efficient calculation of computer generated holograms via parallel computing

15
1 of 24 Efficient calculation of computer generated holograms via parallel computing Vincent Ricardo Daria 1,2 and Andrew Banas 1 1. Instrumentation Physics Laboratory National Institute of Physics University of the Philippines Diliman, Quezon City 2. Computational Science Research Center University of the Philippines Diliman, Quezon City

Upload: xiu

Post on 05-Jan-2016

39 views

Category:

Documents


4 download

DESCRIPTION

Efficient calculation of computer generated holograms via parallel computing. Vincent Ricardo Daria 1,2 and Andrew Banas 1 Instrumentation Physics Laboratory National Institute of Physics University of the Philippines Diliman, Quezon City Computational Science Research Center - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Efficient calculation of computer generated holograms via parallel computing

1 of 24

Efficient calculation of computer generated holograms via parallel computing

Vincent Ricardo Daria1,2 and Andrew Banas1

1. Instrumentation Physics LaboratoryNational Institute of PhysicsUniversity of the PhilippinesDiliman, Quezon City

2. Computational Science Research CenterUniversity of the Philippines Diliman, Quezon City

Page 2: Efficient calculation of computer generated holograms via parallel computing

2 of 24

Holograms: Tailored optical landscapes

Output 3D light fields

LaserOpticssystem of

lenses

Spatial light modulator

Com puter

Hologram

Page 3: Efficient calculation of computer generated holograms via parallel computing

3 of 24

Spatial Light Modulator: a promising tool for dynamic beam control

Digital technology: Computer-addressable

pixel array

Amplitude modulationPhase modulation

http://www.avdeals.comhttp://www.holoeye.com

Page 4: Efficient calculation of computer generated holograms via parallel computing

4 of 24

High-optical throughput tailored optical landscapes

Generalized Phase Contrast

Holographic

Page 5: Efficient calculation of computer generated holograms via parallel computing

5 of 24

Tailored beam shapes via the Holographic method(Gabor, 1948)

Output intensity patternHologram (Records Complex Field)

SignalBeam

ReferenceBeamCom puter

Page 6: Efficient calculation of computer generated holograms via parallel computing

6 of 24

I.F.T.

Intensity constraints

F.T.Intensity

constraints

Phase

Iterative hologram design: Gerchberg-Saxton Algorithm

R.W. Gerchberg, W.O. Saxton, Optik 35, 237 (1972)

Hologram Hologram planeplane

Fourier Fourier planeplane

Page 7: Efficient calculation of computer generated holograms via parallel computing

7 of 24

Parallel Fast Fourier transform

t = 3.85x -0.8133

t = 0.96x -0.9737

0.01

0.1

1

10

0 2 4 6 8 10 12 14 16

Number of nodes, x

Tim

e (

se

co

nd

s)

Athlon Cluster (750 MHz)

Intel Pentium 4 Cluster (2.3 GHz)

Banas and Daria, Proc. Of SPP (2005)

Test cluster

Page 8: Efficient calculation of computer generated holograms via parallel computing

8 of 24

Parallel Gerchberg-Saxton Algorithm

A Banas, A Hilario and V Daria, Proc of SPP 2006

Page 9: Efficient calculation of computer generated holograms via parallel computing

9 of 24

Interface with experimental setup

Spatial light

modulator

Camera

Fast Ethernet100 MBps

Page 10: Efficient calculation of computer generated holograms via parallel computing

10 of 24

The CSRC –High Performance Computing Facility

Nodes: (32)•Intel Pentium 4 3.2 GHz•Gigabit Ethernet

• Servers (4)•Dual processor•Intel Xeon 2.8 GHz (32-bit)•Gigabit Ethernet

Page 11: Efficient calculation of computer generated holograms via parallel computing

11 of 24

Parallel Gerchberg-Saxton algorithm performance with the CSRC HPC

0.1

1

10

100

0 2 4 6 8 10 12 14 16 18 20

Nodes

Ru

n T

ime

(s)

2048 X 2048

1024 X 1024

512 X 512

256 X 256

128X 128

Numerical field reconstruction

Network traffic

Page 12: Efficient calculation of computer generated holograms via parallel computing

12 of 24

Results

TargetNumerical

reconstructionOptical

reconstruction

Page 13: Efficient calculation of computer generated holograms via parallel computing

13 of 24

Summary for Parallel GS algorithm

Developed a parallel Gerchberg-Saxton algorithm for deriving computer generated holograms (CGH).

Large CGH arrays are distributed into computing nodes thereby easing memory allocations for each node

Significant increase in speed of calculation is achieved at larger arrays

Page 14: Efficient calculation of computer generated holograms via parallel computing

14 of 24

Wavefront engineering group

Team Leader Vincent Ricardo Daria, DrEng (Associate Professor)

ResearchersDarwin Palima, PhD (Adjunct Professor)Anthony Montecillo (Research Associate)Godofredo Bautista (PhD Physics student)Jacquiline Romero (MS Physics student)Andrew Banas (BS Physics student)Atchong Hilario (BS Applied Physics student)Reniel Cabral (BS Applied Physics student)

Page 15: Efficient calculation of computer generated holograms via parallel computing

15 of 24

Thank you

Vincent Ricardo DariaNational Institute of PhysicsCollege of ScienceUniversity of the PhilippinesDiliman, Quezon City E-mail: [email protected]