Motivated by the potential of nondiffraction limited, real-time computational image sharpening with neural networks in astronomical telescopes, we studied wavefront sensing with convolutional neural networks based on a pair of in-focus and out-of-focus point spread functions. By simulation, we generated a large dataset for training and validation of neural networks and trained several networks to estimate Zernike polynomial approximations for the incoming wavefront. We included the effect of noise, guide star magnitude, blurring by wide-band imaging, and bit depth. We conclude that the “ResNet” works well for our purpose, with a wavefront RMS error of 130 nm for r0 = 0.3 m, guide star magnitudes 4 to 8, and inference time of 8 ms. It can also be applied for closed-loop operation in an adaptive optics system. We also studied the possible use of a Kalman filter or a recurrent neural network and found that they were not beneficial to the performance of our wavefront sensor.
The Carlina hypertelescope is a planned sparse aperture 100 m telescope with pupil densification. The telescope
has a spherical primary with segments located in a valley between mountains, and additional optical elements in a
gondola suspended in eight cables some 100 m above the primary mirror. The resolution is about 1.2×10-3 arcsec.
It is imperative that the position and attitude of the gondola be maintained within tight tolerances during
observation and star tracking. The present design has servo-controlled winches on the ground for control of
the gondola via the cables. An integrated model of the system, including optics, cables, gondola, position and
attitude control system, and wind disturbances has been set up. The structural and control models are linear.
Calculations in the frequency domain and simulations in the time domain show that the performance of the
telescope with the present design seems adequate for short exposures. However, for long-exposure operation, the
gondola stability should be improved by about two orders of magnitude. Recommendations are given on possible
approaches for performance improvement.
Integrated models including optics, structures, control systems, and disturbances are important design tools
for Extremely Large Telescopes (ELTs). An integrated model has been formulated for the European ELT
and it includes telescope structure, main servos, primary mirror segment control system, wind, optics, wavefront
sensors, deformable mirror, and an AO reconstructor and controller. There are three model phases: Initialization,
execution of a solver to determine time responses, and post-processing. In near future, the model will be applied
for performance studies and design trade-offs for the European ELT.
The next generation of telescopes, the Extremely Large Telescopes (ELTs), will have a multitude of control loops to
maintain nearly diffraction-limited performance in the presence of atmospheric turbulence and external disturbances, for
instance from wind. Integrated simulation models combining structural and optical modeling together with control
system modeling are efficient tools for prediction of performance of ELTs. Such models include submodels of structures,
adaptive optics, atmosphere, wind load, deformable mirrors and a segmented primary mirror. So far the models applied
have been applicable to observations in the K-band. However, there is a desire to also operate the ELTs with adaptive
optics at wavelengths in the visible range. We here give estimates of the feasibility of performing such simulations. We
set up scaling laws for the design parameters as a function of wavelength of operation and we show that the execution
time for an integrated model of an ELT depends dramatically on the operation wavelength. We also discuss the
consequences of different choices of model refinement. Finally we present estimates of the execution time for integrated
models of ELTs. We show that accurate modeling in the K-band calls for long execution times, even with parallel
computers. For wavelengths in the visible range, only the very simplest models are feasible due to execution time
limitations, thereby precluding many interesting studies related to noise sensitivity and limiting magnitude for guide
Observational High Time Resolution Astrophysics differs from conventional astrophysics in regard to the detectors
employed which have a time resolution less than that obtainable through CCD with a normal readout τ < a few
minutes. This paper looks at the implications for HTRA from extremely large telescopes and specifically, as an
exemplar its possible impact on pulsar astrophysics. We demonstrate, by using the derived point-spread-function
from models of the Euro50 telescope, the possible effects active and adaptive mirrors have on observing rapidly
varying astronomical objects.
The Euro50 is a proposed 50m extremely large telescope for optical and infrared wavelengths. To study and predict the performance of the complete telescope system, an integrated model combining the structural model of the telescope, optics models, the control systems and the adaptive optics has been established. Wind and atmospheric disturbances are also included in the model. The model is written in MATLAB and C. It is general and modular and built around dedicated ordinary differential equation solvers. The difference in time constants between subsystems is exploited to speed up calculations. The solvers can handle discontinuities and subsystem mode changes. The high degree of modularity allows different telescope designs to be modelled by rearranging subsystem blocks. Certain subsystems, for instance adaptive optics, can also run in a standalone fashion. Parts of the model are parallelized for execution on a large shared memory machine. The resulting architecture of the integrated model and sample results using the code for different telescope models are presented.
With Euro50 as a convenient telescope laboratory, the Euro50 team has continued development aiming at a European
extremely large telescope (ELT). Here, we give a progress report. The needs of science and instrumentation are briefly
discussed as is the importance of photometric stability and precision. Results are reported from work on integrated
modelling. Details are given concerning point-spread functions (PSFs) obtained with and without adaptive optics (AO).
Our results are rather encouraging concerning AO photometry and compensation of edge sensor noise as well as
regarding seeing-limited ELT operation. The current status of our development of large deformable mirrors is shown.
Low-cost actuators and deflection sensors have been developed as have hierarchic control algorithms. Fabrication of
large thin mirror blanks as well as polishing and handling of thin mirrors has been studied experimentally. Regarding
adaptive optics, we discuss differential refraction and the limitations imposed by dispersive optical path differences
(OPDs) and dispersive anisoplanatism. We report on progress in laser guide star (LGS) performance and a real-time online experiment in multi-conjugate AO (MCAO). We discuss ELTs, high-resolution spectroscopy and pupil slicing with
and without use of AO. Finally, we present some recent studies of ELT enclosure options.
In previous work we have countered computational demands faced in integrated modelling by developing and using a parallel toolkit for MATLAB. However the use of an increasingly realistic model makes the computational requirements of the model much larger, particularly in wavefront sensing, reaching a point where simulations of several real time seconds were no longer practical taking up to 3 weeks per second. In response to this problem we have developed optimised C code to which MATLAB off loads computation. This code has numerous advantages over native MATLAB computation. It is portable, scaleable using OpenMP directives and can run remotely using Remote Procedure Calls (RPCs). It has opened up the possibility of exploiting high end Itanium and Opteron based shared memory systems, optimised 3rd party libraries and aggressive compiler optimisation. These factors combined with hand-tuning give a performance increase of the order of 100 times. The interface to the rest of the model remains the same so the overall structure is unchanged. In addition we have developed a similar system based on Message Passing Interface version 2, (MPI-2) which allows us to exploited clusters. Here we present an analysis of techniques used and gains obtained along with a brief discussion of future work.
The Euro50 is a proposed 50m extremely large telescope for optical and infrared wavelengths. To study and predict the performance of the complete telescope system, an integrated model combining the structural model of the telescope, optics models, the control systems and the adaptive optics has been established. Wind and atmospheric disturbances are also included in the model. The integrated model is written in MATLAB and C. To satisfy memory demands and to achieve acceptable execution times, 64-bit MATLAB is used and part of the model is run on a shared memory machine using OpenMP. We present results from simulations with a complete integrated single conjugate adaptive optics model. Various sensor and actuator geometries are evaluated. A comparison of wind loading and atmospheric turbulence effects is also presented. The model shows that the telescope will be essentially seeing limited under wind load and no AO correction.
The Euro50 is an extremely large telescope for optical and infrared wavelength with a 50 m primary mirror. It has a segmented, aspherical primary mirror and an aspherical, deformable secondary in a Gregorian layout. A tentative conceptual design exists and has been documented in a study report. Recent activities have concentrated on the science case for extremely large telescopes in the 50 m class and on identification of potential technical "show stoppers". The science case investigation has identified four fields of particular interest. The studies of critical technical issues have concentrated on atmospheric dispersion effects for high-resolution adaptive optics for extremely large telescopes, and on the influence of wind and other disturbances on wavefront control. Wind load on the telescope, the primary mirror and the enclosure has been studied using wind tunnel measurements and computational fluid dynamics. The impact of wind on the total system has been investigated using an integrated model that includes the telescope structure, the primary mirror segment alignment system, the secondary mirror alignment system, and single conjugate adaptive optics using the deformable secondary mirror. The first, tentative results show that wind disturbances may be significant and that the task of correcting for wind residuals may be at least as large for the adaptive optics system as that of correcting for atmospheric aberrations. The results suggest that use of extremely large telescopes for observations of earth-like planets around nearby stars may imply a considerable challenge.
The Euro50 is an astronomical extremely large telescope for optical and infrared wavelength with a 50 m primary mirror. The telescope will have an elaborate control system ("live optics") to correct for atmospheric and telescope aberrations. To study and predict performance of the complete telescope system, an integrated model combining the structural model of the telescope, optics models, the control systems, and the adaptive optics has been established. Wind is taken into account on the basis of wind tunnel measurements and computer fluid dynamics calculations. Atmospheric aberrations are included using a seven-layer atmosphere model. The integrated model is written in Matlab and is run on a cluster computer to achieve acceptable execution times. Dedicated ordinary differential equation solvers have been written and a special toolkit for communication between Matlab processes on different nodes of the cluster computer has been set up. Preliminary results from the complete integrated model, including adaptive optics, are shown.
MATLAB and its companion product Simulink are commonly used tools in systems modelling and other scientific disciplines. A cross-disciplinary integrated MATLAB model is used to study the overall performance of the proposed 50m optical and infrared telescope, Euro50. However the computational requirements of this kind of end-to-end simulation of the telescope's behaviour, exceeds the capability of an individual contemporary Personal Computer. By parallelizing the model, primarily on a functional basis, it can be implemented across a Beowulf cluster of generic PCs. This requires MATLAB to distribute in some way data and calculations to the cluster nodes and combine completed results. There have been a number of attempts to produce toolkits to allow MATLAB to be used in a parallel fashion. They have used a variety of techniques. Here we present findings from using some of these toolkits and proposed advances.
The Euro50 is a proposed 50 m optical and infrared telescope. It will have thousands of control loops to keep the optics aligned under influence of wind, gravity and thermal loads. Cross-disciplinary integrated modeling is used to study the overall performance of the Euro50. A sub-model of the mechanical structure originates from finite element modeling. The optical performance is determined using ray tracing, both non-linear and linearized. The primary mirror segment alignment control system is modeled with the 618 segments taken as rigid bodies. Adaptive optics is included using a layered model of the atmosphere and sub-models of the wavefront sensor, reconstructor and controller. The deformable mirror is, so far, described by a simple influence function and a second order dynamical transfer function but more detailed work is in progress. The model has been implemented using Matlab/Simulink on individual computers but it will shortly be implemented on a Beowulf cluster within a trusted network. Communication routines between Matlab on the cluster processors have been written and are being benchmarked. Representative results from the simulations are shown.