Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

Chapter Six

Realization of Discrete-Time Systems


Lecture #11

Rediet Million

AAiT, School Of Electrical and Computer Engineering


rediet.million@aait.edu.et

May,2018

(Rediet Million) DSP-Lecture #11 May,2018 1 / 31


6.1 Introduction

Given a linear-time invariant system, the input -output transformation


can be computed in different ways,each ways is called system
implementation.
A system implementation is a specific description of its internal
computational structure.

The representation often used in the implementation of LTI systems is


the difference equation or the system function.

This difference equation or system function has to be converted to an


algorithm or structure that can be realized in software or hardware.

The choice of an implementation concerns with


I Computational requirements
I Memory requirements,effective of finite precision etc.

(Rediet Million) DSP-Lecture #11 May,2018 2 / 31


Introduction

For example , consider the system described by the system function


b0 + b1 z −1
H(z) = ,|z| > |a|
1 − az −1
-The impulse response and the difference equation can be described as
h(n) = b0 an u(n) + b1 an−1 u(n − 1) and
y (n) = ay (n − 1) + b0 x(n) + b1 x(n − 1)

I Since the impulse response ,h(n) has infinite -duration ,it is impossible
to implement the system by direct convolution ,y (n) = x(n) ∗ h(n).
I However, the LCCDE provide a means for recursive computation of the
output at any time n in terms of the previous output, the current input
and the previous input samples.

(Rediet Million) DSP-Lecture #11 May,2018 3 / 31


6.2 Block Diagram and Signal Flow Graph representation
Block diagram representation

For a LTI system with a rational system function,the input x(n) and
the output y (n) are related by a LCCDE:
M
P M
P
y (n) = bk x(n − k) − ak y (n − k)
k=1 k=0
I The basic computational elements required to find the output at time n
are adders,multipliers, and delays.
I It is often convenient to use a block diagram to illustrate how these
adders,multiplier,and delays are interconnected to implement a given
system.
-The notation that is used for these system is shown bellow.

(Rediet Million) DSP-Lecture #11 May,2018 4 / 31


Block Diagram and Signal Flow Graph representation
Block diagram representation

Example:
Consider a system function ,H(z),described by a second order difference
equation
b0
H(z) = −1
1 − a1 z − a2 z −2
y (n) = a1 y (n − 1) + a2 y (n − 2) + b0 x(n)
-Using a block diagram,this system is represented as follows

(Rediet Million) DSP-Lecture #11 May,2018 5 / 31


Block Diagram and Signal Flow Graph representation
Signal flow graph representation

A computational structure is also represented pictorially using a signal


flow graph,which is a network of directed branches that are connected
at nodes.The node in a SFG correspond to either adders or branch
points.
I Adders correspond to nodes with more than one incoming branch,and
branch points are nodes with more than one outgoing branch,as illustrated
in figure bellow.

- Node j represents an adder ,and node k is a branch point.


(Rediet Million) DSP-Lecture #11 May,2018 6 / 31
Block Diagram and Signal Flow Graph representation
Signal flow graph representation

Example:
Consider a system function ,H(z),described by a first order difference
equation
b0 + b1 z −1
H(z) =
1 − a1 z −1
y (n) = b0 x(n) + b1 x(n − 1) + a1 y (n − 1)
-Using a SFG,this system us represented as follows

(Rediet Million) DSP-Lecture #11 May,2018 7 / 31


6.3 Basic structures for IIR systems

For any given rational system function, a wide variety of equivalent


sets of difference equations or network structures exists.One
consideration in the choice is computational complexity.

I For example, in some digital implementations, structure with a fewest


constant multiplier and the fewest delay branches are often desirable.
I A reduction in the number of constant multipliers means an increase in
speed , and a reduction in the number of delay elements means a
reduction in memory requirements.
I These effects depend on the way in which the structures are organized
i.e on the structure of the block diagram or signal flow graph.
I In this section,we develop several of the most commonly used forms for
implementing a Linear time-invariant IIR system and obtain their block
diagram and flow graph representation.
(Rediet Million) DSP-Lecture #11 May,2018 8 / 31
Basic structures for IIR system
Direct form I

This is a realization of the system is obtained directly from the N th


order difference equation given as :

(Rediet Million) DSP-Lecture #11 May,2018 9 / 31


Basic structures for IIR system
Direct form I

I The direct form realization in figure above has a total of M + N delay


units and adders ,M + N + 1 scalar multipliers .

(Rediet Million) DSP-Lecture #11 May,2018 10 / 31


Basic structures for IIR system
Direct form II(Canonical)

The direct-I form can be rearranged in a variety of ways without


affecting the system transfer function.
I Each rearrangement of the block diagram represents a different
computational algorithm for implementing the same system.
I Canonical form structure is obtained by reversing the order of the
cascade N(z) and 1/D(z).With this implementation ,x(n) is first filtered
with the all-pole filter 1/D(z) and then with N(z).

(Rediet Million) DSP-Lecture #11 May,2018 11 / 31


Basic structures for IIR system
Direct form II(Canonical)

(Rediet Million) DSP-Lecture #11 May,2018 12 / 31


Basic structures for IIR system
Direct form II(Canonical)

1
I If we denote the output of the all-pole filter by w (n), this
D(z)
structure is described by the following pair of coupled difference equations:
N
P
w (n) = x(n) − ak w (n − k)
k=1
M
P
y (n) = ak w (n − k)
k=1
I This structure may be simplified by nothing that the two sets of delays
are delaying the same sequence.
I The computational requirements for a direct form II structures are as
follows:
-Numbers of multiplications: M + N + 1 per output sample
- Numbers of adders :N + M per output sample
- Numbers of delays:Max(N, M)

(Rediet Million) DSP-Lecture #11 May,2018 13 / 31


Basic structures for IIR system
Direct form II(Canonical)

(Rediet Million) DSP-Lecture #11 May,2018 14 / 31


Basic structures for IIR system

Example-1
Consider the LTI system with a system function
1 + 2z −1
H(z) =
1 − 1.5z −1 + 0.9z −2
We find b0 = 1,b1 = 2,a0 = 1,a1 = −1.5 , and a2 = 0.9

(Rediet Million) DSP-Lecture #11 May,2018 15 / 31


Basic structures for IIR system

Example-2
Consider the LTI system with a system function
1 + 2z −1 + z −2
H(z) =
1 − 0.75z −1 + 0.125z −2
We find b0 = 1,b1 = 2,b2 = 1,a0 = 1,a1 = −0.75 , and a2 = 0.125

(Rediet Million) DSP-Lecture #11 May,2018 16 / 31


Basic structures for IIR system
Cascade form

The cascade form realization of a general transfer function H(z) is a


representation of H(z) as cascade of subsystems of smaller dimension.
I The subsystems will normally be arranged as second-order sections.
Thus, the system transfer function is expressed as a product of
second-order sections,i.e with N ≤ 2K

I or, more generally, we may express H(z) in terms of a cascade of first


and second order factors,for N and M even or odd, as

(Rediet Million) DSP-Lecture #11 May,2018 17 / 31


Basic structures for IIR system
Cascade form

IWhere the degree of the numerator polynomial M = M1 + M2 and the


degree of the denominator polynomial N = N1 + N2.
IAlthough there exist different structures for the first and second-order
sections, it is often desirable to implement the cascade realization that
require minimum storage and computation.

(Rediet Million) DSP-Lecture #11 May,2018 18 / 31


Basic structures for IIR system
Cascade form

Example:
Let us consider again the LTI system with a system function
1 + 2z −1 + z −2
H(z) =
1 − 0.75z −1 + 0.125z −2
To illustrate the cascade structure,we can use first-order systems by
expressing H(z) as a product of first-order factors.
1 + 2z −1 + z −2 (1 + z −1 )(1 + z −1 )
H(z) = −1 −2
=
1 − 0.75z + 0.125z (1 − 0.5z −1 )(1 − 0.25z −1 )

(Rediet Million) DSP-Lecture #11 May,2018 19 / 31


Basic structures for IIR system
Parallel form

The parallel form is a realization of the partial fraction expansion of


the rational system function H(z) in the form

where N = N1 + 2N2 . This expression considers all possible options for M


and N. If M ≥ N, then Np = M − N; otherwise, the first summation is not
included.
-If the first term exists, then it represents simple Np scaled delay paths or
branches of the parallel network.
-The second term represents N1 first-order branches.
-The third term represents N2 second-order branches.

(Rediet Million) DSP-Lecture #11 May,2018 20 / 31


Basic structures for IIR system
Parallel form
Consider the parallel form realization, for M = N = 6, of

(Rediet Million) DSP-Lecture #11 May,2018 21 / 31


Basic structures for IIR system
Parallel form

Example:
Consider again the LTI system with a system function
1 + 2z −1 + z −2
H(z) =
1 − 0.75z −1 + 0.125z −2
If we use second-order,
1 + 2z −1 + z −2 −7 + 8z −1
H(z) = = 8 +
1 − 0.75z −1 + 0.125z −2 1 − 0.75z −1 + 0.125z −2
Since all the poles are real, we can obtain an alternative parallel for
realization by expanding H(z) as
18 25
H(z) = 8 + −1

1 − 0.5z 1 − 0.25z −1

(Rediet Million) DSP-Lecture #11 May,2018 22 / 31


Basic structures for IIR system
Parallel form

(Rediet Million) DSP-Lecture #11 May,2018 23 / 31


Basic structures for IIR system
Transposed form

The transposition theorem states that the input-output properties of


a network remain unchanged after the following sequence of network
operations:
1.Reverse the direction of all branches.
2.Change branch points into summing nodes and summing nodes into
branch points.
3.Interchange the input and output.
Example: Transposed form for a first-order system with no-zeros
1
H(z) =
1 − az −1

(Rediet Million) DSP-Lecture #11 May,2018 24 / 31


Basic structures for IIR system
Transposed form

IThe transposed direct form II implements the zeros first and then the
poles, being important effect for finite-precision existing.

(Rediet Million) DSP-Lecture #11 May,2018 25 / 31


6.4 Basic structures for FIR systems

Although the direct and cascade forms for IIR systems include FIR
systems as a special case, there are additional specific forms for FIR
systems.
1.Direct Form :
For causal FIR systems,the system function has only zeros(except for
poles at z = 0), and since the coefficients ak are all zero, the general
Nt h order difference equation reduces to
M
P
y (n) = bk x(n − k)
k=0

I This can be recognized as the discrete convolution of x(n) with the


impulse response
(
bn , n = 0, 1, ..., M
h(n) =
0 , otherwise,
(Rediet Million) DSP-Lecture #11 May,2018 26 / 31
6.4 Basic structures for FIR systems

In this case ,the direct form I and direct form II structure both reduce to
the direct-form FIR structure shown bellow:

I Because of the chain of delay elements across the top of the diagram
,this structure is also referred to as a tapped delay line structure or a
transversal filter structure.

(Rediet Million) DSP-Lecture #11 May,2018 27 / 31


6.4 Basic structures for FIR systems

2. Cascaded Form
The cascade form for FIR systems is obtained by factoring the polynomial
system function.That is , we represent H(z) as
M
X Ms
Y
H(z) = h(n) = (b0k + b1k z −1 + b2k z −2 )
n=0 k=1
Where M is the largest integer contained in Ms = (M + 1)/2.

(Rediet Million) DSP-Lecture #11 May,2018 28 / 31


Chapter Six reading assignment
Structures for Linear-phase FIR system
Lattice structure
Quantization of coefficients

(Rediet Million) DSP-Lecture #11 May,2018 29 / 31


Class exercises & Assignment
1) Consider the causal linear shift-invariant filter with system function

1 + 0.2z −1
H(z) =
(1 − 0.5z −1 + 0.3z −2 )(1 + 0.25z −1 )
Draw a signal flow graph for this system using
i. Direct form I ii. Direct form II
iii. Cascade form using first- and second-order direct form II sections
iv. Parallel form using first and second order direct form II sections
v. Transposed direct form II
2) Consider the filter structure shown in the figure below

Find the system function and the unit sample response of this system.
(Rediet Million) DSP-Lecture #11 May,2018 30 / 31
3) Find the system function and the unit sample response for the following
network, and draw an equivalent parallel form structure

(Rediet Million) DSP-Lecture #11 May,2018 31 / 31

You might also like