Figure 2.31: The total
E
(f) efficiency of the HERCULES
·c
Cl. prism. The absorption from
Cij two passes through the prism
.....
o
I- and the reflection losses from
0.7 four air-glass interfaces have
been included.
400 450 500 550 600 650 700 750 800 850
Wavelength Iv (nm)
70 Chapter 2. Design and performance of HERCULES

Echelle grating and prism summary

The combined efficiency of the HERCULES prisms and echelle grating is shown in Figure
2.32.

0.56
0..
I-.gl 0.54
>- 0.52
<.l
cQ)
'0 0.5
:E
Q)
Q)
>
'iii
....
Q)
c..
(f)
'6

-
til
0
f- Figure 2.32: The total
efficiency of the HERCULES
004
dispersive elements.
400 450 500 550 600 650 700 750 800 850
Wavelength A (nm)

Camera
The HERCULES camera is a folded Schmidt camera (see Section 2.1.2). Both the fold mirror
and the primary mirror have been coated with Laserdyne's uv-enhanced overcoated silver
(identical to that applied to the collimator; see Figure 2.25). The corrector and field-
flattening lenses both have single layer MgF2 anti-reflection coatings. The coating was
applied by Laserdyne, and the transmission is shown in Figure 2.33. The thickness of the
corrector plate is 15.0 mm and the mean ray path length through the field-flattening lens
is 10.5mm .
.--..
°0
II
~
.2:'
'>
.iii
0.99
(f)

'E(f)
c

~
- ~
g' 0.97
Figure 2.33: The transmis-
o
<.l sion of the Laserdyne single
----- - -- - - - - - - -- - - - - - - -- - - - -- - -- - - - - ---
Q) 0.96 layer MgF2 anti-reflection
C
>-
"0 overcoat. The dashed line
ill shows the Fresnel losses of an
0.95
~
-' uncoa.ted surfa.ce.
400 450 500 550 600 650 700 750 800 850
Wavelength A (nm)
2.2. Performance 71

The main source of vignetting in the HERCULES camera is the hole in the fold mirror
through which the rays from the primary mirror are focused (see Figure 2.5). Due to the
fact that the fold mirror is not at the entrance pupil of the camera the amount of vignetting
is field-dependent, where the field angle is a function of order number and wavelength (i.e,
the echelle and cross-dispersion). The effect of beam anamorphism (due to the echelle and
prism dispersion) and the variable path length through the prism complicate the exact
camera vignetting. Both of these effects are readily calculable and the results are shown
in Figures 2.34 and 2.35.

0.95 .'
.2'
f->
c
0.9
0
t5
,g 0.85
· ' . ' . , . i · · · ·..i ..·..· ,/
"0
Figure 2.34: The HERCULES
~ 0.8 " .. ' ,. camera vignetting function
2 ""., ......... .,'
(j)
.0 across an order. The orders
0
c
0.75 .,.,,' """" ........ .,' ,/,/'
~ shown are centred on blaze
:J r----~
wavelengths AB = 380, 490,
"'''' '1 _.,., ",,;
,

0.7 ~.'~.,~___
...~
... ~"..-."-."--~~~~~~~~ --- 150
........... - .. ____ ------......... - 122 610, and 880nm. These are
0.65 93 orders m = 150, 122, 93, and
65 65 respectively.
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Relative order number I::.. m

-AB
0.9 ___ AB +/- 0.5l::..A ~ .. '
fsr
......
1-}'O.85 , ..
," .
c
0
.. .. .
nell 0.8
.;::
"0
OJ
t5 0.75 ~ ~
~~

Figure 2.35: The HER-

2
.. . ' "
~

---
(j)
.0
........... ~,
CULES camera vignetting
0
c
0.7
function at all wavelengths.
:J
r The unvignetted fraction
has been calculated for the
0.65 blaze wavelength and for
A = AB ± !:::.Afsr /2.
400 450 500 550 600 650 700 750 800 850
Wavelength A (nm)

From Figure 2.34 it can be seen that for a majority of wavelengths that lie within
one half of a free spectral range of the blaze wavelength, the vignetting is approximately
constant. This is also shown in Figure 2.35. This is because the dispersed beam at these
wavelengths invariably covers most of the obstruction. The slight decrease in obstruction
across the central part of an order is due to anamorphic magnification, which changes the
relative size of the obstruction. At high dispersion angles the beam becomes displaced
from the obstruction and the unvignetted fraction of rays increases rapidly. This com-
pensates to some extent the decreased blaze efficiency of these wavelengths. The total
72 Chapter 2. Design and performance of HERCULES

efficiency of the HERCULES camera is shown in Figure 2.36.

........ - -------
... - ....... ......... -
~~~

~ ~ ~
E
1-" '" 0.7
e ;~"
o
'c;;
Cf)

'E 0.6r
Cf) I
,'" -------------------
e I
J
jg I

~ I'
Q) I

~ 0.5 "
() I
I Figure 2.36: The total
,
I
efficiency of the HERCULES
0.4 camera. The efficiency at
A = AB ± IJ..Afsr / 2 is also shown.
400 450 500 550 600 650 700 750 800 850
Wavelength A (nm)

CCD
The quantum efficiency (TQE ) of the SITe SI003AB CCD for the HERCULES Series 200 de-
tector is shown in Figure 2.37. When calculating the total efficiency of the OOD the
absorption by an overcoated thin silica window must also be considered.

0.8

0.75

Ql
0.7
1-0"
>. 0.65
()
e
.~ 0.6
:E
Q)
E 0.55
::l
~ 0.5
::l
a Figure 2.37: The quantum
0.45 efficiency of the SITe sI003AB
0.4 COD. The detector has been
overcoated for enhanced uv
0.35 tr ansparency.
400 450 500 550 600 650 700 750 800 850
Wavelength I, (nm)

Spectrograph summary
The total efficiency of HERCULES (Tspc ), which includes everything from the transmission
of the stellar seeing disk through the fibre input to the quantum efficiency of the CCD, is
shown in Figure 2.38. The maximum efficiency in I" seeing is predicted to be just over
20% at a blaze wavelength near /\ = 650 nm. In median seeing conditions (around 2.5/1)
2.2. Performance 73

0.2 -1"
.-.- 2
0.18 ..... .-'-'-, ... ' .... , ....... - _. 3
" .. ,_, .. , ... , ...
g, 0.16 .. .... 4
I-tI) - 5"
>- 0.14
c ---_ ....... -_ .... , ..... ... ... .-.- 6
Q)
.... ...
..,
'5
a=
0.12
," -. ..
Q)
.c
..... .
... ...
.. ..
0.1
c.. '
• ,,,,,,,,,,,,,,.' """""""'" "'111,
",
.. .. ...
e
~

t5
0.08
, , "'/,
...." .. ",./.,
'"
'"
". '" ".
... ..
. ...
...
Q) ". ".
c.. ,'~'~'_'_'_'_'~'-'_'_'_'_'-'-'~'" II" •• ,

C/)
' ..... ' ........... , ".
", .. ,
",,, .... ,

o 400 450 500 550 600 650 700 750 800 850
Wavelength 'A (nm)

Figure 2.38: The total efficiency of the HERCULES spectrograph, including the fibre-feed and the quan-
tum efficiency of the ccd. A 100 J.1.m fibre is assumed and the throughput has been calculated for a range
of seeing conditions.

the spectrograph throughput is expected to be approximately 12%. These efficiencies

compare well with other fibre-fed spectrographs, a few of which will be discussed below.
The predicted throughput of HERCULES is slightly lower than the FEROS fibre-fed spec-
trograph. The system efficiency of FEROS is predicted to be 17% (including the telescope)
(Kaufer et al., 1999) at a resolving power of R = 48 000. Recently, this spectrograph,
had a measured efficiency of approximately 19.5% when coupled to the 2.2 m MPG/ESO
telescope 7 . The HERCULES efficiency is comparable with the HARPS instrument on the
La Silla 3.6 m telescope. Including slit losses, and CCD quantum efficiency, the peak in-
strument efficiency of HARPS is 8.5% for a resolving power of R ~ 100,000 (Mayor et al.,
2003). HERCULES is expected to have somewhat better performance than FOCES for the
2.2 m or 3.5 m Calar Alto Observatory telescopes. At a resolving power of R = 40000 this
instrument was expected to have a throughput of 13% (from the fibre input to the CCD) in
the best seeing conditions (Pfeiffer et al., 1998). The efficiency of ELODIE on the Observa-
toire de Haute-Provence 1.93 m telescope is less than HERCULES. At a resolving power of
R = 42 000 the throughput of the spectrograph is 4.2% (neglecting seeing losses and CCD
quantum efficiency) (Baranne et al., 1996). The Hale 5-m telescope fibre-fed echelle had a
detective-quantum-efficiency of 1.5% at R = 40000 (Libbrecht and Peri, 1995). The early
performance of AFOE was significantly worse. This fibre-fed spectrograph was coupled
to the F. L. VVhipple 1.5 m Tillinghast telescope, and had a measured efficiency of 0.2%
(Brown et al., 1994) for a resolving power of R = 51000. This instrument was seriously
degraded by poor fibre coupling and low CCD quantum efficiency and it is expected that
the performance has since improved considerably.

7http://www.ls.eso.org/lasilla/sciops/2p2/E2p2M/FEROS/TechnicalReports/SN/index.html
74 Chapter 2. Design and performance of HERCULES

2.2.2 Signal to noise predictions

The prediction of the signal-to-noise ratio (SIN) requires that the following quantities
are known:

N number of photons per second per resolution element from object,

ND number of photons per second per resolution element from dark current,
NB number of photons per second per resolution element from sky,
Re rms readout noise per pixel,
nslt number of pixels per resolution element
t exposure time.

That is, the signal to noise ratio can be defined as:

SIN = Signal = Nt (2.6)

Noise J(N + ND + NB)t + -JniliRe

It must be noted that this is the definition for a single resolution element. However, other
definitions may be used (e.g. per extracted pixel) in which case nslt must also change.
The photon flux N can be calculated directly from a known spectrophotometric standard
assuming that the spectrograph's detective quantum efficiency Tdqe is known. Given that

(2.7)

where Tatm , ltel and Tspc are the atmosphere transmission, telescope and spectrograph
efficiencies respectively, it remains to calculate the atmospheric extinction and telescope
efficiency.

Atmospheric extinction and telescope efficiency

The transmission of the local atmosphere can be computed from the extinction coefficient
k. That is,
-k "K
T atm = e 1.086./ (2.8)
where X is the airmass of the object being observed. The following measurements of the
atmospheric extinction at MJUO were made by A. Gilmore on 2001 August 8:

k(V) 0.124
k(B - V) 0.099
k(U - B) 0.260
k(V - R) 0.033
k(V - I) 0.065

These have been used to compute the atmospheric extinction shown in Figure 2.39.
The mirrors of the McLellan telescope are both coated with standard aluminium. The
secondary mirror obstructs 8% of the on-axis rays and therefore Tte1(vig) = 0.91. The total
transmission of the McLellan telescope is predicted to average around 68% across the
wavelengths used by HERCULES.
2.2. Performance 75

0.95

~
f-. 0.9
~
c:
<D
~ 0.85
en
c
~
.2 0.8
Ci3
.J::
Q.
8
E
;;(
Figure 2.39: Atmospheric
extinction over MJUO for an
airmass of X = 1.
~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Wavelength Ie (nm)

Exposure times
It is also possible to compute the exposure time required to reach a given S / N. The
following is a simple Poisson based method for doing this. First we make the following
definitions:

RS/N SIN
NT N +ND+NB

Hence we have
(2.9)

where t is the exposure time required to reach a signal-to-noise ratio R S / N . This expression
may be solved to give,

t = RS/N [p ± v' p2 - 4nN2 R2] (2.10)

2N2 e

where
(2.11)

Signal-to-noise summary
The results of the above calculations have been used to predict the signal-to-noise (S / N)
expected as a function of exposure time and stellar magnitude. The predictions are shown
in Figure 2.40. On this basis it is predicted that HERCULES will obtain a signal-to-noise
ratio of 100/1 in 60 seconds for a GO star with a magnitude mv = 4.7. This assumes
atmospheric seeing of 2.5", an airmass of X = 1.3, and a resolving power of R = 40000.
By way of comparison, under operating similar conditions, FEROS is expected to reach
mv = 6.7 for SIN = 100/1 in 60 seconds on the 2.2m telescope, although SIN = 100/1
would be reached for an mv = 5.1 star in 60 seconds if this instrument were coupled to
aIm telescopes. vVith HERCULES the limiting magnitude of a star of the same spectral
8http://www.lsw.uni-heidelberg.de/cgi-bin/exp-calc.cgi
76 Chapter 2. Design and performance of HERCULES

type for a signal-to-noise of 100/1 should be mv = 7.1 in 10 minutes, and mv = 9.1 in

one hour. On this basis it is also expected that HERCULES should achieve S / N = 10/1 in
half an hour for an mv = 12.7 solar type star, or mv = 13.5 in one hour. However, due
to the limitations of the guide camera, observations of such faint stars are not currently
possible.

"" "" - 60s

" "" ---- 600s
" ...... 1800s
" ""
3
10 " "" - - 3600s
"
---.z
(j)
"
""
"
""

""
"
---
0 "
""
"
"
~ 10 2
" ""
Q) " "-
C/l "-
'0 " "- "- Figure 2.40: HERCULES
c " " "- signal-to-noise (SIN) pre-
"- "-
0
..... "- "-
"- " dictions . The predictions
" " "-
~0> 10 1 "
" "-
are for a GO star at a
wavelength of 550nm.
U5
The SIN is for each
"extracted" pixel at a
resolution of R = 40 ODD.
0 The atmospheric seeing is
10
assumed to be 2.5" at an
0 2 4 6 8 10 12 14 airmass of X = 1.3.
Visual magnitude (m )
v

2.2.3 Efficiency measurements

Observations were made of the spectrophotometric standards given in Table 2.4. The
majority of the observations were made by the author. However, several stars were also
observed by D. Ramm and J. Skuljan. Spectrophotometric data for most ofthe stars have
been taken from either Alekseeva et al. (1997) or from Breger (1976). However, the star
31 Aql was measured using the standard spectral type calibration data of Knyazeva and
Kharitonov (1996).
The spectra of the above spectrophotometric stars were extracted and wavelength-
calibrated. The flux (i.e., the count rate per extracted pixel) was measured at the order
centre nearest each wavelength for which spectrophotometric data exist. The absolute
efficiency of HERCULES was computed by comparing this measured flux with the flux
predicted for stars of the same spectral type and visual magnitude. The results are shown
in Figure 2.4l.
It can be seen that there is significant variation in the measured efficiency of HERCULES
and that the maximum predicted efficiency is never reached. A significant portion of this
variation is possibly due to widely varying atmospheric seeing conditions. Another cause
of variation is likely to be due to inconsistent guiding errors. The effect of guiding can
partially be removed by examining each observation's exposure meter log file (see Figure
2.~1). A guide correction, which is the ratio between the integral of the observed flux
and the maximum flux multiplied by the time of the exposure, is applied to each of the
2.2. Performance 77

HRno. HD no. Name R.A. Dec. V mag. Spec. Rot. vel No. Ref.
Type (km/s) obs.
100 2262 Ii; Phe 0:26:12.2 -43:40:48 3.94 A7V 219 5 P
126 2884 (31 Tuc 0:31:32.7 -62:57:29 4.37 B9V 173 11 B
472 10144 a Eri 1:37:42.9 -57:14:12 0.46 B3Vpe 251 10 P
591 12311 a Hyi 1:58:46.2 -61:34:11 2.86 FOV 153 11 B
674 14228 ¢ Eri 2:16:30.6 -51:30:44 3.56 B8V-IV 247 10 B
705 15008 oHyi 2:21:44.9 -68:39:34 4.09 A3V 163 20 B
919 18978 11 7 3 Eri 3:02:23.5 -23:37:28 4.09 A4IV 144 7 B
1084 22049 f. Eri 3:32:55 -09:27:30 3.73 K2V 5 P
2020 39060 (3 Pic 5:47:17.1 -51:03:59 3.85 A5V 139 33 P
2361 45813 A CMa 6:28:10.1 -32:34:48 4.48 B4V 135 3 P
2451 47670 v Pup 6:37:45.7 -43:11:46 3.17 B8Ill 228 9 P
3165 66811 ( Pup 8:03:35.1 -40:00:12 2.25 05f 211 2 P
3685 80007 (3 Car 9:13:12.0 -69:43:02 1.68 A2IV 133 1 P
5132 118716 f. Cen 13:39:53.2 -53:27:59 2.30 BlllI 159 4 P
5708 136504 f. Lup 15:22:40.9 -44:41:22 3.37 B2IV-V 133 17 P
5812 139365 407 Lib 15:38:39.4 -29:46:40 3.66 B2.5V 149 8 P
5953 143275 70 Sco 16:00:20.0 -22:37:18 2.32 BO.3IV 181 2 B
5993 144470 90 w1 Sco 16:06:48.4 -20:40:09 3.96 B1V 142 1 P
7373 182572 31 Aql 19:24:58 +11:56:40 5.16 G8IV 31 K
8425 209952 a Gru 22:08:14.0 -46:57:40 1.74 B7IV 236 2 P
References: P = Alekseeva 1997, B = Breger 1976, K = Knyazeva 1996.

Table 2.4: Selected spectrophotometric standards used to measure the absolute efficiency of HERCULES.

measurements. The "guide corrected" measurements of HERCULES efficiency are shown

in Figure 2.42.
Even after correcting for possible guiding errors alarge variation in efficiency remains.
The fact that the maximum efficiency of HERCULES is nearly obtained suggests that all the
poor efficiency measurements were due to especially poor seeing and/or uncorrected guide
errors. In Figure 2.43 all the measurements have been divided by the predicted efficiency
and normalized about the mean. It can be seen that the same general trend is apparent
for all measurements. That is, the efficiency appears to be deficient between 450 nm
and 550 nm and there is a possibility that HERCULES is more efficient than predicted at
wavelengths greater than 550 nm.
78 Chapter 2. Design and performance of HERCULES

20 /' -------- ........

--------- .........
"-
'\
/' '\
/ -----~--., \ / 1"
/
/
../
~~----.--/' "\ \/
/ /
\ /
2"
I /
I / \/
/

---
~
..........
15
/ /
/ /
------- -- ------;-!..;-;. ....
:.': : ' .
1/'// II ••• :I!!.!~ _> 3"
:>, // ,/ 1",:1. '11\./
C,) / / • • : I . I I •• v
i:: / / / . : :. • I' I ! I ! .
.....C,)
Q) / // / : •• I : i I !I •• : •• : I I •
I ' • W. . . . _-.--.--.-- •
..... 10 /
/ / ./ / /
// /'
_-----t-!.i!r.
I .. I I;
I : · i : ' I: ............ ,
. :: : : '\ / 4"
.....
..... / / / / .:1.11 I .1'1'1::' :11'e
!::il I // ././ •: • I ill: : I : : • : • i :. ". I ••
1/ / / : . . . . 1.lj-Il'II"·;·, .!..o.JJ.~·_-;..!..t-L.LI-L.t· •
./ / ••••. 4 .-'-::----..!-- ••• ,:,1 . • • • • • •
/1
1./ / • 1..1'-'-:-',1 11. •"'-.'
./ / .: -rll' • • .!' I "
I , • I •
1I I , I I I ' I "
.,,'./ - .- 5"
/
// /
,.,/
,....... U·'11" I I:····
---.:.'. .:r.'
~I'
• I ·'.4-H
. .:. ..
..... L •-• •.. I-"~ -
.
I .:" I I • :
-
•• '.
--.1
•
• •
•

/ / ...... ;,••r.":1>1: I . : ' _--T,r •. a- - -

/
' -t- H ·'
•• ,.......-, ,', ' , ' .! i I '
• • ..........
,,,.......;-~ 6"
5 II·(".-;I.I.I·.:..:. ...... II:;!..!~.:II dlald:'
••••
.:' . . . . . . . :'
• •• : :.
I.'
/
/ /a.:':":,,~.
.t ...... •
.......fi·IIII·'
-, t III II
, :::111
• I"
II::'
I I •
I • • : •
•• '.
•
•
1'..Jn·, =-=--. . . . . . . . . . . :Ii··· " ' I ' : ' . ' : ••
a.I:I.· •• •• •• I" . ' '.'
. ..
,!'"r •...:,-:1.

H.'
"'•• ""-;Y' •• ::,:::'.1
~.{··I.;: ... ••
'..
,:;'" ····U::::·:···· .: •. : .• •.
•••••• I .11'1!,:
• •••••• I'!
'.,I II! I iii
•••••
••. ,:.:" : •••...
.. . . ,
'1':": ", • .::: : : : : : :. ::::::: ••
o
400 500 600 700
Wavelength (nm)

Figure 2.41: The measured efficiency of HERCULES .

.
-------...-- ............ ,
.-- ---
.
"-
20 //
------~ '\
\
/ ..- - . - - .!- - - - - -......... • \ /
1"
/ / ../
-~----~./ "\ \/
/ / \ /' 2"
I / • • \ /
/ / , : /
/ /
.---. 15 I ' : • ___' . ' •
~ / / / I - ______ - ' . ' .- I : - :I - I -: ~~I'"I
'-"
/ / . /
- , 1
! I'
. 'I ~1 \ ' / - . 3"
.: .,
:>, / / /
C,) / / • 'to
i:: / // / ."
Q) / / / . !: I I • .. • ..! •
...... / / / ••••• .:' I 1-t-!.:-i-'--;---:::-;-I-.. . . . , •.
.....
C,)
10 / / . ' . - - - - , - , ........ : ' , : 1 " . , 1 : . , •. = ....... '
4"
.....
..... /"
/
/
/
/
/
'",</
I
'.I.! ; : I
_,.:'':.
' •.I I ' : •
,: 8 "
;I!, • ~'
I,:'! ,,/
, / • , I I : : I 1'1 I ; , I • •• : i ' i ..
!::il
/ /
/ //
/' /
,./
•••) Y • •I I ··.j"·11
' ! ' 2., • I I ' ".L I '1 , :.-!-:...-; ,'--..L'
~
I
:.:.'.
- . - . -'-.
~
. - - . . ,.
' •

/1./
./ /' :':ov-'.~I:!:.'}.""'I!!I
/
/
.. ' / ,I . '- .I ". " "I,i:• I••. i: •.. , : : . . . .• .• , i • i l.J..t-I- ~ :-; - . - ;--1 --t..:,
I •••• , ••• I "
• .'
•
.1.•
I
""
,./ -
'
p
5"
/ /' .I~ • • " " " " •• I • I. t , ___ -'- _--_a.-.--I.-
/ / .......- '::1, J-!" I :: Il-tr',-I-I-L I-' L -+- - ,I- ~ I· .. • • • • • • '~. ___ • • •

1 / I · . ........... •••. ,'" I ' ,0'• •,1:.

............ ....... p' .f-'~Ii
-- .• ' ,111,11::1.
1 ,....... • '.---- 6"
5
/ I..J'';-'-'';''!':':'-/
/
••••••••.--;
i
'IM'
u·.?·...- " •••••/'1
li
I 'I'
'i.,,·I:.II:·
I."
.
• "
....
• • • • •
......
•••• , '.
'• . n'
•

.1·" : : .: •• , : . III

••.,.,.:::..1'
•• ••••• __
•",".:1.;
~

• ..... -..-
h! .... ,,11
••• Ii· ••••••
: ' ,.,
. . . . . ~.----
••.••
. .••
I
.

:::::::! .....
•
I .• :.;· ••.
, •••• , ••••••
i·.···.
'.:11 &
' .:·:·······
......
•
,- ' I"':!: -'•
a'
I •
...
1111'1
.'
1
I' ' I II I • I I
.: •••••••
: I
.'
•
...
, • ' :
.. I I

1.: 1:....... • '::::::: •••••• ,.

o
400 500 600 700
Wavelength (nm)

Figure 2.42: The "guide corrected" efficiency of HERCULES. See text for details.
2.2. Performance 79

I I I I I I I I I I I I I I I I I I

1.5 I- .,I , I

i ' I I, I I
-

',I I:' :!
I I
, I
,
",1 ! I.
-
,i1 I' !
I- :' '. • II ..
I
I
: ' I ' •• -
» .:

r :. ':." -.. : .!••: It...11 .. ; I :

-.
I''
I

II II ' '
• =.
-
• J " • "j:tI'1
'''II . ,." !'ill .I " .... I
1f--: ill',.
..

, '! i: ":'1
11

r ,.,~!: . : 1'1, I : ' : '

II·1 : 1
.1 6
.".". ••• I
....... I:.. .... .
r I··.. II .. ,. ,11·• t , .....
. . . II•
-
j~) ~. t
.. It II
e. .. :
,....
.. • ..
,.
"..
.. •
=: ! : 1"1 I I"
•

1-;':'
I-
'1111111
I! II
!il!lIllil! I,
-

-
-.... . ,III' ,. •
.:
-

I I I I I I I I I I I I I I I

400 500 600 700

Wavelength (nrn)

Figure 2.43: Relative efficiency of HERCULES. The efficiencies shown in Figure 2.42 have been divided
by the predicted efficiency and then normalized about the mean for each observation.
80 Chapter 2. Design and performance of HEROULES

2.2.4 Environmental stability

The HERCULES vacuum tank was sealed in 2001 April and later evacuated in 2001 Septem-
ber. Since this time the instrument has remained unaltered, except for detector replace-
ment and repositioning.

Pressure and temperature

The pressure inside the HERCULES tank over a period of 2i years is shown in Figure 2.44.
The vacuum is reestablished whenever the pressure rises above approximately 5 mmHg.
It can be seen that the rise of pressure after each pump-down is quite steady and on
average the vacuum is reestablished every 130 days. It is interesting to note that the rate
of decay of the vacuum has slightly reduced over time. Initially the vacuum decayed at
around 0.035 mmHg/d. However, toward the end of the 2003, the rate of decay became
0.027mmHg/d. This is probably due to reduced out-gassing ofthe components inside the
vacuum tank, which has not been opened since 2001 April.

2001 2002 2003

Oct Jan Apr Jul Oct Jan Apr Jul Oct

5 I!J)
0
0
0
0
0

0
<D
ID
.
0
0
0
0
0
0
0

0
0
0

!Ill
0
0
0

0 0 0 0
0 0 CJD
0 0 0
0 0 0
0 0 0
t;D4 0 !Ill
<Ill
[]) 0 ..,.0 0
0
0 0 0 m
::c: 0
0 ID
[]l 0
0 0
0 (;

8 0 0
0
0 0

-
III 0
8 0
0
!Ill
0
0
0
([l)
0 0 ClIl 0

-
3
-
Q)
0 0 <Ill 0 0
0 0 0 <ID 0 0
r... 0 0
ID
0
0 0
;:l 0 0
VI 0 0 ID 0
VI
Q)
r...
0.. 2 [])

0
0

(II)
0
0
0

0
0
0
0
0
0
IIIIl
CJD
.
0
0
0
<Ill

0 (JIJ
0 0 0 0
0 0

0 0
0
0
0

0 0
""
0
0
0
0 0
0
0 0 0 (!D 0 0
1 0 0
0
0 0
0 0
0
0
0

2200 2400 2600 2800

JD - 2450000

Figure 2.44: The pressure of the HERCULES vacuum tank.

The temperature of the spectrograph is monitored in several locations throughout the

structure. The temperature of the mid-section for the past 2 years is shown in Figure
2.45. The room in which HERCULES is located is not temperature-controlled, and this is
reflected by the clear seasonal variation of the temperature. The considerable short-term
scatter of temperatures is real. This can be seen in Figure 2.46, where the temperatures
over 40 days in 2002 Feb to Mar are plotted.
2.2. Performance 81

2001 2002 2003

Oct Jan Apr Jul Oct Jan Apr Jul Oct
I I I I

20 I- -

a 15
QJ

.....
I- -
«l
I-.
QJ
0..
eQJ
E-

10- -

I I
2200 2400 2600 2800
JD - 2450000

Figure 2.45: The temperatures inside HERCULES. The short-term scatter is real (see Figure 2.46).

19
,---....
u ~g
0.0
OJ
'"d o
§
~
~8 .0 ,~'\ ~
\~ @,~o
'-" 0
18
OJ § B
~
;j ~ U 0
\\0 9 00
...,...l
8
Cd
~
~ 0
8 ~

,·rt
OJ 17 0
0... 0 8
8 8
S
OJ
E-< ~\ \~ 0

16

2330 2340 2350 2360

JD - 2450000

Figure 2.46: The temperatures inside HERCULES during 2002 Feb-Mar.

82 Chapter 2. Design and performance of HEROULES

Mechanical stability
The mechanical stability of HERCULES is assessed by observing the positional stability
of the thorium-argon spectra. The relative positions of selected lines of thorium-argon
spectra have been measured by D. Ramm for the period from 2001 July to 2004 Jan
and are shown in Figures 2.47 and 2.48. It is apparent that the thorium images are

2001 2002 2003 Figure 2.47: The x-shift

JuJ Oct Jan Apr JuJ Oct Apr JuJ Oct
Jan
of thorium lines (data
15 " from D Ramm).
10 11, o
5
x
:e, 0
c:
o
E -5 .
'"o
?- -10
><
-15

-20
-25

-30

2100 2200 2300 2400 2500 2600 2700 2800 2900 3000
JD -2450000
I

I JuJ
2001
Apr
2002
JuJ Oct Apr
2003
Jui Oct
I I Figure 2.48: The y-shift
30~
OCI Jan Jan
0
of thorium lines (data
25 from D Ramm).
20
15 6' o
g
.3 10
c:
0 5 .,
~ 0 o o
0 o to
0-
J
>- -5
0
\) ., Q
Of!'
0

-10
-15
-20
-25
-30

2100 2200 2300 2400 2500 2600 2700 2800 2900 3000
JD - 2450000

far from stable. The abrupt shift in 2001 September occurred when the CCD cradle
was repositioned in order to correct the tilt of the focal plane. The other significant
change which occurs throughout 2002 December to about 2003 April is more difficult to
explain; however it is noted that the cradle was dismantled during this period while ad-
justments to the CCD focus mechanism were being made. During the period 2001 October
to 2002 November the shift (at least in x) was quite gradual at around 0.02 pixels/day (or
0.5 J.Lm /day), although there was considerable scatter from night to night.
No attempt can be made to disentangle the relative stability of the various optical
components using these data. Most significantly, it is not possible to discern whether the
motions observed are due to the physical shifting of optical elements within HERCULES or
simply to a gradual shift in the location of the CCD cradle. It is most likely that the latter
2.2. Performance 83

plays a significant role in the cumulative shift, as it must be realised that the CCD dewar
is frequently removed from the cradle. The CCD dewar is quite heavy and in practice the
relocation is never a smooth operation. The observer inevitably knocks the CCD cradle
and a shift of half a micron (the average daily shift) is therefore quite probable. That the
shift (during 2001 October to 2002 November) is generally in the same direction reinforces
this speculation, as the CCD dewar is always relocated from the same direction. It should
also be noted that the short-term stability is much better. There is a discernible shift
throughout each night, which is likely to be due to evaporation of the CCD dewar's liquid
nitrogen coolant. The shift is highly correlated for every thorium line, but it can be as
much as 0.01 pixels per hour. This shift equates to around 3-5 mls during 10 to 15 minute
exposures. However, the effect on radial velocity precision will be considerably less than
this if the dispersion solution is interpolated between two (or more) thoriums.

Radial velocity precision

HERCULES has proven to be capable of excellent radial-velocity precision. Ramm et al.
2004 have made observations of the 7th magnitude double-lined spectroscopic binary
HD 181958 and have obtained an RMS of 50 ms- I . A radial velocity standard (HD 45067)
observed as part of this program had a scatter of 20 mls from a total of 76 measure-
ments. Another (single-lined) spectroscopic binary (TrA had its very nearly circular
orbit determined with high precision. An RMS scatter of 14 ms- I was determined from
200 observations after iterative fitting to orbital parameters was performed (Skuljan et al.,
2004). A dedicated reduction software package called the HERCULES Reduction Software
Package (HRSP) has been written by J. Skujlan for the purpose of obtaining precise radial
velocities (see op. cit. for details). Ramm and Skujlan have observed there is no sig-
nificant difference between the short-term and long-term precision of HERCULES (private
communication). It has also been observed by these authors that day-time sky spectra can
achieve an RMS of 2 ms- I over the short-term; however a much larger scatter is observed
over longer periods of time. This scatter is assumed to be due to atmospheric winds, but
could also be the result of variable atmospheric transmission across the face of the Sun
as suggested by Brown et al. (1994).
84 Chapter 2. Design and performance of HERCULES

2.3 HERCULES in the future

Several possibilities exist for improving the performance of HERCULES. The first, and
most obvious of these, is a larger format CCD. Other improvements would include a more
flexible fibre feed and collimator arrangement. The use of image slicers could improve
throughput and fibre double scrambling would improve the radial velocity precision. The
final upgrade possibility would be an improved guide and acquisition camera. Each of
these possibilities is discussed below.

2.3.1 CCD
As described above, HERCULES was originally designed for a 2k x 2k CCD with 24 f.-Lm
square pixels. At the time of construction a Series 200 CCD with 1024 x 1024 pixels each
24 J.Lm square was the only detector available to be used with HERCULES. It is therefore
not possible to observe the entire spectral range simultaneously. In order to observe all
of the spectral format a detector cradle was constructed which has four discrete CCD
positions (see Section 2.1.2). The design was intended to cover the spectral regions shown
in Figure 2.9. However, during assembly the spectral format was slightly altered to give a
more centrally located CCD position which is better suited to precise radial velocities. The
CCD positions currently available with HERCULES are shown in Figure 2.10. Note that it
is currently impossible to observe spectra above 720 nm and the lower limit of 370 nm is
a result of the fall off in instrument efficiency (i.e., fibres, mirrors and CCD).
It appears unlikely that a 2k x 2k CCD with 24/lm pixels can be acquired for HER-
CULES. Vlhen HERCULES was being designed Scientific Imaging Technologies (sITe 9 ) were
producing 2k x 2k CCDS with 24J.Lm pixels. However, these have since been discontinued.
Without considering mosaicking options (e.g. a mosaic of four lk square CCDS) other
possible CCD formats for HERCULES include:
• 2048 x 4096 with 15-f.-Lm pixels: These chips are available from E2v 10 . The area
of such a chip (30.7 x 61.4mm) is still not sufficient to fully cover the HERCULES
spectral format, and several positions would still be required as well as a possible
rotation (see Figure 2.49).
• 2048 x 4608 with 13.5-J.Lm pixels: This chip is also available from E2v. The imaging
area is 27.6 x 62.2 mm. Although the smaller pixels would improve the sampling of
the smallest resolution element, the previous chip is obviously a better choice.
• 4096 x 4096 with 15 f.-Lm pixels: Fairchild Imagingll manufacture these chips. The
imaging area (61.4 x 61.4 mm) is well matched to the available spectral format of
HERCULES (see Figure 2.50). The CCD would require a single position only.
The CCDS mentioned above are available in "scientific" grades of the highest quality, and
can be thinned, back-illuminated and overcoated for high quantum efficiency. The theo-
retical efficiency of HERCULES with an upgraded CCD is shown in Figure 2.51. Although
the primary motivation for an upgraded CCD is increased spectral coverage, either of the
above-mentioned detectors will double the detective quantum efficiency of HERCULES at
380 nm.
9http://www.site-inc.com
lOhttp://e2vtechnologies.com
11 http://www.fairchildimaaging.com
 ~

~
Figure 2.49: The HERCULES
::x:
Order As l1y spectral format showing the pos- trl
il:!

30 59 ,
., ~
I
------- Py -.I,; '96 5 ,.
17.9" #
sible locations of a single 2k x
4k CCD with 15/-lm pixels. Note
0
eto<
trl
(JJ

63
67 ;
904
85~
, " .3"
16.9"
that a fourth horizontal centrally
located position is also probably
5'
g.
(1)
71 ~'
102 16.8"
... ~
I; 759
desirable. S'
16.9"
2"
20
,
""1
'I ~~-
Pd-
-.:::
\
-- Gil ~~) ~
721 17.1" (1)

83 \ ° 2(B) , I 686 17.4"

All Lil
87 - --.;
\
~~ '4fC
Hel
- Nil
SII
Ha
SII
~II ,-. I 655 17.8"
626 18.2"
600 18.8"
19.3"
10 99 75
103 53 20"
,-...,
107 32 20.6"
E
E
..........
111 513 21 .3"

.-
- - --
C 115 495 22.1"
0
+=i 0 119
I 1 479 22.8"
'00
0 123
- 463 23.6"
0... I Ba
I 127 _MgII 449 24.4"
>- 131
I
FFMI(~) 435 25.3"
G"T(
-10 135
H(G) I
422 26.2"

C"ItT,
139 ~ (O , l)~ 410 27.1"

.- •.•
He
143 ' ,~, 398 28"
" ~

Cali f)
- 20 147 CN(l, eN(o,~f8 29"

.-
Hel 'w 'v . I
'v
151 Hl0 3n 30"
~
.
~

Oil
155 • • 368 31 "
J
•
- 30 159 _ t _ t _ ,--&iii"", _t_
- I
_1_1_ ,_1_1_ 1. 358 32.1 "

FSR
FWHM
- 30 - 20 -10 0 10 20 30
X- position (mm) 00
ell
 00
O'l

Figure 2.50: The HERCULES

Order spectral format with a 4k x 4k
CCD with 15 pm pixels. Full
30 ti Py
spectral coverage from 360 nm to
59 , , ~ '965 17.~ #

lJJ
63 , ., 904 " .3" 1 pm. is possible in a single ex-
67 .', 850., ~ 16.9" posure.
71 lJ(J2 16.8"
, 759 16.9"
20 79 '-~~ - _ _ ell ·- ~ - __,-~ 721 17.1"
83 \ \ °2(B) I , 686 17.4"
\ - Hel All _."11 Lil ?II SII
87 1____ ' - ,, -~- NI HeY ~'I ' _ 17.8"
626 18.2"
600 18.8"
99 75 19.3"
10
103 ::>53 20"

E 107 e 32 20.6"

E 111 513 21.3"

-----
c
0
:.t=' 0
115

119
, , liB , I liB
495

479
22.1"

22.8"
·w
0 123 'Jill 14111
, Ilil
PJIII pml
I
463 23.6"
0.. Barr
I 127 MglI 449 24.4"
>- 131 435 25.3"
-10 135 422 26.2"
0
P"
."
C1"Il' [PI'I
"d
c+
139 'v ~ (O,1)~ 410 27.1" (l)
"1

Ilel I ;t!!1 1 ~
Het~
143
Gall(lE} 8'8:M~I~, I '
Ilel 398 28"
tj
.:..
-20 147 CN(1 , t ~ PJeIlIGPJ(1 ,1) GPJ(9,9) ~ 1181 , 29"
(l)
CIJ
PJel1l ::J(1,1) Qq'
, ,,
Ilel 119 ' ,
Hel : IIHl 1119 I lei ::l
151 H10 30" ."

155
: 011
I
011

! ,
Oil
368 31 "
::l
0-
"d
(l)
: I I
"1
0'
• •
I
- 30 159 358 32.1 "
"1
S
."
FSR ::l
n
(l)

FWHM 0
.....,
::r:
trl
-30 -20 -10 0 10 20 30 ::0
0
C!
X-position (mm) t"
trl
rJJ
 Figure 2.51: The HERCULES
Order spectral format with a 4k x 4k
CCD with 15 J.Lm pixels. The ef-

1 ;'rillllllll.j~'96 17.~ #
30 ficiency (in %) of HERCULES (in
63 , .~ 904 " .3" median seeing at R = 40 000)
67
71 .. '
" 850
1/02
., ~ 16.9"
16.8"
is shown by the contour lines .
.. ~ , 75 16.9"
A Fairchild CCD with a broad-
20 band overcoat is assumed. Wave-
~ \" \ ~\ \ i. ,. ...., ~
83 ',\""
87
~
\
~.
\
\_
j
~\ 6.. \ \
~;:"" W \. I
\\
\ .,.
1,,\'!
-'-!\ I
t
;:/ / r /" ...,----:2--1Q--Tf." \
f!
I
r
/' I
J- /
""\
,
\
~
\ I
\
I
I
\
~ I
;
I
' --i L -r'
J
.... " '
f, /
i, ; ,
72
686
65
17.1"
17.4"
17.8"
lengths that are vignetted are
not plotted.
62 18.2"
600 18.8"
99 19.3"
10
103 20"

E 107 20.6"

--
E
c
o
:;::; o
111

115
51

49
21.3"

22.1"

Ow 119 47 22.8"

o 123 46 23.6"
0.
I 127 44 24.4"
>- 131 43 25.3"
-10 135 42 26.2"

139 I 410 27.1"

143 • 39 28"
L ./ /

-20 147 38 29"

. · - -,.,
OJ

151 377 30"

• ·• " .
·
./
155
. "
./

.. a/ 36 31"

-30 159 ." I

./
/
I

FSR
35 32.1"

FWHM
-30 -20 -10 0 10 20 30
X-position (mm)
88 Chapter 2. Design and performance of HERC ULES

2.3.2 Collimator and fibre feed

While the current HERCULES collimator and fibre feed system allow for very stable and
relatively efficient observations, it is impossible to implement any other fibre feed modes
apart from the three currently in use. An alternative collimator design is proposed which
will remove such limitations and hence allow for a variety of fibre feed modes to be
implemented. The essential feature of this upgrade is that the spectrograph's entrance
slit is placed outside the main HERCULES vacuum tank. Any number of collimator designs
could achieve this. The collimator must be off-axis, thereby eliminating at least one source
of light loss (the obstruction due to the fibre feed exit). It will also be folded so that the
entrance slit can be conveniently placed, and hence will require at least one or possibly
two additional mirror reflections. The entire collimator should be able to be placed within
the fixed section of the HERCULES vacuum chamber. This limits the positioning of the
entrance slit and vacuum tank window. A suggested design is shown in Figure 2.52. The
design is a Dall-Kirkham which uses an off-axis ellipsoidal primary mirror and a spherical
secondary mirror. The third mirror folds the focus at right angles to the collimator's
optical axis. The effective focal ratio is shown as f/15. However, this could be readily
altered to suit various fore-optics. It would also probably be necessary to provide both
an intermediate focus (where a slit mask could be placed) and to pass a (near-) collimated
beam through the vacuum entrance window in order to minimize possible flexure issues.
The bare fibres currently used could therefore be placed inside the vacuum tank at the
location of the intermediate slit.

Entrance
slit /vacuum window
Secondary .-----I~

mirror
Primary
mirror

Figure 2.52 : The pro-

posed upgraded collima-
tor is an off-axis Dall-
Kirkham design. See text
for details.

This alternative collimator design would allow the use of fibre image slicers similar
to those being considered for use on the SALT HRS (see Section 3.2.2). These could be
used to improve the efficiency at the highest resolving powers, especially during times of
poorer seeing. A high resolution mode of nearly R = 10 5 would become available with
the combination of fibre image slicing and improved detector sampling. It remains to be
shown definitively that the improved efficiency and/or resolving power justifies the new
collimator design and use of image slicers.
The use of an iodine cell (for improved radial velocities) would be possible in the
collimated space before the vacuum window. However, the radial velocity precision would
2.3. HERCULES in the future 89

probably be most significantly improved (without the associated losses of an iodine cell)
through the use of a fibre double-scrambler (Brown, 1990). It should be noted that the
use of a double-scrambler does not require an upgraded collimator design and a double-
scrambled fibre feed mode could be implemented immediately alongside the currently
available modes.

2.3.3 Guiding
At present it is not possible continuously to guide on the faintest stars which HERCULES is
capable of observing. There is however an intermittent guide mode available which uses a
fully reflective fold mirror, but this mode does not ensure the stability ofthe stars position
on the fibre entrance and is also inefficient. A more efficient guide camera is required to
improve the performance of HERCULES near the faint star limit. That is, a more sensitive
detector is required to replace the image-intensified camera that is currently used. Several
commercial possibilities exist which have improved quantum efficiency and are capable of
on-chip binning and/or longer exposure times. The guide camera optics would probably
need modification in order to accommodate a different detector. A method for more
reliably centring the object onto the fibre entrance should also be investigated. Currently
it is not possible to view the fibre entrance from the guide camera, and it is probable
that the mis-centring of an object is one of the largest contributors to the radial velocity
precision of HERCULES.

2.3.4 Mechanical stability

It was shown above that the temperature stability of HERCULES is not adequate. While
it is not obvious that temperature changes significantly affect the radial-velocity precision
(due to the presence of the vacuum) it would be prudent to reduce the large seasonal
temperature fluctuations. A simple air-conditioning scheme could reduce the amplitude
of the temperature changes by an order of magnitude; however care should be taken to
avoid periodic changes in temperature.
It was also seen above that the repositioning of the CCD appears to be causing the
slow shift of wavelengths. The upgraded CCD proposed, if it were dedicated to HERCULES,
would entirely eliminate this effect. The CCD being considered would use a closed-cycle
cryo-cooler, which would also eliminate the short-term flexure caused by the evaporation
of the liquid coolant.

2.3.5 Efficiency

As discussed above, the use of fibre image slicers will enhance the throughput of HER-
CULES at the highest resolving powers, however the lower resolving powers will remain
unchanged. The upgraded CCD would however enhance the efficiency at all wavelengths.
The enhancement in the blue will be particularly noticeable. Three other options for in-
creasing the efficiency of HERCULES at all wavelengths and resolving powers are as follows:

l. Improved fibres. The FBP fibres from Polymicro have improved uv transmission
(although they are not significantly better than CeramOptec's Optran uv fibres),
and they do not display the OH dip at 720 nm.
90 Chapter 2. Design and performance of HERCULES

2. Telescope secondary mirror coating. The change from standard aluminium to a

high efficiency (uv-enhanced) coating such as Laserdynes overcoated silver would
improve the efficiency by as much as 10% at all wavelengths.

3. Adaptive optics. The implementation of adaptive optics would be the most signifi-
cant of any improvement. Halving the median seeing at MJUO from 3"to 1.5"would
improve the throughput of HERCULES by over 25% (see Figure 2.22). It is also pos-
sible that the ability to reliably maintain the centring of an object on the fibre input
will improve the radial velocity precision. A program is underway at the University
of Canterbury to develop an adaptive optics system for use at MJUO (Mohr et al.,
2004).

It should be noted that while some of the suggested improvements (such as the use of fibre
double-scramblers) would degrade the throughput of HERCULES in order to improve other
functions (e.g., radial velocity precision), this degradation would be largely mitigated if
all (or even some) of the above efficiency upgrade options are implemented.

2.4 Summary
The design of HERCULES has been discussed in detail. The vacuum mounted, R2 echelle
grating, and prism double-pass instrument, has proved to have excellent throughput and
radial velocity stability when fibre-fed by the MJUO l-m telescope. Future upgrades,
including a larger format CCD and/or improved fibre feed and guiding will significantly
improve the performance of HERCULES.
While HERCULES has been designed for use on our l-m telescope, the spectrograph
would be capable of excellent performance on any 2 to 4-m telescope located at a site with
reasonable seeing conditions. HERCULES would also be capable of performing competi-
tively on even larger telescopes at sites with world.,class seeing conditions. In the following
chapter the design of a spectrograph for a 10-metre class telescope will be discussed.
Chapter 3

The design of SALT HRS

3.1 Introduction
The following sections detail a time series of concept and detailed optical design documents
for a high resolution spectrograph (HRS) for the Southern African Large Telescope (SALT).
Except where acknowledged in the text, the optical design work was done substantially by
the author. However the dioptric cameras for the initial R4 design are primarily the work
of D. Jones, with significant input by the author. The SALT HRS Principal Investigator
(P. L. Cottrell) and Project Scientist (M. D. Albrow) provided advice and assistance on
some aspects, most notably Albrow on the fibre feed arrangements. J. B. Hearnshaw,
G. M. Kershaw and P. J. Macqueen also provided advice.
Details of SALT are given in Section 3.1.1 and a summary of the fibre feed options
is given in Section 3.2. Section 3.3 details the development of an R2 HRS design which
culminated in a design which was presented during a Preliminary Design Review (PDR) on
2003 September 4. The optical design document is included in Appendix D. Section 3.4
describes subsequent R4 designs as a response to the 2003 September PDR. This resulted
in the detailed development of a design which was presented at a second PDR in 2004 July.
This optical design document is also included as an appendix (Appendix E). Both of the
optical design appendices form the core of a more extensive series of documents provided
for the SALT consortium.

3.1.1 SALT

The Southern African Large Telescope (SALT) is located in Sutherland in the Northern
Cape, South Africa. It is situated at an altitude of 1798 m above mean sea level. The
telescope (Figure 3.1) has a spherical ll-metre primary mirror comprised of 91 identical
hexagonal segments, which is mounted at a fixed altitude of 37° from vertical. Details
of the optical design are shown in Figure 3.2. A 4-mirror reflective spherical aberration
corrector (SAC) provides a science field of view of 8 arcmin over a declination range from
_75° to 10°. The telescope is able to rotate in azimuth between observations and during
observations objects are tracked by the moveable SAC. During an exposure the illumina-
tion of the SALT entrance pupil will vary (see Figure 3.3). The length of time an object
can be tracked depends upon the object's altitude. This varies from around 2 hours to
about 45 minutes. The operational wavelength range is from 320 nm to 2500 nm. Further
details of the telescope can be found in Stobie et al. (2000), Swat et al. (2003) and on the
web l .

Ihttp://www.salt.ac.za/

91
92 Chapter 3. The design of SALT HRS

Figure 3.1: The SALT

telescope.
3.1. Introduction 93

(a) (b)

/ Ptimary mirrof (M I)

Spherical Aberration
Corrector (SAC)

Figure 3.2: The SALT telescope (a) and detail (b) of the spherical aberration corrector (SAC).

Figure 3.3: The variable illumination of the SALT entrance pupil. The figure is actually of the RET
telescope.
94 Chapter 3. The design of SALT HRS

The parameters of the SALT telescope that have been used throughout the design of
the SALT HRS are given in Table 3.1. In median conditions at Sutherland the seeing is
FWHM = 0.9" (Buckley, 1995) and Erasmus (1999) gives quartile values of 0.74, 0.92 and
1.16" respectively for the distribution. When added in quadrature with the expected
image quality of the SALT optics the total SALT image quality error budget predicts that
the encircled energies EE(50) and EE(80) will be

EE(50) = 1.29" and EE(80) = 2.15"

during median seeing conditions at Sutherland (Swart, 2001) (see Figure 3.4 for details).
The entrance aperture throughput as a function of fibre diameter and seeing conditions
is shown in Figure 3.5. Throughout the rest of this document all calculations of fibre
throughput will be made assuming median seeing conditions.

Parameter Specification
Primary mirror diameter (D) 11.0m
Focal ratio 1/4.2
Effective focal length 46.2m
Image scale 224 fLm / arcsec Table 3.1: SALT parameters.

SALT total:
EEo(50) EEo(80)
1.293 2.150

I
I
SALT performance: Seeing:
EE/50) EE,(80) EE,(50) EE 2(80)
0.590 0.992 1.151 1.908

I
Optical petformance: Dome/facility seeing: Payload positioning:
EE,/50) EE,/80) EE,,(50) EE'2(80) EE,K50) EE,i80)
0.497 0.839 0.200 0.371 0.247 0.377

I
I I I I
Primary minor: Tracker payload: Design residual: Tracker: Structure:
EE,.,..(50) EEI.I.,(80) E£"2(50) EE"l80) E£,.,(50) EE"l80) EE'3,(50) EEl' ,(80) EE 132(50) EE'3,(80)
0.436 0.733 0.230 0.398 0.062 0.091 0.225 0.345 0.100 0.153

Figure 3.4: The SALT optical error budget. The site median seeing has been converted to EE(50) and
EE(80) for a zenith angle of 37° and all calculations assume a wavelength of 633 nm. The calculations
were made by G. Swart (2001).
3,1. Introduction 95

100 .......... - .. - - - - - - - - - - • 0,5"

,, .
.... .. "

, ,
80 ,,
,
;g I
~ I
I
c:
0 60 I
I

'iii I Figure 3.5: Fibre entrance aperture

C/l I
'E I
I efficiencies for SALT, The bold line
C/l
c:
40
I shows the throughput assuming that the
~ I
I-
I stellar PSF is added in quadrature with
the expected SALT image quality. The
dashed line shows the throughput of the
PSF alone, The stellar PSF is a Moffat
function as described by Racine (1996)
and references therein.
200 300 400 500 600
Fibre diameter (/Am)

The expected throughput of SALT is as follows:

340 nm <). < 450 nm: >49%
450 nm <). < 800 nm: >53% and
800 nm <). < 2500 nm: >60%
where the total pupil obscuration due to the SAC structure vignetting and gaps between
the mirrors is no more than 25% (Buckley, 2000). The throughput of the telescope
has been estimated using witness measurements from the four SAC mirrors and a single
aluminium surface and is shown in Figure 3.6. Included in the total efficiency is a fifth
fold mirror (assumed to have the reflectivity of the average of the four SAC mirrors) and
the vignetting due to to the SAC. The efficiency of the ADC is not included. Its efficiency
is expected to be greater that 95%.

68 Figure 3.6: The trans-

mission of SALT. Included
66 are the efficiencies from
the four SAC mirrors, the
primary mirror, and the
SAC vignetting. A fifth
§ 62
'(ij
(Jl
fold mirror is also as-
'E 60 sumed. The ADC was not
(Jl
c: included. (After Buckley
~
.:: 58 et al. 2004.)
~
f- 56

54

52
400 500 600 700 800 900 1000
Wavelength (nm)
96 Chapter 3. The design of SALT HRS

Prime Focus Imaging

Spectrograph (PFIS)

Mounting ring on tracker he xap 0 d

I Fibre feed
3.2 m Imaging and acquisition camera
.......- - - - i (SALTICAM)
PFIS slit-viewing mirrors

Atmospheric Dispersion
Corrector (ADC)
Moving pupil
baffle
Spherical Aberration Corrector (SAC) I

Figure 3.7: The SALT prime focus payload. The payload includes the spherical aberration corrector (SAO)
and the atmospheric dispersion corrector (ADO). The location of the prime focus imaging spectrograph
(PFIS) [ref) and the imaging and acquisition camera (SALTICAM) [ref) is also indicated.

3.1.2 Fibre Instrument Feed

The location of the fibre instrument feed (FIF) in the prime focus payload is shown in
Figure 3.7 and a schematic diagram of the proposed instrument is shown in Figure 3.8.
The FIF has been designed by Buckley and Sessions and details can be found in Buckley
and Sessions, 2004. The FIF will accommodate up to 12 individual fibres which are
mounted in two rows of 6 fibres each. The two rows can be moved apart along a set of
rails. A second pair of rails provides motion in the orthogonal direction. This allows a
pair of fibres accurately to be centred and to be separated anywhere from approximately
15/1 to 3.7'. Five pairs of fibres are available for use by SALT HRS. The input telecentric
angle varies considerably as a function of field angle (see Figure 3.9) and therefore in order
to limit the effects this will have on the fibre's output focal ratio degradation the useable
field for SALT HRS fibres will be limited to ±1'.
The SALT imaging camera (SALTICAM) will be used for initial acquisition of targets,
where fiducials will be defined by peaking up the SALT HRS signal (using an exposure
meter) on bright stars. A guidance stage is mounted directly to the FIF stage which
contains a coherent fibre bundle. Again, fiducials will be predefined for each fibre input
and guiding will continue on a suitable nearby field star. Given the SALT offset pointing
specification of 0.1/1, and a FIF fibre repositioning specification of 27 11m (i.e., 0.12/1) RMS,
this will allow targets to remain accurately centred throughout an exposure. Provision
has not been made for directly viewing the fibre input.
3.1. Introduction 97

Figure 3.8: Model of the

SALT fibre instrument feed
(FIF). Up to 6 pairs of
fibres can be accommo-
dated. The pairs of fibres
is shown at their minimum
separation. (From Buck-
ley and Sessions, 2004.)

5
4.5 ../
~
1P 4
./'
:2. 3.5
ib ./'
.;j 3 ./'
.g 2.5 ~
'6 2 ./'
~ 1.5
~ /""
0.5 ./'
o ./'
o 0.5 1 1.5 2 2.5 3 3.5 4 Figure 3.9: The telecentric angle at
Field Angle [aroninutel the SALT focal plane as a function of
field angle. (From Buckley and Ses-
sions, 2004.)
98 Chapter 3. The design of SALT HRS

3.1.3 SAC calibration optics

A calibration system designed by Swat and Esterhuyse (see Buckley et. al 2004 or Meiring
and Buckley, 2004) is able to mimic the illumination of the SALT pupil (i.e., Figure
3.3). This will ensure that the calibration source is injected into the spectrograph in
a manner which is as nearly identical to the stellar source as is possible. The system is
still under design; however a schematic of one possible concept is shown in Figure 3.10.
The calibration optics can be moved into the beam either immediately prior to or after
an observation, or during the daytime.
4 - -_ _ _ _ _ _ _ ;]5E'lMN

J<1l o:n=l=-u sa;:; J:l

SCArTbR~O R~VS RR~ NOT s~UN

Figure 3.10: A possible SALT prime focus calibration system. The calibration light exits an 8mm
diameter fibre bundle.

3.1.4 HRS science requirements

SALT currently has two first light instruments under construction, an imaging camera
(SALTICAM, see O'Donoghue et al. 2003) and a medium-resolution imaging spectrograph
(PFIS, see Nordsieck et al., 2001). PFIS will cover resolving powers 500 < R < 10000 with
a wavelength range 320 < .A < 900 nm. The proposed HRS complements PFIS by providing
resolving powers 17000 < R < 85000 and wavelength coverage 370 < .A < 890 nm.
SALT HRS will address many of the fundamental questions that drove the construction
of SALT. A list of more than 50 science drivers for SALT HRS arose from a potential user
survey carried out by G. Mackie in late 2000 and early 2001 (Hearnshaw et al., 2001).
The broad categories of science to be addressed by this instrument will be:

48 element abundance studies in local group galaxies;

• extra-solar planet detection;

• stellar internal structure and dynamics;

.. star cluster and galaxy dynamics;

.. outflow and accretion studies;

• high- and moderate-redshift galaxy spectroscopy.

3.1. Introduction 99

Some of the desired capabilities of SALT HRS are:

• resolving power R = 30000 to 70000;

• wavelength coverage ).. = 380-880 nm;

• high mechanical and thermal stability;

• some limited multi-object capability.

SALT HRS is designed to be competitive with other high resolution spectrographs on large
telescopes (see Table 3.2).

THE LIBRARY
UNIVERSITY OF CANTERBURY
CHRISTCHURCH, N.Z.
Telescope: VLT Keck Subaru HET Gemini S. LBT
....0
0
Spectrograph: UVES HIRES HDS HRS bHROS PEPSId
D teI 804m 10m 8.2m 9.2m 804m 2x8Am
Wavelength 300-500 420-1100 320-1100 320-1100 420-1100 400-1000 390-580 580-1050
range (nm):
Echelle: R4 R4 R2.8 R2.9 R4 R2 R4
gv/mm: 41.59 31.6 52.6 31.6 31.6 87.0 31.6

Beam 200 200 305 305 260 200 200

size (mm):
Resolution- 41400 38700 39000 38000 30000 21000 40000
slit product:
Maximum 80000 110000 67000 165000 120000 150000 120000 120000
Resolving power
Pupil 1 1 n/a n/a 1 n/a 1.5 1.5
mag.:
Cross- SR grating SR grating SR grating SR grating Prism VPH grisms
dispersion 1000gv/mm 600gv/mm ?? gv/mm 400gv/mm 600gv/mm 2 x silica 9401/mm 4601/mm
660gv/mm 312gv/mm 250gv/mm 250gv/mm 316gv/mm 60°
Camera: dioptric dioptric catadioptric catadioptric dioptric reflecting dioptric dioptric
1/1.8 1/2.5 1/1.0 1/0.96 1/2.8 1/0.96 1/ 2.3 1/ 2.3

CCD: EEV EEV & MIT/LL SITe EEV Marconi Marconi

Format: 2k x 4k 2k x 4k 2k x 2k 2x (2k x 4k) 2k x 4k 2k x 4k 4k x 4k 4k x 4k
Pixels: 15J-Lm 15J-Lm 24J-Lm 13.5J-Lm 15J-Lm 15J-Lm 15J-Lm 15J-Lm

Max. A 85/126nm 200/403nm ~250nm ~400nm 380nm ~160nmc 660nm 0

coverage: i:l""
fj
<+
DQEa: 12% 14% 10% 13% ~7%b 12.6% co
'i

A 400nm 600nm 600nm 600nm 600nm 600nm :-"

References: Dekker et al. 2000 Vogt et al.1994 Noguchi et al. 2002 Tull1998 DArrigo et al.2000b e Pallavicini et al. 2003 f-j
i:l""
co
a. DQE'S are from top of telescope with "wide slit" 0.
co
to
b. HET HRS predicted efficiency aq.
P
c. bHROS coverage not continuous ...,
0
en
d. PEPSI parameters based on 8/1 order separation and complete wavelength coverage ;,.
e. The reference is actually for HROS ~
~
::0
en

Table 3.2: High resolution spectrographs on other large telescopes.

3.2. SALT HRS fibre feed 101

3.2 SALT HRS fibre feed

The various fibre feed modes that were investigated during the design of SALT HRS are
discussed below. It should be noted that while some options are particular to a spectro-
graph design, given, in particular, the inter-order separation, all of the options can readily
be adapted to any design. Section 3.2.1 introduces the proposed observing modes and in
Section 3.2.2 the options for improving throughput at the highest resolving powers are
discussed. M. Albrow provided significant input to the latter.

3.2.1 The fibre modes

Three fibre feed modes have been considered during the design of SALT HRS. Each of these
will be discussed briefly below.

"Fixed object and sky"

Because of the large SALT aperture it will be important to observe sky spectra along with
every object. In the "fixed object and sky" mode a pair of fibres will be used at the
telescope focal plane to observe the object along with a single patch of nearby sky. In
principle this same mode could be used simultaneously to observe a calibration source.
However, it may not be useful to do so, unless a third fibre is also used to capture sky.

"N od and shuffle"

The technique of "nod and shuffle" is widely used in infra-red spectroscopy for precise
subtraction of the sky background and has been proposed to increase the yield of high
density multi-object surveys (see (Cuillandre et al., 2003) and (Glazebrook and Bland-
Hawthorn, 2001)). It is also shown (op. cit.) that nod and shuffle spectroscopy would
allow much deeper observations on large telescopes such as SALT where the sky emission
is significant.
The nod and shuffle mode proposed for the SALT HRS is outlined schematically in
Figure 3.11. It is assumed that the two object spectra can be shuffled so that they
overlap in the focal plane, and therefore at the end of a nod and shuffle exposure the
spectrum will comprise an object order and two adjacent sky orders. Each of the sky
orders will have been captured by the respective fibres. It follows that the observations
of the wavelength calibration and flat-fielding sources must be obtained using the same
technique.
102 Chapter 3. The design of SALT HRS

Input Output CCD Figure 3.11: The nod

(i) A 0 Sky AUSky =----rC;'\)- - : Sky A and shuffle concept for

B • Object B • Object =__-AIII.r---=- Object B

fibre-fed spectrographs.
(i) Fibre "A" observes
the sky while fibre "B"
is located on the object.
I
-, \
I

'-'
I
(ii) The telescope is then
nodded so that fibre "A"
Nod Shuffle (up) captures the object while
fibre "B" now observes
, .... \ -------i"'"'~----
----- the sky. To avoid con-
_____ _ Sky A

=---.----:
(ii) 'I
I
'-'
I
------'-'---- fusing these spectra, the
charge on the OOD is
A . Object A . Object Object A shuffled upwards so that
the object order "A" is
BO Sky BOSky =--~C)\.- - : Sky B now in the position of
object order "B" in the
previous step. (iii) Next
Nod Shuffle (down) the telescope is nodded so
that the sky returns to
(iii) A@ Sky _--(r)'\.- - - : Sky A fibre "A" and the object

..----=
A@Sky
- to fibre "B". The charge
B . Object B • Object =
___----.1
_
•
-
Object A + B on the OOD is shuffled
down so that now the
I
-,
\
.............. ;v-'~ SkyB object spectra from fibres
-~-~-----
I I
'-' "A" and "B" overlap.

Multi-object capability
In order to satisfy the requirements of the nod and shuffle mode the inter-order spacing
must be sufficient to allow at least three objects to be observed at any resolving power.
Assuming this is possible with bare fibres at the lowest resolving powers then the spacing
between orders will be sufficient for the following multi-object possibilities:

1. R ~ 20000 (400-500 /-tm fibres) - a total of three 1.79-2.23/1 fibres;

2. R ~ 40000 (200/-tm fibres) - a total of six 0.89/1 fibres;

3. R ~ 80000 (100/-tm fibres) - a total of twelve 0.45/1 fibres.

Unless the SALT image quality is improved and/or adaptive optics are used, the high
resolution multi-object modes will be extremely inefficient. The multiplex gain is probably
not sufficient mitigation against this inefficiency and a multi-object mode would also
require a considerably more complicated FIF. The ability to compensate for the variation
in telecentric input angle is probably vital. A SALT high resolution multi-object mode is
not currently being considered.
3.2. SALT HRS fibre feed 103

3.2.2 Fibre slicing options

It has been shown already that if a single small diameter fibre is used to obtain the very
highest resolving powers then the transmission of the stellar PSF would be very small. A
number of fibre slicing options have been explored, and each of them is readily applicable
to any of the concept designs.

Micro-slits
In order to obtain a higher resolving power than is possible with a simple bare fibre a
smaller micro-slit can be imprinted on the fibre's output face. This option was explored
for the CELESTIA design which is discussed below (see Section 3.3.1) and has already
been used successfully on HERCULES (see Section 2.1.3). As was shown in Figure 3.5 the
throughput of the smallest fibres is extremely poor. If it is assumed that a 100 11m fibre
can be used to obtain a resolving power of R = 100000, then the optimal configuration
of a range of fibres and micro-slits and their throughputs in median seeing are given in
Table 3.3. While the throughput at the highest resolving power is quite poor, it is a
factor of 2.2 times better than using bare fibres. It was also shown that for resolving
powers less than R ~ 30,000 there is no advantage to using micro-slits. Otherwise, the
significant advantage of micro-slits is that they require no additional inter-order spacing
than is available at the lowest resolving power. Other methods of increasing the resolving
power involve some form of image slicing and are therefore only possible if the inter-order
separation is significantly greater. These options will be discussed below.

Fibre diam. Micro-slit Resolving power Throughput

(11 m ) width (11m) (A/SA) (%)
400 22900 67.8 Table 3.3: The optimal con-
350 200 37800 41.9 figuration of fibres and micro-
75500 21.9 slits and their throughputs in
300 100
median seeing. The through-
400 70 108000 13.9 put is purely geometric.

Fibre bundles
The concept of using fibre bundles is similar to integral field spectroscopy. However, for
spectroscopy of a point source, only in the spectral and not the spatial resolution is of
interest. A fibre bundle is used to sample a stars seeing disk and then the fibre exits are
reformatted to form the spectrograph's entrance slit. Each of the nod and shuffle and
fixed object plus sky modes requires different optimal fibre bundles. The order spacing in
the fixed object plus sky mode must be sufficient to place the sky from the same number
of fibres (which needn't be in a single input bundle) between adjacent object orders. The
optimal fibre arrangements for the fixed object plus sky and their geometric efficiencies
are given in Table 3.4 and the order profiles are shown in Figure 3.12. These profiles were
created by convolving a uniformly illuminated fibre exit with a PSF and then collapsing in
one-dimension. The PSF image quality is sufficient to just support the highest resolving
power shown. The nod and shuffle mode must allow for spacing for three objects per
order. The optimal fibre configurations (not shown) are therefore slightly different and
104 Chapter 3. The design of SALT HRS

the efficiencies are somewhat lower. All calculations assume a minimum order separation
of 13".

Fibre diam. Resolving No. of Efficiency

({tm) power fibres (%)
100 80000 14 58.4
200 40000 7 71.6 Table 3.4: Fibre bundle efficiencies for fixed
object mode. The bundles are arranged in the
500 16000 1 87.0 formats shown in Figure 3.12.

Fibre image slicer

The final solution to obtaining efficient high resolution spectra is to use an optical im-
age slicer. A design similar to that used on FEROS (Kaufer et al., 1999) is considered.
This is an adaptation of the classical Bowen-Walraven slicer (Bowen 1938, Walraven and
Walraven 1972) that allows two fibres to be sliced with identical optical path lengths. A
similar concept was used for the UVES image slicers (Dekker et al., 2003). It may also
be possible to use "focusing" image slicers similar to that described by Richardson et al.
(2000). Details for this type of image slicer are limited and no attempt to compare its
performance with the Bowen-Walraven type of image slicers will be made. As for the
fibre bundles, the throughput of the image slicer is limited by the amount of inter-order
spacing available to stack sliced images. Assuming a minimum order separation of 13",
the geometric throughput of image slicers as a function of resolving power is shown in
Figure 3.13. The slicer parameters are given in Tables 3.5 and the geometry for selected
slicers is shown in Figure 3.14. The parameters and geometry for nod and shuffle modes
are not shown.

Fibre diam. Resolving No. of Efficiency

({tm) power slices (%)
350 80000 4 54.0
500 38000 3 78.6 Table 3.5: Image slicer parameters for fixed
object plus sky mode. Figures by M. Albrow.
500 17000 1 87.0

Based on the geometrical throughput alone, the image slicer has slightly lower through-
put than fibre bundles at high resolving powers. However, it is possible that the packing
density of the 100 {tm fibres required to give the highest resolving power has been over-
estimated. It may be the case that 100 {tm diameter fibres require a 20% cladding rather
than 10% in order to retain high throughput in the near infrared (generally a cladding 8
to 10 times greater than the longest wavelength to be carried is desirable). If this extra
cladding is required, then the throughput of a 14 x 100 {tm bundle reduces from 58.4%
to 52.6%. It has also been suggested (R. Content, private communication) that the focal
ratio degradation of fibre bundles may be significantly larger due to the stress imposed by
gluing many fibres together at their input. The enhanced efficiency of the image slicers
at lower resolving powers may be further decreased by the efficiency of the image slicer
optics which hitherto have not been considered. The measured losses of the UVES image
slicers is 21-40%, while a 10% upper limit is measured for the FEROS image slicer (Kaufer,
1998).
3.2. SALT HRS fibre feed 105

1 .5 rr-ro..,.,,~-,.-r-r-r-r~-,--r->......,~'l Pixels
1 .0
o 20 40 60 80 100
1.0 ,,-~,.-r-~,.-,-~-,--.-~~~-.--~

Object
0.5 0.8
0.0 ~
c:
0.6
-0.5 ~ 0.4
-1.0 0.2
0.0 "----'----~___'__'_~L-...L~~~~~
-1.5 L.c.......----'--'~-'-'-'-~~~__'_'_'.~.w
-1.5-1.0-0.50.0 0.5 1.0 1.5 o 500 1000 1500
Arcseconds in focal plane Spacing on detector (Ilm)

(a) Fourteen 100 {-tID fibres (b) Fourteen 100 {-tID fibres

1.5 Pixels
1.0
o 20 40 60 80 100
1.0
Object
0.5 0.8
~
0.0 '(j) 0.6
c:
-0.5 ~ 0.4
-1.0 0.2 Sky
-1 .5 ~-'-'-'-~LLL..c-'--'--'-~'--'---'--'-'-'-'-J....LU
0.0 U-.-~~.I.-.--.c-~---'-'--"--'---"-'--''-'--'-'--'---'-'--.l...L-'-'
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 o 500 1000 1500
Arcseconds in focal plane Spacing on detector (Ilm)

(c) Seven 200 {-tID fibres (d) Seven 200 {-tID fibres

1. 5 ,---.--ro-.--,r~,...,...,.,-,-,-r~,.,...,..,-,-,-r--r-r-"r1 Pixels
o 20 40 60 80 100
1.0 1.0
Object
0.8

o
0.5
£' 0.6
0.0 t./)
c
-0.5 ~ 0.4
-1.0 0.2 Sky
0.0 '------"--'--'---'----"---'-'--~~'---'--'~"'------'----"------'--.___'____'__'
-1. 5 '--'--'--'-'-'-'-~'--'--'--'-'-'-'-~~--'--'---'-~
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 o 500 1000 1500
Arcseconds in focal plane Spacing on detector (Ilm)

(e) One 500 {-tID fibre (f) One 500 {-tID fibre

Figure 3.12: Fixed-object observing IDode inputs and outputs. Fibre bundle inputs for (a) 100 {-tID ,
(c) 200 {-tID and (e) 100 {-tID fibres. Black circles represent fibres and their claddings, red and blue circles
are EE(50) and EE(80) expected in IDedian seeing conditions. (b), (d) and (f) show the corresponding
output intensity profiles (one set for the object, one set for the sky) on the detector across a single order.
106 Chapter 3. The design of SALT HRS

100
~
!;2
'-...../
90
Figure 3.13: The geometrical
-+-'
throughput of a Bowen-
:J
D-
80 Walraven type image slicer as
...r: a function of resolving power.
0>
:J
0
70 ~. -, The throughput for fixed
object plus sky (solid line)
L
L

lr·",···. differs from the nod and shuffle

-+-'

E
:J
60
fl throughput (dashed line) owing
to the reduction in available
inter-order spacing for stacking
E 50 image slices. The steps in
x this function arise from the
0
:2 40
U"·' .... combination of discrete fibre
diameter sizes and the integral
number of slices. The chosen
30 modes are indicated. Calcula-
20000 40000 60000 80000 tions courtesy of M. Albrow.

Resolving power

w - 0.525 h 8.086 w 0.525 h 5.041

10

-2

-4

-L! -2 0 2 -4 -2 0 2 4
(mm) (mm}

Figure 3.14: The slice geometry for fixed object plus sky mode at high (left) and medium (right)
resolving powers. Figures by M. Albrow.
3.2. SALT HRS fibre feed 107

A schematic diagram of a proposed image slicer for SALT HRS and the output sliced
image are shown in Figure 3.15. The design is based on a suggestion by R. Bingham
(privat e communication). The slicer will require additional fore-optics in order to convert
the f /3.8 output of the fibres to the ~ f /20 required by the image slicer. This slow focal
ratio ensures that the defocus is kept to a minimum along the length of the sliced image
while also allowing the size of the image slicer to be scaled.

Figure 3.15 : Image slicer concept for SALT HRS (left). The output from a pair of sliced fibres is shown
on the right. The design is based on a suggestion by R . Bingham (private communication).
108 Chapter 3. The design of SALT HRS

3.3 R2 and R3 designs

3.3.1 CELESTIA optical design

An initial concept design for SALT hrs was presented to the SSWG in 2001 October (Hearn-
shaw et al., 2001). This instrument was referred to as the Canterbury Extremely Large
Echelle Spectrograph on a Telescope in Africa CELESTIA. The optical design of CELES-
TIA is shown in figure 3.16. CELESTIA is a fibre-fed spectrograph which uses a mosaic
of two echelle gratings, has two prisms used in double-pass, and an on-axis all spherical
catadioptric camera. The design assumes a CCD detector mosaic of two CCDS each with
2048 x 4096 15/-lm square pixels, although of course a single 4k x 4k detector would
be preferable. The parameters of individual components of the spectrograph, and some
motivation for the choice, will be outlined in the following sections. A summary of the
CELESTIA parameters can be found in Appendix A.2.2.

I
Primary
mirror

Figure 3.16: CELESTIA optical layout.

Echelle grating
In order for the HRS to be well matched to SALT image quality and to achieve the required
resolving powers, a large beam size is needed, which in turn means a large (mosaic) grating.
The choices of large echelle gratings (W > 200 mm) are quite limited and those available
from Richardson Grating Laboratory are listed in Table 3.6. We have not considered
custom gratings at all throughout the design of SALT HRS for reasons of financial constraint
and delivery lead times.
Initially a design which used a mosaic of two R3 (BB = 71.54° i.e, RGL grating no.
53.453) gratings with 31.6lines/mm was considered, but the small wavelength extent of
each order meant that a large number of orders would be needed to cover the visible spec-
trum. This would make the inter-order spacing intolerably small, especially if interleaving
sky or reference-object orders is to be considered.
3.3. R2 and R3 designs 109

Catalogue Ruling Density Blaze angle R# Ruled size (mm)

no. (lines/mm) BB (tan BB) w L
53.408 79 62° 1.9 210 411
53.401 79 63° 2.0 204 408
53.127 87 63° 2.0 308 413
53.451 316 63° 2.0 204 408
53.411 31.6 63.9° 2.0 204 408
53.121 110 64° 2.1 310 413
53.424 52.67 65° 2.1 204 408
53.417 52.67 69° 2.6 308 408
53.453 31.6 71° 2.9 308 408
53.414 31.6 76° 4.0 204 408
53.425 41.59 76° 4.0 204 410
53.113 94.13 79° 5.1 206 413

Table 3.6: Properties of the large echelle gratings available from Richardson Grating Laboratory RGL.

A grating that was felt to deliver a good compromise between (almost) complete
wavelength coverage and sufficient inter-order spacing was the R2.8 (e B = 70.45°) grating
with 52.67 lines/mm2 . A mosaic of two such gratings is required and they are to be
aligned with a 25 mm spacing. The echelle grating is illuminated at an angle of incidence
B = 2.75°which is a compromise between reasonable path lengths, feasible camera sizes
and echelle blaze efficiency.
The use of R4 gratings was not considered at this stage in the design, for reasons similar
to those given during the design of HERCULES (see Section 2.1.2). While a detailed design
was not attempted, it was realized that prism cross-dispersion would have to be abandoned
(because of the need for large angular dispersion); a white pupil design would probably
be necessary (to limit the size of the optics); and complete wavelength coverage could
only efficiently be obtained by splitting the instrument and using separate blue and red
cross-dispersion gratings. Two very fast, wide field, dioptric cameras would be required.

Collimator

The collimator is an on-axis paraboloid which has a focal length of 1160 mm which gives a
beam diameter of B = 305 mm. This assumes that the fibre input is approximately f / 4.2
and that the focal ratio degradation factor is ¢ = 1.10. That is, the output focal ratio is
f /3.8. The collimator may be truncated to match the pupil on the echelle grating and to
avoid conflict with the camera lenses. The collimated beam overfills the echelle grating by
14%. It should be noted that this collimator design could have easily been adapted to a
slower off-axis design which would place the fibre exits out of the beam. This would allow
for greater flexibility in the fibre feed design. However the initial concept (see below) was
designed to accommodate only a set of bare fibres and/or micro-slits and was therefore
extremely simple (and low cost).

2This is RGL grating no. 53.417 which was ruled for the Keck HIRES. The RGL catalogue erroneously
gives a size of 204 x 408. The blaze angle also differs slightly from the catalogue value.
110 Chapter 3. The design of SALT HRS

Fibre micro-slits
Given the above echelle and collimator properties a 100/Lm fibre must be used to obtain
a resolving power of 100000. As shown in Figure 3.5 the throughput of such a fibre would
be extremely poor. For this concept design it is proposed that micro-slits be imprinted
directly onto the output face of the larger diameter fibres. The concept was discussed
above (Section 3.2.2) and optimal configuration of a range of fibres and micro-slits and
their throughputs in median seeing was calculated. The results are repeated here in Table
3.7.

Fibre diam. Micro-slit Resolving power Throughput

(/Lm) width (/Lm) ()../o)..) (%)
400 22900 67.8 Table 3.7: The optimal con-
350 200 37800 41.9 figuration of fibres and micro-
300 100 75500 21.9 slits and their throughputs in
median seeing. The through-
400 70 108000 13.9 put is purely geometric.

Cross dispersion
Like the HERCULES instrument, CELESTIA uses prisms in double-pass for cross-dispersion.
Prisms have a clear advantage over gratings in that they make better use of the CCD
detecting area and they are also significantly more efficient over a large wavelength range
(see Figure 3.17).

------
1001'------,------~--r===~==~====~
- Prism x-dispersion
90 1 - - - - Grating x-dispersion

rr
80
70 ,-- ---- __
, --
;g
o
60 ,/ ---""-
-- ----- .........
I .......
'-" I ......
~
offi 50
'0 --.
if] 40
30 Figure 3.17: The efficiency oftwo prisms used
20 in double pass compared to the efficiency of a
surface relief grating (blazed at A = 550 nm).
10
The prisms, made of BK7 (or equivalently,
O~------~----~----~------~--~ Ohara BSL7), are assumed to have broad-band
400 500 600 700 800
Wavelength A (nm) anti-reflective coatings on each face.

The prism parameters (Table 3.8) have been chosen so that the cross-dispersion will
allow the entire visible spectrum to be captured on a 61 mm square frame which would
be a mosaic of two adjacent 2kx4k CCDS. This provides for a minimum order separation
of 7". At all but the far red wavelengths it is possible to have one sky fibre at R = 37800
when using 350/Lm fibres. At R = 22900 the 400/Lm fibre will only allow a sky fibre
to be used up to ).. ~ 650 nm. However, a 300 - 350/Lm micro-slit aligned horizontally
may be used to minimize the order height to allow a sky fibre to cover the entire visible
3.3. R2 and R3 designs 111

spectrum3 . The prism apex angles have been chosen so that the prism masses are nearly
equal. To make the manufacture of these prisms feasible each prism could be formed from
two right angle prisms which would have masses similar to the HERCULES prism.

Prism 1 Prism 2
Apex angle 81 = 41.50° 82 = 44.50°
Angle of incidence 01 = 33.05° O2 = 34.06°
Base b1 = 300mm b2 = 312mm
Height hI = 393mm h2 = 379mm
Length it = 440mm 12 = 400mm
Table 3.8: The CELESTIA prism
Mass Ml = 61.9 kg M2 = 57.4 kg parameters.

Camera
In order to deliver a resolving power of R = 100000 the focal length of the camera must
be fearn = 650 mm4. With this camera focal length the spectral format is well suited to
a 61.4 x 61.4 mm detector; i.e., a mosaic of two 2k x 4k CCDS with 15/km pixels (see
Figure 3.18). The constraints of the echelle grating/collimator geometry, coupled with
the desire to minimize the angle of illumination of the echelle (0), place the camera 3.50 m
from the entrance pupil (the echelle grating). In order to capture all the dispersed light
the first element of the camera, which is of course a catadioptric design, must be at least
850 mm in diameter. The size of the camera primary mirror has been limited to 1 m for
this design. The extremely fast nature of this camera (rv f /0.65 in white light) prohibits
an off-axis or folded design. Schmidt camera designs were tried, but none was found to
produce the required image quality. It was then realized that Epps and Vogt (1993) came
to the same conclusion when designing the camera for the Keck HIRES instrument. After
exploring several options involving a variety of aspheric elements, these authors found
that an all-spherical design produced satisfactory images. Their design, which involves
two large lenses (a biconvex and a meniscus) has been adapted to the much larger pupil
distance and faster focal ratio required by CELESTIA and optimized for the use of BK7
glass. The design is essentially a derivative of earlier designs by Houghton (op. cit. and
references therein).
The optical design of the CELESTIA camera is shown in Figure 3.19 and the parameters
of the camera are given in Table 3.9. Assuming the final element of the camera (the field-
flattening lens) is only slightly smaller than the overall cryostat dimensions, the cryostat
will vignette at most 30% of the rays for wavelengths near the centre of the CCD frame.
Those wavelengths nearer the edge of the field should experience little or no vignetting.
The optical performance is superb across the entire visible spectrum (see Figures 3.20 and
3.21).

3These calculations were presented to SSWG in October 2001. It was subsequently realized that the
effective height of the extracted fibre profile was over-estimated. In fact it is possible to observe two
400 f-Lm fibres at all wavelengths. However, the calculations are still approximately valid when micro-slits
are used.
4The maximum resolving power obtainable with this focal length is actually R = 108000
 m D.y

20
,, -------- Nil 7~
57

~1
~ -~,
Sill
---- -- --
- I';t;:!f)-N;:!E)-
- 0
, I __ ,
e 5B6 7.8"
-
10
- . ~
--
~
,

E -
",
, 0
~ 8.6"
g 69 .. '1:I'\Y, I
-
c 73 -
"
"
.." , ,n '-'
,~,
,....
'-' : --
:-'189 9"
0 ,
..
0 -
~
'en 77
"
, \
,~

,II r.
-=- 464 9.5"
0 .n · '-'
• -
Q..
I
-
'" .
i" ~"
" f,
• -
--
'''' ~"

>- 81 "'" , _0 -r, r. '1:1

• I 441 10"
" "'IV, '" ,..
I~
.. ,..
.,. '"
-1 0
85
-
1'-' ,~,.c,

'-' "
\
,
, . _

'\"',.
-,.
· M. " '-' '\~/~
'-'

.,.
;
'.
- .
I, --
420 10.6"

.
,~

~ r-. '\'!J
'-' ,~
". '''' ", ,.... -~
'-'
,'-' IV
-- , '''''' . -

. .•
,,·,1,
.
'-'
r; '\'
"
89 - 'v
'-'
I '-'
402 11 .1 "
- I
"';.,
'"'' '''' -
- '-'~,
~~U\"'~

"...
~

.~u"\"1 .
I I

93
'v'
I .1
f>,

"I I~
~.
'-' '\V ,V,
~

- I .11 , 384 11.7"

-- , "" .
, · } '
"
V
''''
•
· "
•I
·
I
- '-' r.
'-'
97 - I I 368 12.3"
-
FSR
FWHM

-30 -20 -10 0 10 20 30

X-position (mm)

Figure 3.18: CELESTIA spectral format.

3.3. R2 and R3 designs 113

~
~
~

~
nD T /' L
L :.l
L.

-'
..I
...l
~

.....
- ++-: ..c::::

..; ,...~ ~ ~

~- ;,

V TG- :"I

Figure 3.19: CELESTIA camera.

Lens 1: Radius of curvature Rl = 2794 mm

R2 = -3446mm
Thickness l = 56 mm
Lens 2: Radius of curvature Rl = 1194 mm
R2 = 666mm
Thickness l = 25 mm Table 3.9: Parameters of the
CELESTIA camera. All surfaces
Primary mirror: Focal length f = 650 mm
are spherical and both lenses
Diameter D = 1.0 m are BK7.

Mechanical design
The mechanical design of CELESTIA is shown in Figure 3.22. The spectrograph is a
completely static structure apart from the ability to focus the CCD. A framework of
trusses is used to support the various optical elements and the entire instrument is to
be enclosed inside a sealed tank. The tank is completely non-structural and the two end
sections are able to be removed by means of an over-head gantry. It was proposed to fill
the section containing the dispersive elements with helium in order to avoid temperature
and pressure-dependent wavelength shifts. The camera would remain at atmospheric
pressure in order to make access to the detector easier.
114 Chapter 3. The design of SALT HRS

A- 1.411111 A- !)J.51ll1l A- · 'r':.7UIll

.....

" ": \f(i~{:,< : "

."

" ">H"1~ii{ " .: ' .' .' ~

I, .'

Onil'T 41):
I'r ms = 5.DJIlll. r "I'O = 11.!JpLll

A-!i J 1.2uIlI t\ - 5Ll1. 1Il11 t\ - [)07. 0ulll

·..S.,
Orclt·r 70:

,\ - 3!J 1.9 tull A - 392. Illl 1 >. - 300.6nm

Figure 3.20: Spot

diagrams for CELES-
TIA. Wavelengths
are given in the cen-
tre and extreme edge
of representative or-
Ordf'r 91: ders. All boxes are 2
rrm - J.7/1 111. rg - 12.5fJlfl rJln. - ;3.9/011. r e - 1 ..It/In pixels (Le. , 30 Mm.

0.9

0.8
>-
~
~0 . 7
(!)
'0
~0 . 6
o
(3
~0 .5
'0
- t.. = 390 .6. order = 91
.§0.4
- t.. = 392 .8. order =91
t3
- t.. = 394.9, order = 91
J: 0.3 - t.. = 507 .0. order = 70
- t.. = 510.6, order = 70
t.. = 514.2, order = 70
- t.. = 885.7, order = 40
- t.. = 893.5, order = 40 F igure 3.21: CELESTIA geometric encir-
- t.. = 901.4, order = 40
cled energies. The same wavelengths used
5 10 15
Radius from centroid ()..t.m) in Figure 3.20 are shown.
3.3. R2 and R3 designs 115

Figure 3.22: Mechanical design of CELESTIA.

116 Chapter 3. The design of SALT HRS

3.3.2 Alternative designs

Following SALT meetings in April 2002 a subcommittee of the SALT Science Working
Group (SSWG) reviewed the CELESTIA concept and resolved that:

1. As accurate a measurement of the sky background as possible is essential for SALT HRS
science. This requirement can be satisfied with one star and one sky fibre.

2. There is sufficient motivation for increasing the inter-order spacing to warrant re-
questing that the instrument designers investigate this possibility, either with prism
or grating cross dispersion.

3. It is imperative for the instrument designers to investigate the effects of a changing

pupil on stability of HRS radial-velocity measurements, and weigh the relative merits
of an encapsulating tank against other alternatives, e.g., iodine cell, interferometric
comb, etc.

Items 1. and 2. reflect the fact that the CELESTIA optical design did not allow suffi-
cient inter-order spacing for an additional sky fibre to be used at all resolving powers and
wavelengths. However, as noted above, although the inter-order separation was misrepre-
sented to SSWG this conclusion does not change. The absence of a sky fibre in the red-most
wavelengths was felt to be a particular problem as this is where night-sky contaminations
becomes particularly severe. The third item is simply a misunderstanding of the principle
of enclosing the spectrograph inside a helium or vacuum chamber. This measure is only
one aspect of an attempt to make the HRS as stable as it can possibly be, and it does
not preclude the use of other measures (e.g., iodine cell and/or fibre double-scrambling)
which can further enhance the instrument's stability. Both these options remain possible.
With these resolutions in mind we have explored a number of design concepts below.

Echelle grating and cross-dispersion options

The choices of echelle grating and method of cross-dispersion are intimately connected
to the amount of inter-order spacing. In order to find an optical design which increased
the inter-order spacing, several echelle grating options among those detailed in Table 3.6
were explored. These gratings were:

Catalogue Ruling density Blaze angle R# Ruled size (mm)

no. (lines/mm) eB (taneB ) w L
53.417 52.67 70.45° 2.8 308 408
53.121 110 64° 2.1 310 413
53.127 87 63° 2.0 308 413

Assuming a beam size B = 308 mm, grating 53.417 gives a resolving power of R ~ 20000 if
500/Lm (1.8") fibres are used. Because gratings 53.127 and 53.121 are ruled at a shallower
blaze angle, the same fibre will give R ~ 20000 only if a beam size of B = 350 to 400 mm
is used. This beam will overfill both the length and width of the grating mosaic. However
a 365 mm beam overfills the R2 gratings by approximately the same amount (!"V15%) as a
300 mm beam overfills an R2.8 grating with the same dimensions.
3.3. R2 and R3 designs 117

The use of BK7 prisms continues to be assumed for cross-dispersion and for each echelle
grating an optimum prism apex angle was found which allowed between two and three
objects per order to be observed. Three objects are required in order to allow a "nod and
shuffle" mode (see Section 3.2). The criteria for sufficient inter-order spacing is that the
distance between adjacent objects (or orders) should be at least 3 x h where h is the height
of the fibre image on the detector. This allows an optimal extraction slit of 2.5 x h to be
used while still allowing the background to be sampled. It is important to note that the
extracted height of a fibre is somewhat less than the height of the dispersed fibre. This is
for the same reason that a fibre will give a greater resolving power than a slit (see Section
1.2.10)5. The minimum apex angles of two identical prisms in double-pass for each of the
grating options are given in Table 3.10. Figures 3.23 to 3.33 show the spectral format for
some of these grating/prism configurations. A discussion of the various options follows.

Catalogue Ruling density T Blaze angle Prism apex angle (ccp)

no. lines/mm eB Number fib. = 2 Number fib. =3
53.417 52.67 70.45° 47.7° 57.4°
53.121 110 64° 23.3° 32.8°
53.127 87 63° 28.3° 38.8°
Table 3.10: Minimum prism apex angles for various echelle gratings. The use of two prisms in double-
pass is assumed. For echelle grating no. 53.417 up to three prisms may be required.

Figures 3.23 and 3.24 show the spectral format for the R2.8 echelle grating with 52.67
lines/mm with a pair of 47.7° and 57.4° prisms respectively. The first configuration is
essentially the OELESTIA spectral format (Figure 3.18), but with slightly more cross-
dispersion so that two fibres can be observed at all wavelengths. The camera focal length
is also slightly reduced to allow the entire spectrum to fit on a mosaic of two 2k x 4k
OODS with 15/Lm pixels. This would reduce the maximum possible resolving power from
around Rrnax = 100,000 to Rrnax = 85,000. The increased cross-dispersion which allows
three objects to be observed places a large portion of the spectrum off the detector. These
wavelengths could be recovered by either moving the detector or by rocking the prism(s).
Alternatively, the camera focal length could be reduced (to fearn = 500 mm) which would
limit the maximum resolving power to Rrnax ::;::j 80,000. Such a camera would be even
faster than the one proposed for OELESTIA and is possibly not feasible.
The spectral formats shown in Figures 3.25 and 3.26 are for an R2 echelle grating with
110 lines/mm with pairs of 32.8° and 40° prisms respectively. The pair of 32.8° prisms
could be replaced by a single 53.8° prism, while still allowing three objects to be observed
per order. A more natural format is obtained with the 40° pair of prisms (Figure 3.26).
This allows nearly 4 objects to be observed per order. However, because of the high line
density of this echelle grating the angular extent of the orders is much greater than can be
captured by a 60-mm wide OOD. The missing wavelengths could be recovered by tilting
the echelle grating; however this is not an attractive solution for reasons of stability. A
larger format OOD (i.e., a 3 OOD mosaic) would reduce the lost wavelengths. However a
tilt able grating would still be required to obtain complete coverage. The large inter-order

5This aspect was neglected when the spectral formats were presented in October 2002 (Barnes and
Albrow, 2002) but this has been corrected here.
118 Chapter 3. The design of SALT HRS

E
5
c
o
~
?--1
>-

95 o 376 14.3'

Figure 3.23: The spectral

format for an R2.8 echelle
grating with 57lines/mm.
Cross-dispersion is with two
-30 -20 -10 0 10 20 30 47.7° prisms.
X-position (mm)

. . .

~
Order

30

20 "'3.2'
13.7'
E
5 10 3 14.2"

c 503 14.8"
..g 0 7
- 476 15.5"
'8 79 452 16.2Q
?- -10
>- 83 430 17.1"

-20 87 411 18'

I
91 393 19"
-30
- I
I I
I
95
== I I 376 20.1Q

Figure 3.24: The spectral

format for an R2.8 echelle
grating with 57lines/mm.
Cross-dispersion is with two
-30 -20 -10 0 10 20 30 57.4° prisms.
X-position (mm)
3.3. R2 and R3 designs 119

. .

Order As !1y
30
905 12.8"
20

E 10
5-
c
.,8 0
'iii
o
?- -10
>- - 8
-
-20
-
-30

Figure 3.25: The spectral

format for an R2 echelle
grating with 110lines/mm.
Cross-dispersion is with two
-30 -20 -10 0 10 20 30 32.8° prisms.
X-position (mm)

Order As !1y
30 904 16.9"

20

E 10 -
---:'
5- ~
c .::::
g 0- 3
-
.~
\ 36

?- -10 .-- 38 29
>- -- 26.
-
-20 - --
\
28.1"

..
2\
I

-30 -- 2 •.• •

Figure 3.26: The spectral

format for an R2 echelle
grating with 110 lines/mm.
Cross-dispersion is with two
-30 -20 -10 0 10 20 30 40.0° prisms.
X-position (mm)
120 Chapter 3. The design of SALT HRS

separation makes this option extremely efficient at high resolutions as the space between
orders could be used to place either fibre bundles or sliced fibre images.
The final echelle grating investigated was an R2 with 87 lines/mm. The spectral
formats with this gratings are shown in Figures 3.27 and 3.28. These spectral formats
are obtained using a pair of 38.8° prisms in double-pass. The requirement that up to
three objects be observable at all wavelengths is met. However, the wavelength coverage
is not complete above 550 nm if a mosaic of two 2k x 4k CCDs with 15J-lm pixels is used.
Complete wavelength coverage may be obtained if the echelle grating can be tilted (Figure
3.28). A camera capable of capturing these spectral formats was developed and is shown
in Figure 3.29. The camera design is an all-spherical catadioptric system, with three
large lenses and has been derived from the CELESTIA camera. The optical performance
is superior to the earlier two-lens design and is a precursor to later designs which require
multiple elements in order to obtain good image quality over much larger field angles.
The spot diagrams for a range of wavelengths are shown in Figure 3.30. At nearly all
wavelengths the encircled energy is better than 80% within one pixel.
3.3. R2 and R3 designs 121

. .

Order AB !>y
30 887 12.8"

20

E 10
.sc -
g 0 - -
.~

?- -10 -
>- - -... D.G Il
-20 -...- '1.6"
- --". 22.7"
-30

Figure 3.27: The spectral

format for an R2 echelle
grating with 87lines/mm.
Cross-dispersion is with two
-30 -20 -10 0 10 20 30 38.8° prisms.
X-position (mm)

Order AB 'Y Order AB 'Y

30 30
63' 12.9' 893 12.7'

E
-S
20

10 E
-S
20

10
IIIII
c: c:
a 0'- a 0
~ ~
a
a c.
:u:"
c.
~ -10 ~ -10 18.4'

-:<.. 49 19.4'

-20 "-,;- -20

.51

3 , 3M
20,3"

21.4"

4 22.5'
-30 -30

-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

X-position (mm) X-position (mm)

Figure 3.28: The spectral format for an R2 echelle grating with 87lines/mm, Cross-dispersion is
with two 38.8° prisms. Complete wavelength coverage can be obtained by tilting the echelle grating by
tlo = ±O.7°.
3.3. R2 and R3 designs 123

3.3.3 SALT HRS R2

At SALT meetings in April 2003 the final design discussed above was presented. The
option of allowing a tilt able echelle grating was rejected in favour of increasing the field
of the camera and using a mosaic of three 2k x 4k CCDS. This spectrograph option,
hereinafter referred to as SALT HRS R2, was developed in detail and was presented for a
Preliminary Design Review (PDR) in 2003 September. The PDR Optical Design Definition
Document can be found in Appendix D. The design of SALT HRS R2 necessitated the use
of large prism cross-dispersers and the development of a very large catadioptric camera
(see Figures 3.31 and 3.32). The spectral coverage is however complete (see Figure 3.33).

/
Collimator

Figure 3.31 : Plan and elevation views of SALT HRS R2. The collimator is shown on axis with a focal
ratio of f /3.8.
124 Chapter 3. The design of SALT HRS

CCD

\ t!
Corrector lenses
obstruction

CCD and field-

Primary

1.\

Figure 3.32: The SALT HRS R2 camera. The CCD obstruction approximates the position of the cryostat.
 I I I I I I I I I I I

m AB fly

-
30 ~

"'"
;s-~
r
r-
21

? 7-
.,..
... .....
"
n '"
ro.
v
-
972 13.8"
~ 13.2"
.
~
,-
,,
ro .. ~~
20 '}Q
~ ~
~

..12 "l'!: --=

~
SII I ~02\B} -,
I- 31
.33. ~ '.
I Hel ' , All Lil SII
-- ~i:
- .. --
I
NMill
Ha.
SiB -
Nil
- -- r- - .. Hel , All
~~

L~I , ' 619

14.2" -

10 ~
~:.t5 .. Nl'INal
,
" ~
. I"
i - -68£ ::I.4:8"-.
_ ~
-J7 "
, ,,"
'. ~
1~
-
39_ ..
' ' ~.
.t..
I
i
... 11.,. ~
i
.1 "
--
g
c
-
,," II~ '
o o I- 41 ' ..v,
" ~
' oy ~ 16.9"
~
'00 - \ ,n
"
43
.-' .::J75 17.7"
~

o I
...
. -
" ,"
\ "" "
-- ,
~
0-
45
..":. ...-". "'" - 454 18.5"

--
I \.

>- -10 I- 47 -- ~
..
, ~

"'''+:
r.

~
:.".:
' r .,.,
I r. 434 19.4" -
49 - v,
\.
"'-
. r .. "

'''"' C"II ... • ...

1-, ,ro
· ro. 417 20.3"
"'~,
• co ,, '.
.v .\ '. ~ ~.
"" " - ~

-
~

r . .f
-20 I- 1
51 't '.
.~\ .'. . s. ~ I on ..;".. ,, rll
400 21.3"

53
-
.~ ] . ': '}
'v, '.
-''-

,
~ ,~
,~

-
' / t.n.ItlJ

lr>. '
!~. " ' ' ' '
':"' . : , ..In
385 22.3"

-30 l- SS - ~

.IL
I
I 371 23.3"
-
I_II
FSR
- - FWHM
I I I I I I I I I I I

-50 -40 -30 -20 -10 0 10 20 30 40 50

X-position (mm)

Figure 3.33: SALT HRS R2 spectral format. The echelle grating has 87lines/mm and cross-dispersion is with two 40 0 prisms used in double-pass. The
wavelength coverage obtainable with a mosaic of three 2k x 4k CCDS is almost complete (apart from the gap between CCDS). The inter-order separation is
sufficient for more than three fibre at the lowest resolving power.
126 Chapter 3. The design of SALT HRS

Preliminary Design Review result

Formally the SALT HRS R2 did not pass the Preliminary Design Review (PDR) held in
September 2003. This was because of concerns about the management (the schedule was
perceived somewhat short) and potential costs (some of which may have been under-
estimated). The optical design itself did formally pass, but there were concerns. These
have been summarized by the SALT project scientist (Buckley, 2003); namely:

1. whether the current design was significantly better than a potentially simpler, less
risky and less costly alternative design proposed by one of the external reviewers
(B. Delabre);

2. the current design involves handling some large optics. There are potentially signif-
icant handling risks, particularly in coating and mounting the lenses, and breakage
risks in their transportation;

3. the prisms will be the largest ever produced for astronomy. Optical requirements
for homogeneity are quite demanding. Delivery of suitable material blanks is a very
long lead item, and a major potential single-point failure in either breakage, schedule
or cost risk;

4. the camera design involves mounting large diameter lenses (three) in a cell(s), which
were poorly specified at the time of PDR;

5. use of a BK7 cryostat window is to be avoided due to the inherent radioactivity of

the material and the consequently high cosmic ray rate on the CCD detectors;

6. the heat pipe arrangement for the CCD cryostat was identified as a difficult and
risky area needing more detailed attention.

Items 2, 3 and 4, relate to concerns about the apparently novel optical design. The
proposed instrument would have exceeded the size of any previous high resolution spec-
trograph by a considerable factor. However, similar design have been built. For instance,
the Subaru HDS (Noguchi et aI., 2002) and Hectochelle on the MMT (Szentgyorgyi et aI.,
1998) both have cameras which are nearly identical in form although slightly smaller.
During the PDR it was noted that a two-corrector lens camera design would be possible
with a slightly hyperboloid mirror. Along with a reduction in unvignetted field cover-
age (perhaps motivated by the use of a smaller 4k x 4k CCD), these measures had the
potential to bring down the size of the camera considerably.
The prisms, while being the largest ever proposed at that time for use in astronomy, if
made as four pairs, would have been only slightly larger than the HERCULES prisms and
those used in the Keck ESI (Sheinis et aI., 2002). Recently, even larger prisms have been
proposed (Fabricant et aI. 2003, Szentgyorgyi et aI. 2003).
The issue of the CCD (and its cyrostat window) certainly required further attention.
Again numerous working examples exist which suggest that the CCD concept was techni-
cally feasible, although this would only have been demonstrable after a not inconsiderable
effort.
However, in light of these comments, in particular item 1., an R4 dual beam design
has been developed and is presented in the following section.
3.4. R4 designs 127

3.4 R4 designs
In order to make a direct comparison with SALT HRS R2 (see Section 3.3.3) the same
baseline functional requirements are demanded of any alternative R4 design. That is,
the wavelength coverage, resolving powers and stability requirements are assumed to be
unchanged. It is possible that along with a change in the form of the spectrograph some
additional functionality could be provided; or that with a change in requirements of either
design, the performance could be substantially improved. Such possibilities are beyond
the scope of this investigation and will not be discussed in any detail here. In Section 3.4.1
the initial R4 concept is detailed. Section 3.4.2 provides a comparison of the SALT HRS R2
with the R4 design described below.

3.4.1 Conceptual design

The form of the SALT HRS R4 was suggested by one the PDR external reviewers, B. Delabre
and is based in part on a design for an HRS for the 4-metre SOAR telescope (Castilho et al.,
2003). The design is also quite similar to the UVES instrument on the VLT (Dekker et al.,
2000) which uses the white pupil concept of Barrane as adapted by Delabre (op. cit. and
references therein). Unlike UVES however, we are assuming that both the blue and red
arms use a single echelle grating. A dichroic filter will be used to split the instrument
into two beams after echelle dispersion, and each beam will use dedicated cross-dispersers
and dioptric cameras.
An initial design, shown in Figure 3.34, uses a single echelle grating blazed at BB =
76.0° 'with a groove density of 41.5 grooves/mm . .A single parabolic mirror is used off-axis,
first as the collimator and then again, after echelle dispersion, to form an intermediate
focus. At this point the red and blue arms are split by a dichroic at a nominal wavelength
of 550 nm. The blue arm then uses the collimator mirror to form an image of the echelle
grating ("the white pupil") at the point of cross-dispersion, while the red arm uses a
smaller spherical mirror. A novel feature of the design is the assumption that volume
phase holographic (VPH) grisms can be used for cross-dispersion. This is only possible
because of the recent development of such gratings with high efficiencies over broad bands
(see Barden etal., 2000). A pair of fiat fold mirrors allows convenient placement of the
cameras.

Figure 3.34: The SALT HRS R4 conceptual design. The red camera is on the left and the blue camera is
on the right. Both cameras are depicted as paraxial elements.
128 Chapter 3. The design of SALT HRS

Fibre input
The use of image slicers is proposed for SALT HRS R4. The form of the fibre input and
image slicer (including transfer optics) is assumed to be similar to the R2 design (see
Section 3.2) except that the minimum inter-order spacing is different. The parameters
of the image slicers and their geometrical throughputs are given in Table 3.12 (see also
Appendix E.2.2).

Collimator
The collimator uses a portion of an 800 mm diameter parabola that also serves (twice) as
the blue arm white pupil mirror. It would be possible to use a smaller off-axis element,
which would also double as the first white pupil mirror. However this mirror would be
quite large (i.e., > 400 mm) and would probably necessitate the manufacture of the large
mirror in any case. It would then also be necessary to manufacture an additional blue
white pupil mirror (which may be spherical). The choice offocal ratio (J /2.5) is somewhat
arbitrary, although a slower design would require a larger mirror and a faster design would
be increasingly difficult to manufacture.

Echelle grating
The splitting of the spectrograph into the red and blue arms occurs near the intermediate
focus following the first white pupil mirror. This enables the use of a single echelle
grating mosaic, where each grating has the parameters given in Table 3.11. This is not
necessarily ideal and more efficient use of the cross-dispersers and camera/detector could
possibly have been made if the choice of echelle grating were optimized for each of the
red and blue arms. It was proposed to assemble the grating from two individual gratings
which will be aligned mechanically into a single mosaic with dimensions 855 x 204 mm.
This would leave a gap of 35 mm between the gratings. With a collimated beam size of
B = 200 mm there is no overfilling.

Parameter Specification
Blaze angle, ()B 76.0°
Groove density, T 41.5 grooves/mm
Grating ruled width 204mm Table 3.11: Echelle grating parameters
Grating ruled length 410mm for SALT HRS R4.

Dichroic
The dichroic has a nominal wavelength division of 555 nm. It is placed at the f /2.5
focus of the intermediate echelle spectrum. To capture all the light it must be at least
50 x 350 mm in size.

Pupil mirrors
The pupil imaging in the blue arm is performed by the large parabolic mirror that also
serves as the collimator. The red arm white pupil mirror is a 500 mm diameter spherical
3.4. R4 designs 129

mirror with a 3 m radius of curvature. This allows the white pupil to be demagnified by
1.33 (from 200 mm to 150 mm), which is better suited to the use of VPH gratings.

The fold mirrors

A pair of fold mirrors is placed just in front of the white pupil to make room for the pair
of cross-dispersers and cameras. The blue arm fold mirror must be 320 mm in diameter
while that for the red arm is 280 mm in diameter. Although the footprint on the mirror
will be elliptical, it is assumed that they would be manufactured circular. These mirrors
can be repositioned in order to make the camera focal planes approximately coplanar.

VPH cross-dispersers

The development of VPH gratings and their potential for use in astronomical instrumen-
tation has been described by Barden et al. (2000). A report by Clemens and Seagroves
(1999) gives an overview of the theory of VPH gratings and an extensive list of technical
information can be found on the NOAO VPH website6 .
It is possible to tune the wavelength of peak efficiency by altering the angle of incidence
onto the grating. However, if the grating is rotated, the camera must either be articulated,
so that it can move into the dispersed beam, or a system of two counter-rotating mirrors
employed (i.e., the "butterfly mount", see Bernstein et al., 2000). Another possibility is
to immerse the grating inside a prism with apex angles chosen so that the mean deviation
is zero. In the case of a minimum deviation grism the efficiency can be tuned simply by
rotating the grism about its central axis. This would not significantly alter the spectral
format, and the camera may remain fixed (except, perhaps, for a small focus correction).
The second mirror of the butterfly mount is therefore not required. It is also possible to
increase the overall efficiency of a VPH grating by ensuring that every wavelength reaches
the grating at close to the ideal angle of incidence. This may be achieved by using a
prism to disperse each wavelength before the grating, although this dispersion must be in
the opposite direction to the grating dispersion (Delabre, personal communication). This
technique has not been investigated further.
The choice of cross-dispersers depends entirely on the desired wavelength range and
order separation. In order to be able to make a direct comparison with the R2 design the
R4 design should be capable of at least the same wavelength coverage. Because the amount
of CCD real estate is fixed, it is this constraint that limits the order separation. With unity
pupil magnification, the blue grating requires T = 950 lines/mm with Oi = 12.3° (Amid =
450nm). This provides for a nominal wavelength range from 370 to 565nm in.44 orders
(m = 125 to 82) and a minimum order separation is 11.5/1. A red cross-disperser used
with unity pupil magnification would require a 450-line/mm grating in order to achieve
the same minimum order separation. This is considered a low density for the efficient
use of VPH gratings. Hence a demagnification of 1.33 in the pupil is required in order to
allow a slightly higher groove density (i.e., T = 600 line/mm) to be used. The short focal
length transfer mirror, used only in the red arm, produces this demagnification.

6http://www.noao.edu/ets/vpgratings/
130 Chapter 3. The design of SALT HRS

Camera and detectors

The blue camera will be limited to a maximum resolving power of R = 80000 with
two (15 Mm) pixel sampling. The focal length required is fearn = 301 mm which gives a
monochromatic focal ratio of f /1.5. With a 30.5 x 61 mm focal plane (i.e., a single 2k
by 4k CCD wit h 15Mm pixels) a single FSR can be observed at all wavelengths from 380
to 560 nm (see Figure 3.35). After a 1.33 pupil demagnification the focal length required
for two pixel sampling of R = 100000 is fearn = 283 mm. This is also a potentially very
challenging focal ratio (f /1.4) with a 12.2° field of view - assuming that at least two
2k by 4k CCDS (with 15 Mm pixels) are used. The spectral format of the blue camera is
shown in Figure 3.36. It may also be possible to use a single 4k by 4k detector with 15 Mm
pixels. Cameras for each of the blue and red arms were designed by D. Jones 7 (with some
assistance by t he author) and are described further in Appendix E.

Order ~y

30 I ~~
85 I 23.9"
I
87 \ 22.9"
.J.
20 89 I
~.

~
~-~ 21.9"
.'" ..,
'L v I
-91 -5""1"r 21"
I

E 10
--g3 ,~

, ~
'"'." 20.1"

95 ....,
....oft">
19.2"

--
E
c
o
..;::: o
97

W
I

I
't-'

'~
I
.
I
1f82

--:!!2
18.4"

17.7"
1U1 I ~63 16.9"
·00 ,J.., .

o 103
I - - 454 16.2"
a.. 105
,~

'''' "'" .:::. 445 15.6"

I ~. "c.,. T :

>- -10
-=--
107 " IV!
- 437 15"
'"', ~

109 " I" 429 14.4"

.'-':'!
111
- 421 13.8"

---
,~,

113 '''I'
,...
,~

413 13.3"

-20 ----
~

115 ~
406 12.8"
117 399 12.3"
119 393 11.8"
:.:;=)
121
123 -- 386
380
11.4"
10.9"
-30 125 374 10.5"

-20 -10 0 10 20
X-position (mm)

Figure 3.35: The SALT HRS R4 blue camera spectral format. The orders are plotted over two free spectral
ranges and the dot-dashed line shows the extent of one free spectral range. The outline of a single 2k by
4k CCD with 15 f.Lm pixels is shown in bold.

7Prime Optics, Queensland, Australia.

3.4. R4 designs 131

Order As ~y

53 882 27.3"
30 \
~-
55 \ ~

~ 849 25.5"
-I--
20 57
\
- 820 23.8"

- ----
-!--
\

-EE 10
59

61
\
,n I
1-
792

766
22.3"

20.9"

-
' /1

\
2

I
--
-- 742 19.6"

---
63
C
0 7
0 65 \
2
(\/ ."
719 18.4"
~
'00
0
67
"

n"
2
---- 697 17.3"

0..
I
69 J
--
--
677 16.3"

>- -10 71
73 ~ . . --
-
658
640
15.4"
14.6"
- - -\- - ~p----
75 -- -- - I
I I
623 13.8"
-20 77 I I
I I I 607 13"
79 I1 I I
I I I
I
1 591 12.3"
81 I
_ 1-
II
II
_--- 1-: 577 11.7"
83 ":'--=--.1 563 11.1"
-30 85 550 10.5"

-30 -20 -10 0 10 20 30

X-position (mm)

Figure 3.36: The SALT HRS R4 red camera spectral format. The orders are plotted over two free spectral
ranges and the dot-dashed line shows the extent of one free spectral range. The outline of a mosaic of
two 2k by 4k CCDs with 15 f..Lm pixels is shown.
132 Chapter 3. The design of SALT HRS

3.4.2 Comparison of efficiencies: R2 vs. R4

Fibre and collimator

The transmission of the fibres, transfer and fold optics, image slicers and collimator are
assumed to be identical to the R2 design, except that the geometrical transmission of the
image slicer depends on the inter-order spacing. The parameters of the image slicers, which
have been optimized for an 11.5" order separation (which allows complete wavelength
coverage), are given in Table 3.12.

Resolving Fibre diameter Number of Geometrical

power (microns) slices transmission
R Fixed N&S Fixed N&S Fixed N&S
Table 3.12: Fibre
17000 500 0 82%
and image slicer
40000 500 400 3 2 70% 57% properties for the
80000 300 300 4 3 48% 38% SALTHRSR4.

Echelle grating

The near Littrow illumination of the R4 echelle grating results in a peak theoretical blaze
efficiency of nearly 75%. This is somewhat higher than the R2 grating which is illuminated
at 4.5 There is no overfilling of the grating, although there is a 5.5% loss due to the gap
0
•

between the two gratings. This also occurs in the R2 design.

VPH gratings

Some examples of the theoretical efficiency of VPH gratings are shown in Figures 3.37
and 3.38. Also shown in Figure 3.38 is the change in efficiency that results from a small
change in the angle of incidence on the grating. These, and other efficiency predictions,
have been used to predict the performance of the R4 SALT HRS VPH gratings.

HG-T-435-11 ...... ·1 Order

ReWA Theoretical Performance - 0 Order
gOO I/mm - +1 Order
Unpolarized Incident Light @ 11.30 - +2 Order
100,---,----.----,----,----,----.----,-..'::::::::···:::··..;:::
.. =+1:::0r:::de;::r@:::10:::,6=;'

90 -:.:,::p""
......... • • •--•• >ot •••• ,,=~
..... ~

~
80
r"""'~
_<1""""'-
.... .......
~

t 70 ~

.
~60~-+--+--+--~-4_-~-~-~-~-__1
'0
5 50+---+--+--+--4--4--~-~-~-_r-__1
5
~ 40+---+--+--+--4--4--~-~-~-_r-__1
e
~ 30~-+--+--+--~-4_-~-~-~-~-__1
Figure 3.37: The theoretical effi-
20 r---. ciency of a 900-line/mm VPH grating
10 +-1- -----f""'--=;;!:::c---t---+---+-----i--t----:::!=---="'f-.......------__1
from 350nm to 550nm (from Kaiser
.. . . ·=. .
oIr===:
350 370 390
..
410 430
.. .. .. ..·=.. ..=.. . . --
~-·=·l·~ ·~·~·~~-·~·~·l··=-=-~~~===-~-=-~~~E~~··= ·~·-= -=.. -~·=·=·~ Optical Systems Inc.). The efficiency
450 470 490 510 530 550
Wavelength (nm) at two angles of incidence is shown.
3.4. R4 designs 133

HG-T -100-18
ReWA Theoretical Performance
900 I/mm I~-"'-
- ·+1
lOrder
Order
Un polarized Incident Light @ 18.36° - o Order

--
- +2 Order
100

90

V-
~ - --...
...............
80
.......
_ 10
",
./ i'-......
t .............
It 60
.;
I ~o
j
II ~o

i 30 ./'

20 V Figure 3.38: The theoretical effi-

10 ~~
.... .. .... ' "
~ ciency of a 900-line/mm VPH grat-
..
o
550
'

600
~ ...
650 100
-~150 eoo
.-.----------- ---- -.. _------
850 900
ings from 550 nm to 900 nm (from
W.v.... ngth (n"') Kaiser Optical Systems Inc. ) .

Dichroic
For optimum efficiency the dichroic will transmit the longer wavelengths and reflect the
shorter wavelengths. An example of the efficiency of a dichroic is shown in Figure 3.39.
The crossover wavelength can be tuned. Numerically this is done simply by shifting the
efficiency curves. The mean efficiency is 99.3% (below 510 nm) in reflection and 94% in
transmission (above 560 nm). Up to five orders on the red and blue arms will have rapidly
reducing efficiencies due to the dichroic's response .

r .A"

0.8 0.8

~ ~
§ 0.6 § 0.6
.~ .~

'E 'E
~
~
04
. ~ 0.4
~
f- f-

0.2 0.2

\..-...,
J
o 400 500 600 700 BOO 900
O b===~~----~----~----~~
540 550 560 570 5BO
Wavelength (nm) Wavelength (nm)

Figure 3.39: The dichroic efficiency. The blue wavelengths are reflected while the red wavelengths are
transmitted. A dose-up of the crossover region is shown on the right.

White pupil mirrors and camera

The red and blue arms share the same first white pupil mirror (which also doubles as the
collimator). In total, this mirror is used three times in the blue and twice in the red. Hence
it will need to be coated with a high-efficiency broadband coating - although emphasis
should be placed on its blue performance. Alternatively the upper half of the mirror,
which is used only by the blue arm, could be coated separately with a blue-optimized
coating.
The red arm pupil mirror may be optimized for a more limited wavelength range,
as can the two fold mirrors. These coatings have not been specified and the current
134 Chapter 3. The design of SALT HRS

calculations assume a broad-band coating similar to that used on the SALT SAC mirrors.
This coating has a reflectivity of 97% in the red. It is probable that an enhanced silver
coating could be used instead, which has a reflectivity of between 98% and 99% over
these wavelengths. Hence the red efficiencies could be improved by some 3% (i.e., after
two reflections).
The transmission of the cameras has been calculated using the UVES cameras as mod-
els. 8 Standard catalogue data has been used for the absorption of the different glasses.
All air/glass surfaces are assumed to have broad-band anti-reflection coatings applied.
For the purposes of calculation only these are assumed to be Solgel plus MgF 2 .

Summary
A comparison between the R2 and R4 efficiencies (at the blaze peak in fixed fibre mode) is
given in Figure 3.40 and their relative efficiencies are given in Figure 3.41. The quantum
efficiencies of the CCDS are not included as the efficiencies in the red and blue are assumed
to be identical; that is, the R4 design will have blue and red optimized coatings on their
respective CCDS, while the R2 design could employ graduated coatings with red and blue
optimized regions.

20
.........

--
(fl

0
c::
'(j)
en
'E
en
c::
co 10
....
I-

5
R = 17,000
R = 40,000
R = 80,000
OL-~----~------~----~~--~~~==~~
400 500 600 700 800 900
Wavelength (nm)

Figure 3.40: Efficiencies of the R2 and the R4 SALT HRS designs. The black line is the R2 design, while
the blue and red lines are for the blue and red arms of the R4 design. The efficiencies at higher resolving
powers for fixed fibre mode are also shown.

It appears that the blue arm of the R4 design is superior over much of the wavelength
range by as much as 40 to 60%. However, the R4 red efficiency is everywhere poorer (by
between 10 and 20%) mainly because of the additional reflections required and the poorer

8These calculations were made before the cameras were fully designed.
3.4. R4 designs 135

,
1~
\
~'--"
~ .,
___________'.-__ _________R_2__________________________
~ ~

., fIIi ...... - - .. - - . - . .
,
R4 (red)
.,
, fill' . . . . . . . -
.".

400 500 600 700 800 900

W av elength (nm)

Figure 3.41: Relative efficiencies of the R2 and R4 SALT HRS designs. The dashed lines show the mean
relative efficiencies at higher resolving powers in both fixed and nod and shuffle mode.

cross-dispersion efficiency. The efficiencies at higher resolving powers can be found by

multiplying by the ratios of the sliced and unsliced efficiencies. With an order separation
of 11.5" the relative efficiency of the R4 design at a resolving power of R = 40000 is a factor
0.80 and 0.66 lower than at R = 17000 in fixed and nod and shuffle mode respectively.
At R = 80000 the factor is 0.59 and 0.44 in fixed and nod and shuffle mode respectively.
Because the R2 design has a minimum order separation of 13", the sliced efficiencies are
on average 10% better than the R4 design. This mitigates to some extent the decreased
blue performance so that it is now only significantly superior (i.e., by more than 10%) at
wavelengths less than 420 nm.
It should be noted that the objective of obtaining complete wavelength coverage from
370 to 890 nm, coupled with the fixed amount of CCD real estate and realistic camera
fields of view, limits the inter-order spacing. If the wavelength coverage were to be re-
duced, either by reducing the red or blue coverage (or both), the order separation may be
increased, thereby increasing the efficiency at high resolving powers. This applies equally
to bot h the R2 and R4 designs, although only the R4 design will permit interchangeable
cross-dispersers. However, it should also be noted that if superior blue performance is
required by the R2 design at high resolving powers it would be possible to use dedicated
image slicers that can make use of the much larger inter-order spacing provided by prism
cross-dispersion in the blue orders. This would be possible without permanently losing
wavelength coverage, and could make this design extremely attractive for use in the near
UV.
136 Chapter 3. The design of SALT HRS

3.4.3 SALT HRS R4

The design of SALT HRS R4 was developed in detail and a second PDR was held in 2004
July. The design is essentially as described above, except that the use of VPH grisms was
abandoned, and the spectrograph layout was altered to incorporate a 45° dichroic mirror
(see Figure 3.42). T his permitted the fold mirrors before the cameras to be removed. The
optical design document presented at the 2004 July PDR is given in Appendix E and a
summary of t he design is given in Appendix A.2.4.

Blue pupil
mirror

Red pupil
mirror

CollimatorlPupil
transfer mirror

Figure 3.42: The SALT HRS R4 2004 July design. The cameras are shown as paraxial elements.

Post PDR changes

While the 2004 July review was successful, as a result of feedback received at this time
some aspects of the design have been subsequently altered. In particular, the large dioptric
camera required (especially for the blue arm) was found to be extremely expensive and one
of the external reviewers (B. Delabre) pointed out that smaller cameras were possible if
the entire optical train was optimized as a whole and if the VPH gratings were immersed
inside weak meniscus lenses. The use of smaller cameras is made possible by further
demagnifying the pupil in both arms (from an initial 200 mm and 150 mm in the blue and
red arms respectively to 100 mm in both arms). This requires that higher line density VPH
gratings be used. A second aspect of the design that was found in need of improvement
is the dichroic's high angle of incidence. A lower incidence angle dichroic has proved
to reduce the extent of the cross-over region and to eliminate the fringing that would
otherwise occur at red wavelengths.
3.4. R 4 designs 137

The redesign of SALT HRS R4 has allowed a range of spectrograph layouts to be con-
sidered. The design chosen is shown in Figure 3.43. The large off-axis mirror is retained
as both a collimator and first pupil mirror. Both the blue and red pupil mirrors are
spherical and have focal lengths of 1000 mm. These mirrors are slightly displaced from
the int ermediate focus so that the white pupil (on the VPH gratings) is not exactly col-
limated. This is corrected by immersing the gratings inside weak meniscus lenses. Fold
mirrors are inevitable with a design that uses a low angle of incidence dichroic. There
is considerable flexibility in the location of these mirrors, and the chosen option was to
use relatively small (130 mm diameter) mirrors immediately prior to the VPH gratings.
Finally, two entrance slit locations are shown. The first, which directs light toward a
fold mirror, is actually the location of an intermediate slit mask and a set of fore-optics
have been designed which transfer the image of sliced fibres to this point . The other slit
location can be used by removing the fold mirror, and will be the location of the bare
fibres which deliver the lowest resolving power. This will enhance the efficiency of this
mode. The entire instrument will be enclosed inside a pair of intersecting cylinders that
form a vacuum chamber. The lid of the chamber that surrounds the cameras will provide
support for the detectors and fibre fore-optics. The interior bench on which all optical
components are mounted will be rigidly coupled to this lid and will be immune to pressure
and temperature changes. The spectrograph will in turn be housed inside a temperature
stabilized environment.

VPH
gratings
Entrance slit(s)

Red pupil
mirror

~l
Blue pupil if--==......;;;..;=-
mirror

Figure 3.43: The revised SALT HRS R4 design.

The design is currently being finalized and it is expected that a critical design review
will be held in 2005 April. Completion of t~e spectrograph is expected approximately two
years following this date.
138 Chapter 3. The design of SALT HRS

3.5 Summary
The design of SALT HRS has evolved considerably since its inception. The early design,
was in large part very similar to HERCULES, although considerably larger. The R2 double-
pass prism design was conceptually rather simple, and while aspects of the design were
technically challenging, it was far from apparent (at least to the author) that it was an
unrealistic design option for SALT HRS. The alternative R4 design, also described here in
detail, while having potentially equal or even slightly better efficiency, has proved to be
somewhat more complex. However, the R4 optics are considerably smaller and the CCD
options are much more flexible. Both of these facts result in an instrument design about
which there can be little doubt in terms of performance and/or manufacturability.
Chapter 4

Conclusion

In Chapter 1 the theoretical background needed for the design of a high resolution fibre-
fed spectrograph in astronomy was outlined. The relative merits of various spectrograph
forms were explored, and it was made apparent that, for many modes of operation, there
is no unique design solution for an HRS.
Chapter 2 describes the design, construction, and performance of the HERCULES in-
strument in use on the MJUO I-m telescope. The fibre-fed spectrograph has been in
operation for over 3 ~ years (since 2001 April), and in that time has proved to be capable
of excellent performance. The vacuum mounted design allows extremely precise radial
velocity measurements, while the double-pass prism and folded Schmidt camera optical
layout delivers excellent throughput, image quality and stability. The performance of
HERCULES could be improved in a number of ways. A larger format CCD is currently
being acquired and extended wavelength coverage will probably be possible by mid-2005.
Improved fibres, image slicers, and a more efficient detector are all possibilities for enhanc-
ing the efficiency of HERCULES. Greater wavelength coverage, a more stable detector, and
improved throughput will all contribute to further improving the radial velocity precision.
However, probably the greatest improvement in radial velocity precision can only occur
if the effects of telescope guiding errors and incomplete fibre scrambling are removed.
Suggestions for improving the guiding include a more sensitive detector, the redesign of
the fibre input optics, and possibly the use of adaptive optics (or active fibre positioning).
A fibre feed mode incorporating the use of a double-scrambler should also be considered.
Several high resolution spectrograph designs for the SALT ll-m telescope were de-
scribed in Chapter 3. That quite different instrument designs are capable of comparable
performance was demonstrated by the sequential development of first an R2 (and R2.8) de-
sign and then an R4 design. These designs were shown to have (on average, over all modes
of operation) nearly equivalent performance. It was argued that the R2 optical design,
while requiring very large optics, also conveyed considerable simplicity to the mechani-
cal layout, was capable of complete wavelength coverage on a single detector (mosaic),
and would have therefore been operationally quite simple. Concerns about the size (and
cost) of the optics and/or their handling and coating, could readily have been mitigated
against a small loss in throughput (over a small wavelength range). Detector issues were
significant. However, several options were possible, and given sufficient resources a solu-
tion (most probably based on an existing design on another instrument) would have been
possible.
The alternative SALT HRS R4 design has proved to be an attractive solution. The use
of the white pupil layout permits a large reduction in the size of the R4 optics. How-
ever, the number and complexity of the optical surfaces has increased considerably. The
use of volume-phased holographic gratings is vital to making this dual-beam instrument
competitive with prism-based spectrographs. The R4 instrument is extremely compact
(although space constraints should be of minor concern for a fibre-fed spectrograph) and

139
140 Chapter 4. Conclusion

will be mounted inside a light vacuum in a temperature-stabilized environment. Apart

from the fibre-feed and focus control, the spectrograph will have no moving parts. The
design is currently progressing toward a critical design review scheduled to be held in 2005
April. Upon completion, SALT HRS will have the capability to be equivalent to or even
to exceed the performance of all other high resolution spectrographs currently on other
large telescopes.
Appendix A

ECHMOD - a Matlab tool for echelle spectrograph

modelling

A.1 Basic outline

ECHMOD is a tool for modelling echelle spectrographs written in the MATLAB1 software
environment. The program was writ en primarily to investigate fibre feed spectrograph
solutions for SALT, however it has evolved into a tool that is capable of evaluating and
comparing arbritary telescope, slit optics, and spectrograph configurations. Simple ray
tracing simulations are possible and these have been used throughout this thesis to com-
pute spectral formats. However, the techniques are not rigourous and an interface to the
ray-tracing software Zemax has also been developed for verification and detailed optical
performance characterization. Presently this interface is limited to file input/output and
exact spectral formats are computed using code written in the Zemax programing lan-
guage (ZPL). Detailed efficiency models have been constructed. Where possible, optical
prescription data (including coatings) are used along with manufacturers data. Appro-
priate substitutions are made if such data are not available. The efficiency calculations
also form the basis of an exposure time calculator.

A.2 Example input/output files

The telescope and spectrograph (including slit optics) properties are defined by instrument
specific input files. Example input files for HERCULES, CELESTIA and the SALT HRS R2 and
R4 spectrographs are given on the following pages.

IThe MathWorks, Inc; www.mathworks.com

141
142 Appendix A. ECHMOD - a Matlab tool for echelle spectrograph modelling

A.2.1 HERCULES

[ tel, slt, spc ] = hercules_2001apr

tel.name 'MJUO 1m' ;
slt.name { 'HERCULES fibre feed' } ;

spc.lam.minmax [ 377 887 ] *1e-9

spc.R.minmax [ 35000 76300 ]
spc.samp = 2.0 ;
spc.B 0.210 ;
spc . lam . eff { [J }
spc.nwav = 21 ;
spc.dlam '2fsr'

col.name 'HERCULES collimator';

col.F1 0;

ech.name '53. 411-mr152'

ech.type 'user defined echelle'
ech.num = 1 ;
ech.gap = 0 ;
ech.theta 3.0
ech. dtheta = 0
ech.dgamma = 0 ;

xdp.meth 'prism'
xdp.type 'dbl' ;
xdp.num [ 1 ]
xdp.glass { 'BK7' }
xdp.angle [ 49.5 ]
xdp.lam [ 370 ] *1e-9
xdp. coat { 'mgf2 © 40deg' }
xdp.t [ 128 ]
xdp.T [ ]

cam. name { 'HERCULES folded Schmidt' }

cam.dmag [ 1 ] ;

ccd.type { 'tk1024_sq' }
ccd.coat { 'SITe_SI003ABuv' }
ccd.num [ 1 , 1 ]
ccd.orient [ 'y' ]
ccd.psi [ 0.5 ]
ccd.absoff [ 0 8 ]
ccd.offset [ 0 0
-15.3 , 6.7
-2.6 6.7
5.2 , 6.7
-4.9 , -17 .3 ]
ccd.offset = ccd.offset - ccd.absoff
A.2. Example input/output files 143

==============HERCULES Properties============
Telescope: McLellan telescope
D = 1.0m, f 4.55m (f/D = 4.55)
Plate scale = 22.1 um/arcsec (45.33 arcsec/mm)

Beam size, B = 210.0 mm

R_max = 76300.0
Wavelength range: 374.8 to 890.1 nm
(in 89 orders: m = 152 to 64)

Echelle grating:
Blaze angle 64.33 deg (R2.08)
Littrow angle 3.00 deg
T 31.6 grooves/mm
L x W 408.0 x 204.0 mm
[T_blz,T_ove,T_gapJ = [80.6,84.9,0.OJ

Cross-dispersion/order separation:
min = 17.5 arcsec, max = 31.5 arcsec
Method - Prism:
Glass = BK7
alphaP 49.50 deg
thetai = 40.00 deg

Collimator:
F 3.73, f 783.2 mm
(assumes B = 210.0, f/D_tel 4.5 and FRD rho = 1.22)
Plate scale = 22.1 um/arcsec (45.33 arcsec/mm)

Camera:
F = 4.65, f 976.1 mm
(assumes 2.0 (24.0um) pixel sampling of R_max
and pupil demagnification cam_dmag = 1.0)
Plate scale 27.5 um/arcsec (36.37 arcsec/mm)
CCD:
Type tkl024_sq
Coat SITe_SI003ABuv
n_pix [x,yJ [1024,1024J pixels
s_pix 24 microns
num [x,yJ [l,lJ (y)
Pixel scale 1.1 pix/arcsec (0.87 arcsec/pix)

Fibre diameter Slice properties Resolving Power Transmission (pc)

microns arcsec width height num R R_tot PSF Geom Total
=============================================================================
100.0 4.53 100.0 100.0 1 43000 77.9 100.0 77.9
50.0 2.27 50.0 50.0 1 86100 36.2 100.0 36.2
100.0 4.53 50.0 100.0 1 73200 77.9 60.9 47.4

Assumes ...
Seeing median FWHM = 2.50 arcsecs
144 Appendix A. EOHMOD a Matlab tool for echelle spectrograph modelling

A.2.2 CELESTIA

[tel, slt, spc] = salthrs_celestia

tel.name 'SALT'
slt.name { 'SALT HRS fibre feed' } ;

spc.lam.minmax = [370 890 ]*1e-9

spc.R.minmax = [ 17000 , 108000 ]
spc.samp 2.0 ;
spc.B 0.305
spc.lam.eff { [ ] }

col.name 'SALT HRS R2 collimator'

col.F1 o;
ech.name 'R2.8.T57'
ech.type 'user defined echelle'
ech.num 2 ;
ech.gap 0.035
echo theta 2.75
ech. dtheta = 0
ech.dgamma = 0 ;

xdp.meth 'prism'
xdp.type 'dbl' ;
xdp.num = 1
xdp.glass { 'BK7' 'BK7' } ;
xdp.angle [ 41.5 44.5 ] ;
xdp.lam [ 370 370 ] *1e9
xdp.coat = { 'Solgel + MgF2' 'Solgel + MgF2' }
xdp.t = [ 150 156 ]
xdp.T [ ]

ccd.type { 'ccd44_82' }
ccd.coat { 'E2V DDSi astroBB' }
ccd.num [ 1 , 2 ]
ccd.psi [ 1.0 ]
ccd.orient [ 'x' ]
ccd.offset [ o , -0.5 ]
cam. name { 'SALT HRS R2 camera' }
cam.dmag [ 1 ] ;
A.2. Example input/output files 145

==============SALT HRS CELESTIA Properties============

Telescope: SALT
D = 11.0m, f = 46.20m (f/D = 4.20)
Plate scale = 224.0 um/arcsec (4.46 arcsec/rom)

Beam size, B = 305.0 rom

R_max = 108000.0
Wavelength range: 368.4 to 893.4 nm
(in 58 orders: m = 97 to 40)

Echelle grating:
Blaze angle 70.45 deg (R2.82)
Littrow angle 2.75 deg
T 52.7 grooves/rom
Lx W 847.8 x 304.8 rom
(no. gratings = 2; each 406.4 x 304.8 rom, with 35.0rom gap)
[T_blz,T_ove,T_gapJ = [75.7,85.6,4.2J

Cross-dispersion/order separation:
min = 6.9 arcsec, max = 12.2 arcsec
Method - Prism:
Glass = BK7 and BK7
alphaP [41.50,44.50J deg
thetai = [32.95,35.55J deg

Collimator:
F = 3.82, f = 1164.5 rom
(assumes B = 305.0, f/D_tel = 4.2 and FRD rho = 1.10)
Plate scale = 223.5 um/arcsec (4.47 arcsec/rom)

Camera:
F = 2.14, f = 653.1 rom
(assumes 2.0 (15.0um) pixel sampling of R_max
and pupil demagnification cam_dmag = 1.0)
Plate scale 125.6 um/arcsec (7.96 arcsec/rom)
CCD:
Type ccd44_82
Coat E2V DDSi astromid
n_pix [x,yJ [4096,2048J pixels
s_pix 15 microns
num [x,yJ [1,2J (x)
Pixel scale 8.4 pix/arcsec (0.12 arcsec/pix)

Fibre diameter Slice properties Resolving Power Transmission (pc)

microns arcsec width height num R R_tot PSF Geom Total
=============================================================================
400.0 1. 79 400.0 400.0 1 25300 70.5 100.0 70.5
350.0 1.56 200.0 350.0 1 37900 62.2 68.6 42.7
300.0 1.34 100.0 300.0 1 75800 52.4 41.6 21.8
400.0 1. 79 70.0 400.0 1 108400 70.5 22.2 15.6

Assumes ...
Seeing median FWHM = 1.12 arcsecs
146 Appendix A. ECHMOD - a Matlab tool for echelle spectrograph modelling

A.2.3 SALT HRS R2

[ tel, slt, spc J

tel.name 'SALT'
slt.name { 'SALT HRS R2 fibre feed' }

spc.lam.minmax = [ 377 , 950 J*1e9

spc.R.minmax [ 17000 , 80000 J
spc. samp 2.0;
spc.B 0.365
spc. lam. eff { [ J }
spc.nwav =11;
spc.dlam '2fsr' ;

col. name 'SALT HRS R2 collimator'

col.F1 0;

ech.name '53.127'
ech.type 'RGL standard echelle'
ech.num 2 ;
ech.gap 0.035
ech.theta 4.5
ech.dtheta = 0
ech.dgamma = 0 ;

xdp.meth 'prism'
xdp.type 'dbl' ;
xdp.num 1
xdp.glass { 'BK7' 'BK7' } ;
xdp.angle [ 40.0 40.0 J ;
xdp.lam [ 370 370 J *1e-9
xdp.coat = { 'Solgel + MgF2' 'Solgel + MgF2' }
xdp.t [ 150 180 J
xdp.T = [ J

cam. name { 'SALT HRS R2 camera' }

cam.dmag 1

ccd.type { 'ccd44 82' }

ccd.coat { 'E2V astroBB-mid graduated' }
ccd.num [ 3 , 1 J
ccd.psi = [ 2.5 J
ccd.orient = [ 'y' J
ccd.offset = [ 0 , 0 J
A.2. Example input/output files 147

==============SALT HRS R2 Properties============

Telescope: SALT
D = 11.0m, f = 46.20m (f/D = 4.20)
Plate scale = 224.0 um/arcsec (4.46 arcsec/mm)

Beam size, B = 365.0 mm

R_max = 80000.0
Wavelength range: 371.3 to 972.4 nm
(in 35 orders: m = 55 to 21)

Echelle grating:
Blaze angle 63.00 deg (Rl.96)
Littrow angle 4.50 deg
T 87.0 grooves/mm
Lx W 861.0 x 308.0 mm
(no. gratings = 2; each 413.0 x 308.0 mm, with 35.0mm gap)
[T_blz,T_ove,T_gap] = [72.8,85.2,3.9]

Cross-dispersion/order separation:
min = 13.0 arcsec, max = 22.8 arcsec
Method - Prism:
Glass = BK7 and BK7
alphaP [40.00,40.00] deg
thetai = [31.68,31.68] deg

Collimator:
F = 3.82, f = 1393.6 mm
(assumes B = 365.0, f/D_tel 4.2 and FRD rho = 1.10)
Fl = 10.00, fl = 3650.0 mm
Plate scale = 586.6 um/arcsec (1.70 arcsec/mm)

Camera:
F = 1.93, f = 705.9 mm
(assumes 2.0 (15.0um) pixel sampling of R_max
and pupil demagnification cam_dmag = 1.0)
Plate scale 113.4 um/arcsec (8.81 arcsec/mm)
CCD:
Type ccd44 82
Coat E2V astroBB-mid graduated
n_pix [x,y] [2048,4096] pixels
s_pix 15 microns
num [x,y] [3,1] (y)
Pixel scale 7.6 pix/arcsec (0.13 arcsec/pix)

Fibre diameter Slice properties Resolving Power Transmission (pc)

microns arcsec width height num R R_tot PSF Geom Total
=============================================================================
500.0 2.23 500.0 500.0 1 17300 82.2 100.0 82.2
500.0 2.23 160.0 500.0 3 40400 82.2 94.6 77.8
400.0 1. 79 160.0 400.0 2 40400 70.5 87.9 61. 9
300.0 1.34 80.0 300.0 4 80900 52.4 98.5 51. 6
300.0 1.34 80.0 300.0 3 80900 52.4 81.7 42.8

Assumes ...
Seeing median FWHM = 1.12 arcsecs
148 Appendix A. ECHMOD - a Matlab tool for echelle spectrograph modelling

A.2.4 SALT HRS R4

[ tel, slt, spc ]

tel.name 'SALT'
slt.name { 'SALT HRS R4 fibre feed' } ;

spc.lam.minmax = [ [375, 555] [555, 880] ]*le-9

spc.R.minmax [ [17000, 80000] [17000, 100000 ] ]
spc.samp [ 2.0 2.0 ]
spc.B 0.200
spc.lam.eff = { [] [] }

col.name 'SALT HRS R4 collimator' ;

col. Fl 10 ;

ech.type 'RGL standard echelle'

ech.name '53.425'
ech.num = 2 ;
ech.gap 0.016 ;
ech.theta 0.05;
ech.dtheta = 0 ;
ech.dgamma = 2.5 ;

xdp.meth 'grating'
xdp.type 'vph' ;
xdp.glass { 'bk7' 'bk7' }
xdp.angle [ 0 0 ]
xdp.coat { 'mgf2 © 30deg' 'mgf2 © 30deg' }
xdp.lam [ 462 700 ] *ie-9
xdp.T [ 1050 650 ]
xdp.t [ 5 8 ] ;

cam. name { 'HRS-Blue-CM-2-06' 'HRS-Red-CM-2-01' }

cam.dmag [ 1 1.5 ]

ccd.type { 'ccd44 82' 'ccd44_82' }

ccd.coat { 'E2V StdSi astroBB' 'E2V DDSi astromid' }
ccd.num [ [1,1] [2,1] ]
ccd.orient [ 'y' 'y' ]
ccd.psi = [ 0 0 ]
ccd.offset [ [0,0] [0,0] ]
A.2. Example input/output files 149

==============SALT HRS R4 Blue Properties============

Telescope: SALT
D = 11.0m, f = 46.20m (f/D = 4.20)
Plate scale = 224.0 um/arcsec (4.46 arcsec/mm)

Beam size, B = 200.0 mm

R_max = 80000.0
Wavelength range: 373.3 to 555.5 nm
(in 42 orders: m = 125 to 84)

Echelle grating:
Blaze angle 76.00 deg (R4.01)
Littrow angle 0.05 deg
T 41. 6 grooves/mm
Lx W 836.0 x 204.0 mm
(no. gratings = 2; each 410.0 x 204.0 mm, with 16.0mm gap)
[T_blz,T_ove,T_gapJ = [82.2,97.5,2.5J

Cross-dispersion/order separation:
min = 11.2 arcsec, max = 24.3 arcsec
Method - Holographic grating:
Glass bk7
T = 1050 lines/mm
thetai = 14.04 deg (lamB 462.0 nm)
coat = mgf2 © 30deg

Collimator:
F = 3.82, f = 763.6 mm
(assumes B = 200.0, f/D_tel = 4.2 and FRD rho = 1.10)
Fl = 10.00, fl = 2000.0 mm
Plate scale = 586.6 um/arcsec (1.70 arcsec/mm)

Camera:
F = 1.50, f = 300.2 mm
(assumes 2.0 (15.0um) pixel sampling of R_max
and pupil demagnification cam_dmag = 1.0)
Plate scale 88.1 um/arcsec (11.36 arcsec/mm)
CCD:
Type ccd44_82
Coat E2V StdSi astroBB
n_pix [x,yJ [2048,4096J pixels
s_pix 15 microns
num [x,yJ [l,lJ (y)
Pixel scale = 5.9 pix/arcsec (0.17 arcsec/pix)

Fibre diameter Slice properties Resolving Power Transmission (pc)

microns arcsec width height num R R_tot PSF Geom Total
=============================================================================
500.0 2.23 500.0 500.0 1 16400 82.2 100.0 82.2
500.0 2.23 160.0 500.0 3 38400 82.2 81.4 66.9
400.0 1. 79 160.0 400.0 2 38400 70.5 76.1 53.6
300.0 1.34 80.0 300.0 4 76800 52.4 91.4 47.9
300.0 1.34 80.0 300.0 3 76800 52.4 68.5 35.9

Assumes ... Seeing median FWHM = 1.12 arcsecs

150 Appendix A. BeHMon - a Matlab tool for echelle spectrograph modelling

==============SALT HRS R4 Red Properties============

Telescope: SALT
D = 11.0m, f = 46.20m (f/D = 4.20)
Plate scale = 224.0 um/arcsec (4.46 arcsec/mm)

Beam size, B = 200.0 mm

R_max = 100000.0
Wavelength range: 548.9 to 880.4 nm
(in 33 orders: m 85 to 53)

Echelle grating:
Blaze angle 76.00 deg (R4.01)
Littrow angle 0.05 deg
T 41.6 grooves/mm
Lx W 836.0 x 204.0 mm
(no. gratings = 2; each 410.0 x 204.0 mm, with 16.0mm gap)
[T_blz,T_ove,T_gapJ = [75.1,97.5,2.5J

Cross-dispersion/order separation:
min = 10.0 arcsec, max = 25.0 arcsec
Method - Holographic grating:
Glass bk7
T 650 lines/mm
thetai = 13.15 deg (lamB 700.0 nm)
coat = mgf2 @ 30deg

Collimator:
F = 3.82, f 763.6 mm
(assumes B = 200.0, f/D_tel 4.2 and FRD rho = 1.10)
Fl = 10.00, fl = 2000.0 mm
Plate scale = 586.6 um/arcsec (1.70 arcsec/mm)

Camera:
F 1.88, f 250.2 mm
(assumes 2.0 (15.0um) pixel sampling of R_max
and pupil demagnification cam_dmag = 1.5)
Plate scale 110.1 um/arcsec (9.08 arcsec/mm)
CCD:
Type ccd44_82
Coat E2V DDSi astromid
n_pix [x,yJ [2048,4096J pixels
s_pix 15 microns
num [x,yJ [2,lJ (y)
Pixel scale 7.3 pix/arcsec (0.14 arcsec/pix)

Fibre diameter Slice properties Resolving Power Transmission (pc)

microns arcsec width height num R R_tot PSF Geom Total
=============================================================================
500.0 2.23 500.0 500.0 1 16400 82.2 100.0 82.2
500.0 2.23 160.0 500.0 3 38400 82.2 81.4 66.9
400.0 1. 79 160.0 400.0 2 38400 70.5 76.1 53.6
300.0 1. 34 80.0 300.0 4 76800 52.4 91.4 47.9
300.0 1.34 80.0 300.0 3 76800 52.4 68.5 35.9

Assumes ... Seeing median FWHM = 1.12 arcsecs

Appendix B

Optical prescriptions

The optical prescriptions of the spectrographs described in the previous appendix are
given below. Only summary surface data are given. Coordinate breaks and apertures are
not detailed here. Full Zemax prescriptions can be obtained from the author.

151
 >-'
B.1 HERCULES 01
."

Surf Type Comment Radius Thickness Glass Diameter Conic

OBJ STANDARD FIBRE Infinity 0 0.1 0
1 STANDARD Infinity 783.3 0.1 0
STO STANDARD COLLIMATOR -1566.6 -783 MIRROR 210 -1
3 STANDARD FIBRE OBSTRUCTN Infinity -1667 0 0
4 COORDBRK 0
5 COORDBRK ANGLE OF INCIDENC 0
6 STANDARD PRISM FACE 1 Infinity 0 BK7 273.4955 0
7 COORDBRK -129.575
8 COORDBRK 0
9 STANDARD PRISM FACE 2 Infinity 0 273.5579 0
10 COORDBRK ECHELLE/PRISM ANG -300
11 COORDBRK 0
12 DGRATING ECHELLE GRATING Infinity 0 MIRROR 565.5365 0
13 COORDBRK 300
14 COORDBRK 0
15 STANDARD PRISM FACE 3 Infinity 0 BK7 305.7749 0
16 COORDBRK 129.575
17 COORDBRK 0
18 STANDARD PRISM FACE 4 Infinity 0 312.6507 0
19 COORDBRK 0
20 COORDBRK 2461.5
21 COORDBRK 0
22 STANDARD COLL OBSTRUCTION Infinity 0 0 0
23 STANDARD COLL OBSTRUCTION Infinity 0 397.2517 0 >-
'd
'd
24 COORDBRK 113.5 co
::l
0..
25 COORDBRK 0 x'
**26 EVENASPH SCHMIDT CORR. Infinity 15 BK7 406.291 0 ?J
27 STANDARD Infinity 777.679 406.8081 0 0
'd
28 COORDBRK PRE FF HOLE 0 <"I-
r;'
29 STANDARD FOLD-FLAT HOLE Infinity 0 459.1752 0 e:.
...,
'd
30 STANDARD FOLD-FLAT MIRROR Infinity 0 MIRROR 459.1752 0 co
en
31 COORDBRK -859.79 ...,co
32 STANDARD SCHMIDT PRIM. 1946.26 859.79
.s'
MIRROR 495.5923 0 ,,;-
o·
::l
33 COORDBRK 0 en
 tJ:j
34 STANDARD FOLD-FLAT HOLE Infinity 0 103.6216 0 I-'

35 COORDBRK 75.0488 ~
t'.1
36 STANDARD FEILD-FLATTENING 327.6568 10.5 BK7 85 0 ;D
0
~
37 STANDARD 10310.53 25.0008 85 0 t"'
t'.1
38 COORDBRK 0 '"
IMA STANDARD CCD Infinity 50 0
**Surface 26 EVEN ASPHERE
Coeff on r 2 4.76e-006
Coeff on r 4 -6.478798ge-011
Coeff on r 6 -2. 238092ge-017
 t-'
B.2 CELESTIA C1l

"'"

Surf Type Comment Radius Thickness Glass Diameter Conic

oBJ STANDARD Infinity 0 0.35 0
1 STANDARD Infinity 1170 10 0
STo STANDARD COLLIMATOR -2340 0 MIRROR 300 -1
3 CoORDBRK -3000
4 CooRDBRK ANGLE OF INCIDENC 0
5 STANDARD PRISM 1 FACE 1 Infinity 0 BK7 407.6306 0
6 CooRDBRK -124
7 CooRDBRK 0
8 STANDARD PRISM 1 FACE 2 Infinity 0 407.1239 0
9 CooRDBRK -255
10 CooRDBRK 0
11 STANDARD PRISM 2 FACE 1 Infinity 0 BK7 395.178 0
12 CooRDBRK -141
13 CooRDBRK 0
14 STANDARD PRISM 2 FACE 2 Infinity 0 395.2307 0
15 COORDBRK -520
16 CoORDBRK DGAMMA 0
17 DGRATING ECHELLE Infinity 0 MIRROR 1054.307 0
18 CooRDBRK 520
19 CooRDBRK 0
20 STANDARD PRISM 2 FACE 2 Infinity 0 BK7 0 0
21 CooRDBRK 141
22 CooRDBRK 0
23 STANDARD PRISM 2 FACE 1 Infinity 0 0 0 >
1:1
1:1
24 CooRDBRK 255 '"
::J
25 CooRDBRK 0 8:
?<
26 STANDARD PRISM 1 FACE 2 Infinity 0 BK7 595.8322 0 ~
27 CooRDBRK 124 0
1:1
28 CooRDBRK 0 c+
n"
29 STANDARD PRISM 1 FACE 1 Infinity 0 638.7895 0 e:.
1:1
,..,
30 CooRDBRK 3000 '"
CIl

31 STANDARD COLLIMATOR oBSTR Infinity 0 0 0 ,..,

()

>8"
32 CooRDBRK 50 c+
0"
::J
33 CoORDBRK CAMERA CENTRE 0 CIl
 tJj
34 STANDARD Infinity 0 200 0
35 STANDARD CORRECTOR ONE 2794.205 55.59809 BK7 750 0
""
Q
to
36 STANDARD -3444.976 0 750 0 l:"'
to
UJ
37 CooRDBRK 4.205722 >-l
;;
38 CooRDBRK 0
39 STANDARD CORRECTOR TWO 1194.175 24.30712 BK7 730 0
40 STANDARD 665.7248 0 730 0
41 CooRDBRK 1215.911
42 CooRDBRK 0
43 STANDARD CCD OBSTRUCTION Infinity 0 0 0
44 CooRDBRK 825.5883
45 CooRDBRK 0
46 STANDARD PRIMARY MIRROR -1443.243 0 MIRROR 1000 0
47 CooRDBRK -675.5883
48 CooRDBRK 0
49 STANDARD FIELD-FLATTENER -165.21 -33.94388 BK7 170 0
50 STANDARD -680.328 -11.84338 170 0
51 CooRDBRK -5
IMA STANDARD Infinity 0 0
B.3 SALTHRSR2
I-'
01
en

Surf Type Comment Radius Thickness Glass Diameter Conic

OBJ STANDARD Infinity 0 0 0
1 STANDARD Infinity 1340.2 0 0
STO STANDARD COLLIMATOR -2680.4 -2420 MIRROR 351.001 -1
3 COORDBRK THETA (& ECH CEN) 0
4 COORDBRK ANGLE OF INCIDENC 0
5 STANDARD PRISM 1 FACE 1 Infinity 0 BK7 656.5564 0
6 COORDBRK PRISM APEX ANGLE -188
7 COORDBRK 0
8 STANDARD PRISM 1 FACE 2 Infinity 0 637.3285 0
9 COORDBRK -115
10 STANDARD REF Infinity -115 0 0
11 COORDBRK 0
12 STANDARD PRISM 2 FACE 1 Infinity 0 BK7 601.0753 0
13 COORDBRK -165
14 COORDBRK 0
15 STANDARD PRISM 2 FACE 2 Infinity 0 584.2432 0
16 COORDBRK -500
17 COORDBRK BLAZE AND DECENT 0
18 DGRATING ECHELLE Infinity 0 MIRROR 920.8308 0
19 STANDARD ECHELLE GAP Infinity 0 0 0
20 COORDBRK 500
21 COORDBRK 0
22 STANDARD PRISM 2 FACE 3 Infinity 0 BK7 0 0
23 COORDBRK ;:t>
165 >0
>0
24 COORDBRK 0 CD
::J
0.
25 STANDARD PRISM 2 FACE 4 Infinity 0 0 0 x'
26 COORDBRK 115 SO
27 STANDARD REF Infinity 115 0 0 0
>0
28 COORDBRK 0 ~.
(")

29 STANDARD PRISM 1 FACE 3 Infinity 0 BK7 567.1906 0 e

>0
30 COORDBRK 188 '"'
CD

31 COORDBRK 0 '"
(")
::t
>0
32 STANDARD PRISM 1 FACE 4 Infinity 0 595.8706 0 o·'"'"
33 COORDBRK 0 ::J
'"
 to
34 COORDBRK 0 ~
35 COORDBRK 2420
'"
>
36 STANDARD COLLIMATOR OBSTR Infinity 0 0 0 S;
~
37 COORDBRK CAMERA XY OFFSET 50
38 COORDBRK CAMERA ROTATION 0 '"'"
~
tv
39 COORDBRK CAMERA ANG OFFSET 0
40 STANDARD CORRECTOR ONE 5296.991 75 BK7 870 0
41 STANDARD -7872.096 5 870 0
42 STANDARD CORRECTOR TWO 938.3651 75 BK7 860 0
43 STANDARD 744.8469 281.0331 830 0
44 STANDARD CORRECTOR THREE -744.8469 75 BK7 840 0
45 STANDARD -938.3651 400 870 0
46 STANDARD CCD OBSTRUCTION Infinity 876.0455 0 0
47 STANDARD PRIMARY MIRROR -1505.958 0 MIRROR 1200 0
48 COORDBRK -776.0455
49 COORDBRK FF TILT 0
50 STANDARD FIELD FLATTENING -365.2816 -22.09066 BK7 162 0
51 STANDARD LENS 1101.869 0 162 0
52 COORDBRK -3.184869
53 COORDBRK CCD TILT 0
IMA STANDARD Infinity 106.1381 0
B.4 SALTHRSR4
t-'
01
00

Note that the prescription for SALT HRS R4 is actually the 2005 April design.

BA.l SALT HRS R4 - Blue arm

Surf Type Comment Radius Thickness Glass Diameter Conic

OBJ STANDARD INTERMEDIATE INJECT Infinity 534.3569 21.2132 0
1 STANDARD FC1.l 96.81967 15 N-FK51 60 0
2 STANDARD FC1.2 -85.49425 15 S-LAL7 60 0
3 STANDARD -563.0025 10 60 0
4 STANDARD VACUUM WINDOW Infinity 10 SILICA 60 0
5 STANDARD Infinity 10 60 0
6 STANDARD FC2.1 180.6662 15 S-LAL7 50 0
7 STANDARD FC2.2 49.29438 15 N-FK51 50 0
8 STANDARD -212.0699 174.6397 50 0
9 STANDARD "SLIT PLATE" Infinity 0.5 8.596469 0
10 STANDARD TC 1.1 18.0863 3 BK7 12.5 0
11 STANDARD 39.198 -2.53 12.5 0
12 COORDBRK OFF-AXIS ANGLE 0
13 COORDBRK Y DECENTRE 267.47
14 COORDBRK 0
15 COORDBRK 0
16 COORDBRK INJECTION ANGLE 0
17 STANDARD FOLD MIRROR Infinity 0 MIRROR 187.7476 0
18 COORDBRK 0
~
19 COORDBRK 0 '0
'0
20 COORDBRK 267.47 co
i='
0.
21 STANDARD "DIRECT INJECTION" Infinity -267.47 20 0 x·
22 COORDBRK -1735 SO
23 STANDARD Ml PASS 1 4000 2000 MIRROR 800 -1 0
'0
24 COORDBRK GRATING DECENTRE 0 C".
()

25 COORDBRK BLAZE ANGLE 0 e:.

26 COORDBRK THETA 0 ..,co
'0
en
27 COORDBRK ROTATE GROOVES 0
()
::J.
'0
STO DGRATING ECHELLE Infinity 0 MIRROR 887.1472 0 ,.;-
o·
i='
29 COORDBRK UNROTATE 0 en
 b:I
30 COORDBRK UNDO THETA 0 ~
31 COORDBRK BLAZE ANGLE UNDO 0
'"t""'
;,-
32 COORDBRK BACK TO GLOBAL-Z -2000 >-l
~
33 STANDARD Ml PASS 2 4000 0 MIRROR 800 -1 ~

34 STANDARD Ml PARENT Infinity 2000 690 0 '"~

35 COORDBRK 0 """
36 STANDARD INT. IMAGE Infinity 150 187.7652 0
37 STANDARD DICHROIC Infinity 0 MIRROR 201.3004 0
38 STANDARD Infinity -984.8725 152.0777 0
39 COORDBRK FOCUS 0
40 STANDARD 2ND PUPIL MIRROR 2222.222 0 MIRROR 440 0
41 COORDBRK 1075.613
42 COORDBRK DEC. TO WHo PUPIL 0
43 STANDARD WHITE PUPIL Infinity -200 144.5143 0
44 COORDBRK 0
45 STANDARD FOLD MIRROR Infinity 0 MIRROR 176.0127 0
46 COORDBRK -175
47 COORDBRK AOI 0
48 STANDARD VPH LENS 1 -5117.579 -15 BK7 160 0
49 STANDARD VPH LENS 1 Infinity 0 BK7 160 0
50 STANDARD VPH Infinity -10 BK7 160 0
51 DGRATING VPH Infinity -10 BK7 160 0
52 STANDARD VPH Infinity 0 BK7 160 0
53 STANDARD VPH LENS 2 Infinity -15 BK7 160 0
54 STANDARD VPH LENS 2 -5117.579 0 160 0
55 COORDBRK -40
56 COORDBRK CAMERA OFFSET 0
57 STANDARD "APERTURE STOP" Infinity 0 168.0042 0
58 STANDARD BCM1.l -1500 -35 S-FPL51Y 150 -181. 6
59 STANDARD 295.4211 -17 .24081 150 0
60 STANDARD BCM2.1 190.1559 -10 PBM2Y 150 0
61 STANDARD BCM2.2 -143.4802 -40 S-FSL5 180 0
62 STANDARD -1385.66 -1 180 0
63 STANDARD BCM3.1 -275.7727 -50 S-FPL51Y 200 0
64 STANDARD 282.9042 -114.8481 200 0
65 STANDARD BCM4.1 -323.1683 -42.5 PBM2Y 220 0
66 STANDARD 777.002 -1 220 0 >-'
CJl
<0
 >-'
67 STANDARD BCM5.1 -105.2192 -55 PBM2Y 185 0 0>
0

68 STANDARD -79.51249 -18.16142 135 0

69 STANDARD BCM6.1 -104.7826 -55 PBM2Y 130 0
70 STANDARD -196.8763 -17 .5 90 0
71 CooRDBRK 0
72 STANDARD BCM7.1 154.3904 0 SILICA 85 0
73 TOROIDAL Infinity -13 SILICA 95 0
74 TOROIDAL -567.5836 -6.41121 80 0
IMA STANDARD DETECTOR Infinity 70.03411 0

BA.2 SALT HRS R4 - Red arm

Surf Type Comment Radius Thickness Glass Diameter Conic
oBJ STANDARD INTERMEDIATE INJECT Infinity 534.3569 21.2132 0
1 STANDARD FC1.l 96.81967 15 N-FK51 60 0
2 STANDARD FC1.2 -85.49425 15 S-LAL7 60 0
3 STANDARD -563.0025 10 60 0
4 STANDARD VACUUM WINDOW Infinity 10 SILICA 60 0
5 STANDARD Infinity 10 60 0
6 STANDARD FC2.1 180.6662 15 S-LAL7 50 0
7 STANDARD FC2.2 49.29438 15 N-FK51 50 0
8 STANDARD -212.0699 174.6397 50 0
9 STANDARD "SLIT PLATE" Infinity 0.5 8.596829 0
10 STANDARD TC 1.1 18.0863 3 BK7 12.5 0
11 STANDARD 39.198 -2.53 12.5 0
12 CooRDBRK OFF-AXIS ANGLE 0
~
13 CooRDBRK Y DECENTRE 267.47 >0
>0
14 CooRDBRK 0 co
::J
0.
15 CooRDBRK 0 x'
16 CooRDBRK INJECTION ANGLE 0 ~
17 STANDARD FOLD MIRROR Infinity 0 MIRROR 188.858 0 0
>0
18 CooRDBRK 0 ::1",
(")

19 CooRDBRK 0 e:..
>0
...,
20 CooRDBRK 267.47 co
en
21 STANDARD "DIRECT INJECTION" Infinity -267.47 20 0
(")
::l,
>0
22 CooRDBRK -1735 <+
0'
23 STANDARD Ml PASS 1 4000 2000 MIRROR 800 -1 ::J
en
 b:J
24 COORDBRK GRATING DECENTRE 0 ~
25 COORDBRK BLAZE ANGLE 0 en
;.-
26 COORDBRK THETA 0 ~
~
27 COORDBRK ROTATE GROOVES 0 ;Xl
en
STO DGRATING ECHELLE Infinity 0 MIRROR 861.1699 0 ::0
29 COORDBRK UNROTATE 0 """
30 COORDBRK UNDO THETA 0
31 COORDBRK BLAZE ANGLE UNDO 0
32 COORDBRK BACK TO GLOBAL-Z -2000
33 STANDARD M1 PASS 2 4000 0 MIRROR 800 -1
34 STANDARD M1 PARENT Infinity 2000 690 0
35 COORDBRK 0
36 STANDARD INT. IMAGE Infinity 150 304.1606 0
37 STANDARD DICHROIC Infinity 15 F_SILICA 319.3556 0
38 STANDARD Infinity 988.0404 212.0425 0
39 COORDBRK FOCUS 0
40 STANDARD 2ND PUPIL MIRROR -2222.222 0 MIRROR 440 0
41 COORDBRK -1050
42 COORDBRK DEC. TO WHo PUPIL 0
43 STANDARD WHITE PUPIL Infinity 150 147.716 0
44 COORDBRK 0
45 STANDARD FOLD MIRROR Infinity 0 MIRROR 204.4704 0
46 COORDBRK 125
47 COORDBRK AOI 0
48 STANDARD VPH LENS 1 4300.759 15 BK7 160 0
49 STANDARD VPH LENS 1 Infinity 0 BK7 160 0
50 STANDARD VPH Infinity 10 BK7 160 0
51 DGRATING VPH Infinity 10 BK7 160 0
52 STANDARD VPH Infinity 0 BK7 160 0
53 STANDARD VPH LENS 2 Infinity 15 BK7 160 0
54 STANDARD VPH LENS 2 4300.759 0 160 0
55 COORDBRK 40
56 COORDBRK CAMERA OFFSET 0
57 STANDARD "APERTURE STOP" Infinity 0 161. 9181 0
58 STANDARD BCM1.1 Infinity 0 200 0
59 STANDARD Infinity 0 200 0
60 STANDARD RCM1.1 309.3243 60 S-FSL5 170 -2.763853 >-'
0>
>-'
 f-'
61 STANDARD RCM1.2 -151.3385 20.5 S-TIHl 170 0 0'>
I:..:>

62 STANDARD 415.2096 57.31691 170 0

63 STANDARD RCM2.1 1284.695 55 S-BAH11 220 0
64 STANDARD -242.0536 47.51831 220 0
65 STANDARD RCM3.1 148.8329 55 BK7 210 0
66 STANDARD 652.7292 84.59567 210 0
67 STANDARD RCM4.1 199.7302 40 S-BAH11 140 0
68 STANDARD 261.3614 25 120 0
69 STANDARD BCM6.1 Infinity 0 150 0
70 STANDARD Infinity 0 150 0
71 COORDBRK 0
72 STANDARD RCM5.1 -126.7565 0 SILICA 105 0
73 TOROIDAL Infinity 13 SILICA 115 0
74 TOROIDAL 693.3784 8.341316 100 0
IMA STANDARD DETECTOR Infinity 84.55915 0
Appendix C

HERCULES observing manual

C.1 Initializing HERCULES

C.1.1 Before observing begins

• Check the CCD supply dewar (located in the main corridor outside the HERCULES
room) is at least 30% full. This will allow for at least 2 automatic fills of the CCD
dewar. These are usually scheduled to occur at 6 am and 6 pm each day. To refill
the dewar:

- Unscrew the black plug on the dewar lid.

- Insert the filling funnel.
- Fill the dewar from a 10 or 20 litre dewar.

Carry out a manual fill immediately afterward to check that there are no leaks and
the filling system is working correctly. The CCD control box temperature reading
should be approximately -85°C. The display may be viewed by removing the small
cover directly above the supply dewar between the CCD controller vents.

• Make certain all HERCULES room lights are off and that there are no light leaks into
the HERCULES room.

• Turn on the exposure meter PMT power supply. This is located on a shelf in the
data room beside the computer Octans. NEVER switch the HERCULES room lights
on while the PMT is switched on. Check that the voltage is set to 1125 V.

e Turn on the exposure meter electronics (the red switch marked MAINS on the box
with the exposure meter analogue display).

C.1.2 Initializing the HERCULES fibre-feed control

The HERCULES fibre-feed control software operates on the PC running Windows situated
beside Hydrus (the Linux computer on which the CCD controller runs). This software
controls all aspects of the fibre-feed module including the shutter, the exposure meter,
and auto-guiding. It also monitors the temperature and pressure inside the HERCULES
spectrograph. To start the HERCULES fibre-feed control:

• In the dome:

- Ensure that the fibre-feed module is plugged into the telescope pier and turned
on.

163
164 Appendix C. HERCULES observing manual

- Ensure that the telescope drive speed on the main telescope controls (in a small
drawer under the displays) is set to the lowest speed (1). This determines the
auto-guide slewing rate and is also the manual telescope control on the fibre-
feed control.
- Turn on the thorium-argon lamp power supply. The Photron power supply is
located next to the fibre-feed module on the telescope.
1. Before turning on ensure that the current is zero. The current controller
should be fully turned anti-clockwise.
2. Turn mains switch on.
3. Adjust current control to 7mA.
(Note: The Photron power supply replaces the original power supply. The
small switch by the green LED is now redundant.) The power supply should
be turned on an hour or so before observations begin in order to ensure that
the lamp has stabilized.
- Turn on the image intensifier. There is a small switch on the fibre-feed module
by a red LED.
- The only variable control on the fibre-feed module is the fibre number. Choose
between:
* fibre #1 - a 100 /-Lm fibre,
* fibre #2 - a 50 /-Lm fibre, and
* fibre #3 - a 100 /-Lm fibre with a 50/-Lm micro-slit.
See Section C.2.3 for details on the fibres.

• In the control room:

- To start the HERCULES fibre-feed control double-click on the icon named

Hercules fibre-feed control. The initialization process takes a few sec-
onds. The graphical user interface shown in Figure C.l should now appear.
Figure C.l: The HERCULES fibre-feed control graphical user interface.
166 Appendix C. HERCULES observing manual

C.1.3 Using MoJo

The Mount John CCD image acquisition software runs only on the Linux machine Hy-
drus. First start an Xwindows session (i.e., logon) then type moj 0 at a command prompt.
Three windows (see Figure C.2) will open:
1. the main control panel,

2. the instrument control panel, and

3. a message window.

Figure C.2: The Mojo control panel.

In the main control panel the observer first needs to set the instrument parameters. The
observer should do the following:
.. Enter the observers initials or name in the Observer box .

• The Site should be Mount John University Observatory.

• Set the Telescope to MJUO McLellan 1 m f / 13 . 5.

C.l. Initializing HERCULES 167

.. Set the Instrument to the HERCULES spectrograph. In the HERCULES instrument

control panel the following information must be manually entered:

1. CCD position (see Section C.2.1)

2. CeD focus. See Appendix C.5 for instructions on obtaining the best focus .

.. The only Detector available is the Site Si003ab. Set the Gain to 4. This
is the appropriate setting for spectroscopy. The inverse gain at this setting is
1.22 e- / ADD, and the base level noise is 2.59 e-.

.. The External Command should be enabled and set to the system command
getheaderfiles [obs] , where Cobs] should be the same as entered in the Ob-
server box.

Next, the Exposure type should be set. The possibilities are:

.. The exposure can be either Timed or Bulb. The latter allows the user to stop
the exposure and read-out the CCD at any time. This is particularly useful when
using the HERCULES spectrograph as the observer may stop the exposure when the
exposure-meter records sufficient flux .

.. The possible exposure types are bias, dark, th_arc, smooth_field, and object.
To ensure that the fits image headers are meaningful the object type should be
changed on Mojo when the HERCULES object is changed.

The Target can be set while t~e Exposure Type is set to object. The system Mojo
target list may be found at

/usr/local/mojo-[latest_version_numberJ/targets.dat.

The targets. dat file contains the following objects with declinations less than +25°:

1. 1208 stars with Bayer designations (e.g. f3 Corvi)

2. 1612 stars with Flamsteed numbers (e.g. 9 Corvi)

3. 6431 HR stars

4. 28763 HD stars

When entering these stars into Mojo remember always to use the underscore C) in the
star names. E.g., 9_crv, hr _4786, or hd_109379. The Bayer names (Greek letters) in
the star list are up to three characters long, so that, for example, alp should be used
for "alpha", bet for "beta", etc. The constellations are given as standard three-letter
abbreviations. The search is case insensitive. Additional stars can be included in the
observers personal targets. dat file. The mo j 0 . rc file should then be edited to show the
location of this file. For example,

Targets: :-/.mojo/targets.dat
168 Appendix C. HERCULES observing manual

Be sure that the format of your personal targets file is identical to the system targets. dat
file.

To start an exposure press Start. This sends a command to the fibre-feed control to
open the shutter and to start the exposure meter. If the exposure type is Timed then the
exposure will continue for the Exposure Duration or until Stop or Abort is pressed.
Otherwise a Bulb image will continue until Stop or Abort is pressed. If the Display
Image option is on then after the readout the Mojo image window will display the current
image (see Figure C.3). An aborted exposure will not be read out. To save an image click
ARCHIVE. Archiving is complete when the ARCHIVE letters turn grey. It is also
possible to Automatically Archive.

Figure C.3: The Mojo

image display.
C.2. Observing 169

C.2 0 bserving
C.2.l CCD position
The CCD position is the only user adjustment inside the HERCULES spectrograph room.
There are presently four discrete positions for the CCD which are defined by the brackets
for the CCD cradle (see Figure C.4). To reposition the CCD remove the locking nut on
the front of the cradle and lift the cradle into the appropriate slot. Be sure that all three
locating pins have entered the same slot. This will be easier if the CCD cradle is held
steady and level whilst being inserted. The CCD should not be shut down or have the
electronics plug removed. Note that it should not be necessary to refocus if the CCD has
been repositioned, but the observer would be prudent to check this. Spectra taken with
the CCD in each of the four regions may be viewed at:
http://www.phys.canterbury.ac.nz/research/astronomy/hercules/CCD_regions.shtml

Figure C.4: The CCD

positioning template. The
three locating pins on the
COD cradle can be posi-
tioned in one of the four
numbered slots. The direc-
tions of main dispersion and
cross-dispersion are shown

C.2.2 CCD focus

The CCD focus control is in the data room near the computer Hydrus. See Appendix C.5
for instructions on how to focus HERCULES.

C.2.3 Fibre choice

There are three different fibres are available on HERCULES. Changing the fibre type will
give the resolving powers given in Table C.1. The relative throughput of the three fibres

Fibre no. Fibre type Image scale Resolving power

arcsecs )../0)..
1 100 p,m 4.50 41000
2 50p,m 2.25 82000
3 100 p,m with 50p,m microslit 4.50 70000

Table C.l: Fibre type and resolving power

is shown in Figure C.6. When the highest resolving powers is desired the choice between
fibres "2" and "3" will depend upon the seeing (see section C.3.1 for instructions on how
to compute the seeing). It can be seen that if the seeing is better than about 2" then it
would be better to use the 50 p,m fibre. If the seeing is worse than this then the 100 p,m
with 50p,m microslit will give superior throughput for about the same effective resolving
170 Appendix C. HERCULES observing manual

(a) Fibre 1 (b) Fibre 2 (c) Fibre 3

Figure C.5: The observer can choose between three fibres; a lOOl1m fibre (fibre 1), a 50 11m fibre (fibre
2), and a lOOl1m fibre with a 50l1m micro-slit (fibre 3)

power. Because the exits of all three fibres are in the same plane it is not necessary to
refocus the spectrograph every time the fibre is. changed however the telescope may need
to be repositioned and/or refocused.

100~--'-----'-----'--r==~=====c====~
- 100)..tm
90 .--_. 100)..tm with 50)..tm slit
- 50)..tm
80

~ 70
o
'-'

'5
0.
60
..c
g> 50
o
....
..c
Figure C.6: [
~ 40
.0 The tllTOughput of the
iL 30 HERCULES nbres.] The
throughput of the
20
three HERCULES n-
10 bres. Entrance, exit
and absorption losses
OL---~------~------~------~------~------~
1 234 5 6 are a.ll a.ccounted for.
Seeing fwhm (arcsecs)
C.2. Observing 171

C.2.4 Calibration spectra

White-light spectra
Set the turntable position on the HERCULES fibre-feed to "WHITE". This will automat-
ically turn on the white lamp. Individual observers will require a different set of white
light images depending on what their requirements are. For instance, with the CCD in
positions "1", "2", and "3" at least different exposure lengths will be required to record
a reasonable signal in each third of the CCD. The ratio of these exposures will be of
the order 1:3.3:13.3. The red orders will become increasingly saturated as the exposure
time increases, but these orders will not be used when fiat-fielding. For the purposes of
fiat-fielding the observer should take at least 9 exposures at each exposure time. However,
it should be noted that a single white light spectra suffices for order definition and for
approximately normalizing the extracted spectra.
Now on MoJo, select smooth as the "Object", set the "Exposure Type" to timed and
enter the required exposure duration. Click "Start" to begin the exposure. Archive the
resulting image. It may be more convenient to use MoJo's "SEQUENCE" function and
archive the images automatically.

Thorium-argon spectra
The thorium-argon lamp should already have been turned on at the fibre-feed module.
(Note that because the power supply is now external the calibration lamp state will not
be displayed on the fibre-feed control.) Set the turntable position to "THORIUM". Again
users will have to determine the optimum length of exposure. When the CCD is in position
"2" , the MIDAS command

@c check_thorium [thorium file name]

can be used to compute the ideal exposure length. The purpose of this program is to
ensure that every thorium image is exposed for a similar duration. It also monitors
changes in the output of the thorium lamp which has been observed to change because of
some unknown cause.
Now on MoJo set the "Exposure Type" to timed th_arc and enter the required
exposure duration. Click "Start" to begin your exposure. Archive the resulting image.

C.2.5 Stellar spectra

The guide and acquisition camera
The guide and acquisition camera works by inserting a diagonal mirror or beam-splitter
into the telescope beam which directs some or all of the light to an image intensified
camera. The acquisition and guide camera has two modes of operation:

1. "COARSE" (f/S.5) - FOV = 5.3' X 3.7' and

2. "FINE" (f 121.9) - FOV = 2.1' X 1.4'.

The auto-guiding routines will work in either mode.

172 Appendix C. HERCULES observing manual

The exposure meter

The exposure meter uses a small fraction of light that would otherwise be lost to the hole
in the flat fold-mirror of the camera. A diagonal mirror and lens array directs the light to
a PMT situated on the outside of the HERCULES tank. It is very useful to have some idea
of the total fl. ux the exposure meter needs to record in order to reach a S / N ratio which
satisfies the observer's requirements. This can be easily determined by observing a bright
star. With the CCD in position "2" an accumulated count of approximately 1.4 x 10 6 is
sufficient to give a S / N of 100 in the middle of the spectrum for K and G type stars. Note
that the PMT is very sensitive to small changes in voltage and this value may change.

Locating the fibre

It is not possible simultaneously to view both the star and the fibre input. Therefore the
fibre-feed module contains a set of transfer lenses which rei mage the fibre input onto the
camera. A set of LEDs inside the spectrograph provide the back-light for this operation.
To find the approximate location of the fibre set the turntable position to "LED". This
will automatically turn on the LEDs and set the camera position to "FINE". The fibre
should now be visible. Mark the fibre location by clicking "Centre cross on object" then
dragging a box over the fibre image and clicking "OK". The box will be centred over the
fibre.

Acquiring a star
First, in MoJo, enter the "Target" name. This is useful to do first because MoJo will
display the R.A. and Dec. of the star, as well as its airmass. To acquire a new star set
the HERCULES camera mode to "COARSE" and drive the telescope to the appropriate
coordinates. The pointing of the McLellan telescope is not perfect and a small offset
must be used to locate an object. This offset can easily be determined by observing a
bright nearby star. Generally, R.A. is the most reliable adjustment - point the telescope
approximately 20 s west of true R.A. Declination is more uncertain but observers note that
it is usually 3 to 7' north of the true declination.

IMPORTANT: If the movement of the telescope is relatively large the ob-

server should do this while in the dome in order to avoid damaging the fibre.
Under NO circumstances should the observer drive the telescope more than
once around the polar axis! Not only does this have the potential to damage
the fibre, but all the electronic connections in the pier will be destroyed. A
small box in the data room under the dome window indicates the telescope's
relative position. It has 5 LEDs of which the central one is green and the
others are red. If any of the red LEDs are on then the telescope needs to
be driven in the opposite direction to unwind the fibre and electronics. The
observer should do this while in the dome!

Once a star has been roughly acquired and focused (using the image on the screen),
the camera mode may be set to "FINE" so as to precisely place the star. However, it is
common practice to use the "COARSE" setting for both acquisition and guiding. The
"Filter wheel setting" may need to be changed to brighten or darken the image.
C.2. Observing 173

Although the relative alignment of the fibre image and the star image should be
sufficiently accurate that starlight will now enter the spectrograph it is possible to improve
the pointing of the telescope. This is done using the exposure meter. The following steps
are necessary:

• Set fibre-feed position to "BEAM SPLITTER".

• Using the fibre-feed controls only, start a manual exposure. Be sure to set the guide
mode to "MANUAL".

• While watching either the analogue meter or the display on the fibre feed control
make small adjustments in RA and DEC in order to maximize the light entering
the spectrograph. The observer should make adjustments in one axis at a time in
order to maximize the count rate.

• It may also be necessary to refocus the telescope. Be warned that this will also
move the stellar image. The previous step will then have to be repeated.

• When the flux has been maximized click "Centre Cross On Object". Drag a drag
the box over the star and click "OKAY". The guiding box will centre itself on the
image.

• Be sure to stop the manual exposure before acquiring a spectrum.

C.2.6 Guiding
Initially the beam-splitter was not installed and acquisition and guiding were performed
using the diagonal mirror (i.e., the turntable position "CAMERA"). The installation of
a beam-splitter now means that acquisition of an object and guiding during exposures
both use the same turntable position. In practice the observer will find this the most
useful mode for guiding on all but the faintest stars (mv > 10). However, for the sake of
completeness, the following will describe both the original "INTERMITTENT" guiding
(which uses only the "CAMERA" turntable position) and "CONTINUOUS" (which will
be used with the "BEAM SPLITTER" ).

Manual guiding
With some patience it is possible to guide a star manually. First the star should be located
using the method described above. After making the necessary calibration images the
star should be recentred using the procedure described above. Set turntable position to
"BEAMSPLITTER" or "FIBRE DIRECT" and the guide mode to "MANUAL". Note
that if the latter option is used, it will not be possible to see the star.
Decide whether you want the Mojo exposure type to be "Timed" or "Bulb". If the
exposure type is "Bulb" the exposure should end only when the desired exposure meter
count has been reached. Click "Start" to begin the exposure. Watch the exposure meter
and from time to time make small adjustments in order to maximize the count rate. It
should only be necessary to make adjustments every minute or two. It will probably be
the case that corrections will always be in the same direction indicating that the tracking
rate of the telescope is not perfect and/or that uncorrected telescope flexure is present.
174 Appendix C. HERCULES observing manual

Continuous guiding

Again, centre the star as described in section C.2.5. The box should be set to a minimum
size as it is not possible for the star to drift outside its bounds. Set the turntable position
to "BEAMSPLITTER" and the guiding mode to "CONTINUOUS". Guiding will begin
immediately and will continue until the guide mode is set to manual or the guide star is
lost.
Next, the guide parameters should be adjusted. The integration time changes the
number of images that are co-added before an auto-guiding correction is computed. The
integration time should be more in times of bad seeing in order to average out large shifts
in the centroid however 1 or 2 seconds is generally sufficient. Note that the filter wheel
should be set at the lowest possible setting that still allows the star to be seen. This is
to avoid saturating the camera, which could affect the centroid accuracy. The guiding
accuracy indicates the tolerance on the centroiding. A correction of the telescope position
will be made if the current centroid is greater than the stated number of pixels from the
centre of the guide box. This number should be less in good seeing and more in poor
seeing. Typically 2 to 4 pixels (each 0.2/1) is sufficient. A green light appears in the
integration time window to indicate that the camera is exposing. An adjacent red light
indicates that the telescope is moving.
Click "Start" on Mojo to begin the exposure.

Intermittent guiding
The star should be centred as described in Section C.2.5 except that the fine-tuning of the
star's location should be done with the turntable on "FIBRE DIRECT" and the marking
of the guide box with the turntable on "CAMERA". The guide mode should be set to
"INTERMITTENT" and before an exposure begins the turntable should be returned to
"CAMERA".
Again, adjust the integration time and guiding accuracy. During intermittent guiding
the turntable will periodically move from "FIBRE DIRECT" to "CAMERA" and if nec-
essary an auto-guiding correction will be made. This correction will take a few seconds
and the guide interval is the interval between successive corrections. If the telescope is
tracking well then this interval could be anywhere from one to several minutes. A lower
limit of 30 seconds is possible.
Click "Start" on Mojo to begin the exposure.

C.3 Miscellaneous additional information

C.3.1 Computing atmospheric seeing
It is possible to use the fibre-feed module's guide camera to compute the seeing. To do
this:

• Select "BEAMSPLITTER" or "CAMERA" and set the lens position to "FINE".

Find a star. Choose a filter which prevents the star from saturating the camera .

• Click "Centre cross on object" and drag a box over the star. Click "OK".
C.3. Miscellaneous additional information 175

• Set the integration time to several seconds to compute the average or integrated
seeing.

• Click "Save Image". The button is in the top left hand corner of the guide camera
display. The image will be saved as /rnnt/herc/IMAGEFILE.DAT. The pixel scale of
this image is 0.205".

• Finally, at a command prompt on Hydrus type:

cornpute_hercules_seeing
The terminal will display the FWHM of a fitted 2-dimensional gaussian. The cen-
tral position and x and y FWMH values will also be displayed. The average see-
ing for that night should be recorded in the observer's log book which is found
in the data room. If the profile (which is displayed briefly and saved to the file
hercules_seeing. ps) appears flattened about the centre then the camera has been
saturated. Either change the filter and/or switch from/to "BEAM SPLITTER" or
"CAMERA" and try again. Otherwise, try a dimmer star.

Alternatively the observer may wish to save the image file /rnnt/herc/IMAGEFILE. DAT to
a permanent location; e.g.,
cp /rnnt/herc/IMAGEFILE.DAT [filename.dat].
This image may also be used to compute the "seeing". Simply type:
cornpute_hercules_seeing [filename]
where the filename is entered without the . dat extension.

C.3.2 HERCULES log files

The Hercules computer will supply at the end of an exposure the following three files:

1. /rnnt/herc/HercHeaderFile. dat
This file contains all the information about the status of the fibre-feed controller at
the end of each exposure. An example of such a file follows:

STARTED: :15:23:41 UT
STOPPED: :16:11:42 UT
ExposureType: : Stellar
Fibre type::1 (100-MICRON NO SLIT)
Temp in deg C: : Collimator 10.95 : :Echelle 11.95 ::Camera 13.85 ...
Tank pressure: :2.6 mm
Exposure meter flux::1023707
Mean count: : 356
Flux-weighted mid-exposure: :1416.681

2. /rnnt/herc/HercExposureFiles.dat
This file contains a continuous log of the exposure meter readings every second.

3. /rnnt/herc/HercTernperatureFile.dat
This file contains a continuous log of the temperatures.

The following two subsections describe how the information contained in these files is
saved for later use.
176 Appendix C. HERCULES observing manual

The HERCULES FITS header

The images produced by MoJo are in the standard FITS (Flexible Image Transport Sys-
tem) format. See NASA's HEASARC webpage
http://heasarc.gsfc.nasa.gov/docs/heasarc/fits_overview.html
for detailed information on the FITS image format. Of relevance here is the fact that
every FITS image contains an ASCII header unit which has the general form:
KEYNAME = value / comment string

This is known as the "FITS header". MoJo currently fills the FITS header of each image
with a variety of information about the current image, including some details about the
instrument on which they were captured. The majority of the information contained in
the file /mnt/herc/HercHeaderFile. dat is automatically read by MoJo into the current
image's FITS header. An example of the how the above file would be included in the
FITS header of the current image follows:
INSTRUME= 'HERCULES' / Hercules
HERCUTC1= '15:23:41' / Hercules START UTC
HERCUTC2= '16:11:42' / Hercules STOP UTC
HERCEXPT= 'STELLAR' / Hercules Exposure Type
HERCFIB = '1 (100-MICRON NO SLIT)' / Hercules Fibre 1/2/3
HERCT1 = 1.0950000000000E+01 / Hercules Temperature Collimator
HERCT2 1. 1950000000000E+01 / Hercules Temperature Echelle
HERCT3 = 1.3850000000000E+01 / Hercules Temperature Camera
HERCT4 O.OOOOOOOOOOOOOE+OO / Hercules Temperature Spare_i
HERCT5 = 1.3550000000000E+Oi / Hercules Temperature Room_Nth
HERCT6 1.2350000000000E+01 / Hercules Temperature Room_Mid
HERCT7 1.2150000000000E+01 / Hercules Temperature Room_Sth
HERCT8 O.OOOOOOOOOOOOOE+OO / Hercules Temperature Spare_2
HERCP 2.5000000000000E+00 / Hercules Pressure (mm)'
HERCFTC = 1023707 / Hercules flux meter total counts
HERCFMC = 356 / Hercules mean count
HERCFWMT= 1.4166810000000E+03 / Hercules flux-weighted mean exp time (mins)
The data manually entered in the "Hercules instrument controller" window (see Section
C.1.3) are also included. That is the following two fields are also found in the FITS
header:
HERCCCDP= 2 / Hercules CCD Position [1,2,3 or 4J
HERCCCDF= 830 / Hercules CCD Focus (mm)
The FITS header of an image can be viewed in a number of ways. A simple way to view
the FITS header is to use the unix command
more [filename] .fit
Another way is to use Gaia and choose "View" then "fits header" .

Archiving header files

It is useful to retain copies of the exposure meter and temperature log files as they are
NOT automatically saved. All users should be encouraged to follow the following few
instructions in order to ensure that all these files will be available for quick and easy
examination by anyone who wishes to examine the performance of HERCULES.
It is possible to archive these header files by running a short shell script at the end of
each exposure. This can be done automatically as follows:
C.3. Miscellaneous additional information 177

• Turn ON Mojo's external command

• Enter the following command in the space provided:

getheaderfiles [observers initials] [yin]
where the observers initials must match those entered in Mojo's "Observer" field.
Enter n if "Automatic Archiving" is turned OFF and y if it is turned ON.

All three files will now be saved to the directories:

/dos/d/mjuo/si003ab/[ccyymmdd]/Headerfiles_[yymmccdd]
and
/dos/e/[obs]/si003ab/[ccyymmdd]/Headerfiles_[ccyymmdd]
where Cobs] are the observers initials and [ccyymmdd] specifies that night's archive
directory. The files will also be named according to their parent archive image. That
is, they will have the names:

• HercHeaderFile.dat -+ [prefix] [file#]_header.dat

• HercExposureFiles.dat -+ [prefix] [file#]_exp.dat

• HercTemperatureFile.dat -+ [prefix] [file#]_temp.dat

where [prefix] is the night's image prefix (e.g., f2078) and [file#] are the successive
archived file numbers (e.g., 001, 002, etc).

Monitoring header files

The information contained in the header files for a particular night can be examined by
using the command:
monitorhercules [ccyy] [mm] [dd]
This information will be displayed on the observer's terminal. Use the Unix pipe command
>! to direct the output to a file; i.e.,
moni torhercules [ccyy] [mm] [dd] > ! [f ilename . dat] .
178 Appendix C. HERCULES observing manual

C.4 Trouble shooting

Note: These trouble shooting notes relate only to problems encountered (so far!) with
HERCULES. Check Alan Gilmore's trouble shooting orange pages in the I-metre Infor-
mation Folders in the data room and dome for more general telescope/dome problems.
Please report any additional problems you encounter and potential solutions to them.

C.4.1 The CCD dark readout is not what was expected

The following problems may have occurred (see C.4.2 and C.4.3 for other CCD problems):
• The CCD is completely saturated.
If the chip has been saturated by bright light a few darks may be needed to clear it.
Check the SITe control box temperature is about -94.5°C and that the dewar is not
empty.
Check that the fibre back-light LEDs have not remained on, even though the HER-
CULES status shows they are off. This known problem is solved by turning the
LEDs on and off.
If all else fails, try reinitializing the fibre-feed control or turning off the SITe control
box for a few minutes.
NEVER unplug the CCD electronics from the dewar in the hope that it may solve
a readout problem. The plug is already damaged and a careless observer may break
it entirely.
• A bias or short exposure has diagonal waves of light and dark across it.
This is apparently due to some sort of mains power interference. The observation is
that it is not a significant source of noise (try plotting a cross-section of an image).
• The chip has irregular darker vertical bands through it.
Sometimes this is cleared by turning off the SITe control box for a minute or so. It
doesn't seem to affect the chip pixel dark values nor have any effect on subsequent
spectra.

C.4.2 CCD is contaminated with thorium or white light

After taking a stellar exposure, make certain that Mojo has successfully closed the HER-
CULES shutter before changing the turntable to another position. This is because the
shutter is actually located in the fibre-feed module and not at the CCD.
Check that the shutter state starts at "CLOSED" and that it then opens and closes
when a Mojo exposure begins and ends.
Ensure that a "Manual Exposure" has not been previously started on the Hercules
control.

C.4.3 Stellar CCD signal is not what expected from exposure meter counts
If the CCD signal is more than you expected then why are you complaining? Are you
sure you have the right star? Check the telescope's pointing offset against a nearby bright
isolated star. Otherwise, check for cloud (especially light cloud) and check fibre centring.
CA. Trouble shooting 179

C.4.4 Exposure times much longer than expected

Check you have chosen the correct fibre. The fibres look noticeably different when using
"LED". Check that the guide-box has been centred for maximum counts. Check the dome
rotation, cloud, fog, poor seeing.

C.4.5 Hercules screen locks

This has been observed to occur when either

• an intermittent auto-guide sequence being interfered with an attempt by the ob-

server to do something, such as the observer asking for an exposure to stop or

• occasionally when a manual exposure start/stop sequence on HERCULES is not prop-

erly completed.

The fibre feed status (top left corner) will probably show "BUSY".
Check all the status markers (black dots and red lights) are correctly paired off. Set
the guide mode to "MANUAL" then fix any un-matched black dots and red lights by
clicking on the black dot command line (this is what the fibre-feed control is trying to
do).
When fixed, the status should be "READY". If not, reinitialize the fibre-feed control.

C.4.6 Filter wheel or turn-table out of alignment.

The filter wheel may occasionally get out of alignment so that an incorrect filter is been
used. The only solution to this is to reinitialize the fibre-feed control.
It has also been observed that the mirror turret gets slightly out of alignment. This
causes significant problems when using intermittent guiding. It would be better to use
manual or continuous guiding only until the problem is fixed. File a fault report to have
this done.

C.4.7 The fibre-feed control is acting "strangely" ...

If anything genuinely puzzling is happening it will probably be solved by exiting the
fibre-feed control software and restarting it. If this fails, try exiting the software then
switching the fibre-feed module off at the mains (by unplugging the black power cable
from the white power box). Then restart the fibre-feed control.
If you hear a noise continuously from the fibre-feed module (meaning the mirror
turntable or camera lenses are rotating continuously) switch off the module at the tele-
scope (or pull out the plug) and restart the fibre-feed control.

C.4.8 Auto-guide fails

If a previous good exposure meter count rate suddenly drops following an auto-guide then
one of the following has probably occurred:

• The telescope slew speed on the control under the dome desk is not on speed "1".

• The star has moved from within the guide box.

180 Appendix C. HERCULES observing manual

• The "scope relative to pier" is on the wrong setting; i.e., east or west .

• The mirror turret may be misaligned. See C.4.6 above.

It may also be the case that the star was too faint because of cloud etc. Change the
camera filter to a lighter filter if this is possible. Generally, if you can't see the star (with
the beam-splitter) then it will probably be too faint to get a reasonable signal. Otherwise,
if you wish, try manual or intermittent guiding using the direct "CAMERA" position.

C.4.9 Exposure meter dark count high

Check for any sources of light either in the dome or in the HERCULES room. Any lights
along the fibre's path should be off (or minimized). This includes the data room, passage-
way, and dome. Note that the fluorescent lights continue to glow for some minutes after
they are switched on. Unless you really have to - don't switch them on while observing.
You should switch the PMT off first.
C.5. Focusing HERCULES 181

C.5 Focusing HERCULES

C.5.1 Introduction
A procedure has been written by J. Skuljan to enable observers to determine the best
CCD focus. The following description was initially written by J. Skuljan.

C.5.2 The focuser

HERCULES is focused by adjusting the location of the CCD cradle. The focus control is
in the data room near the Hercules control computer and the focus display is on the wall
above this same computer. The control has three speed settings:

1. slow,

2. medium, and

3. fast.
The display is in millimeters, however, for historical reasons, the following discussion will
assume that 1 unit equal 0.01 mm.

C.5.3 Collecting the images

In order to collect a series of thorium images suitable for focusing; do the following:

1. On Hercules computer:

• Make sure that the thorium lamp is on (the switch is located on the fibre-feed
module on the telescope, next to the image intensifier switch).
It Choose a fibre. Check that the correct fibre number is displayed. (Note that
the actual choice of fibre is unimportant as they are all located at the same
focal plane.)
• Select "THORIUM"

2. On Hydrus computer:

.. Start Mojo.
• Check your usual settings (observer name, instrument, focal ratio, gain) .
.. Select OBJECT, and then set the target name to focus (this is only to keep
the FITS headers tidy - target name is optional).
• Select TH-ARC.
• Set exposure to "Timed", "5-sec".
II Set "Full Chip" to OFF.
ell Select a 200 x 200 sub-frame around the centre of the CCD chip:
Pixels: Start: Binning: OverScan:

~ I~
Columns: 200 400
Rows: 1
200 1
400 1
182 Appendix C. HERCULES observing manual

A larger frame may be also be useful. However, read-out times will be longer.
• Finally, collect a series of images at different focus positions. This may be
done as follows. The focus should be set to about 50 units (0.5 mm) below
the expected value. For example, if you expect the best focus to be at 570,
start from 520. Step through the focus position by 10 units while exposing
and archiving the images. Take e.g. 11 images, so that you end up at 50 units
above the expected value (e.g. if you start from 520, stop at 620). The focus
position MUST increase linearly for each successive image.

C.5.4 Running focus_hercules

Once you have collected the thorium images (as described above), open an X-terminal on
Hydrus and change the directory to your usual working directory. Then copy the FITS
files from your observing archive. If your observer name is obsname, and if the current
date is 31 May 2001, then you can do something like this:

cp /dos/e/obsname/SI003AB/20010531/*.fit .

Alternatively, switch to this directory and run the command there; i.e:

cd /dos/e/obsname/SI003AB/20010531/ .

The procedure to compute the best focus may be run from a command line as follows

focus_hercules [P1] [P2] [P3] [P4] [P5] [P6]

where the six input parameters are:

• P1 - Image prefix, i.e. first five characters of the FITS file name (e.g. f2061).

• P2 - First image number.

e P3 - Last image number.

• P4 - First focus position (in units of 0.01 mm).

• P5 - Focus step (in units of 0.01 mm).

• P6 - 'Reference' image number. This will normally be an image in the middle of the
sequence, where you expect the focus to be good. 'Whatever number you specify, the
corresponding image will be displayed first, and you will have to choose an isolated
spectral line from that image.

If your thorium images are numbered from 1 to 11 (e.g., f2061001. fit ... 2061011. fit
on 31 May 2001), and if they correspond to focus positions starting from 720 in steps of
10, then your command line will look like this:

focus_hercules f2061 1 11 720 10 6

C.5. Focusing HERCULES 183

Although it is essential that the focus position increases linearly as the image number
increases (the program will calculate the focus position for every image from its sequence
number) the image sequence does not have to start from 1. The procedure will accept
images in any given range. For example, ifthe first thorium image is 27, then the command
line becomes:

focus_hercules f2061 27 37 720 10 32

Note how the last parameter (the reference image) is now changed from 6 to 32, to stay
in the middle of the sequence.
When the procedure is started, a display window will appear on your monitor (see
Figure C.7) and you will use the box cursor to select a single unsaturated line. The
box can be moved using the mouse and its size altered using the arrow keys (use help
extract/ cursor in MIDAS if you are not sure how to use the box cursor). The program
will then go through all your thorium images and examine the same spectral line every
time (the display will flash as each image is loaded). Finally, a graphics window will appear
with a plot showing how the half-width of your line changes with the focus position (see
Figure C.S). A parabola is fitted to the points and the best focus is displayed, as calculated
from the fit. It is a good idea to repeat the procedure on several different spectral lines
(by restarting focus.1lercules) before adopting the best focus value. Note that only one
line can be selected during one session .

Figure C.7: Choosing an isolated spec-

tralline for focus determination.

C.5.5 Some useful tips

1& Make sure that your box cursor is not too high (across the orders). The program
will construct an average line profile from all pixel rows extracted from the box and
a gaussian will be fitted to that profile. If the box is higher than the spectral orders,
the average line profile may become degraded.
184 Appendix C. HERCULES observing manual

Best focus: 760 •

4.5

4
-x
0.

L
I
:3
LL
3.5 ~
\\
I
)
3
~. ............-~§!.---.....,..
/" Figure C.8: Determining the best focus
position of the OOD camera using a series
of thorium exposures. The best focus is
calculated from a parabolic fit to the data
720 740 760 780 800 820 points.
Focus position (mm/100)

• The box width (along the orders) should be large enough to include a few pixels of
the continuum on either side. However, make sure that no other strong lines are
found in the same box.

• Always examine the plot (Figure C.8) before you adopt the best focus position
returned by the procedure. If the data points are scattered significantly more than
in the example shown in Figure C.8, try another line. A parabolic fit to the measured
FWHM values around the best focus should have a clear minimum which will not
vary from one line to another.

• Do not forget actually to set the CCD focus!

• The MoJo focus CCD focus setting should also be changed.

• Remember to set 'Full Chip' to ON before you start observing.

Appendix D

SALT HRS R2 optical design

This appendix was prepared for the Preliminary Design Review held in 2003, September in
Southampton (Barnes et al., 2003). All aspects ofthe optical design are the authors' while
the remaining aspects were developed in conjunction with the coauthors. The document
is complete except for the details of SALT which have been discussed at length in Section
3.1.1.

D.l Scope
This document provides details of the SALT HRS optical design. It provides details of the
spectrograph design, beginning at the spectrograph entrance slit. The fibre feed input
details are given in 3400AEOOXX and the fibre output, and slit optics, are described in
3230AEOOO1.
The performance requirements of the SALT HRS are described in the Functional Per-
formance Requirements Document (3200AE0001) , and the science requirements are dis-
cussed in the Operational Concepts Definition Documents (3200AE0005).

D.2 SALT HRS optical design

D.2.1 Overview
The SALT HRS is a fibre-fed spectrograph designed for maximum stability and high through-
put. It will be capable of resolving powers up to R = 80000 with a wavelength range
of ,\ = 370 nm to ,\ = 890 nm. In "fixed position" mode a single object can be observed
with simultaneous sky monitoring. A "nod and shuffle" mode is also provided to allow
accurate sky subtraction. The spectrograph uses a mosaic of two R2 echelle gratings with
87 grooves/mm for the main dispersion and two large BK7 prisms with 40 apex angles,
0

which are used in double-pass, for cross-dispersion. The collimated beam size is 365 mm.
The camera, which has a focal length of 706 mm, is an all-spherical catadioptric design
with a primary mirror 1.2 m in diameter and a detector which uses a mosaic of three 2k
by 4k CCD s. The spectrograph will be housed inside an evacuated vessel, and the entire
instrument will be kept inside a temperature-stabilised environment.

D.2.2 Fibre input

The fibre instrument feed (FIF) will allow a single object to be observed concurrently with
a nearby patch of sky. The relative location of the sky can be selected in azimuth and
radial separation. The FIF will allow for the five different observing modes, which are
described in the following section. See document series 34XX for further details of the
fibre instrument feed.

185
186 Appendix D. SALT HRS R2 optical design

/
Collimator

Echelle

Figure D . l: Plan and elevation views of SALT HRS. The collimator is shown on-axis with a focal ratio
of f /3.8.

D.2.3 Fibre outp ut, collimator and vacuum window

A more detailed description of the fibre injection into the spectrograph is given in the
Fibre Injection Design Document (3230AE0001). A range of fibre diameters from 300/-Lm
to 500/-Lm will be used with and without fibre slicers for resolving powers ranging from
R = 17000 to 80000. A 500/-Lm fibre with no fibre slicer will be used for the lowest
resolving powers in both the fixed position and nod and shuffle modes. A pair of 500 /-Lm
and 300/-Lm fibres will be used in conjunction with fibre slicers for resolving powers of
R = 38000 and 80000 respectively in fixed position mode, while the node and shuffle
mode requires pairs of 400 /-Lm and 300/-Lm fibres for resolving powers of R = 33000 and
80000 respectively.
Those fibres that will be sliced will have their exit faces reimaged onto the image-slicer.
The fibre-slicer is required to operate at a relatively slow focal ratio (f'V f /17). This will
require the use of t ransfer optics to convert the f /3.8 beam emerging from the fibres to
an f /17 beam. Because the unsliced 500/-Lm fibres will be placed in the same plane as
the fibre-slicers, their focal ratio must be made to match. This will be done with a single
micro-lens on each fibre. This micro-lens could be used to transfer the fibre near-field
onto the spectrograph pupil. The form of the transfer optics is to be determined (TBD).
It is not possible to accommodate an f /17 collimator within the mechanical envelope
of the spectrograph. Hence, the focal ratio must be sped up after the fibre slicer. The
optics to perform t his will be incorporated into the fold mirror, which will now be a prism
with a 45° surface of internal reflection, with lenses cemented onto the input and/or
output face . T his is a modification of the HROS fold-mirror and focal modifier (D' Arrigo
D.2. SALT HRS optical design 187

et al., 2000a). The lens on the front surface will also serve as the vacuum window. Given
that a vacuum window is necessary, the fold prism and focal modifier should not be seen
as introducing any additional optics. The collimator will be an off-axis paraboloid that
operates at around f /6 to f /8. A schematic of this concept is shown in Figure D.2. The
form of the collimator optics is TBD.

Fold prism, focal

modifier

- .£/8 to f/6 COllimato~

mirror
I
Shutter
f/17
Vacuum tanle wall

Spectrograph
.......-- entrance slit INot to scale. I

Figure D.2: Schematic of the collimator fold prism and focal modification optics. The scale of the fold
prism is greatly exaggerated, as is the distance from the collimator optical axis to the entrance slit,

D.2.4 Dispersive system

The dispersive system comprises a mosaic of two large echelle gratings and two BK7
prisms for cross-dispersion. The prisms are used in double-pass before and after the
echelle dispersion.

Echelle gratings
A mosaic of two of the largest gratings available from the Richardson Grating Laboratory
(RGL) is required. The parameters of the echelle gratings are given in Table D.l. Note
that only standard catalogue gratings are being considered owing to the prohibitive cost
of have a grating custom ruled.

Table D.l: SALT HRS grating parameters.

Parameter Specification
RGL catalogue number 53045ZDOl-127E
Blaze angle, BB 63.0°
Groove density, T 87.0 grooves/mm
Grating ruled width 308mm
Grating ruled length 413mm
188 Appendix D. SALT HRS R2 optical design

The relatively fine ruling of the grating produces orders that have a large angular
spread. However, the total number of orders is small and therefore the amount of inter-
order spacing is large. This is the reason for rejecting an earlier design that called for a
R2.8 grating (BB = 70.45 with 52.7 grooves/mm. The dispersed wavelengths fall in 33
0
)

orders from m = 55 for A = 370 nm to m = 23 for A = 900 nm (see Figure D.9 and Table
D.5).
The gratings will be mechanically aligned with a 35 mm gap between the ruled regions
of each grating, which allows for a gap of 25 mm between the grating substrates. This gives
a total grating length of L = 861 mm. The facets of the echelle grating are illuminated at
0
a Littrow angle of B = 4.5 with respect to the facet normal. The camera is located at a
distance of 3.5 metres from the centre of the dispersive system (i.e., the echelle grating).
This distance is a compromise between making efficient use of the echelle grating (i.e.,
a small Littrow angle) and having an excessively large spectrograph, while also ensuring
that the collimator obstructs none of the dispersed beam at all wavelengths within one
half of a free spectral range from the blaze wavelength.

Resolving power and beam size

If the angular size of a fibre projected onto the sky is Bs , and the fibre degrades the
telescope focal ratio by p, then a grating which receives a collimated beam of diameter B
is capable of a resolving power R = A/oA given by

R= 2B tan BB
(D.1)
pBsD(l - tan Btan BB)

where D is the telescope primary mirror diameter. It should be noted that the effective
resolving power R of a fibre is somewhat greater than that of a (uniformly illuminated)
slit that has the same angular width on the sky. A fibre with a uniformly illuminated
output will, after convolution with the point spread function of the spectrograph and
extracted to a one-dimensional profile, have a FWHM of between 0.70 to 0.80 times the
projected fibre diameter, depending on the image quality. Only the worst image quality
(relative to the projected fibre diameter) will result in profiles that can be approximated
by a gaussian. It is assumed that R' = R/0.75 which concurs with previous results from
fibre-fed spectrographs, where the FWHM is universally used as a measure of resolving
power.
Assuming that the focal ratio degradation is 10% (i.e., p = 1.1), and a collimated
beam size B = 365 mm, the resolving powers obtained with a variety of fibre diameters
are shown in Table D.2. The transmission through the entrance aperture of an individual
fibre under median seeing conditions is also given.
The resolving power/fibre diameter product is RBs = 28800 arcsec, or in terms of
effective resolving power R'Bs = 38600 arcsec. This rather large product is required in
order to ensure that the spectrograph is well matched to image quality delivered by SALT.
Thus it can be seen that up to 82% of the light is accepted by the fibre entrance aperture
for a resolving power of R = 17000. Clearly for resolving powers greater than R = 20000
to 25000 some form of image slicing is required (see 3230AE0001).
D.2. SALT HRS optical design 189

Table D.2: Fibre diameters, resolving powers, and entrance aperture transmissions.

Fibre diameter Resolving powers Transmission

d (/Lm ) es (arcsec) R' R Tsee (%)
100 0.45 64700 86300 8.6
200 0.89 32300 43100 29.3
300 1.34 21600 28700 52.1
400 1.79 16200 21600 70.2
500 2.23 12900 17300 82.0

I I

200 l- .
.... -~---- .. -~ .. ----~---~.-------- ... ... ....
......, .. ... .. ...., ...
~ ","

E 100 ",' .
E ,/ ,.
c I
.I .,,
o
E
o , I
I :-
en
o .., \

...
~
.I
I

0. -100 I- ..........
..
... ... ; " -
>- ...... .........
~
... .. ...

-200~~__~__~__"'~"'-_--_-_--~--_-_--_-~·-_--_-~--~--_-_--_--~"__~__~__~j
~~

-500 -400 -300 -200 -100 0100 200 300 400 500
X position (mm)

Figure D.3: A schematic of the footprint of the collimated beam on the echelle gratings. Note that this
does not depict the effect of cross-dispersion before the gratings.

Overfilling of the echelle grating

The collimated beam overfills the echelle grating in both directions. The fraction of the
collimated beam accepted by the echelle grating mosaic is about 85%, which includes
a 4% loss due to the gap between echelle gratings. Note that this calculation assumes
a uniformly illuminated footprint on the echelle (see Figure D.3), and in reality there
will be less light near the edges due to fibre FRD effects. This will mitigate the effect
of geometrical overfilling. Note that an overfilling of 85% is comparable to the amount
that would be expected with a mosaic of two R3 gratings (i.e., eB = 71.5°) with the same
dimensions that is illuminated with a 300 mm beam.

Cross-dispersion prisms
Two large BK7 prisms with an apex angle of ap = 40.0 are used in double-pass for cross- 0

dispersion. This amount of dispersion allows for complete wavelength coverage from
190 Appendix D. SALT HRS R2 optical design

..\ = 370nm to..\ 900nm with a 705mm focal length camera and a 61.4mm high
detector. The orders in which these wavelengths appear are near the bottom and top of
the OOD (see Figure D. 9).
The angle of incidence of the collimated beam on the first prism is Bi = 31.6 This 0
•

is the angle for minimum deviation for ..\ = 370 nm, which keeps the plane of echelle
dispersion of the order in which this wavelength appears in the same plane as the optical
axis of the collimator. Because the angular spread of this order is the smallest this allows
the angular separation of the collimated and dispersed beams to be minimized without
obstructing wavelengths that fall within central free spectral range.

Prism homogeneity

The effect of refractive index inhomogeneity within the cross-dispersing prisms has been
estimated. A Monte-Carlo simulation of the cross-dispersion was performed where the
refractive index of the prisms at each air-glass interface was varied with a normal distri-
bution about the nominal refractive index. A total of 105 rays was used in each simulation.
The deviation of the angle of dispersion from the mean was computed and the results were
transformed to the image plane for a camera focal length of 705 mm. This allows the effect
of prism inhomogeneity on the line profile in the direction of the cross-dispersion to be
estimated. It is assumed that the effect in the direction of echelle dispersion will be of
a similar order of magnitude. Examples of these profiles are shown in Figure D.4. The
FWHM of the cross-dispersion profile for a point source was calculated. The results are
given in Table D.3. Given that the mean RMS image quality is 7 to 8/-hm the refractive
index homogeniety of less than 1e-6 will degrade the order profile by less than 10%. Hence
the prism glass must be H4 or better. This is well within the capability of major glass
manufactures (e.g. Schott or Ohara) who routinely supply blanks of BK7 greater than 1 m
in diameter with this homogeneity.

0.1 0.1

o -5 5 10 - 0 -5 o 5 10
!;y (~lm)

(a) Homogeneity class H4 (6.11, = ±10-6 ) (b) Homogeneity class H5 (6.11, ±5 x

10- 7 )

Figure D.4: The effect of prism inhomogeneity on image quality.

D.2. SALT HRS optical design 191

Table D.3: Effect of prism homogeneity on image quality.

Homogeneity Refractive index (/-Lm)

FWHM
class variation ±n A = 370nm A = 900nm
H3 2e-6 5.3 4.9
H4 1e-6 2.7 2.5
H5 5e-7 1.4 1.3

The effect of cross-dispersion on line tilt and spectral format

There are some significant effects of using prisms for cross-dispersion on the tilt of spectral
lines that must be considered. The first is that due to the dispersion before the echelle.
This causes a variable angle of illumination of the echelle facets (in the x-z plane), which
has a total range that is equal to one half the total dispersion angle of both passes through
the prism. The angle ranges from 'Y = 00 at A = 370 nm to 'Y = 2.5 0 at A = 890 nm. The
fact that 'Y = 0 at A = 370 nm is a consequence of choosing this wavelength to be minimally
dispersed. It can be shown from the grating equation,

rnA = a(sin a + sin (3) cos I (D.2)

that due to a small change in 'Y from the bottom of the slit to the top, the echelle angular
dispersion changes slightly. The effect of this is to tilt the lines by an amount given by

tan ¢ = 2 tan OB tan 'Y (D.3)

in the order centre. Hence the line tilt, due to this effect, ranges from ¢ = 00 at A = 370 nm
to ¢ = 10 0 at A = 890 nm. The tilt also varies across each order. This variation is around
±0.3° in order 55 and ±2.3° in order 23.
The other effect is due to the fact that the angle of incidence on the prisms in the
y - z plane (i.e, the plane of echelle dispersion) is not zero. On the first pass through the
prisms the angle of incidence is the Littrow angle (0 = 4.5 0). On the second pass through
the prism the angle of incidence varies according to the echelle dispersion. The result is
that the prism contributes to the echelle dispersion in such a way that the ratio of echelle
dispersion to cross-dispersion varies considerably across an order. This gives rise to the
considerable order curvature, which would be absent if all the prism cross-dispersion were
done at normal incidence. Finally, this last effect also causes the line tilt to increase by
around 6 degrees across the entire field. This is due to the ratio of prism dispersion in the
x-z and y-z planes. Hence the line tilt will range from ¢ = 60 at A = 370 nm to ¢ = 16 0
at A = 890nm.
It will be possible to orientate the fibre-slicers in order to set the line tilt to zero in
the middle of the central order. The range of line tilt will then be between _50 and 5°.
Because the line tilt pattern is always the same, it will be possible to remove its effect
during the reduction process. This may involve, for instance, a modification of techniques
commonly used for the reduction of long slit spectra.
192 Appendix D. SALT HRS R2 optical design

Line tilt and nod and shuffle

The line tilt also has potential consequences for the proposed nod and shuffle mode of
operation (see the Operational Concepts Definition Document, 3200AE0005). In this
mode, the science target is moved between object and sky fibres during an exposure while
simultaneously charges on the detector are shuffled along its columns. The resulting
imaged spectrum for a given order is the sum of two spectra, one of which has been
shifted in the dispersion direction by an amount equal to the shuffle distance (normally
halfthe order separation) multiplied by the line tilt angle in radians. We have modeled this
effect by shifting and adding synthetic line profiles of various widths. By construction,
the equivalent width is always preserved by this operation, but very narrow lines can
become noticeably shallower and broader. Since the two added spectra are of almost
equal intensity and dispersion solutions will be available for each component spectrum,
the dispersion solution for the combined spectrum will be close to the mean of that of the
two components. Radial velocity precision may in practice be degraded slightly for spectra
acquired in nod and shuffle mode. Extremely sharp emission lines (widths of the order
of the resolution limit) will appear doubled near the edges of the spectral format where
the line tilt is most severe. Sharp night-sky emission lines will still subtract-out with very
high accuracy (even if the profiles are doubled) because the sky for each component of
the co-added spectrum is recorded separately.

D.2.5 Camera

The camera is required to be capable of critically sampling the smallest resolution element.
This occurs when Rrna:x. = 80000 and the required focal length of the camera is given by

learn = ns;rnp RrnaxSpix cot 8B (1 + tan 8B tan 8) (D.4)

where Spix is the pixel size. The number of elements required in order to sample critically
a resolution element is n sarnp = 2. Assuming 15/-lm pixels, the focal length required is
learn = 705mm, which gives a monochromatic focal ratio of f / D = 1.9. The diameter of
the camera primary mirror has been limited to 1.2 m (for practical and budgetary reasons)
and hence the white light focal ratio is f / D = 0.6.
The camera is a catadioptric design with three large corrective elements (a bi-convex
and two meniscus lenses), a deep mirror and a small field-flattening lens that also serves
as the cryostat window. All surfaces are spherical. The design evolved from the camera
for the Keck HIRES instrument (Epps and Vogt, 1993). However the HIRES design, which
uses two large corrective lenses, is incapable of delivering satisfactory image quality, given
the large dispersive angles and pupil distance of SALT HRS. This type of camera, which
was reoptimized to use BK7 glass, was considered during earlier conceptual designs, when
a reduced field of view of the camera allowed for a less demanding camera design.
The final design, which is shown in Figure D.5, is quite similar to the original HROS
camera for the Gemini telescope (D'Arrigo et al., 2000b) and the HDS1 camera for Subaru
(N oguchi et al., 2002), although considerably larger than either. Some details of the
optical elements will be discussed below.

lThe PDR document incorrectly stated this was the HIDES camera.
194 Appendix D. SALT HRS R2 optical design

Figure D.6: Footprint on the camera primary mirror. The blue wavelengths are at the bottom and
the red wavelengths are at the top. The direction of echelle dispersion is from left to right. Note that
the centroid of dispersion does not coincide with the optical axis of the camera. The camera central
obstruction is clearly seen.

The camera mirror

As stated above, the diameter of the primary mirror was limited to 1.2 metres. It was
found that this allowed an almost entirely unvignetted view of all dispersed wavelengths
with the free spectral range from A = 370 nm to A = 890 nm. Because the angular spread
of the orders in the red is significantly larger than in the blue, it was also found that the
camera could be used more efficiently if the centroid of the dispersed rays was slightly
displaced from the optical axis. This is shown in Figure D.6. The dimensions of all the
other optical elements, including the prisms, have been determined so as not to cause any
additional vignetting.

The large camera lenses

During optimization of the design it was noted that the radii of the two meniscus lenses
were quite similar. After coupling the radii of these lenses, and reoptimizing the design,
D.2. SALT HRS optical design 195

the performance was only marginally degraded. However, there will be significant savings
in the manufacture and testing of these lenses if they have identical figures.

The field-flattening lens

The field-flattening lens will be used as the CCD cryostat window. Initially it was unde-
cided whether to use BK7 or fused silica for the field-flattening lens. Silica is commonly
used for CCD windows. However, this is generally because of the good transmittance of
this glass at short wavelengths. The decision to use BK7 came after discussions with the
manufactures of the CCD cryostat. Although BK7 has a larger coefficient of thermal ex-
pansion, materials are readily available with matching CTES. Materials with lower CTE are
more costly in larger dimensions. A possible reason for favouring silica may have been its
ease of manufacturability. However, the manufacture of a BK7 lens is not overly difficult,
so this was not considered to be relevant. The strength of BK7 and silica is comparable.
Table D.4 lists the physical properties of the two materials. 2

Table D.4: The physical properties of BK7 and fused silica.

Property BK7 Silica

3
Density 2.51 g/cm 2.20 g/cm 3
Young's modulus 82.0 x 10 9 N /m 2 72.1 x 10 9 N/m2
Poisson's ratio 0.206 0.179
CTE (-30 C to 70 C)
D D
7.1 x 1O- 6 /K 0.55 x 10- 6 /K
Heat capacity 0.858 J / (g.K) 0.741 J/(g.K)
Thermal conductivity 1.11 W /(m.K) 1.38W/(m.K)

The location of the field-flattening lens relative to the CCD will be fixed. This allows
the entire cryostat to be used for focusing. Appropriate tip/tilt and focus adjustments
have been allowed for. The size of the field-flattening lens is quite critical as this dictates
the size of the CCD cryostat and hence the size of the central obstruction. It was found
that a lens that is 130 x 100 mm rectangular would not obstruct any light that falls on
the three CCDS (see Figure D. 7). A complicating factor in the design of this lens is the
necessity for it to be tilted (about the x-axis) with respect to the camera optical axis (i.e.,
the axis of the camera mirror and three large lenses). The image surface, which is now
planar, is also tilted with respect to the optical axis of the field-flattening lens. In order to
minimize the size of this lens it must also be decentred by 13 mm in y. This corresponds
to the centroid of the echelle spectrum on the CCD mosaic.

D.2.6 Detector
The detector is a mosaic of three E2V 42-82 CCDS. Each detector has 2048 x 4096
pixels that are 15 /--lm square. In order to allow charge shuffling in the direction of cross-
dispersion, the detectors must be aligned with the columns being vertical. As shown in
2During the PDR it was pointed out by S. Shectman that because BK7 contains significant quantities of
potassium the decay of K-40 will produce an unacceptable level of background radiation. For this reason
BK7 should not be used as a CCD window.
196 Appendix D. SALT HRS R2 optical design

130mm

I" ~I

r-
o
o
§ Figure D. 7: The foot-
print diagram on the field-
flattening lens. Note that
the entire width of the lens
is used by wavelengths
that lie near the edge of
the detector

Figure D.S, this provides a total imaging area of 95.4 x 61.4mm, where the gap between
eeDs is 1.6 mm. This allows for an inactive are on each side of the chip of 0.45 mm and
a 0.7mm gap between the detector packages.

95.16 mm

1.5 mm 1.5mm

-11-4- ~I
-r--
-1r-
CCD 1 CCD2 CCD3

A~

.p-
§ AXD o
\D

'" Figure D.S: The format

'E.
~
to of the SALT HRS CODs.
The direction of echelle

AECH
.... dispersion (ECH) and
cross-dispersion (XD)
are shown. The CCDs
are aligned vertically
-'--- - to enable charge shuf-
/ .. 2048 pixels
~I fle in the direction of
cross-dispersion.
D.3. Instrument performance 197

D.3 Instrument performance

D.3.1 Spectral format
Table D.5 gives the blaze wavelengths and order numbers for the SALT HRS spectral format
that is shown in Figure D.9. The wavelength coverage is from A = 370 nm to A = 890 nm
which corresponds to order numbers m = 55 to m = 23 respectively. With a mosaic of
three CCDs the wavelength coverage is almost complete across this entire range. Above
A = 750 nm some wavelengths within the free spectral range cannot be captured on the
CCDS. The gaps between the CCDS also means there will be gaps in every order. However,
for orders above m = 41 (i.e., A < 500 nm) the wavelengths which are lost because of the
gaps may be retrieved from either the preceding or following orders.

Table D.5: Order numbers and wavelengths for SALTHRS. The wavelengths at the extent of the free
spectral range are also shown. That is, + / - AFSR = AB ± AFSR/2, where AFSR is the wavelength extent
of one free spectral range. Amin and Amax are the wavelengths at the edge of the CCD mosaic.
Order Wavelength (nm) Order Wavelength (nm)
m AB -AFSR +AFSR Amin Amax m AB -AFSR +AFSR Amin Amax
55 371.3 367.8 374.6 364.4 378.2 38 537.2 529.9 544.0 526.1 546.5
54 378.1 374.6 381.6 370.8 385.1 37 551.7 544.0 558.9 540.3 561.3
53 385.3 381.5 388.8 377.8 392.3 36 567.0 558.8 574.6 555.2 576.8
52 392.7 388.8 396.4 385.0 399.8 35 583.2 574.5 591.2 571.1 593.3
51 400.4 396.3 404.2 392.5 407.6 34 600.3 591.1 608.8 587.8 610.7
50 408.4 404.2 412.3 400.3 415.8 33 618.5 608.7 627.4 605.6 629.1
49 416.7 412.3 420.8 408.4 424.2 32 637.8 627.4 647.3 624.4 648.7
48 425.4 420.8 429.7 416.9 433.0 31 658.3 647.2 668.5 644.5 669.6
47 434.4 429.7 438.9 425.7 442.2 30 680.3 668.4 691.1 666.0 691.9
46 443.9 438.9 448.5 434.9 451.8 29 703.7 691.0 715.2 688.9 715.7
45 453.7 448.5 458.6 444.6 461.8 28 728.8 715.1 741.1 713.4 741.2
44 464.0 458.6 469.1 454.6 472.2 27 755.8 741.1 769.0 739.8 768.6
43 474.8 469.1 480.1 465.2 483.2 26 784.8 768.9 799.1 768.2 798.1
42 486.1 480.1 491.7 476.2 494.6 25 816.2 798.9 831.5 798.8 830.0
41 497.9 491.6 503.8 487.8 506.7 24 850.1 831.4 866.8 832.0 864.5
40 510.4 503.8 516.5 499.9 519.3 23 887.0 866.6 905.1 868.1 902.0
39 523.4 516.5 529.9 512.7 532.6

Figure D.10 shows the position of thorium-argon lines that are considered useful for
the purposes of wavelength calibration (Ramm, 2003). It can be seen that even when
the image plane is divided into three, there are at least 10 or 15 lines per order. Hence
there will be some 400 or 500 lines available on each CCD for two-dimensional wavelength
calibration.
 40
Order
---------- -------------------------
- .... ...- ... ~ ".- ~ -~ ---~----------- --~---..-

30 23
_.. _.-
.pttj-
Bill
-~-et'·-···
25

20

E 10
E
'-"

55 371

-30 ••••• - . . . . . . -- • • • < ____ - - ______________ __ --1_- ___ LL_

_~_...,t.--

-40

-60 -40 -20 o 20 40 60

X-position (mm)

Figure D.9: The SALT HRS spectral format. The orders from m = 23 to m = 55 are plotted as solid lines. Orders above and below these are shown as
dashed lines. The two curved solid lines show the extent of the free spectral range, and the full width at half maximum is shown by the curved dashed lines.
The outline of the three CCDs is also shown. Note that some wavelengths that fall in the gaps between CCDS will appear in the preceding or following order.
The dot-dashed lines show the position of these wavelengths.
 t:::I
'"
tj
CIl
....,
",.

C
S
(D
::<
0+

40 '0
(D
>-j

.. .
8'

30
Order

23 ...
r=~----.~ _=
__~__=__~==========~====~ ~'~"=~ . . . .=. .=
- --- ~-..:~- ...........-_.................__..._.._._--_..... _ . - ............:::::::::::=:--- . - .._-
AB
887
>-j

S
P'
::<
()
25
27
._----- 816 (D

--~-

20

10
E
E
'-" 41
s:::
0 0 43
~
(f) 45
0
c..
>- -10 47

49 417
··········'000· .
51 •~~~~++~~4-~~~*-~__ 400
-20 -V'OO'·-·0····
......+!-<>---f"---B-/",*'---w.'>,....{:-_..--$-Q.
53 385
0':0 _:...-...~.o.--E>--~~~·-+I>-"~~~<'--./f---e--· 0

55 --e- _ .' 000

371
" ":"~l---·-HL--->-----:-o ...__ .... -0 ...
. --~...•...
-30 L -_ _ _ _~~~~~--~
-Q-'.()'

-40

-60 -40 -20 o 20 40 60

X-position (mm)

Figure D.lO: The position of thorium-argon calibration lines. The input catalogue (Ramm, 2003) is incomplete beyond A = 750nm. The plotted lines are
those that are considered "good" for the purposes of calibrating the HERCULES spectrograph. Depending on the operating current of the calibration lamp,
some of the lines may be either too faint or too bright to be used. This will eliminate around 10% to 20% of lines.
200 Appendix D. SALT HRS R2 optical design

D.3.2 Image quality and vignetting

Spot diagrams for wavelengths at the centre and edges of orders for three orders (at the
bottom, middle, and top of the camera field of view) are shown in Figure D.ll and en-
squared energies at these wavelengths are shown in Figure D.12. The requirement is that
the ensquared energy within one pixel is greater than 80%. However in recognition of the
fact that some wavelengths towards the edge of the field have large dispersion angles, and
hence have poorer image quality, this requirement was relaxed to the following:
• A < 700 nm - greater than 80% ensquared energy within 15/-Lm x 15/-Lm .
• A> 700 nm - greater than 80% ensquared energy within 15/-Lm x 15/-Lm at most
wavelengths with greater than 70% at all wavelengths.

(a) A = 866.7nm (b) A = 886.8nm (c) A = 90S.2nm

(d) A = S03.7nm (e) A=S10.3nm (f) A = S16.4nm

C" .
" .
"
"

(g) A=367.8nm (h) A = 371.2 nm (i) A = 374.6nm

Figure D.ll: Spot sizes of representative wavelengths. The wavelengths are from positions near the
edges and centre of orders m = 23, Ll0 and 55. The box size is 30 f-Lm .
D.3. Instrument performance 201

(a) A = 866.7nm (b) A = 886.8nm (c) A = 905.2nm

(d) A = 503.7nm (e) A = 510.3nm (f) A = 516.4nm

(g) A = 367.8 nm (h) A = 371.2 nm (i) A = 374.6nm

Figure D.12: Ensquared energies of representative wavelengths. The wavelengths and order numbers
are the same as in Figure D .11.

Figure D.13 shows the ensquared energy within one pixel at all wavelengths. It can
be seen that the image quality is excellent at all wavelengths within the central CCD. At
larger dispersion angles there is a gradual degradation of image quality. The image quality
rapidly deteriorates at wavelengths where the dispersed beam is not completely captured
by the primary mirror. The extreme wings ofthe blue orders have the worst image quality.
However the image quality within the central free spectral range is satisfactory.
The vignetting function of the spectrograph is shown in Figure D.14. It can be seen
that a significant proportion of wavelengths are vignetted to some extent by the CCD
cryostat. This obstruction contributes between 10% and 12% to the overall vignetting.
 100

40

30

20 85
,-...,
E 10 80
E
"-'"
c:
.Q 0
~
en
o
9- -10
>-
-20

-30

-40

-60 -40 -20 o 20 40 60

X-position (m m)

Figure D.13: The ensquared energy (%) within one pixel at all wavelengths.
 40

30

20
",-.....

E 10 75
E
-...-
c
.Q 0
:t::::
en
0
9- -10
>-
-20

-30

-40

-60 -40 -20 o 20 40 60

X-position (mm)

Figure D.1 4: The vignetting function of the spectrograph. Shown is the unvignetted percentage of rays. The vignetting due to the central obstruction
with the camera is clearly visible. The increase in vignetting towards the edges of the field is due to the limited size of the primary mirror. Note that the
calculation of the vignetted fraction of rays assumes a uniformly illuminated pupil.
204 Appendix D. SALT HRS R2 optical design

D.3.3 Throughput
The throughput calculations have been done assuming mean seeing conditions. All optics
are assumed to have single layer MgF2 anti-reflection coatings, except the fibre slicer
transfer optics, and the focal modifier. The mirrors are assumed to have uv-enhanced
silver reflection coatings. There is a possibility that some, or all, of the large optics will
also be coated with Solgel. The transmission of a Solgel + MgF 2 coating is around 0.995
(at 600 nm). Therefore, the throughput may be improved by a factor (0.990/0.98)14 to
(0.995/0.98)14 = 1.15 to 1.23, where 14 is the number of vacuum/glass surfaces (i.e., two
prisms in double-pass, and three lenses). The application of Solgel is therefore to be
highly recommended.

Fibre feed and image slicers

The length of fibre needed is about 35 m (TBe). The transmission through a low OR silica
fibre at ). = 400, 600 and 850 nm, is 80.6%, 93.9% and 90.9% respectively.
The geometrical throughput of the fibre feed input and fibre image slicer is given in
Table D.6. Included in the calculation are the entrance aperture losses due to the stellar
PSF, and the geometrical losses of the fibre image slicer. The efficiency of the fibre slicer
is estimated to be 85%.

Table D.6: Geometrical throughput of the fibre feed and image slicer. The modes are F = fixed object
plus sky, N = nod and shuffle. L, M, and H, represent low, medium and high resolution respectively (see
Section 3.3). The throughput relative to FL (which is equivalent to NL) is also shown.

Mode: FL FM FH NL NM NH
Geometric throughput (%) 87.2 78.6 54.0 87.2 69.2 42.8
Slicer efficiency (%) 100.0 85.0 85.0 100.0 85.0 85.0
Total (%) 87.2 66.8 45.9 87.2 58.8 36.3
Relative throughput 1.00 0.76 0.52 1.00 0.67 0.41

The fibre image slicer will require some transfer optics, the form of which is TBD. It
is assumed that these lenses will have high efficiency broad-band anti-reflection coatings
applied. A throughput of 97% is assumed. The total throughput of the fibre feed and
image slicers are given in Table D.7.

Table D.7: Fibre feed and image slicer throughput.

Mode Resolving Throughput (%)

Power ). = 400 nm ). = 600 nm ). = 800 nm
FL 17,000 65.3 76.3 74.0
FM 38,000 48.3 56.9 55.4
FH 80,000 33.3 39.1 38.0
NM 33,000 42.5 50.1 48.7
NH 80,000 26.3 31.0 30.2
D.3. Instrument performance 205

Fold mirror/focal modifier and collimator

The reflectivity of the fold mirror is assumed, for the moment, to be that of a standard
mirror (with enhanced silver coatings). The focal modifier (whose design is TBD), is
assumed to have only two air/glass surfaces (of which one will be inside the vacuum tank),
and to be made from BK7 glass with a total glass length of around 100 mm. Of course, the
final optical elements will be formed from several glass types, but the overall transmittance
will not change substantially. Broad-band anti-reflection coatings are assumed. The
throughput at ,\ = 400, 600, and 850 nm is 92.4%, 95.9%, and 95.0% respectively.
The overfilling of the collimator mirror due to focal ratio degradation in the fibre is
assumed to be 5%. Combined with the reflectivity of the mirror, the transmittance is
92.0%, 93.5%, and 91.9% at ,\ = 400, 600, and 850 nm respectively. The total throughput
of the fold mirror/focal modifier and collimator is given in Table D.8.

Table D.8: Fold mirror/focal modifier and collimator throughput.

Throughput (%)
400 nm 600 nm 800 nm
Fold mirror/focal modifier 92.4 95.9 95.0
Collimator 92.0 93.5 91.9
Total 85.0 89.6 87.3

Prism and echelle

The throughput oftwo prisms, used in double-pass is 63.5%, 76.9%, and 80.0% at ,\ = 400,
600, and 850 nm respectively. The blaze efficiency of the echelle grating is approximately
73.8% at these wavelengths (assuming a Littrow angle of 4.5°). It will be assumed that the
grating is coated with standard aluminium. This gives an absolute efficiency of the echelle
grating of 62.7%, 65.7%, and 62.5%. The overfilling of the grating gives a transmission of
85.2% (including light lost to the gaps).

Table D.9: Prisms and echelle throughput.

Throughput (%)
400 nm 600 nm 800 nm
Prisms 63.5 76.9 80.0
Echelle grating 62.7 65.7 62.5
Overfilling 85.2 85.2 85.2
Total echelle and prisms 33.9 43.0 42.6

Camera
The efficiency calculations of the camera includes the following:
206 Appendix D. SALT HRS R2 optical design

• three large lenses (average thickness 75 mm),

• the primary mirror,

• a field-flattening lens, and

• the central obstruction due to the CCD cryostat.

A summary of the calculated efficiencies is given in Table D.10.

Table D.lO: Camera throughput.

Throughput (%)
400 nm 600 nm 800 nm
Lenses (x 3) 79.5 88.9 86.4
Primary mirror 96.8 98.4 96.7
Field-flattening lens 94.5 97.0 95.7
CCD cryostat obstruction 84.8 78.8 78.5
Total camera 61. 7 66.8 62.7

CCD
The CCDS will be E2V 42-82 chips. The most likely coating will be E2V's "astro-mid"
coating. At A = 400, 600, and 850 nm, these CCDS have quantum efficiencies of 60.1 %,
89.0% and 71.8% respectively. The possibility has been raised, with the support of Dr
Paul Jorden from E2V, that the lower half ofthe three CCDS could have "astro-BB" (broad-
band) coatings. At a wavelength of A = 400 nm the quantum efficiency increases to 80.1 %
(a 33% increase in efficiency). The gain at A = 370 nm is even greater. Here the quantum
efficiency would increase from 35.1% to 61.7% (a gain of 75%!). Clearly this would greatly
enhance the blue-wavelength performance of SALT HRS.

Summary
The total spectrograph throughput, from the fibre feed entrance to the CCD detector is
given in Table D.ll.
The efficiency of the SALT telescope has been estimated to be 56.8%,66.2%, and 66.4%
at the three representative wavelengths. This gives a total detective quantum efficiency
of the SALT HRS and telescope as given in Table D.12.
Finally, as noted in the introduction to this section, a dramatic improvement in
throughput is to be expected if all large refracting optics have Solgel coatings applied
in addition to MgF 2 . The efficiency of the SALT HRS and telescope with such coatings is
given in Table D.13.

D.3.4 Stray light and ghosts

A complete analysis of the stray light properties of the spectrograph has not yet been
completed. It should be pointed out that the use of prisms for cross-dispersion ensures
D.3. Instrument performance 207

Table D.ll: Total SALTHRS throughput. The efficiencies at A = 400nm for "astro-BE" are in brackets
in this and all subsequent tables.

Mode Resolving Throughput (%)

Power 400nm 600nm 800nm
FL 17,000 7.0 (9.3) 18.0 12.4
FM 38,000 5.2 (6.9) 13.4 9.3
FH 80,000 3.6 (4.7) 9.2 6.4
NM 33,000 4.6 (6.1) 11.8 8.2
NH 80,000 2.8 (3.8) 7.3 5.1

Table D.12: SALT HRS and telescope detective quantum efficiency.

Mode Resolving Throughput (%)

Power 400 nm 600 nm 800 nm
FL 17,000 4.0 (5.3) 11.9 8.2
FM 38,000 2.9 (3.9) 8.9 6.2
FH 80,000 2.0 (2.7) 6.1 4.2
NM 33,000 2.6 (3.4) 7.8 5.4
NH 80,000 1.6 (2.1) 4.8 3.4

that the amount of stray light is significantly less than if a grating had been used. It is
expected that the most significant source of stray light will be from the echelle grating.
Based on previous experience (e.g., the HERCULES spectrograph), and the work of others
(c.f. D'Arrigo et al., 2000a) it is expected that stray light will contribute approximately
1% to 2% of the local continuum. The variation in the intensity of this stray light across
the CCD is also expected to be smooth. Again, this is based on prior experience with
similar instruments.
The impact of ghosts is also expected to be minimal. This is due in part to the very
fast nature of the camera, which will ensure that any ghosts will be significantly out of
focus. For a preliminary analysis of the effects of ghosts we refer the reader to a study
of ghosting in the original HROS spectrograph (D'Arrigo et al., 2000a). The format of

Table D.13: SALT HRS and telescope detective quantum efficiency assuming Solgel coatings.

Mode Resolving Throughput (%)

Power 400nm 600nm 800nm
FL 17,000 4.9 (6.5) 14.6 10.1
FM 38,000 3.6 (4.8) 10.9 7.6
FH 80,000 2.5 (3.3) 7.5 5.2
NM 33,000 3.2 (4.2) 9.6 6.7
NH 80,000 2.0 (2.6) 6.0 4.1
208 Appendix D. SALT HRS R2 optical design

the camera is very similar, and hence their conclusions are expected to be valid for the
present spectrograph. Here it was shown that even the strongest ghosts will contribute
no more than 0.02% with respect to the local continuum. However, it was noted that
"picket-fence" ghosts are produced with intensities 10 to 20 times larger. These ghosts
are due to light being reflected off the CCD, recollimated by the camera, and then being
redispersed by the echelle.

D.4 Exposure meter

The exposure meter has been designed (see Figure D.15) to use light that would otherwise
be obstructed by the CCD cryostat. A fold mirror will be placed between the last large
camera lens and the rear of the cryostat. A doublet placed before the fold mirror will
bring the light to a focal plane near the vacuum tank wall. A singlet, which will also
act as a window, is used to reduce the scale of the field to a reasonable size. At most
wavelengths around 3% of the light is captured by the exposure meter.

Figure D.15: The expo-

sure meter. A fold mirror
and doublet lens are used
to bring the light to a con-
venient location near the
tank wall. A singlet lens
will then be used as a win-
dow into the tank and the
detector will be placed ex-
ternally.

Because the metered light has been dispersed, the exposure meter optics have been
designed to produce a spectrum on the image plane. This allows the possibility of using
a CCD detector to capture this spectrum, where the image quality is sufficient for low-
resolution extraction. The detector under consideration has 1024 x 1024 pixels, each
24 Mm square, and possibly frame transfer capability. This will permit two modes of
operation. The first vvill be a traditional metering of spectrograph throughput, where the
CCD is read out semi-continuously, and the accumulated signal is calculated. The other
mode will use a less frequent readout, and the spectra will then be rapidly reduced and
displayed. After appropriate corrections are made, this will allow continuous assessment
of the signal to noise of the current exposure. Comparative measurements of the sky and
object signal will also be possible.
D.5. Opto-mechanical tolerances 209

D.5 Opto-mechanical tolerances

For an analysis of the opto-mechanical tolerances, and an analysis of manufacturability,
of the spectrograph we refer the reader to the study by A. Rakich from KiwiStar Optics
(321 OAA 0003).

D.6 Procurement
D.6.1 Optical components
We are currently investigating three sources for the procurement of the main optical
elements. These are:

• GlassFab, Rochester, NY,

• LZOS Optics Group, Moscow region and

• Ohara GmbH, Hofheim, Germany.

Inquiries have been made with wzw-Optic to supply the fibre slicers. The alternative
design of Robert Content (see 3230AA0002) could be manufactured by the University of
Durham.
No vendor has been identified for the supply of miscellaneous optics (e.g., the fold-
prism/focal-modifier and the exposure meter optics).

D.6.2 Figuring
Our preferred vendor for the figuring of the optics is KiwiStar Optics, Wellington, New
Zealand. We have also sent an RFQ to Sagem-REosc.

D.6.3 Coatings
Cleveland Crystals (Ohio, USA) has received an RFQ to supply Solgel coatings to all large
refractive optical elements. They may also supply MgF 2 through a vendor they have
identified. Sagem-REosc also have the facilities to provide AR coatings to the optics and
have been sent an RFQ.
The preferred coating for the reflective optics is enhanced and overcoated silver. No
specific vendor has been identified to coat the camera primary mirror. Laserdyne Tech-
nologies (Queensland, Australia) are able to coat the collimator mirror.
Appendix E

SALT HRS R4 optical design

This appendix is the optical design document for the SALT HRS R4 instrument which was
presented to the SALT Science Working Group on 2004 July 29, in Gottingen, Germany.
The document forms part of a larger package of documents that were presented on this
date. The document is essentially complete, except that the introductory material on
SALT has been removed. Details of the telescope can be found in Section 3.1.1.

E.1 Scope
This document provides details of the SALT HRS R4 optical design, including a small
amount of information on the telescope design, the fibre feed and slit optics. More spe-
cific details of the fibre feed can be found in 3400AEOOXX (FIF) and the fibre exit and
image slicers are also described in 3230AE0003 (R4 Fibre Injection Design). The performance
requirements of the SALT HRS R4 are described in 3200AE0015 (R4 FPRD), and the science
requirements and operation modes are discussed in 3200AE0018 (R4 OCDD). Some possible
upgrade paths are also described in this last document.

211
212 Appendix E. SALT HRS R4 optical design

E.2 SALT HRS R4 optical design

The complet e optical design of the SALT HRS R4 will be discussed in detail in this section.
Section E.3 will contain details of the spectrograph's performance. A discussion of the
merits of the design and justification for some of the design choices will be contained in
Section E.4. This section will also contain suggestions for further improving the design in
terms of either performance and/or cost.

E.2.1 Overview
The ray diagram of SALT HRS R4 is shown in Figure E.l and Figure E.2.

Blue pupil
mirror

Red pupil
Slit~ fID converter
mirror Entrance
/
CollimatorlPupil

j transfer mirror

Figure E.l: T he ray diagram of SALT HRS R4. The slit area is accessed by a fold mirror across the
echelle grating. The collimator serves as the first pupil mirror for both arms. A dichroic located just
after the int ermediate focus splits the spectrograph into red and blue arms. Each arm will have its own
VPH cross-disperser and camera (shown here as paraxial elements). The slit fore-optics are not shown.
E.2. SA LT HRS R4 optical design 213

Red pupil
mirror

Entrance slit

Figure E.2: A solid model view of the SALT HRS R4 optics. The red and blue cameras are shown as
paraxial elements.

The spectrograph is a dual beam white pupil design, with a single R4 echelle grating, a
dichroic beam-splitter, and VPH cross dispersers. The echelle grating has 41.6 grooves/mm
and is illuminated with a 200 mm diameter beam. Two fully dioptric cameras are used
to aquire complete wavelength coverage from 370 nm to 890 nm at a maximum resolving
power of R = 80000 and in a fixed spectral format. Up to two objects can be observed
simult aneously with a minimum separation between adjacent orders of II". The highest
resolving powers are obtained by using dual fibre image slicers.
Apart from shutters, the camera focusing, and the fibre interchange mechanism,
SALT HRS R4 will contain no moving parts. In order to provide complete immunity from
pressure and temperature changes the dispersive elements will be enclosed with a light
(2 hPa) vacuum. The entire instrument will in turn be housed in a temperature-stabilized
environment .

E.2.2 Fib re injection design

The details of the fibre injection design are described in a separate document (3230AE0003).
This document contains details of the fibres (including the possibility of operating with
micro-lenses) the slit fore-optics and the spectrograph's focal conversion optics. A brief
summary (of the optical design, which is the work of the current author) is given below.

Fibre modes
A summary of the SALT HRS R4 fibre modes, and their transmission (due to stellar PSF
overfilling and geometrical slicing) is given in Table E.1. Each of the fibre modes is as
follows:
1. low resolution fixed object and sky (FL); also used in nod and shuffle mode (NL),
2. medium resolution fixed object and sky (FM),
3. medium resolution nod and shuffle (NM),
4. high resolution fixed object and sky (FH), and
5. high resolution nod and shuffle (NH).
214 Appendix E. SALT HRS R4 optical design

The low resolving power mode (L) will deliver R = 17000, while the medium (M) and high
(H) resolving powers give R = 38400 and R = 76800 respectively. The "fixed" object
modes (F) will allow a single object to be observed with simultaneous sky. The "nod and
shuffle" modes (N) will allow more precise subtraction of the background sky.

Table E.l: Summary of the SALT HRS R4 fibre modes. The transmission has been calculated assuming
median seeing conditions and include fibre and image slicer vignetting only. (See 3230AE0003 for details.)

Fibre Fibre Slice Slice Resolving Trans-

mode diameter width number Power mission
(/Lm) (arcsec) (/Lm) (A/6A) (%)
1. FL/NL 500 2.23 16400 82
2. FM 500 2.23 160 3 38400 67
3. NM 350 1.57 160 2 38400 56
4. FH 350 1.57 80 3 76800 48
5. NH 250 1.12 80 3 76800 37

Slit fore-opt ics

The slit fore-o ptics convert the the fibre output f /3.8 into f /20 as accepted by the image
slicers. A triplet/doublet pair (LAL7 and N-FK517, all 20 mm in diameter) provides the
focal conversion, and a meniscus corrects the telecentricity. The design shown in Figure
9 allows for a field of ±1.22 mm at the fibre exit plane, or ±5.7"on the sky, which covers
the available inter-order space. The spectrograph's entrance slit is at the far right of
Figure E.3. T his will be the location of the Walraven-type image slicers. The fibre to slit
distance is 350 mm. The fibres in Figure 9 are depicted butted up to a flat glass plate
which can be easily anti-reflection overcoated.

Figure E.3: T he slit fore-optics convert from f /3.8 to f /20. The spectrograph's entrance slit is at the
right.

Focal conversion optics

A second set of focal conversion optics converts from f /20 back to f /10 as accepted by
the spectrograph's collimator while also maintaining telecentricity. A triplet consisting
of BK7/N-FK51/caF2 is shown in Figure E.4. However, it is probable that a doublet of
some other combination would be adequate. The first element (BK7) will also serve as
the window into the spectrograph's vacuum. There is 150 mm of path length between the
entrance slit and the vacuum window. As shown in Figure E.1, a small fold mirror located
approximately 350 mm after the vacuum window folds the light towards the collimator.
E.2. SALT HRS R4 optical design 215

-.-~------~

Figm:e E.4: The focal conversion optics provide the conversion from f /20 to f /10. The entrance slit
(on the left) is the same as depicted at the right of Figure E.3.

E.2.3 Collimator and blue pupil mirror

The collimator (M 1 ) is an f /10 off-axis parabaloid. This choice of focal length is a

compromise between an excessively large instrument (i.e., slow focal ratio), and a tolerably
fast focal ratio which can be readily manufactured. In order to accommodate the dual
use of the collimator as the first pupil transfer mirror, the total aperture must be 270 mm
x 570mm. Both this mirror and the blue pupil mirror (M 2 ), which is 220mm x 570 mm,
can be sourced from a single 720 mm diameter parabolic parent with a 4000 mm radius of
curvature (see Figure E.5).

720mm
I 5mm chamfers on all edges
--- ------ - - - ----
1-----
1
-- -- - - - --1
I
I
I
I
I
I

1 I
I I

+
1 N
I N

~
\
o
I
I
I
\
\ R=4000mm
§
I \
M2
§ I
I
I \
\

I I
I \ I I
I \ I I
I I

+
I I I I
I I CT:::75mm I I

I --Jo.I
I
:-0lil-----
I
I I I
I I
I I I I
I

§ \
\
\ I
I
I

I
\
\
Ml I
I

\ I R:::4000mm

+
\ I I
\
\
I
I
I
~ §
I I
I I
I I
I I

- - -------- - -
I I
------- ------- L ___ I

I. 570mm
.1
Figure E.5: Collimator (M 1 ) and blue pupil mirror (M2 ) dimensions.
216 Appendix E. SALT HRS R4 optical design

640mm
I 5mm chamfers on all edges I
- - - - - - -- - - -- ,,
r---'"
,
,,,
./
./ -- -- " "-
"-
,,
,/ "- ,
r - - - I
I I
I I

~ I
I
I
/
/
/

+ M3
\
\
\
\
\
\
~
R=3000mm
,,
§
I \
\
,,, ,,
I

+
,
I I CT=75mm , ,
\
I ~,
,, ' ...
I
,
,,,
\ I
\ I
\ I
520mm I
\ I
\

\ /
\ /

"" /
/

" -'
/
/
J
J
J
J

------- J

Figure E.6: Red pupil mirror (M3) dimensions.

E.2.4 Red pupil mirror

The red arm white pupil mirror is a 170 mm x 520 mm off-axis paraboloid with a 3000 mm
radius of curvature. This allows the red arm white pupil to be demagnified by a factor
of 1.33 (from 200mm to 150mm), which is better suited to the use of VPH gratings. The
red pupil mirror can be sourced from a 640mm diameter mirror (see Figure E.6).

E.2.5 Echelle grating

The splitting of the spectrograph into red and blue arms occurs just after the intermediate
focus following the first white pupil mirror. This enables the use of a single R4 echelle
grating mosaic, where each grating has the parameters given in Table E.2. The mosaic
has dimensions 855 mm x 204 mm, and has a gap of 35 mm between the gratings. -VVith
a collimated beam size of B = 200 mm, and FRD of 10%, there is no overfilling and about
5% of the light is lost to the gap between gratings. Each grating will be replicated on a
220 x 420 x 74mm block of Zerodur and the coating will be aluminium.

Parameter Specification
Blaze angle, BB 76.0°
Groove density, T 41.5 grooves/mm
Grating ruled width, Till 204mm Table E.2: The SALT HRS R4 grating
Grating ruled length, L 410mm parameters.
E.2. SALT HRS R4 optical design 217

The gratings are illuminated in quasi-Littrow mode; i.e e ~ 0 and I ~ O. The angle
e
of illumination with respect to the grating facet normal is = 0.35°. This allows a more
centred blaze function on each of the CCDS, and is in fact representative of the tolerance
in the blaze angle of each replicated grating.
A feature of large blaze angle echelle gratings is the considerable anamorphic magni-
fication they introduce. In the blue arm the anamorphic magnification (r) is in the range
0.82 < r < 1.22 from one side of the free spectral range to the other. In the red arm, the
spread is 0.76 < r < 1.41. This effect is relatively unimportant, except that it will cause
a variation in the sampling of each resolution element, and will lead to a small fraction of
the wavelengths being undersampled at the highest resolving powers.

E.2.6 Dichroic
The dichroic has a nominal wavelength division of 550 nm. It will be located a short
distance after the intermediate echelle spectrum. To capture all the light it must have a
clear aperture of 75 mm x 360 mm.

E.2.7 The cross-dispersers

The parameters of the VPH gratings have been optimized in order to provide the max-
ium possible order separation while maintaining complete wavelength coverage. Without
de magnifying the red arm pupil an rv450-line/mm grating would be required. This is
considered a low density for the efficient use of VPH gratings. Hence a demagnification
of 1.33 of the red pupil has been specified in order to allow a higher groove density to
be used. The short focal length transfer mirror used only in the red arm produces this
demagnification. Finally, while each grating is situated at the "white pupil", because of
the echelle anamorphic magnification each grating must have a clear aperture somewhat
larger than the projected beam size. The properties of the blue and red VPH gratings are
summarized in Table E.3. The gratings will be manufactured on a substrate of BK7 which
is 15 mm thick for the blue grating and 12 mm thick for the red. Identical cover glasses
will be applied to each grating. Each grating air/glass surface will be post-polished to at
least A/4 and anti-reflection coatings will be applied.

Grating Line Wavelength Clear Substrate

Table E.3: Parameters
density T range aperture diameter of the VPH gratings for
(lines/mm) (nm) (mm) (mm) SALT HRS R4.
Blue: 1050 370 < A < 560 260 280
Red: 650 550 < A < 890 195 215
218 Appendix E. SALT HRS R4 optical design

E.2.8 Cameras
Two cameras have been designed for SALT HRS R4 by D. Jones of Prime Optics. These are
described in detail in the document 3210AA0007 (R4 Camera). Ray diagrams of the two
cameras are shown in Figures E.7 and E.8.

Figure E.7: The SALT HRS R4 blue arm camera.

Figure E.8: The SALT HRS R4 red arm camera.

E.2. SALT HRS R4 optical design 219

The blue camera (Figure E.7) has a focal length of 300 mm and an effective focal ratio
of f /1.5. The diameter of the largest element is 356 mm, and the field size is 30.72 mm x
61.44 mm (i.e., a single 2k x 4k CCD) or 5.8° x 11.6°. The camera's optical performace
allows the Nyquist sampling to be limited by the 15J.lm CCD pixels.
The red camera (Figure E.8) has a focal length of 280 mm, which with a 150 mm
entrance pupil gives an effective focal ratio of f /1.88. The diameter of the largest element
is 278 mm. With a 61.44 mm x 61.44 mm field (i.e., either a 4k x 4k CCD, or a mosaic
of two 2k by 41<: CCDS) the camera's field of view is 12.3° x 12.3°. Again, the camera's
optical performance is sufficient (across most of the field) to allow the Nyquist sampling
to be detector-limited.
As described in the camera design document (3210AA0007), the combination of white
pupil optics and cross-dispersion at the white pupil causes the ideal focal plane to be
cylindrical. As shown in Figures E.g and E.I0, this curvature may be corrected by a
combination of spherical and cylindrical lenses. As depicted in these figures, each field-
flattening lens will be manufactured in two parts, the first element of which will act as
the CCD cryostat vacuum window.

IOuun lDmm

L,
4mm
41lun
R1 ;::-232.7 mm

.,.,o

70mm

95mm
I
Cemented
70llun

1100101
I
Cemented

Figure E.9: Blue camera field-flattening lens. Figure E.I0: Red camera field-flattening lens.

E.2.9 CCDS

The specifications of the CCDS are described in detail in the document series 3297AEOOXX.
It is assumed that a single 21<: x 41<: chip with 15 J.lm pixels is used in the blue camera.
This is sufficient to just capture a single free spectral range. The red camera will use a
mosaic of two 21<: x 4k CCDS, again with 15 J.lm pixels. In order to accommodate a nod
and shuffle mode, the red camera mosaic must have the columns aligned in the direction
of cross-dispersion. This means a small fraction of each order will be lost to the gap.
An alternative would be to use a single 4k x 4k CCD. There would be some advantage
in using such a chip in the blue as well. CCDS of this size are available from Fairchild
Imaging and Semiconductor Technology Associates. The suitability of these devices is yet
to be assessed.
220 Appendix E. SALT HRS R4 optical design

E.2.10 Exposure meter

It is proposed to insert a fold mirror in the gap between the echelle gratings. This
mirror will direct light towards a focusing element, which will in turn illuminate either
a photoelectric photometer or an avalanche photodiode. It is probable that a pair of
detectors will be used, so that optimization for each of the blue and red arms can occur.
This would also facilitate the use of different exposure times for each camera.
E.3. Performance 221

E.3 Per formance

E.3.1 Spectral formats
The spectral formats in each of the two spectrograph arms are shown in Figures E.ll and
E.12. The choice of cross-dispersion line density, and dichroic crosss-over wavelength has
been optimized to ensure that complete wavelength coverage from 370 nm to 890 nm is
obtained while maintaining the same minimum order separation in each arm. The spectral
formats are based on detailed ray-tracing using the complete Zemax optical models. The
exact CCD orientation (i.e., rotation) is yet to be fully optimized.

Order

30
85 I
~--
-
I
----
- ---
~y

23.9"
,,-,
87 \ ~ 22.9"
.1
20 89 I
r-
-"":' -52s- 21.9"
~
'",'luJ I
91 ~ 21 "
I
.-... -93 ,.."

~ .- ~

-
OJU' 20.1"

E 10 ~

~
I
-zJ9'2 19.2"
E
1JS
' I-'
,
II 18.4"
c
lJ7
gg- , , --zr82
.. 2 17.7"
o o ,,-, .
...... 101
~

" I
'-'
::2'63 16.9"
en .I. .
...... ,
o
Cl..
103
10S
I
,- ."'"
,-
.-
I
,~~"

'''''
---
-
454
445
16.2"
15.6"
..
>-
I
-10 101
10~
<7.
\'~ "~:

'"'
""
'" ,..... J
-
-== 437
429
15"
14.4"
111
113 ''i'
I",';/'~
"' ....,\"',.........
'"'
,- r-
- ,-,
"'.
-
----
421
413
13.8"
13.3"
,... '" '-'.......",,~'!' 'c- '-'

---
'v

-20
'-'
115 '-'
406 12.8"
117 ',-,' ''''' r- 399 12.3"
119 u,
~:=)
- 393 11.8"
121 -y

" - 386 11.4"

123 ''<-
- 380 10.9"
-30 125 374 10.5"

-20 -10 0 10 20
X-position (mm)
Figure E.1l: SALT HRS R4 blue camera spectral format . The extent of one free spectral range is shown by
the dashed lines. A single 2k by 4k detector with 15 p,m pixels is depicted by the rectangle. Wavelengths
correspond to the order centre.
 Order As t:,.y

30 53

55
\
\ ,...
'-'
.. '
~~"
-r
-!- ---
-r--
882

849
27.3"

25.5"

20
'-
\
57 ~ 820 23.8"
\ /

-E 10
59
\ I

-- --
-- -
792 22.3"

-
61 If"ItA\ I 766 20.9"
E r 2 "

I
-
--
--
63 \ 742 19.6"
C r:
0 7
.-t=
0 65 \ I-l ()("'\
2
- 719 18.4"
CJ) 67 1 I - - 697 17.3"

- --
0 f"I It
2
0... 69 "
1'-' "
'-'
I 677 16.3"

>-
I
-10 71
' .':'.
." ~"

--
-- 658 15.4"
-
~
.~
I
\
73
75
- t - - - -'- -
-- \
'H -
-
- -
--_n- -=--=-=-
II
81 1 -- -=1" 640
623
14.6"
13.8"
-20 77 II
II
II
607 13"
79 Hel Na~aD l II 591 12.3"
I I
81
83
I I
-- - '-'- - - - ---- 577
563
11.7"
11.1 "
-30 85 550 10.5"

-30 -20 -10 0 10 20 30

X-position (mm)

Figure E.12: SALT HRS R4 red camera spectral format . A pair of 2k by 4k detectors with 15 m pixels is depicted
by the rectangles. Wavelengt hs correspond t o the order centre.
E.3. Performance 223

E.3.2 Image quality

Slit fore-optics
The image quality of the slit fore-optics is shown in Figure E.13. The RMS image quality,
as measured at the spectrograph's f /20 entrance slit, is less than 20 p,m at all wavelengths
from 380 nm to 890 nm. This corresponds to a diameter at 80% encircled energy of around
25 MID. This equates to about 5% of the smallest resolution element, and hence the effect
on image quality will be negligible.

Figure E.13: Spot diagrams of the slit

fore-optics. The image is at the spec-
trograph's entrance f /20 slit and spots
at ±1.2 mm are shown. The scale bar is
200 pm and the bold circle represents the
diffraction limit.

The focal conversion optics

The image quality of the focal conversion optics is shown in Figure E.14. The optics
introduce considerable lateral colouration. However, this is not of concern as the final
imaging will occur after dispersion within the spectrograph. The RMS spot size is less
than 24 Mm across a field ±6 mm at the f /20 entrance slit and the diameter at EE(80) is
less than 20 Mm at each individual wavelength. This is less than 5% of the projected size
of the smallest resolution element.

Figure E.14: The image quality of the

focal conversion optics. There is consid-
erable lateral colouration. However, this
will not be perceived by the spectrograph.
The scale bar is 100 pm and the bold circle
represents the diffraction limit.

E.3.3 The white pupil optics

The image quality of the white pupil optics can be assessed by replacing the red and blue
cameras with paraxial cameras and by using a toroidal detector surface. The results for
the blue and red anus are given in Figures E.15 and E.16.
224 Appendix E. SALT HRS R4 optical design

25,----r----,----.-----,----,----,----,-----,----r----~~

~ 20
E

i:
~ 5r--\~\~0\\\\\\~\\I\\\1\\1\\f\~\
lli~'0.,lli),);!D~WjJJJ .
O~--~-----L----~--~----~----~----L---~----_L

360 380 400 420 440

llfu\
lliU/v\,. .
460
I' m~\
--\--\/.' V\·
480
.~"i ~.'.'
500 520
..J\_; T\J~--
~.~,) .~/.i
____L_~
540
b 560
Wavelength (nm)

~ 30
E
2.
~ 20 .... -- -.. . -- _......- _.- . -- ._- - _.. -._. . . . _.. -- -- . .- -. -- -" ._.- -........- -.. . -- -.- -..- .... . . - . ; '-- '\ --b"-'
--- . . .

110 c - - - ~. ~ ~ ~ ~ ~ ~ \-1 ~ ~ \-\ -\ \ \ \-\-\-\ -\ \ \ \ ~ \ ~ ~\~ i --

- - -i. \ \

\-illl~1illl\i~D,,\jjjJjj)Jjjv4\Jv\11JJ~)J)J,\'
OL---~-----L----~--~----~----~----L- __ ~ ____ ~ ____ L_~

360 380 400 420 440 460 480 500 520 540 560
Wavelength (nm)

Figure E.15: The RMS spot diameters and the diameter at 80% encircled energy for the SALT HRS R4
blue arm white pupil relay. The slit optics are not included and the camera is paraxial. Two free spectral
rangeH are covered by each order.

550 600 650 700 750 800 850 900

Wavelength (nm)

~ 30
E
2.

O~-L------~----~~----~-------L------~----~------~
550 600 650 700 750 800 850 900
Wavelength (nm)

Figure E.16: The RMS spot diameters and the diameter at 80% encircled energy for the SALT }IRS R4
red arm white pupil relay. The slit optics are not included and the camera is paraxial. Two free spectral
ranges are covered by each order.
E.3. Performance 225

It can be seen that the blue arm white pupil optics are essentially diffraction limited
(i.e., dRMS < 3 pm ) everywhere within one half of a free spectral range of the blaze
wavelength in each order. The red arm white pupil optics have slightly worse image
quality, but the RMS diameter is still less than 5 pm everywhere within one half of a free
spectral range of the blaze wavelength.

The spectrograph with cameras

The image quality of the cameras was assessed by the camera optical designer at selected
wavelengths (see 3210AA0007) and spot diagrams are shown in Figures E.17 and E.1S.

566.5811111 570.40 11111 572.9611111

"'.

rad =0.010
. :"
[!J
! ~ ~••• )

RMS rad =0.005 RMS rad =0.005

Order 82

' .. RMS
14.893,27.192 0.012,28.122 -11.884,29_048

464.6511111 467.72 11m 470.26 nm

Order 100

RMS rad = 0.008 RMS rad = 0.005 RMS rad = 0.005

14.558, -3.347 0.055,-2.669 -14.529,-1.685

371.9011m • 374.1811111 376.3311111

.. , .. ,..-----.-'-,
.... ".
Order 125
Figure E.17: Spot diagrams for the
SALT HRS R4 blue camera (v2.08). The boxes
RMS rad = 0.006 RMS rad = 0.006 are two pixels square. Figure courtesy of D.
0.030.-29.460 -15.274,-28.595 Jones.

905.54J'lfl!

Order 52

RMS rad =0.012 RIv1S rad =0.008 RIv1S rad = 0.008

28.431.24.269 '0.144,25.069 -23.130.26.700

734.29 11m 742.42 11m 748.42 11m

Ol'der 63

RIv1S rad = 0.008 RIv1S rad = 0.006 RIv1S rad = 0.006

28.101. -2.733 -0.145,-2.206 -28.572,-0.272

563.52 11m 568.12 11m

r------] Order 83 Figure E.18: Spot diagrams for the

L_.___·. SALT HRS R4 red camera (v2,01). The boxes
RIv1S rad = 0.011 RIv1S rad = 0.004 RIv1S rad = 0.007 are two pixels square. Figure courtesy of D.
24.801 ,-32.616 -0.142,-32.355 -28.832,,30.752 Jones.
226 Appendix E. SALT HRS R4 optical design

The image quality at all wavelengths from 370 nm to 890 nm is shown in Figures E.19
and E.20.
The image quality of the blue camera is excellent; at all wavelengths that fall on a
single 2k by 4k CCD the encircled energy within one pixel is greater than 80%. This is
sufficient to ensure that the Nyquist sampling limited resolving power (R = 80000) is
possible with minimal degradation due to the optics.
The image quality of the red camera is not quite as good .. From A = 625 nm to
A = 725 nm the encircled energy within one pixel is around 70%, which is below the
specified 80%. However, because the sampling limited resolving power of the red arm is
in fact R = 100000, the slight degradation due to image quality C. . . . 15 to 20%) will lower
the effective resolving power to match that of the blue arm.
E.3. Performance 227

380 400 420 440 460 480 500 520 540 560

!1::-=~~,mffmTf!(!rtr(rfrr r~-'f1=-
w

O~--~----~----~--~----~----~---J-----L----~--~~
360 380 400 420 440 460 480 500 520 540 560

.~
I30
20 -, .,. 1·\ \-\ H··
~. ~ V, \.~.\ \-\. .\ -\-. \-\ '\ ~~.
\. \_...:\ ~. \. \. . \--\ .\ . c, -~.
.1 -\ -~.\\- V:' . . ,
\ i

~
CD 10
r- - Yfj)(!fJC' \\iV~WJJ}(y.tw ~"'}i\v~~7\-. - r\r ;-\r.," +
~./'
-
W
W O~--~----~----~--~----~----~--~-----L----~---J~
360 380 400 420 440 460 480 500 520 540 560
Wavelength (nm)

Figure E.19: The RMS spot diameter, the encirled energy within one pixel, and the diameter at 80%
encircled energy for the SALT HRS R4 blue arm. The slit fore-optics and focal conversion optics are not
included.

OL-~ ____ ~L- ____-L______L -_ _ _ _ ~ _ _ _ _ _ _J -_ _ _ _ ~ ______ ~

550 600 650 700 750 800 850 900

OL-~ ____ ~L- ____ -L______L -_ _ _ _ ~ _ _ _ _ _ _J -_ _ _ _ ~ ______ ~

550 600 650 700 750 800 850 900

550 600 650 700 750 800 850 900

Wavelength (nm)

Figure E.20: The RMS spot diameter, the encirled energy within one pixel, and the diameter at 80%
encircled energy for SALT HRS R4 red arm, The slit fore-optics and focal conversion optics are not included.
228 Appendix E. SALT HRS R4 optical design

E.3.4 Efficiency
Wherever possible the measured efficiencies of equivalent components have been used.
Other efficiences are based on theoretical measurements supplied by the coating vendor,
or other coatings have been scaled according to the vendor's minimum specification.

SALT

The reflectance of each SALT mirror has been measured. The SAC reflectivities are from
witness samples made at the time of coating, while the primary mirror (aluminium)
reflectivity is that of a standard coating. The image quality is assumed to be EE(80) =
2.15".

FIF and fibres

For these efficiency calculations Optran uv fibres from Ceramoptec are assumed. It is
likely that either Polymicro FBP or an equivalent which uses Hearaus STU preforms will
be used in the final design. This will improve performance of the fibres at wavelengths
shorter than 400 nm and will also remove the presence of the OH feature at ,-,,;600 nm.
The fibres are assumed to have overcoated micro-lenses at both the input and output. A
multi-layer broad-band coating supplied from measurements made by Fisba Optik for a
HERCULES micro-lens has been assumed.

Slit fore-optics, image slicers and focal conversion optics

The material of all of the slit fore-optics, image-slicers and focal conversion optics is in
most cases that of the prescriptions described above. However, in some cases the glass
has been substituted for another with better uv transmissIon (e.g., silica or CaF 2 ) and
therefore the material absorption is insignificant. Each face of all the elements (including
the image slicers) is assumed to have a broadband coating from Denton (UBBAR).

Mirrors
The reflectivity of a selection of mirror coatings from Laserdyne Technologies (Queens-
land, Australia) is shown in Figure E.21. Each of the coatings, apart from the new uv
enhanced silver coat are from historical measurements. Data for the new uv enhanced
silver coat is from theoretical predictions supplied by Laserdyne. Each of Laserdyne's his-
torical coatings also match the minimum specification supplied by Spectrum Thin Films,
while the new uv coating is equivalent to the SAC mirror coatings. Currently the coatings
specified for each of the mirrors are:

Ml (collimator and first pupil mirror): uv enhanced silver,

lv/b (blue pupil mirror): enhanced aluminium, and
M3 (red pupil mirror): enhanced silver.
E.3. Performance 229

_.i.,~>::" .. ·~·"''''1'''·''~::;~-;,",-:':'-;':.:O:'.~:-::~:.7~·.7::~:~~-::.~~~:~~-:~-;~;:
95
: ", ~

:I
90 :/
(J) ;/
(.)
C :!
(1j 85 Of
1:5 1
(J) I

1i5 80.f
a: f
r~---- __----__--__----,
- Enhanced aluminium Figure E.21: The reflectivities
75 ._.- Enhanced silver of various coatings by Laserdyne .
.,',., UV enhanced silver
- _. New UV enhanced silver
The new UV enahnced coating is
70~X===C======C==~==~~__~__~__L-~ theoretical only.
400 450 500 550 600 650 700 750 800 850
Wavelength (nm)

98
96
94
~92
~
(J) 90
(.)

§ 88
g 86
:j::
QA
a:
(J)
U"T

82 - Enhanced aluminium
80 ,-, - Enhanced silver
" "" UV enhanced silver
78 Figure E.22: The UV close-up
- -. New UV enhanced silver
of the Laserdyne mirror reflectivities.
370 380 390 400 410 420 430 440 450
Wavelength (nm)

The new uv coating from Laserdyne has the potential to improve the response at
370 nm to 380 nm by over 10% if it were used on both the collimator and blue pupil
mirrors (see Figure E.22). However, the response between 450nm and 600nm would
actually decrease. Discussions are continuing with Laserdyne with regard to improved
mirror coatings. It is assumed that the small input fold mirror will have a very efficient
multi-layer coating.
230 Appendix E, SALT HRS R4 optical design

Dichroic
The transmission of a high efficiency dichroic supplied by Barr Assoicates is shown in
Figure E.23. The transmissions have been scaled to 95% of those shown here to reflect
the minimum specifications indicated by Barr Associates for the SALT HRS dichroic.

100 100
, ... ..,,,--_ .... _-- - -- - - - - ----- .... ~ ... _... ... -_ ............ -
90 " 90
J
" ~
I
80 80 I
I
I
#: 70 ~ 70 I
~
c 60 C
60
0 0
'iii 'iii
,!!l
E 50 'E'" 50
'"c 40 '"cf!! 40
~ .f-
I
30 30 I
I
20 20 I

I
10 10 , I

0 o520- _... - ...... .-

400 500 600 700 800 900 530 540 550 560 570 580 590
Wavelength (nm) Wavelength (nm)

Figure E.23: The dichroic efficiency, The blue wavelengths are reflected while the red wavelengths are
transmitted. A close-up of the crossover region is shown on the right,

VPH gratings
The theoretical efficiencies of the two VPH gratings required have been computed by
Wasatch Photonics (Figures E.24 and E.25).

lJro
o.oco
ow
{;'D.iW
c
.~ D£(O

""~
0
Ofro

~ 0.4:0
:!=
;S O;:ro
Figure E.24: Theoretical effi-
o~r(l
ciencies of a l050line/mm VPH
0,100 grating from Wasatch Photon-
OJXO ics, The calculations assume
370 :111 <Y.(, 4i(1 400 MO 8% Fresnel losses from un-
'--_ _ _ _ _ _ _ _ _ _ _ _ _ _~)_m_
...."'_~:1h
_ _ _ _ _ _ _ _ _ _ ___' coated surfaces.
E.3. Performance 231

1.(((1

O.qx, f--'''--;-';'''';;~+'~~

(,.O.;W ~.4'''~::':'-+'';'''';'';'''';'';'''';+-:-;;-';;;'''''+4+~~...;."t';;;''''';'''';'';'''';,.....,F,,,,,
c
·~o.m) ¥'4;"""~+;';"';+':"':"+h....;;;o~+4""":,"';';""";"~~"""';;";;";""""F"""''f.-...,-70:.r-:-""",,....:.;.;;:;.;
'l=
~o.~'f0~~~~~~~4f~~~7+±+~~~~~~~~~~~~~--~
o
~ o..w-f-',-'--,-"'-io-'-;++:..,:.,;.;...;....,-:'-:'-+-~"'+~"*~?~.;..,o.j,.:"...;..,.,.,.;..,....,;....;t-.....,.~~*:"-..:.::.,.,,;;.:
~ 1~~~~~~~~~~~~+2~~~~~~~~~
;So,J:(,+
Figure E.25: Theoretical effi-
0,10 +",,,;,;,,;;,;,;;,,,,;';;;;"++F ciencies of a 650 line/mm VPH
O. t(l +-.c.:....;:,;:..;;.";;.+...;;.".,,,,,,,.....,.,.,. grating from Wasatch Photon-
0.0:0 +--=~~-+~Z2¥:-2=.i24z~§12=2±1i~:2i.:.2:~~iJ ics, The calculations assume
7(()
/JX)
8% Fresnel losses from un-
~,
coated surfaces,

A comparision with a number of measured VPH gratings is given in Figure E.26. The
gratings are assumed to be anti-reflection coated with Rave < 0.7%.

100 ~-"""""--"-'~"""""-""'l~""'~~"'--~'-~'- ..... -..,-~""'---~~ ...."'--........... ~.....-'----'t'"...........,..,·C~-!

: : I .. _ t . ~ . l' , . . - ,'- . j _. , !
, I l ~ t i t ~
i
.... :..- ..... ~ .............. -:.............. t-- _........ : ..... _...... -:- ............. -t . . . . . . . '"':"' . . . . . . . . :........... -"'1
1 I I ~ 1 , ~

90
, I I I ~ I I 1 !
:tt ". .. - t : ?- ~ : : : !
80 ---- --~- --... -:.. ---- .~--- -. :~. Pr-i
70 --=:k--jd;~CL-----Lh~_~.'.!
~.~~:~ l '
J I I
•
i
.

i
~\ ...............
t
.J~_
:
........ "' ... I ...... ~ .........
I :
J. .. "' .......... 4 .. .,. ..... __ ... \.. ..........
:
~
I
... " ........ _ ... ...- ... 1.........
I
60 '" ......... ""

.,
() J 1 I , ~ j ,

c ( I I I I ,
Figure E.26: The measured
.~
,g
.i
,,
I I i
,, l l

50 ................ 1_ "' ............ J. ... ~ ... '" .. _--: ............ '" .. ~ ... ~ I ............... ",~ ......... ., ...... ~ efficiencies of VPH gratings sup-
:w >o. .............. ""'" ....

cCl

~ 40 .:-o;:,.".,..,_:;:,..;~.- -- --~- - ----

. . -------:-----.. --
-r--_ -~~ -~-- --~
.
plied by three different vendors
A, Band C (Dekker 2004) .
,;; , ,
is
30
., ,
...... - -- ~ .. - - . . -- -:- .. _. . _.... +, ......... ,.
.,
l ____ . . _~ ___ . . __ -~------~
, The efficiencies have been cor-
rected for reflection losses due
,, ,,
,
,...;:;;::-......::::""'-.. ., -... -... ~
to the grating substrate. Mea-
20 ------~-----~-j------~~-.--
: j
.. -~-------~------~--
j
.. -~
I surements at the blaze wave-
, I I ( T t 1 i
length are marked. The results
1 (I ~ ~ ~ ~ ~
.. _ .......... ............. _ ...... :_ . . . . . . . . . . . . . . . .. . . . . . . . . . _: .......... 00-" ... } . ' " _ ........ .;. . . . . . . . "' ..... _:_ ........... - } - - __ _

show quite a large variation in

I
,
k j Y •

., absolute grating efficiency be-

tween vendors. However, they
demonstrate that high through-
Incident angle (degrees) put is achievable.
232 Appendix E. SALT HRS R4 optical design

Cameras
Standard absorption data are used for each of the camera lens elements. A generic broad-
band coating has been specified on each element apart from the field-flattening lens which
is single-layer MgF 2 . Both Laserdyne and Spectrum Thin Films have supplied theoretical
data for the air/glass overcoated surfaces. Laserdyne's data are shown in Figure E.27
and Figure E.2S. The transmission of the oil interfaces has been computed using Fresnel
losses, where the oil's refractive index is assumed to be an average of that of the two
glasses being interfaced. The resultant losses are negligible.

- Laserdyne BSM51Y blue

, - ,. Laserdyne FPL51 Y blue
""" Laserdyne FPL51Y red
- - Laserdyne LAL7 blue
- Laserdyne LAL7 red

£
~ 0.6
~ ,

~ .~; n/\:""'"
0.4
\
\
' I \ ',.
0.2 . '''''' _ .~. ~(': ;.,........ .
'0 ._.- '.:k.J' ...... ~ "-
o'----40O-"-'----50 O---'--'---60'-0---7-"0'-0---8--'0-0----"""--900
L L
Figure E. 2 7: The reflectivity ofthe
Wavelength (nm) multi-layer coatings from Laserdyne.

3 - Laserdyne SAR on BSM51Y blue

,-,. Laserdyne SAR on FPL51Y red
.. ,. " Laserdyne SAR on silica blue
2.5 - - Laserdyne SAR on silica red

0.5 "" ........... _.,.."..'"

"', .... , .. ,.,., W " ,.. ... ~----
........... ,.. ""'" __ ~ ~ Figure E.28: The reflectivity of
L
o'----40-'-0--'-"-'---50 O---6-0'-0--=---'-"'--7-"0'-0---8--'0-0-----'900 the single-large MgF 2 coatings from
Wavelength (nm) Laserdyne.
E.3. Performance 233

Summary
The computed efficiencies of SALT HRS R4 are given in the following tables. The spec-
trograph efficiencies (spc) include everything except the telescope (TEL) and slit optics
(SLT). The slit optics include fibres, slit fore- optics, and (where appropiate) image slicers.
The other items are the collimator (COL), which includes FRD, the echelle grating (ECH),
the cross-dispersers (XDP), the camera (CAM), and the CCD. The camera includes the
white pupil mirrors and dichroic.

Table E.4: Detailed efficiciencies of the SALT HRS R4 blue arm in "Fixed Object" mode at the lowest
resolving power
Wavlength Totals
and Component by component SPC+
Order efficiencies SPC+ SLT+
A m TEL SLT COL ECH XDP CAM CCD SPC SLT TEL
380 123 0.622 0.570 0.737 0.553 0.804 0.611 0.674 0.135 0.077 0.048
480 97 0.642 0.688 0.900 0.605 0.910 0.800 0.803 0.319 0.219 0.141
540 86 0.630 0.681 0.890 0.589 0.737 0.799 0.795 0.245 0.167 0.105

Table E.5: Detailed efficiciencies of the SALT HRS R4 red arm in "Fixed Object" mode at the lowest
resolving power.
Wavlength Totals
and Component by component sPc+
Order efficiencies SPC+ SLT+
A m TEL SLT COL ECH XDP CAM CCD SPC SLT TEL
560 83 0.631 0.684 0.890 0.584 0.835 0.788 0.923 0.315 0.216 0.136
650 72 0.617 0.708 0.884 0.571 0.951 0.816 0.898 0.352 0.249 0.154
800 58 0.602 0.724 0.884 0.565 0.596 0.794 0.794 0.188 0.136 0.082
234 Appendix E. SALT HRS R4 optical design

Fibre Resolving Transmissions

mode power
A/OA SPC SPC + SLT SPC + SLT + TEL
FL/NL 16400 31.9% 21.9% 14.1%
FM 38400 31.9% 17.0% 10.9% Table E.6: Summary of ef-
13.6% 8.7% ficiencies of the SALT HRS R4
NM 38400 31.9%
blue arm at all resolving pow-
FH 76800 31.9% 12.1% 7.8% ers and modes at a wave-
NH 76800 31.9% 9.1% 5.8% length of 480 nm.

Fibre Resolving Transmissions

mode power
A/OA SPC sPc+ SLT SPC + SLT + TEL

FL/NL 16000 35.2% 24.9% 15.4%

FM 38000 35.2% 19.3% 11.9% Table E.7: Summary of ef-
ficiencies of the SALT HRS R4
NM 38000 35.2% 15.4% 9.5%
red arm at all resolving pow-
FH 77000 35.2% 13.8% 8.5% ers and modes at a wave-
NH 77000 35.2% 10.3% 6.4% length of 650 nm.

E.3.5 Signal to noise predictions

Using the efficiency calculations from Section E.3.4 the signal to noise expected for a
range of observing conditions has been computed. The calculations assume a G dwarf
star and a telescope airmass of X = 1.3. The Moon is assumed to be at quarter phase.
The calculated signal-to-noise ratios (S/N) are for each extracted half-resolution element
(to allow correct Nyquist sampling).
The predicted signal to noise ratio of the red arm (at 650 nm) at the lowest resolving
power is shown in Figure E.29. With a signal to noise ratio of S/ N = 10, the limiting
magnitudes in all other observing modes for a 3600-second exposure are as follows:
Blue arm (A = 480nm):
FL/NL: 11 = 20.7,
FM: 11 = 19.7,
NM: 11 = 19.5,
FH: 11 = 18.7, and
NH: 11 = 18.4.
Red arm (A = 650nm):
FL/NL: 11 = 21.0,
FM: 11 = 20.0,
NM: 11 = 19.8,
FH: 11 = 19.0, and
NH: 11 = 18.7.
E.3. Performance 235

At a wavelength of ,,\ = 650 nm, at the lowest resolving power, it will be possible to
reach a signal to noise of SIN = 100 in 5 minutes for a 11 = 14.4 magnitude object, or If
= 15.6 in 15 minutes, and 11 = 16.3 in a half-hour exposure.
At the highest resolving powers, at a wavelength of ,,\ = 650 nm, a signal to noise ratio
of SIN = 100 will be achieved after 5 minutes for 11 = 12.1, after 15 minutes for 11 =
13.4, after 30 minutes for 11 = 14.0 and after one hour for If = 14.8.

". - 300s
......... ' .................
- - 900s
3
.-. 10 1800s
Z
........ - - 3600s
(f) '.
........
0
'';::
en
.... 10
2
Q)
C/)
'0' '. Figure E.29: The
c
0 predicted signal to
...... 1 '. , , noise ratio (SjN)
en 10 " '-
c ". "',
'-
'- "-
. of SALT HRS R4 at
OJ
U5 '- A = 650 nm in "Fixed
object" fibre mode at
10° the lowest resolving
powers.
10 12 14 16 18 20 22
Visual magnitude (mJ
236 Appendix E, SALT HRS R4 optical design

EA Discussion
EA.1 The white pupil optics
A suggestion by B. Delabre made at the time of the 2004 September PDR for the
SALT HRS R2 design was that each of the white pupil transfer mirrors be made spheri-
cal. This possibility was explored. However the large beam size and short focal length
of SALT HRS R4 prevents such a system from achieving the required image quality. The
system was found to be adequate only when a refractive correcting element was intro-
duced prior to the VPH gratings. Such a catadioptric white pupil relay is in fact being
used on PEPSI for the LBT (see Figure E.30) where each white pupil mirror has a com-
panion correcting lens. The complication to the optical layout, the added expense, and
the additional light losses from between 4 and 6 air/glass reflections support the use of
the marginally more difficult off-axis parabolic mirror system.

Cl

",,'tt::::

._-- ~~-~"~~;~ ,_,~,_",,,_",:':].~3~w)

C2

Cl' M~11l CQrI'.Ct(!f "nll. Mit· 1'01o:l109114r mirror M5· f<lIdIMQ lJ4t mirror
~1' Main $p""tlC~l rnlrror M:) • sllW trdnsf~r cohlmator M';: • R"" trl:lnst"r co Ulmalor
M<I • FQIGI1l9 mlrrQf Cl • UI!.I\' earm-roil c;: • f.t"o (:4!1WrCl
Flo 'l'i<>kt~lliiC<

Figure E.30: The PEPSI catadioptric white pupil relay (Pallavicini et al., 2003),

EA.2 VPH gratings

The 2004 January concept for the SALT HRS R4 used VPH grisms for cross-dispersion. A
small increase in the average order separation can be achieved by this means. However,
the gain was not considered sufficient to warrant the added complexity to the design and
manufacture of the cross-dispersers. Nevertheless, it will be necessary to use grisms if
VPH gratings other than those installed at the time of commissioning are to be considered
in future. This will avoid the need to articulate the camera.

EA.3 Cameras
This section explores some of the camera design issues. The effect of using the first
element as a vacuum window is discussed first, and the remainder of the issues relate to
E.4. Discussion 237

the various means by which the cameras could be made simpler, smaller and possibly less
expensive. The options presented here will only impact on the total system throughput,
and are not expected to degrade the image quality. There is in fact some scope for further
improving the camera's imaging performance.

Vacuum windows
The elements of the spectrograph camera will be held at atmospheric pressure in order
to simplify the maintanance of the oil coupling between the various multiplets that make
up each camera. Therefore, the first lens will be used as a window between the partial
vacuum inside the spectrograph and the external atmosphere. This lens will be deformed
by this pressure difference and there will be some effect on image quality.
The deformation of the lens can modelled using finite element analysis (FEA). A
constant force (pressure) is assumed to be acting across the surface of the lens which
is held rigidly against an annulus (the "o-ring"). The deformed surface can then be
described by a general asphere, where the aspheric terms come from a polynomial fit to
the deviation of the surface from the undeformed sphere. At the time of writing only
the red camera vacuum window has been modelled in detail. The deformation of the red
camera window is shown in Figure E.3l.

Figure E.31: The FEA of the red camera vacuum window. The mesh is an exaggerated profile of the
deformed lens. The total deformation is nearly 3 pm and can be well modelled by a 4th order polynomial
(see Figure E.32). This surface was used to assess the effect on image quality and spot diagrams are
shown in Figure E.33. The forces within the glass « 2 N /mm 2 ) are far from the theoretical yield point
( <10%).

The effect on image quality is small (Figure E.33). This can be explained by noting
that while the total deformation of each lens appears quite large (nearly 3 Mm, or several
wavelengths) the deformed surface is itself still quite well described by a displaced sphere.
Figure E.34 shows the residuals of the deformed surface and the best fit sphere. The RMS
residual is around ),,/8, which is considerably less than the surface quality required on
individual camera elements.
In principle the camera could be slightly reoptimized in order to compensate entirely
for the vacuum window deformation. This could be done by ensuring that the best fit
238 Appendix E. SALT HRS R4 optical design

E 0
.6
§ -1
~
E -2
~
o -3
0 . 20 40 60 80 100 120 140

E 0.05
-S
(f)
"iii
:::J
'1:l
Figure E.32: The deformation of the red
'iii camera vacuum window with respect to
OJ
a: -0.05 the original sphere (upper panel) and the
0 20 40 60 80 100 120 140 residual of a 4th order polynomial fit to
Radius (mm) this deformation (lower panel).

sphere of the deformed surface is actually the prescribed sphere. However, in the case of
the red camera the radius of curvature of the best fit sphere is less than 0.1 mm different
from the original sphere. This is well within the ±0.7 mm allowed by manufacturing tol-
erances alone (see 3210AA0007 R4 Camera) and hence is not worth considering.

Figure E.33: Spot diagrams showing the ef-

fect on image quality of the deformation of the
camera vacuum window. Spots are shown for
wavelengths in the centre of orders at the bot-
tom (blue), middle (green), and top (red) ofthe
field. Each box is two pixels square.
E.4. Discussion 239

0.5

0.4

0.3

0.2 ...-
-..
50
CJ)

...
.s::::.
C)

E 0.1 cCl)
E
"- >
Cl)

~
(J)
:::l
0 0 ~
-C ~
m :::J
0:: -0.1 "'C
"Ci)
Cl)
-50 -0.2 0::

-0.3
-100
-0.4

-0.5
-100 -50 0 50 100
Radius (mm)

Figure E .34 : The residuals of the red camera vacuum window with respect to the best fit sphere.
240 Appendix E. SALT HRS R4 optical design

Camera aper tures

Currently the cameras are able to accept 100% of the light from any wavelength within
one half a free spectral range of the blaze wavelength in each order. The camera apertures
therefore become quite large because of the effect of echelle anamorphic magnification.
If the cameras have their apertures reduced to approximately 85% of the design values,
the resultant vignetting will be as shown in Figures E.35 and E.36. A veraged over all
wavelengths within the free spectral range the total light loss is about 2%. Wavelengths
near the short er wavelength end of each order suffer worst. In the blue camera the
maximum loss is 5%, while in the red camera the maximum loss is as much as 15%.
However, the significant savings in cost and manufacturability of each camera suggest
that this is a reasonable tradeoff to make.

100

30 95

20
90
..-
E 10
--
E
c
85
o
:E o
U)
o 80
0..
~ -10

75
-20

70
-30

65
-20 -10 0 10 20
X-position (mm)

Figure E.35: The vignetting of the blue camera with reduced apertures.
E.4 . Discussion 241

100

95

90

--EE 10
---c
0
85

:E 0
CJ)
0 80
c.
I -10
>-
-20 75

-30 70

65
-30 -20 -10 0 10 20 30
X-position (mm)

Figure E.36 : The vignetting of the red camera with reduced apert ures .
242 Appendix E. SALT HRS R4 optical design

Field-flattening lens
The current field-flattening lens is shown in Figure E.37a. This design requires that the
distance between the field-flattening lens and the CCD be as little as 1.5 mm. This is
smaller than current ly perceived possible by the preferred detector subcontractor and an
alternative solution is shown in Figure E.37b. This design requires that the field-flattening
lens be constructed from at least three pieces of two types of glass (e.g., lal7 and silica).
Only the last element (silica) is cylindrical. The distance between the CCD and the lens
can be increased to more than 3 mm before the image quality deteriorates significantly.

(a) (b)

Figure E.37: T he current camera field-flattening lens (a) and an alternative design (b) which would
increase the spacing between the CCD and this lens.
Bibliography

Alekseeva, G. A.; Arkharov, A. A.; Galkin, V. D.; Hagen-Thorn, E. 1.; Nikanorova, 1. N.; Novikov,
V. V.; Novopashenny, V. B.; Pakhomov, V. P.; Ruban, E. V. and Shchegolev, D. E.: "Pulkovo
Spectrophotometric Catalog (Alekseeva+ 1997)", VizieR Online Data Catalog, 3201, pp. 0-+, 1997
Sep.
Angel, J. R P.; Adams, M. T.; Boroson, T. A. and Moore, R L.: "A very large optical telescope array
linked with fused silica fibers", ApJ, 218, pp. 776-782, 1977 Dec.
Baranne, A.; Queloz, D.; Mayor, M.; Adrianzyk, G.; Knispel, G.; Kohler, D.; Lacroix, D.; Meunier, J.-P.;
Rimbaud, G. and Yin, A.: "ELODIE: A spectrograph for accurate radial velocity measurements.",
At9AS, 119, pp. 373-390, 1996 Oct.
Barden, S. C.; Arns, J. A.; Colburn, W. S. and Williams, J. B.: "Volume-Phase Holographic Gratings
and the Efficiency of Three Simple Volume-Phase Holographic Gratings", PASP, 112, pp. 809-820,
2000 Jun.
Barden, S. C. and Wade, R A.: "DensePak and spectral imaging with fiber optics", in "ASP Conf. Ser.
3: Fiber Optics in Astronomy" , pp. 113-124, 1988.
Barnes, S. and Albrow, M.: "Revised SALT HRS concept proposal", A presentation to the SALT Science
Working Group, 2002 Oct.
Barnes, S.; Cottrell, P.; Albrow, M. and Kershaw, G.: "3210AEOOOl: SALT HRS optical design defintion
document", A presentation to the SALT Science Working Group, 2003 Sep.
Barnes, S. 1.; Clark, M.; Cottrell, P. L.; Hearnshaw, J. B.; Petterson, O. K. L.; Pollard, K. R.; Pritchard,
J. D.; Richards, A. and Tobin, W.: "Charac.teristics ofthe Mt John Series 200 CCD system", Southern
Stars, 39, pp. 1-8,2000.
Bernstein, G. M.; Athey, A. E.; Bernstein, R.; Gunnels, S. M.; Richstone, D. O. and Shectman, S. A.:
"Volume-phase holographic spectrograph for the Magellan telescopes", in "Proc. SPIE Vol. 4485, p.
453-459, Optical Spectroscopic Techniques, Remote Sensing, and Instrumentation for Atmospheric
and Space Research IV, Allen M. Larar; Martin G. Mlynczak; Eds." , pp. 453-459, 2002 Jan.
Bingham, R G.: "Grating spectrometers and spectrographs re-examined", QJRAS, 20, pp. 395-421,
1979 Dec.
Bottema, M.: "Echelle efficiencies - theory and experiment; comment", Appl. Opt., 20, pp. 528-530,1981
Feb.
Bowen, 1. S.: "The Image-Slicer a Device for Reducing Loss of Light at Slit of Stellar Spectrograph." ,
ApJ, 88, pp. 113-+, 1938 Sep.
Breger, M.: "Catalog of spectrophotometric scans of stars.", Astrophys. J., Suppl. Ser., 32, pp. 7-87,
1976 Sep.
Brown, T. M.: "High precision Doppler measurements via echelle spectroscopy", in "ASP Conf. Ser. 8:
CCDs in astronomy" , pp. 335-344, 1990.
Brown, T. M.; Noyes, R. W.; Nisenson, P.; Korzennik, S. G. and Horner, S.: "The AFOE: A spectrograph
for precise Doppler studies", PASP, 106, pp. 1285-1297,1994 Dec.
Buckley, D. A. H.: "Annual Report of the SAAO", p. 35, 1995.
Buckley, D. A. H.: "SALT Science Requirements", SALT Document DB000531, 2000.
Buckley, D. A. H.: "Summary of SALT HRS Preliminary Design Review", SALT internal document, 2003
September.
Buckley, D. A. H.; Cottrell, P. L.; Nordsieck, K. H.; O'Donoghue, D. E. and Williams, T. B.: "The
first-generation instruments for the Southern African Large Telescope (SALT)", in "UV and Gamma-
Ray Space Telescope Systems. Edited by Hasinger, Gunther; Turner, Martin J. L. Proceedings of the
SPIE, Volume 5492, pp. 60-74 (2004).", pp. 60-74,2004 Sep.
Buckley, D. A. H. and Sessions, N.: "FIF pre-CDR", SALT document SALT-3400-AE-0024, 2004.
Carrasco, E. and Parry, 1. R: "A method for determining the focal ratio degradation of optical fibres for
astronomy", MNRAS, 271, pp. 1-12,1994 Nov.
Castilho, B. V.; Delabre, B.; Gneicling, C. D. and Tull, R G.: "New solution in echelle crossdispersing
- the soar telescope echelle spectrograph", Bv,lletin of the Astronomical Society of Bmzil, 23, pp.

243
244 BIBLIOGRAPHY

195-195,2003 Aug.
Clemens, J. C. and Seagroves, S.: "Volume holographic gratings", Tech. rep., Goodman Laboratory, 1999.
Comte, A.: Gours de Philosphie Positive, chap. II, 19th lesson, 1835.
Cuillandre, J.; Fort, B. P.; Picat, J. and Soucail, G.: "Back and forth spectroscopy: optimization of an
optical nod-and-shuffle technique to reach fainter objects and increase the multiplex gain on multi-
object spectrographs", in "Instrument Design and Performance for Optical/Infrared Ground-based
Telescopes. Edited by lye, Masanori; Moorwood, Alan F. M. Proceedings of the SPIE, Volume 4841,
pp. 1531-1538 (2003).", pp. 1531-1538,2003 Mar.
Cummings, 1. N.; Hearnshaw, J. B.; Kilmartin, P. M. and Gilmore, A. C.: "High-Precision Radial-Velocity
Measurements of Late-Type Evolved Stars", in "ASP Conf. Ser. 185: IAU Colloq. 170: Precise Stellar
Radial Velocities" , pp. 204-210, 1999.
D'Arrigo, P.; Bingham, R.; Diego, F.; Crawford, 1.; A., C.; Savidge, T. and Percival, J.: "Optical Design
Definition Document for High Resolution Optical Spectrograph (HROS) (CDR version)", Tech. rep.,
University College London, 2000 Apr.a.
D'Arrigo, P.; Bingham, R. G.; Charalambous, A.; Crawford, 1. A.; Diego, F.; Percival, J. F. and Savidge,
T. E.: "Design of the high-resolution optical spectrograph (HROS) for the Gemini telescope", in
"Proc. SPIE Vol. 4008, p. 159-166, Optical and IR Telescope Instrumentation and Detectors, Masanori
lye; Alan F. Moorwood; Eds.", pp. 159-166,2000 Aug.b.
Dekker, H.; D'Odorico, S.; Kaufer, A.; Delabre, B. and Kotzlowski, H.: "Design, construction, and
performance of UVES, the echelle spectrograph for the UT2 Kueyen Telescope at the ESO Paranal
Observatory", in "Proc. SPIE Vol. 4008, p. 534-545, Optical and IR Telescope Instrumentation and
Detectors, Masanori lye; Alan F. Moorwood; Eds.", vol. 4008, pp. 534-545,2000 Aug.
Dekker, H.; Nissen, P. E.; Kaufer, A.; Primas, F.; D'Odorico, S. and Hanuschik, R. W.: "High SIN, high
resolution Image Slicer observations with UVES" , in "Specialized Optical Developments in Astronomy.
Edited by Atad-Ettedgui, Eli; D'Odorico, Sandro. Proceedings of the SPIE, Volume 4842, pp. 139-150
(2003)." , pp. 139-150, 2003 Feb.
Diego, F. and Walker, D. D.: "On the possibility of increasing the throughput of astronomical spectro-
graphs by overfilling the dispersing element", MNRAS, 217, pp. 347-354, 1985 Nov.
Epps, H. W. and Vogt, S. S.: "Extremely achromatic f/1.0 all-spherical camera constructed for the high-
resolution echelle spectrometer of the Keck telescope" , Appl. Opt., 32, pp. 6270-6279, 1993 Nov.
Erasmus, A.: "SALT Site Survey Results", SAAO Newsletter, p. 32, 1999.
Fabricant, D. G.; Szentgyorgyi, A. and Epps, H. W.: "Segmented Zero-Deviation Cross-Dispersion Prisms
for the Hectochelle Multiobject Spectrograph", PASP, 115, pp. 235-242,2003 Feb.
Glazebrook, K. and Bland-Hawthorn, J.: "Microslit Nod-Shuffle Spectroscopy: A Technique for Achieving
Very High Densities of Spectra", PASP, 113, pp. 197-214, 2001 Feb.
Harrison, G. R.: "The production of diffraction gratings: 1. the design of echelle gratings and spectro-
graphs", J. Opt. Soc. Am., 39, pp. 522-528, 1949.
Harrison, G. R.; Loewen, E. G. and Wiley, R. S.: "Echelle gratings: their testing and improvement",
Appl. Opt., 15, pp. 971-976, 1976 Apr.
Heacox, W. D.: "Radial image transfer by cylindrical, step-index optical waveguides", Optical Society of
America, Journal, A: Optics and Image Science (ISSN 0740-3232), vol. 4, March 1987, p. 488-493.,
4, pp. 488-493, 1987 Mar.
Hearnshaw, J.: The analysis of stari'tght: one hundred and fifty years of astronomical spectroscopy, Cam-
bridge University Press, 1986.
Hearnshaw, J.; Barnes, S.; Cottrell, P. and Kershaw, G.: "Draft Preliminary Design Document for
CELESTIA", A presentation to the SALT Science Working Group, 2001 Oct.
Hearnshaw, J.; Rumsey, N. and Nankivell, G.: "Some comments on the design of echelle spectrographs
using R2 or R4 gratings for precise radial velocity measurements", in "ASP Conf. Ser. 185: lAD
Colloq. 170: Precise Stellar Radial Velocities", pp. 29-35, 1999.
Hearnshaw, J. B.: "The Cassegrain Echelle Spectrograph at Mt. John Observatory", Proceedings of the
Astronomical Society of Austral,ta, 3, pp. 102-103, 1977 Sep.
Hearnshaw, J. B.: "An echelle spectrograph", Sf3T, 56, pp. 6-8, 1978 Jul.
Hearnshaw, J. B.; Barnes, S. 1.; Kershaw, G. M.; Frost, N.; Graham, G.; Ritchie, R. and Nankivell, G. R.:
"The Hercules Echelle Spectrograph at Mt. John", Experimental Astronomy, 13, pp. 59-76,2002.
Hill, J. M.; Angel, J. R. P.; Scott, J. S.; Lindley, D. and Hintzen, P.: "Multiple object spectroscopy -
The Medusa spectrograph", ApJ, 242, pp. L69-L72, 1980 Dec.
BIBLIOGRAPHY 245

Hubbard, E. N.; Angel, J. R P. and Gresham, M. S.: "Operation of a long fused silica fiber as a link
between telescope and spectrograph", ApJ, 229, pp. 1074-1078,1979 May.
Huggins, W.: The Nineteenth Century Review, 1897 Jun.
Hunter, T. R and Ramsey, L. W.: "Scrambling properties of optical fibers and the performance of a
double scrambler", PASP, 104, pp. 1244-1251,1992 Dec.
Jacquinot, P.: "The lminosity of spectrometers with prisms, gratings, or Fabry-Perot etalons",
J. Opt. Soc. Am., 44, pp. 761-765, 1954.
Kaufer, A.: "A Two-Beam Two-Slice Image Slicer for Fiber-Linked Spectrographs", in "ASP Conf. Ser.
152: Fiber Optics in Astronomy III", pp. 337-+, 1998.
Kaufer, A.; Stahl, 0.; Thbbesing, S.; N0rregaard, G., P.and Avila; Francois, P.; Pasquini, L. and Pizzella,
A.: "Commissioning FEROS, the new high-resolution spectrograph at La Silla", Messenger, 95, pp.
8-18,1999.
Kershaw, G. M. and Hearnshaw, J. B.: "High precision radial velocities using an optical fibre feed",
Southern Stars, 33, pp. 89-97, 1989.
Kirchhoff, G. and Bunsen, R: Poggendorf/'s Ann., 110, 1860.
Kitchin, C. J.: Optical Astronomical Spectroscopy, Institute of Physics Publishing, 1995.
Knyazeva, L. N. and Kharitonov, A. V.: "Energy distribution in stars at 320-760 nm (Knyazeva+, 1996)",
VizieR Online Data Catalog, 3186, pp. 0-+, 1996 Nov.
Libbrecht, K. G. and Peri, M. L.: "A fiber-fed echelle spectrograph for the hale 5-m telescope", PASP,
107, pp. 62-67, 1995 Jan.
MacQueen, P. J.: Solid-state image detector development: a Linear Diode Army for astronomical spec-
troscopy, Ph.D. thesis, University of Canterbury, 1986.
Mayor, M.; Pepe, F.; Queloz, D.; Bouchy, F.; Rupprecht, G.; Lo Curto, G.; Avila, G.; Benz, W.; Bertaux,
J.-L.; Bonfils, X.; dall, T.; Dekker, H.; Delabre, B.; Eckert, W.; Fleury, M.; Gilliotte, A.; Gojak, D.;
Guzman, J. C.; Kohler, D.; Lizon, J.-L.; Longinotti, A.; Lovis, C.; Megevand, D.; Pasquini, L.; Reyes,
J.; Sivan, J.-P.; Sosnowska, D.; Soto, R.; Udry, S.; van Kesteren, A.; Weber, L. and Weilenmann, U.:
"Setting New Standards with HARPS", The Messenger, 114, pp, 20-24,200.3 Dec.
Meiring, J. G. and Buckley, D. A. H.: "Southern African Large Telescope (SALT) project progress and
status after four years", in "Proceedings of the SPIE, Volume 5489, pp. 592-602 (2004).", pp. 592-602,
2004 Oct.
Merrill, P. W.: "A Plane-Grating Spectrograph for the Red and Infra-Red Regions of Stellar Spectra",
ApJ, 74, pp. 188-+,1931 Oct.
Michelson, A. A.: "The Echelon Spectroscope", ApJ, 8, pp. 37-+, 1898 Jun.
Moffat, A. F. J.: "A Theoretical Investigation of Focal Stellar Images in the Photographic Emulsion and
Application to Photographic Photometry", A8A, 3, pp. 455-+, 1969 Dec.
Mohr, J. L.; Johnston, R A.; Lane, R G. and Cottrell, P. L.: "Adaptive optics at Mount John University
Observatory", in "Proceedings of Image and Vision Computing New Zealand 2004", pp. 233-236,
Akaroa, New Zealand, 200421-23 November.
Murdoch, K. A.; Hearnshaw, J. B. and Clark, M.: "A search for substellar companions to southern
solar-type stars", ApJ, 413, pp. 349-363, 1993 Aug.
Nankivell, G. R and Rumsey, N. J.: "The Optical System of the MT.JOHN I-METRE Telescope", in
"lAD Symp. 118: Instrumentation and Research Programmes for Small Telescopes", pp. 101-+, 1986.
Noguchi, K.; Aoki, W.; Kawanomoto, S.; Ando, H.; Honda, S.; Izumiura, H.; Kambe, E.; Okita, K.;
Sadakane, K.; Sato, B.; Tajitsu, A.; Takada-Hidai, T.; Tanaka, W.; Watanabe, E. and Yoshida, M.:
"High Dispersion Spectrograph (HDS) for the Subaru Telescope", PASJ, 54, pp. 855-864,2002 Dec.
Nordsieck, K. H.; Burgh, E. B.; Kobulnicky, H. A.; Williams, T. B.; O'Donoghue, D.; Percival, J. W. and
Smith, M. P.: "The Prime Focus Imaging Spectrograph for the Southern African Large Telescope",
Bulletin of the American Astronom'tcal Society, 33, pp. 1465-+, 2001 Dec.
o 'Donoghue, D.; Bauermeister, E.; Carter, D. B.; Evans, G. P.; Koorts, VlT. P.; O'Connor, J.; Osman,
F.; van der Merwe, S. and Bigelow, B. C.: "SALTICAM: $0.5M acquisition camera: every big
telescope should have one", in "Instrument Design and Performance for Optical/Infrared Ground-
based Telescopes. Edited by lye, Masanori; Moorwood, Alan F. M. Proceedings of the SPIE, Volume
4841, pp. 465-476 (2003).", pp. 465-476,2003 Mar.
Pallavicini, R; Zerbi, F. M.; Spano, P.; Conconi, P.; Mazzoleni, R; Molinari, E. and Strassmeier, K. G.:
"The ICE spectrograph for PEPSI at the LBT: preliminary optical design", in "Instrument Design
and Performance for Optical/Infrared Ground-based Telescopes. Edited by lye, Masanori; Moorwood,
246 BIBLIOGRAPHY

Alan F. M. Proceedings of the SPIE, Volume 4841, pp. 1345-1356 (2003).", pp. 1345-1356,2003 Mar.
Palmer and Verrill: "Diffraction gratings", Contemp. Phys., 9, pp. 257-276, 1968.
Palmer, C.: Diffraction Grating Handbook, Richardson Grating Laboratory, 2000.
Pfeiffer, M. J.; Frank, C.; Baumueller, D.; Fuhrmann, K. and Gehren, T.: "FOCES - a fibre optics
Cassegrain Echelle spectrograph", A€9AS, 130, pp. 381-393, 1998 Jun.
Racine, R.: "The Telescope Point Spread Function", PASP, 108, pp. 699-706, 1996 Aug.
Ramm, D.: "Atlas of thorium and argon spectra", , 2003 March.
Ramm, D. J.; Skuljan, J. and Hearnshaw, J. B.: "The orbit of a new double-lined spectroscopic binary:
HD 161958", The Observatory, 124, pp. 167-173,2004 Jun.
Ramsey, L. W.: "Focal ratio degradation in optical fibers of astronomical interest", in S. C. Barden
(ed.), "Fiber Optics in Astronomy, Astronomical Society of the Pacific Conference Series", vol. 3, pp.
26-40, 1988 Aug.
Richardson, E. H.; Moore, A.; Tilleman, T. and Crampton, D.: "Focussing Image Slicers: Refractive
and Reflective (Poster)", in "ASP Conf. Ser. 195: Imaging the Universe in Three Dimensions", pp.
540-+,2000.
Schroeder, D. J.: "Echelle efficiencies - theory and experiment; author's reply to comment", Appl. Opt.,
20, pp. 530-531, 1981 Feb.
Schroeder, D. J.: Astronomical Optics, Academic Press, 2nd ed., 2000.
Schroeder, D. J. and Hilliard, R. L.: "Echelle efficiencies - theory and experiment", Appl. Opt., 19, pp.
2833-2841, 1980 Aug.
Sheinis, A. I.; Bolte, M.; Epps, H. W.; Kibrick, R. I.; Miller, J. S.; Radovan, M. V.; Bigelow, B. C. and
Sutin, B. M.: "ESI, a New Keck Observatory Echellette Spectrograph and Imager", PASP, 114, pp.
851-865,2002 Aug.
Skuljan, J.; Hearnshaw, J. B. and Cottrell, P. L.: "Absolute Radial Velocities by Cross-Correlation with
Synthetic Spectra", in "ASP Conf. Ser. 185: IAU Colloq. 170: Precise Stellar Radial Velocities", pp.
91-101, 1999.
Skuljan, J.; Ramm, D. J. and Hearnshavv, J. B.: "J.A.Lccurate orbital parameters for the bright southern
spectroscopic binary ( Trianguli Australis - an interesting case of a near-circular orbit", MNRAS,
352, pp. 975-983, 2004 Aug.
Stobie, R.; Meiring, J. and Buckley, D. A.: "Design ofthe Southern African Large Telescope (SALT)", in
"Proc. SPIE Vol. 4003, p. 355-362, Optical Design, Materials, Fabrication, and Maintenance, Philippe
Dierickx; Ed." , pp. 355-362, 2000 Jul.
Swart, G.: "SALT Specification and Error Budget Status", A presentation to SSWG, 2001.
Swat, A.; o 'Donoghue, D.; Swiegers, J.; Nel, L. and Buckley, D. A. H.: "The optical design ofthe South-
ern African Large Telescope", in "Large Ground-based Telescopes. Edited by Oschmann, Jacobus M.;
Stepp, Larry M. Proceedings of the SPIE, Volume 4837, pp. 564-575 (2003).", pp. 564-575, 2003 Feb.
Szentgyorgyi, A. H.; Cheimets, P.; Eng, R.; Fabricant, D. G.; Geary, J. C.; Hartmann, L.; Pieri, M. R.
and Roll, J. B.: "Hectochelle: a multiobject echelle spectrograph for the converted MMT", in "Proc.
SPIE Vol. 3355, p. 242-252, Optical Astronomical Instrumentation, Sandro D'Odorico; Ed.", pp.
242-252, 1998 Jul.
Szentgyorgyi, A. H.; Fabricant, D. G.; Brown, W. R. and Epps, H. W.: "Cross dispersion and an integral
field unit for Hectochelle", in "Instrument Design and Performance for Optical/Infrared Ground-
based Telescopes. Edited by lye, Masanori; Moorwood, Alan F. M. Proceedings of the SPIE, Volume
4841, pp. 1026-1035 (2003).", pp. 1026-1035, 2003 Mar.
Tobin, W.: "Gain, noise and related characteristics ofthe Mt John Photometrics CCD system", Southern
Stars, 34, pp. 421-429, 1992.
Tobin, W.; Hearnshaw, J. B.; Kershaw, G. M.; Nankivell, G. R.; Persson, S.; Rumsey, N. J. and Thirkettle,
R.: "A focal reducer for the Mt John echelle spectrograph", Southern Stars, 37, pp. 197-205, 1998.
Tull, R. G.: "High-resolution fiber-coupled spectrograph of the Hobby-Eberly Telescope", in "Proc. SPIE
Vol. 3355, p. 387-398, Optical Astronomical Instrumentation, Sandro D'Odorico; Ed.", pp. 387-398,
1998 Jul.
Vaughnn, D.: "What's wrong with the throughput-resolution product? A fiber-fed spectrograph forces a
reevaluation of instrument design parameters", in "Proc. SPIE Vol. 2198, p. 31-43, Instrumentation
in Astronomy VIII, David L. Crawford; Eric R. Craine; Eds.", vol. 2198, pp. 31-43,1994 Jun.
Vogt, S. S.; Allen, S. L.; Bigelow, B. C.; Bresee, L.; Brown, B.; Cantrall, T.; Conrad, A.; Couture, M.;
Delaney, C.; Epps, H. W.; Hilyard, D.; Hilyard, D. F.; Horn, E.; Jern, N.; Kanto, D.; Keane, M. J.;
BIBLIOGRAPHY 247

Kibrick, R. 1.; Lewis, J. W.; Osborne, J.; Pardeilhan, G. H.; Pfister, T.; Ricketts, T.; Robinson, 1. B.;
Stover, R. J.; Tucker, D.; Ward, J. and Wei, M. Z.: "HIRES: the high-resolution echelle spectrometer
on the Keck 10-m Telescope", in "Proc. SPIE Vol. 2198, p. 362-375, Instrumentation in Astronomy
VIII, David L. Crawford; Eric R. Craine; Eds.", pp. 362-375, 1994 Jun.
Walker, D. D. and Diego, F.: "Design philosophy of the forthcoming echelle spectrographs for the AAT
and LPO", MNRAS, 217, pp. 355-365, 1985 Nov.
Walraven, T. and J.H, W.: in S. Lausten and A. Riez (eds.), "Aux. Instrumentation for Large Telescopes",
p. 175, 1972.
Wood, R. W.: "The echelette grating for the infra-red", Phil. Mag., 20, pp. 770-778,1910.