You are on page 1of 13

DSP 1 Lecture 2.

3
Discrete Fourier Transform

1

RECAP Frequency of Interest BIN Frequency Peak BIN f frequencies i Spectral leakage Scalloping effect Spectral leakage Spectral leakage fM f exact f M −1 fs N − f M +1 N-Points N Points DFT Frequency Estimations 1 fs 1 fs + f BIN _ PEAK < f exact < + f BIN _ PEAK 2N 2N 2 .

M 2 • Improves the DFT frequency resolution.1..RECAP xn → n = 0. fs M = L× N n = 0.. fs fs fs .... M − 1 ⎧ xn → n = 0.1...... N − 1 N .. 2. • Minimizes scalloping effects ≠0 N =0 2N N M =3 =0 3N N 3 .. N − 1 xn = ⎨ 0 → n > N −1 ⎩ fs fs .. N 2 M ..

N-Point DFT f s .Minimizes Scalloping ? POSSIBLE ANSWER -.Convolution !! Step 1 --.Adding Zeros …M-Point DFT (M = LxN) f s .0 -. L × N ⎛ 1 ⎞ fs Spectral resolution !!! ⎜ ⎟ ⎝L⎠ N 4 .Discrete Fourier Transform Question 1. N fs p N Spectral resolution !!! Step 2 --.

Discrete Fourier Transform fj f j −1 ⎛ 1 ⎞ fs ⎜ ⎟ ⎝L⎠ N ⎛ 1 ⎞ fs ⎜ ⎟ ⎝L⎠ N ⎛ 1 ⎞ fs ⎜ ⎟ ⎝L⎠ N ⎛ 1 ⎞ fs ⎜ ⎟ ⎝L⎠ N EXACT DTFT f j +1 Original BIN Frequencies Additional BIN Frequencies Due adding zeros fs N Observations fs N 5 • L-1 extra BIN frequencies will results in between the OLD BIN Frequencies • The original N BIN frequencies remain intact (including amplitude) • Better Fidelity on Amplitude resolution is achieved (Therefore minimizing scalloping to a degree) .

Discrete Fourier Transform We need an example !! 6 .

Let us consider the following situation !! p g x 'n → 0 ≤ n ≤ N − 1 wn = 1 → 0 ≤ n ≤ N − 1 DTFT x xn = x 'n wn → 0 ≤ n ≤ N − 1 1 X (e jω ) = 2π π − ∫π X '(e jθ )W (e j (ω −θ ) )dθ 7 DTFT x’ DTFT w .Convolution !! Step 1 --.Discrete Fourier Transform POSSIBLE ANSWER -.

Attributes !! 2π 2π − ≤ω ≤ N N N 8 Main l b M i lobe width idth Max Value .Compute DTFT of w wn = 1 → 0 ≤ n ≤ N − 1 sin(ω N ) 2 1 sin(ω ) 2 W (e jω ) = ∑ wn e − jω n = n =0 N −1 1− e =e − jω 1− e − jω N N −1 −1 − jω ( ) 2 Step 3--.Discrete Fourier Transform Step 2 --.

Discrete Fourier Transform 2π N 2π 2× N 2π 3× N 9 .

Discrete Fourier Transform fs 2π f s × = N 2π N fs N fs 2× N fs 3× N 10 .

Discrete Fourier Transform Step 4 --.Compute DTFT of x X (e jω ) = 1 2π π − ∫π X '(e jθ )W (e j (ω −θ ) )dθ = 1 2π π −π ∫ Scalloping Due to window Convolution sinusoid N ⎤ ⎡ ) N −1 sin(( ω − θ ) − j (ω −θ ) ( ) jθ ⎢ 2 ⎥ dθ 2 X '(e ) ⎢e 1 ⎥ ⎢ sin((ω − θ ) ) ⎥ ⎣ 2 ⎦ Exact DTFT Convolution sinusoid f ( m − 1) s N m fs N Exact Frequency ( m + 1) fs N ( m + 2) fs N 11 fs N .

Discrete Fourier Transform Let us consider the following situation !! ∑ n =0 N −1 1 X [k ] = N 2 N −1 n =0 ∑ x [n] = rms 2 n =0 −j 2π n⋅k N N −1 2 → ⎡volts 2 ⎤ ⎣ ⎦ 1 X [k ] = N rms 2 = ∑ n =0 N −1 ∑ x[n]e → Δf = fs N 2 X [k ] ⎡ volts 2 ⎤ 2 →⎢ X [k ] = ∑ P(k )Δf → P(k ) = ⎥ Δf n =0 ⎣ Hz ⎦ N −1 X [k ] ⎡ volts 2 ⎤ P(k ) = →⎢ ⎥ Δf ⎣ Hz ⎦ 2 Power Spectral Density Power Spectrum X [k ] → ⎡volts 2 ⎤ ⎣ ⎦ 2 12 .

FIN !!!! 13 .