Update to End to End LSST Science Simulation Garrett Jernigan and John Peterson December, 2004 Status of the Science End-to-End Simulator: 1. Sky Models.

Update to End to End LSST Science SimulationGarrett Jernigan and John PetersonDecember, 2004

Status of the Science End-to-End Simulator:

1. Sky Models (two modes)

Grids of stars / FITS interface for arbitrary image

2. Atmospheric Model

Kolmogorov refractive layer models

3. Optics and Deformations

Geometric ray trace with perturbations

4. Detector Model

Conversion depth/Diffusion (Andy Rasmussen)

Telescope diffraction

1. Component Design, Modeling, and Simulation: (a required routine engineering activity)

2. Science End-to-End Simulator:(early; informs design; 10% accuracy)

3. Engineering End-to-End simulator:(late; follows design in detail, <1% accuracy)

Three Types of Simulators:

Atmospheric Models

Raytrace Code:

- Monte Carlo of photons through Atm. (also optics and detector: end-to-end) - Multi-layer Atmospheric Model (each layer a frozen screen) - Modified Kolmogorov Model for each layer (Random Gaussian with outer scale) - Now: Each Layer contains 256x256x256 cube (3D Kolmogorov Model) - Soon: Each layer: 2048x2048x16 (also 3D Kolmogorov Model) - Refractive Approximation for Raytrace with Phase Screen - Modeling ground/dome effects (not currently included)

Validation Code:

- PSF determined from multiple phase screen with full diffraction (FFTs required) - Non-Kolmogorov Models (atmospheric wedge, wind sheer driven flows) - Time Dependent Kolmogorov Models (drop frozen screen assumption) - Numerical Hydrodynamic Simulations

Near Term Goals (Science Applications)

- Results on PSF atmosphere only (raytrace and validation code results) - Distribution of seeing and ellipticity of PSF - Semi-analytic form for e1 and e2 de-correlation versus angle for stars - Numerical simulation to verify semi-analytic model - Determine the effects of LSST optics (Zernike perturbations only) - Simple examples of the shear of ideal galaxies (PSF corrected) - Some simple tests to estimate effects of non-Kolmogorov models - Validate with real data (Guide stars ?; large aperture telescopes ?)

Kolmogorov Model

Numerical simulation (Porter)

3-D atmospheric density Density Slice

Single layer phase screen based on Kolmogorov spectrum Refraction raytraced

Phase MapVector Perturbations

Altitude

Structure Function Wind Speed

Multilayer Models

Vernin et al., Gemini RPT-A0-G0094from Sebag

QuickTime™ and aYUV420 codec decompressor

are needed to see this picture.

Telescope Diffraction

AtmosphericDiffraction

Detector Model: (Rasmussen)Refraction for light entering the Si surface reduces the cone angle of the incident beam (cf. Radeka)Finite electric field at point of interaction leads to a lateral diffusion during drift time to the channel.

Electric field function is dependent on doping density profile in the Si and bias voltage.

Interaction length into Si is strongly wavelength dependent, and also temperature dependent, particularly at long wavelength.

Photon detection by CCD alters the position of best focus and also the PSF.

Telescope Raytrace + Perturbations:

Fast Geometric Optics codeFinds ray intercept / refraction or reflectionHandles non-sequential straylightHas arbitrary rotations, translations,and perturbations

Perturbations: Residual wavefront Zernike coefficients as deformations and vary as function of time

Telescope (No perturbations)PSFs separated by 0.6 degrees centered at (0,0)

Dotted grid is 10 microns PSFs separated by 10 arcseconds centered at ( +1.5, 0) degrees

Dotted grid is 10 microns

Optics+Perturbations on PrimaryPSFs separated by 0.6 degrees centered at (0,0)

Dotted grid is 10 microns PSFs separated by 10 arcseconds centered at ( +1.5, 0) degrees

Dotted grid is 10 microns

Optics+Perturbations on Secondary

PSFs separated by 0.6 degrees centered at (0,0)

Dotted grid is 10 microns PSFs separated by 10 arcseconds centered at ( +1.5, 0) degrees

Dotted grid is 10 microns

Optics+Perturbations on TertiaryPSFs separated by 0.6 degrees centered at (0,0)

Dotted grid is 10 microns PSFs separated by 10 arcseconds centered at ( +1.5, 0) degrees

Dotted grid is 10 microns

Telescope: Optics+ 10 realizations of zernike perturbations (all mirrors)PSFs separated by 0.6 degrees centered at (0,0)

Dotted grid is 10 microns PSFs separated by 10 arcseconds centered at ( +1.5, 0) degrees

Dotted grid is 10 microns

Telescope (w/ Perturbations)+AtmospherePSFs separated by 0.6 degrees centered at (0,0)

Dotted grid is 10 microns PSFs separated by 10 arcseconds centered at ( +1.5, 0) degrees

Dotted grid is 10 microns

Telescope+Perturbations+Atmosphere+WindPSFs separated by 0.6 degrees centered at (0,0)

Dotted grid is 10 microns PSFs separated by 10 arcseconds centered at ( +1.5, 0) degrees

Dotted grid is 10 microns

HDF galaxies

Image raytraced+ perturbations+atmosphere+wind

Sky Image Simulations

Perfect telescope Zernike Pert. on all mirrors

Ellipticity Residual Studies

Ellipticity Vectors: computed from weighted Qij momentsellipticity shot noise ~ 1/sqrt(N)

Ellipticity Residuals as a function of separationPreliminary

Many investigations continuing to understand ellipticity changes:Telescope: (current sims give e=0.3 for perfect telescope

e=0.1 for 10 realizations of up to 5th order pert.decorrelated=a degree for all mirrors)

# of Zernikes (more reduces corr.)amplitude of Zernikes (affects rel. importance)which mirror (should be less correlated w/ tertiary,

but not so obvious?)rate of changes of Zernikes (affect corr.)

Atmosphere: (current sims give e=0.05, decorrelated=40”)seeing/structure function (affect rel. importance)outer Kolmogorov scale (increase e)wind (reduces e)layer height (reduces correlation)non-Kolmogorov effects (ellip?)

Three Types of Simulators Garrett Jernigan December 9, 2004 Component Design, Modeling and Simulation: * High fidelity engineering calculations for targeted components

of the system. * Often complex, inefficient, interactive general purpose codes

(example include optics design, finite element mechanical modeling, thermal modeling, etc).

• Results from these types of simulations inform the design of

efficient end - to - end simulators for the targeted component (examples included normal modes of mirrors, geometric motions, fabrication accuracy, etc).

• Code developed for this type of simulator is rarely used in raw

form as part of an end - to - end simulator due to computational inefficiency and specialized form (often propriet ary).

* Early effort required for expensive and key components

especially those items that are identified in the critical path. * This type of component simulation is required for success (as

contrasted with efficient end - to - end simulators that are not necessarily needed for success).

Science End - to - End Simulation: * Early effort likely by scientists not professional software engineers with a "level of effort" work and management approach. * The goal is to address the most important component s of the system that determine the quality of the science. * Rapid coding and prototyping using any tools without much formal review. * Primary goal is to build a tool that answers science performance questions (10% accuracy) very early in the p roject life cycle. * Secondary goal is to inform design decisions (more difficult to achieve in a timely manner).

* The atmospheric model and the model of the telescope/camera control system are the two most important components for LSST. At a minimum both of these components must be in place to even call the simulator an end-to-end simulator (initial REV 0 version).

* The goal of REV 0 is to facilitate use of the model by other

members of the LSST team but without costly software infrastructure but with some significant and useful documentation and science description.

Engineering End-to-End Simulation: * Carefully planned and reviewed approach (managed like hardware). * Mostly implemented by software engineers not scientists. * Design is informed by the Science End-to-End simulator (The

algorithms and not the code in the Science end-to-end simulator have value for the Engineering End-to-End simulator).

* The goal is to develop a 1% accurate model of the

Telescope/Camera that realizes the actual detailed engineering design (all details including those details that are not science drivers).

* Clearly completion will be too late to inform design. * Cost and schedule controlled (likely an order of magnitude

larger than the Science End-to-End simulation effort.) * Priority goal is to be a component of or an interface to the

Data Management System. Ideally the Engineering end-to-end simulator should be completed several years prior to LSST first light so that the full up initial Data Management System can be tested early without any LSST hardware.