Control algorithms for dual-wavefront sensor single-conjugate adaptive optics

TL;DR

This work tackles non-common-path aberrations in high-contrast imaging by integrating focal-plane WFS information into a dual-WFS, single-DM adaptive optics loop. It develops a frequency-domain control framework with multiple controller families, including a double integrator with a high-pass filter, and validates these designs against time-domain end-to-end simulations. The key finding is that applying a fast-arm high-pass filter to suppress inter-arm NCP transfer yields substantial improvements in the science-plane error X, especially when NCP is strong; simple LQG-based schemes offer limited gains in this dual-WFS setting. Practically, the study demonstrates that FP-WFS concepts can be implemented within existing AO systems without extra hardware, enabling rapid testing and deployment for direct imaging of exoplanets, with a path toward more advanced control strategies such as MPC in future work.

Abstract

High-contrast imaging systems using active control with adaptive optics (AO) are often limited by non-common path (NCP) aberrations that are seen only at the final science image. AO systems employing focal-plane wavefront sensors (FP-WFSs) are able to simultaneously correct NCP aberrations and measure science images, but they typically require a second stage of control that adds system cost and complexity. We present control algorithms to augment AO systems with FP-WFSs within their existing control setup. We demonstrate inter-arm NCP aberration transfer can be mitigated through temporal filtering, present frequency- and time-domain validation of controller stability and performance, and discuss the optimality of the chosen controllers. This work will enable the development, testing, and installation of FP-WFS technologies for direct imaging of exoplanets.

Figures (10)

• Figure 1: The dual-WFS single-DM control scheme for a representative astronomical AO system, indicating the introduction of common- and non-common-path disturbances. The common optical path is shown in purple and the non-common optical paths are shown in blue and red. NCP disturbances are shown in localized points in each beam for visualization purposes, but can in general arise anywhere along the non-common paths. *The slow WFS is also the science imager. †If only the fast WFS is used for feedback, this block represents the standard single-WFS control computation and corresponds to the schematic switch shown in the open state. ‡Here latency represents all delays, including servo, processing, network/propagation, and mechancial latency of the DM.
• Figure 2: A block diagram of two-WFS control, indicating the different frame rates for the two WFSs, to enable controller analysis and optimization. Disturbances are shown in red and relevant intermediate and output signals are shown in purple.
• Figure 3: The estimated open-loop PSDs for the atmosphere, non-common-path aberrations, and noise as seen in the time-domain AO simulation (§\ref{['sec:2dsims']}). We take $r_0 = 1$ m for NCP aberrations and $f_\text{cross} = 50$ Hz. The dashed lines show the analytic power spectrum models considered for frequency-domain optimization, and the solid lines show the PSDs of a representative 1-second time series.
• Figure 4: The single IC controller is unable to correct NCP aberrations seen only by the slow WFS. Since there is no slow WFS, there is no attenuation of $L_{\text{slow}}$ (solid green line in Figure \ref{['fig:errx_singlewfs']}), and therefore NCP error at $X$ is a large part of the residual error (dark green line in Figure \ref{['fig:integrands_singlewfs']}). Not attenuating slow NCP error results in a higher closed-loop error at $X$ than $Y$ (purple and pink lines in Figure \ref{['fig:psds_singlewfs']}).
• Figure 5: The double IC controller is able to correct slow NCP aberrations, but suffers from inter-arm NCP transfer. There is attenuation of slow NCP aberrations (solid green line in Figure \ref{['fig:errx_doublewfs']}) but they still contribute strongly to the residual error (dark green line in Figure \ref{['fig:integrands_doublewfs']}) and closed-loop performance at $X$ is still marginally worse than $Y$.