Introduction to Adaptive Optics - Durham Universitycommunity.dur.ac.uk/t.j.morris/AstInst4f.pdf ·...
Transcript of Introduction to Adaptive Optics - Durham Universitycommunity.dur.ac.uk/t.j.morris/AstInst4f.pdf ·...
Introduction to Adaptive Optics
Tim Morris
Contents
• Definitions and introduction
• Atmospheric turbulence
• Components of an AO system
• Wavefront Sensing
• Wavefront Correction
• Turbulence Conjugation
• Laser Beacons
• AO Modelling
Adaptive Optics (AO)Real-time correction of wavefront distortion
• The diffraction limit of an 10m telescope in in the visible is approximately 0.01” FWHM
• At the very best astronomical sites in the world, you’ll very rarely see images much better than 0.4” FWHM.
Why?!?
• Atmospheric turbulence distorts stellar wavefronts– Turbulence results in blurred images
• Two solutions:– Put your telescope in space
• Limited to a small mirror
• Expensive
– Correct for the atmospheric distortion• ADAPTIVE OPTICS!
Know Your Enemy
AO Conference
(me)
Enemies
The Atmosphere
What does turbulence look like?
• Kinetic energy in large scale turbulence cascades to smaller scales
• Inertial interval – Inner scale l0 2mm. Outer scale L010 to 100 m
• Turbulence distributed within discrete layers
• The strength of these layers is described by a refractive index structure function:
The AtmosphereKolmogorov model of turbulence
J. Vernin, Universite de Nice.
Cerro Pachon for Gemini IGPO
2
2 2 / 3
( ( ) ( )
( )
n
n n
D r n r n
D r C r
• Strength of turbulence can be described by a single parameter, r0 , Fried’s parameter
• Fried’s parameter is the diameter of a circular aperture over which the wavefront phase variance equals 1 rad2
Isoplanatic angle, temporal variation
• Angle over which wavefront distortions are essentially
the same:83
32 5
2 5 / 3
0
22.91 sec ( )nC h h dh
• It is possible to perform a similar turbulence weighted
integral of transverse wind speed in order to derive an
effective wind speed and approximate timescale of
seeing
• τ0 is the characteristic timescale of turbulence
• Note the importance of Cn2(h) in both cases
Atmospheric Seeing - Summary
Dependence on Wavelength
56
0 r 56
0 56
0
=0.55m =1.6m =2.2m
r0 10cm 36 53
0 10ms 36ms 53ms
0 5’’ 18’’ 27’’
Image quality
• Image quality is determined by the wavefront variance across the telescope pupil
• The above equation gives the phase variance over a telescope of diameter DT
• A phase variance of less than ~ 0.2 gives diffraction limited performance
• There are 3 regimes
– DT < r0 Diffraction dominates
– DT ~ r0 - 4r0 Wavefront tilt (image motion) dominates
– DT >> r0 Speckle (multiple tilts across the telescope aperture) dominates
4m telescope:
D/ r0(500nm)=20
D/ r0(2.2 m)=3.5
2
0
2
35
030.1 radiansr
DT
Strehl ratio
• There are two components of the PSF for 2 < 2 radians2
• So ‘width’ of the image is not a useful parameter, use height of PSF:
• Strehl ratio:
– For small 2 : R ~ exp (-2)
– Note that 2 should be expressed in radians2
R = Peak intensity in a (un)corrected image
Peak intensity in a diffraction limited image
MARTINI
WHT, K-band
Uncorrected
0.49” FWHM
Corrected
0.20” FWHM
AO Performance
• RMS error terms in AO add in quadrature
– Easiest to perform as nanometers RMS wavefront error
• Many sources of error
– Temporal
– DM sampling
– Anisoplanatism
– WFS noise
• Once added in quadrature, the RMS wavefront error can be converted to a Strehl Ratio
Wavefronts
• Zernike polynomials are normally used to
describe the actual shape of an incoming
wavefront
• Any wavefront can be described as a
superposition of zernike polynomials
Atmospheric Wavefront Variance after Removal of Zernike Polynomials
j n m Zernike Polynomial Name Resid. Var. (rad2)
1 0 0 1 Constant 1.030 (D/r0)5/3
2 1 1 Tilt 0.582 (D/r0)5/3
3 1 1 Tilt 0.134 (D/r0)5/3
4 2 0 Defocus 0.111 (D/r0)5/3
5 2 2 Astigmatism 0.0880 (D/r0)5/3
6 2 2 Astigmatism 0.0648 (D/r0)5/3
7 3 1 Coma 0.0587 (D/r0)5/3
8 3 1 Coma 0.0525 (D/r0)5/3
9 3 3 0.0463 (D/r0)5/3
10 3 3 0.0401 (D/r0)5/3
23 2 1
2 cos
2 sin
28 3 2 sin
26 sin 2
26 cos2
28 3 2 cos
38 sin3
38 cos3
Components of an AO System
High order AO architecture
• Wavefront controller
– Typically a deformable mirror (DM)
– May not be optically conjugate to an image of the primary
• Wavefront sensor (WFS)
– Shack Hartmann (WFS) or Curvature Sensor (CS)
• Beamsplitter
– Dichroic, multi-dichroic, intensity, spatial or combination
• Controller
– Typically multi-processor or multi-DSP
• Interfaces
– Can be complex and include removal of non-common path errors to science instrumentation (hence an interface to science data path)
• Laser beacons
• Multi-conjugate AO: many beacons, DMs
4/12/2010 16
Astronomical Adaptive Optics
Telescope
Science
target
Laser
Natural
Guide
Star
Corrected
Science
focus
dichroic
beamsplitter
IR
light
Visible
light
Wave-
front
Sensor
Adaptive
Mirror
Control
System
wavefront
information
control
signals
atmospheric
turbulence
*
*
Uncorrected
image
Corrected
Image
Uncorrected
wavefront
Corrected
wavefront
Correcting the fluctuating
aberrations caused by
atmospheric turbulence
above ground-based
optical and near-infrared
telescopes.
Wavefront Sensing
Wavefront Sensing
• Types of Adaptive Optics Wavefront Sensor
(WFS)
– Shack-Hartmann WFS
– Curvature Sensor
– Interferometers
– Others
• Performance comparison of Shack-
Hartmann (SH) and Curvature Sensor (CS)
Shack-Hartmann
Wavefront Sensor (WFS)
Microlens Array
Wavefront
Detector
Each xy offsetmeasures the local wavefrontslope across thecorrespondinglenslet.
Curvature
Wavefront SensorFocal PlaneInput Wavefront
Sensing Planes
Wavefront Sensors and Detectors
• The curvature sensor minimises the number
of pixels required to remove a given
wavefront variance
– the use of noiseless fibre-coupled avalanche
photo-diodes is therefore feasible
• Shack-Hartmann requires more pixels so a
CCD is normally employed
– low read-noise multi-port specialised devices
Comparison of SH and CS
(~0.5” seeing)(Pete Doel, University of Durham)
Comparison of SH and CS
(~1” seeing)(Pete Doel, University of Durham)
Wavefront Control
Wavefront Control
• Deformable Mirror (DM) types:
– Continuous
– Bimorph
– Segmented
• Hysteresis
• System order
Types of Adaptive Mirror(J.C.Dainty, Imperial College)
Deformable Mirror
Flexible continuous phase sheet
reflectivesurface
Actuators:typically PZT or PMNthrow: 2-20 microns
Minimum physicalactuator separation ~ 1mm
Fitting error:2
f =k (rs/r0)5/3 rad2
rs= projected actuatorseparation on sky
k = fitting coefficientfor DM type.(continuous facesheet: 0.35-0.4)
One type of Deformable Mirror (DM):
Continuous Face-sheet
Deformable Mirror
Bimorph Mirror(J.C.Dainty, Imperial College)
Bimorph
Deformable Mirror
The ELECTRA Segmented Adaptive Mirror
(76 tip-tilt-piston segments)
built by ThermoTrex, San Diego
228 degreeof freedomadaptivemirror
Wavefront Fitting Error
Comparison
0.5 1 1.5 2 2.5 3 3.5 4
Wavelength microns
0.2
0.4
0.6
0.8
1
ev
it
al
eR
la
rt
ne
Cy
ti
sn
et
nI
Comparative AO Technology Limits for WHT r0 =0.155 ,t0 =1000 ,t=1,d0 =1000
Segmented 10, 0%
Segmented 0%
Segmented 0.2 % hysteresis
Continuous face sheet 8,0.4
Bimorph n=7
Bimorph n=5
tip - tilt
Actuator Hysteresis
PMN
(electrostrictive)
Low (<2%) hysteresis
at 20o C;
High (~40%) at 0o C
Low drive voltage
(<100V)
soft PZT ~15% but
temp. stable
low drive
voltage
hard PZT ~2%, stable high drive voltage
(1000V)
Hysteresis
• Effect of high hysteresis:
– Continuous mirror: 2-3 times more WFS samples
required
– Segmented mirror: makes piston control hard
• Solutions:
– low hysteresis actuators
– linearise with motion sensor (e.g., strain gauge)
– linearise with figure sensor
– Example: ELECTRA: has strain gauges (with
temperature compensation) which reduce hysteresis
from ~15% to <0.1%
Sky CoverageThe big problem with AO
You can’t observe off-axis!
• Angle over which wavefront distortions are essentially
the same:83
32 5
2 5 / 3
0
22.91 sec ( )nC h h dh
– This is a very small angle
~5” in the visible
– It means that if you look at
an object that’s a large
angular distance away from
your guide star, you get
poor correction!
Guide Star Availability.
All sky.Model: D. Simons, Gemini
10
12
14
16
18
20
R mag
0
20
40
60
radius (arcsec)
0
0.25
0.5
0.75
1
prob >=1 stars
10
12
14
16
18
20
R mag
Guide Star Availability.
(Galactic Latitude > 30 degrees)Model: D. Simons, Gemini
10
12
14
16
18
20
R mag
0
20
40
60
radius (arcsec)
0
0.2
0.4
0.6
0.8
prob >=1 stars
10
12
14
16
18
20
R mag
NGS sky coveragemodel for ING by Remko Stuik, Leiden Observatory
Remember the Enemy?
The Atmosphere
Physicist+Enemy
leads to…
Complicated plan to defeat enemy
Part of plan that requires
a really big laser
Adaptive Optics is no different…
Laser Guide StarsCreating an artificial wavefront reference
Really Big
LaserPhysicist
GLAS LGS AO system commissioning 2007
Laser Guide Star• Purpose of a laser guide star (LGS) is to increase the sky
coverage by creating a bright wavefront reference
anywhere in the sky to replace the natural guide star
• Two types of LGS:
– Rayleigh (Green or UV)
• uses Rayleigh backscatter
• beacon height up to ~20km
• requires time-gating to set beacon height
– Sodium D (Orange)
• uses excitation of mesospheric sodium atoms
• beacon height 80-90km
• no time-gating required
• tuned to sodium D line at 589nm
Comparison of Rayleigh backscatter
and sodium-resonance backscatter.
(Courtesy of MIT Lincoln Lab.)
Rayleigh and Sodium Guide Stars at La Palma(IC Applied Optics Group + Tom Gregory, ING)
Durham’s Rayleigh Laser Guide Star
Other LGS Systems
KeckSubaru (Keck)
Other LGS Systems
VLT
WHT
US Military
LGS sky coveragemodel for ING by Remko Stuik, Leiden Observatory
LGS sky coveragemodel for ING by Remko Stuik, Leiden Observatory
Laser Beacon Limitations
• Tilt reciprocity
– no tip-tilt signal from laser beacons
• must use a natural guide star
– focus is complicated for sodium beacons
• low frequency atmospheric focus may be masked by
changes in effective beacon height
• Focus Anisoplanatism (cone effect)
• Sodium layer saturation
• Safety/site issues
Tilt Reciprocity(J.C.Dainty, Imperial College)
Multiple Tip/Tilt NGS’s?
– Consider a turbulence profile with focus aberrations at two ranges (blue)
– LGS measurements (yellow)cannot determine range of the aberration
• Tip/tilt information lost
• Equal focus measurement from each LGS, regardless of aberration range
– Tip/tilt NGS measurements can determine range from the differential tilt between stars
– Three tip/tilt NGS’s needed for all three quadratic modes
– Alternate approaches: Rayleigh LGS’s, or a solution to the LGS tilt indeterminacy problem
fr)=ar2
fr)=a(cr+d)2
=ac2r2+2acdr+ad2
~ ac2r2
After tilt removal
Angular and Focal
Anisoplanatism
Focal Anisoplanatism
0 0.5 1 1.5 2 2.5
Wavelength (microns)
0
0.2
0.4
0.6
0.8
Strehl Ratio
Effect of Na beacon Focal Anisoplanatism (Cone Effect)
d0=3m
d0=5m
d0=8m
Schemes for the use of multiple
laser beacons(J.C.Dainty, Imperial College)
Turbulence Conjugation(if normal AO is just a bit too easy)
Multiple Conjugate AO
• Putting a second
DM in a plane
conjugated to a
higher layer of
turbulence allows
off-axis correction
• Requires multiple
guide stars
Multi-Conjugate AO
(MCAO)Multiple LGS, Multiple DM: Wide
corrected FOV
Conventional AO MCAONo AO
Courtesy of GEMINI
Strehl Uniformity vs. FOV
• 0 degree zenith angle, 50% Cerro Pachon Turbulence Profile• 5 LGS, 16 by 16 subapertures, 3 DM’s• No WFS noise or servo lag
Courtesy of GEMINI
Tip/Tilt
Focus+Astig.
Cubic and Above
MCAO Control Loop Architecture
Focus
Adjustment
Integrator
SM
Tip/Tilt
Offload
Low-order
Offload to TCS
Integrator
Boresight
Adjustments
BTO Tip/
Tilt Loops
Average
Differential
LGS Tracking
(option)
(Option)
(Option)
Integrator
Boresight
Adjustment Primary
NGS WFS
LGS WFS’s
DM 1
DM 2
DM 3
Wave Front
Tomography
Auxiliary
NGS WFS’s
TTM
OIWFS
Blend
Blend
DMFS 2
DMFS 3
Blend
Blend
Piston and Waffle
Courtesy of GEMINI
AO Modelling(or AO on a budget)
AO Modelling
• Computer Modelling
– Required for performance prediction, instrumentation choices, instrumentation and AO systems engineering, detailed design.
– 8m Monte Carlo models using 10-12 processor Beowulf clusters are available, examples:
• ESO/RTN (LeLouarn et al)
• Durham (Wilson et al)
• Ellerbroek/Rigaut:
– Memory requirements scale as D4
– CPU requirements scale as D6
• D > L0 poses new challenges for optimisation of WFS sample rate and control law (both in performance model and implementation)
Durham 12-processor cluster simulations (Richard Wilson)
• DCAO I-band simulation (the full Monte Carlo):• D Tel diam (m) WFS order Tmx (s) Tloop (s)
4 8x8 18 0.14
8 16x16 140 0.52
12 24x24 847 1.4 (MOVIE)
16 32x32 4067 4.1– Tmx is the time taken to produce the poke/control matrix, which as expected goes as something like D4.
• Assuming that the wavefront reconstruction calculation does not take over as the slowest component (ie. we use sparse matrix techniques), then we can project the timings to higher orders assuming that Tloop goes as D2 and Tmx as D4: Projecting from the 24x24 case gives:
32 64x64 42831 (12 hours) 9.9
64 128x128 685296 (190 hours) 39.6
128 256x256 107 (3045 hours) 158.4 [ELT]
• 5000 loops required for 10 seconds of seeing
• Need a factor of ~100 speedup.– Assuming better parallelisation, this could be accomplished with order of magnitude larger
cluster of up-to-date CPUs, and hardware acceleration.
Durham 12-
processor cluster
simulations
(Richard Wilson)
• 24 x 24 WFS
• r0=20cm at V
• Science at 1μ.
• Cn2 is just 2 layers
(0km, 5km)
• 500Hz simulated sample rate
• Top left: 24x24 WFS
• Top right: phase map at science pupil
• Bottom left + right: science PSF at 2 field points: 30 arcsec apart.
Does it actually work on-sky?
AO Scientific Potential
0.1” slittip/tilt simulation of Galactic Center image(K) at CFHT
Actual AOimage
Doug Simons
Gemini
PUEO image: fov 10x10”, resolution~0.13”
NGC7469 - Starburst galaxy
Io imaged with Keck AO
INGRID J-band image of M15, 20s exposure, 20” diameter FOV, Open
loop FWHM ~0.45”, Closed loop ~0.2” (Moffat fit to PSF)
The GLAS LGS AO System
What does this mean to an
astronomer?
Observing with an AO System• Position of target in the sky?
– Nearer zenith is better (less atmosphere to correct)
• Is there a suitable guide star near your target?
• What wavelength do you want to observe in?– Longer is better for AO as turbulence is weaker
• What field of view do you require?
– Current facility-class AO systems are not multiconjugate
• What performance can you expect?
– Highly dependent on weather
• How long does it take to set-up the AO system?
• Will a variable PSF across the field affect your results?
• What is the throughput to the Science CCD with the AO system?
– Extra surfaces in the optical path lower efficiency
End