On the Classication of Series-Parallel Electrical and Mechanical

Networks
Jason Zheng Jiang and Malcolm C. Smith
AbstractThe purpose of this paper is to continue to develop
the recently introduced concept of a regular positive-real func-
tion and its application to the classication of low-complexity
two-terminal networks. Previous work showed that series-
parallel ve-element networks with two reactive elements are
always regular and that there are four quartets of series-parallel
networks with three reactive elements and three resistive ele-
ments which may realise non-regular biquadratic immittances.
In this paper we consider six-element networks with four
reactive elements. We describe a classication procedure to
nd an efcient subset of such networks which may realise any
non-regular biquadratic that can be synthesised by this class of
networks. We also describes a method to derive the necessary
and sufcient realisablity conditions for a given network. This
paper builds on previous work of the authors as well as the
work of Vasiliu. The results are motivated by an approach
to passive mechanical control which makes use of the inerter
device.
I. INTRODUCTION
The notion of a positive-real function was introduced
in the seminal paper of Brune [1] where it was shown
that the impedance of any passive two-terminal electrical
network must be positive-real and, conversely, any positive-
real function could be realised as the driving-point im-
mittance of a network consisting of resistors, capacitors,
inductors and transformers. Later, Bott and Dufn showed
that transformers were not necessary [2]. The construction
in [2] had the appearance of being wasteful in terms of the
number of elements used, and much effort in the electrical
circuit literature was applied to shed light on this matter,
without a full resolution being obtained. This paper is a
continuation of [3], [4], [5] which has revisited this question.
In particular, we exploit the concept of a regular positive-
real function to study the classication of low-complexity
network realisations.
The paper is motivated by a new network element (the
inerter) which was introduced for the purpose of passive
control of mechanical systems [6]. It has revived interest
in passive network realisations. The inerter is a mechanical
two-terminal element with the property that the applied force
at the terminals is proportional to the relative acceleration
across the terminals and is the exact analogue of the ca-
pacitor for network synthesis. Applications of the method to
vehicle suspension [7], [8], control of motorcycle steering
instabilities [9], [10] and vibration absorption [6] have been
This work was supported by the Engineering and Physical Sciences
Research Council grant number EP/G066477/1.
Both authors are with the Department of Engineering, University of Cam-
bridge, Cambridge CB2 1PZ, U.K.; zj219@cam.ac.uk, mcs@eng.cam.ac.uk
identied. The inerter has been successfully deployed in
Formula One racing since 2005 [11].
The present paper is a successor to [3], [4] which for-
malised the concept of a regular positive real function,
introduced the terminology of a network quartet, and gave
a complete reworking and characterisation of the class of
transformerless networks containing two reactive elements
and ve-, six-element series-parallel networks with (at most)
three reactive elements. This paper is also a continuation of
the work of Vasiliu [12], [13]. The present paper consid-
ers six-element series-parallel networks with four reactive
elements. We describe a classication procedure to nd an
efcient subset of such networks which may realise any non-
regular biquadratic that can be synthesised by this class of
network. It is veried that there are ve network quartets
which may realise non-regular positive-real biquadratic func-
tions.
II. REGULAR POSITIVE-REAL FUNCTIONS
In this section we recall the concept of regularity and its
properties given in [3].
Denition: A positive-real function Z(s) is dened to be
regular if the smallest value of Re (Z(j)) or Re

Z
1
(j)

occurs at = 0 or = .
Lemma 1: Let Z(s) be a regular positive-real function.
Then Z (s), Z (s), Z

s
1

, Z
1
(s) are all regular, where
, > 0.
Lemma 2: Let Z(s) be a regular positive-real function.
Then Z(s) +R and Z
1
(s) +R
1
are both regular, where
R is nonnegative.
Lemmas 1 and 2 imply that if a network can only realise
regular immittances, then so will the other networks in the
quartet and also the networks obtained by adding a resistor
in series or in parallel with the original one.
The next lemma follows from the fact that the impedance
Z (s) or admittance Y (s) of any network that has all reactive
elements of the same kind has Re (Z(j)) and Re (Y (j))
monotonic ([14, Chapter 2.2]).
Lemma 3: Any network that has all reactive elements of
the same kind can only realise regular immittances.
For a general connected graph a cut set is a collection of
branches whose removal results in a disconnected graph [15].
For a two-terminal electrical network, a cut set is dened to
be a set of branches whose removal places the two terminals
1 and 1

in disconnected parts.
Lemma 4: i) Any network that has a path between the
two external terminals 1 and 1

consisting only of
2010 American Control Conference
Marriott Waterfront, Baltimore, MD, USA
June 30-July 02, 2010
WeB18.2
978-1-4244-7427-1/10/$26.00 2010 AACC 1416
one type of reactive element can only realise regular
immittances.
ii) Any network that has a cut set consisting only of
one type of reactive element can only realise regular
immittances.
We now focus attention on biquadratics
Z(s) =
As
2
+Bs +C
Ds
2
+Es +F
, (1)
where A, B, C, D, E, F 0. It is well known [16], [17],
[18] that Z (s) is positive real if and only if
= BE

AF

CD

2
0. (2)
We will make use of the resultant of the numerator and
denominator in (1) which is given by
K = (AF CD)
2
(AE BD)(BF CE). (3)
Lemma 5: A positive-real biquadratic impedance (1) is
regular if and only if the conditions of at least one of the
following four cases are satised:
Case 1. AF CD 0 and

1
= E (BF CE) F (AF CD) 0,
Case 2. AF CD 0 and

2
= B(AE BD) A(AF CD) 0,
Case 3. AF CD 0 and

3
= D(AF CD) E (AE BD) 0,
Case 4. AF CD 0 and

4
= C (AF CD) B(BF CE) 0.
Lemma 6: A positive-real biquadratic impedance (1) with
Z (0) = Z () = 0, (which implies AF CD = 0) is
regular.
Lemma 7: A positive-real biquadratic impedance (1) with
the resultant K 0 has Re (Z(j)) and Re (Y (j))
monotonic, hence is regular.
Lemma 8: A positive-real biquadratic impedance (1) with
any of the parameters A, B, C, D, E and F equal to zero
is regular. Moreover, any such impedance can be realised by
a series-parallel network with at most two reactive elements
and two resistive elements.
III. A CANONICAL FORM FOR BIQUADRATICS AND
NETWORK QUARTETS
The classication of networks is facilitated by the follow-
ing transformations on the impedance Z (s):
1) Multiplication by a constant ,
2) Frequency scaling: s s,
3) Frequency inversion: s s
1
,
4) Impedance inversion: Z Z
1
.
In network realisations, the rst two transformations corre-
spond to element scaling, the third to replacing inductors
with capacitors of reciprocal values (and vice versa), and
the fourth to taking the network dual. The third and fourth
transformations together allow networks to be arranged into
groups of four, which we call network quartets. Such families
have appeared in [19] with the terminology Untergruppe.
Fig. 1. Transformations relating members of a network quartet and
corresponding transformation in the canonical form (4).
It should be noted that a network quartet may sometimes
reduce to two or even one distinct network(s).
For the study of networks that may realise biquadratics,
based on Lemma 8, we can restrict attention to the case
that A, B, C, D, E, F > 0 in (1). Using the rst two
transformations, Z (s) can be reduced to a canonical form
Z
c
(s) =
s
2
+ 2U

Ws +W
s
2
+

2V/

s + 1/W
, (U, V, W > 0) . (4)
This canonical form was rst considered in [20].
Next we observe that frequency inversion

s s
1

cor-
responds to the transformation W W
1
in the canonical
form and impedance inversion (duality) corresponds to the
transformations: U V , W W
1
. This is illustrated in
the network quartet shown in Fig. 1.
It follows from (2) that Z
c
(s) is positive real if and only
if

c
= 4UV + 2 (
1
W
+W) 0.
The resultant of the numerator and denominator of Z
c
(s) is
K
c
= 4U
2
+ 4V
2
4UV (
1
W
+W) + (
1
W
W)
2
.
For any rational function (U, V, W) we introduce the
notation

(U, V, W) =

U, V, W
1

and

(U, V, W) =
(V, U, W). We observe that

c
=

c
=
c
and K

c
=
K

c
= K
c
. Let

c
= 4UV 4V
2
W (
1
W
W).
We can now restate Lemmas 57 for the canonical form (4):
Lemma 9: Let Z
c
(s) be a positive-real biquadratic de-
ned in (4). Then:
1) Z
c
(s) is regular if and only if at least one of the
following four conditions is satised:
Case 1. W 1 and
c
0,
Case 2. W 1 and

c
0,
Case 3. W 1 and

c
0,
Case 4. W 1 and

c
0.
2) When W = 1, Z
c
(s) is regular.
3) When K
c
0, Z
c
(s) is regular.
Lemma 9-1) denes the regular region of a biquadratic in
the (U, V )-plane, as shown in Fig. 2. In [3] the following
theorem was shown.
1417
c < 0
Kc < 0

c
> 0
c > 0
c = 0
c = 0

c
= 0
Kc = 0
U
V
W=0.5
c < 0
Kc < 0

c
> 0
c > 0
c = 0
c = 0

c
= 0
Kc = 0
0 1 2
0
1
2
Fig. 2. The regular impedances of (4) with W (0, 1).
aZ1
bZ1
cZ2
b

Z1
a

Z1
c

Z2
(a) (b)
Fig. 3. Two equivalent networks related by the transformations in Lemma
10.
Theorem 1: A biquadratic impedance (1) can be realised
by series-parallel ve-element networks with two reactive
elements if and only if it is regular.
IV. SIX-ELEMENT FOUR-REACTIVE ELEMENT
SYNTHESIS OF THE BIQUADRATIC
A. Elimination of the Networks that Can Only Realise Reg-
ular Biquadratics
In [12], Vasiliu claims there are six seires-parallel net-
works with four reactive elements which can realise classes
of biquadratic immittances that otherwise would require a
full Bott-Dufn synthesis. Vasiliu [13] later claims that there
are 12 more such networks. In this section, we will provide a
complete elimination process of the six-element four-reactive
element networks which can only realise regular biquadratics
(Theorem 2). It will be shown that only the eighteen networks
identied by Vasiliu remain, which can be grouped into ve
network quartets of Fig. 6.
Lemma 10: ([21]) For arbitrary impedances Z
1
(s), Z
2
(s)
and positive constants a, b, c, the networks of Figs. 3(a)
and (b) are equivalent under the transformations: a

=
Passive
Network
Passive
Network
Lossless
Network
Lossless
Network
(a) (b)
Fig. 4. Two network structures which can only realise regular biquadratics
based on Lemma 11.
a (a +b) /b, b

= a + b, c

= c ((a +b) /b)

2
[a =
a

/ (a

+b

), b = b
2
/ (a

+b

), c = c

(b

/ (a

+b

))
2
].
Lemma 11: Networks with the structures of Fig. 4(a) or
(b) can only realise regular biquadratic impedances (1).
Proof: Consider Fig. 4(a). Let us denote the impedance
function for the passive network and lossless network as Z
1
and Z
2
respectively, and Z = Z
1
+ Z
2
. Based on [22], Z
2
has all its poles on the imaginary axis or at . These must
be poles of Z also, since the residues of such poles are real
and positive for positive-real Z
1
and Z
2
[23]. Hence if Z(s)
reduces to a biquadratic, it must be regular by Lemma 8. A
similar argument holds for Fig. 4(b).
Theorem 2: A non-regular biquadratic immittance can be
realised by a series-parallel network with four reactive and
two resistive elements if and only if it is realisable by some
network in the ve quartets of Fig. 6.
Proof: If no distinction is made among the elements,
there are 66 distinct two-terminal series-parallel structures
with six elements [24]. These structures may be divided into
two classes, any structure in one class having its dual in the
other. Based on Lemma 1, the analysis may be performed
upon only one class. In Fig. 5, all the series-parallel six-
element structures in one class are presented according to a
numbering of Vasiliu [12]. Based on Lemma 3, we only need
to investigate the networks containing both kinds of reactive
elements. Furthermore, it is sufcient to consider only the
assignments of elements with two capacitors and two induc-
tors or three capacitors and one inductor by Lemma 1.
It is apparent that structures 16 must reduce to a net-
work with fewer than six elements. It can be checked that
structures 724 will either reduce to a network with at most
ve elements or can only realise regular immittances by
Lemmas 2 and 4. The only series-parallel networks with ve
elements which can realise non-regular immittances belong
to the network quartet of Fig. 3 in [4] and are special cases
of quartets II-V in Fig. 6.
For structure 25, if it does not reduce to a ve-element
network the upper branch contains all three kinds of elements
and the lower branch has a resistor in series. So if the
network has three capacitors, it must reduce to a ve element
network. If the network has two capacitors, the conditions of
Lemma 6 apply. Hence, it can only realise regular biquadratic
immittances.
1418
For structure 26, if it does not reduce to a ve-element
network the three parallel elements in the lower branch must
be different. By Lemma 4, the series element in the lower
branch must be a resistor. So the upper branch is lossless.
Based on Lemma 11, it can only realise regular biquadratic
immittances.
For structure 27, since there are only two resistors, at least
one of the three branches is lossless. Based on Lemma 11,
it can only realise regular biquadratics.
For structure 28, based on Lemma 11 the lower branch
cannot be lossless. When the network contains three capac-
itors, it will either reduce to a ve-element network or can
only realise regular biquadratic immittances by Lemma 4.
When the network contains two capacitors, we only need
to consider the case of a resistor and capacitor in the lower
branch (the case of a resistor and inductor in the lower branch
can be covered by the s s
1
transformation). The element
in series in the top branch cannot be a capacitor otherwise
Lemma 4 applies. If the series element is an inductor, based
on Lemma 4 it must take the form of either Fig. 6-II(a) or
III(a). If the series element is a resistor, for it not reduce
to a ve-element network, it takes the form of Fig. 6-I(c)
after applying the transformation of Lemma 10 to part of
the upper branch.
For structure 29 the upper branch must be a resistor by
Lemma 4. The remaining part contains only one resistor, for
which the immittance function has a zero or pole at s = 0
or , or the conditions of Lemma 6 apply. Then based on
Lemma 2 the network can only realise regular biquadratic
immittances.
For structure 30, if it does not reduce to a network with
fewer than six elements and also does not have a path of one
kind of reactive element, the two parallel-pairs in series in the
upper branch must contain one inductor and one capacitor
respectively, and each contains a resistor. So the lower branch
is lossless. Based on Lemma 11, it can only realise regular
biquadratic immittances.
For structure 31 the element in series must be a resistor
by Lemma 4. The argument is now the same as for structure
29.
For structure 32, based on Lemma 11, the lower branch
cannot be lossless. The network cannot have three capacitors,
otherwise either it reduces to a ve-element network or
Lemma 4 applies. When the network has two capacitors,
we only need to consider the case where the lower branch
consists of a resistor and inductor (the case where the lower
branch contains a resistor and capacitor can be covered by the
s s
1
transformation). Based on Lemma 4 the two series
elements in the upper branch are a resistor and a capacitor
and the network takes the form of Fig. 6-I(a).
For structure 33, based on Lemma 11, we need only
consider the case where both the upper and lower branches
contain a resistor. Each branch must contain both kinds of
reactive element, otherwise either the network reduces to
fewer than six elements or Lemma 4 applies. If the series
elements in both branches are resistors, the conditions of
Lemma 6 applies. If the series elements in both branches are
Fig. 5. One-Half of the Six-Element Series-Parallel Structures.
reactive elements of the same kind, Lemma 4 applies. If the
series elements in the two branches are reactive elements of
different kinds, the network takes the form of Fig. 6-IV(a). If
the series elements in the two branches are a resistor and an
inductor, the network takes the form of Fig. 6-V(a). The case
where the series elements in the two branches are a resistor
and a capacitor is covered by the s s
1
transformation of
the previous case.
B. The Non-Regular Realisable Region of the Network Quar-
tet Fig. 6- II
The realisability conditions given by [13] for the network
shown in Fig. 6-II(c) are not correct. In this section, we state
the corrected necessary and sufcient realisablity conditions
for this network (Theorem 3). We also give the range of
parameter values corresponding to non-regular biquadratics
in the canonical form which this network quartet can realise
(Theorem 4 and Table I).
Theorem 3: The biquadratic impedance (1) with
A, B, C, D, E, F > 0 can be realised by the network
of Fig. 6-II(c) with R
1
, R
2
, L
1
, L
2
, C
1
and C
2
positive
and nite, if and only if

4
< 0, (5)
CD
4
+B

EFB
2
BAF
2
BE
2
C +EDC
2

= 0.
(6)
1419
R
1
C
1
C
2
R
2
L
1
L
2
dual
dual
ss
1
ss
1
(a)
(b) (c)
(d)
I
V
R
1
C
1
C
2
R
2
L
1
L
2 dual
dual
ss
1
ss
1
(a)
(b) (c)
(d)
II
V
R
1
C
1
C
2
R
2
L
1
L
2
dual
dual
ss
1
ss
1
(a)
(b) (c)
(d)
III
V
R
1
C
1
C
2
R
2
L
1
L
2
dual
(a) (b)
IV
R
1
C
1
C
2
R
2
L
1
L
2
dual
dual
ss
1
ss
1
(a)
(b)
(c)
(d)
V
V
Fig. 6. The ve network quartets of the series-parallel four-reactive six-
element networks that can realise non-regular biquadratics.
TABLE I
THE REALISABILITY CONDITIONS OF THE NETWORK QUARTET OF
FIG. 6-II FOR THE BIQUADRATIC IMPEDANCE (4).
Networks Nec. and Suff. con-
ditions for (4) to be
realisable
Range of W for which
non-regular region is non-
empty
Fig. 6-II(a)
10
= 0,

c
< 0 W (0, 1)
Fig. 6-II(b)

10
= 0, c < 0 W (0, 1)
Fig. 6-II(c)

10
= 0,

c
< 0 W (1, )
Fig. 6-II(d)

10
= 0,

c
< 0 W (1, )
where
4
is dened in Lemma 5.
If conditions (56) are satised, then
R
1
=
A
D
, (7)
R
2
=
C
F
, (8)
L
1
=
AB

C +aB +a
2
A

D(C +aB)
2
, (9)
L
2
=
C
aF
, (10)
C
1
=
D(C +aB)
(C +aB +a
2
A) B
, (11)
C
2
=
D(C +aB)
aAC
, (12)
where
a = C

F/
4
. (13)
Proof: See [25].
It can be written down directly from Theorem 3 that the
realisability conditions of the network Fig. 6-II(a) for the
biquadratic canonical form (4) are

c
< 0 and
10
= 0,
where

10
= 16 W
2
U
2
V
2
8 UW

1 + 2 U
2
W
2

V
+ 1 W
2
+ 4 W
2
U
2
+ 4 W
4
U
2
.
The curve
10
= 0 has two branches which satisfy

c
> 0
and

c
< 0. (We can check that the resultant of
10
and

c
in V is 16U
2
W
4

1 W
2

, which does not vanish for

W = 1.) Only the latter curve is shown in Fig. 7.
Theorem 4: A biquadratic impedance (4) is realisable by
the network of Fig. 6-II(a) (Fig. 6-II(b)) if and only if

c
< 0
and
10
= 0 (
c
< 0,

10
= 0). (See dashed curves of Fig. 7.)
In both cases, non-regular impedances (4) are realisable if
and only if W (0, 1).
Proof: See [25].
Based on Theorem 4, the conditions for a non-regular
biquadratic impedance (4) to be realisable by the network
quartet of Fig. 6-II are summarised in Table I. The necessary
and sufcient realisability conditions and the non-regular
realisable regions of the biquadratics for the network quartets
shown in Figs. 6-I, III, IV and V are given in [25].
1420
U
V
W=0.6
c = 0

c
= 0

c
< 0,
10 = 0
c < 0,

10
= 0
P1
P2
0 1 2
0
1
2
(a)
U
V
W=2

c
= 0

c
= 0

c
= 0

c
< 0,
10 = 0
c < 0,

10
= 0
0 1 2
0
1
2
(b)
Fig. 7. The impedances (4) that can be realised by the network shown in
Fig. 6-II(a) and II(b) with (a) W (0, 1), (b) W (1, ).
V. CONCLUSION
This paper has studied six-element series-parallel networks
with four reactive elements and presented a complete clas-
sication process using the recently introduced concept of
regular positive-real functions. Five network quartets have
been identied which can realise non-regular biquadratic
functions.
REFERENCES
[1] O. Brune, Synthesis of a nite two-terminal network whose driving-
point impedance is a prescribed function of frequency, J. Math. Phys.,
vol. 10, pp. 191236, 1931.
[2] R. Bott and R. J. Dufn, Impedance synthesis without use of
transformers, J. Appl. Phys., vol. 20, p. 816, 1949.
[3] J. Z. Jiang and M. C. Smith, Regular positive-real functions and
passive networks comprising two reactive elements. European
Control Conference (ECC), August 2009, pp. 219224.
[4] , Synthesis of positive-real functions with low-complexity series-
parallel networks. Conference on Decision and Control (CDC), Dec.
2009, pp. 70867091.
[5] , Regular positive-real functions and the classication of trans-
formerless series-parallel networks, in Perspectives in Mathematical
System Theory, Control, and Signal Processing. Springer Verlag, to
appear.
[6] M. C. Smith, Synthesis of mechanical networks: the inerter, IEEE
Trans. Automatic Control, vol. 47, no. 10, pp. 16481662, 2002.
[7] C. Papageorgiou and M. C. Smith, Positive real synthesis using
matrix inequalities for mechanical networks: application to vehicle
suspension, IEEE Trans. on Contr. Syst. Tech., vol. 14, pp. 423435,
2006.
[8] M. C. Smith and F. C. Wang, Performance benets in passive vehicle
suspensions employing inerters, Vehicle System Dynamics, vol. 42,
pp. 235237, 2004.
[9] S. Evangelou, D. J. N. Limebeer, R. S. Sharp, and M. C. Smith,
Control of motorcycle steering instabilities - passive mechanical com-
pensators incorporating inerters, IEEE Control Systems Magazine, pp.
7888, October 2006.
[10] , Mechanical steering compensation for high-performance mo-
torcycles, Transactions of ASME, J. of Applied Mechanics, vol. 74,
pp. 332346, 2007.
[11] Cambridge University Engineering Department news item, 19 August
2008. [Online]. Available: http://www.eng.cam.ac.uk/news/stories/
2008/McLaren/
[12] C. G. Vasiliu, Series-parallel six-element synthesis of the biquadratic
impedances, IEEE Trans. on Circuit Theory, pp. 115121, 1970.
[13] , Four-reactive six-element biquadratic structures, IEEE Trans.
on Circuit Theory, pp. 530531, September 1972.
[14] E. A. Guillemin, Synthesis of Passive Networks. John Wiley & Sons,
1957.
[15] F. Harary, Graph Theory. Addison-Wesley, Reading, 1969.
[16] R. M. Foster and E. L. Ladenheim, A class of biquadratic
impedances, IEEE Trans. on Circuit Theory, pp. 262265, 1963.
[17] M. E. V. Valkenburg, Introduction to Modern Network Synthesis. John
Wiley & Sons, 1960.
[18] M. Z. Q. Chen and M. C. Smith, A note on tests for positive-real
functions, IEEE Transactions on Automatic Control, vol. 54, no. 2,
pp. 390393, 2009.
[19] G. Dittmer, Zur realisierung von RLC-Br uckenzweipolen mit zwei
Reaktanzen und mehr als drei Widerst anden (on the realisation of
RLC two-terminal bridge networks with two reactive and more than
three resistive elements), Nachrichtentechnische Zeitschrift, vol. 23,
pp. 225230, 1970.
[20] M. Reichert, Die kanonisch und ubertragerfrei realisierbaren
Zweipolfunktionen zweiten Grades (transformerless and canonic reali-
sation of biquadratic immittance functions), Arch. Elek.

Ubertragung,
vol. 23, pp. 201208, Apr. 1969.
[21] P. M. Lin, A theorem on equivalent one-port networks, IEEE Trans.
on Circuit Theory, vol. CT-12, pp. 619621, Dec. 1965.
[22] R. Foster, A reactance theorem, Bell Syst. Techn. Journ., vol. 3, p.
259, 1924.
[23] B. D. O. Anderson and S. Vongpanitlerd, Network Analysis and
Synthesis. Upper Saddle River, NJ: Prentice-Hall, 1973.
[24] J. Riordan and C. E. Shannon, The number of two-terminal series-
parallel networks, J. Math. Phys., vol. 21, 1942.
[25] J. Z. Jiang, Passive electrical and mechanical network synthesis,
Ph.D. dissertation, University of Cambridge, March 2010.
1421