CEDC606: Digital Signal Processing

Lecture Notes 4: Transform analysis of LTI systems

Ramez Koudsieh, Ph.D.

Faculty of Engineering
Department of Robotics and Intelligent Systems
Manara University
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 1/36

 Chapter 4
Transform analysis of LTI systems

1. Frequency-Domain Characteristics of LTI Systems

2. Frequency Response of LTI Systems
3. Linear Time-Invariant Systems as Frequency-Selective Filters
4. Invertibility and minimum-phase systems

https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 2/36

 1. Frequency-Domain Characteristics of LTI Systems
Response to Complex Exponential and Sinusoidal Signals
▪ The response of any relaxed LTI system to an arbitrary input signal x[n] is
given by the convolution sum formula: y[n ] = k =− h[k ]x[n − k ]


▪ If the input is a complex exponential sequence x[n] = Aejn, −∞  n  ∞

k =− 
 
y[n ] = h[k ]Ae j  (n − k )
= A  h[k ]e − j k )  j n
e
 k =− 
H (e ) = k =− h[k ]e − j k is the Fourier transform of h[k] of the system
j 

n =− h[n ]  

H(ej) exists if the system is BIBO stable, that is, if
y[n ] = AH (e j  )e j n x[n] = Aejn is an eigenfunction of the system
H(ej) is the corresponding eigenvalue of the system
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 3/36

 ▪ Since H(ej) is the Fourier transform of {h[k]}, it follows that H(ej) is a periodic
function with period 2.

 k =− h[k ]e

−
j − j k
H (e ) = , h [k ] = 1
2
H (e j  ) e j k d 

H (e j  ) = |H (e j  )|e j  ( ) = H R (e j  ) + jH I (e j  ), ( ) = H (e j  )
x [n ] = Ae j (n + ) ⎯⎯⎯
T
→ y[n ] = A|H (e j  )|e j [n + + ( )]
If h[k] is a real-valued, then X R (e − j  ) = X R (e j  ), X I (e − j  ) = −X I (e j  )
H (e − j  ) = H (e j  ) , H (e − j  ) = −H (e j  )

▪ Suppose that the input is a real sinusoidal sequence

Ax j x Ax
x [n ] = Ax cos(n + x ) = 2
e e j n
+ 2
e − jx e − j n = x 1 [n ] + x 2 [n ]
Ax Ax
y1 [n ] = 2
|H (e j  )|e jx e j [n + ( )] , y2 [n ] = 2
|H (e − j  )|e − jx e j [ −n + ( )]
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 4/36

 Ax Ax
y [n ] = 2
|H (e j  )|e jx e j [n + ( )] + 2
|H (e − j  )|e − jx e j [ −n + ( )]
If the impulse response h[k] is real-valued,
y[n ] = Ax |H (e j  )|cos[n + x + ( )] = Ay cos(n + y )
where Ay = Ax |H (e j  )|, y = x + ( )
Similarly, if the input is x [n ] = Ax sin(n + x ), the output is y[n ] = Ay sin(n + y )
▪ Note: The quantity |H(ej)| is known as the magnitude response or gain of the
system, and y is called the phase response of the system.
▪ Example 1: A stable system described by the first-order difference equation
y[n ] = ay[n − 1] + bx [n ], 0  a  1
j b
The frequency response function is: H (e ) =
1 − ae − j 
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 5/36

 j |b | j −1a sin( )
|H (e )| = , H (e ) = b − tan
1 + a 2 − 2a cos( ) 1 − a cos( )
choosing b so that max H(ej) = 1. since a  0, the denominator of H(ej) attains
its min at  = 0. |H(0)| = |b|/(1 − a) = 1 ⇒ b = ±(1 − a). We choose b = 1 − a.
Figure below shows plots of magnitude and phase response functions for
a = 0.8 and an input–output pair for the frequency  = 2/20. (x[n] = cos(0.1n))

https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 6/36

 The response of the system to x[n] = 5 + 12sin(n/2) −20cos(n + /4)
y [n ] = 5|H (e j 0 )| + 12|H (e j  / 2 )|sin[ 2 n + ( 2 )] − 20|H (e j  )|cos[ n + 4 + ( )]
1−a
j0
|H (e )| = 1, H (e j / 2
)=  0.074, ( 2 ) = − tan −1a = −0.733,
1 + a2
j 1−a
H (e ) =  0.053, ( ) = 0
1+a
y [n ] = 5 + 0.888sin( 2 n − 0.733) − 1.06cos( n + 4 ), −   n  
Steady-State and Transient Response to Sinusoidal Input Signals
▪ The eigenfunction property holds if the input signal x[n] is a complex
exponential and sinusoidal and applied to the system at n = −∞. In such a case,
the response of the LTI system is the steady-state response. There is no
transient response in this case.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 7/36

 ▪ However, in practice every input starts at a finite time. Consider a complex
exponential starting at time n = 0, that is, x[n] = ejnu[n], the response of a
causal system (h[n] = 0, n  0) to the causal input x[n] is:
n n
  − j k  j n   − j k  j n
y[n ] =  h[k ]x[n − k ] =  h[k ]e j  (n − k )
=   h[k ]e
 k =0
e

−   h[k ]e
 k =n +1
e

k =0 k =0

j
y[n ] = H (e )e j n
− ( k = n + 1

h[k ]e − j k
)e j n

yss [n ]
ytr [n ]

k =n +1 h[k ]  k = 0 h[k ]   .
 
If the system is stable, we have: ytr [n ] 
For large values of n the transient response of a stable system decays
towards zero, that is, lim y[n ] = H (e j  )e j n = yss [n ]
n →
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 8/36

 ▪ Example 2: A system described by the first-order DE y[n] = ay[n − 1] + x[n]
This system’s response to any input x[n] applied at n = 0 is given as:
+ k = 0 a k x[n − k ], for n  0
n +1 n
y[n ] = y[ − 1]a
Now, let the input to the system is the complex exponential x[n] = Aejn, n ≥ 0
y[n ] = y[ − 1]a n +1 + Ak = 0 a ke j  (n −k ) = y[ − 1]a n +1 + A  k = 0 (ae − j  )k  e j n , n  0
n n
 
n + 1 − j  (n + 1)
Aa e j n A j n
y[n ] = y[ − 1]a n +1 − − j
e + − j
e , for n  0
1 − ae 1 − ae
n +1 Aa n +1e − j  (n +1) j n
The system is BIBO stable if |a|  1. ytr [n ] = y[ − 1]a − − j
e → 0
1 − ae n →

A j n j j n
yss [n ] = lim y[n ] = e = AH (e )e
n → 1 − ae − j 
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 9/36

 ▪ Example 3: A causal and stable system described by y[n] = 0.8y[n − 1] + x[n]
The response y[n] of the system to
the input: x[n] = cos(0.05n)u[n]:
y[n ] = −2.515(0.8)n u[n ] +
y [n ]
tr

4.093cos(0.05 n − 0.538)u[n ]
y [n ]
ss

|H (e j 0.05 )| = 4.093, H (e j 0.05 ) = −0.538

Steady-State Response to Periodic Input Signals

▪ Let the input to a stable LTI system is a periodic signal x[n] with fundamental
period N.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 10/36

 ▪ Since such a signal exists from −∞  n  ∞, the total response of the system
at any time instant n is equal to the steady-state response.
2

N −1 (x ) j N kn
▪ Using of the FS representation of the periodic signal x[n ] = c e
k =0 k
2 2 2
N −1 (x ) j N kn j kn N −1 (y ) j N kn
  
T N −1 (x ) 2
x [n ] = c e
k =0 k
⎯⎯⎯→ y [n ] = c
k =0 k
H (j N
kn )e N = c e
k =0 k
2
ck(y ) = ck(x )H (j N
kn )
(y ) 2 (x ) 2
 n = 0 y[n ]   k = 0 H (j
1 N −1 2 N −1 N −1 2 2
Py = N
= c
k =0 k
= N
kn ) ck

▪ Example 4: Zero-state and steady-state responses

y[n ] = 0.9y[n − 1] + 0.1x [n ], y[ − 1] = 0
1, 0  0  6
The system is excited by a periodic sequence, with N = 10, x [n ] = 
0, 6  n  10
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 11/36

 Response to Aperiodic Input Signals
▪ The convolution theorem provides the desired frequency-domain relationship
for determining the output of an LTI system to an aperiodic finite-energy
signal. Y(ej) = H(ej)X(ej)
▪ The LTI system, through its frequency response function, attenuates some
frequency components of the input signal and amplifies others.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 12/36

 ▪ If the input signal spectrum is changed by the system in an undesirable way,
we say that the system has caused magnitude and phase distortion.
▪ The output of a linear time-invariant system cannot contain frequency
components that are not contained in the input signal.

−
j
−1
y [n ] = F {Y (e )} = 1
2
Y (e j  ) e j n d  x[n] LTI system y[n] = x[n] ∗ h[n]
j
 2 2 X(ej) h[n], H(e ) Y(ej) = X(ej)H(ej)
 n =− y[n ]

− X (e
2 j j
Ey = = 1
2
) H (e ) d 

2. Frequency Response of LTI Systems

Frequency Response of a System with a Rational Transfer Function

M −k
b z B (z )
y[n ] = −k =1 aky[n − k ] + k = 0 bk x[n − k ]  H (z ) =
N M k =0 k
=
1 + k = 1 a k z A(z )
N −k

https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 13/36

 For a stable system, the transfer function converges on the unit circle.
k = 0 k
M − j k
B (z ) b e
H (e j  ) = =
A(z ) z =e j 1 + N a e − j k
k = 1 k
k = 1 
M −1 M − j
(1 − z z ) (1 − z e )
H (e j  ) = b0
k k
= b0 k =1

k = 1 k = 1
N −1 N − j
(1 − pk z ) (1 − p k e )
z =e j 
where the {ak} and {bk} are real, but {zk} and {pk} may be complex valued.
j 2
H (e ) = H (e j  ) H (e j  )

k =1 (1 − k = 1
M M
z ke j  ) (1 − z 
kz )
H  (e j  ) = b0 H  (1/z ) = b0
k =1 (1 − k = 1
N N
pke j  ) (1 − p 
kz )
It follows that H (ej) is obtained by evaluating H (1/z∗) on the unit circle.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 14/36

 When h[n] is real or, equivalently, the coefficients {ak} and {bk} are real,
complex-valued poles and zeros occur in complex-conjugate pairs.
In this case, H (1/z∗) = H(z−1). Consequently, H (ej) = H(e−j), and
j 2
H (e ) = H (e j  ) H (e j  ) = H (e j  ) H (e − j  ) = H (z )H (z −1)
z =e j
H(z)H(z−1) is the z-transform of the autocorrelation sequence rhh[n] of h[n].
It follows that |H(ej)|2 is the Fourier transform of rhh[n].
3. Linear Time-Invariant Systems as Frequency-Selective Filters
▪ Filters are usually classified according to their frequency-domain
characteristics as lowpass, highpass, bandpass, and bandstop filters.
▪ These ideal filters have a constant-gain (usually taken as unity-gain) passband
characteristic and zero gain in their stopband, and a linear phase response.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 15/36

 ▪ Another characteristic of an ideal filter is a linear phase response. For
example, a filter with frequency response:
j Ce − jn0 , 1    2
H (e ) = 
0, otherwise
where C and n0 are constants
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 16/36

 Y (e j ) = X (e j )H (e j ) = CX (e j )e − jn0 , 1    2  y[n ] = Cx [n − n 0 ]
The filter output is simply a delayed and amplitude-scaled version of the
input signal.
A pure delay is usually tolerable and is not considered a distortion of the
signal. Neither is amplitude scaling.
▪ Therefore, ideal filters have a linear phase characteristic within their
passband, that is, () = −n0. d ( )
 gd ( ) = − group delay of the filter
d
▪ We interpret gd() as the time delay that a signal component of frequency 
undergoes as it passes from the input to the output of the system.
▪ Linear phase = group delay is constant.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 17/36

 For example, an ideal lowpass filter (LPF) with linear phase is defined by:
j e − jn0 ,   c sinc (n − n 0 )
H lp (e ) =  hlp [n ] =
0, otherwise  (n − n 0 )
▪ Note: this filter is not causal and it is not absolutely summable and therefore it
is also unstable. Consequently, this ideal filter is physically unrealizable.
▪ We can design some simple digital filters by the placement of poles and zeros
in the z-plane. The location of poles and zeros affects the frequency response
characteristics of the system.
▪ The basic principle underlying the pole–zero placement method is to locate
poles near points of the unit circle corresponding to frequencies to be
emphasized, and to place zeros near the frequencies to be deemphasized.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 18/36

 • All poles should be placed inside the unit circle in order for the filter to be
stable. However, zeros can be placed anywhere in the z-plane.
• All complex zeros and poles must occur in complex-conjugate pairs in order
for the filter coefficients to be real.
Usually, b0 is selected such that
k = 0 bkz k = 1
M −k −1 M
(1 − z k z )
H (z ) = = b0 |H( = 0)| = 1. 0 in the
1 + k =1 ak z −k k =1 (1 − pkz ) passband of the filter, N is
N N −1

selected to equal or exceed M.

Lowpass, Highpass, and Bandpass Filters
▪ Design of lowpass filters: poles should be placed near the unit circle at points
corresponding to low frequencies (near  = 0), zeros should be placed near or
on the unit circle at points corresponding to high frequencies (near  = ).
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 19/36

 ▪ The opposite holds true for highpass filters.
A single-pole filter with Transfer function:
1−a
H 1(z ) =
1 − az −1
The filter has unity gain at  = 0. The
gain of this filter at high frequencies is
relatively small.
The addition of a zero at z = −1 further
attenuates the response of the filter at
high frequencies.
1 − a 1 + z −1
H 2 (z ) =
2 1 − az −1
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 20/36

 In this case the magnitude of H2(ej) goes to zero at  = .
We can obtain simple highpass filters by reflecting a = 0.9
the pole–zero locations of the lowpass filters about
the imaginary axis in the z-plane.
1 − a 1 − z −1
H 3 (z ) =
2 1 + az −1

https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 21/36

 ▪ Basically, the bandpass filter should contain one or more pairs of complex-
conjugate poles near the unit circle, in the vicinity of the frequency band that
constitutes the passband of the filter.
▪ Example 5: Bandpass filter
Design a two-pole bandpass filter that has the center of its passband at  = /2,
zero in its frequency response characteristic at  = 0 and  = , and a
magnitude response of 1/ 2 at  = 4 /9.
The filter must have poles at p1,2 = re±j/2 and zeros at z = 1 and z = −1.
(z − 1)(z + 1) z2 − 1  2 1 − r2
H (z ) = G =G 2 , H ( = ) = G = 1G =
(z − jr )(z + jr ) z +r 2 2 1−r 2 2
2
4 (1 − r 2 )2 2 − 2cos(8 /9) 1
H ( = ) = =  r 2
= 0.7
9 4 1 + r + 2r cos(8 /9)
4 2 2
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 22/36

 1 − z −2
H (z ) = 0.15
1 + 0.7z −2

▪ A Simple Lowpass-to-Highpass Filter Transformation:

Frequency translation of  rad: Hhp() = Hlp( − )
hhp[n] = ejnhlp[n] = (−1)nhlp[n]
Lowpass filter by difference equation: y[n ] = −k =1 aky[n − k ] + k = 0 bk x[n − k ]
N M

k = 1 k = 0
N k k M
A highpass filter is derived: y[n ] = − ( − 1) a k y[n − k ] + ( − 1) bk x[n − k ]
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 23/36

 ▪ Example 6: Highpass filter
Convert the lowpass filter described by the DE y[n] = 0.9y[n − 1] + 0.1x[n] into
a highpass filter.
0.1
y[n] = −0.9y[n − 1] + 0.1x[n] ⇒H hp (e j  ) =
1 + 0.9e − j 
Digital Resonators
▪ A digital resonator is a special two-pole bandpass filter with the pair of
complex-conjugate poles located near the unit circle.
▪ A resonant peak at or near  = 0, we select the complex-conjugate poles at
p1,2 = re  j 0 , 0  r  1
▪ We can select up to two zeros. One choice is to locate the zeros at the origin.
The other choice is to locate a zero at z = 1 and a zero at z = −1.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 24/36

 ▪ Digital resonator with zeros at the origin:
j b0
H (e ) = j 0 − j  − j 0 − j  we select b0 so that |H( = 0)| = 1
(1 − re e )(1 − re e )
b0 = (1 − r ) 1 + r 2 − 2r cos 20

▪ Digital resonator with zeros z = 1 and z = −1

https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 25/36

 j (1 − e − j  )(1 + e − j  )
H (e ) = b0
(1 − re j 0e − j  )(1 − re − j 0e − j  )

All-Pass Filters
▪ The frequency response of an all-pass filter has constant magnitude (G  0) at
all frequencies, that is, |H(ej)| = G (G = 1).
▪ The simplest example of an all-pass filter is a pure delay system: H(z) = z−k.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 26/36

 ▪ A more interesting all-pass filter is described by the transfer function
k = 0 k
N −N + k
−1 −N + 1 −N a z
aN + aN −1z + + a1z +z
H (z ) = = , a0 = 1

−1 −N N
1 + a1z + + aN z az −k
k =0 k
−1
−N A(z )
, A(z ) = k = 0 ak z −k , a 0 = 1
N
H (z ) = z
A(z )
where all the filter coefficients {ak} are real.
j 2
H (e ) = A(z )A(1/z ) z =e j = 1
If z0 is a pole of H(z), then 1/z0 is a
zero of H(z).
The poles and zeros are reciprocals
of one another. a first-order and a second-order all-pass filter
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 27/36

 NR N
z −1 − k C (z −1 − k )(z −1 − k)
H ap (z ) =  −1  −1  −1
k =1 1 −  k z k =1 (1 −  k z )(1 −  kz )

where there are NR real poles and zeros and NC complex-conjugate pairs of
poles and zeros. For causal and stable systems, −1  k  1 and |k|  1.
− j j
e − re
▪ For a single pole–single zero all-pass system we have: H ap (e j  ) =
1 − re je − j 
r sin( −  ) d  ( ) 1 − r 2
ap ( ) = − − 2tan −1   gd ( ) = − ap =
1 − r cos( −  ) d 1 + r 2 − 2r cos( −  )
For a causal and stable system, r  1 and hence gd() ≥ 0. Since the group
delay of a higher-order pole–zero system consists of a sum of positive terms,
the group delay will always be positive.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 28/36

 4. Invertibility and minimum-phase systems
▪ An LTI system H(z) with input x[n] and output y[n] is said to be invertible if we
can uniquely determine x[n] from y[n]. T is one-to-one x[n] = T −1{T {x[n]}}
▪ The system Hi(z) that produces x[n] when excited by y[n] is called the inverse
system. h[n] ∗ hi[n] = [n] ⇒ H(z)Hi(z) = 1
1 A(z )
For rational system function: H i (z ) = =
H (z ) B (z )
The zeros of H(z) become the poles of the inverse system, and vice versa.
If H(z) is an FIR system, then Hi(z) is an all-pole system, and vice versa.
▪ Example 7: Determine the inverse of the system with h[n ] = ( 21 ) u[n ]
n

1 1
H (z ) = −1
, ROC: z  This system is both causal and stable
1 − 0.5z 2
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 29/36

 Since H(z) is an all-pole system, its inverse is FIR and is given by the system
function H i (z ) = 1 − 21 z −1  hi[n ] =  [n ] − 21  [n − 1]
▪ Example 8: Determine the inverse of the system with h[n ] =  [n ] − 21  [n − 1]
This is an FIR system and its transfer function is
−1 1 z
H (z ) = 1 − 2 z , ROC: z  0  H i (z ) =
1 = Two possible ROCs
1− 2z
1 −1 z−21

z  21  hi [n ] = ( 21 ) u[n ]
n
This system is both causal and stable
 hi [n ] = − ( 21 ) u[ − n − 1] This system is anticausal and unstable
n
z  1
2

▪ Since h[n] ∗ hi[n] = [n] does not, in general, possess a unique solution, we
assume that the system and its inverse are causal. Then,
k = 0 h[k ]hi[n − k ] =  [n ]
n

https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 30/36

 By assumption, hi[n] = 0 for n  0. For n = 0 we obtain hi[0] = 1/h[0].
The values of hi[n] for n ≥ 1 can be obtained recursively from the equation:
n h[n ]hi [n − k ]
hi [n ] = k =1 , n  1 ()
h[0]
There are two problems associated with (). First, the method does not work if
h[0] = 0. Second, the recursion gives rise to round-off errors which grow with n.
▪ Example 9: Determine the causal inverse of the FIR system with
h[n ] =  [n ] −  [n − 1]
Since h[0] = 1, h[1] = −, and h[n] = 0 for n ≥ 2, we have hi[0] = 1/h[0] = 1.
and hi[n] = hi[n − 1], n ≥ 1. hi[1] = , hi[2] = 2, ..., hi[n] = n
which corresponds to a causal IIR system as expected.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 31/36

 Minimum-Phase, Maximum-Phase, and Mixed-Phase Systems
▪ The invertibility of a linear time-invariant system is intimately related to the
characteristics of the phase spectral function of the system. Consider two FIR
systems: H1(z) = 1 + ½z−1 = z−1(z + ½), H2(z) = ½ + z−1 = z−1(½ z + 1).
H 1 (e j  ) = H 2 (e j  ) = 5
4
+ cos 
The magnitude are equal because the zeros of H1(z) and H2(z) are reciprocals.
−1 sin  −1 sin 
1 ( ) = − + tan 1 , 2 ( ) = − + tan
2 + cos 
2 + cos 
The phase for the first system
H1(z) with its zero (−½) inside the
unit circle undergoes a net phase
change 1() − 1(0) = 0 radians.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 32/36

 On the other hand, the phase for the second system H2(z) with its zero (−2)
outside the unit circle undergoes a net phase change 2() − 2(0) =  rad.
As a consequence, we call the first system a minimum-phase system and the
second system a maximum-phase system.
▪ An FIR system of length M + 1 that has M zeros. Its frequency response can
be expressed as: H(ej) = b0(1 − z1e−j)(1 − z2e−j) ··· (1 − zMe−j).
When all zeros are inside the unit circle, Minimum-phase: () − (0) = 0;
When all zeros are outside the unit circle, Maximum-phase: () − (0) = M.
▪ If the FIR system with M zeros has some of its zeros inside the unit circle and
the remaining zeros outside the unit circle, it is called a mixed-phase system.
▪ Extend to IIR systems that have rational transfer functions H(z) = B(z)/A(z).
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 33/36

 It is minimum-phase, if all its poles and zeros are inside the unit circle.
▪ For a stable and causal system, the system is maximum-phase if all the zeros
are outside the unit circle, circle, and mixed-phase if some, but not all, of the
zeros are outside the unit circle.
A stable pole-zero system that is minimum-phase has a stable inverse
which is also minimum-phase.
Maximum-phase systems and mixed-phase systems result in unstable
inverse systems.
▪ Decomposition of nonminimum-phase (mixed-phase) pole–zero systems. Any
nonminimum-phase pole–zero system can be expressed as:
H(z) = Hmin(z)Hap(z)
where Hmin(z) is a minimum-phase system and Hap(z) is an all-pass system.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 34/36

 Demonstration for the class of causal and stable systems with H(z) = B(z)/A(z).
B(z) = B1(z)B2(z), where B1(z) has all its roots inside the unit circle, B2(z) has
all its roots outside the unit circle, then
B1(z )B2 (z −1) B2 (z )
H min (z ) = , H ap (z ) =
A(z ) B2 (z −1)
Hap(z) is a stable, all-pass, maximum-phase system.
▪ Group delay of nonminimum-phase system. Based on the decomposition of a
non-minimum-phase system, the group delay of H(z) as:  gd ( ) =  gd ( ) +  gd ( )
min ap

Since  gd
ap
( )  0 for 0 ≤  ≤ , it follows that  gd ( )   gd
min
( ), 0 ≤  ≤ .
We conclude that among all pole–zero systems having the same magnitude
response, the minimum-phase system has the smallest group delay.
https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 35/36

 ▪ Example 10: Minimum-phase/allpass decomposition
1 + 5z −1
H (z ) =
1 + 21 z −1
z −1 + 15 z −1 + 15 1 + 15 z −1 1 + 15 z −1 z −1 + 15
H (z ) = 5 −1
=5 −1 −1
=5
1+ 2z
1 1+ 2z 1+ 5z
1 1 1 + 21 z −1 1 + 51 z −1
H min (z ) Hap (z )

▪ Note: a quick way to obtain the minimum-phase system is by replacing each

factor (1 + az−1), where |a|  1, by a factor of the form .

https://manara.edu.sy/

Transform analysis of LTI systems https://manara.edu.sy/ 2024-2025 36/36