Skip to content

AEiE0606 Digital Signal Processing

\(z\)-transform and region of convergence

Definition:

\[ X(z)=\sum_{n=-\infty}^{\infty}x[n]z^{-n}. \]
  • ROC is the set of \(z\) values for which the series converges.
  • ROC contains no poles.

ROC, causality, and stability

For rational systems or sequences:

  • Causal right-sided sequence: ROC is outside the outermost pole.
  • Anti-causal left-sided sequence: ROC is inside the innermost pole.
  • Stable LTI system: ROC includes the unit circle.

Critical exam statement:

  • Stability requires the unit circle to lie in the ROC.
  • Causality for a rational system means ROC extends outward from the outermost pole.
  • A causal system is stable only if the unit circle is also outside all poles and included in the ROC.

Thus, for a causal stable rational system,

\[ \operatorname{ROC}: |z|>r_{\max}, \qquad r_{\max}<1. \]

“Outside a circle” expresses causality; stability supplies the additional unit-circle condition.

Inverse \(z\)-transform

  • Methods include power-series expansion, partial fractions, inspection, and contour integral definition.
  • The same algebraic expression with different ROCs can correspond to different time sequences.

System function and responses

\[ H(z)=\frac{Y(z)}{X(z)} \]

for zero initial conditions.

  • Poles govern natural or transient behavior.
  • Zeros shape frequency response and can cancel spectral components.
  • Steady-state sinusoidal response is obtained by evaluating on the unit circle if the unit circle is in ROC:
\[ H(e^{j\omega}). \]

Convolution and Parseval

  • Time-domain convolution corresponds to multiplication of transforms.
  • Parseval relates energy in time domain to energy in transform domain.

DFT and properties

The \(N\)-point DFT is:

\[ X[k]=\sum_{n=0}^{N-1}x[n]e^{-j2\pi kn/N}. \]

Inverse DFT:

\[ x[n]=\frac{1}{N}\sum_{k=0}^{N-1}X[k]e^{j2\pi kn/N}. \]

Property cues:

  • Linearity.
  • Periodicity in both time-indexed extension and frequency-indexed repetition.
  • Circular time shift corresponds to phase factor.
  • Conjugate symmetry for real sequences.

Multiplication of DFTs and circular convolution

  • Multiplication in the DFT domain corresponds to circular convolution in the time domain.
  • Linear convolution equals circular convolution only when proper zero-padding avoids aliasing.

Trap:

  • DFT-based fast convolution often requires zero-padding specifically to reproduce linear convolution.

FIR versus IIR

Filter type Key feature
FIR Finite impulse response, can achieve exact linear phase, usually non-recursive
IIR Infinite impulse response, recursive, efficient for sharp responses but phase usually nonlinear

IIR design by impulse invariant method

  • Start from analog prototype.
  • Sample the analog impulse response to obtain digital equivalent.
  • Preserves impulse-shape relation but can suffer aliasing.

IIR low-pass discrete filter design cue

  • Often begins with an analog low-pass prototype such as Butterworth or Chebyshev and maps it to digital form.

FIR design by Fourier approximation and window method

  • Ideal desired frequency response is approximated and truncated.
  • Windowing truncates the ideal impulse response and trades main-lobe width, which controls transition width, against side-lobe level, which controls ripple/leakage.
  • A window reduces Gibbs oscillation and controls sidelobes; it does not simply reduce transition bandwidth.
Window Typical trade-off cue
Rectangular Narrow main lobe, high sidelobes
Hamming / Hann Better sidelobe reduction
Blackman Wider main lobe, lower sidelobes

DSP examples

  1. If a rational ROC is outside the outermost pole, the sequence is causal.
  2. If stability is asked, first check whether the unit circle lies in the ROC.
  3. If DFT-domain multiplication is mentioned, answer circular convolution unless zero-padding converts it to linear convolution.

AEiE0606 revision box

  • Causal rational ROC: outside outermost pole.
  • Stable system: unit circle included in ROC.
  • Same poles with different ROC can mean different signals.
  • DFT multiplication <-> circular convolution.
  • FIR: finite, often linear phase. IIR: recursive, efficient, infinite response.