Date post: | 18-Jan-2016 |
Category: |
Documents |
Upload: | anabel-gibson |
View: | 215 times |
Download: | 0 times |
* Work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Security, LLC, Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344
Agenda
• Introduction to NIF and automatic alignment
• Control system architecture
• Coordination and scaling
— 3,800 closed loop adjustments using 12,000 devices
— Management of shared laser components
• Image processing
— Algorithm robustness in a laser environment
— Subpixel accuracy
— Reliability and off-normal image detection
NIF concentrates all the energy in a football stadium-sized facility into a mm3
Beampath Movie
In an analogy to baseball, this requirement is like hitting the strike zone with a pitch thrown from 350 miles away
Beams are aligned along NIF’s 500-meter path to focus on the mm-size target within 10 microns
Target Alignment Sensor
Alignment uses sensors and actuators within Line Replaceable Units
Symmetry & nomenclatureNational Ignition Facility
ArchiveVerify
ShotShot Preparation
Amplifier cooling Shot
Fire pre-amplifier
Archive
30-m
Automated Shot Cycle
• Input shot goals from laser physics model
• Perform automatic alignment
• Configure diagnostics and laser settings
• Conduct countdown (SW: 4-m, HW: 2-s)
• Assess shot outcome and archive data
20-m
Fire main amplifiers
2-s
4 hours
Shot
ICCS shot cycle automatically aligns, fires and diagnoses laser shots every 4 hours
Automatic Alignment
30-m
The automatic alignment system compensates for optical system drift
ISP cw @ISP 032603
-6.00E+00
-5.00E+00
-4.00E+00
-3.00E+00
-2.00E+00
-1.00E+00
0.00E+00
1.00E+00
2.00E+00
3.00E+00
4.00E+00
5.00E+00
0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00
Minutes from starting time
Po
intin
g e
rro
r (u
R)
x error
y error
Required Tolerance
Pointing error at the Input Sensor
5
4
3
2
1
0
-1
-2
-3
-4
-5
-6 0 100 200 300 400 500 600 700
Minutes from starting time
Micro radians
Requirements led to significant technical challenges
• Minimize operator effort— Situational awareness— Manual controls
• Deliver subpixel accuracies
• Algorithms tolerant to— Varying light levels— Laser diffraction effects— Equipment anomalies
• Reject off-normal images
• Rapidly align 192 beams— Parallel operation— Computational resources— Many devices are shared
Single operator oversees Beam Control
Replicated bundles are activated by starting new processes in the database
The bundle architecture assures full-scale performance of the alignment control system
SupervisoryLayer
Front-EndLayer
ReplicatedBundles
BundleCoordinator
AutomaticAlignment
Bundle 1
VideoSensors
Beam Control
ActuatorControls
BundleCoordinator
AutomaticAlignment
Bundle 24
VideoSensors
Beam Control
ActuatorControls
BundleCoordinator
AutomaticAlignment
Target
VideoSensors
Beam Control
ActuatorControls
Shot Sequencer
Process distribution example for beam control
Common Control System Bundle 11 Controls
ConfigurationLogging
ArchivingReservations
etc.
ShotDirector
Beam ControlSupervisor
AlignmentController
B11Actuators
VideoController
B11Sensors
B11Timing
LINUXCluster
CORBA
B11 Automatic Alignment
B11 Manager
The alignment system maintains the beam within the optical clear aperture
Spatial Filter(Low Pass)
PinholeCentering
PointingMotorized Mirrors
LaserBeam
Lens Beam Tube
Centering definition — Translates beam position without affecting pointing— Sensor configured to near-field mode
Pointing definition— Adjusts angle of propagation without affecting centering— Sensor configured to far-field mode
Generic control loop flow diagram
Done
Adjust Optic
Setup Reference
Locate Reference
Acquire Image
Setup Beam
Locate Beam
Acquire Image
Centering Alignment Principle
The image of the beam reticule (circle) is adjusted to the midpoint between the reference reticules (squares)
Beam Center
CCD Camera
Light Source
ImageAcquisition
ActuatorControls
Network
Automatic
Alignment
Reference Reticule Mask
Motorized Turning Mirrors
Beamline
Beam Reticule Mask
Fixed SamplingMirror
Firewire
Near-field camera image
Control
Far-field camera image
Pointing Alignment Principle
The source sub-image is aligned to the center of the pinhole shadow
Pinhole
CCD Camera
Image Acquisition
Network
Automatic Alignment
ActuatorControls
PinholeLens
Light Source Pinhole
Illuminator
Motorized Turning Mirrors
Firewire
Managers coordinate resources and activities
• Supervisor — Manual controls— Status
• LINUX Cluster— Image processing— Self-leveling
• Segment Manager— Alignment plans— Loop definitions— Sequence control
• Component Manager— Device sharing— Throughput optimization
Bxx Automatic Alignment
Video Sensors
Actuator Devices
Beam Control Supervisor
Segment Manager
Component Manager
Database
- Alignment Plans
- Segments, Loops
- Components
- Devices
LINUXCluster
NIF’s beamline architecture was designed to permit alignment of segments in parallel
Reference Adjust
Optic
Beam
1 minute
CheckSetup Setup LocateLocate
A Segment Manager
Alignment Plan 1
Control Loop 1Control Loop 2…
Loop Timing
PreamplifierSegment Manager
Main LaserSegment Manager
1 Pulsed Source
1 CWSource
1 CW Source
3CW Source
Camera
Camera
Camera Camera
Target AreaSegment Manager
A NIF Beamline
The Segment Manager executes alignment plans efficiently to achieve the required performance
• Component Managers— Orchestrate setting up laser configurations— Contain multiple device grouping called
mediated components
• Mediated components— Manages device positions— Can include other mediated components— Reservations lock the configuration
• Task queues— Requests wait until devices become available— Deadlocks eliminated by enforcing priority— Queue optimization actively minimizes
required movements
One out of every four devices in the alignment system are shared
The Component Manager optimizes task processing throughput
The main laser sensor is a mediated component
Image processing must always be reliable
• Precision and accuracy required to 0.3 pixels
• Robustness is challenged by— Gradient illumination— Noise— Diffraction effects— Defocus— Magnification
• Discard off-normal images— Extraneous blobs— Saturation— Clipping
… while still being robust to varying laser conditions !
Example of a Clipped Image
Many processing algorithms are used …
Crystal reflection
Gaussian Beam
Weighted Centroid
Glass Grid
Crosshair Grid
Crosshair Grid
Hough Transform
Fiber Source Dark Stop
Pinhole
Template Match
Reticule
Matched filter
Shot Mask
Spiral Search
Corner Cube
Images undergo three processing steps
Discard Off-normal Images
Calculate Position
Check Position Uncertainty
Good Bad
Image quality is measured by estimating algorithm uncertainty
• Uncertainty Definition
— Instability in the position determination
— Obtained by varying either thresholds or noise
• Threshold-based method (e.g. weighted centroid)
— Image processed with thresholds at 10 different levels
— Delivers subpixel accuracy as a byproduct
— Uncertainty: variability of the position estimates
• Noise-based method (e.g. matched filter)
— Monte Carlo model of known uncertainty is constructed for the image using prescribed amounts of noise
— Uncertainty: estimated from the model based on noise present in the input image
Low uncertainty confirms a high quality input image, with confidence that algorithm results can be trusted
Example of sub-pixel position estimation confirms accuracy is within tolerance
Gaussian Image
0 Distance (p) 76
Inte
nsity
110%
90%
Lineout
1 pixel (x)
Tolerancelimit
1 pixel (Y)
Centroid Plotted at Various Thresholds
Poor laser wavefront prohibits a successful alignment outcome in this uncertainty example
Defocused Gaussian Image
10%
90%
Lineout
5x5 (x,y) pixel grid
Centroid Plotted at Various Thresholds
Tolerancelimit
OUT OF TOLERANCE!OUT OF TOLERANCE!
High uncertainty halts the automatic alignment process
Architectural enhancements achieved the required performance
0
10
20
30
40
50
60
70
80
90
100
2000 2002 2004 2006 2008 2010 2012
4 beam sequentially aligned to Target Chamber
Year
Required alignment time
Target camera bottleneck
Alignment Time(Minutes)
Sequential Mediated
Full automation of all 192 beams is moving rapidly toward completion
0
500
1000
1500
2000
2500
3000
3500
4000
2000 2002 2004 2006 2008 2010 2012
1/3
of
1 b
eam
1/2
of
1 b
eam
4 b
eam
s
4 b
eam
s
2/3
of
8 b
eam
s
2/3
of
16 b
eam
s
2/3
of
96 b
eam
s
192
bea
ms
144
bea
ms
Year
Number Control Loops
We’ve exceeded performance goals with a automatic system that is robust and reliable
Remaining work extends automation of all 192 beams to the target for ignition experiments beginning in 2010
NIF Automatic Alignment
Video Sensors
Actuator Devices
LINUXCluster
Beam Control Supervisor
Segment Manager • Sequence
Component Manager
Database
- Alignment Plans
- Segments
- Loops
- Components
- Devices
Queues
Queues
Many processing algorithms are used …
Crystal Reflection
a) Weighted Centroid
Crosshair Grid
b) Hough Transform
Dark Stop
c) Template Match d) Matched Filter
Shot Mask
5x5 (x,y) pixel grid
Tolerancelimit
OUT OF TOLERANCE!OUT OF TOLERANCE!