Chapter 3

Network Reailzability
Outline
3.1. Introduction

3.2. Causality and Stability

3.3. Hurwitz Polynomials

3.4. Positive Real Functions

3.5. Sturm’s Test

• Reference: F.F. Kuo, "Network Analysis and Synthesis" Chapter 10

V.K. Aatre, "Network Theory and Filter Design“: Chapter 9

AAU, AAiT, 2019 2

Introduction
• In the last chapter we were concerned with the problem of identifying the
response given the excitation and network.
• When we discuss about synthesis we are concerned with the problem of
constructing a network given the excitation and response.
• The starting point for any synthesis is the system function
R( s)
H (s) 
E (s)
• The first step in synthesis procedure is determining whether the system
function can be realized with a physical passive network
• There are two considerations
• Causality
• Stability
AAU, AAiT, 2019 3
Causality
• By causality we mean that a voltage doesn’t appear between any terminals in
the network before a current/voltage is applied.
• In other words, the impulse response of the network must be zero for 𝑡 < 0.
h(t )  0 for t  0
• Some examples:
h(t )  e  at u(t) : is causal because for t  0 u(t)  0 , hence, h(t)  0.
a t
h(t )  e : is not causal because for t  0 , h(t ) is not zero.

AAU, AAiT, 2019 4

Causality…
• In certain cases, the network can be made causal by delaying the impulse
response by a certain time delay.

• If we delay ℎ(𝑡) by 𝑇, then ℎ(𝑡 − 𝑇) will be causal.

AAU, AAiT, 2019 5

Paley-Wiener criterion
• In the frequency domain causality is implied when the Paley-Wiener
criterion is satisfied.
• The Paley-Wiener criterion states that a necessary and sufficient condition
for causality is

log H ( j )

  12
d  

• The physical implication of the Paley-Wiener criterion is that the amplitude

response of a causal network can’t be zero over a finite band of frequency.
• Another way of looking at the Paley-Wiener criterion is that the amplitude
function cannot fall off to zero faster than exponential order.

AAU, AAiT, 2019 6

Examples
• The ideal low pass filter
• Is not causal because it is zero
for all frequencies 𝜔 > 𝜔𝑐 .

• The Gaussian filter

H ( j )  e  
2

• Is not causal because


log H ( j ) 
2

  2 1
d   2
   1
d

is not finite.

AAU, AAiT, 2019 7

Examples…
• The amplitude function
1
H ( j ) 
 2 1
• Is causal because


log H ( j ) 
log( )

  1
2
d  
  1
2
d  

AAU, AAiT, 2019 8

Stability
• If a network is stable, then for a bounded excitation 𝑒(𝑡) the response will
also be bounded. | e(t ) | C 0  t  
1

| r (t ) | C2 0  t  
• Where 𝐶1 and 𝐶2 are real, positive and finite numbers.
• If a linear system is stable, then we get from the convolution integral and the
above definition of stability 
| r (t ) | C1  h( ) d  C2
0
• The above equation implies that the impulse response be absolutely
integrable. 

 h( ) d  
0

• One important requirement for ℎ(𝑡) to be absolutely integrable is, ℎ(𝑡)

approach zero as 𝑡 increases to infinity.
AAU, AAiT, 2019 9
Stability…
• Note that our definition of stability implies systems with sin 𝜔𝑡 impulse
response are not stable because


 sin t dt  
0

• However, a simple 𝐿𝐶 network has such an impulse response.

• Since we don't want to call these systems unstable, we call them marginally
stable if they satisfy the following criterion.

lim h(t )  0
t 
and h(t )  C all t

• Stability in the frequency domain implies that the system function should
only have poles on the left had side of the ‘𝑠’ plane or simple poles on the 𝑗𝜔 axis.
AAU, AAiT, 2019 10
Stability…
• This is because if we have a pole on the right hand side, then the impulse
response will have an exponentially increasing term, 𝑒 𝜎𝑡 .
• Hence, our response will not be bounded.
• If there is a double pole on the 𝑗𝜔 axis, then the impulse response of the
network will have a term 𝑡sin 𝜔𝑡.
• This will not be bounded.

AAU, AAiT, 2019 11

Stability Criterion
• If 𝐻(𝑠) is given as bm s m  bm 1 s m 1  ...  b1 s  b0
H (s) 
an s n  an 1 s n 1  ...  a1 s  a0

• Due to the requirement of simple poles on the 𝑗𝜔 axis, the order of the
numerator shouldn’t exceed the order of the denominator by more than 1.
That is m  n  1
• If m  n  1 then there would be multiple poles on the 𝑠 = 𝑗𝜔 = ∞.
• To summarize, for a network to be stable the following three conditions
must be satisfied
𝐻(𝑠) can’t have poles on the right side of the ‘𝑠’ plane.
𝐻(𝑠) can’t have multiple poles on the 𝑗𝜔 axis.
The degree of the numerator of 𝐻(𝑠) can’t exceed that of the denominator by
more than 1.
AAU, AAiT, 2019 12
Hurwitz Polynomials
• We mentioned in the previous section that in order for a system to be stable,
its poles must lie in the left side of the ‘𝑠’ plane; moreover the poles on the 𝑗𝜔
axis must be simple.
• The denominator polynomial of a system function 𝐻(𝑠) that satisfies these
criteria belongs to a class of polynomials called Hurwitz polynomials.
• A polynomial 𝑃(𝑠) is said to be Hurwitz if it satisfies
• 𝑃(𝑠) must be real if 𝑠 is real.
• The real part of its roots must be negative or zero.
• As a result of these conditions, if 𝑃(𝑠) is a Hurwitz polynomial given by
P(s)  an s n  an1s n1  ...  a1s  a0
• Then all coefficients 𝑎𝑛 must be real and if 𝑠𝑖 = 𝛼𝑖 + 𝑗β𝑖 is root of 𝑃(𝑠), then 𝛼𝑖
must be negative.
AAU, AAiT, 2019 13
Examples
1. The polynomial
 
P(s)  (s  1) s  1  j 2 s  1  j 2 
is Hurwitz because
• For real s 𝑃(𝑠) is real, 𝑃(𝑠) = (𝑠 + 1)(𝑠 2 + 3𝑠 +2)
• None of the roots lie on the right hand side of the ‘𝑠’ plane
2. G ( s )  ( s  1)(s  2)(s  3)

is not Hurwitz
• The root 𝑠 = 1 lies on the positive ‘𝑠’ plane.

AAU, AAiT, 2019 14

Properties of Hurwitz polynomial
i. All the coefficients of the polynomial are non negative.
• This is readily seen by examining the types of terms 𝑃(𝑠) can have

P( s)  s   i  s 2  i
2
s      
i
2
i
2
s   i
s   j i
Simple pole on
Simple Complex s   i   i
real pole the jw axis conjugate roots

• The multiplication of these non negative coefficients can only give non
negative coefficients.

AAU, AAiT, 2019 15

Properties of Hurwitz polynomial…
ii. The even and odd parts of 𝑃(𝑠) have roots on the 𝑗𝜔 axis only.
• If we denote the odd and even parts of 𝑃(𝑠) as 𝑛(𝑠) and 𝑚(𝑠)

P ( s )  n ( s )  m( s )

ODD EVEN

• If 𝑃(𝑠) is either even or odd, all its roots are on 𝑗𝜔 axis.

AAU, AAiT, 2019 16

Properties of Hurwitz polynomial…
iii.The continued fraction expansion of
𝜓 𝑠 = 𝑛(𝑠)/𝑚(𝑠) or 𝜓 𝑠 = 𝑚(𝑠)/𝑛(𝑠)
of a Hurwitz polynomial yields only positive quotient terms.
n( s ) 1
 (s)   q1 s 
m( s ) 1
q2 s 
1
q3 s 
1
q4 s 
.....
1
qn s
• All the 𝑞𝑖 ’s must be real and positive. (𝑞1 ≥ 0)

AAU, AAiT, 2019 17

Properties of Hurwitz polynomial…
• A theorem in the theory of continued fractions which states that the
continued fraction expansion of the even to odd or odd to even parts of a
polynomial must be finite in length.
• Another theorem states that, if the continued fraction expansion of the odd
to even or even to odd parts of a polynomial yields positive quotient terms,
then the polynomial must be Hurwitz to within a multiplicative factor 𝑊(𝑠).
F ( s )  W ( s ) F1 ( s )
• 𝐹(𝑠) is Hurwitz, if 𝑊(𝑠) and 𝐹1 (𝑠) are Hurwitz.

AAU, AAiT, 2019 18

Example
• Test whether the polynomial F ( s)  s  s  5s  3s  4 is Hurwitz.
4 3 2

• Solution:
m( s )  s 4  5s 2  4 and n( s )  s 3  3s

• Perform a continued fraction expansion of 𝜓(𝑠) = 𝑛(𝑠)/𝑚(𝑠) or 𝜓(𝑠) = 𝑚(𝑠)/𝑛(𝑠)

• Since the order of 𝑚(𝑠) is (one) higher than 𝑛(𝑠), we start with 𝑚(𝑠)/𝑛(𝑠).
m( s ) s 4  5s 2  4 2s 2  4 1 1
  s   s   s 
n( s ) s 3  3s s 3  3s s 3  3s 1 s
s 
2s 2  4 2 2s 2  4
1 1 1
 s  s  s
1 1 1 1 1 1
s s s
2 2 s 2
 4 2 4 2 1
2s  2s 
s s s
• Note that all the coefficients of the quotients are positive. 4
• Hence 𝐹(𝑠) is Hurwitz.
AAU, AAiT, 2019 19
Exercise
• Test whether the following polynomial are Hurwitz using continued
fraction expansion.
(a) F ( s )  s 4  16 s 3  86 s 2  176 s  105
(b) F ( s )  s 5  12 s 4  45s 3  60 s 2  44 s  48
(c) G ( s )  s 3  2 s 2  3s  6
(d) F ( s )  s 7  2 s 6  2 s 5  s 4  4 s 3  8s 2  8s  4

AAU, AAiT, 2019 20

Positive Real Functions
• A fundamental theorem, in passive network synthesis states that “a rational
function of 𝑠 is the driving point immittance of a passive network if, and
only if, it is positive real.”
• A rational function 𝐹(𝑠) is a positive real function (prf) if it satisfies the constraints
• F(s) is analytic in 𝑠 plane
• 𝐹(𝑠) real for real 𝑠 Brune's Conditions
• 𝑅𝑒 𝐹 𝑠 ≥ 0 𝑓𝑜𝑟 𝑅𝑒 𝑠 ≥ 0
• If 𝐹(𝑠) is a prf then the reciprocal of 𝐹( 𝑠) (i.e. 1/𝐹(𝑠)) and 𝐹(1/𝑠) are also prf.
• If 𝐹( 𝑠) is expressed as : N (s)
F (s) 
D( s)
where 𝑁(𝑠) and 𝐷(𝑠) are polynomials in 𝑠, and both 𝑁(𝑠) and 𝐷(𝑠) must have real coefficients.
• This ensures that 𝐹(𝑠) is real when 𝑠 = 𝜎, 𝑁(𝑠) and 𝐷(𝑠) being real for 𝑠 = 𝜎, complex roots
must appear in conjugate pairs.
AAU, AAiT, 2019 21
Positive Real Functions…
• A prf 𝐹(𝑠) cannot have any poles and zeros in the right half (RH) 𝑠-plane.
• Any roots of 𝑁(𝑠) and 𝐷(𝑠) on the 𝑗𝜔-axis must be simple.
• The residues of 𝐹(𝑠) and 1/𝐹(𝑠) at the simple poles on the 𝑗𝜔-axis must be
real and positive.
• Thus the numerator and denominator polynomials of a prf must be Hurwitz.
• A polynomial 𝐹(𝑠) is :
• Strictly Hurwitz polynomial if all its roots are in left half (LH) 𝑠-plane.
• Hurwitz if its roots are in LH plane or simple on the 𝑗𝜔-axis
• The analyticity of 𝐹( 𝑠) and 1/𝐹(𝑠) or the Hurwitz nature of 𝑁 𝑠 and 𝐷(𝑠)
can be tested in a number of ways.
• Routh-Hurwitz array, Continued fraction expansion and Strum’s Test

AAU, AAiT, 2019 22

Routh-Hurwitz array
• Let P ( s )  an s n  an 1 s n 1  ...  a1 s  a0
• For 𝑃(𝑠) to be Hurwitz all the coefficients 𝑎𝑖 . 𝑖 = 0 … 𝑛, must be of the same sign.
• No coefficient should be missing for a strictly Hurwitz polynomial.
• For a polynomial which has all its zeros on the 𝑗𝜔-axis alternate coefficients
must be zero, i.e. 𝑃(𝑠) is either even or odd polynomial.
• The Routh-Hurwitz array is constructed as follows:
sn an an  2 an  4 
s n 1 an 1 an  3 an  5 
s n2 bn bn 1 bn  2
s n 3 cn cn 1
 
s1
s0
AAU, AAiT, 2019 23
Routh-Hurwitz array…
an an  2 an an  4
 
an 1 an  3 an 1 an  5
where bn  , bn 1 
an 1 an 1
an an  6

an 1 an  7
bn  2 
an 1
an 1 an  3 an 1 an  5
 
bn bn 1 bn bn  2
cn  , cn 1 
bn bn

AAU, AAiT, 2019 24

Routh-Hurwitz array…
• For a strictly Hurwitz polynomial
• No change of sign in the coefficients of the first column, i.e. 𝑎𝑛 , 𝑎𝑛−1 , 𝑏𝑛 , 𝑐𝑛 , …
• No vanishing row
• A vanishing row in the array indicates that the polynomial cannot be strictly
Hurwitz.
• In this case, the vanishing row is replaced by an auxiliary polynomial.
• Let the 𝑘𝑡ℎ row, the row corresponding to 𝑠 𝑘 , be the row prior to the
vanishing row,
Pk ( s )  ak s k  ak  2 s k  2  ...  a0
will always be an even polynomial.

AAU, AAiT, 2019 25

Routh-Hurwitz array…
• A vanishing row shows the presence of equal and opposite roots and the order of
the polynomial 𝑃𝑘 (𝑠) indicates the number of equal but opposite roots.
• An auxiliary polynomial is defined as
dPk ( s )
Pak ( s )   k k s  (k  2)ak  2 s  ...
k 1 k 3

ds
• The coefficients of (𝑘 + 1)𝑡ℎ row are replaced by the coefficients of the auxiliary
polynomial and the array is now continued.
• If there is a change of sign in the first column then the polynomial 𝑃(𝑠) is not
Hurwitz.
• If there are no changes of sign in the first column then the polynomial 𝑃𝑘 (𝑠) is
tested for its zeros and if the zeros are simple on the 𝑗𝜔-axis then 𝑃(𝑠) is Hurwitz

AAU, AAiT, 2019 26

Examples
Check whether
(a) P ( s )  s 4  16 s 3  86 s 2  176 s  105
(b) P ( s )  s 5  12 s 4  45s 3  60 s 2  44 s  48
(c) P ( s )  s 5  20 s 4  147 s 3  484 s 2  612 s  336

are strictly Hurwitz using Routh-Hurwitz array?

AAU, AAiT, 2019 27

Strum's Test
• A rational function 𝐹(𝑠) with real coefficients is a prf if, and only if,
• 𝐹(𝑠) is analytic in RH 𝑠-plane with at most simple poles on the 𝑗𝜔-axis
• The residues at any poles on the 𝑗𝜔-axis are real and positive.
Re{F ( j )}  0, for all 

A( s ) Ae ( s )  Ao ( s )
• Let F (s)  
B ( s ) Be ( s )  Bo ( s )
• if 𝐹(𝑠) is even ⟹ 𝐹 𝑗𝜔 = REAL
• else if 𝐹(𝑠) is odd ⟹ 𝐹 𝑗𝜔 = IMAGINARY

AAU, AAiT, 2019 28

Strum's Test…
• On the 𝑗𝜔-axis the Re 𝐹 𝑗𝜔 is the even part of 𝐹 𝑗𝜔 .

 Ae ( j )  Ao ( j )  Be ( j )  Bo ( j ) 
F ( j )    ; with Be ( j )  Bo ( j ) for 
 Be ( j )  Bo ( j )  Be ( j )  Bo ( j ) 
 Ae ( j ) Be ( j )  Ao ( j ) Bo ( j )   Ao ( j ) Be ( j )  Ae ( j ) Bo ( j ) 
  
 


     
2 2 2 2
 B e
( j ) Bo
( j )   B e
( j ) Bo
( j ) 
 ReF ( j ) j ImF ( j )

• For 𝐹(𝑠) to prf: Re{F ( j )}  0, for all 

• Thus: Check if N ( 2 )  Ae ( j ) Be ( j )  Ao ( j ) Bo ( j )  0
• Note: Be 2 ( j )  Bo 2 ( j )  0 is positve and real for 

AAU, AAiT, 2019 29

Strum's Test…
• If all the coefficients of 𝑁(𝑥) where 𝑥 = 𝜔2 are positive then 𝑁(𝑥) is positive
for all values of 𝑥 between 0 and ∞.
• When all the coefficients of 𝑁(𝑥) are not positive we can resort to Sturm's
test.
 N 0 ( x)  N ( x)
dN ( x)
 N1 ( x ) 
dx
 1 x   0  
N i  2 ( x) N i ( x)
 , i2
N i 1 ( x) N i 1 ( x)
where  0 & 1 are constants.
• This procedure is continued till 𝑁𝑛 is a constant.

AAU, AAiT, 2019 30

Strum's Theorem
• The number of zeros of 𝑁(𝑥) in the interval 𝑎 = 0 ≤ 𝑥 ≤ (𝑏 → ∞) is equal
to 𝑆𝑏 − 𝑆𝑎 , where 𝑆𝑎 and 𝑆𝑏 are the number of sign changes in the set
𝑁0 , 𝑁1 , 𝑁2 , … , 𝑁𝑛
evaluated at 𝑥 = 𝑎 and 𝑥 = 𝑏, respectively.
• By selecting 𝑎 = 0 and 𝑏 = ∞, we can find the number of zeros of 𝑁(𝑥) from
0 to ∞.
• Every time 𝑁(𝑥) goes through a zero, the sign of 𝑁(𝑥) changes.
• If there are no zeros of 𝑁(𝑥) in the interval 0 ≤ 𝑥 ≤ ∞, then 𝐹(𝑠) is prf.

AAU, AAiT, 2019 31

Examples
1. Check whether s 3  9 s 2  23s  15
(a) F ( s ) 
s 2  6s  8
2 s 4  7 s 3  11s 2  12 s  4
(b) F ( s )  4
s  5s 3  9 s 2  11s  6
s 4  2 s 3  3s 2  s  1
(c) F ( s )  4 3
s  s  3s 2  2 s  1
are prf using Sturm’s test.
2. Determine the minimum value of the real positive constant 𝛼 for which the
following network functions is positive real and hence realizable.
2s 2  s  2
F (s)  2
2 s  1  2 s  

AAU, AAiT, 2019 32