Signall

Si
Processing
First

Copyright Monash University 2009

Lecture11
Linearity
TimeInvariance
Convolution

READING ASSIGNMENTS
ThisLecture:
Chapter5,Sections55and56
p 5,
55
5

OtherReading:
Recitation:Ch.5,Sects56,57&58
R i i Ch S
6 & 8
CONVOLUTION

NextLecture:startChapter6

Copyright Monash University 2009

LECTURE OBJECTIVES
GENERAL PROPERTIES of FILTERS
LINEARITY
TIME-IINVARIANCE
CONVOLUTION

LTI SYSTEMS

BLOCKDIAGRAMREPRESENTATION
Componentsfor
Hardware
p
ConnectSimpleFiltersTogethertoBuildMore
ComplicatedSystems
Copyright Monash University 2009

OVERVIEW
IMPULSERESPONSE, h[n ]
FIRcase:sameas
FIR

{{bbk }

CONVOLUTION
GENERAL: y[ n ] = h[ n ] x[ n ] = k h[ k ] x[ n k ]
GENERALCLASSofSYSTEMS
GENERALCLASS fSYSTEMS
LINEARandTIMEINVARIANT

ALLLTI systemshaveh[n]&useconvolution
Copyright Monash University 2009

DIGITAL FILTERING
x(t)

A-to-D

x[n]

FILTER

y[n]

D-to-A

y(t)

(
)
CONCENTRATEontheFILTER(DSP)
DISCRETETIMESIGNALS
FUNCTIONSofn,thetimeindex
f
h
d
INPUTx[n]
OUTPUTy[n]
Copyright Monash University 2009

GENERAL FIR FILTER

FILTERCOEFFICIENTS{bk}
M
DEFINETHEFILTER
y[n ] = bk x[n k ]
(k=0 CAUSAL
k =0
M<
FIR)
M
Forexample,bk = {3, 1, 2,1}

y[n ] =

bk x [ n k ]

Difference Equation

k =0

= 3 x [ n ] x [ n 1] + 2 x [ n 2 ] + x [ n 3 ]
Copyright Monash University 2009

3/23/2011

IMPULSE SIGNAL
x[n]hasonlyoneNONZEROVALUE

1 n = 0
[n ] =
0 n 0
UNIT-IMPULSE
UNIT
IMPULSE

1
n

Copyright Monash University 2009

FIR IMPULSE RESPONSE

Convolution=FilterDefinition
FilterCoeffs=ImpulseResponse
p
p

M
M

hy[n ] = bkkx[n k ]
kk==00

Copyright Monash University 2009

Use SHIFTED
IMPULSES to
write h[n]
8

MATH FORMULA for h[n]

UseSHIFTED IMPULSEStowriteh[n]
h[n ] = [n ] [n 1] + 2 [n 2] [n 3] + [n 4]
2

h[n ]
1

1
1

3
2

bk = { 1, 1, 2, 1, 1 }
y [ n ] = x[ n ] x[ n 1] + 2 x[ n 2] x[ n 3] + x[ n 4]
Copyright Monash University 2009

Convolution Sum
Output=Convolutionofx[n]&h[n]
y[n ] = h[n ] x[n ]
NOTATION:
HereistheFIRcase:
FINITE LIMITS

y[n ] =

* is convolution
not multiplication

h[ k ] x[ n k ] =: h[ n ] * x[ n ]
k =0

Same as bk
Copyright Monash University 2009

FINITE LIMITS
10

CONVOLUTION Example
p
M

y[n] = h[k ] x[n k ] = h[0] x[n] + h[1] x[n 1] + L + h[ M ] x[n M ]

k =0

Finite

Finite y[n]
Copyright Monash University 2009

3/23/2011

11

CONVOLUTION Commutes
M
M

y[n] = hx[k ]hx[n k ] = hx[[0]]hx[n] + hx[[1]]hx[n 1]] + L + hx[ M ]hx[n M ]

k =0

y [ n ] = h[ k ] x[ n k ] = h[ n ] * x[ n ]
x[0]h[n-0]
[ ] [ ]
x[1]h[n-1]
x[2]h[n-2]
x[3]h[n-3]
x[3]h[n
3]
x[4]h[n-4]

|
|
|
|
|

k =0
6

-2
12

4
-4
18

2
8
-6
2

4
12
-4
4
6

= x[ k ]h[ n k ] = x[4n ] * h[ n ]
k =0

6
8
-2

3/23/2011
Copyright Monash University 2009

12

h[ n ] = [ n ] [ n 1] + 2 [ n 2] [ n 3] + [ n 4]
x[ n ] = u [ n ]

CONVOLUTION Example
4

y[ n ] = h[ n ] * x[ n ] = h[ k ]u[ n k ]
k =0

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
0 1 1
1 1 1 1 1 1
h[0]x[n ]
0 0 1 1 1 1 1 1 1
h[1]x[n 1]

h[2]x[n 2]
h[3]x[n 3]
h[4]x[n 4]
y[n]

0 0
0 0
0 0
0 1

0
0
0
0

IInfinite
fi i
x[n]

2 2 2 2 2 2
0 1
1 1
1 1
1 1
1 1
1
0 0
1 1 1 1 Infinite
2
1 2 2 2 ... y[n]
3/23/2011

Copyright Monash University 2009

13

GENERAL FIR FILTER

SLIDEaLengthLWINDOWoverx[n]
g
L=M+1

x[n-M]
Copyright Monash University 2009

x[n]
14

MATLAB for FIR FILTER

xx)
yy = conv(bb
conv(bb,xx)
VECTORbb containsFilterCoefficients
SPFirst:yy = firfilt(bb,xx)
(
)
Previousexample
bb=[3
[ -2 2 1];
] xx=[2
[ 4 6 4 2];
]

yy = firfilt(bb,xx)
ans = [6 10 18 16 18 12 8 2]
Copyright Monash University 2009

15

Polynomial
y
product
p
and convolution
p ( x ) = 1 2 x + x 2 deg( p ) = 2
q ( x ) = 1 x deg(
d ( p) = 1
p ( x ) q ( x ) = 1 x + 3 x 2 x 3 deg( pq ) = 3
x[ n] = {..., 0, 0, 1, 2,1, 0, 0,...}
h[ n] = {..., 0, 0,1, 1, 0, 0,...}

L1 = 3 ( M 1 = 2)

L2 = 2 ( M 2 = 2)

y[ n] = h[ n]* x[ n] = h[ k ]u[ n k ] = {..., 00, 00,1,

1 11, 33, 11, 00, 00,...}
k =0

L1 + L2 1 = 4

( M 1 + M 2 1 = 3)

([ -2 1],[1
],[ -1])
]) =>
>> conv([-1
Copyright Monash University 2009

[ -1

-1

-1]]
16

DCONVDEMO: MATLAB GUI

3/23/2011
Copyright Monash University 2009

17

POP QUIZ
FIRFilterisFIRSTDIFFERENCE
FIRFilteris
FIRSTDIFFERENCE
y[n]=x[n] x[n1]

INPUTisUNITSTEP
1 n 0
u[n ] =
0 n < 0

Find y[n]

y[n ] = u[n ] u[n 1] = [n ]

3/23/2011

Copyright Monash University 2009

18

HARDWARE STRUCTURES
M

x[n]

FILTER

y[n]

y[n ] = bk x[n k ]
k =0

INTERNALSTRUCTURE
ofFILTER
S
C
f
WHATCOMPONENTSARENEEDED?
HOWDOWEHOOKTHEMTOGETHER?

SIGNALFLOWGRAPHNOTATION

Copyright Monash University 2009

19

HARDWARE ATOMS
Multiply,Store&Add

y[n ] = bk x[n k ]
k =0

y[n] = x[n]

y[n] = x1[n] + x2[n]

y[n ] = x[n 1]
Copyright Monash University 2009

20

FIR STRUCTURE
S UC U
DirectForm
SIGNAL
FLOW GRAPH

y[n ] =

bk x[ n k ]

k =0

Stored sequence

Copyright Monash University 2009

21

FIR TRANSPOSE
S OS S
STRUCTURE
UC U
y[n ] =

bk x[ n k ]

k =0

x[n]
X

b3

b2
Unit
Delay

Copyright Monash University 2009

b1

Unit
Delay

b0

Unit
Delay

y[n]

22

SYSTEM PROPERTIES
x[n]

SYSTEM

y[n]

MATHEMATICALDESCRIPTION
TIMEINVARIANCE
TIME
LINEARITY
CAUSALITY
Nooutputpriortoinput

Copyright Monash University 2009

23

TIME INVARIANCE
TIME-INVARIANCE
IDEA:
TimeShiftingtheinputwillcausethesame
g
p
timeshiftintheoutput

EQUIVALENTLY,
Wecanprovethat
Thetimeorigin(n=0)ispickedarbitrary

Copyright Monash University 2009

24

Time Invariance
TESTING Time-Invariance

Copyright Monash University 2009

25

1, n = 0
,
Q [n] =
0 n0
0,

2003, JH McClellan & RW Schafer

n [n] = 0 for all n

3/23/2011
Copyright Monash University 2009

26

LINEAR SYSTEM
LINEARITY=TwoProperties
SCALING
Doublingx[n]willdoubley[n]

SUPERPOSITION:
Addingtwoinputsgivesanoutputthatisthe
sumoftheindividualoutputs
sumoftheindividualoutputs

Copyright Monash University 2009

27

TESTING LINEARITY

3/23/2011
Copyright Monash University 2009

28

Ex: time-flip system y[n]=x[-n] is a

linear system

3/23/2011
Copyright Monash University 2009

29

LTI SYSTEMS
LTI:Linear&TimeInvariant
COMPLETELYCHARACTERIZEDby:
COMPLETELYCHARACTERIZEDby
IMPULSERESPONSE h[n]
CONVOLUTION:y[n]=x[n]*h[n]
The
TheruledefiningthesystemcanALWAYSbere
rule definingthesystemcanALWAYSbere
writtenasconvolution

FIRExample:h[n]issameasbk
Exampletransparence

Copyright Monash University 2009

30

LTI syste
system a
and
d co
convolution
o ut o
y[n]=LTI(x[n])h[n]=LTI([n])
y[n] LTI(x[n])h[n] LTI([n])
yy[n]=LTI(
x[k][nk])impulse
p
repres.of
p
x[n]
y[n]=LTI(x[k] [nk])linearity
y[n]=x[k] LTI([nk])linearity
y[n]=x[k] h[nk]time invariance
y[n]=x[n]*h[n]
Copyright Monash University 2009

3/23/2011

31

CASCADE SYSTEMS
DoestheorderofS1 &S2 matter?
NO,LTISYSTEMScanberearranged!!!
,
g
WHATARETHERESULTINGFILTERCOEFFS?
{bk}ORh[n]

S1
Copyright Monash University 2009

S2
32

CASCADE
C
SC
EQUIVALENT
QU
Findoverallh[n]foracascade?

S1

S2

h[ n ] = h1 [ n ] * h2 [ n ]
= h2 [ n ] * h1 [ n ]

S2
Copyright Monash University 2009

S1

3/23/2011

33

03
1
h1[n]
0 otherwise

Example

1 3
1
h 2 [[n]]
h
i
0 otherwise
h[n]=h
[ ] 1[[n]*h
] 2 [[n]]
6

y[n]= h[k]x[n-k]
k=0

or
3

w[n]= x[n-k]
k=0

3/23/2011
Copyright
Monash University 2009

y[n]= w[n-k]
k=1

5 +, 4 x
5 + only
only, no x
34

BUILDING BLOCKS
OUTPUT

x[n]
FILTER

FILTER

y[n]

INPUT

FILTER

BUILDUPCOMPLICATEDFILTERS
FROMSIMPLEMODULES
Eg

FILTER = 3x[n]-x[n-1] +2x[n-2]+x[n-3]

Copyright Monash University 2009

35