Inverse z-Transform and Difference Equations

6.003 Fall 2016 Lecture 12

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 1 / 27
Inverse z-Transform and Difference Equations

Motivation: efficient analysis of difference equations

Which functions of z are actually z-transforms?
Partial Fraction Expansion
Inverse z-transform of (z z0 )k
Stability analysis in terms of pole locations and causality

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 2 / 27
z-Transform and LTI Difference Equations
Consider difference equation
Pm Pq
(1) k=0 ak y [n k] = k=p bk v [n k] for all n,
and associated functions of complex argument z:
a(z) = m k , b(z) = qk=p bk z k
P P
k=0 ak z (z 6= 0).

If signals v and y have z-transforms V (z) and Y (z) respectively, and

satisfy (1) then
(2) a(z)Y (z) = b(z)V (z) for all z ROC (y ) ROC (v ), z 6= 0

If signals v and y have z-transforms V (z) and Y (z) respectively, and

satisfy (2) on a circle in ROC (y ) ROC (v ), then they also satisfy the
difference equation (1).

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 3 / 27
Example: Solving Difference Equation Using z-Transform

Equation: y [n] 2y [n 1] = u[n 1], solving for a right-sided y

U(z) =

a(z) =

b(z) =

Y (z) =

Partial fraction expansion: Y (z) =

Inverse z-transform: Y (z) is z-transform of

Hence y [n] =

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 4 / 27
Example: Solving Difference Equation Using z-Transform

Equation: y [n] 2y [n 1] = u[n 1], solving for a right-sided y

z
U(z) = z1 , ROC {|z| > 1}

2
a(z) = 1 z

1
b(z) = z

b(z)U(z) z
Y (z) = a(z) = (z1)(z2) , outer-most ROC: {|z| > 2}

z z
Partial fraction expansion: Y (z) = z2 z1 (|z| > 2)

Inverse z-transform: Y (z) is z-transform of 2n u[n] u[n]

Hence y [n] = 2n u[n] u[n]

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 4 / 27
A Side-Track: Which Functions Of z Are z-Transforms?

Every right-sided signal growing not faster than exponentially has a

z transform with ROC including the exterior of some circle

Which functions of z are z-transforms of signals?

Example 1: function V (z) = sin(sin(z)) is z-transform of some signal

v = v [n], with ROC (v ) = C.
Example 2: for arbitrary 0 < r < R, the function V (z) = Re(z) is not a
z-transform of a signal v = v [n] with ROC (v ) {z C : r < |z| < R}.

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 5 / 27
Functions that are z-transforms
A complex function f = f (z) with domain Ar ,R = {z : r < |z| < R}
equals (on Ar ,R ) z-transform V = V (z) of some signal v = v [n] with
ROC (v ) Ar ,R , if and only if f is complex differentiable in Ar ,R , in the
sense that the limit
f (z) f (z0 )
f 0 (z0 ) = lim
zz0 z z0
exists for all z0 Ar ,R .

Example: function V (z) = sin(z) is complex differentiable, with

sin(z) sin(z0 )
lim = cos(z) for all complex z.
zz0 z z0
All elementary functions are complex differentiable. Composition of
complex differentiable functions via addition, multiplication,
inversion results in complex differentiable functions.
6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 6 / 27
Example: function V (z) = Re(z) is not complex differentiable anywhere,
because
Re(z) Re(z0 ) Re(z z0 )
=
z z0 z z0
converges to different limits depending on the direction from which z
approaches z0 .

In particular:
the limit is 0 when z = z0 + j, where is a real number converging
to zero;
the limit is 1 when z = z0 + , where is a real number converging
to zero.

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 7 / 27
Understanding Difference Equations Via z-Transform

m
X q
X
ak y [n k] = bk v [n k] (a0 6= 0, am 6= 0)
k=0 k=p

v y
- F -

b(z)
Y (z) = V (z) for z ROC (v ) ROC (y ).
a(z)

When can we be sure that ROC (v ) ROC (y ) is not empty?

What does z-transform tell about y = y [n]?

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 8 / 27
Existence of z-Transforms
m
X q
X
ak y [n k] = bk v [n k] (a0 6= 0)
k=0 k=p

z-transform of v
If v is right-sided and grows not faster than an exponent then v has
z-transform with ROC containing a region {z C : |z| > rv } for some
rv > 0.

z-transform of y
If v is right-sided and grows not faster than an exponent then the (unique)
right-sided solution of the difference equation also grows not faster than
an exponent, i.e., has z-transform with ROC containing a region
{z C : |z| > ry } for some ry > 0.

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 9 / 27
Why?
by assumption, v [n] = 0 for n < n0 , and |v [n]| cv rvn for some
rv , cv > 0
let M = |a0 |1 (|a1 | + + |am |), D = |a0 |1 (|bp | + + |bq |), and
select arbitrary r > max{M, rv }
since y is right-sided, p[n] = r n max{|y [k]| : k n} is finite for all n
since
y [n] = -a01 (a1 y [n-1]+ . . . +am y [n-m]+bp v [n-p]+ . . . +bq v [n-q]), we
have

p[n + 1] r n1 (Mr n p[n] + Dcv rvn ) Mr 1 (p[n] + Dcv )

M
hence for every R the inequality p[n] RDcv implies
r M
 
M M M
p[n + 1] (RDcv + Dcv ) + 1 Dcv RDcv ,
r r r M

which proves that p[n] will be uniformly bounded.

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 10 / 27
Rational functions
A function f : C {} C {} is called rational when it can be
represented in the form
r d
g (w ) X
i
X
f (z) = lim : g (z) = gi z , q(z) = qi z i ,
w z q(w )
i=0 i=0

where gi , qi are constant complex coefficients, qd 6= 0.

The limit is needed to treat the case when q(z) = 0 (and, possibly,
g (z) = 0), otherwise simply f (z) = gq(z)
(z)
.
Different pairs of polynomials g = g (z), q = q(z) may define same
rational functions, as in
z 1 1
f (z) = is the same rational function as f (z) = .
z2 z z

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 11 / 27
Examples: Rational Functions

z
f (z) = z+1 is

f (z) = Re z1 is

z2
f (z) = z+1 is

1
f (z) = z is

f (z) = z 2 + z + 1 is

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 12 / 27
Examples: Rational Functions

z
f (z) = z+1 is not rational, because of z

f (z) = Re z1 is not rational, because of Re

z2 z 3 z 2
f (z) = z+1 is rational, same as z 2 1
etc.

1 z2
f (z) = z is rational, same as z 2 2z
etc.

z 3 1
f (z) = z 2 + z + 1 is rational, same as z1 etc.

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 12 / 27
Rational Functions Terminology

g0 + g1 z + + gr z r
f (z) = , qd 6= 0.
q0 + q1 z + + qd z d

f is called proper when f () is finite (i.e., r d);

f is called strictly proper when f () = 0 is finite (i.e., r < d);
the points z where f (z) = 0 are called zeros of f (z);
the points z where f (z) = are called poles of f (z);
when gr 6= 0, the number d r is called relative degree.
rational function with a non-negative (positive) relative degree is
called proper (respectively, strictly proper)
lets agree that f (z) = 10 0 has no zeros, no poles, and an
undefined relative degree

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 13 / 27
Multiplicity of Zeros and Poles
g0 + g1 z + + gr z r
f (z) = , qd 6= 0, gr 6= 0.
q0 + q1 z + + qd z d

Every rational function f = f (z) can be represented in the form

(z z1 )r1 (z z2 )r2 . . . (z z )r
f (z) = f0 ,
(z p1 )d1 (z p2 )d2 . . . (z p )d

where f0 is a complex constant, (z1 , . . . , z ) is the list of all finite zeros of

f , (p1 , . . . , p ) is the list of all finite poles of f , and ri , di are positive
integers (called the multiplicities of the corresponding zeros/poles)

When d > r , the multiplicity of the zero at z = is d r .

When d < r , the multiplicity of the pole at z = is r d.
Hence the sum of multiplicities of all poles equals the sum of
multiplicities of all zeros.
6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 14 / 27
 z1
Example: f (z) = z 2 +1.5z

poles p = 0 and p = 1.5 (both multiplicity 1)

zeros z = 1 and z = (both multiplicity 1)
strictly proper (relative degree 1)

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 15 / 27
 z 2 +1
Example: f (z) = z 4 z 2

poles p = 0 (multiplicity 2), p = 1, and p = 1 (both multiplicity 1)

zeros z = (multiplicity 2), z = j, and z = j (both multiplicity 1)
strictly proper (relative degree 2)

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 16 / 27
 1
Example: f (z) = (z 2 z+0.5)(z+1)

poles p = 0.5 + 0.5j, p = 0.5 0.5j, and p = 1 (all multiplicity 1)

zeros z = (multiplicity 3)
strictly proper (relative degree 3)

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 17 / 27
Example: f (z) = (z 2 z + 0.5)(z + 1)

poles z = (multiplicity 3)
zeros p = 0.5 + 0.5j, p = 0.5 0.5j, and p = 1 (all multiplicity 1)
not proper (relative degree -3)

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 18 / 27
Partial Fraction Expansion

Every rational function

g0 + g1 z + + gr z r
H(z) = (gr 6= 0)
(z p1 )d1 (z p2 )d2 . . . (z p )d

can be represented in the partial fraction expansion form

r d i 1
dX
X X ci,l
H(z) = c0,l z l + ,
(z pi )l+1
l=0 i=1 l=0

where d = d1 + d2 + + d , and ci,l are complex coefficients.

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 19 / 27
Example: Partial Fraction Expansion
z3 c1,0 c2,0 c3,0
H(z) = = c0,0 + + +
(z 1)(z 2)(z 3) z 1 z 2 z 3

H(z) = 1 @ z = c0,0 = 1
(z 1)H(z) = 0.5 @ z = 1 c1,0 = 0.5
(z 2)H(z) = 8 @ z = 2 c2,0 = 8
(z 3)H(z) = 13.5 @ z = 3 c3,0 = 13.5

0.5 8 13.5
H(z) = 1 + +
z 1 z 2 z 3
h[n] = [n] + 0.5u[n 1] 8 2n u[n 1] + 13.5 3n u[n 1].

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 20 / 27
Inverse z-Transforms of Partial Fraction Expansion Terms

Vl (z) = z k is z-transform of
v [n] = [n + k] (ROC=C for k 0, ROC=C{0} for k < 0).

+
Va,0 (z) = (z a)1 (|z| > |a| > 0) is z-transform of
v [n] = an1 u[n 1]

Va,0 (z) = (z a)1 (|z| < |a|) is z-transform of
v [n] = an1 u[n]

+
Va,k (z) = (z a)k1 (|z| > |a| > 0) is z-transform of
(n1)(n2)...(nk) nk1
v [n] = k! a u[n 1]
k1
Va,k (z) = (z a) (|z| < |a|) is z-transform of
v [n] = (n1)(n2)...(nk)
k! ank1 u[n]

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 21 / 27
Time Domain Implications of Partial Fraction Expansion

Every rational function

g0 + g1 z + + gr z r
H(z) = (gr 6= 0),
(z p1 )d1 (z p2 )d2 . . . (z p )d

restricted to {z C : |z| > R0 } with R0 = max{|p1 |, . . . , |p |}, is

z-transform of the signal
r d
X i 1
dX
X
h[n] = c0,l [n + l] + ci,l nl pn u[n 1],
l=0 i=1 l=0

where d = d1 + d2 + + d , and ci,l are complex coefficients.

Note: the coefficients ci,l are not necessarily the same as the coefficients
ci,l of the corresponding partial fraction expansion.

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 22 / 27
Example: Another Partial Fraction Expansion

1 c1,0 c1,1
H(z) = 2
= c0,0 + +
z(z 1) z 1 (z 1)2

zH(z) = 1 @ z = 0 c0,0 = 1
2
(z 1) H(z) = 1 @ z = 1 c1,1 = 1
d
[(z 1)2 H(z)] = 1 @ z = 1 c1,0 = 1
dz

1 1 1
H(z) = +
z z 1 (z 1)2
h[n] = [n-1] u[n-1] + (n-1)u[n-1] = [n-1] + (n-2)u[n-1].

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 23 / 27
Poles to Time Domain

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 24 / 27
Poles to Time Domain

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 25 / 27
Poles to Time Domain

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 26 / 27
Relative Degree and Latency

Every rational function H(z) coincides with a z-transform of a unique

right-sided signal h = h[n] on its ROC. Moreover, h[n] = 0 for all n < 0
(for all n 0) if and only if H(z) is proper (respectively, strictly proper).

Recall that the convolution system v 7 y = h v is

causal if and only if h[n] = 0 for all n < 0;
strictly causal if and only if h[n] = 0 for all n 0.

Examples:
z2
H(z) = z 3 z1
corresponds to a strictly causal system, latency -1
z 2
H(z) = z1 does correspond to a finite latency system, latency 1.

6.003 Fall 2016 Lecture 12 Inverse z-Transform and Difference Equations October 21, 2016 27 / 27