Discrete Fourier Transform (DFT)
and FFT Algorithms
• Subtopics:
• - DTFT vs DFT
• - DFT Definition & Properties
• - DFT and Circular Convolution
• - FFT (Radix-2) Algorithms
DTFT and DFT – Relationship
• - DTFT: Infinite-length signal, continuous
frequency spectrum
• - DFT: Finite-length signal, discrete frequency
spectrum
• Relationship:
• DFT samples the DTFT at ω_k = 2πk/N
DFT Definition
• DFT:
• X[k] = Σ (n=0 to N-1) x[n] e^(-j2πkn/N)
• Inverse DFT:
• x[n] = (1/N) Σ (k=0 to N-1) X[k] e^(j2πkn/N)
DFT Properties
• 1. Periodicity:
• X[k+N] = X[k], x[n+N] = x[n]
• 2. Linearity:
• DFT{a·x[n] + b·y[n]} = a·X[k] + b·Y[k]
• 3. Symmetry:
• For real x[n], X[N-k] = X*[k]
DFT as a Linear Transformation
• DFT is a matrix-vector multiplication:
• X = W_N · x
• W_N is the DFT matrix transforming time-
domain to frequency-domain
Multiplication & Circular
Convolution
• - Time-domain circular convolution ↔
Frequency-domain multiplication
• - z[n] = Σ (m=0 to N-1) x[m] · y[(n - m) mod N]
Efficient Computation – Why FFT?
• - DFT direct computation: O(N²)
• - FFT algorithm: O(N log₂N)
• Advantage: Drastically reduces computation
time
Radix-2 FFT Algorithms
• Decimation-in-Time (DIT):
• - Split input into even/odd samples
• - Recursive
• Decimation-in-Frequency (DIF):
• - Split DFT output
• - Butterfly structure
Butterfly Diagram
• Butterfly structure:
• X[k] = x₁ + W_N^k·x₂
• X[k+N/2] = x₁ - W_N^k·x₂
• Used in both DIT and DIF FFT algorithms
Summary
• - DFT: Finite-length frequency analysis
• - Related to DTFT by frequency sampling
• - FFT (DIT & DIF): Fast DFT computation
• - Widely used in DSP and communications

DFT_and_FFT_Presentation.pptx .com dft and fft power point presentation

  • 1.
    Discrete Fourier Transform(DFT) and FFT Algorithms • Subtopics: • - DTFT vs DFT • - DFT Definition & Properties • - DFT and Circular Convolution • - FFT (Radix-2) Algorithms
  • 2.
    DTFT and DFT– Relationship • - DTFT: Infinite-length signal, continuous frequency spectrum • - DFT: Finite-length signal, discrete frequency spectrum • Relationship: • DFT samples the DTFT at ω_k = 2πk/N
  • 3.
    DFT Definition • DFT: •X[k] = Σ (n=0 to N-1) x[n] e^(-j2πkn/N) • Inverse DFT: • x[n] = (1/N) Σ (k=0 to N-1) X[k] e^(j2πkn/N)
  • 4.
    DFT Properties • 1.Periodicity: • X[k+N] = X[k], x[n+N] = x[n] • 2. Linearity: • DFT{a·x[n] + b·y[n]} = a·X[k] + b·Y[k] • 3. Symmetry: • For real x[n], X[N-k] = X*[k]
  • 5.
    DFT as aLinear Transformation • DFT is a matrix-vector multiplication: • X = W_N · x • W_N is the DFT matrix transforming time- domain to frequency-domain
  • 6.
    Multiplication & Circular Convolution •- Time-domain circular convolution ↔ Frequency-domain multiplication • - z[n] = Σ (m=0 to N-1) x[m] · y[(n - m) mod N]
  • 7.
    Efficient Computation –Why FFT? • - DFT direct computation: O(N²) • - FFT algorithm: O(N log₂N) • Advantage: Drastically reduces computation time
  • 8.
    Radix-2 FFT Algorithms •Decimation-in-Time (DIT): • - Split input into even/odd samples • - Recursive • Decimation-in-Frequency (DIF): • - Split DFT output • - Butterfly structure
  • 9.
    Butterfly Diagram • Butterflystructure: • X[k] = x₁ + W_N^k·x₂ • X[k+N/2] = x₁ - W_N^k·x₂ • Used in both DIT and DIF FFT algorithms
  • 10.
    Summary • - DFT:Finite-length frequency analysis • - Related to DTFT by frequency sampling • - FFT (DIT & DIF): Fast DFT computation • - Widely used in DSP and communications