EEO 401

Digital Signal Processing

Prof. Mark Fowler

Note Set #9
• Using ZT to Analyze DT LTI Systems
• Reading Assignment: Sect. 3.5 of Proakis & Manolakis

1/12
 Transient and Steady-State Responses

We can use the ZT to get an idea of what to expect the output of an LTI
system will look like.

N M Contains x[n], x[n-1],…

y[n ] + ∑ ai y[n=
− i] ∑ b x[n − i ]
i This system is causal, we won’t
=i 1 =i 0
have x[n+1], x[n+2], etc. here

Az ( z ) = z N + a1 z N −1 + ... + a N −1 z + a N
N −1 N −M B z ( z)
B ( z=
z
) b0 z + b1 z
N
+ ... + bM z H ( z) = z
z

A ( z)

Assuming zero ICs and using the convolution property:

B z ( z) z General result: Get output’s ZT

=
Y ( z ) H=
z z
( z) X ( z) z
z
X ( z) by multiplying TF by input’s ZT.
A ( z)
But… we can study this further
H(z) = transfer function to get key insight!!
2/12
 For simplicity assume Xz(z) = Ez (z)/Fz (z)

For simplicity… assume distinct poles

z z z
Some
B ( z) E ( z) c1 z c2 z cN z D ( z)
Y (= = + ++ + z
z Polynomial
z)
A ( z ) F ( z ) z − p1 z − p2
z z
z − pN F ( z ) that “falls
out of” PFE

y=
zs [ n ] c p
1 1
n
u[ n ] + c p
2 2
n
u[ n ] +  + c N N u[n ] + y ss [n ]
p n

Transient Response…Decays if |pi| < 1 Steady State

Response

So… if all the poles are inside the UC then the Transient response decays and
is not that interesting for most signal processing applications.

3/12
Causality & ROC

Recall that a causal system with 0 ICs can not give a non-zero output until the
input becomes non-zero.

Since h[n] is the output due to δ[n] with zero ICs we can see that a causal
system must have
h[n ] = 0 ∀n < 0

We also know that for a causal sequence the ROC of its ZT is the exterior of a
circle. Thus the ROC of Hz(z) for a causal system is the exterior of a circle.

An LTI System is causal if an only if the ROC of its transfer

function Hz(z) is the exterior of a circle of radius R < ∞,
including the point z = ∞.

4/12
Stability & ROC
We’ve already discussed that a necessary and sufficient condition for an LTI
system to be stable is for the impulse response to be absolutely summable:
∞

∑ h[n] < ∞
n = −∞
From this we can derive insight into the impact of the ROC on stability. Start
from definition of Hz(z):
∞ ∞ ∞
H ( z) =
z
∑ h[n]z
n = −∞
−n
H ( z) ≤
z
∑
n = −∞
h[n ]z −n
∑
= h[ n
n = −∞
] z −n

Now evaluating this inequality on the unit circle shows

∞
z
H ( z)
z =1
≤ ∑ h[n]
n = −∞
H z ( z)
z =1
≤∞ for a stable system

Thus… If the system is BIBO stable then the UC is in the ROC.

Can prove the reverse is true as well.

An LTI System is BIBO Stable if an only if the UC is in the ROC.

5/12
For a causal system… we now the ROC is the outside of a circle.

Thus, that circle must be inside the Unit Circle…

And… since the poles of a stable system must be outside the ROC…

An LTI Causal System is BIBO Stable if an only

if all the poles are strictly inside the UC.

Causal System ROC

Im{z}
ROC
UC
Rc
Re{z}
Poles will be inside this circle…
with at least 1 pole on the circle. For
stability the poles must be strictly
inside the UC… not on the UC.
6/12
Stability & Poles on the Unit Circle
On the previous slide we said explicitly that the poles must be strictly inside
the UC for stability. To show that we can’t have a pole on the unit circle and
still have BIBO all we need is one counter example.
z
Counter Example: Consider a system with a single pole H ( z) =
z z−a
Consider an input with ZT X ( z) =
z−b
For simplicity let a,b be real and positive But let 0 < b ≤ 1 so the input is bounded
 z  z 
Then the output ZT=
is Y ( z ) H=
( z) X ( z) 
 z − a   z − b 
For a ≠ b we find that the IZT of this will have the form
Remains bounded even if
=
y[n ] Aa u[n ] + Bb u[n ]
n n
pole is on the UC

Bounded only if pole is

For a = b we find that the IZT of this will have the form strictly inside the UC

=
y[n ] Aa n u[n ] + Bna n u[n ] Unbounded if pole is
on the UC

Exercise: Consider the case where there are two poles @ z = 1

7/12
Summary of Stability Results
A causal system is
• …BIBO stable if all poles lie strictly inside the UC
• …BIBO marginally stable if there are some single poles on the UC but
no poles outside the UC
• …BIBO unstable if there is at least one pole outside the UC and/or at
least one multiple pole on the UC

8/12
Pole – Zero Cancellation
Sometimes when finding Yz(z) the combination of z-transforms leads to a pole
and a zero at the same location in the plane.
This can happen either from:
• the interaction of Hz(z) and Xz(z)
• when cascading multiple systems together
− which we saw leads to a composite transfer function that is the
product of the cascaded transfer function
z ( 2 z + 1) ( z − 2)
Example: Cascade of two systems H1 ( z ) =
z
H z
( z ) =
( z − 0.3)( z − 2) ( z + 0.5)
2

First system is NOT stable!!!

The total transfer function is
 z ( 2 z + 1)   ( z − 2)  z ( 2 z + 1)
=
H ( z ) H=
z z
( z)H ( z)  z
   =
1
 ( z − 0.3)( z − 2)   ( z + 0.5) 
2
( z − 0.3)( z + 0.5)

• In theory this kind of “stabilization” by pole cancellation works…

• But… in practice, numerical precision issues in the implementation may
result in imperfect cancellation… so beware of this!
9/12
Example: Similarly we can have pole-zero cancellation between the system
and the input
y[n ] − 6 y[n − 1] + 6 y[n − 2] =
5 1
x[n ]
x[n ] =δ [n ] − 13 δ [n − 1]
1 1
=
The transfer function is: H z ( z ) =
1 − 65 z −1 + 16 z −2 (1 − 1
2 z −1 )(1 − 13 z −1 )

−1
The ZT of the input is: X ( z )= 1 − 13 z
z

1
So… after cancellation the output ZT is Y z ( z) =
1 − 12 z −1

And.. then the time-domain form for the output signal is

y[n ] = ( 12 ) u[n ]
n

Note… from our earlier studies we would expect the output to include an
exponential term for each pole in its transient response. But here we only see
one of the pole’s exponential…
This input only “excites” one of the two poles!
10/12
Stability of 2nd-Order Systems
As we’ll see later… 2-pole sub-systems form the building blocks for realizing
higher-order systems. Thus, having explicit insight into the stability of 2nd order
systems is very useful.

2nd Order Diff. Eq.: y[ n ] =−a1 y[n − 1] − a2 y[n − 2] + b0 x[n ]

2
b b z
=
2nd Order Transfer Function: H ( z) = 0 0
1 + a1 z −1 + a2 z −2 z 2 + a1 z + a2

2nd Order Poles: −a1 a12 − 4a2

p1 ,=
p2 ±
2 4

− ( p1 + p2 )
a1 = =
a2 p=
1 p2 p1 p2 < 1 These provide the
a2 = p1 p2 a1 < 1 + a2 “Stability Triangle”

a12
a12 − 4a2 =
0 a2 = … provides the dividing curve between real roots
4 and complex roots
11/12
12/12