30.03.2014 Views

Wavefront Sensor Design - Center for Adaptive Optics

Wavefront Sensor Design - Center for Adaptive Optics

Wavefront Sensor Design - Center for Adaptive Optics

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Wavefront</strong> sensing <strong>for</strong><br />

adaptive optics<br />

Richard Dekany<br />

Caltech Optical Observatories<br />

2009


• Thanks to:<br />

Acknowledgments<br />

– Marcos van Dam – original screenplay<br />

– Brian Bauman – adapted screenplay<br />

– Contributors<br />

• Richard Lane, Lisa Poyneer, Gary Chanan,<br />

Jerry Nelson, and others;<br />

Elements of this presentation were prepared under the auspices of the U.S. Department of Energy<br />

by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.


Outline<br />

• <strong>Wavefront</strong> sensing<br />

– Shack-Hartmann<br />

• Hartmann test<br />

• History of Shack-Hartmann WFS<br />

• Centroid estimation<br />

• SH WFS design<br />

– Pyramid<br />

– Curvature<br />

• Not covered here<br />

– Shearing interferometers<br />

– Direct phase/interferometric measurements<br />

– Phase retrieval<br />

– Double curvature sensing<br />

– PIGS, SPLASH, L-O PWFS, etc.<br />

• Topics are covered with a bit of the Optical<br />

Engineer’s point of view<br />

– “the Boss : Springsteen :: the Optical Engineer : Bauman”


Hartmann test<br />

• Be<strong>for</strong>e there was Shack, there was Hartmann (1900,<br />

1904 (in German)).<br />

• Used <strong>for</strong> testing figure of optics<br />

reference spots<br />

wavefront screen Detector (film/CCD)


Hartmann test<br />

• Be<strong>for</strong>e there was Shack, there was Hartmann<br />

(1900, 1904 (in German)).<br />

• Used <strong>for</strong> testing figure of optics<br />

z<br />

Δy<br />

Slope= Δy/z<br />

wavefront screen Detector (film/CCD)


Hartmann masks<br />

• Originally, polar array of holes to sample aperture; suffered<br />

from sparse sampling at outer edge (or over-dense sampling near<br />

center), radial patterns hard to see<br />

• Holes sized according to power, diffraction size<br />

• Helical pattern <strong>for</strong> testing Lick 3-meter mirror (Mayall &<br />

Vasilevskis, 1960)<br />

• Square grid was introduced in early ’70’s<br />

Malacara


Hartmann test<br />

• Be<strong>for</strong>e there was Shack, there was Hartmann<br />

(1900, 1904 (in German)).<br />

• Used <strong>for</strong> testing figure of optics<br />

z<br />

Δy<br />

slope= Δy/z<br />

wavefront screen Detector (film/CCD)


Shack-Hartmann test<br />

• Be<strong>for</strong>e there was Shack, there was<br />

Hartmann (1900, 1904 (in German)).<br />

z<br />

Δy<br />

Slope= Δy/z<br />

wavefront lenslets Detector (film/CCD)


History of Shack’s/Platt’s modifications<br />

• Original application was <strong>for</strong> measuring atmospheric<br />

distortions to deconvolve images of satellites<br />

• Replaced holes with lenslets to maximize throughput<br />

(application was measuring atmosphere-distorted<br />

wavefronts) and to reduce spot size<br />

– Made lenslets by polishing glass with 150-mm-long cylindrical nylon<br />

mandrel sliding on steel shaft until cylindrical divot was desired<br />

width (1 mm), then shifted the mandrel by the lenslet pitch<br />

– Cylinders polished to λ/20<br />

– Used glass cylinders as master in molding process<br />

– Plexiglass was molded between crossed cylindrical sets to <strong>for</strong>m<br />

spherical lenslets on square grid<br />

• Molds <strong>for</strong>med in Platt’s kitchen oven, softening<br />

plexiglass until it slumped between masters; trimmed<br />

plexiglass with electric kitchen knife<br />

Platt and Shack, J. Refractive Surgery, vol. 17, p. S573 –S577 (2001)


Shack-Hartmann spots


Shack-Hartmann spots<br />

45-degree astigmatism


Lenslets-to-CCD: The “dot relay”<br />

• Lenslets available generally only in fixed sizes; CCD pixels<br />

available in fixed sizes; but can adapt lenslet pitch to CCD pixel<br />

pitch via relay, often two-lens “4-f” telescope <strong>for</strong> low<br />

aberrations/geometric distortion;<br />

– Relay often necessary anyway because of short lenslet focal lengths<br />

and clearance issues<br />

– Modeled as separate imaging system; “dots” are objects; entrance<br />

pupil is at infinity (telecentric)<br />

• There is rarely any optical design advantage in modeling the<br />

lenslet array as such. Divide design into “be<strong>for</strong>e-lenslets” and<br />

“after-lenslets”.<br />

Dot<br />

plane<br />

Good place <strong>for</strong> filters<br />

CCD<br />

plane


Dot relay design considerations<br />

• Once wavefront is sampled by lenslets, the game is over; the<br />

wavefront measurement has been made.<br />

– The relay need only to not blur spot too much, and not introduce<br />

unacceptable distortion, which is interpreted as a wavefront error by the<br />

WFS.<br />

– f# is generally slow (e.g., f/20 to f/50), so re-imaging dots is not difficult.<br />

• For quad-cell systems, spacing between lenses needs to be<br />

perturbed from “4f” otherwise there is no dot magnification<br />

adjustment possible<br />

• There is a pupil wrt imaging the dots—this is a good place <strong>for</strong><br />

filters as it after the measurement of the wavefront and<br />

affects all subapertures equally.<br />

Dot<br />

plane<br />

Good place<br />

<strong>for</strong> filters<br />

CCD<br />

plane


Spot size/subaperture size<br />

• Spot size<br />

– ~λ/d, where d is the subaperture size.<br />

– Typically, d is on the order of the actuator pitch (often<br />

exactly the actuator pitch—Fried geometry), and is on the<br />

order of r 0 to a few r 0 at the science wavelength.<br />

– For λ≈0.8μ and d=40 cm, the spot size is approximately 0.8<br />

μ/0.4m=2 μrad=0.4 arcsec<br />

• Spot size trade-off<br />

– bigger subapertures => more light, better SNR in centroid<br />

measurement, but poorer fit to wavefront.<br />

– If subapertures are too small, then spot size increases due<br />

to diffraction, degrades spot centroid estimate (proportional<br />

to spot size)<br />

• In example above, 5% spot-size displacement => 0.1<br />

μrad => 0.1 μrad * 0.4 m = 40 nm tilt across<br />

subaperture


Plate scale<br />

• “Plate scale” refers to the size in arcsec (on the<br />

sky) of a pixel<br />

– For SH WFS, this is often ~1-2 arcsec/pixel.<br />

• Plate scale trade-off<br />

– Bigger pixels (in arcsec)<br />

• More dynamic range w/o “crosstalk” between subapertures,<br />

but lets in more sky background photons (noise)<br />

– More pixels per subaperture<br />

• Increases linearity, at cost of more read noise and dark<br />

current in slope measurement


Typical vision science WFS<br />

Lenslets<br />

CCD<br />

Many pixels per subaperture


Typical Astronomy WFS<br />

Former Keck AO WFS sensor<br />

2 mm<br />

21 μ pixels<br />

3x3 pixels/subap<br />

200 µ<br />

lenslets<br />

relay lens<br />

CCD<br />

3.15 × reduction<br />

Low- or zero-noise detectors are starting to<br />

change astronomical WFS thinking (more pixels)


Centroiding<br />

• Once you have generated spots, how do you<br />

determine their positions?<br />

– The per<strong>for</strong>mance of the Shack-Hartmann<br />

sensor (the quality of the wavefront estimate)<br />

depends on how well the displacement of the<br />

spot is estimated.<br />

• The displacement is usually estimated using<br />

the centroid (center-of-mass) estimator.<br />

– This is the optimal estimator <strong>for</strong> the case<br />

where the spot is Gaussian distributed and the<br />

noise is Poisson.


Centroiding noise<br />

• Due to read noise and dark current, all pixels are<br />

noisy.<br />

• Pixels far from the center of the subaperture are<br />

multiplied by a large number:<br />

• The pixels with the most “leverage” on the<br />

centroid estimate are the dimmest (there<strong>for</strong>e,<br />

the pixels with the least in<strong>for</strong>mation), and… there<br />

are lots of dim pixels<br />

• The more pixels you have, the noisier the centroid<br />

estimate!


Weighted centroid<br />

• The noise can be reduced by windowing<br />

the centroid:


Weighted centroid<br />

• Can use a square window, a circular window:<br />

• Or better still, a tapered window


Correlation (matched filtering)<br />

• Find the displacement of the image<br />

that gives the maximum correlation:<br />

=<br />

!


Correlation (matched filtering)<br />

• Noise is independent of number of<br />

pixels<br />

• Much better noise per<strong>for</strong>mance <strong>for</strong><br />

many pixels<br />

• Estimate is independent of uni<strong>for</strong>m<br />

background errors<br />

• Estimate is relatively insensitive to<br />

assumed image<br />

• Computationally more expensive<br />

– This used to matter more.


Quad cells<br />

• In astronomy, wavefront slope<br />

measurements are often made using a quad<br />

cell (2x2 pixels)<br />

• Quad cells are faster to read and to<br />

compute the centroid and less sensitive to<br />

noise


Quad cells<br />

• The estimated centroid position is linear with displacement<br />

only over a small region (small dynamic range)<br />

• Sensitivity is proportional to spot size<br />

Estimated centroid position vs. displacement <strong>for</strong><br />

different spot sizes<br />

Centroid<br />

estimated<br />

position<br />

Displacement


Denominator-free centroiding<br />

• When the photon flux is very low,<br />

noise in the denominator increases the<br />

centroid error variance<br />

• Centroid error can be reduced by<br />

using the average value of the<br />

denominator


Laser guide elongation<br />

• Shack-Hartmann subapertures see a line not a spot<br />

– Length Θ ≈ t*s / h 2 ; where t is Na layer or range gate thickness<br />

• Depends on projector offset, not viewing direction


LGS elongation at Keck II<br />

Laser projected from right


A possible mitigation <strong>for</strong> LGS elongation<br />

• Radial <strong>for</strong>mat CCD<br />

– A specially oriented<br />

array of CCDs on one<br />

chip<br />

• Arrange pixels to be<br />

at same orientation<br />

as spots<br />

• Currently hardware<br />

testing this design<br />

<strong>for</strong> TMT<br />

laser


Dynamic refocusing <strong>for</strong> pulsed lasers<br />

• Powered mirror on<br />

mechanical resonator<br />

(U of A)<br />

• Segmented MEMS,<br />

one segment per<br />

subaperture (Bauman;<br />

Baranec)<br />

• Rotating phase plates<br />

(e.g., Alvarez lens)<br />

(Bauman)


Problems with SH WFS<br />

• Spot size is large (~ λ/d)<br />

• Crucial measurement is made at junction<br />

between pixel boundaries, which are indistinct<br />

(has been reported at ~1/3 pixel charge<br />

diffusion)<br />

• Worst of all worlds: photons near knife-edge<br />

generate all the noise and none of the signal!


Foucault knife-edge test<br />

• Foucault (1858, 1859 (in French))<br />

Knife-edge test <strong>for</strong> perfect lens<br />

(top), and one with spherical<br />

aberration (bottom). At right are<br />

observer views of pupil in each<br />

case.<br />

An irregular mirror<br />

tested with knifeedge<br />

test


Foucault test with mirror


Pyramid WFS<br />

• Simultaneous implementation of 4 Foucault knife-edge<br />

measurements<br />

– SH WFS divides aperture into subapertures (via lenslets), then field into<br />

quadrants (via pixels)<br />

– PWFS does in reverse order: pyramid divides field into quadrants (via<br />

pyramid) then aperture into subapertures (via pixels)<br />

pyramid<br />

field lens<br />

pupils with CCD<br />

pixels demarking<br />

subapertures<br />

incoming<br />

beam<br />

CCD at<br />

pupil plane<br />

image plane


Pyramids…<br />

• … are naturally quite small<br />

– Size of pyramid ~ n * (λ*f#), where n is # of subapertures<br />

(natural spatial filter)<br />

• … have tight fabrication tolerances:<br />

– Edge precision is a fraction of the full-aperture diffraction<br />

spot size (e.g., λ=1μ, f/15 ⇒ sub-micron precision required.<br />

Can make beam slower to relax edge requirements, but at<br />

cost of length.<br />

• … can be made of glass, using cemented facets.<br />

– It is difficult to make sharp edges<br />

• … can be based on lenslets (coming up)<br />

– Advantage: if edges of pyramid can be sharp, then centroid<br />

measurement can be quite precise; indistinct CCD pixel<br />

boundaries relegated to subaperture division—not crucial<br />

• … can transmogrify<br />

– As wavefront slopes becomes small, the PWFS becomes a<br />

direct phase measuring device


SH WFS vs. PWFS<br />

• Geometrically, identical—just remapping<br />

of pixels.<br />

• Diffractive advantage appears in high-<br />

Strehl regime.


Pyramid wave-front sensor<br />

non-linearity<br />

• When the aberrations are large (e.g.,<br />

defocus below), the pyramid sensor is very<br />

non-linear (reaches saturation).<br />

4 pupil images x- and y-slopes estimates.


Modulation of pyramid sensor<br />

Without modulation:<br />

Linear over spot width<br />

With modulation:<br />

Linear over modulation width


Another pyramid implementation<br />

Pyramid + lens = 2x2 lenslet array<br />

•Bauman (2003 (in English))<br />

•Lenslets are<br />

inexpensive and easily<br />

replicated.<br />

•The right<br />

manufacturing<br />

technique produces<br />

sharp boundaries<br />

between lenslets (where<br />

all the action is).<br />

pyramid<br />

field lens<br />

lenslets<br />

Bauman dissertation


• Brightening of rim is<br />

real effect—PWFS is<br />

not quite a slope<br />

detector, but a<br />

derivative detector<br />

(effect also seen in<br />

knife-edge tests)<br />

• There is a large<br />

derivative (in amplitude)<br />

at the edge of an<br />

aperture<br />

• Pupils should not be too<br />

close to avoid<br />

contamination between<br />

pupil images<br />

Image of PWFS<br />

Johnson, et al., 2006


Why is a PWFS/Foucault test a<br />

slope sensor?<br />

• Use Fourier optics!<br />

focal plane<br />

mask<br />

W (x,y)<br />

p<br />

CCD<br />

pupil<br />

relay<br />

lenses<br />

pupil


1/x is poor man’s approximation<br />

to δ (1) (x)


How to convert SHWFS to PWFS<br />

• This works only<br />

when the # of<br />

subapertures is<br />

approximately<br />

equal to the # of<br />

pixels per<br />

subaperture<br />

• Otherwise, other<br />

optical changes<br />

need to be made<br />

image<br />

plane<br />

"pupilet plane"<br />

lenslet<br />

array<br />

collimating<br />

lens<br />

demagnified<br />

"pupilet plane"<br />

relay<br />

lenses<br />

WFS<br />

"dot plane"<br />

lenslet<br />

array<br />

demagnified<br />

"dot plane"<br />

relay<br />

lenses<br />

Figure 3- 17: Conversion of a SH WFS to a PWFS. Top figure: Converging<br />

light from the left comes to a focus and is then collimated by a collimating<br />

lens. The collimating lens creates a pupil downstream, where the lenslet<br />

array is placed. The lenslets produce a series of images or dots at the focal<br />

plane of the lenslets or “dot plane”. Subsequent relay optics scale the dots<br />

as appropriate <strong>for</strong> the WFS CCD. For clarity, light from only one dot is<br />

shown after the dot plane. Bottom figure: to convert SH WFS to PWFS,<br />

remove collimating lens and translate lenslet array, relay optics, and WFS<br />

CCD upstream until the lenslet array is at the focus of the incoming beam.<br />

The lenslets now produce “pupilets” at the lenslet focal plane, i.e., where the<br />

dots were in the top figure. Thus, the relay optics will relay the pupilets to<br />

the WFS CCD.<br />

WFS


Curvature sensing<br />

-z<br />

Image 2<br />

Aperture<br />

Wave-front at aperture<br />

z<br />

Image 1


Curvature sensing<br />

• Developed by Roddier <strong>for</strong> AO<br />

in 1988.<br />

• Linear relationship between<br />

the curvature in the aperture<br />

and the normalized intensity<br />

difference:<br />

• Broadband light helps reduce<br />

diffraction effects.<br />

• Tends to be used in lowerorder<br />

systems (i.e., fewer<br />

subapertures/actuators,<br />

because of higher 1 error<br />

propagation<br />

1<br />

Still an area of active research


Curvature sensing<br />

• Using the irradiance transport<br />

equation,<br />

where I is the intensity, W is the wavefront<br />

and z is the direction of propagation,<br />

we obtain a linear, first-order<br />

approximation, which is a Poisson equation<br />

with Neumann boundary conditions.


Solution at the boundary<br />

If the intensity is constant at the aperture,<br />

I<br />

I<br />

1<br />

1<br />

!<br />

+<br />

I<br />

I<br />

2<br />

2<br />

=<br />

H<br />

H<br />

( x<br />

( x<br />

!<br />

!<br />

R<br />

R<br />

!<br />

!<br />

zW<br />

zW<br />

x<br />

x<br />

)<br />

)<br />

!<br />

+<br />

H<br />

H<br />

( x<br />

( x<br />

!<br />

!<br />

R<br />

R<br />

+<br />

+<br />

zW<br />

zW<br />

x<br />

x<br />

)<br />

)<br />

I 1<br />

I 2<br />

I 1 - I 2


Solution inside the boundary<br />

I<br />

I<br />

1<br />

1<br />

!<br />

+<br />

I<br />

I<br />

2<br />

2<br />

= ! z(<br />

W xx<br />

+ W yy<br />

)<br />

Curvature<br />

• There is a linear relationship between<br />

the signal and the curvature<br />

• The sensor is more sensitive <strong>for</strong> large<br />

effective propagation distances


Curvature sensing<br />

As the propagation distance,<br />

z, increases, sensitivity<br />

increases.<br />

• Spatial resolution<br />

decreases.<br />

• Diffraction effects<br />

increase.<br />

• The relationship between<br />

the signal, (I 1 - I 2 )/(I 1 + I 2 )<br />

and the curvature, W xx +<br />

W yy , becomes non-linear


Curvature sensing<br />

• Practical implementation uses a<br />

variable curvature mirror (to obtain<br />

images below and above the aperture)<br />

and a single detector.


Curvature sensor subapertures<br />

• Measure intensity in each subaperture with an<br />

avalanche photo-diode (APD)<br />

• Detect individual photons – no read noise

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!