www.rejinpaul.

com

UNIT I

www.rejinpaul.com
 www.rejinpaul.com

SIGNAL
Signal is a physical quantity that varies with
respect to time , space or any other
independent variable
Eg x(t)= sin t.
the major classifications of the signal
are:
(i) Discrete time signal
(ii) Continuous time signal

www.rejinpaul.com
 www.rejinpaul.com

Unit Step &Unit Impulse

Discrete time Unit impulse is defined as
[n]= {0, n 0
{1, n=0
Unit impulse is also known as unit sample.
Discrete time unit step signal is defined by
U[n]={0,n=0
{1,n>= 0
Continuous time unit impulse is defined as
(t)={1, t=0
{0, t 0
Continuous time Unit step signal is defined as
U(t)={0, t<0
{1, t0

www.rejinpaul.com
 www.rejinpaul.com

SIGNAL

Periodic Signal & Aperiodic Signal

A signal is said to be periodic ,if it exhibits periodicity.i.e.,
X(t +T)=x(t), for all values of t. Periodic signal has the
property that it is unchanged by a time shift of T. A signal
that does not satisfy the above periodicity property is
called an aperiodic signal
even and odd signal ?

A discrete time signal is said to be even when, x[-n]=x[n].
The continuous time signal is said to be even when, x(-
t)= x(t) For example,Cosn is an even signal.

www.rejinpaul.com
 www.rejinpaul.com

Energy and power signal

A signal is said to be energy signal if it

have finite energy and zero power.
A signal is said to be power signal if it
have infinite energy and finite power.
If the above two conditions are not
satisfied then the signal is said to be
neigther energy nor power signal

www.rejinpaul.com
 www.rejinpaul.com

Fourier Series
The Fourier series represents a periodic signal in terms of
frequency components:

p 1
ik0n
x(n) = Xk e x( t ) = Xk e ik0 t
k =0 k =
We get the Fourier series coefficients as follows:

1 p 1 1p
Xk = x ( n )e ik 0n X k = x ( t )e ik 0 t dt
p n=0 p0
The complex exponential Fourier coefficients are a sequence of
complex numbers representing the frequency component 0k.

www.rejinpaul.com
 www.rejinpaul.com

Fourier series

Fourier series: a complicated waveform analyzed into a

number of harmonically related sine and cosine functions

A continuous periodic signal x(t) with a period T0 may be

represented by:

X(t)=k=1 (Ak cos k t + Bk sin k t)+ A0

Dirichlet conditions must be placed on x(t) for the series

to be valid: the integral of the magnitude of x(t) over a
complete period must be finite, and the signal can only
have a finite number of discontinuities in any finite
interval

www.rejinpaul.com
 www.rejinpaul.com

Trigonometric form for Fourier series

If the two fundamental components of a

periodic signal areB1cos0t and C1sin0t,
then their sum is expressed by trigonometric
identities:
X(t)= A0 + k=1 ( Bk 2+ Ak 2)1/2 (Ck cos k t-
k) or
X(t)= A0 + k=1 ( Bk 2+ Ak 2)1/2 (Ck sin k t+
k)

www.rejinpaul.com
 www.rejinpaul.com

UNIT II

www.rejinpaul.com
 www.rejinpaul.com

Fourier Transform

Viewed periodic functions in terms of frequency components (Fourier

series) as well as ordinary functions of time
Viewed LTI systems in terms of what they do to frequency
components (frequency response)
Viewed LTI systems in terms of what they do to time-domain signals
(convolution with impulse response)
View aperiodic functions in terms of frequency components via
Fourier transform
Define (continuous-time) Fourier transform and DTFT
Gain insight into the meaning of Fourier transform through
comparison with Fourier series

www.rejinpaul.com
 www.rejinpaul.com

The Fourier Transform

A transform takes one function (or signal)
and turns it into another function (or signal)
Continuous Fourier Transform:

H ( f ) = h (t )e 2ift dt


h (t ) = H ( f )e 2ift df

www.rejinpaul.com
 www.rejinpaul.com

Continuous Time Fourier Transform

We can extend the formula for continuous-time Fourier series
coefficients for a periodic signal

1p 1 p/2
X k = x ( t )e ik 0 t dt = x ( t ) e ik 0 t
dt
p0 p p / 2
to aperiodic signals as well. The continuous-time Fourier
series is not defined for aperiodic signals, but we call the
formula

X() = x( t )e i t dt the (continuous time)
Fourier transform.

www.rejinpaul.com
 www.rejinpaul.com

Inverse Transforms
If we have the full sequence of Fourier coefficients for a periodic
signal, we can reconstruct it by multiplying the complex
sinusoids of frequency 0k by the weights Xk and summing:

p 1
x(n) = Xk e ik0n x( t ) = Xk eik0 t
k =0 k =

We can perform a similar reconstruction for aperiodic signals

1 i n 1 i t
x(n) =
2
X()e d x( t ) =
2
X()e d

These are called the inverse transforms.

www.rejinpaul.com
 www.rejinpaul.com

Fourier Transform of Impulse Functions

Find the Fourier transform of the Dirac delta function:

it it i0
X() = x ( t )e dt = ( t )e dt = e =1

Find the DTFT of the Kronecker delta function:

in in i0
X( ) = x(n)e = (n)e =e =1
n = n =
The delta functions contain all frequencies at equal amplitudes.
Roughly speaking, thats why the system response to an impulse
input is important: it tests the system at all frequencies.

www.rejinpaul.com
 www.rejinpaul.com

Laplace Transform
Lapalce transform is a generalization of the Fourier transform in the sense
that it allows complex frequency whereas Fourier analysis can only
handle real frequency. Like Fourier transform, Lapalce transform allows
us to analyze a linear circuit problem, no matter how complicated the
circuit is, in the frequency domain in stead of in he time domain.
Mathematically, it produces the benefit of converting a set of differential
equations into a corresponding set of algebraic equations, which are much
easier to solve. Physically, it produces more insight of the circuit and
allows us to know the bandwidth, phase, and transfer characteristics
important for circuit analysis and design.
Most importantly, Laplace transform lifts the limit of Fourier analysis to
allow us to find both the steady-state and transient responses of a linear
circuit. Using Fourier transform, one can only deal with he steady state
behavior (i.e. circuit response under indefinite sinusoidal excitation).
Using Laplace transform, one can find the response under any types of
excitation (e.g. switching on and off at any given time(s), sinusoidal,
impulse, square wave excitations, etc.

www.rejinpaul.com
 www.rejinpaul.com

Laplace Transform

www.rejinpaul.com
 www.rejinpaul.com

Application of Laplace Transform to

Circuit Analysis

www.rejinpaul.com
 www.rejinpaul.com

system

A system is an operation that transforms

input signal x into output signal y.

www.rejinpaul.com
 www.rejinpaul.com

LTI Digital Systems

Linear Time Invariant
Linearity/Superposition:
If a system has an input that can be
expressed as a sum of signals, then the
response of the system can be expressed as
a sum of the individual responses to the
respective systems.
LTI

www.rejinpaul.com
 www.rejinpaul.com

Time-Invariance &Causality

Ifyou delay the input, response is just a delayed

version of original response.

X(n-k) y(n-k)

Causality could also be loosely defined by there is

no output signal as long as there is no input
signal or output at current time does not depend
on future values of the input.

www.rejinpaul.com
 www.rejinpaul.com

Convolution
The input and output signals for LTI
systems have special relationship in terms
of convolution sum and integrals.

Y(t)=x(t)*h(t) Y[n]=x[n]*h[n]

www.rejinpaul.com
 www.rejinpaul.com

UNIT III

www.rejinpaul.com
 www.rejinpaul.com

Sampling theory
The theory of taking discrete sample values (grid of color
pixels) from functions defined over continuous domains
(incident radiance defined over the film plane) and then
using those samples to reconstruct new functions that are
similar to the original (reconstruction).
Sampler: selects sample points on the image plane
Filter: blends multiple samples together

www.rejinpaul.com
 www.rejinpaul.com

Sampling theory
For band limited function, we can just
increase the sampling rate
However, few of interesting functions in
computer graphics are band limited, in
particular, functions with discontinuities.
It is because the discontinuity always falls
between two samples and the samples
provides no information of the discontinuity.

www.rejinpaul.com
 www.rejinpaul.com

Sampling theory

www.rejinpaul.com
 www.rejinpaul.com

Aliasing

www.rejinpaul.com
 www.rejinpaul.com

Z-transforms
Fordiscrete-time systems, z-transforms play
the same role of Laplace transforms do in
continuous-time systems
Bilateral Forward z-transform Bilateral Inverse z-transform

1
H [ z] =
n =
h[n ] z n h[n] =
2 j R
H [ z ] z n +1dz

Aswith the Laplace transform, we compute

forward and inverse z-transforms by use of
transforms pairs and properties

www.rejinpaul.com
 www.rejinpaul.com

Region of Convergence
Region of the complex Four possibilities (z=0
z-plane for which is a special case and
forward z-transform may or may not be
converges Im{z} included) Im{z}

Entire Disk
plane Re{z} Re{z}

Im{z} Im{z}

Intersection
Complement of a disk and
of a disk Re{z} complement Re{z}

of a disk

www.rejinpaul.com
 www.rejinpaul.com

Z-transform Pairs
h[n] = [n] h[n] = an u[n]
0

[n] z = [n] z = 1

a u[n] z
n n
H [ z] = H [ z] = n n
n = n =0
n =
Region of convergence:
a
n

= a n z n =
entire z-plane n =0 n =0 z

1 a
= if <1
h[n] = [n-n1]1 1
a z
H [ z ] = [n 1] z = [n 1] z n = z 1 z
n = n =1

Region of convergence: Region of

entire z-plane convergence: |z|
h[n-1] z-1 H[z] > |a| which is
the complement
www.rejinpaul.com
of a disk
 www.rejinpaul.com

Stability
Rule #1: For a causal sequence, poles are
inside the unit circle (applies to z-transform
functions that are ratios of two polynomials)
Rule #2: More generally, unit circle is
included in region of convergence. (In
continuous-time, the imaginary axis would
be in the region of convergence
n
Z 1
of the
a u[n] for z > a
Laplace transform.) 1
1 a z

This is stable if |a| < 1 by rule #1.

It is stable if |z| > |a| and |a| < 1 by rule #2.
www.rejinpaul.com
 www.rejinpaul.com

Inverse z-transform
c + j
1
f [n] =
2j c j
F [z ]z n 1dz

Yuk! Using the definition requires a contour

integration in the complex z-plane.
Fortunately, we tend to be interested in only
a few basic signals (pulse, step, etc.)

Virtually all of the signals well see can be built
up from these basic signals.

For these common signals, the z-transform pairs
have been tabulated (see Lathi, Table 5.1)
www.rejinpaul.com
 www.rejinpaul.com

Example
X [ z] =
z 2 + 2z +1 Ratio of polynomial z-
3 1
z2 z +
2 2
domain functions
1 + 2 z 1 + z 2 Divide through by the
X [ z] =
3 1
1 z 1 + z 2 highest power of z
2 2
Factor denominator into
1 + 2 z 1 + z 2
X [ z] = first-order factors
1 1 1
1 z 1 z( )
2 Use partial fraction
X [ z ] = B0 +
A1
+
A2 decomposition to get
1 1 1 z 1
1 z
2
first-order terms

www.rejinpaul.com
 www.rejinpaul.com

Example (cont)
2
1 2 3 1
z z + 1 z 2 + 2 z 1 + 1 Find B0 by
2 2
z 2 3z 1 + 2 polynomial division
5 z 1 1
1 + 5 z 1
X [ z] = 2 +
1 1 1
Express in terms of
1 z 1 z ( )
2 B0
1 + 2 z 1 + z 2 1+ 4 + 4
A1 = = = 9
1 z 1 z 1 = 2
1 2
Solve for A1 and A2
1 + 2 z 1 + z 2 1+ 2 +1
A2 = = =8
1 1
1 z 1
2 z 1 =1 2

www.rejinpaul.com
 www.rejinpaul.com

Example (cont)
Express X[z] in terms of B0, A1, and A2
9 8
X [ z] = 2 +
1 1 1 z 1
1 z
2
Use table to obtain inverse z-transform
n
1
x[n] = 2 [n] 9 u[n] + 8 u[n]
2

With the unilateral z-transform, or the

bilateral z-transform with region of
convergence, the inverse z-transform is
unique
www.rejinpaul.com
 www.rejinpaul.com

Z-transform Properties
Linearity
a1 f1 [n] + a2 f 2 [n] a1 F1 [z ] + a2 F2 [z ]

Right shift (delay)

f [n m] u[n m] z m F [z ]
m
f [n m] u[n ] z F [z ] + z f [ n]z n
m m

n =1

www.rejinpaul.com
 www.rejinpaul.com

Z-transform Properties
Convolution definition

f1 [n] f 2 [n] = f [m] f [n m]
m =
1 2
Take z-transform

Z { f1 [n] f 2 [n]} = Z f1 [m] f 2 [n m] Z-transform definition
m =


= f1 [m] f 2 [n m] z n Interchange summation
n = m =

Substitute r = n - m
= f [m] f [n m]z
m =
1
n =
2
n

Z-transform definition
=
m =
f1 [m] f 2 [r ]z (r + m )
r =

m


= f1 [m]z f 2 [r ]z r
m= r =
= F1 [z ]F2 [z ]

www.rejinpaul.com
 www.rejinpaul.com

UNIT IV

www.rejinpaul.com
 www.rejinpaul.com

Introduction
Impulse response h[n] can fully characterize a LTI
system, and we can have the output of LTI system as

y[n] = x[n ] h[n]

The z-transform of impulse response is called transfer or
system function H(z).
Y (z ) = X (z )H (z ).
Frequency response at H e j = H (z ) z =1 is valid if
( )
ROC includes
z = 1, and
Y (e j ) = X (e j )H (e j )

www.rejinpaul.com
 www.rejinpaul.com

5.1 Frequency Response of LIT

System
j j jX ( e j ) j j jH ( e j )
Consider X (e ) = X (e ) e and H (e ) = H (e ) e
, then

magnitude
Y (e j ) = X (e j ) H (e j )

phase
Y (e j ) = X ( e j ) + H ( e j )

We will model and analyze LTI systems based on the

magnitude and phase responses.

www.rejinpaul.com
 www.rejinpaul.com

System Function

General formN of LCCDE M

a y[n k ] = b x[n k ]
k =0
k
k =0
k

N M

k
a z
k =0
k
Y ( z ) = k X (z )
b z k

k =0
Compute the z-transform
M

Y (z ) k
b z
k =0
k

H (z ) = = N
X (z )
k
a z
k =0
k

www.rejinpaul.com
 www.rejinpaul.com

System Function: Pole/zero

Factorization
Stability requirement can be verified.
Choice of ROC determines causality.
Location of zeros and poles determines the
frequency response
M
1 and phase
zeros : c , c ,..., c .
k
b0
(1 c z ) 1 2 M
k =1
H (z ) = N
a0 1 poles : d1 , d 2 ,..., d N .
(1 d z )
k =1
k

www.rejinpaul.com
 www.rejinpaul.com

Second-order System
Suppose the system function of a LTI system is
(1 + z 1 )2
H ( z) = .
1 1 3 1
(1 z )(1 + z )
2 4

Tofind the difference equation that is satisfied by

the input and out of this system
(1 + z 1 )2 1 + 2 z 1 + z 2 Y ( z)
H ( z) = = =
1 3 1 3
(1 z 1 )(1 + z 1 ) 1 + z 1 z 2 X ( z )
2 4 4 8
1 3
y[n ] + y[n 1] y[n 2] = x[n ] + 2 x[n 1] + 2 x[n 2]
4 8

Can we know the impulse response?

www.rejinpaul.com
 www.rejinpaul.com

System Function: Stability

Stability of LTI system:

h[n] <
n =

This condition is identical to the condition

that
h[ n
n =
] z n
< when z = 1.

The stability condition is equivalent to the

condition that the ROC of H(z) includes the unit
circle.
www.rejinpaul.com
 www.rejinpaul.com

System Function: Causality

If the system is causal, it follows that h[n] must be a right-
sided sequence. The ROC of H(z) must be outside the
outermost pole.
If the system is anti-causal, it follows that h[n] must be a
left-sided sequence. The ROC of H(z) must be inside the
innermost pole.
Im Im Im

a 1 a 1 Re a b
Re Re

Right-sided Left-sided Two-sided

(causal) (anti-causal) (non-causal)
www.rejinpaul.com
 www.rejinpaul.com

Determining the ROC

Consider the LTI system
5
y[ n ] y[n 1] + y[n 2] = x[n ]
2

The system1 function is obtained as

H ( z) =
5 1 2
1 z +z
2
1
=
1
(1 z 1 )(1 2 z 1 )
2

www.rejinpaul.com
 www.rejinpaul.com

System Function: Inverse Systems

H i (z ) H (z )
is an inverse system for , if

G ( z ) = H ( z ) H i ( z ) = 1 g [n ] = h[n ] hi [n ] = [n ]

1 1
Hi ( z) = j
H i (e ) =
H ( z) H ( e j )
The ROCs of H ( z ) and H i ( z ) overlap.
must
Useful for canceling the effects of another system
See the discussion in Sec.5.2.2 regarding ROC

www.rejinpaul.com
 www.rejinpaul.com

All-pass System
A system of the form (or cascade of these)
1
z a pole : a = re j
H Ap (Z ) =
1 az 1 zero : 1 / a* = r 1e j

j j
e a 1 a * e
H Ap (e j ) = j
= e j
j
1 ae 1 ae

H Ap (e j ) = 1

www.rejinpaul.com
 www.rejinpaul.com

All-pass System: General Form

In general, all pass systems have form

Mr 1 Mc 1 * 1
z dk ( z e )( z ek )
k
H Ap (z ) = 1 1 * 1
k =1 1 d k z k =1 (1 ek z )(1 ek z )

real poles complex poles

Causal/stable: ek , d k < 1

www.rejinpaul.com
 www.rejinpaul.com

All-Pass System Example

Im
Unit
circle z-plane
M r = 2 and M c = 1

0.8
Re
4 3 0.5
2
3 4

pole : re j reciprocal
& conjugate
zero : r 1e j

This all - pass system has M = N = 2 M c + M r = 4 poles and zeros.

www.rejinpaul.com
 www.rejinpaul.com

Minimum-Phase System
Minimum-phase system: all zeros and all poles are
inside the unit circle.
The name minimum-phase comes from a property of the
phase response (minimum phase-lag/group-delay).
Minimum-phase systems have some special properties.
When we design a filter, we may have multiple choices to
satisfy the certain requirements. Usually, we prefer the
minimum phase which is unique.
All systems can be represented as a minimum-phase
system and an all-pass system.

www.rejinpaul.com
 www.rejinpaul.com

UNIT V

www.rejinpaul.com
 www.rejinpaul.com

Example
y[n] = a1y[n 1] + a2y[n 2] + b0x[n]

Block diagram representation of

www.rejinpaul.com
 www.rejinpaul.com

Block Diagram Representation

LTI systems with
rational system
function can be
represented as
constant-coefficient
difference equation
The implementation of
difference equations
requires delayed
values of the

input www.rejinpaul.com
 www.rejinpaul.com

Direct Form I
N M

a y[n k ] = b x[n k ]
k =0
k
k =0
k

General form of difference equation

N M
y[n] a y[n k ] = b x[n k ]
k k
k =1 k =0

Alternative equivalent form

www.rejinpaul.com
 www.rejinpaul.com
Direct Form I M

k
b
k =0
z k

H(z ) = N
1 ak z k
Transfer function
can
be written as
k =1

M
1 k
H z = H2 z H1 z =
( ) ( ) ( ) bk z
N
k k = 0
1 ak z
k =1
M k
V z = H1 z X z = bk z X(z )
( ) ( ) ( ) M

k =0 v[n] = b x[n k ]
k

Direct Form I Represents

k =0
N

Y (z ) = H2 (z )V (z ) =
1 V(z )
y[n] = a y[n k ] + v[n]
k =1
k

N
1 ak z
k

k =1

www.rejinpaul.com
 www.rejinpaul.com

Alternative Representation
Replace order of cascade
LTI systems

M
1
H(z ) = H (z )H (z ) = b z k
1
1 2 k N
k =0


az

k =1
k
k

N
w[n] = a w[n k ] + x[n]
k
k =1
1
W(z ) = H2 (z )X(z ) = X(z ) M

N y[n] = b w[n k ]
k
1 ak z
k
k =0
k =1
M
Y (z ) = H1 (z )W(z ) = bk z k W(z )
k =0

www.rejinpaul.com
 www.rejinpaul.com

Alternative Block Diagram

We can change the order of the cascade
systems
N
w[n] = a w[n k ] + x[n]
k =1
k

M
y[n] = b w[n k ]
k =0
k

www.rejinpaul.com
 www.rejinpaul.com

Direct Form II
No need to store the same data
twice in previous system
So we can collapse the delay
elements into one chain
This is called Direct Form II or
the Canonical Form
Theoretically no difference
between Direct Form I and II
Implementation wise

Less memory in Direct II

Difference when using
finite-precision arithmetic

www.rejinpaul.com
 www.rejinpaul.com

Signal Flow Graph Representation

Similar to block diagram representation

Notational differences
A network of directed branches connected
at nodes

www.rejinpaul.com
 www.rejinpaul.com

Example
Representation of Direct Form II with signal
flow graphs w1 [n] = aw4 [n] + x[n]
w2 [n] = w1 [n]
w3 [n] = b0w2 [n] + b1w4 [n]
w4 [n] = w2 [n 1]
y[n] = w3 [n]

w1 [n] = aw1 [n 1] + x[n]

y[n] = b0w1 [n] + b1w1 [n 1]

www.rejinpaul.com
 www.rejinpaul.com
Determination of System
Function from Flow Graph
w1 [n] = w4 [n] x[n]
w2 [n] = w1 [n]
w3 [n] = w2 [n] + x[n]
w4 [n] = w3 [n 1]
y[n] = w2 [n] + w4 [n]
W1 (z ) = W4 (z ) X(z ) X(z ) z 1 1
( )
W2 (z ) =
W2 (z ) = W1 (z ) 1 z 1 Y (z ) z 1
H(z ) = =
W3 (z ) = W2 (z ) + X(z ) X(z )z 1 (1 ) X(z ) 1 z 1
W4 (z ) =
W4 (z ) = W3 (z )z 1 1 z 1 h[n] = n 1u[n 1] n +1u[n]
Y (z ) = W2 (z ) + W4 (z ) Y (z ) = W2 (z ) + W4 (z )

www.rejinpaul.com
 www.rejinpaul.com

Basic Structures for IIR Systems:

Direct Form I

www.rejinpaul.com
 www.rejinpaul.com

Basic Structures for IIR Systems:

Direct Form II

www.rejinpaul.com
 www.rejinpaul.com

Basic Structures for IIR Systems:

Cascade Form
General form for cascade implementation
M1 M2
1 1 1
(1 f z ) (1 g z )(1 g z )
k =1
k
k =1
k

k
H(z ) = A N1 N2
1 1 1
(1 c z ) (1 d z )(1 d z )
k =1
k
k =1
k

k

M1
b0k + b1k z 1 b2k z 2
H(z ) = 1
k =1 1 a1k z a2k z 2

More practical form in 2nd order systems

www.rejinpaul.com
 www.rejinpaul.com

Example
1 2 1 1
H(z ) =
1 + 2z + z (1 + z =
)(1 + z )
1 2 1 1
1 0.75z + 0.125z (1 0.5z )(1 0.25z )
1 1
=
(1 + z ) (1 + z )
1 1
(1 0.5z ) (1 0.25z )

Cascade of Direct Form I subsections

Cascade of Direct Form II subsections

www.rejinpaul.com
 www.rejinpaul.com

Basic Structures for IIR Systems:

Parallel Form
Represent system function using partial fraction expansion
NP NP NP
Ak Bk 1 ek z 1
( )
H(z ) = Ckzk
k =0
+
k =1 1 ck z
1
+
(
k =1 1 dk z
1
1 dk z 1
)( )
NS
NP
e0k + e1k z 1
H(z ) = C z
k =0
k
k
+
k =1 1 a1k z
1
a2k z 2
Or by pairingthe real poles

www.rejinpaul.com
 www.rejinpaul.com

Example
Partial Fraction Expansion
1 + 2z 1 + z 2 18 25
H(z ) = = 8 +
1 0.75z 1 + 0.125z 2 (1 0.5z 1
) ( 1 0.25z 1 )

Combine poles to get

7 + 8z 1
H(z ) = 8 +
1 0.75z 1 + 0.125z 2

www.rejinpaul.com
 www.rejinpaul.com

Transposed Forms
Linear signal flow graph property:

Transposing doesnt change the input-output
relation 1
H(z ) = 1
Transposing:1 az

Reverse directions of all branches

Interchange input and output nodes
Example:

www.rejinpaul.com
 www.rejinpaul.com

Example
Transpose

y[n] = a1y[n 1] + a2y[n 2] + b0x[n] + b1x[n 1] + b2x[n 2]

Both have the same system function or

difference equation

www.rejinpaul.com
 www.rejinpaul.com

Basic Structures for FIR Systems: Direct Form

Special cases of IIR direct form structures

Transpose of direct form I gives direct form II

Both forms are equal for FIR systems
Tapped delay line
www.rejinpaul.com
 www.rejinpaul.com

Basic Structures for FIR Systems:

Cascade Form
Obtainedby factoring the polynomial
system function
M MS
H(z ) = h[n]z
n=0
n
= (b
k =1
0k + b1k z 1 + b2k z 2 )

www.rejinpaul.com
 www.rejinpaul.com

Structures for Linear-Phase FIR

Systems
Causal FIR system with generalized linear phase are
symmetric:
h[M n] = h[n] n = 0,1,..., M (type I or III)
h[M n] = h[n] n = 0,1,..., M (type II or IV)
Symmetry means we can half the number of multiplications
Example: For even M and type I or type III systems:

M M / 2 1 M
y[n] = h[k ]x[n k ] = h[k ]x[n k ] + h[M / 2]x[n M / 2] + h[k ]x[n k ]
k =0 k =0 k = M / 2 +1
M / 2 1 M / 2 1
= h[k ]x[n k ] + h[M / 2]x[n M / 2] + h[M k ]x[n M + k ]
k =0 k =0
M / 2 1
= h[k ](x[n k ] + x[n M + k ]) + h[M / 2]x[n M / 2]
k =0
www.rejinpaul.com
 www.rejinpaul.com
Structures for Linear-Phase FIR
Systems
Structure for even M

Structure for odd M

www.rejinpaul.com