Grasping the Potencial of

1

• Grasping the Potencial of
• Digital Signal Laboratory Processing
• through Real-Time DSP Laboratory Experiments
• Aníbal J. S. Ferreira, Francisco J. O.
  Restivo
• Faculdade de Engenharia da Universidade do
  Porto
• Departamento de Engenharia Electrotécnica e
  Computadores
• Rua Dr. Roberto Frias, 4200-465 Porto,
  Portugal
• INESC Porto, Portugal
• ajf_at_inescporto.pt, fjr_at_fe.up.pt

2
Abstract

• A new DSP laboratory course has been included in
  the Electrical and Computer Engineering
  curriculum at the Faculdade de Engenharia da
  Universidade do Porto, in Portugal, since the
  school year of 1999/2000. This paper addresses
  the context and motivation underlying this new
  course, outlines its structure and methodology,
  highlights the design and goals of all DSP
  experiments currently proposed for the 13 weeks
  of the semester, and reports on the receptivity
  students have expressed to this elective course.
  The course is based on the TI C31 Starter Kit and
  tries to combine full use of its resources with a
  representative diversity of efficient digital
  signal processing techniques and associated
  application scenarios. A perspective is also
  given on current plans to reinforce DSP expertise
  at the graduate level.

3
Summary

• EEC at FEUP-DEEC
• basic structure and topics basic to DSP
• EEC4162 (PDS)
• EEC5274 (PDSTR)
• focus, rational, history
• C31 starter kit
• course structure and organization
• DSP laboratory experiments
• students feed-back
• Looking Forward
• new challenges
• Conclusion

4
EEC at FEUP-DEEC

• basic structure
• 1st and 2nd year core
• 3rd year students select a branch
• APEL industrial automation, production and
  electronic systems
• E energy systems
• TEC telecommunication, electronic and computer
  systems
• 4th and 5th year mandatory elective
  disciplines per branch

3rd Year APEL
5th Year APEL
4th Year APEL
1st Year
2nd Year
3rd Year E
5th Year E
4th Year E
3rd Year TEC
5th Year TEC
4th Year TEC
PDS PDSTR
5
EEC at FEUP-DEEC

• basics to DSP during the first 3 years
• programming
• Fourier transform / Fourier analysis
• Laplace transform
• Z transform
• sampling / modulation
• filtering
• random processes
• noise
• other topics

• digital systems
• signal theory
• circuit theory
• systems theory

• algorithms and data structures
• microprocessors
• probability and statistics
• telecommunications

6
EEC4162 (PDS)

• syllabus
• 3-hour class / week theory illustrative
  problems
• 2-hour class / week problems, application of
  theory,
• 6 individual work assignments / semester (e.g.,
  using Matlab)
• main topics
• discrete signals and systems
• sampling and reconstruction of analogue signals
• linear-time invariant systems
• structures for the realization of LTI systems
• FIR and IIR filter design
• finite word length effects
• decimation and interpolation
• the discrete Fourier transform
• overlap-add and overlap-save methods of FFFD
• response of LTI systems to random discrete
  signals
• FFT and its implementation

7
EEC4162 (PDS)

• teaching experience reveals that
• typically, a pencil and paper approach, even if
  complemented with simulation exercises using
  Matlab, is not enough for the student to grasp
  the advantage and potential of digital signal
  processing in many application areas including
  multimedia, telecommunications, control, and
  consumer electronics
• as a consequence, a hands-on DSP laboratory
  course has been included, since 1999, in the EEC
  curriculum, so as to motivate students to explore
  DSP based solutions to practical problems such as
  filtering, Fourier analysis or Single Side Band
  Modulation (SSB)

8
EEC5274 (PDSTR)

• profile
• elective course based on DSP laboratory
  experiments
• offered during the 8th semester of the EEC
  curriculum
• part of the EEC curriculum since 1999
• student preference
• 1999/2000 16 students
• 2000/2001 32 students
• 2001/2002 22 students
• focus
• practical digital signal processing issues and
  applications
• efficient realization structures
• real-time processing constraints
• approach
• advantages of DSP are demonstrated by lab
  examples covering a representative diversity of
  application scenarios
• student is challenged with specific DSP design
  and realization issues.

9
EEC5274 (PDSTR)

• selected DSP laboratory platform C31 starter Kit
• DSP initialization kit able to realize many
  different laboratory experiments running in
  real-time
• includes assembler and (windows-based) debugger
  environment (Go-DSP Code Explorer)
• availability of many demonstration code examples
  and support (C31 teaching kit), and literature
• kind support of the Texas Instruments European
  University Programme

10
EEC5274 (PDSTR)

• modus operandi
• 1.5 hour class / week
• theory, demonstration of concepts
• 2.5 hour class / week
• laboratory work design, realization and
  performance assessment of algorithms running in
  real-time
• students are encouraged to keep the kit between
  classes
• eases preparation for the lab work
• opportunity for students to explore beyond the
  strict realization goals of each lab work
• self-learning

11
EEC5274 (PDSTR)

• theory and laboratory classes
• first part (theory)
• use of the development environment of the C31
  assembler, debugger, C31 architecture,
  peripherals and instruction set
• second part (theory)
• real-time realization of algorithms for FIR IIR
  and FIR-adaptive filtering, multirate processing
  using polyphase decomposition, FFT and spectral
  analysis, SSB modulation (Hilbert Transform)
• laboratory classes
• 10 lab experiments and reports during the 13-week
  semester
• each report main results and conclusions of each
  lab work
• early feed-back is given to students

12
EEC5274 (PDSTR)

• syllabus
• presumes the core knowledge of the pre-requisite
  DSP course (EEC 4162), including efficient
  realization structures and FFT
• new topics
• filter banks, uniform filter banks and their
  relation to the DFT
• half-band filters, M-band filters, power
  complementary filters
• the QMF filter bank, design and implementation
  issues, multi-resolution analysis using the QMF
• adaptive filtering
• polyphase decomposition of interpolation and
  decimation filters and their efficient
  realization
• Hilbert Transformer and SSD modulation

13
EEC5274 (PDSTR)

• laboratory experiments

caracterization of the problem DSP conceptual
approach
validation in Matlab identification of
opportunities for efficient realization
assembly code debugging performance assessment
14
EEC5274 (PDSTR) - Lab experiments

• 1. verification of the aliasing in sampling
• goal estimate the sampling frequency just by
  hearing the result at the output of the system,
  due to a sinusoid of varying (and known)
  frequency that is injected at the input

15
EEC5274 (PDSTR) - Lab experiments

• 2. waveform generation and converter testing
• goal take advantage of circular addressing in
  order to synthesize different waveformss, and of
  the lookp-back mode of AIC in order to evaluate
  the accumulated quality of the D/A and A/D
  conversion (delay, noise floor, differential
  non-linearity)

16
EEC5274 (PDSTR) - Lab experiments

• 3. fixed point processing vs. floating point
  processing
• goal identify issues of fixed representation of
  numbers (namely the need for scaling) in
  recursive processing, versus floating-point

sin(n?)
sin(n1)?
c1
c2
cos(n?)
cos(n1)?
c3
c4
17
EEC5274 (PDSTR) - Lab experiments

• 4. FIR filtering
• goal 1 H(z)1-z-15, compare theoretical H(ej?)
  vs. experimental
• goal 2 realization of FIR equiripple h(n),
  2h(n)cos n?/2, (-1)nh(n)

?
?/2
?
?
?/2
?
?
?/2
?
18
EEC5274 (PDSTR) - Lab experiments

• 5. IIR filtering
• goal 1 6th order IIR, type 2 realization
  structure
• goal 2 compare simulated response vs.
  experimental response

19
EEC5274 (PDSTR) - Lab experiments

• 6. five vowel synthesizer
• goal synthesize /à/, /é/, /i/, /ó/, /u/ on a
  DSK using three formants
• didactic application allows to give each
  implementation a personal flavor (high motivation
  impact)
• quality improvements by modulating pitch

20
EEC5274 (PDSTR) - Lab experiments

• 7. interpolation using polyphase filters
• goal efficient realization of interpolation
  filter, experimental evaluation of of the sinc
  function associated to the D/A reconstruction
  when 4-fold interpolation is used and when not

?
A/D
D/A
AAF
AIF
C31
21
EEC5274 (PDSTR) - Lab experiments

• 8. adaptive filtering
• goal 1 realization and assessment of the
  operation of an adaptive filter (32 tap FIR, LMS)
• goal 2 implementation of the configuration
  insuring real-time echo canceling

what are the differences ?
22
EEC5274 (PDSTR) - Lab experiments

• 9. FFT assembly implementation
• goal 1 assembly implemention of a radix-2 FFT
  based on a C-like optimized (Matlab) code (N64)
• goal 2 assessment of the implementation when
  used as a simple real-time spectrum analyser

23
EEC5274 (PDSTR) - Lab experiments

• 10. single side band modulation
• goal design and realization of a band-pass
  filter, Hilbert transformer, SSB modulator based
  on analytic signal generation

?
?
down-shift
?
?
up-shift
?
?
spectral-inversion
?
?
24
EEC5274 (PDSTR)

• students feed-back
• plus
• tangible and intuitive linking between theory and
  practice
• methodology elicits insight and promotes
  application
• minus
• more time to prepare and to play
• sooner indication of lab work description
• lab classes gt 2,5 hours
• regarding possibility of additional DSP lab
  course
• 50 say yes if addressing more application
  scenarios and combining video and audio

25
Looking Forward

• new challenges
• new starter kit ? VLIW TMS320C6711 ?
• new DSP issues
• utilization of cache
• utilization of DMA
• combined C, assembly code optimization

26
Conclusion

• PDSTR an ECE advanced undergraduate level DSP
  laboratory course
• all course material (in Portuguese) is available
  on the Web
• hands-on DSP laboratory experience
• consolidates knowledge
• stimulates criativity
• helps to develop a rewarding sense of achievement
• motivates final-year course projects using DSP
  technology
• plans for a new DSP lab (C6711) at the graduate
  level
• focus on complex and complete algorithm
  implementation