Distributed control in adaptive optics: Deformable mirror and turbulence modeling
Summary (2 min read)
1. INTRODUCTION
- Over the years, adaptive optics (AO) has grown from wild ideas to a proven technology that is indispensable for any large telescope.
- Standard control approaches for AO consist of operations that typically scale with the number of actuators squared, which means that the total computational complexity of the controller increases with the telescope diameter to the fourth power.
- Many results towards realizing this are already available,1,2 but few methods implemented for AO fully exploit all available wavefront information.
- Advances in the design of the DM are discussed in,4 while this paper focusses on the controller design, for which a fully distributed control approach will be introduced.
- 6–8 Often, the atmospheric disturbance is assumed to show a frozen flow characteristic, which makes it well predictable.
2. DISTRIBUTED CONTROL DESIGN FRAMEWORK
- The distributed controller framework consists of a grid of processors called Distributed Processing Units (DPU) that each control one actuator of the deformable mirror.
- These DPU’s have thus the same regular geometric grid layout as the actuators.
- Wavefront slope information extracted from each spot is known only to the four surrounding DPU’s.
- This links locality directly to scalability : if a system has a high locality then its behavior can be well described using a limited set of local information that does not grow with the dimensions of the system.
- When a computation requires not only information from the DPU’s connections, but also from a limited set (i.e. whose size is independent of the system dimensions) of other neighboring DPU’s, this can also be fit into the distributed framework.
3. MIRROR MODELING
- The DM for which the distributed controller is designed,4 is of the continuous face-sheet type.
- This section comprehends the DM modeling that is relevant for controller design and will be used to answer the following questions:.
- The general mirror design is depicted in figure 2 and consists of a thin deformable reflective face-sheet that is supported by electromagnetic push-pull actuators.
- They are connected to the reflective face-sheet with thin rods, Proc. of SPIE Vol.
- The implications of the result for a distributed controller design will be evaluated.
3.1. Actuator modeling and identification
- By applying a current through the coil which is situated around the magnet, this force is influenced, providing movement of the ferromagnetic core.
- This movement is transferred via a rod imposing the out-of plane displacements in the reflective deformable membrane.
- In the actuator design a match is made between the negative stiffness of the magnet and the positive stiffness of the membrane suspension.
- The actuator can be straightforwardly modeled as a linear mass-spring-damper system.
3.2. Modeling the deformable face-sheet
- It still has a considerable out-of-plane stiffness when and should be modeled as a thin plate.
- This magnitude decreases also very fast with the distance and is only slightly influenced by ka.
- Further, since both plate mass as well as actuator stiffness are added, the lowest eigenfrequency should not decrease much when the DM design is extended to larger sizes.
- It can be safely concluded that the eigenfrequencies of the DM will not affect the control system performance as long as they are damped sufficiently.
- Therefore, the system should have a suitable amount of electromagnetic or air damping to have a fast, but well damped step-response.
4. TURBULENCE MODELING
- Characterization and modeling of atmospheric turbulence has a long history, e.g. Fried15 in 1965.
- Many approaches towards realistic simulation models have since then been recorded.
- Models exploiting the fractal nature of turbulence11 seem better at this, and even speculate on the existence of good linear predictor models.
- On the other hand, continuing on results by11,18 the remainder of this section discusses a moving average predictor filter that is designed to fit within the distributed framework and can in the same framework be made adaptive.
4.1. A distributed moving average filter
- In open loop, 10000 samples were collected using a 127-spot hexagonal SHWFS.
- As the FIR filter is both local in time and in space, this suggests that the atmospheric disturbance has a high locality.
- Therefore, the next subsection discusses possibilities for making this filter adaptive.
4.2. Adaptive distributed predictor
- Adaptive linear filters have been around for a long time and their properties thoroughly studied.
- The filter coefficients of an adaptive filter are not fixed, but are updated each sample to either converge to their globally optimal values or to track changes in the statistical properties of the input signal.
- Note that all calculations of the filter update equations can be performed within the distributed framework.
- Results for both algorithms are shown in terms of J2 in figure 9 together with the corresponding results from the optimal static predictor and the random walk predictor.
5. WAVEFRONT RECONSTRUCTION
- The last, but probably most important component of the controller is the wavefront reconstructor that reconstructs the wavefront phase from the gradients measured by a SHWFS.
- Note here that the matrix GT G is singular due to the two unobservable modes of the SHWFS,5 which means that a pseudo-inverse21 must be employed to solve the system directly.
- Thus, two options can now be considered: do the reconstruction by a dedicated external processor or find a reconstruction procedure that does fit into the framework.
- This means that one iteration of SOR can be performed with O(nw) sequential steps.
- This is then also the case for the required processing speed of the DPU’s and the inter-DPU communication bandwidth.
Did you find this useful? Give us your feedback
Citations
38 citations
25 citations
Cites background from "Distributed control in adaptive opt..."
...The relation can then be inverted to yield: [ Fζ M ] = [ Ωζζ Uζ U ζ 0 ]−1 [ Hζ 0 ] = [ Km Kr ] Hζ , (2)...
[...]
...Substitution of (2) for Fζ into the force equilibrium in (3) yields Fa −CaHζ −KmHζ = 0 and thus the mirror deflection Hζ at the actuator positions is related to the actuator forces Fa via the influence matrix Bζ as:...
[...]
...The results from (2) can be substituted into the plate equation in (1) together with (4), yielding Hz = BzFa where Bz = (ΩzζKm+UzKr)Bζ ....
[...]
17 citations
17 citations
Cites background from "Distributed control in adaptive opt..."
...2004, 2005a,b, 2006), (Ellenbroek et al. 2006) proposed a distributed control framework for AO in which each actuator has a separate processor that can communicate with a few direct neighbors....
[...]
14 citations
References
2,549 citations
"Distributed control in adaptive opt..." refers background in this paper
...(20) As both Rk and ṽk converge to the best estimates for the covariance matrix R and correlation vector ṽ respectively for k → ∞, the adaptive filter should converge to the static solution as long as the step size α is chosen properly....
[...]
1,944 citations
911 citations
Additional excerpts
...h(F, z, ζ) = F 16πR { (1 − r(2))(1 − ρ(2)) + [r2 + ρ(2) − 2rρ cos(φ − ψ)] ln r(2) + ρ(2) − 2rρ cos(φ − ψ) 1 + r2ρ2 − 2rρ cos(φ − ψ) } , (5)...
[...]
Related Papers (5)
Frequently Asked Questions (8)
Q2. How much damping is required to suppress the undamped system?
The undamped system has a highly oscillating step response, requiring over 100 times more actuator damping than that estimated in section 3.1 to suppress.
Q3. What is the effect of the magnetic force on the ferromagnetic core?
The actuators consist of a closed magnetic circuit in which a permanent magnet provides a static magnetic force on a ferromagnetic core which is suspended in a membrane.
Q4. Why is the forward substitution problem written as a series of nf n?
But again due to the structure of A when the points in the phase grid are suitably ordered, this forward substitution problem can be written as a series of nf ≈ nw/2 elimination steps – where nw is the number of phase points over the diagonal of the grid.
Q5. What is the simplest way to model a thin plate?
(4)Although the face-sheet has only a slight thickness, it still has a considerable out-of-plane stiffness when and should be modeled as a thin plate.
Q6. What is the DM influence function of a mirror?
For the static case, all time-derivative terms in (8) are zero and the deflection h can be directly expressed in terms of the forces Fa, from which the mirror influence matrix B containing the DM influence functions in its columns can be derived:B = (K−1 + Ca)−1. (9)The influence function of a mirror’s central actuator has been plotted for four values of ka in figure 3.
Q7. What is the purpose of this paper?
This paper reports on work that has been done in a joint project aimed at designing a new AO system that has an extendible design, which means that the same design should be applicable when the number of actuators is increased.
Q8. What is the way to model the evolution of the disturbance?
Models based on Kolmogorov statistics such as16,17 work well, but apart from pure frozen flow behavior it appears difficult to model the evolution of the disturbance over time.