Abstract
The Real Time Controllers (RTCs) for the W. M. Keck Observatory Adaptive Optics (AO) systems have been upgraded from a Field Programmable Gate Array (FPGA) to a Graphics Processing Unit (GPU) based solution. The previous RTCs, operating since 2007, had reached their limitations after upgrades to support new hardware including an Infra-Red (IR) Tip/Tilt (TT) Wave Front Sensor (WFS) on Keck I and a Pyramid WFS on Keck II. The new RTC, fabricated by a Microgate-led consortium with SUT leading the computation engine development, provides a flexible platform that improves processing bandwidth and allows for easier integration with new hardware and control algorithms. Along with the new GPU-based RTC, the upgrade includes a new hardware Interface Module (IM), new OCAM2K EMCCD cameras, and a new Telemetry Recording Server (TRS). The first system upgrade to take advantage of the new RTC is the Keck I All-sky Precision Adaptive Optics (KAPA) Laser Tomography AO (LTAO) system, which uses the larger and more sensitive OCAM2K EMCCD camera, tomographic reconstruction from four Laser Guide Stars (LGS), and improvements to the IR TT WFS. On Keck II the new RTC will enable a new higher-order Deformable Mirror (DM) as part of the HAKA (High order Advanced Keck Adaptive optics) project, which will also use an EMCCD camera. In the future, the new RTC will allow the possibility for new developments such as the proposed 'IWA (Infrared Wavefront sensor Adaptive optics) system. The new RTC saw first light in 2021. The Keck I system was released for science observations in late 2023, with the Keck II system released for science in early 2024.

Abstract
METIS, the Mid-infrared ELT Imager and Spectrograph, will operate an internal Single Conjugate Adaptive Optics (SCAO) system, which will mainly serve the science cases targeting exoplanets and disks around bright stars. The Extremely Large Telescope (ELT) is expected to have its first light in 2028, and the entire instrument recently passed its final design phase. The adaptive optics (AO) of METIS SCAO is designed to correct for atmospheric distortions, and is essential for diffraction-limited observations with METIS. The computational and data transfer requirements for these next generation ELT AO Real-Time Computers (RTCs) are enormous, and require advanced data processing and pipelining techniques. METIS SCAO will use a pyramid wavefront sensor (WFS), which captures incoming wavefronts at 1 kHz with a raw throughput of 148 MB/s. The RTC will ingest these WFS images on a frame-by-frame basis, compute the corrections and send them to the deformable mirror M4 and the tip/tilt mirror M5. The RTC is split up into two distinct systems: the Hard Real-Time Computer (HRTC) and the Soft Real-Time Computer (SRTC). The HRTC is responsible for computing the time sensitive wavefront control loop, while the SRTC is responsible for supervising and optimising the HRTC. A working prototype for the HRTC has been completed and operates with an RTC computation time of roughly 372 mu s. This computation is memory limited and runs on two NVIDIA A100 GPUs. This paper shows a breakdown of the HRTC on a CUDA kernel level, focusing on the tasks that run on the GPUs. We also present the performance of the HRTC and possible improvements for it.

Abstract
Context. The deployment of meter-scale (hitherto pre-focal) adaptive deformable mirrors finds some prominent examples in the leading ground-based visible to near-infrared facilities (e.g. the Very Large Telescope (VLT), the Large Binocular Telescope (LBT), or the Magellan Telescope) and is being adopted by several others (e.g. the Multiple Mirror Telescope (MMT) or Subaru). Furthermore, two out of the three giant segmented-mirror telescopes now under design will feature them. In all these cases, the proprietary technology is based on voice-coils and is limited in force, stroke, and velocity. Aims. Because of the nature of their purpose, that is, adaptive wave-front correction, any kind of optimality relies on the control of a subset of principal wave-front components or eigenmodes, for short, a basis of functions in a mathematical sense. Here we provide algorithmic procedures for generating such eigenbases, also called Karhunen–Loève (KL) modes, that integrate force limitations in their definitions whilst maintaining standard orthonormality, statistical independence, and deformable mirror span. Methods. The double-diagonalisation method was revisited to build KL modes ranked by the force applied on the actuators. Results. We analysed this new KL basis for von Kármán turbulence statistics and present the fitting error and the distribution of positions and forces. We further illustrate their use in the case of the quaternary mirror control for the European Extremely Large Telescope, and we include the outer actuator minioning and force policy constraints. © The Authors 2024.

Abstract
Capturing high-resolution imagery of the Earth's surface often calls for a telescope of considerable size, even from low Earth orbits (LEOs). A large aperture often requires large and expensive platforms. For instance, achieving a resolution of 1 m at visible wavelengths from LEO typically requires an aperture diameter of at least 30 cm. Additionally, ensuring high revisit times often prompts the use of multiple satellites. In light of these challenges, a small, segmented, deployable CubeSat telescope was recently proposed creating the additional need of phasing the telescope's mirrors. Phasing methods on compact platforms are constrained by the limited volume and power available, excluding solutions that rely on dedicated hardware or demand substantial computational resources. Neural networks (NNs) are known for their computationally efficient inference and reduced onboard requirements. Therefore, we developed a NN-based method to measure co-phasing errors inherent to a deployable telescope. The proposed technique demonstrates its ability to detect phasing errors at the targeted performance level [typically a wavefront error (WFE) below 15 nm RMS for a visible imager operating at the diffraction limit] using a point source. The robustness of the NN method is verified in presence of high-order aberrations or noise and the results are compared against existing state-of-the-art techniques. The developed NN model ensures its feasibility and provides arealistic pathway towards achieving diffraction-limited images. (c) 2024 Optica Publishing Group

Abstract
METIS is the European Extremely Large Telescope (ELT) 1st-generation Mid-Infrared ELT Imager and Spectrograph. It will offer spectroscopic, imaging and coronagraphic capabilities from 3 up to 13 microns with Adaptive-Optics correction. With its Final Design Review due late 2022 we report on the wavefront control strategy devised to meet the METIS science and technological requirements. Such strategy addresses challenging aspects as i) the appearance of differential petal piston modes in the presence of secondary mirror support struts caused either by numerical processing or the actual, physical low-wind effect, ii) the numerical pupil derotation and mis- reg compensation, iii) the adaptation to transient disturbance signals such as telescope-to-instrument handover control and iv) the compliance with constrained modal control of the pre-focal beam corrector mirrors (M4/M5). The overall METIS wavefront control strategy consists in a split approach cemented in a sequence of steps: 1) Tikhonov-regularised spatial wavefront estimation/reconstruction on a zonal Cartesian coordinate system tied to the pyramid (P-WFS) sampling pixel grid, 2) the regularised projection onto a global modal control space including correction of mis-registrations and rotation between the P-WFS coordinate grid and the ELTs M4/M5, and 3) the time-filtering through the application of proportional-integral control before converting to actuator commands readied for the ELTs collaborative TT off-loading scheme whilst avoiding hitting the mirrors constraints in amplitude, speed and force. We present physical-optics simulation results of the whole AO system obtained with prototyped instances of the real-time and soft-real-time computers including sensitivity analysis with respect to observational, atmospheric, non-atmospheric (telescope-intrinsic such as wind-induced low-order modes comprising tip-tilt) and instrument-specific conditions and disturbances. An error budget is put together that meets the METIS science requirements in terms of wavefront error with reassuring margins thus endorsing the strategy devised.

Abstract
The infrared low order sensor (IRLOS) upgrade project was recently launched to increase the sky coverage of GALACSI narrow-field mode (NFM). While the baseline is to perform low-order wavefront sensing with a 2x2 Shack-Hartmann wavefront sensor (SHWFS) operating in the J+H band, a full-pupil mode was proposed to address the faintest end of the magnitude range by concentrating the photons from the full aperture in a single point spread function (PSF). In this context, we have investigated the wavefront sensing approach called the linearized focal-plane technique (LIFT). It enables the retrieval of low-order modes such as tip/tilt, defocus, astigmatism (and possibly more) from a single focal-plane PSF of a very faint natural guide star (NGS) target. LIFT is a phase diversity technique based on introducing a known amount of astigmatism into the optical path. The morphological change induced by the astigmatic shift allows encoding information about the phase aberrations into the PSF morphology. In this work, we discuss the linearity and flux sensitivity of the method and present experimental on-sky results obtained at the VLT. We discuss the applicability of this method in realistic conditions and the limitations that this method can face while operated on-sky.