Chapter 2

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Chapter 2

Realizability Theory and Positive Real Functions

2.1 Realizability criteria (Passive Networks)

Deriving-point functions of networks made up of linear, lumped, passive elements (resistors,


inductors, capacitors and transformers) are rational functions of s. But not all rational functions
of s describe a realizable network of RLCM elements.
According to Otto Brune1, a deriving-point function (Z(s) or Y(s)) is realizable using lumped
passive elements (R, L, C, M) elements if it is positive real (PR) rational function of s. i.e if it
satisfies the following conditions.

Or

1) H(S) (Z(s) or Y(s)) is a real rational function of s where all coefficients (of the numerator
and denominator polynomials) are real and positive.
2) If s (which is generally complex, s = σ + jω) has non-negative real part, H(s) should not
have negative real part. i.e. if Re[s] > 0, then Re[H(s)] > 0.

2.2 Positive Real Conditions

Equivalent positive-real conditions


A. H(s) should be quotient of polynomials with positive real coefficients
B. For all real ω, Re[H(jω)] ≥ 0
C. All poles and zeros of H(s) should lie in the closed Left-hand S-plane. All jω-axis poles
and zeros must be simple (multiplicity of one) with positive residues.

2.3 Hurwitz and Strictly-Hurwitz Polynomials


Definition:
A polynomial with its zeros restricted to the closed Left-hand S-plane is called Hurwitz
polynomial.
A polynomial with its zeros restricted to the inside if Left-hand S-plane (excluding the jω –axis)
is called Strictly Hurwitz polynomial.
A polynomial that is not Hurwitz is called Non-Hurwitz polynomial.

1
Otto Walter Heidrich Oscar Brune (1901– ) who found the necessary and sufficient conditions (positive real
conditions for realizable passive network functions at Massachusetts Institute of Technology (MIT) during his
doctorial thesis.

Page 1
Therefore, the positive real condition restricts the polynomials of H(s) to be Hurwitz or
strictly Hurwitz polynomial; if they are Hurwitz only, their jω -axis zeros must be simple.

The following are zero locations of examples of strictly Hurwitz, Hurwitz and non-Hurwitz
polynomials.

a) P(s) = (s + 3)(s2 + 2s + 2)

b) P(s) = s(s +1)(s2 + 4)

c) P(s) = (s+2)(s2 – 4s + 5)

Properties

A strictly Hurwitz polynomial has the form:

Page 2
P(s) = ansn + an-1sn-1 +… + a1s1 +a0
Where, ai > 0 (no missing terms)
A Hurwitz polynomial may have missing terms (every second term) due to zeros of the form (s2
+ ω2) that lie on the jω-axis.

2.4 Tests for Hurwitz nature of polynomials ( P(S))

Method 1 – FACTORIZATION

Factorize P(s), find all zeros (roots of the polynomial) and inspect their location on the s-plane.

Method 2– Continued Fraction Expansion (CFE)

 Separate P(s) in to even (Pe(s)) and odd (Po(s)) parts.


 Form φ(s) = Pe(s)/Po(s) or Po(s)/Pe(s). Select the higher degree to be
numerator.
 Expand φ(s) using CFE (Continued Fraction Expansion)

The expansion stops when remainder of the subsequent division is zero. If there
is no premature termination, denominator of the last division is a constant.
If the last denominator is not a constant, then we say there is premature
termination.

The polynomial P(s) is strictly Hurwitz if the all coefficients, αi, are real and positive, and if
there is no premature termination.
If the coefficients are real and positive, then P(s) is Hurwitz. (*** there can be premature
termination)
Otherwise, the polynomial is non-Hurwitz.

Example 1
Test for Hurwitz nature of the following polynomial using CFE.
P(s) = (s2 + 2s + 1)(s2 + s + 1)(s2+4)
= s6 + 3s5 + 8s4 + 15s3 + 17s2 + 12s + 4

Page 3
Solution:

 all are real positive coefficients


 there is premature termination; if there were no premature termination,
there would be 6 (the degree of P(s)) terms in the expansion
 Therefore P(s) is simply Hurwitz. This can also be verified from the zero
plot of P(s) shown above.

Problem 1
Test for Hurwitz nature of the following polynomial using CFE

P(s) = s4 + s3 + 5s2 +3s+2

Method 3– Routh-Hurwitz Array

For P(s) = ansn + an-1sn-1 +… + a1s1 +a0


Construct a triangular array as follows:

Page 4
Where the ai’s are the polynomial coefficients, and the coefficients bi, ci,… are computed as
follows:

Special cases:

when all coefficients in a row are zero ( Vanishing row)


This occurs when the polynomial has roots on the jw-axis. The polynomial with coefficients in a
row prior to the vanishing row is a factor of P(s), and it has roots on the jw-axis.
k+1 k
If the vanishing row is at s , create a polynomial, Pk(s), using coefficients of a row at s
(prior to the vanishing row)

Pk(s) = u1sk + u2sk-2 + u3sk-4 + u4sk-6 +…

Define an auxiliary polynomial, dPk(s)/ds, and replace the vanishing row with coefficients of this
polynomial (derivative of Pk(s))
Continue the process until you reach the last row at s0.

the first element of a row is zero, and there is at least one non-
zero coefficient in a row

Page 5
If a row begins with a zero and the row has some non-zero coefficients, replace the leading zero
(the first zero) with ε (assume a small positive number) and find the remaining coefficients (in
terms of ε).

Cases
 If there is no sign change in coefficients of the first column, and if there is no vanishing
row, the polynomial is strictly Hurwitz. (If there are coefficients that contain ε,
evaluate sign of the coefficients by taking the limit ε→ 0+).

 If there is at least one sign change, the polynomial is non-Hurwitz.

 If there exists vanishing row (root on the jw-axis), the polynomial is not strictly-
Hurwitz, but if there is still no sign change, the polynomial is Hurwitz. i.e. The
polynomial is Hurwitz if there is no sign change. (Vanishing rows are allowed)

Example 2
Test Hurwitz nature the following polynomial using Routh-Hurwitz array
a) P(s) = s4 + s3 + 5s2 +3s+2

Solution:

b) P(s) = s4 + 4s3 + 5s2 +8s+6

Solution:

Page 6
2.5 Sturm’s Theorem and Sturm’s Test

Recall that there are three conditions for a network function to be positive real.
A. H(s) should be quotient of polynomials with positive real coefficients
B. For all real ω, Re[H(jω)] ≥ 0
C. All poles and zeros of H(s) should lie in the closed Left-hand S-plane. All jω-axis poles
and zeros must be simple (multiplicity of one) with positive residues.

The first condition can be checked by inspection, and the last condition restricts denominator of
H(s) to be strictly Hurwitz or simply Hurwitz. If it is Hurwitz only its jω-axis roots must be
simple.
Sturm’s test is useful for testing condition B. Consider a network function H(s) = N(s)/D(s),
where N(s) is the numerator and D(s) is denominator of H(s).
First, separate both polynomials in to even and odd parts.

Property D(-s) = De(s) – Do(s)

Now, multiply H(s) by D(-s)/D(-s)

Page 7
= He(s) + Ho(s)

Where, De(s)2 – Do(s)2 = D(s)D(-s)


For s = jω
He(jω) – is purely real
Ho(jω) – is purely imaginary.
De(s)2 – Do(s)2 = D(s)D(-s) = D(jω)D(-jω) = |D(jω)|2 which is always positive.
Ne(jω)De(jω) – No(jω)Do(jω) = P(ω2) is an even polynomial function of ω

Therefore, Re[H(jω)] = He(jω)


The second positive real condition (condition B) states that:
For all real ω, Re[H(jω)] ≥ 0

►He(jω) ≥ 0

where He(jω) = P(ω2)/ |D(jω)|2


Now condition B can be restated as:
P(ω2) ≥ 0 for all real ω OR for ω2 ≥ 0
2
Let x =ω

P(x) = anxn + an-1xn-1 + …+ a1x + a0 ≥ 0 for x ≥ 0

This implies that the graph of P(x) should not cross the x – axis in the right half of the xy – plane
(positive values of x)

Cases:

 If all coefficients ai are positive, then P(x) ≥ 0 for x ≥ 0 (condition met)


 P(x) = a0 at x = 0 hence, ao ≥ 0 (necessary)
 P(x) → an as x → ∞ hence, an ≥ 0 (necessary)
 If a0 ≥ 0, an ≥ 0 but ai < 0 for some i, then P(x) ≥ 0 for x ≥ 0 provided that P(x) does not
have odd-ordered zeros on the positive x-axis (except at x = 0 and x → ∞). If the
multiplicity of zeros of P(x) is odd, then the graph of P(x) crosses the x-axis at those
points (condition not met); otherwise, if multiplicity of zeros of P(x) is even, the graph
of P(x) touches the x-axis at those points. Therefore, even-ordered (even multiplicity)
zeros (x-intercepts) are allowed, but odd-ordered zeros are not.

Page 8
Plot the following polynomials and observe the nature of zeros.
a) P(x) = (x-1)2
b) P(x) = (x-1)3
c) P(x) = x(x2 - 4x + 4)

Test for condition B is reduced to two steps


1. Check that both a0 and an are non-negative.
2. Check that P(x) has no odd-ordered zeros on the positive x-axis. This can be done in
three ways:

Method 1 – Factorization
Method 2 – Plot P(x) over a sufficient range of x
Method 3 – Sturm’s test

Sturm’s Test

Step 1: Develop a sequence of polynomials Po(x), P1(x), P2(x)… called Sturm’s


functions as follows:

Define P0(x) = P(x)


P1(x) = P’(x)
To find P2(x), divide P0(x) by P1(x) to get a two term quotient and a remainder. The remainder is
negative of P2(x).
i.e. P0(x) / P1(x) = b1x + c1 + [-P2(x)] / P1(x)
Repeat this stem for P3, P4 …

Euclid Algorithm:
P0(x) = P(x)
P1(x) = P’(x)
P0(x) = q1(x)P1 + [-P2(x)]
P1(x) = q2(x)P2 + [-P3(x)]


Pk-2(x) = q1(x)Pk-1 + [-Pk(x)]
The process stops when the remainder Pk(x) becomes a constant (when k = n) or zero ( when k ≤
n premature termination)

Step 2:
Case 1: Pk(x) = constant when k = n

Sturm’s Theorem

Page 9
The number of odd-ordered zeros which Pk(x) has in the interval a ≤ x ≤ b is equal to |Sb – Sa|
where Sa and Sb are the number of sign changes in the test (P0, P1,P2,…) evaluated at x=a and
x=b respectively.
Here, we are interested in the presence of odd-ordered zeros on the positive x-axis, hence we
take a = 0 and b → ∞.

Case 2: Pk(x) = 0 for some k ≤ n (Premature termination)


This shows that Pk-1(x) is a factor of P(x) so that all zeros of Pk-1 are zeros of P(x) and the
multiplicity of these zeros in P(x) is one higher than their multiplicity in Pk-1(x). The test
continues by taking polynomials P0 to Pk-1. The zero count |Sb – Sa| in this case is the sum of
number of odd-ordered zeros plus multiple zeros (due to Pk-1(x) each counted once.

Example 3
Test whether the following polynomials satisfy the condition P(x) ≥0 for all x ≥ 0.
a) p(x) = x2 – 4x + 3

Solution:
Method – 1
P(x) = (x – 1)1(x – 3)1
 two odd-ordered zeros on the positive x – axis (x =1 and x = 3)
 Therefore, P(x) does not satisfy the condition.

Method – 2
Method – 3 Sturm’s test n=2

P0(x) = x2 – 4x + 3

P1(x) = P1’(x) = 2x – 4

Page 10
P2(x) = - (-1) = 1

P2(x) = 1 ← constant when k = 2 = n

|Sb – Sa| = 2 ← the number of odd-ordered zeros on the positive x – axis


Condition not satisfied.

b) p(x) = x4 – 8x3 + 23x2 – 28x + 12

Solution:
Sturm’s test n = 4

P0(x) = x4 – 8x3 + 23x2 – 28x + 12


P1(x) = P0’(x) = 4x3 – 24x2 + 46x – 28
Using the Euclid algorithm:
P2(x) = ½x2 – 2x + 2
P3(x) = 2x – 4
P4(x) = 0 k=4 ≤ n = 4 ← premature termination.

 P3(x) = 2(x-2)1 is a factor of P(x)


 The multiplicity of the zero (x-2) in P(x) is one higher than in P3(x)
 (x-2)1+1 = (x-2)2 is a factor of P(x)
 P(x) has one double (even-ordered) zero

 P(x) has 3 zeros on the positive x – axis among which the double zero (x-2)2 is the one.

 Therefore, P(x) has 3-1 = 2 odd-ordered zeros.

 Hence, P(x) 0 for x ≥ 0

Page 11
Problem 2

Check whether p(x) = x4 – 15x2 + 10x + 24 satisfies the condition P(x) ≥0 for all x ≥ 0.

2.6 Testing Deriving-point functions

Recall that according to Otto Brune2, a deriving-point function (Z(s) or Y(s)) is realizable using
lumped passive elements (R, L, C, M) elements if it is positive real (PR) rational function of s.
If Z(s) is positive real, Y(s) (=1/z(s)) is also positive real.
Let us consider Z(s) = N(s)/D(s) (the same conclusion can be reached for Y(s))

The following condition must be satisfied.

A. Z(s) should be quotient of polynomials with positive real coefficients


B. For all real ω, Re[Z(jω)] ≥ 0
C. All poles and zeros of Z(s) should lie in the closed Left-hand S-plane. All jω-axis
poles and zeros must be simple (multiplicity of one) with positive residues.

2.6.1 General realizability criteria


The following are procedures for testing positive-real nature of Z(s) with all common factors in
the numerator and denominator been removed.

1. Inspection test for necessary conditions


a) all polynomial coefficients are real and positive
b) Degree of N(s) and D(s) differ at most by1

2
Otto Walter Heidrich Oscar Brune (1901– ) who found the necessary and sufficient conditions (positive real
conditions for realizable passive network functions at Massachusetts Institute of Technology (MIT) during his
doctorial thesis.

Page 12
c) Lowest degrees of N(s) and D(s) differ at most by 1
d) There should be no missing terms in N(s) and D(s) unless all even or
odd terms are missing.
e) jω – axis poles and zeros must be simple

2. Test for necessary and sufficient conditions


a) Re[Z(jω)] ≥ 0 for all real ω
Or P(x) ≥ 0 for all x ≥ 0
b) N(s) + D(s) must be strictly Hurwitz

Example 4
Test positive real nature of the following Dp impedance function.

Solution:
1. Inspection test.
 Real and positive coefficients
 Highest degrees differ by 1-1 = 0 ≤ 1
 Lowest degrees differ by 0-0 = 0 ≤ 1
 No missing terms
 No imaginary axis poles and zeros
Ok!
2. Necessary and sufficient conditions

P(ω2) = NeDe – NoDo |s = jω


= (3)(1) – (j2ω)(jω)
= 2ω2 + 3

P(x) = 2x + 3 ← all coefficients are positive


 P(x) ≥ 0 for all x ≥ 0
 Re[z(jω)] ≥ 0 for all real ω Ok!

 N(s) + D(s) = 3s + 4 = 3(s + 4/3) ← zero at s = -4/3


 N(s) + D(s) is strictly Hurwitz
 All condition met

Z(s) is positive real

Problem 3
Test the following admittance function.

Page 13
Example 5
Test positive realness of the following impedance function

Solution:
1. Inspection test.
 Real and positive coefficients
 Highest degrees differ by 3-3 = 0 ≤ 1
 Lowest degrees differ by 0-0 = 0 ≤ 1
 No missing terms
 Test for imaginary axis poles and zeros.
 This can be done using Routh-Hurwitz array; there will be vanishing row if a polynomial
has imaginary axis (jω – axis) roots.

Pk(s) = 3s2 + 3 = 3(s2 + 1)1 is a factor of P(s)

Page 14
P(s) has one simple root.
 Z(s) has one simple zero in the jω – axis

Ok!

2. Necessary and sufficient conditions

P(ω2) = NeDe – NoDo |s = jω


= (3s2 + 3)(3s2 + 1) – (2s3 + 2s)(s3 + 4s)| s = jω
= 2ω6 – ω4 - 4 ω2 + 3
P(x) = 2x3 – x2 – 4x + 3

 Sturm’s Test
P0(x) = 2x3 – x2 – 4x + 3
P1(x) = 6x2 – 2x – 4
P2(x) = 25/9x – 25/9
P3(x) = 0 ← premature termination
P2(x) = 25/9(x – 1)1
 (x-1)1+1 = (x-1)2 is a factor of P(x)
 P(x) has one double zero

 P(x) has one zero on the positive x – axis among which the double zero (x-1)2 is the one.
 Therefore, P(x) has 1-1 = 0 odd-ordered zeros.
 No odd-ordered zeros on the positive x - axis
 Hence, P(x) ≥0 for x ≥ 0

N(s) + D(s) = 3s3 + 6s2 + 6s + 4

Z(s) is positive real

Page 15
Problem 4
Show that the following impedance function is positive real

Page 16

You might also like