35017 - Electronic Processing of Digital Signals (Graduate Course)

Learning outcomes

The aim of the course is to illustrate the basic methodologies for the treatment of digital signals in modern electronic systems operating in real time, with special reference to telecommunications, industrial automation and, more generally, digital processing of information. The common trait of these systems is the use of a programmable DSP, i.e. a special processor whose architecture is optimized to efficiently carry out repetitive arithmetic operations, while being at the same time sufficiently flexible to be programmed for the execution of highly diversified functions. The course organization includes, in addition to room lectures, a cycle of laboratory training lectures where the students will be given the opportunity to design a module of a digital-signal processing system based on a programmable DSP. Typical examples may include: 1) Waveform generation; 2) Design and implementation of finite-impulse response (FIR) digital filters; 3) Design and implementation of infinite-impulse response (IIR) digital filters; 4) Design and implementation of adaptive digital filters; 5) Generation of pseudo-random binary sequences; 6) Echo-cancellation systems; 7) Frequency conversion with complex modulation signals; 8) Frequency demodulation; 9) Clock syncronization in TLC circuits.

Course contents

Signal analysis. Introduction on analog, discrete-time and digital signals. Discrete Fourier transform of a sampled signal. Relationship between the Fourier transform and the discrete Fourier transform of a sampled signal. The Shannon theorem. Time-frequency duality. Z-transform and its properties. Relationship between the Z-transform and the discrete Fourier transform. Inverse Z-transform. The residue theorem. Linear transformations between time-discrete signals. Transfer function. Real convolution. Signal correlation.
  
Design and realization of FIR filters. Finite-impulse response (FIR) filters and their realization. Transfer function of FIR filters. Filter properties related with parity conditions of the coefficients. Filter specs. FIR filter design: alternation theorem and Remez algorithm. MATLAB design of FIR filters. Implementation experiments of FIR filters on the TMS320C6711 DSP.
  
Design and realization of IIR filters. Infinite-impulse response (IIR) filters and their realization. Transfer function of IIR filters. Stability theorems and stability criteria for IIR filters. Realization of IIR filters: canonical form; series and parallel decomposition of 2nd order canonical blocks; ladder filters. Analog filters: Butterworth, Chebishev, Bessel and elliptic filters. Conversion of an analog filter into a discrete-time filter with approximately the same transfer function. The invariant impulse-response method and its extension to filters with poles and zeros. The bilinear transformation method. Constantinides transformations. Quantization effects: precision, instability, dead-band effects and limit cycles. Design of IIR filters using MATLAB. Implementation experiments of IIR filters on the TMS320C6711 DSP.
 
Discrete-time stochastic processes. Definition of stochastic process. First- and second-order probability distributions. Autocorrelation function. The Wiener-Kintchine theorem. Relationship between input and output spectral densities. Quantization noise. Influence of the coefficient quantization on the filter transfer function. Product quantization and its influence on the filter transfer function. Fixed-point implementation and determination of the scale factor of quantized values. Dead-band effects and limit cycles for purely-recursive filters of 1st and 2nd order. Calculation of the extrema of the limit cycle.
 
Fast Fourier Transform (FFT) and its applications . Discrete Fourier transform (DFT) of sampled periodic signals: definitions and properties. Inverse Fourier transform (IDFT). Relationship between the DFT, the discrete Fourier transform of an aperiodic signal and the Z-transform. Relationship between the DFT and the series expansion of a continuous-time periodic signal. Sampling in the frequency domain. Fast Fourier transform (FFT). Time-domain and frequency-domain decimation. MATLAB implementation of the FFT.
 
Custom implementation of filters. Filter representation by data-flow (DFG) and signal-flow (SFG) graphs. Definitions of iteration bound and critical path. Transposition theorem. Pipeline and parallel implementations of FIR filters. Supply-voltage scaling and its impact on energy dissipation. Retiming techniques. Shortest-path algorithm between two DFG nodes. Minimization of the critical path by filter retiming. Folding transformation meant to reduce the number of arithmetic units in an integrated VLSI filter implementation. Solution method of an inequality set.
 
Adaptive filtering. Adaptive filters and optimization methods: “steepest-descent” (SD) and “least mean square” (LMS) algorithms. Performance analysis: stability, convergence, mean-square error. Applications: adaptive system identification; adaptive linear prediction; adaptive noise cancellation; adaptive channel equalization.
 
Speech processing. Definitions of complex and real “cepstrum”. Generalized superposition principle and homomorphic deconvolution. Canonical form of homomorphic systems and its representation. Deconvolution based on the generalized superposition principle. Minimum-phase and maximum-phase sequences. Hilbert transformation. Homomorphic filtering. Speech model. Application of the complex cepstrum to the speech analysis and synthesis.
 
Architecture of the TMS320C6711 DSP.Schematic of the TMS320C6711 DSP. Software development tools: C compiler, assembler, linker. Addressing methods: direct, indirect, absolute, circular methods. Addressing memory-mapped registers. Pipeline of the TMS320C6711 DSP. Parallel execution. Instruction set: arithmetic, logic, bit-level instructions. Program-flow control instructions. Mixed programming in C language and linear assembler. Applications and experiments.

Readings/Bibliography

A. Antoniou: Digital Filters: Analysis and Design, Mc Grow-Hill, 1979.

A.V. Oppenheim, R.V. Shafer: Discrete Time Signal processing, Prentice Hall, 1995.

K.K. Parhi: VLSI Digital Signal Processing Systems, Wiley, 1999.

S. M. Kuo, B. H. Lee: Real-Time Digital Signal Processing, Wiley, 200

E. C. Ifeachor, B. W. Jervis: Digital Signal Processing, Prentice Hall, 2002.

Teaching methods

The course is delivered with room lectures (5 hours per week over 14 weeks)

Assessment methods

The assessment is based on an oral colloquium on the contents of the lectures.

Teaching tools

The teaching material of this course can be found in the web site: http://didattica.arces.unibo.it/

Links to further information

http://didattica.arces.unibo.it/user/view.php?id=4&course=1

Office hours

See the website of Giorgio Baccarani