48x96 poster template

1
An SRAD Image Processor as a Thermal-Aware SoC
Designed for Low-Power Operation Wei Huang,
Garrett S. Rose, Yan Zhang, Wenqian Wu, Adam C.
Cabe, Zhenyu Qi, and Mircea R. Stan Electrical
and Computer Engineering Dept., University of
Virginia, Charlottesville, VA 22904
ALGORITHM
DESIGN METHODOLOGY AND TOOLS
BACKGROUND
Equations

• Thermal and Power Considerations in SoC Design
• High power kills battery (thus portability)
• High temperature kills functionality (thermal
  hazards)
• Positive loop between power and temperature
  Thermal Runaway
• SoC Design for SRAD
• SRAD Speckle Reducing Anisotropic Diffusion - a
  novel ultrasound imaging technique to reduce the
  speckles in an ultrasound image
• Offering a compact solution for an ultrasound
  image processing unit
• Lower power and faster communications among
  system modules
• Purposes
• Image processing system-on-a-chip
• Hardware implementation of a novel imaging
  algorithm
• Incorporating Dynamic Thermal Management (DTM)
  and Dynamic Voltage Scaling (DVS) in our design
• Y.Yu, S.T.Acton, Speckle reducing anisotropic
  diffusion.IEEE Trans. on Image Processing,
  11(11)12601270,2002.

Calculation of Diffusion Coefficient
Calculation of New Pixel Values
Specifications and descriptions
MATLAB simulations
VHDL development and ModelSim simulation
(Skipped, future work)
Cadence RTL Compiler
Cadence Virtuoso
ModelSim, prove netlist equivalency
Cadence Encounter (including floorplan)
BLOCK DIAGRAM SRAD UNIT
MATLAB Simulation Results
SIMULATION and TESTING
Global control unit

• The new value of I is finally calculated as
  follows
• Repeat steps 3 to 7 for 70 to 200 iterations
• Transform the pixel value by
• Rescale the pixel value to the range
• of 0255

Original Image
After 50 Iterations
Video input controller
SYSTEM-ON-CHIP LAYOUT
After 100 Iterations
After 200 Iterations
Sensors
Ring-oscillator temperature sensor
KEY FEATURES OF THE DESIGN
SRAM arrays
SRAD and control logic
Floorplan and sensor locations
Micro-photo and wire-bonding