READING ASSIGNMENTS

Signal Processing First

ƒ This Lecture:
ƒ Chapter 6, Sections 6-6, 6-7 & 6-8

Lecture 13 ƒ Other Reading:

Digital Filtering ƒ Recitation: Chapter 6
of Analog Signals ƒ FREQUENCY RESPONSE EXAMPLES
ƒ Next Lecture: Chapter 7

10/6/2003 © 2003, JH McClellan & RW Schafer 1 10/6/2003 © 2003, JH McClellan & RW Schafer 3

LECTURE OBJECTIVES TIME & FREQUENCY

M M
ƒ Two Domains: Time & Frequency
ƒ Track the spectrum of x[n] thru an FIR
y[n ] = ∑ bk x[n − k ] = ∑ h[k ] x[n − k ]
Filter: Sinusoid-IN gives Sinusoid-OUT k =0 k =0
FIR DIFFERENCE EQUATION is the TIME-DOMAIN
ƒ UNIFICATION: How does Frequency
Response affect x(t) to produce y(t) ? M
H ( e jωˆ ) = ∑ h[k ] e − jωˆ k
H ( e jω̂ )
x(t) x[n] y[n] y(t) k =0
A-to-D D-to-A

FIR H ( e jωˆ ) = h[0] + h[1] e − jωˆ + h[2] e − j 2ωˆ + h[3] e − j 3ωˆ + "
ω̂ ω̂
10/6/2003 © 2003, JH McClellan & RW Schafer 4 10/6/2003 © 2003, JH McClellan & RW Schafer 5
Ex: DELAY by 2 SYSTEM DELAY by 2 SYSTEM

Find h[n ] and H (e jωˆ ) for y[n ] = x[n − 2] Find h[n ] and H (e jωˆ ) for y[n ] = x[n − 2]
x[n] y[n] bk = { 0, 0, 1 } x[n]
δ [n − 2] y[n]
h[n ] M

h[n ] = δ [n − 2] H ( e ) = ∑ δ [k − 2] e − jωˆ k
jωˆ

H ( e jω̂ )
k =0

H ( e jω̂ )
x[n] y[n]
e − j 2ω̂
x[n] y[n] k = 2 ONLY

ω̂ ω̂ ω̂ ω̂
10/6/2003 © 2003, JH McClellan & RW Schafer 6 10/6/2003 © 2003, JH McClellan & RW Schafer 7

GENERAL DELAY PROPERTY FREQ DOMAIN --> TIME ??

Find h[n ] and H (e jωˆ ) for y[n ] = x[n − nd ] ƒ START with H (e jω̂ ) and find h[n ] or bk
h[n ] = δ [n − nd ] x[n]
h[n ] y[n]
h[n ] = ?
M
H ( e ) = ∑ δ [k − nd ] e − jωˆ k = e − jωˆ nd
jωˆ
H ( e jωˆ ) = 7e − j 2ωˆ cos(ωˆ )
k =0
x[n] y[n]
ONLY ONE H ( e jω̂ )
non-ZERO TERM
for k at k = nd ω̂ ω̂
10/6/2003 © 2003, JH McClellan & RW Schafer 8 10/6/2003 © 2003, JH McClellan & RW Schafer 9
 FREQ DOMAIN --> TIME PREVIOUS LECTURE REVIEW

H ( e jωˆ ) = 7e − j 2ωˆ cos(ωˆ ) EULER’s Formula ƒ SINUSOIDAL INPUT SIGNAL

ƒ OUTPUT has SAME FREQUENCY
− j 2ωˆ jωˆ − jωˆ
= 7e (0.5e + 0.5e ) ƒ DIFFERENT Amplitude and Phase

= (3.5e − jωˆ + 3.5e − j 3ωˆ ) ƒ FREQUENCY RESPONSE of FIR

MAG
ƒ MAGNITUDE vs. Frequency
h[n ] = 3.5δ [n − 1] + 3.5δ [n − 3] ƒ PHASE vs. Freq PHASE
ƒ PLOTTING jωˆ

bk = { 0, 3.5, 0,3.5 } H ( e jωˆ ) = H (e jωˆ ) e j∠H ( e )

10/6/2003 © 2003, JH McClellan & RW Schafer 10 10/6/2003 © 2003, JH McClellan & RW Schafer 11

FREQ. RESPONSE PLOTS PLOT of FREQ RESPONSE

{bk } = {1,2,1}

π to +π
ƒ DENSE GRID (ww) from -π π
ƒ ww = -pi:(pi/100):pi;
ƒ HH = freqz(bb,1,ww) jωˆ
H (e ) = ( 2 + 2 cos ωˆ )e − jωˆ
RESPONSE at π/3
ω̂

ƒ VECTOR bb contains Filter Coefficients

ƒ DSP-First: HH = freekz(bb,1,ww)
M
H ( e ) = ∑ bk e − jωˆ k
jωˆ

k =0
π
−10/6/2003 π
10/6/2003 © 2003, JH McClellan & RW Schafer 12
ω̂ (radians)
© 2003, JH McClellan & RW Schafer 13
EXAMPLE 6.2 EXAMPLE 6.2 (answer)

Find y[n ] when H (e jωˆ ) is known Find y[n ] when x[n ] = 2e jπ / 4e j (π / 3) n

and x[n ] = 2e jπ / 4e j (π / 3) n One Step - evaluate H ( e jωˆ ) at ωˆ = π / 3
x[n] jω̂ y[n] H ( e jωˆ ) = ( 2 + 2 cos ωˆ )e − jωˆ
H (e )
ω̂ ω̂ H ( e jωˆ ) = 3e − jπ / 3 @ ωˆ = π / 3
H ( e jωˆ ) = ( 2 + 2 cos ωˆ )e − jωˆ y[n ] = (3e − jπ / 3 ) × 2e jπ / 4e j (π / 3) n = 6 e − jπ /12e j (π / 3) n
10/6/2003 © 2003, JH McClellan & RW Schafer 14 10/6/2003 © 2003, JH McClellan & RW Schafer 15

EXAMPLE: COSINE INPUT EX: COSINE INPUT (ans-1)

Find y[n ] when H ( e jωˆ ) is known Find y[n ] when x[n ] = 2 cos( π3 n + π4 )
and x[n ] = 2 cos( π3 n + π4 ) 2 cos( π3 n + π4 ) = e j (π n / 3+π / 4 ) + e − j (π n / 3+π / 4 )
x[n] y[n] ⇒ x[n ] = x1[n ] + x2 [n ]
H ( e jω̂ )
ω̂ ω̂ y1[n ] = H ( e jπ / 3 )e j (π n / 3+π / 4 )
H ( e ) = ( 2 + 2 cos ωˆ )e − jωˆ
jωˆ y2 [n ] = H ( e − jπ / 3 )e − j (π n / 3+π / 4 )
10/6/2003 © 2003, JH McClellan & RW Schafer 16 10/6/2003
⇒ y[n ] = y1[n ] + y2 [n ]
© 2003, JH McClellan & RW Schafer 17
 EX: COSINE INPUT (ans-2) SINUSOID thru FIR

Find y[n ] when x[n ] = 2 cos( π3 n + π4 ) *

ƒ IF H ( e ) = H ( e
jωˆ
) − jωˆ

ƒ Multiply the Magnitudes

H ( e jωˆ ) = ( 2 + 2 cos ωˆ )e − jωˆ
y1[n ] = H ( e jπ / 3 )e j (π n / 3+π / 4 ) = 3e − j (π / 3)e j (π n / 3+π / 4 ) ƒ Add the Phases
y2 [n ] = H ( e − jπ / 3 )e − j (π n / 3+π / 4 ) = 3e j (π / 3)e − j (π n / 3+π / 4 ) x[n ] = A cos(ωˆ1n + φ )

y[n ] = 3e j (π n / 3−π / 12 )
+ 3e − j (π n / 3−π / 12 ) ⇒ y[n ] = A H ( e jωˆ1 ) cos(ωˆ1n + φ + ∠ H ( e jωˆ1 ))
⇒ y[n ] = 6 cos( π3 n − 12
π )
10/6/2003 © 2003, JH McClellan & RW Schafer 18 10/6/2003 © 2003, JH McClellan & RW Schafer 19

LTI Demo with Sinusoids

DIGITAL “FILTERING”
y[n]
x[n]
FILTER

H ( e jω̂ )
x(t) x[n] y[n] y(t)
A-to-D D-to-A

ω ω̂ ω̂ ω
ω ƒ SPECTRUM of x(t) (SUM of SINUSOIDS)
ƒ SPECTRUM of x[n]
ω̂ ƒ Is ALIASING a PROBLEM ?
ƒ SPECTRUM y[n] (FIR Gain or Nulls)
ω ƒ Then, OUTPUT y(t) = SUM of SINUSOIDS
10/6/2003 © 2003, JH McClellan & RW Schafer 20 10/6/2003 © 2003, JH McClellan & RW Schafer 21
FREQUENCY SCALING 11-pt AVERAGER Example

x(t)
A-to-D
x[n]
H (e jω̂ ) y[n]
D-to-A
y(t) x(t)
A-to-D
x[n]
H (e jω̂ ) y[n]
D-to-A
y(t)

10
ω ωˆ ωˆ ω
ω ωˆ ωˆ ω
y[n ] = ∑ 111 x[n − k ]
ƒ TIME SAMPLING: t = nTs 250 Hz
k =0

ƒ IF NO ALIASING:
ƒ FREQUENCY SCALING
ωˆ = ωTs = ω
25 Hz H (e ) =
sin( 112 ωˆ ) − j 5ωˆ
jωˆ
11sin( 2 ωˆ )
1
e ?
fs
10/6/2003 © 2003, JH McClellan & RW Schafer 22
x (t ) = cos(2π (25)t ) + cos(2π ( 250)t − 12 π )
10/6/2003 © 2003, JH McClellan & RW Schafer 23

D-A FREQUENCY SCALING TRACK the FREQUENCIES

x(t)
A-to-D
x[n]
H (e jω̂ ) y[n]
D-to-A
y(t) x(t)
A-to-D
x[n]
H (e jω̂ ) y[n]
D-to-A
y(t)

ω ωˆ ωˆ ω ω ωˆ ωˆ ω

ƒ 250 Hz π H ( e j 0.5π ) ƒ 0.5π

ƒ 0.5π π ƒ 250 Hz
ƒ TIME SAMPLING: t = nTs ⇒ n ← t f s
ƒ RECONSTRUCT up to 0.5fs ƒ 25 Hz π H ( e j 0.05π ) ƒ .05π
ƒ .05π π ƒ 25 Hz
ƒ FREQUENCY SCALING ω = ωˆ f s Fs = 1000 Hz NO new freqs
10/6/2003 © 2003, JH McClellan & RW Schafer 24 10/6/2003 © 2003, JH McClellan & RW Schafer 25
 11-pt AVERAGER EVALUATE Freq. Response
jωˆ sin( 112 ωˆ ) − j 5ωˆ
NULLS or ZEROS H (e ) = e
11sin( 2 ωˆ )
1
At ωˆ = 0.5π
jωˆ sin( 112 (0.5π )) − j 5( 0.5π )
H (e ) = e
ωˆ = 0.05π 11sin( 12 (0.5π ))
ωˆ = 0.5π
sin(2.75π ) − j 2.5π
= e
11sin(0.25π )
= 0.0909e − j 0.5π
10/6/2003 © 2003, JH McClellan & RW Schafer 26 10/6/2003 © 2003, JH McClellan & RW Schafer 27

EVALUATE Freq. Response DIGITAL FILTER

H ( e jω̂ )

H ( e j 2π ( 25) /1000 )
MAG SCALE
fs = 1000
PHASE CHANGE EFFECTIVE RESPONSE
H ( e j 2π ( 250) /1000 ) LOW-PASS FILTER

10/6/2003 © 2003, JH McClellan & RW Schafer 28 10/6/2003 © 2003, JH McClellan & RW Schafer 29
FILTER TYPES B & W IMAGE
ƒ LOW-PASS FILTER (LPF)
ƒ BLURRING
ƒ ATTENUATES HIGH FREQUENCIES
ƒ HIGH-PASS FILTER (HPF)
ƒ SHARPENING for IMAGES
ƒ BOOSTS THE HIGHS
ƒ REMOVES DC
ƒ BAND-PASS FILTER
(BPF)

10/6/2003 © 2003, JH McClellan & RW Schafer 30 10/6/2003 © 2003, JH McClellan & RW Schafer 31

B&W IMAGE with COSINE FILTERED B&W IMAGE

FILTERED: 11-pt AVG

LPF:
BLUR

10/6/2003 © 2003, JH McClellan & RW Schafer 32 10/6/2003 © 2003, JH McClellan & RW Schafer 33
ROW of B&W IMAGE FILTERED ROW of IMAGE
BLACK = 255

WHITE = 0
ADJUSTED DELAY by 5 samples
10/6/2003 © 2003, JH McClellan & RW Schafer 34 10/6/2003 © 2003, JH McClellan & RW Schafer 35