READING ASSIGNMENTS

Signal Processing First

ƒ This Lecture:
ƒ Chapter 5, Sections 5-5 and 5-6
ƒ Section 5-4 will be covered, but not “in depth”

Lecture 11
Linearity & Time-Invariance ƒ Other Reading:
ƒ Recitation: Ch. 5, Sects 5-6, 5-7 & 5-8
Convolution
ƒ CONVOLUTION
ƒ Next Lecture: start Chapter 6

8/22/2003 © 2003, JH McClellan & RW Schafer 1 8/22/2003 © 2003, JH McClellan & RW Schafer 3

LECTURE OBJECTIVES OVERVIEW

ƒ GENERAL PROPERTIES of FILTERS ƒ IMPULSE RESPONSE, h[n ]
ƒ LINEARITY LTI SYSTEMS ƒ FIR case: same as {bk }
ƒ TIME-INVARIANCE
ƒ CONVOLUTION
ƒ ==> CONVOLUTION
ƒ GENERAL: y[n ] = h[n ] ∗ x[n ]
ƒ BLOCK DIAGRAM REPRESENTATION ƒ GENERAL CLASS of SYSTEMS
ƒ Components for Hardware ƒ LINEAR and TIME-INVARIANT
ƒ Connect Simple Filters Together to Build More
Complicated Systems
ƒ ALL LTI systems have h[n] & use convolution

8/22/2003 © 2003, JH McClellan & RW Schafer 4 8/22/2003 © 2003, JH McClellan & RW Schafer 5
DIGITAL FILTERING BUILDING BLOCKS

x(t) x[n] y[n] y(t) x[n] OUTPUT y[n]

A-to-D FILTER D-to-A FILTER FILTER +
INPUT +

ƒ CONCENTRATE on the FILTER (DSP) FILTER

ƒ DISCRETE-TIME SIGNALS
ƒ FUNCTIONS of n, the “time index” ƒ BUILD UP COMPLICATED FILTERS
ƒ INPUT x[n] ƒ FROM SIMPLE MODULES
ƒ OUTPUT y[n] ƒ Ex: FILTER MODULE MIGHT BE 3-pt FIR

8/22/2003 © 2003, JH McClellan & RW Schafer 6 8/22/2003 © 2003, JH McClellan & RW Schafer 7

GENERAL FIR FILTER MATLAB for FIR FILTER

ƒ FILTER COEFFICIENTS {bk} ƒ yy = conv(bb,xx)
ƒ DEFINE THE FILTER M
ƒ VECTOR bb contains Filter Coefficients
y[n ] = ∑ bk x[n − k ]
k =0 ƒ DSP-First: yy = firfilt(bb,xx)

ƒ For example, bk = {3, − 1, 2,1} ƒ FILTER COEFFICIENTS {bk} conv2()

3 for images
M
y[n ] = ∑ bk x[n − k ]
k =0
y[n ] = ∑ bk x[n − k ]
= 3x[n ] − x[n − 1] + 2 x[n − 2] + x[n − 3] k =0
8/22/2003 © 2003, JH McClellan & RW Schafer 8 8/22/2003 © 2003, JH McClellan & RW Schafer 9
 SPECIAL INPUT SIGNALS FIR IMPULSE RESPONSE

ƒ x[n] = SINUSOID FREQUENCY RESPONSE ƒ Convolution = Filter Definition

ƒ x[n] has only one NON-ZERO VALUE ƒ Filter Coeffs = Impulse Response
1 n = 0
δ [n ] = 
0 n ≠ 0
UNIT-IMPULSE
1 M
M
hy[n ] = ∑ bkkδx[n − k ]
n kk==00
8/22/2003 © 2003, JH McClellan & RW Schafer 10 8/22/2003 © 2003, JH McClellan & RW Schafer 11

MATH FORMULA for h[n] LTI: Convolution Sum

ƒ Use SHIFTED IMPULSES to write h[n] ƒ Output = Convolution of x[n] & h[n]
h[n ] = δ [n ] − δ [n − 1] + 2δ [n − 2] − δ [n − 3] + δ [n − 4] ƒ NOTATION: y[n ] = h[n ] ∗ x[n ]
ƒ Here is the FIR case:
2 FINITE LIMITS
h[n ]
1 M

0 4 n
y[n ] = ∑ h[k ]x[n − k ]
k =0
–1
bk = { 1, − 1, 2, − 1, 1 } Same as bk FINITE LIMITS
8/22/2003 © 2003, JH McClellan & RW Schafer 12 8/22/2003 © 2003, JH McClellan & RW Schafer 13
CONVOLUTION Example GENERAL FIR FILTER
h[n ] = δ [n ] − δ [n − 1] + 2δ [n − 2] − δ [n − 3] + δ [n − 4] ƒ SLIDE a Length-L WINDOW over x[n]
x[n ] = u[n ]
n −1 0 1 2 3 4 5 6 7
x[n] 0 1 1 1 1 1 1 1 ...
h[n] 0 1 −1 2 −1 1 0 0 0
h[0] x[n ] 0 1 1 1 1 1 1 1 1
h[1] x[n − 1] 0 0 −1 −1 −1 −1 −1 −1 −1
h[2]x[n − 2] 0 0 0 2 2 2 2 2 2
h[3]x[n − 3] 0 0 0 0 −1 −1 −1 −1 −1
h[4]x[n − 4] 0 0 0 0 0 1 1 1 1 x[n-M] x[n]
8/22/2003 y[n] 0 1 0 2 1 2
© 2003, JH McClellan & RW Schafer 2 2 ... 14 8/22/2003 © 2003, JH McClellan & RW Schafer 15

DCONVDEMO: MATLAB GUI POP QUIZ

ƒ FIR Filter is “FIRST DIFFERENCE”
ƒ y[n] = x[n] - x[n-1]
ƒ INPUT is “UNIT STEP”
1 n ≥ 0
u[n ] = 
0 n < 0

ƒ Find y[n]
y[n ] = u[n ] − u[n − 1] = δ [n ]
8/22/2003 © 2003, JH McClellan & RW Schafer 16 8/22/2003 © 2003, JH McClellan & RW Schafer 17
HARDWARE STRUCTURES HARDWARE ATOMS
M ƒ Add, Multiply & Store M
x[n]
FILTER
y[n]
y[n ] = ∑ bk x[n − k ] y[n ] = ∑ bk x[n − k ]
k =0
k =0

ƒ INTERNAL STRUCTURE of “FILTER”

ƒ WHAT COMPONENTS ARE NEEDED?
y[n ] = x1[n ] + x2 [n ]
ƒ HOW DO WE “HOOK” THEM TOGETHER? y[n ] = β x[n ]
ƒ SIGNAL FLOW GRAPH NOTATION

8/22/2003 © 2003, JH McClellan & RW Schafer 18 8/22/2003

y[n ] = x[n − 1]
© 2003, JH McClellan & RW Schafer 19

Moore’s Law for TI DSPs

FIR STRUCTURE

ƒ Direct Form M LOG SCALE

y[n ] = ∑ bk x[n − k ]
SIGNAL
FLOW GRAPH k =0

Double every
18 months ?

8/22/2003 © 2003, JH McClellan & RW Schafer 20 8/22/2003 © 2003, JH McClellan & RW Schafer 21
SYSTEM PROPERTIES TIME-INVARIANCE

x[n]
SYSTEM
y[n] ƒ IDEA:
ƒ “Time-Shifting the input will cause the same
time-shift in the output”
ƒ MATHEMATICAL DESCRIPTION
ƒ TIME-INVARIANCE
ƒ EQUIVALENTLY,
ƒ LINEARITY
ƒ We can prove that
ƒ CAUSALITY ƒ The time origin (n=0) is picked arbitrary
ƒ “No output prior to input”

8/22/2003 © 2003, JH McClellan & RW Schafer 22 8/22/2003 © 2003, JH McClellan & RW Schafer 23

TESTING Time-Invariance LINEAR SYSTEM

ƒ LINEARITY = Two Properties

ƒ SCALING
ƒ “Doubling x[n] will double y[n]”

ƒ SUPERPOSITION:
ƒ “Adding two inputs gives an output that is the
sum of the individual outputs”

8/22/2003 © 2003, JH McClellan & RW Schafer 24 8/22/2003 © 2003, JH McClellan & RW Schafer 25
TESTING LINEARITY LTI SYSTEMS

ƒ LTI: Linear & Time-Invariant

ƒ COMPLETELY CHARACTERIZED by:
ƒ IMPULSE RESPONSE h[n]
ƒ CONVOLUTION: y[n] = x[n]*h[n]
ƒ The “rule”defining the system can ALWAYS be re-
written as convolution
ƒ FIR Example: h[n] is same as bk

8/22/2003 © 2003, JH McClellan & RW Schafer 26 8/22/2003 © 2003, JH McClellan & RW Schafer 27

POP QUIZ CASCADE SYSTEMS

ƒ FIR Filter is “FIRST DIFFERENCE”
ƒ y[n] = x[n] - x[n -1]
ƒ Does the order of S1 & S2 matter?
ƒ Write output as a convolution ƒ NO, LTI SYSTEMS can be rearranged !!!
ƒ Need impulse response ƒ WHAT ARE THE FILTER COEFFS? {bk}
h[n ] = δ [n ] − δ [n − 1]
ƒ Then, another way to compute the output:

y[n ] = (δ [n ] − δ [n − 1]) ∗ x[n ]

S1 S2

8/22/2003 © 2003, JH McClellan & RW Schafer 28 8/22/2003 © 2003, JH McClellan & RW Schafer 29
CASCADE EQUIVALENT
ƒ Find “overall” h[n] for a cascade ?

S1 S2

S2 S1
8/22/2003 © 2003, JH McClellan & RW Schafer 30