ELT -HIRES, the high resolution spectrograph for the ELT; end-to-end simulator: design approach and results

(1)

2018 Publication Year

2021-02-26T09:12:42Z Acceptance in OA@INAF

Title

Genoni, Matteo; LANDONI, Marco; RIVA, Marco; PARIANI, Giorgio; MASON, Elena; et al. Authors 10.1117/12.2312105 DOI http://hdl.handle.net/20.500.12386/30628 Handle PROCEEDINGS OF SPIE Series 10705 Number

(2)

PROCEEDINGS OF SPIE

SPIEDigitalLibrary.org/conference-proceedings-of-spie

ELT -HIRES, the high resolution

spectrograph for the ELT; end-to-end

simulator: design approach and

results.

Genoni, M., Landoni, M., Riva, M., Pariani, G., Mason, E.,

et al.

M. Genoni, M. Landoni, M. Riva, G. Pariani, E. Mason, P. Di Marcantonio, O.

A. Gonzalez, P. Huke, H. Korhonen, M. Xompero, Christophe Giordano, I. Di

Varano, Gianluca Li Causi, E. Oliva, Thomas Marquart, A. Marconi, "ELT

-HIRES, the high resolution spectrograph for the ELT; end-to-end simulator:

design approach and results.," Proc. SPIE 10705, Modeling, Systems

Engineering, and Project Management for Astronomy VIII, 1070514 (10 July

2018); doi: 10.1117/12.2312105

Event: SPIE Astronomical Telescopes + Instrumentation, 2018, Austin, Texas,

United States

(3)

ELT -HIRES the High Resolution Spectrograph for the ELT: End to

End simulator. Design approach and results.

M. Genoni*

ab

, M. Landoni

b

, M. Riva

b

, G. Pariani

b

, E. Mason

c

, P. Di Marcantonio

c

, O. A. Gonzalez

f

,

P. Huke

d

, H. Korhonen

g

, M. Xompero

l

, Christophe Giordano

l

, I. Di Varano

e

, Gianluca Li Causi

h

, E.

Oliva

l

, Thomas Marquart

i

and A. Marconi

l,m

a

_{Univeristà degli Studi dell’ Insubria, Dipartimento di Scienza ed Alta Tecnologia, via Valleggio 11,}

I-22100 Como, Italy;

b

_{Istituto Nazionale di Astrofisica – Osservatorio Astronomico di Brera-Merate, via E. Bianchi 46,}

I-23807 Merate (LC), Italy;

c

_{Istituto Nazionale di Astrofisica – Osservatorio Astronomico di Trieste, via Tiepolo 22, I-34131}

Trieste;

d

_{Georg-August Universität Göttingen, Institut für Astrophysik, Friedrich-Hund-Platz 1,}

Göttingen Germany;

e

_{Leibniz-Institut für Astrophysik (AIP), An der Sternwarte 16, 14482 Potsdam,}

Germany;

f

_{UK Astronomy Technology Center (ATC) - Royal Observatory, Edinburgh, Blackford Hill –}

Edinburgh;

g

_{Dark Cosmology Center, Niels Bohr Institute, University of Copenhagen,}

Juliane Maries Vej 30, 2100 København Ø;

h

_{Istituto Nazionale di Astrofisica – Istituto di Astrofisica e Planetologia Spaziali, via Fosso del}

Cavaliere 100, Roma – Italy;

i

_{Department of Physics and Astronomy, Uppsala University, Box 516, SE-751 20 Uppsala, Sweden}

l

_{Istituto Nazionale di Astrofisica – Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, I-50125,}

Firenze, Italy;

m

_{Dip. di Fisica e Astronomia, Univ. di Firenze, Via Giovanni Sansone 1, 50019 Sesto Fiorentino,}

Italy;

ABSTRACT

We present the updated design and architecture of the End-to-End simulator model of the high resolution spectrograph HIRES for the future Extremely Large Telescope (ELT). The model allows to simulate the propagation of photons starting from the scientific object of interest up to the detector, allowing to evaluate the performance impact of the different parameters in the spectrograph design. The model also includes a calibration light module, suitable to evaluate data reduction requirements. In this paper, we will detail the architecture of the simulator and the computational model which are strongly characterized by modularity and flexibility that will be crucial in the next generation instrumentation for projects such as the ELT due to of the high complexity and long-time design and development. We also highlight the Cloud Computing Architecture adopted for this software based on Amazon Web Services (AWS). We also present synthetic images obtained with the current version of the End-to-End simulator based on the requirements for ELT-HIRES (especially high radial velocity accuracy) that are then ingested in the Data reduction Software (DRS) of CRIRES+ as case study.

Keywords: End-to-End simulators – Instrument modeling – Parallel computing – Cloud Computing Architecture – HIgh REsolution Spectrograph (ELT-HIRES) – ELT telescope.

*matteo.genoni@brera.inaf.it; or matteo.genoni@gmail.com

Modeling, Systems Engineering, and Project Management for Astronomy VIII edited by George Z. Angeli, Philippe Dierickx, Proc. of SPIE Vol. 10705

Proc. of SPIE Vol. 10705 1070514-1

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 11 Jan 2021 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

(4)

1. INTRODUCTION

Front line researches in di erent fields of astrophysics are entering a new era, where the signal detection capabilities will be pushed to unprecedented sensitivity. In the context of exoplanetary science, the atmospheric characterization through high resolution spectroscopy, with the ultimate goal of detecting the few ppm amplitude signal of an Exo-Earth orbiting Solar-like star, and furthermore the possible detection of life bio-signature on exoplanets are only two of the most fascinating examples.

However, the measurement of these properties is challenging both for the physics behind them and for the precise design, optimization, characterization and complexity of the required instrumentation. This unprecedented sensitivity, in fact, will be achieved thanks to more and more complex observation facilities in term of technologies, instrumentations, operative modes and procedures. This arises both from the infrastructure required by large aperture telescope and even more heterogeneous and complex instrumentation. In order to assess the impact of design architecture on the instruments behavior and to accurately predict instruments performances, end-to-end simulators have become key tools2,3.

End-to-End instrument models (E2E) are simulators which allow physical modeling of the whole system: from the light source to the raw-frame data. Synthetic raw-frames are the output, ingested by the Data Reduction Software (DRS), to be analyzed in order to assess if the scientific requirements (in terms of spectral resolution, SNR, Radial Velocity accuracy and precision, etc.), related to the specific science drivers, are satisfied with the specific instrument design and architecture. Moreover, E2E are valuable tools to be exploited to optimize performances, calibration procedures and observation plans, with the ultimate goal of maximizing the scientific return, before the on sky operations.

In this paper we present the current design and development status of the E2E simulator for ELT-HIRES, the high resolution spectrograph for the ELT. The development of the simulator directly during the design phase will benefit the whole ELT-HIRES project, in addition to the aspects mentioned before, giving reliable simulations to gain a deep understanding of the instrument design, improving the capability of early identification of system level problems. This is fundamental to properly characterize error budget and systematic e ects at system engineering level.

2. ELT-HIRES OVERVIEW

The huge photon collecting power of the 39 m primary mirror diameter ELT coupled with a High Resolution Spectrograph (ELT-HIRES) will allow to make fundamental discoveries in a wide range of astrophysical areas, outlined by the Science Team of the ELT HIRES consortium4,5,6,8_:

• The study of Exo-planetary atmospheres and the detection of signatures of life on rocky exo-planets. • The chemical composition, atmospheres, structures and oscillations of stars.

• The spectroscopic study of the galaxies evolution as well as the three dimensional IGM reconstruction at high redshift.

• Fundamental constants (such as the fine-structure constant α and the proton-to-electron mass ratio µ) variation and the related cosmology.

The E-ELT will be the largest telescope to observe in visible and infra-red light; the baseline of the optical design (see Figure 1) is five mirror solution1_{: aspherical (almost paraboloid) primary mirror M1, a convex secondary mirror M2 with}

4 m diameter, concave tertiary mirror M3 with 3.75 m diameter, and two flat mirrors (called M4 and M5). These two latter mirrors have the purpose to feed two Nasmyth focal stations and for adaptive optics; below each Nasmyth platform a Gravity Invariant focal station, fed by a steerable and removable mirror (M6), will be located (see Figure 1). In addition the M6 mirror and a Coudè-train relay optics will allow to feed a Coudè focal station, which will be specialized to host instruments requiring very high long term stability in terms of thermal and mechanical perturbations. The telescope structure will be alt-azimuth type.

Proc. of SPIE Vol. 10705 1070514-2

(5)

Quaternary miner (M4) F4ih mirror mirror (MS) Tertiary mora (M3)

Nasmyth local surface Primary mera (Me) Gravrty++vananl focal mataco

BBfibers

(L <fO rn) Eibe

HR and UHR

front -end _Calibration

Unit r to fiber interface Cr.Disp. Cr.Disp. CCr.Disp\ Cr.Disp Camera Warm spectrograph

, \

0.40 -0.46 0.46 -0.55 0.55 -0.67 0.67 -0.83 Ó _ SCAO-IFU front-end

J

w

Other front -ends

--

á J Cold IR ó cLi p Cr.Disp.H Came

L

0.83 -0.95 0.95 -1.13 1.13 -1.35 1.44 -1.80 l 1 E-ocd cLi

j

w

- Cr.Disp. - Camer

_i

Camera Y

I

fiber to fiber

interface _{- Cr.Disp. - Camera}

i

_J

/

Telecom fibers (L <200 m) ` J spectrograph

/

1 Cr.Disp. - Camera

i

H )

Figure 1. E-ELT optical layout: Nasmyth focus (left) and Gravity invariant focus (right), taken from [1]

HIRES is a modular fiber-fed cross-dispersed echelle spectrograph that will operate both at a Nasmyth and at Coudé focus of the ELT. The baseline architecture, defined in the Phase A of the project, consists of two separate spectrographs operating in the visible (BVRI) and infrared (ZYJH) bands, (see Figure 2, and for details reference21_{). A more complete}

solution consisting of four separate spectrographs with the addition of the U and K bands respectively is also taken into account in the instrument design (and handled by the software, described in reference7_{) even though this solution due to}

the cost cap is currently termed as “add-on”.

Figure 2. Schematic of the proposed ELT-HIRES functional architecture concept. The di erent sub-systems of the instrument (front-end, fiber link, calibration unit, spectrometers) are indicated, as well as the possible length of the fiber bundles, according to the specific wave-band. The wavelength splitting in the di erent spectrograph modules is reported (in µm).

Proc. of SPIE Vol. 10705 1070514-3

(6)

The light from the telescope is split, via dichroics in the Front-End (FE), in the two wavelength channels. The FE sub-system also provides: atmospheric dispersion correction, field stabilization and guiding. Each wavelength channel interfaces with several fiber bundles that feed the corresponding spectrograph module (VIS and NIR). Each fiber-bundle corresponds to an observing mode. All spectrometer modules have a fixed configuration, i.e. no moving parts. They include a series of parallel entrance slits consisting of linear micro-lenses arrays each glued to the fiber bundles. In order to accomplish the specific scientific goals (see refrence8_{, as well as for a description of the observing modes), for the}

di erent main observing modes two possible resolution capabilities (HR with R = 100000 and UHR with R = 150000, see Figure 2) are foreseen, which can be implemented by feeding the spectrograph modules with di erent fibers bundles. The technique used by the Front-End and Fiber-Link sub-systems to feed the huge AΩ-product at the spectrograph entrance in the selected architecture is the field dicing (see for details reference9_{), in which each fiber of the bundle is}

looking at a slightly di erent part of the object.

The spectrometers can be ideally divided according to their specific function into two units: the pre-slit unit, a re-imaging system which collects the light from the fiber optics and feeds the spectrometer unit, which has the usual purpose of separate the light into its constitutive wavelengths and then refocus them onto the detector surface. The spectrometer units’ optical configuration (see for details21_{) is the white pupil layout with an o -plane echelle grating; while the current}

camera optical configuration is the Schmidt camera.

The Calibration Unit, located in the Coudé room, is connected via fibres to the Front End and to the Fibre Links. The foreseen calibration sources are intensity calibration sources (laser driven light source, halogen lamps, light emitting diodes, light bulbs) and spectral sources for simultaneous calibration (AstroComb, Faby-Perot, single wavelength laser, hollow-cathode lamps).

ELT-HIRES has a Polarimeter arm, which is composed by two modules: Intermediate Focus Polarimetric Module (IFPM) and Front End Polarimetric Module (FEPM). The IFPM is located in the Adaptive Optics Tower of the telescope and is in charge of splitting the incoming beam into the ordinary and extraordinary beams via a birefringent prism (double Wollaston). This module uses also the calibration light via a fiber bundle coming from the Calibration Unit. The FEPM splits, via dichroics, the ordinary and extraordinary beams in the two wavelength channels (VIS and NIR) and provides usual Front-End functionalities listed above.

3. END-TO-END SIMULATOR DESIGN PHILOSOPHY AND ARCHITECTURE

The design philosophy of the E2E simulator is characterized by three fundamental aspects:

• Modularity: the system is organized in Modules and Units, each in charge of its own responsibility, with specified interfaces and possibly with the maximum degree of independence. Since the instrument is itself complex (telescope, fiber-link, spectrograph, detectors) one of the key step is to identify and define the different Modules (and Units) and their interfaces. This allows to well characterize the behavior and the functionalities of each of them;

• Flexibility: Since Units and Modules are, at first level of approximation, independent the system should allow to choose what to simulate or not, or to by-pass some units (e.g. simulate only emission from the object without atmospheric contamination or, instead, simulate the echellogram without taking into account PSF diffraction); • Speed: It is straightforward to expect that a certain amount of simulations will be performed during the whole

design phase of the instrument. For this reason, it is crucial to design a solution that could achieve a fast computational time by adopting the proper state-of-the-art technologies. For this reason, the system has been design by the adoption of Cloud-Computing that allow to speed up by various orders of magnitude the computational time.

The End-to-End simulator architecture, shown in the schematic of Figure 3, is highly modular, composed by different modules each one with specific tasks, units-functionalities and interfaces. In details, a Unit is a piece of software, function or interface of a part of the HIRES spectrograph and could be considered as a black-box that exchanges data

Proc. of SPIE Vol. 10705 1070514-4

(7)

both with oth of software th Energy distrib the Image Sim expected outp Figure 3. data-&-lo orange, w units whic

4. END-T

In this section 4.1 Science M In the science source. Curre specified e e source, when flux and the s accordingly. wavelength, i

her units of th hat model eith bution of the o mulator modu puts in the exp

General schem gic path throug while in blue oth ch compose the

TO-END SI

n, the descript Module e module, a sy ently the mod

ective temper the magnitud spectrum is no This module in Angstroms he same modu her an externa object to simu ule). Each mod pected format

matic diagram o gh the di erent her units simula e Instrument Mo

IMULATO

tion of each bl ynthetic spectr ule is set for rature and su de is provided ormalized at th e provides an

, and the seco

ule, and even w al component ulate, or a piec dule or unit is

(like tfits tabl

of the End-to-E t units are show ated by the End odule are group

OR BREAK

lock reported rum is created stellar science urface gravity) in a given ph he reference w n output file, ond one is the

with other mo (w.r.t. the HI ce of software identified and le, image, etc.

End simulator m wn. Instrument r d-to-End model ped by the grey

K-DOWN

in the schema d from a set of e cases, for w ), magnitude, otometric ban wavelength of a FITS tabl e flux, in CGS

odules (as rep IRES spectrog e that has not a d characterize ). modules, with th related units of l but independe box. atic view of Fi f input parame which the inpu

radial veloc nd, the observe f the provided le, composed S units, or pho resented in Fi graph) such as a counterpart o ed by: main ta he related inter f the End-to-End

ent from ELT-H

igure 3 is repo eters related t ut parameters ity and expos ed energy den d magnitude pa d by two colu oton flux. The

igure 3). Mod s Telescope, o on the actual h asks, inputs req

rfaces. The ligh d simulator are HIRES Spectrog

orted. to a specific sc

are: the spect sure time. Fo nsity is conver ass-band and umns: the fir e resolution ca

dules are piece or the Spectra

hardware (e.g quired and the

ht path and colored in graph. The

cience case, or tral type (with or each stellar rted to photon then re-scaled rst one is the an be set as an e al g. e r h r -d e n

Proc. of SPIE Vol. 10705 1070514-5

(8)

output-simula 1D spectrum The module w same strategy for QSO and

Figure 4. e ective

4.2 Atmosph The main aim introduce it t page11_{), whi} 0.006 Angstr ELT-HIRES. synthetic spec SkyCalc give fraction) at P general extinc is obtained b component, o The sky radia the units are c

ation paramet is reported in will also be s y, it can be ext Ly-α forest fo Synthetic emi temperature 50 heric Module m of this modu to the simulat ich is based o roms, correspo

The sky spe ctrum. es two differen Paranal. For t ction curve, an by a spline fit on the other ha ance is calcula converted in p

ter, and curren n Figure 4.

set for both to tended for the or the Sandage ssion spectrum 000 K, surface g e ule is to mode ted science ob on the Cerro P onding to at ectrum obtain nt outputs: sk the End-to-en nd telluric abs t to the data, and, is given b ated at the area photons/(s m2 ntly it is such o stellar popu e whole scienc e Tests simula m from a G2V gravity log(g)=4 el the scatterin bservations. T Paranal Adva least spectral ned from SkyC ky radiance (in nd simulation sorption lines from which by removing t a of one indiv µm).

h that the reso ulations and ex ce cases of HI ations). star obtained b 4.5 and solar m ng, absorption This is done b anced Sky Mo resolving po Calc is re-bin n photons/(s m n purposes the (see Figure 5 the saturated the extinction vidual HIRES lving power i xtragalactic sc IRES wheneve by the science etallicity. and emission by using the E odel. The Sky ower of about nned to the sa m2_{μm arcsec}2 e sky transm 5). Extinction d absorption l from the total fiber (0.17 arc is R=550000. cience (QSO, er the theoreti module. Typic n occurring in ESO SkyCalc yCalc is queri t 550000 with ame wavelen

2_{)) and sky tra}

mission is divi curve, shown lines are exclu

l sky transmis csec diameter

An example etc.). In fact ical spectrum

cal main sequen

the Earth's atm c tool (availab

ied using reso h the wavelen gth scale as i ansmission (in ided into two

in the examp uded. The ab ssion spectrum r of sky projec of a synthetic t, adopting the is known (e.g nce star of mosphere, and ble at the web olution step o ngths range o is used in the n transmission o components le of Figure 6 bsorption lines m.

ction) and then c e g. d b f f e n s, 6, s n

Proc. of SPIE Vol. 10705 1070514-6

(9)

Figure 5. Figure 6. 4.3 Telescop The aim of th system corre PASSATA so adaptive opti Example of tell Atmospheric ex pe Module his module is ctions, the im oftware11_{, dev} ics simulation luric spectrum r xtinction curve s to predict th mage at the te veloped by th ns and manag recovered adop adopted. he telescope p elescope foca e Adaptive O ged by the Te

pting the proced

point spread fu al plane and Optics group o elescope Simu dure reported in unction (PSF) the telescope of the Arcetri ulation and A

n the text (y axis

) with and wi throughput. observatory f AO group in th s is in relative u ithout adaptiv This module for Monte-Car he framework units). ve optics (AO relies on the rlo end-to-end k of the ELT ) e d

-Proc. of SPIE Vol. 10705 1070514-7

(10)

HIRES proje simulations o The input pa atmospheric reflectivity pr parameters (l frequency) ar for what conc concern trans An example o (Single Conj performance source magni Figure 7. distance = profiles, a 4.4 Instrum This module the instrumen • • •

ect. The intera of this indepen arameters are layers), telesc rofile. The W like actuators re internal par cern the telesc smission. The

of the PSF pr ugate Adapti is simulated i itude, for zeni

Mean profiles = 0 at λ = 100 at a certain dista ment Module is in charge to nt PSF and ins The Front En The Calibrati The Spectrog effects of d compute the plane. The tra blaze functio calculation. action with th ndent software : zenith angle cope wave-fro WFEM and tel

number and f rameters used cope PSF (bo output telesco rofile (non-nor ive Optic) op in di erent c th distance = of the non-no 00 nm. The co ance from centr

o generate dat strument throu nd (FE) Unit a

ion Unit produ graph Unit mo dispersive ele whole point s ansmissivity a on and cross his work-pack e and the E2E e, seeing, the ont error matr lescope transm frequency, nu d by the PASS th with and w ope PSF is giv rmalized) sim perative mode onditions of s 0 at λ =1000 n ormalized PSFs omparison betw ral core, drop to

ta for renderin ughput. The m and Fiber Link uces the SED odels the opti ements and th spread functio and reflectivit disperser di

kage has been simulator. e estimated a rices (WFEM mission profil umber of wave SATA softwar without the AO

ven with a sam mulated by the e for ELT-HI seeing (best s nm. under di eren ween the AO co o the seeing lim

ng the geomet module is orga k (FL) Unit mo

of the selected ical path alon he optical qua on for differen ty of the diffe iffraction effi

n set for the verage wind M), the AO sy les are gather e-front modes re. The outpu O correction), mpling of 0.00 Telescope M IRES is show eeing = 0.47 nt seeing condi orrection and s mited case becau

trical images o anized in units

odels the link d calibration s ng the optical ality in differe nt wavelength erent optical c iciency are m definition of speed (from ystem paramet ed from ESO s to be contro ut interfaces ar , and Image S 08 arcsec/pixe Module, both in wn in Figure arcsec; media

itions and sour seeing limited use not corrected

of the instrum : between the t sources. train taking ent condition hs and position omponents as modeled for the interfaces wind profile ters and teles O, while the s

lled and syste re with Instru Simulation Mo el. n seeing limit e 7. The SCA an seeing = 0 rce magnitude, case shows tha d.

ment focal plan telescope and into account s are evaluate ns on the inst s well as the e the instrumen s between the s of di eren scope coatings et AO system em acquisition ument Module

odule for wha ted and SCAO AO correction .8 arcsec) and for zenith at the PSF ne, to evaluate instrument. distortion, the ed in order to trument image echelle grating nt throughpu e nt s m n e, at O n d e e o e g ut

Proc. of SPIE Vol. 10705 1070514-8

(11)

• 4.4.1 Front-The task of t required inpu and optical p FE throughpu selected) with bundle (part o In order to pr with the estim resolution of Figure 8. 0.025 arcs The dicing-un FE focal plan taken by each so that it is po the modeling and related fib

The Polarime flux at the po -End Unit

this unit is to uts are: telesco

arameters like ut). The FE fo h a micro-lens of the Fiber-L redict the light

mated PSF of 0.008”/pixel (

Front-End foc sec. Right pane

nit micro-lens ne according t h micro-lens i ossible to estim of the di ere bers, and FE t eter Unit is ai olarimetry arm o predict the ope focal plan e the proposed ocal plane is th s array and sep Link unit).

t distribution f the Front-E (see Telescop

cal plane. Left p

el: Portion of th

s array has be to a hexagona s computed by mate the amou ent glasses an throughput as imed at evalu m output. light distribut ne image (inte d glasses and he plane wher

parates the dif at the FE foca nd (2D imag e Module sect

panel: the teles

e FE unit focal een modeled b l packing. An y evaluating h unt of light pr nd coatings tra function of w

uating the tele

tion at the FE ernal interface coatings in o re the dicing-u fferent parts o al plane (see F e FITS). The tion). The FE

scope focal pla plane light dist

by simulating n example with how much of l rocessed by ea ansmissivity a wavelength are scope polariz E focal plane e in the E2E rder to estima unit samples th of the field fee

Figure 8) the t e telescope im estimated PS ane image is co tribution that fe the different h 30 fibers is light with resp ach of them. F and reflectivity e transmitted t zation effects e and to predi simulator), th ate the FE poi he field (in ca eding the corre telescope foca mage in input F size is 0.025

onvolved with eeds the dicing u

micro-lenses reported in Fi pect to the tot Finally the FE

y. The output to the Image S

and the final

ict the FE thr he Front-End

int spread fun ase of field-dic esponding fibe al plane image is sampled w 5 arcsec. the estimated F unit aperture. as sub-circles igure 9. The a tal is captured throughput is tables with am Simulator Mod

odd and even

roughput. The optical design nction (and the cing technique ers in the fiber e is convolved with a defined FE PSF of s placed in the amount of flux by each fiber s calculated by mount of ligh dule. n e n e e r d d e x r, y ht

Proc. of SPIE Vol. 10705 1070514-9

(12)

rnalized flux dis pixel id]sa}ewpi000 eya uoi;nqua! 150 ix] FE focal plane 1 o mo Figure 9. 4.4.2 Fiber-L This unit has e ects that c current versio to be a top-ha simulation (w and the opera should be tak disk. The mo (see reference

The free shap amount of di model with th physical e e mentioned be

Dicing unit arra

Link Unit s the purpose could occurs w on of the simu at function. N which will be ative conditio ken into accou ost straightfor e15_{for details)} pe parameter M raction e e he confidence ects observed efore. ay example: 30 to predict the while light pa ulator, a perfe evertheless, th part of the ne ons in term of

unt so this pro rward equation

):

M was tuned t ect coming fro e level of 0.05 d in optical fi 0 micro-lenses m e light distribu asses through ect scrambling he whole set o ext E2E simu f temperature ofile could be n that is able ,

to M=30, as sh om the finite a 5. In the next fibers, like th modeled as circ ution at the en h the fiber (lik g system is as of physical pa ulator versions and mechanic safely approx to capture th

with

hown in Figur aperture of th versions of th he modal nois

les and packed

nd of the fiber ke modal nois sumed so that arameters fore s) are: fibers-b cal stresses. A ximated by a t hese e ects is

,

re 10, in order he system. Wi he simulator, se, taking int

in hexagonal la rs by taking in se and non-un t the output en seen to be con bundles layou According to top-hat functio s the so-called / r to recover th th this set, thi this module s to account th ayout. nto account a niform illumin nd of each fib nsidered for a ut, fibers type physical optic on plus the Ai d Super-Lore / he proper dime is function fit should be able he set of inp

all the possible nation). In the ber is expected more detailed and geometry cs, di raction iry di raction ntian function

ension and the ts the physica e to reproduce ut parameters e e d d y n n n e al e s

Proc. of SPIE Vol. 10705 1070514-10

(13)

0.035 0.03 0.025 0.02 0.015 0.01 0.005 0 12 11 10 -9 8 7- 3-2r 935.6

Ideal light distribution prol

linear coordinate [om]

OOylJUl

935.8 150 200

JUvvU l

936 ! os OAS 0.9 0.35 03 025 02 0.15 0.1 0.05 '1'30 -10 936.2 936, 1

Airy Diffraction profile

to u

linear coordinate x [um)

lu

V y 1J .4 936.6 JVavelength (Angst' 20 30 0.035 Disbibution4 0.03 °0.025 Fri ó, 0.02 0.015 .6 0.07 0.005 00 937

comparison: convolution and

I

convoi Super.

J

50 100 linear coordinate x [um]

Accepte( a signific C-937.2 Super -Lorentz n Mon ,Lorentzian M =30 150 200 3 with :ance a =0.05 -FP spectrum M U 937.4

u

T Figure 10 the outpu Lorentian 4.4.3 Calibr The calibratio the same way lamps are to b that contains Figure 11, wh Figure 11 4.4.4 Polarim The Polarime extraordinary simulator no

. Left panel: top ut fiber. Right p n function. ration Unit on unit is in c y of the Scienc be simulated ( the spectral e here it is show . Example of SE meter Unit eter Unit is y flux at the po modeling is im p hat function t panel: compar charge to simu ce module. Bo (flat field, ThA energy distribu wn a portion of ED produced b aimed at eva olarimetry arm mplemented f

that model the o rison between

ulate the Spec oth arc lamp a Ar, Fabry Pero ution of the re

f a Fabry-Pero

by the calibratio

aluating the t m output, as w for this Unit.

output of a perf the top-hat fun

ctral Energy D and flat field a

ot, etc.), the in equired source ot etalon used

on unit (Fabry-P

telescope pola well as PSFs an

fect optical fibe nction convolv Distribution (S are simulated. nstrument max e in the prope d as a wavelen Perot etalon). arization effe nd overall effi

er. Center panel ved with the A

SED) of the c This unit requ ximum resolu r units. An ex gth calibration ects and to g ficiency. In the l: di raction pa

Airy disk and t

calibration sou uires to know ution and retur xample of this

n source.

ive the final e current versi attern from the Super-urces acting in which kind o rns a tfits table s is reported in ordinary and ion of the E2E n f e n d E

Proc. of SPIE Vol. 10705 1070514-11

(14)

4.4.5 Spectr The purpose the final aim distribution o spectrum due already been instrument in The spectrog Version. The equations and commercial o a complete ru detailed optic consideration software. In a spectrograph While the R software pac instrument fo needs some i designer take of the End-to resolving pow Detector pixe elements wor and pressure) The echellogr fiber/lens-let synthetic di echellogram f Figure 12 rograph Unit of this Unit is m of predict t of the camera e to pupil obs n taken into n term of ambi raph can be m e former is b d relations w optical design un of the sim cal ray-tracing ns regarding th addition this a design choice ay-Tracing V kage Zemax, ocal plane, the

input paramet es into account -End simulato wer, wavelen el size) and I rking angle an ). ram in output on the detect racted spectr format retriev . Echellogram f s to simulate t the echellogr entrance pup struction and account at th ient, mechanic modeled using ased on a ph which characte software (e.g mulator, even g files and do he orders curv approach allo es and architec Version takes and uses so e throughput ters to model t in the design or architecture ngth coverage Input variable nd line density t from this Un tor surface; th ra (which will ved by the Inst

format example

the physical e am (spectral pil (this is exp non-uniform his current si cal and therma g two di eren hysical param

erize the opti g. Zemax). The

if the optical ocumentation vature, lines til ows for quick

ctures. as input dire ome specific s and the light the spectrog n and optimiza e the inputs ha e and samplin e parameters ( y, glasses and nit is compose hese are passe l be comment trument Modu

e from the Instru

e ects of the format) at in ploited to sim illumination) imulator vers al e ects will nt alternative metric model ( ical elements, e Parametric V l design of th

are not alread lt, fibers align parametric ev ectly the

ray-software tool distribution o raph; these p ation of the sp ave been divid ng), inputs p (e.g. number coatings prop ed by the coo ed as input to ted in details ule is shown in ument Module di erent opti nstrument foc mulate/estimate ). Aberrations ion, while th l be introduced approaches: t (see for exam , while the la Version is a u he instrument dy complete a nment which a valuations and -tracing optica s to extract t of the camera arameters are pectrograph w ded in the foll parameters fo of optical fib perties and op rdinates of th o the Image S in the Image n Figure 12. Paraxial Param cal componen cal plane, the e the second , distortion an he physical o d in the future the Parametric mple refrence1 atter is a ray seful tool whi is only at a and available, are relevant as d analyses co al design file the required o entrance pup e the same pa with the ray-tra

lowing types: rm other mo bers, optical e erative condit he projected re Simulation Mo Simulator M metric Model co nts of the spec e throughput order e ects nd di raction operative cond e versions. c Version and 14_{), built with} y-tracing versi ich can be exp preliminary p , with the pur spects for the oncerning di es, built in th

outputs (spec pil), the Param arameters whi

acing software top level requ odules (e.g. te elements F-rat tions in term o esolution elem odule for the Module). An e

omputation.

ctrometer with and the ligh on the objec n e ects have

ditions of the d Ray-Tracing h the physica

ion built with ploited to have phase and the rpose of doing data reduction erent possible he commercia ctral format a metric Version

ich the optica e. In the frame uirements (e.g

elescope size tio, dispersing of temperature ments per each generation o xample of the h ht t e e g al h e e g n e al at n al e g. e, g e h f e

Proc. of SPIE Vol. 10705 1070514-12

(15)

4.5 Detector Module

The task of this module is to simulate, on the rendered data, the e ects produced by the detector. Besides the nominal Quantum E ciency response and the usual noises (like photon noise, Read-Out-Noise, Dark Current, pixel non uniformity, etc.) that are well characterized and, in most cases, straightforward to simulate, this module will also have the aim to add other e ects that could be relevant for high precision radial velocity measurement. In particular, additional blurring types of the Spectral Resolution Elements (SRE) caused by the detector itself, are planned to be taken into account in future versions of the E2E simulator; these are: defocusing due to the not ideal planarity of the detector surface, optical defocusing of the diverging beam after focal point and Gaussian blurring due to the dispersion of electrons across the sensitive material.

4.6 Image Simulator Module

This portion of the simulator is the kernel of the whole system. This piece of software, which runs heavily in parallel in a cloud distributed environment (see next section), is responsible for rendering the photons distribution of each resolution element for each fiber, for each order for each wavelength as should be detected at the level of the detector. A portion of this module, written in MATLAB, is also in charge to glue the different Units and Modules to produce the actual echellogram.

The perfect image of the fibers as seen at the level of the focal plane is estimated as explained in the Fiber-Link Unit subsection. Then, this distribution must be convolved with the PSF of the instrument that includes both aberration, diffraction and effects from the optics of the spectrograph. This is referred as 1st order PSF effect, and it could be carried out in two ways:

• Analytically, using only the spectrograph paraxial model. In this scenario, the net effect is a Gaussian function whose FWHM represents the quality of the system.

• Using ray tracing, by polling Zemax to recover the map of the PSF that could be used as a kernel to be convolved with the light distribution coming from the fiber.

For the current version of the E2E simulator the first mode (analytically) has been adopted, although the whole architecture of the system is able to handle also the second one.

The actual PSFs are also affected by the diffraction coming from the obstruction present in Schmidt cameras proposed in the current spectrograph optical design. Diffraction spikes in the PSF are generated by the presence of sharp edges within the beam shape at pupil position. Each spike extends towards orthogonal direction with respect to the edge that produced it and its spatial distribution is broader than the Airy diffraction halo, as deeply described in reference16_{. The importance}

of simulating diffraction spikes for a spectrograph like ELT-HIRES is that, if a spike happens to extend along the spectral direction in the final image, it will spread the flux of any bright spectral line onto the adjacent spectral intervals, thus increasing the local stray light contamination. This effect is especially important around the narrow and intense sky emission lines (see again reference16_{for details). In this context, the Image Simulator Module (2}nd_{order PSF effect) first}

computes an high resolution model of the beam obstruction, which depends on the beam shape, angle and position when the beam intersects the obstructing elements (different for each field and wavelength, and passed as inputs to this module, coming from the Spectrograph Unit of the Instrument Module). Then, it computes the diffraction component of the PSF shape by means of a Fast Fourier Transform of the obstruction model. The output of this module is a 2D image of the diffraction component computed on a sub-pixel sampling, combined with the 1st order optical aberration PSF (explained before). Finally, the Super-Lorentian illumination profile coming from fiber output end is convolved with the whole estimated PSF model, and the surface integral of the convolution is computed in each pixels to obtain the photons distribution for each SRE.

5. COMPUTATIONAL ARCHITECTURE

For a high resolution and high radial velocity precision spectrograph like ELT-HIRES, the wavelength image barycenter reconstruction accuracy will be the primary performance; to give an order of magnitude the spectral resolution element barycenter reconstruction accuracy should be better than ∼ 1nm. This requires, as expected, that the computation of the photons distribution for each SRE should be performed with high precision.

For this reason, we have developed our own integral computation procedure using an innovative approach based on heavy parallel computing CUDA by NVIDIA for the evaluation of single point convolution value.

Proc. of SPIE Vol. 10705 1070514-13

(16)

v

X (D -< Block E I Threád

3lock Block Bli

Pixel)I

Block

Thread Thee

ock Block Blo

Blo !ad ó I. I'1 I I I I Pixel fraction Y k hread thread th Pixe Block Thread Thread cread thread th I Fraction X cread thread thread The procedur exploited to d integral calcu (assuming on The adoption a huge amou computed is simulator. Th Lorentzian fu the PSF of th The computa the schematic 100 sub-pixe parallel 10.00 each SRE. Si capability of computer ava additional ke independent o throughput pr applied locall Figure 13 integral on

6. RESUL

6.1 Perform As stated in t especially in with the adop approximatel allow to conf be used in the of 2nd_order

illumination s (far all the sc less than the o

re is based on directly contr ulation and the nly 1st PSF eff n of parallel co unt of compu of the order he image of sin unction with p e spectrograph ation of each S c of Figure 13 els, a worker t 00 worker thr ince the probl cloud comput ailable to per y point of the of the spectra rofile of the w ly, once theil

3. Schematic of n each detector

LTS

mances the previous s terms of comp ption of the C y 0.5 days. T firm that the c e case of incre PSF effects) solution of th cientific cases ones locally re n standard inte ol the compu e pitch scanni fect and analy omputing is cr utational pow

of 7 x 109_{. It}

ngle resolutio parameter M =

h itself (as des SRE image is . In particular thread compu read on a sing lem is, as exp ting using the rform the wh e computation al energy distr whole instrume lumination so f CUDA Block r pixel. section, one o plexity in tim Cloud Archit This kind of c current archite eased comput ) which shou e echellogram that can be co equired to app egration algor tation accurac ing of the Fibe ytical profile in rucial, since th er. In particu t is thus clea on element is o = 30 (see Fibe scribed before performed on r, for each pix ute the convol gle GPU kerne plained, parall

Elastic Comp hole simulatio nal architectur

ribution of the ent (from tele olution of the e k and Worker of the most im e. This is the ecture (see re computational ecture is suita tational time r uld be roughl m format is ind onsidered), th ply the SED.

rithm and, am cy and cost. T er illumination n the current v he simulations ular the numb r that the com obtained start er-Link Unit d e). n a grid of 10x xel a Computa lution value i el allowing to lelizable and S puting 2 of Am on (see for d re adopted is t e scientific sc escope to detec echellogram f Thread archite mportant issue most critical p eference18_{), th} l time, which able for intens required to add ly of the sam dependent on e number of r ong others, fo These parame n profile and version of the s of synthetic ber of SRE ( mplexity in ti ing from the i description fo x10 pixels sam ation CUDA B

in that point. o decrease the SREs are inde mazon Web S details about

that the instru cenario that ca ctor QE). Thu format is comp

ecture for the c

that should b point that cou he E2E is abl consider mo sive simulatio d other effect me order. Mo the spectral e required iterat oresees two di ters are the si PSF for the p simulator). spectra gather (including sam ime is a key illumination p r details), and mpled in step Block is assign This kind of e time of comp ependent of e ervices (AWS cloud compu umental echel an be simulate us the SED an puted. computation of be investigated uld results in a le to recover ost of the effe n scenarios w (such as mod oreover, as e energy distribu tions on the cl ifferent variab ingle pixel pa oint convolut red by ELT-H mpling anti-a factor in the profile, model d a Gaussian t of 0.1 pixels ned and, for e approach, all mputation up to

ach other, we S) to increase uting, the ref logram forma ed, as well as nd throughput f convolution a d is the comp a showstopper the echellogr ect needed to with large mar

dal noise or th xplained befo ution of whic loud are order

ble parameters artition for the ion evaluation HIRES involve aliasing) to be design of our ed as a Super that represents as depicted in ach one of the lows to run in o 0.001 µs for e exploited the the number o ference18_{). An} at computed is s of the globa profile can be and surface utational cost r. In particular ram format in be simulated rgin that could he contribution ore, since the ch is simulated r of magnitude s e n e e r -s n e n r e f n s al e t, r, n d, d n e d e

Proc. of SPIE Vol. 10705 1070514-14

(17)

6.2 Raw fra One final acti with Data Re strong link b characterize scientific term To demonstra by a DRS pi CRIRES+ Ins fiber-fed and cross disperse Two di eren simulator. Th a scientific fr produces di illuminated b of CRIRES+. Figure 14 tracing fro

The order tra order was cho how the simu

mes renderin ivity required eduction Softw etween the E2 the performa ms, by simulat ate that this sy ipeline, the E

strument and CRIRES+ a t ed, tilted and i nt raw frames he first on is a frame of a G2 erent illumin by the same ob . 4. ELT-HIRES om CRIRES2 p

acing, i.e. find osen for extra ulations were m ng and Data R d in the design ware (DRS), s 2E and DRS ances of the ting all the ma ynergy is settl LT-HIRES D its DRS can b true slit spectr

inclined (pseu (with excepti flat field fram 2V star with nated slits) in bject) with sim

Z Band flat fie pipeline. Bottom

ding the spec action. Perfect made. An exa

Reduction So n of the E2E s since the deliv is crucial esp instrument by ain ELT-HIRE led with prope DRS working-be found here rograph, they udo-)slits, that on of the BIA me in the Z Ban m V = 10 an n the HIRES Z multaneous Fa

eld frame produ

m: recovered fla

ctral orders an tly vertical ali

mple of this, i oftware simulator is ai vered frames o pecially in the y translating ES science cas er interface an -group applied 19_{), which is c} share enough t the recipes co AS trivial ones nd as seen by nd exposure t Z Band (see abry-Perot ref

uced with the E at field 1D whic nd fitting them ignment of the is reported on imed at demo of the simulat e next E2E sim

directly the ses.

nd to see if the d the beta rec currently unde similarities in ould be applie s, not reported HIRES (left p time = 50 s o Figure 15) in ference. These E2E simulator. h shows the cha

m with polyno e fiber-slit to p n the right and

onstrating the tor will act as mulator versio optical and e simulated ec cipes of the C er developmen n the echellog ed without rel d here) have b panel of Figur observed and n polarimetric e frames have Top-Left: RAW aracteristic blaz omials, was s pixel columns bottom panel capability to input for this ons since this mechanical a chellograms c CRIRES+ DR nt. In spite of grams being hi evant modific been produced re 14) and the seeing = 0.8 c mode (both e been ingeste W data. Top-R ze function shap successful. A s was assumed l of Figure 14.

close the loop s pipeline. The s will allow to architecture in can be handled RS (Details on f HIRES being igh-resolution cations. d with the E2E e second one is 85 arcsec (tha fiber-bundles ed by the DRS ight: order pe. well-detected d, since this is . p e o n d n g n, E s at s S d s

Proc. of SPIE Vol. 10705 1070514-15

(18)

hui

gI><I1l1i711

alma ;

Fal

Accor

Objec-2 obje

Even

bry-Perot ding to obs(

t + sky cont

cts contribu

and Odd coi

erving mod( :ribution ition ntribution Figure 15 the observ

For the scient is independen pseudo-slit ca was first extra An order, fro the fiber-slit. calibration is with a simpl function with systematic e a SNR of abo

7. CONC

In this paper for the high purpose (and The modular which compo been detailed produced and probing the fu E2E simulato development instrument op . ELT-HIRES G ving mode is re tific frame, th nt of the slit-i an be extracte acted and reco m the first ava The etalon w feasible. The le fit compute h the recover ect stemming out 40 in a sing

LUSION

we have prese resolution sp ultimate goal r and flexible ose the entire d, highlightin d successfully full chain feasi or and the DRS

of the spectr ptimization (o

G2V star simul call in the figur

he optimal extr llumination fu ed together (se overed the bla ailable E2E sc was extracted

one dimensio ed in MATLA

red one (spe g from the sim gle radial velo

ented the deve pectrograph at l for future de design philo tool have be ng its speed y processed by

ibility and con S pipeline (the rograph and w r to overcome

lated raw frame re. As explained raction was do unction, mean ee panels of F aze function al cience frame, d separately fr onal spectrum AB and then ectrum) by D mplified wavel ocity measurem elopment strat t the ELT (E evelopments) o osophy as wel een described performance. y the beta rec nsistency of th e specific one will enable ast e possible prob

e (see text for d d in the text thi

one by slit-de ning that di e Figure 16). Th long a single o

was extracted rom the scien m has been wav

the theoretic DRS. The ove length solutio ment.

tegy and archi LT HIRES). of modeling th ll as the phys d. Furthermore The results cipes of the C he system. Th e for ELT-HIR tronomers and blems) well b

details). The gen s simulation is composition ( erently illumin he same flat-fie order. d with simulta nce, keeping i velength calib cal spectrum erall dispersio on) is roughly itecture of the The simulato he full observ sical modelin e the computa concerning CRIRES+ inst is is a key step RES) which w d engineers to efore the first

neral possible f for spectro-pola

(see for detail nated fibers (d

eld frame that aneous

etalon-dentical pixel brated (using t has been cro on and uncer few m/s that i e current versi or architecture vation chain as g of the diffe ational archite the synthetic trument DRS p for the futur ill help the fu o perform scie t light.

fiber bundles fl arimetric mode

ls20_{), a robust}

due to the see t was used for

calibration on l scales so th the extracted e oss correlated rtainty (exclu is the expecte ion of End-to-e has bEnd-to-eEnd-to-en sEnd-to-e s much realist erent Module ecture of the echellogram have been p re developmen ture design an ence verificat lux according to e. algorithm tha eing) along the

r order tracing n either side o at wavelength etalon spectra using a CCF uding obvious d accuracy for -End simulator et up with the tic as possible s (and Units) simulator has m (raw frame) presented, thus nts of both the nd architecture ion as well as o at e g f h ) F s r r e e. ), s ) s e e s

Proc. of SPIE Vol. 10705 1070514-16

(19)

2.

IIIII011II

ti Ali fy Y IA I' I 1111111110 V ey i I i I 1 IIIIIIII ii III ,. ,,

V

.v+

IIIIIIIIII

II Figure 16 bottom pa Figure 17 simultane . Order tracing anel: reconstruc 7. Extraction o eous reference e along spectral cted surface. Rig

of the 146th or etalon. direction of the ght panels: Extr der of the EL e order 146 of H racted scientific T-HIRES Z B HIRES Z band. c spectrum and Band. Upper pa

Left upper pan

illumination pr

anel: scientific

el: RAW data. L

rofile of the fibe

spectrum. Low

Left

er-slit.

wer panel:

Proc. of SPIE Vol. 10705 1070514-17

(20)

REFERENCES

[1] European Southern Observatory, [The E-ELT construction proposal], (2011).

[2] Goodwin, M., Smedley, S., Barnes, S., Farrel, T., Barden, S., “Data simulator for the HERMES instrument”, Proc. SPIE 7735, 77357U-1 (2010).

[3] Jarno, A., Bacon, R., Ferruit, P., and Pécontal-Rousset, A., “Numerical simulation of the VLT/MUSE instrument” Proc. SPIE 7017, 701710 (2008).

[4] Maiolino, R., Haehnelt, M., Murphy, M., et al., “A Community Science Case for E-ELT HIRES”, arXiv:1310.3163, (2013).

[5] Zerbi, F. M., Bouchy, F., Fynbo, J., et al., “HIRES: The High Resolution Spectrograph for E-ELT”, Proc. SPIE 9147, 914723-1, (2014).

[6] A. Marconi, P. Di Marcantonio, V. D’Odorico, et al., “EELT-HIRES the high-resolution spectrograph for the E-ELT”. Proc. SPIE 9908, Ground-based and Airborne Instrumentation for Astronomy VI, 990823, 2016.

[7] Di Marcantonio P., Cirami R., et al., “ELT High Resolution Spectrograph: Phase-A software architecture study”, Proc. SPIE 10707-65, 2018.

[8] Marconi A., Di Marcantonio P., Maiolino R., et al., “ELT-HIRES, the high resolution spectrograph for the ELT: results from the Phase A study”, Proc. SPIE 10702-70, 2018.

[9] Parry, I., Bunker, A., Dean, A., et al. “CIRPASS: description, performance and astronomical results”, Proc. SPIE 5492, p. 1135 (2004).

[10] http://www.eso.org/observing/etc/bin/gen/form?INS.MODE = swspectr + INS.NAME = SKY CALC

[11] Agapito G., Puglisi A., Esposito S., “PASSATA - Object oriented numerical simulation software for adaptive optics.” Proc. SPIE 9909, 99097E, 2016.

[12] Di Varano, I., Strassmeier, K. G, et al. “Optical and mechanical architecture for the E-ELT HIRES polarimeter”, this proceeding 10330-12 (2017).

[13] Huke, P., Origlia, L., Riva, M., et al., “Phase A: Calibration Concepts for HIRES”, this proceeding 10329-91 (2017).

[14] Genoni, M., Riva, M., et al., “Optical parametric evaluation model for a broadband high resolution spectrograph at E-ELT (E-ELT HIRES)”, Proc. SPIE 9911-99 (2016).

[15] Shealy, D. L. and Ho nagle J. A., “Beam shaping profiles and propagation”, SPIE Conf. Laser Beam Shaping VI, Proc. 5876-13 (2005).

[16] Li Causi, G., et al, “Virtual MOONS, a focal plane simulator for the MOONS thousand-fiber NIR spectrograph” Proc. of SPIE Vol. 9147, 914764 (2014).

[17] Genoni, M., et al., “The end-to-end simulator for the EELT HIRES high resolution spectrograph,” Proc. SPIE 10329, 103290Z-1 (2017).

[18] Landoni M., Genoni M., Riva M., et al., “Application of cloud computing in astrophysics: the case of Amazon Web Services”, Proc. SPIE 10707-17, 2018.

[19] https://www.eso.org/sci/facilities/develop/instruments/crires_up.html

[20] Piskunov N. E., Valenti J. A., “New algorithms for reducing cross-dispersed echelle spectra.” A&A, 385, 1095, 2002.

[21] Oliva, E., Tozzi, A., Ferruzzi, D., and et al., “ELT-HIRES the high resolution instrument for the ELT: optical design and instrument architecture” SPIE 10702-317, (2018).

Proc. of SPIE Vol. 10705 1070514-18