Waveguidebased Room Acoustics using Graphics Hardware

1
Waveguide-based Room Acoustics using Graphics
Hardware
Niklas Röber, Martin Spindler, Maic
Masuch University of Magdeburg
2
Outline
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Motivation
• Waveguide meshes and sampling lattices
• GPU-based implementation
• Results and discussion
• Summary and future work

3
Outline
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Motivation
• Waveguide meshes and sampling lattices
• GPU-based implementation
• Results and discussion
• Summary and future work

4
Research
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Games Research Group University of Magdeburg
• Tools and techniques for future games
• Non-photorealistic rendering and cinematography
• Interactive digital storytelling and authoring
  techniques
• Audiogames
• 3D virtual auditory environments
• Audiogames and augmented audio reality
• Interactive audiobooks
• Sound rendering and synthesis (3D, room acoustics)

5
Motivation
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Audiogames and auditory displays
• Non-realistic audio environment
• High quality 3D sound and room acoustic
  simulations
• Similarities between sound and light propagation
• Sophisticated and fast algorithms to render
  visual scenes
• Powerful graphics hardware
• AMD 64 ( 8 GFlops) vs. nvidia 7900 GTX ( 250
  GFlops)
• Very fast for parallelizable problems
• Various GPGPU applications

6
Outline
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Motivation
• Waveguide meshes and sampling lattices
• GPU-based implementation
• Results and discussion
• Summary and future work

7
Waveguide meshes
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Time domain finite difference models
• Bi-linear delay lines arranged in a mesh like
  structure
• Scattering junctions are of equal impedance
• Sum of inputs sum of outputs
• Pressure equal at crossings
• Modeling of boundary conditions
• Limitations
• Direction dependent dispersion error
• Finite mesh resolution / sampling
  frequency

8
Optimal Sampling
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Hexagonal lattices have a higher packing density
• More optimal sampling
• Spherically band limited signals
• Unit length increases to
• less samples in 3D
• 8 neighbors with 4 axes of propagation
  and 4 delay lines
  per node
• Used in compression, scientific
  visualization and image processing

9
BCC Waveguide Mesh
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Update frequency with unit length
• One BCC unit cell consist of 2 nodes and 8 delay
  lines
• Frequency dispersion with
  max. error 4.7 , compared to 7.3 (3D CC)
  (Campos and Howard 2005)
• Sampling efficiency

10
Outline
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Motivation
• Waveguide meshes and sampling lattices
• GPU-based implementation
• Results and discussion
• Summary and future work

11
GPU-based Implementation
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Implementation based on 3D textures, fragment
  shaders and framebuffer-objects (fbo)
• Two 32-bit float textures (RGB)
• Waveguide data (t-1, t) (R and B channel)
• Geometry, material and boundary conditions (G
  channel)
• Shader computes / samples texture using screen
  aligned slicing planes

12
Waveguide Shader
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• 1 uniform float layer
• 2 uniform vec3 stepX, stepY, stepZ
• 3 uniform sampler3D tex
• 4 vec3 pos vec3(gl_TexCoord0.xy, layer)
• 5
• 6 vec4 center texture3D(tex, pos)
• 7 vec4 left texture3D(tex, pos - stepX)
• 8 vec4 right texture3D(tex, pos stepX)
• 9 vec4 up texture3D(tex, pos stepY)
• 10 vec4 down texture3D(tex, pos - stepY)
• 11 vec4 front texture3D(tex, pos stepZ)
• 12 vec4 back texture3D(tex, pos - stepZ)
• 13
• 14 float ampl left.r right.r up.r
  down.r
• 15 ampl front.r back.r
• 16 ampl ampl 0.3333 - center.b
• 17 gl_FragColor vec4(ampl, center.g, center.r,
  1.0)

Waveguide fragment shader (Cartesian Lattice)
13
BCC Implementation
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• BCC lattice decomposed into two CC textures
• Base grid (R and G), offset grid (B and A) for
  t-1 and t
• One additional 3D texture (geometry, boundary
  conditions)
• BCC fragment shader
• Indexing adjusted for two textures
• Two nodes computed in one step
• Overall less computations

14
Outline
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Motivation
• Waveguide meshes and sampling lattices
• GPU-based implementation
• Results and discussion
• Summary and future work

15
Results and Discussion
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Benchmarks 2D, 3D CC and 3D BCC
• Software (CPU) and Hardware (GPU)
• Comparison CC and BCC lattices
• Examples
• Wavefield synthesis
• Impulse responses (3D CC, 3D BCC)

16
Benchmarks 2D
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• AMD 64 4000 single core PC with 1 GB main memory
• nvidia GeForce 7800GT with PCIe interface

17
Benchmarks 3D
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• cp with glCopyTexSubImage3D
• ncp without glCopyTexSubImage3D

18
Rectilinear vs. Hexagonal
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion
CC
BCC
19
Wavefield Synthesis
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• 21 sound sources
• Sine wave pulse
• Anechoic walls (top, bottom, right)
• Phase-rev. reflection (left, middle obstacle)

20
Impulse Responses
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Original sound ( The Notwist )
• Low-pass filtered
• 3D CC room acoustics
• 3D BCC room acoustics

21
Outline
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Motivation
• Waveguide meshes and sampling lattices
• GPU-based implementation
• Results and discussion
• Summary and future work

22
Summary
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• Novel and fast technique for wave propagation
  using waveguide meshes
• Exploiting powerful graphics hardware with
  build-in visualization
• Up to 15 20 times faster
• Additional improvements through hexagonal
  sampling
• Open questions / problems
• CC vs. BCC (quality)
• Missing 3D fbo extension

23
Future Work
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• GPU accelerated ray-acoustic simulations
• Global illumination (phonon mapping, spherical
  harmonics, BRDF)
• Raytracing (particles, raycasting)
• Combination of GPU-based wave- and ray-acoustics
• Evaluation of physics hardware (Ageias PPU)
• Mass-spring system
• Particles, raycasting

24
Thank you for your attention!
Outline Motivation Waveguides and Sampling
GPU Implementation Results Conclusion

• More Information
• games.cs.uni-magdeburg.de/acoustics
• niklas_at_isg.cs.uni-magdeburg.de