While it is attractive to integrate a deformable mirror (DM) for adaptive optics (AO) into the telescope itself
rather than using relay optics within an instrument, the resulting large DM can be expensive, particularly for
extremely large telescopes. A low-cost approach for building a large DM is to use voice-coil actuators, and rely
on feedback from mechanical sensors to improve the dynamic response of the mirror sufficiently so that it can
be used in a standard AO control system. The use of inexpensive voice-coil actuators results in many lightly-
damped structural resonances within the desired control bandwidth. We present a robust control approach for
this problem, and demonstrate performance in a closed-loop AO simulation, incorporating realistic models of
low-cost actuators and sensors. The first contribution is to demonstrate that high-bandwidth active damping
can be robustly implemented even with non-collocated sensors, by relying on the "acoustic limit" of the structure
where the modal bandwidth exceeds the modal spacing. Next we introduce a novel local control approach, which
significantly improves the high spatial frequency performance relative to collocated position control, but without
the robustness challenges associated with a global control approach. The combination of these "inner" control
loops results in DM command response that is demonstrated to be sufficient for integration within an AO system.
To improve the mechanical characteristics of actively controlled continuous faceplate deformable mirrors in adaptive optics, a strategy for reducing crosstalk between adjacent actuators and for suppressing low-order eigenmodes is proposed. The strategy can be seen as extending Saint-Venant's principle beyond the static case, for small local families of actuators. An analytic model is presented, from which we show the feasibility of the local control. Also, we demonstrate how eigenmodes and eigenfrequencies are affected by mirror parameters, such as thickness, diameter, Young's modulus, Poisson's ratio, and density. This analysis is used to evaluate the design strategy for a large deformable mirror, and how many actuators are needed within a family.
We study a concept for a low-cost, large deformable mirror for an Extremely Large Telescope. The use of inexpensive voice-coil actuators leads to a poorly damped faceplate, with many modes within the desired control bandwidth. A control architecture, including rate and position feedback to add damping and stiness, for the faceplate has been presented in our previous papers. An innovative local control scheme which decouples adjacent actuators and suppresses low-order eigenmodes is a key feature in our controller. Here, we present an integrated
model of a partially illuminated large deformable mirror in an experimental laboratory setup with a limited amount of actuators. From the model, conclusions are drawn regarding the number of actuators needed to identify the key features, such as local control performance, dynamic range, and controllability and robustness of the deformable mirror.
Large (>1m) deformable mirrors with hundreds or thousands of actuators are attractive for extremely large
telescopes. Use of force actuators coupled to the mirror via suction cups, and electret microphones for position
sensing, has the potential of substantially reducing costs. However, a mirror controlled with force actuators
will have many structural resonances within the desired system bandwidth, shifting the emphasis somewhat
of the control aspects. Local velocity and position loop for each actuator can add significant damping, but
gives poor performance at high spatial frequencies. We therefore introduce a novel control strategy with many
parallel "actuator families", each controlled by
single-input-single-output controllers. This family approach
provides performance close to that of global control, but without the accompanying robustness challenges. Using
a complete simulation model of a representative large deformable mirror, we demonstrate feasibility of the
This paper describes the challenges of non-ideal actuators and sensors. The results presented give an understanding
of the required actuator bandwidth and the effects of the sensors dynamics. The conclusion is that the
introduction of actuator and sensor dynamics does not limit the control system of the deformable mirror.
Planned Extremely Large Telescopes will rely on availability of large Deformable Mirror in the 2-3m class. Design
and construction of such mirrors are challenging and call for powerful simulation tools. We present an evaluation
model which is used to study performance of a large deformable mirror for three actuator topologies.
Back sensors topologies are discussed from the point of view of sensor noise propagation. Two methods for
estimating the deflection at the actuator locations on the basis of sensor signal are presented and compared
regarding the computational power needed.