Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userFocal Plane Wavefront Sensing using Machine Learning: Performance of Convolutional Neural Networks compared to Fundamental Limits
83
0
0.0
(
0
)
Added by
Gilles Orban De Xivry
Publication date
2021
fields
Physics
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
Focal plane wavefront sensing (FPWFS) is appealing for several reasons. Notably, it offers high sensitivity and does not suffer from non-common path aberrations (NCPA). The price to pay is a high computational burden and the need for diversity to lift any phase ambiguity. If those limitations can be overcome, FPWFS is a great solution for NCPA measurement, a key limitation for high-contrast imaging, and could be used as adaptive optics wavefront sensor. Here, we propose to use deep convolutional neural networks (CNNs) to measure NCPA based on focal plane images. Two CNN architectures are considered: ResNet-50 and U-Net which are used respectively to estimate Zernike coefficients or directly the phase. The models are trained on labelled datasets and evaluated at various flux levels and for two spatial frequency contents (20 and 100 Zernike modes). In these idealized simulations we demonstrate that the CNN-based models reach the photon noise limit in a large range of conditions. We show, for example, that the root mean squared (rms) wavefront error (WFE) can be reduced to < $lambda$/1500 for $2 times 10^6$ photons in one iteration when estimating 20 Zernike modes. We also show that CNN-based models are sufficiently robust to varying signal-to-noise ratio, under the presence of higher-order aberrations, and under different amplitudes of aberrations. Additionally, they display similar to superior performance compared to iterative phase retrieval algorithms. CNNs therefore represent a compelling way to implement FPWFS, which can leverage the high sensitivity of FPWFS over a broad range of conditions.
rate research
Read More
Focal plane wavefront sensing is an elegant solution for wavefront sensing since near-focal images of any source taken by a detector show distortions in the presence of aberrations. Non-Common Path Aberrations and the Low Wind Effect both have the ability to limit the achievable contrast of the finest coronagraphs coupled with the best extreme adaptive optics systems. To correct for these aberrations, the Subaru Coronagraphic Extreme Adaptive Optics instrument hosts many focal plane wavefront sensors using detectors as close to the science detector as possible. We present seven of them and compare their implementation and efficiency on SCExAO. This work will be critical for wavefront sensing on next generation of extremely large telescopes that might present similar limitations.
In this article we show that the vector-Apodizing Phase Plate (vAPP) coronagraph can be designed such that the coronagraphic point spread functions (PSFs) can act as a wavefront sensor to measure and correct the (quasi-)static aberrations, without dedicated wavefront sensing holograms nor modulation by the deformable mirror. The absolute wavefront retrieval is performed with a non-linear algorithm. The focal-plane wavefront sensing (FPWFS) performance of the vAPP and the algorithm are evaluated with numerical simulations, to test various photon and read noise levels, the sensitivity to the 100 lowest Zernike modes and the maximum wavefront error (WFE) that can be accurately estimated in one iteration. We apply these methods to the vAPP within SCExAO, first with the internal source and subsequently on-sky. In idealised simulations we show that for $10^7$ photons the root-mean-square (RMS) WFE can be reduced to $simlambda/1000$, which is 1 nm RMS in the context of the SCExAO system. We find that the maximum WFE that can be corrected in one iteration is $simlambda/8$ RMS or $sim$200 nm RMS (SCExAO). Furthermore, we demonstrate the SCExAO vAPP capabilities by measuring and controlling the lowest 30 Zernike modes with the internal source and on-sky. On-sky, we report a raw contrast improvement of a factor $sim$2 between 2 and 4 $lambda/D$ after 5 iterations of closed-loop correction. When artificially introducing 150 nm RMS WFE, the algorithm corrects it within 5 iterations of closed-loop operation. FPWFS with the vAPPs coronagraphic PSFs is a powerful technique since it integrates coronagraphy and wavefront sensing, eliminating the need for additional probes and thus resulting in a $100%$ science duty cycle and maximum throughput for the target.
High quality, repeatable point-spread functions are important for science cases like direct exoplanet imaging, high-precision astrometry, and high-resolution spectroscopy of exoplanets. For such demanding applications, the initial on-sky point-spread function delivered by the adaptive optics system can require further optimization to correct unsensed static aberrations and calibration biases. We investigated using the Fast and Furious focal-plane wavefront sensing algorithm as a potential solution. This algorithm uses a simple model of the optical system and focal plane information to measure and correct the point-spread function phase, without using defocused images, meaning it can run concurrently with science. On-sky testing demonstrated significantly improved PSF quality in only a few iterations, with both narrow and broadband filters. These results suggest this algorithm is a useful path forward for creating and maintaining high-quality, repeatable on-sky adaptive optics point-spread functions.
Direct imaging of Earth-like planets from space requires dedicated observatories, combining large segmented apertures with instruments and techniques such as coronagraphs, wavefront sensors, and wavefront control in order to reach the high contrast of 10^10 that is required. The complexity of these systems would be increased by the segmentation of the primary mirror, which allows for the larger diameters necessary to image Earth-like planets but also introduces specific patterns in the image due to the pupil shape and segmentation and making high-contrast imaging more challenging. Among these defects, the phasing errors of the primary mirror are a strong limitation to the performance. In this paper, we focus on the wavefront sensing of segment phasing errors for a high-contrast system, using the COronagraphic Focal plane wave-Front Estimation for Exoplanet detection (COFFEE) technique. We implemented and tested COFFEE on the High-contrast imaging for Complex Aperture Telescopes (HiCAT) testbed, in a configuration without any coronagraph and with a classical Lyot coronagraph, to reconstruct errors applied on a 37 segment mirror. We analysed the quality and limitations of the reconstructions. We demonstrate that COFFEE is able to estimate correctly the phasing errors of a segmented telescope for piston, tip, and tilt aberrations of typically 100nm RMS. We also identified the limitations of COFFEE for the reconstruction of low-order wavefront modes, which are highly filtered by the coronagraph. This is illustrated using two focal plane mask sizes on HiCAT. We discuss possible solutions, both in the hardware system and in the COFFEE optimizer, to mitigate these issues.
We present a method to calibrate a high-resolution wavefront-correcting device with a single, static camera, located in the focal plane; no moving of any component is needed. The method is based on a localized diversity and differential optical transfer functions (dOTF) to compute both the phase and amplitude in the pupil plane located upstream of the last imaging optics. An experiment with a spatial light modulator shows that the calibration is sufficient to robustly operate a focal-plane wavefront sensing algorithm controlling a wavefront corrector with ~40 000 degrees of freedom. We estimate that the locations of identical wavefront corrector elements are determined with a spatial resolution of 0.3% compared to the pupil diameter.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
