VLSI DESIGN PRESENTATION

1
VLSI DESIGN PRESENTATION

• DSP IMPLEMENTATION ON RECONFIGURABLE COMPUTING

2
Reconfigurable Computing for Digital Signal
Processing

• Steady advances in VLSI technology and design
  tools have extensively expanded the application
  domain of digital signal processing over the past
  decade.While application-specific integrated
  circuits (ASICs) and programmable digital signal
  processors (PDSPs) remain the implementation
  mechanisms of choice for many DSP
  applications,increasingly new system
  implementations based on reconfigurble computing
  is being considered.

3
DSP Implementation Spectrum
4
DSP Implementation Comparison
5
DSP PROCESSORS

• Compared to ASIC or FPGA hardware solutions a DSP
  chip is by far the slowest option.
• DSPs just draw too much power for portable
  applications.
• Power down sides of DSP chips become of too much
  of a deterrent ,most designers have turned to
  ASIC hardware solutions.A hard-wired , cell based
  custom chip.An ASIC solution will be faster ,
  more efficient , and cheaper than its DSP chip
  alternative.

6
PARALLELISM

• Many characteristics of FPGA devices,in
  particular,make them especially attractive for
  use in digital signal processing systems.The
  fine-grained parallelism found in these devices
  is well matched to the high sample rates and
  distributed computation often required of signal
  processing applications in areas such as image
  audio,and speech processing.Given the highly
  pipelined and parallel nature of many DSP
  tasks,such as image and speech processing,these
  implementations have exhibited substantially
  better performance than standard PDSPs

7
Distributed Single-Chip DSP Interconnection
Network
8
Hybrid DSP Architecture

• The parallel execution of the DSP core and the RL
  resources can often be expressed as simple
  extensions of DSP instruction sets.The on-chip
  DMA controllers which many DSPs already have may
  prove useful for configuring the RL without
  burdening the processor with the task.

9
DSP With a Reconfigurable Coprocessor Combination

• The RR both access to the processor register file
  and the four on-chip memory ports.This makes the
  RR a cross between a function unit and a
  coprocessor.It provides flexibility and reduces
  the requirement of using memory as the means of
  communication Between the two data paths,further
  reducing the memory bandwidth burden and allowing
  tighter cooperation between the DSP core and the
  RL

10
The FPGA Based Array Processor for an
Inosphereric-Imaging Radar

• This slide has described an FGPA-based system to
  process signals for a passive radar.The system
  uses an FPGA based array processor and a DSP
  board to compute a cross-ambiguity function and a
  crosscorrelation on the received data.The system
  is dynamically reconfogurable.It allows users to
  modify range,delay and decimation factor.

11
Reconfigurable DSP Processor

• A Reconfigurable architecture which incorporates
  a reconfigurable coprocessor into a DSP can have
  performance benefits for a reasonable increase in
  chip area.In addition,DSPs have many
  architectural features which make a combination
  with reconfiguarble logic feasible and
  beneficial.Despite the raw clock rate
  disadvantage of DSPs when compared with general
  purpose microprocessors,DSPs serve an important
  role in high-performance,embedded computingA
  reconfigure processor on-chip can help DSPs
  exploit more of the parallelism found in digital
  signal processing applications,thus improving the
  processors overall performance

12
Designing Digital Signal Processing with FPGAs

• Designers creating an FPGA implementation begin
  with the knowledge that the multipliers they will
  need are often already fabricated on the
  chip.Whereas DSP processors typically have only 8
  dedicated multipliers at their disposal,a higher
  end FPGA device such as Alteras Stratix offers
  up to 224 dedicated multipliers plus additional
  logic element-based multipliers are needed.

13
Designing Digital Signal Processing with FPGAs

• Complex digital signal processing applications
  such as finite impulse response (FIR)
  filters,modulation-demodulation , and encryption
  call for larger multiplier requirements.Given the
  number of multipliers available with an FPGA ,the
  designers job of defining an architecture for
  these types of digital signal processing
  applications tends to be quick and extremely
  flexible.Higher- end FPGA devices often feature
  multiple DSP blocks that can provide data
  throughput of up to 56 GMACS.

14
Digital Signal Processing Architecture

• A typical digital signal processing architecture
  includes three structures
• Datapath- a collection of arithmetic operators
  such as adders and multipliers
• Controller- finite state machines (FSMs) that
  sequence the actions of the datapath.
• Memory-temporary storage elements used while the
  algorithm executes.

15
What is the major advantage of FPGA over ASIC

• Once a system designer has arrived at an
  architecture that runs the algorithm efficiently
  , she or he typically turns the job over to a
  team of hard ware designers.This team,in turn,
  translates the architecture into a register
  transfer level description in HDL (VHDL or
  Verilog).This translation allows hardware
  designers to consider the design at the
  appropriate level of abstraction.And it is here
  that a major advantage of FPGA over ASIC can be
  found.

16
How FPGA Design Deliver Faster Performance

• During the Place Route phase of FPGA design,all
  logic components are located,wiring connections
  are made,and a final timing analysis is
  performed.Place and route software,such as
  Alteras Quartus II uses the hardware designers
  timing constraints to create optimal logic
  mapping and placement.Critical timing paths are
  optimized first to help achieve timing closure
  faster and deliver faster performance.

17
Implementing DSP on FPGA

• FPGA that has been optimized to perform a digital
  signal processing task,will run anywhere from 10
  times to more than 1000 times faster than a
  single DSP chip.Whereas a DSP processor typically
  employs serial processing,the parallel capacities
  inherent FPGA architecture will always give them
  both a significant edge over DSPs.DSPs just draw
  too much power for portable applications like
  Cellular Mobile Communications and Wireless
  LAN.Texas Instrument doesnt use DSP for WLAN.

18
Implementing DSP on FPGA

• An increasing number of designers are turning to
  programmable logic devices,or FPGA , as a way to
  navigate the software-hardware extremes of DSP or
  ASIC design solutions.An FPGA thats been
  enhanced for digital signal processing gives you
  unlimited customizing options in a chip without
  all the silicon physical-design work required for
  an ASIC solution.

19
RISE OF FPGA

• Key tasks to the rise of FPGA in the signal
  Processing realm could be assigned to hardware to
  take advantage of optimum speed , power
  consumption , and per unit costs,while other
  tasks are performed by software to speed time to
  market and ease legacy compatibility.With digital
  signal processing continuing to be a critical
  component in the evolution of wireless networks
  and in multimedia .FPGA will grow in their
  ability to deliver state-of-art signal processing
  in a fraction of the time than ASIC.

20
Implementing DSP Designs in Altera?Stratix Devices

• The most commonly used DSP functions are FIR
  (Finite Impulse response)filters,IIR(Infinite
  Impulse response filters,FFT(Fast Fourier
  Transform),DCT (direct Cosine Transform),Encoder/D
  ecoder and Error Correction/Detection
  functions.All of these blocks perform intensive
  arithmetic operations such as add,subtract,multipl
  y,multiply-add or multiply-accumulate.

21
Signal flow graph for an 8-trap Direct Form I FIR
filter
22
Simplified View of a Stratix DSP Block
23
VHDL Code Describing Multiply-Accumulate Function
24
Verilog Code Describing Multiply-Add/Subtract
Function
25
VHDL Code That Infers Dual-Port RAM
26
Verilog Description that Infers Shift Registers
27
SPEECH PROCESSING

• To support speech processing,a bus-based multi
  FPGA board,Tabula Rasa,was Programmed to perform
  Markov searches of speech phonemes.This system is
  particularly interesting because it allowed the
  use of behavioral partitioning and contained a
  codesign environment for specification,synthesis,s
  imulation, and evaluation design phase.Moreover
  we can perform Speech Recognition based on
  Speaker Dependent or Speaker Independent.

28
The Changing World of DSP Applications

• DSP is often found in human computer interface
  such as
• Sound Cards
• Video Cards
• Speech Recognition
• DSL Modems
• CDMA Receiver Chips
• Read circuitry of hard drive and CD/DVD storage
  system

29
The Changing World of DSP Applications

• Wireless communications and multimedia
• Cellular Mobile Communications,both GSM and CDMA
• Wireless LANs
• General Purpose Processors, such as Intel
  Pentium, can provide much of the signal
  processing needed for desk-top applications such
  as audio and video processing.

30
FUTURE WORK

• A refinement of the programming methodology for
  the DSP hybrid processor as well as the many
  issues of interfacing the RL with the DSP.The
  intention is to make the programming task
  resemble more of a software creation problem than
  a hardware design exercise,a requirement crucial
  in making the architecture usable by DSP
  programmers and not just hardware designers.
  Future work will also quantify the effects of
  frequent reconfiguration on application
  performance.