International Journal of Control

ISSN: 0020-7179 (Print) 1366-5820 (Online) Journal homepage: http://www.tandfonline.com/loi/tcon20

Conditions for triangular decoupling control

Hung T. Nguyen & Steven W. Su

To cite this article: Hung T. Nguyen & Steven W. Su (2009) Conditions for triangular decoupling
control, International Journal of Control, 82:9, 1575-1581, DOI: 10.1080/00207170802566614

To link to this article: http://dx.doi.org/10.1080/00207170802566614

Published online: 15 Jul 2009.

Submit your article to this journal

Article views: 90

View related articles

Citing articles: 4 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tcon20

Download by: [National Institute of Technology - Trichy] Date: 26 April 2017, At: 01:56
International Journal of Control
Vol. 82, No. 9, September 2009, 15751581

Conditions for triangular decoupling control

Hung T. Nguyen* and Steven W. Su
Faculty of Engineering and Information Technology, University of Technology, Sydney (UTS), Australia
(Received 15 August 2008; final version received 20 October 2008)

The main purpose of this article is to explore the relationship of two existing conditions for the triangular
decoupling problem. The first one is the triangular-diagonal-dominance condition proposed by Hung and
Anderson. The second one is the stable coprime factorisation-described condition proposed by Gomez and
Goodwin, which has been proven as a necessary and sufficient condition for the triangular decoupling problem.
This article proves that the two conditions are actually equivalent. It also provides easy-to-use criteria for
assessment of the solvability of the triangular decoupling problem.
Keywords: linear multivariable systems; triangular decoupling; triangular-diagonal-dominance; stable coprime
factorisation

1. Introduction (Anderson and Hung 1978; Hung and Anderson 1979)

The problem of decoupling linear multivariable sys-
tems has still been receiving great attention after
several decades (Rosenbrock 1969; Anderson and
Hung 1978; Hung and Anderson 1979; Koumboulis
1996; Koumboulis and Skarpetis 2000; Gomez and
Goodwin 2000; Wang and Yang 2002; Chughtai and
Wang 2007). It has been established that the diagonal
decoupling problem is solvable for systems without
coincident unstable poles and non-minimum phase
zeros (Giindeq 1990; Vardulakis 1991; Lin and Hsieh
1991). In terms of the necessary and sufficient
conditions for the solvability of the diagonal decou-
pling problem, there are many approaches that provide
solutions (Wang 1992; Lin 1997; Gomez and Goodwin
2000; Wang and Yang 2002). However, the main
limitations of diagonal decoupling relate to the
possible increase in the number of non-minimum
phase zeros (and/or unstable open-loop poles) and
the resulting designs are often quite complex. As
a consequence, various approaches have been devel-
oped recently to design simpler compensators to
achieve diagonal dominance (Chughtai and Munro
2004; Chughtai and Wang 2007).
On the other hand, there has also been much interest
in triangular decoupling which gives a simpler compen-
sator and allows more design freedom. Earlier works
on triangular decoupling set the multivariable design
problem on the framework of finding canonical forms
of matrices over Euclidean domains. In this framework,
Hung (an author of this article) and Anderson
proposed efficient triangularisation approaches for
a class of multivariable systems which satisfies the
triangular-diagonal-dominance (TDD) condition.
Recently, several new efficient methods, such as the
stable coprime factorisation-based approach (Gomez
and Goodwin 2000) and the static feedback approach
(Koumboulis 1996; Koumboulis and Skarpetis 2000)
have been developed to provide solutions to the
triangular decoupling problem. In particular, Gomez
and Goodwin (2000) systematically examined the
problem of decoupling design based on an algebraic
approach and the use of coprime factorisation over the
ring of proper and stable rational functions (Vidyasagar
1985) and proposed a necessary and sufficient condition
for the solvability of the triangular decoupling problem
(see Problem 1.1 below).
The main purpose of this article is to investigate
the relationship between the TDD condition proposed
in Hung and Anderson (1979) and the stable coprime
factorisation-described condition in Theorem 1 of
Gomez and Goodwin (2000). Gomez and Goodwin
proposed the following definition of the triangular
decoupling problem.
Problem 1.1: Consider a controlled process G(s) as
shown in Figure 1, which is described by a m  m,
full-rank, proper rational matrix. The triangular
decoupling problem is the design of a controller
C(s) for the plant G(s) to make the closed-loop
complementary sensitivity function T(s) G(s)C(s)
(I G(s)C(s))1 and the sensitivity function

*Corresponding author. Email: Hung.Nguyen@uts.edu.au

ISSN 00207179 print/ISSN 13665820 online

 2009 Taylor & Francis
DOI: 10.1080/00207170802566614
http://www.informaworld.com
1576 H.T. Nguyen and S.W. Su
d . R(s): the field of fractions associated with R[s]
r u + y
C(s) G(s)
+
n
Figure 1. A closed-loop system.
S(s) (I G(s)C(s))1 lower triangular, and to guar-
antee the internal stability of the closed loop.
Hung and Anderson (1979) showed that if G(s) can
be precompensated to TDD (see Definition 1), the
triangular decoupling problem is solvable. Hung and
Anderson (1979) also provided systematic approaches
to reduce the multivariable design problem to scalar
design problems. In Gomez and Goodwin (2000),
a necessary and sufficient condition for Problem 1
was provided based on stable coprime factorisation
(Gomez and Goodwin 2000, Theorem 1).
This article will prove the equivalence of the stable
coprime factorisation condition and TDD condition
so that some easy-to-use rules developed in Hung and
Anderson (1979) can be applied to evaluate the
solvability of the triangular decoupling problem, even
without necessarily performing triangularisation. This
article also provides a matrix fractional descriptions
(MFDs) method for readers to assess the solvability of
the triangular decoupling problem as an alternative to
the stable coprime factorisation approach.
This article is organised as follows. All main
results are presented in x 2. The condition provided
by Gomez and Goodwin (2000) is given in
Theorem 2.2. Theorem 2.4 proves the TDD condition
is equivalent to the MFDs condition. The main result
is given in Theorem 2.6, which proves the TDD
condition is actually equivalent to the condition given
by Gomez and Goodwin. Lemma 2.7 proposes
a sufficient condition for both lower and upper
triangular decoupling. In x 3, two illustrative examples
are presented to explain and justify the results.
2. TDD and stable coprime factorisation-based
condition
In order to clarify our discussion, the following
symbols and notations are adopted from literature
(Gomez and Goodwin 2000; Vidyasagar 1985; Hung
and Anderson 1979):
. R[s]: the set of polynomials in the indetermi-
nate s with coefficients in the field R of real
numbers.
(or the set of all possible transfer functions of
scalar, lumped, linear time-invariant systems).
. M(R): the set of matrices with elements
in R[s].
. M(F ): the set of matrices with elements
in R(s).
. S: the subset of R(s) consisting of all rational
functions that are bounded at infinity, and
whose poles all have negative real parts.
. U: the set of units of S, i.e. functions in S
whose reciprocal also belongs to S.
. M(S): the set of matrices with elements in S.
. U M(S): the set of unimodular matrices
of M(S).
. C: the closed right half-plane.
. Ce: the extended right half-plane, i.e. C
together with the point at infinity.
. X(s): the sum of those strictly proper terms in
a partial fraction expansion of X(s) 2 M(F )
with poles in C.
. X(s): the sum of those strictly proper terms in
a partial fraction expansion of X(s) 2 M(F )
with poles in the (open) left half plane.
. jX(s)j: the determination of X(s). That is
jX(s)j det(X(s)).
. [X(s)]: the McMillan degree of a proper
transfer function matrix X(s) 2 M(F ). It is
not hard to prove that [X(s)] [X] [X].
~
. Xs: the diagonal part of the square matrix
~ diagx11 s, . . . , xmm s.
X(s). That is Xs
Definition 2.1 (Hung and Anderson 1979): For
a proper rational matrix G(s) 2 R(s)mm, assume
there exists a unimodular matrix U(s) 2 U M(S),
such that
Ps GsUs,
where P(s) 2 R(s)mm is a lower triangular matrix.
If
P s P~ s 1
then, we say P(s) is triangular-diagonal-dominant
(TDD) and G(s) can be precompensated to TDD.
The following notations are cited from Gomez and
Goodwin (2000) to highlight facts about stable
coprime factorisation:
Assume a stable coprime factorisation of a full-
rank, proper rational matrix G(s) 2 R(s)mm can be
expressed in the following from:
Gs Np  D1
p 2
 International Journal of Control 1577

where Np 2 S mm and Dp 2 S mm are stable transfer Theorem 2.4: For a proper rational matrix
function matrices and right coprime. Suppose that G(s) 2 R(s)mm, assume there exists a unimodular
matrix U 2 U M(S) such that
Np UNp LNp
Ps GsUs,
and
where P(s) 2 R(s)mm is a lower triangular matrix.
D p Dp UNp
Suppose P(s) has the following MFDs form:
where UNp 2 U M(S) (an unimodular matrix of M(S)),
and LNp is lower triangular. Then, Ps NL sD1
L s, 5

Gs LNp U1 1
Np Dp LNp Dp UNp
Further assume
1
LNp D 1 where NL(s) 2 R[s]mm and DL(s) 2 R[s]mm are lower
p :
triangular polynomial matrices and right coprime. Then,
the following two statements are equivalent:

U Dp D p L Dp (i) Each diagonal entry of NL(s) and DL(s) has no

common non-minimal phase zeros.
where U Dp 2 U M(S), and L Dp is lower triangular. Then, (ii) P(s) is TDD (i.e. G(s) can be precompensated
to TDD).
Gs LNp U 1  1 LNp L 1 U Dp PsU Dp ,
Dp LDp Dp

3 Proof: For a square polynomial matrix D(s) 2 M(R)

we denote
where Ps GsU 1Dp is a lower triangular transfer
function matrix.
N R Ds
As claimed in Gomez and Goodwin (2000), LNp and
L Dp are invariants up to postmultiplication by a lower as the number of non-minimal phase zeros (including
triangular unimodular matrix of M(S) and are multiplicity) of jD(s)j.
independent of particular factorisation of G(s). Consider that NL(s) and DL(s) in Equation (5) are
Therefore, the non-minimum phase zeros of right coprime; we have
L Dp 2 MS (including multiplicities) and the zeros at
infinity of each diagonal entry are invariant. P s N R DL s: 6
Gomez and Goodwin proposed a necessary and
From Equation (5), we can get
sufficient condition for Problem 1.1, based on the
stable coprime factorised form (3). ~ N~ L sD~ 1 s:
Ps L
Theorem 2.2 (Gomez and Goodwin 2000): For Now, if condition (i) is satisfied, i.e. each diagonal
a full-rank, proper rational matrix G(s) 2 R(s)mm, the entry of NL(s) and DL(s) has no common non-minimal
following two statements are equivalent:
phase zeros, then NL(s) and D~ L s are coprime over the
(i) the triangular decoupling problem (Problem 1.1) closed right half-plane C. Then, we have
is solvable for G(s);
(ii) each diagonal entry of LNp is coprime with the P~ s N R D~ L s: 7
corresponding diagonal entry of L Dp (LNp and Consider that DL(s) is lower triangular matrix; we
L Dp are as defined in Equation (3)). can see that
Proof: See Theorem 1 of Gomez and Goodwin
(2000). jD~ L sj jDL sj:

Lemma 2.3 (Hung and Anderson 1979): For a lower Thus,

triangular proper transfer function matrix
P(s) 2 R(s)mm, there exist right matrix fraction descrip- N R D~ L s N R DL s: 8
tions (MFDs): Based on Equations (6), (7) and (8), we conclude
Ps NL sD1 4 that
L s,

where NL(s) and DL(s) are lower triangular polynomial P~ s P s:
matrices.
Thus P(s) is TDD.
Proof: See Lemma 2 of Hung and Anderson On the other hand if (i) is incorrect, i.e. some
(1979). diagonal entries of NL(s) and DL(s) have common
1578 H.T. Nguyen and S.W. Su
non-minimal phase zeros, then NL(s) and D~ L s are not
coprime over the closed right half-plane C. Thus,
P~ s 5 N R D~ L s N R DL s: 9
Based on Equations (6) and (9), we have
P~ s 5 P s:
Then, (ii) is also incorrect.
Lemma 2.5: Consider a lower triangular proper trans-
fer function matrix P(s) 2 R(s)mm. Assume P(s) can be
expressed in a stable coprime factorised form:
Ps LNp L 1
Dp ,
where LNp and L Dp are stable lower triangular matrices
and coprime each other.
Then, there is no infinity zero in each diagonal entry
of L Dp .
Proof: See Appendix A.
Now, we prove the equivalence of the stable
coprime described condition and TDD condition.
Theorem 2.6: (Main result) For a proper rational
matrix G(s) 2 R(s)mm, assume there exists a
unimodular matrix U(s) 2 U M(S) such that
Ps GsUs,
where P(s) 2 R(s)mm is a lower triangular matrix.
Suppose P(s) can be expressed in a stable coprime
factorised form:
Ps LNp L 1
Dp ,
where LNp 2 R(s)mm and L Dp 2 Rsmm are stable
lower triangular matrices and coprime each other
(Equation (3)). Then, the following two statements are
equivalent:
(i) Each diagonal entry of LNp is coprime with the
corresponding diagonal entry of L Dp .
(ii) P(s) is TDD (i.e. G(s) can be precompensated
to TDD).
Proof: As
 LNp L 1 ,
Ps GsUs 12
Dp
we have [P(s)] equals to the number of non-minimal
phase zeros of L Dp .
As L Dp is a lower triangular matrix, all its zeros are
contained in its diagonal entries. If we define
N F fas=bsg 13
as the number of remaining non-minimal phase zeros
(including multiplicity) of a(s) 2 R(s) after possible
cancellation with the zeros (including multiplicity) of
b(s) 2 R(s), where a(s) and b(s) are two scalar transfer
functions, then we have P~ s m i1 N F flNp sii =
lDp sii g, where lNp(s)ii and lDp sii are the i-th diagonal
entries of LNp and L Dp respectively.
Now, if P(s) is TDD, then we have
P s P~ s:
That is, the number of non-minimal phase zeros of L Dp
equals
P~ s m

i1 N F flNp sii =lDp sii g:
10 This means that there are no common non-minimal
phase zeros between each diagonal entry of LNp and
L Dp . Furthermore, according to Lemma 2.5, lDp sii has
no zeros at infinity. Thus, we have each diagonal entry
of LNp as coprime with the corresponding diagonal
entry of L Dp .
On the other hand, if each diagonal entry of LNp
is coprime with the corresponding diagonal entry of
L Dp , then there are no common non-minimal phase
zeros between each diagonal entry of LNp and L Dp .
Thus, the number of non-minimal phase zeros of L Dp
equals P~ s m

i1 N F flNp sii = lDp sii g. This ensures
that P(s) is TDD.
Remark 1 : Readers may find it helpful to compare
the statement and proof of Theorems 2.4 and 2.6.
The two theorems are quite similar. The main differ-
ence is that Theorem 2.4 (Equation (5)) is stated in
the Euclidean domain R[s], while Theorem 2.6
(Equation (11)) is in another Euclidean domain R(s).
11
Theorems 2.4 and 2.6 imply that the TDD
condition, stable coprime factorisation-based condi-
tion, and MFDs-based condition are all equivalent.
They are all necessary and sufficient conditions for
triangular decoupling problem. In Hung and Anderson
(1979), some easy-to-use criteria were developed to
verify TDD properties. These criteria can also be
applied to determine the solvability of the decoupling
problem. For example, the following lemma as well as
its proof are mainly from Lemma 6 at page 457 of
Hung and Anderson (1979). In some cases, this lemma
can determine the solvability of both lower and upper
triangular decoupling problems without performing
a triangularisation procedure (see Example 3.1).
Lemma 2.7: Consider an m  m, full-rank, proper
rational matrix G(s). If
jGsj  G s, 14
then both the lower triangular decoupling problem
and upper triangular decoupling problem are solvable
for G(s).
 International Journal of Control 1579

Proof: Assume P(s) G(s)U(s) where U(s) 2 U M(S) Similarly, the upper triangular form of G(s) can be
is a unimodular and P(s) is in either upper or lower calculated:
triangular form. Then it is easy to see
Pup s GsUup s
jGsj  jPsj , 2 1 1 3" #
and 6s 1 s  27 1 0
64
7
5
G s P s: 1 1 1 1
s1 s1
Considering Equation (14), we have 2 3
3 1
jPsj  P s: 6 s  2s 1 s  2 7
6 7
6 7:
This proves P(s) is TDD. 4 1 5
0
s1

3. Example The MFDs form and stable coprime factorisation

form of Pup(s) are shown as follows:
The following two examples are from Hung and
Anderson (1979) and Gomez and Goodwin (2000), 2 3
3 1
respectively. 6 s  2s 1 s  2 7
6 7
Example 3.1: Consider the following transfer func- Pup s 6 7
4 1 5
tion matrix (Hung and Anderson 1979): 0
s1
2 3
1 1 " #" #
6s 1 s  27 3 1 s  2s 1 s 1 1
Gs 6
4 1
7:
1 5 0 1 0 s1
s1 s1 2 3
3 1 2 31
It can be transferred to a lower triangular form: 6 s 12 s 1 7 s  2  s  2
6 76 s 1 s 17
6 74 5 :
Ps GsUs 4 1 5
2 3 0 0 1
1 1 2 3 s1
6s 1 s  27 0 1
64 1
74 s  25
1 5 1 
s1 By checking the TDD condition, the stable coprime
s1 s1
2 3 factorisation condition (Gomez and Goodwin 2000)
1
0 and MFDs described condition, we can see that
6s 2 7
64 1 3
7:
5 both upper and lower triangular decouplings are
2
solvable.
s 1 s 1 It is also easy to see that (Hung and Anderson
The MFDs form and stable coprime factorisation 1979, p. 457):
form of P(s) are shown as follows:
jGsj  1 G s 1: 15
2 1 3
0 Based on Lemma 2.7 and Equation (15), we can easily
6s 2 7
Ps 6
4 1
7
5 conclude that both lower and upper triangular
3
2
decouplings are achievable without necessarily trans-
s 1 s 1
ferring G(s) to its upper and triangular form and
" #" #1 calculating its stable coprime factorisation form.
1 0 s2 0

1 3 s 1 s 12 Example 3.2: Consider the following transfer func-
2 3 tion matrix (Gomez and Goodwin 2000):
1 2 31 2 3
6 s1 0 7 s2
6 74 s 1 0 5 1 0
6 7 : 6 7
4 s2 3 5 Gs 4 s 1 s  2 5:
0 1
s 12 s 12 s  1 s 12
1580 H.T. Nguyen and S.W. Su

G(s) is already in lower triangular form. The upper Just based on Equation (16), Lemma 2.7 cannot draw
triangular form can be calculated as follows: any conclusion about the solvability of the upper or
lower decoupling problem for G(s).
Pup s GsUup s
2 32 3
1 0 s  1s  2
1
4s 1 s  2 54 s 13 5 4. Conclusion
s  1 s 12 1 0 This article has proven that the TDD condition
2 3
s  1s  2 (which was regarded as restrictive in Gomez and
1
6 s 13 7 Goodwin (2000)) is actually equivalent to the
64
7:
s1 5 algebraic condition (Gomez and Goodwin 2000) for
0 the triangular decoupling problem. An MFDs
s  1
described necessary and sufficient condition is also
Gomez and Goodwin (2000) showed that the lower provided. Based on these criteria, we can simplify the
triangular decoupling problem is not solvable for this assessment of the solvability of the triangular
system, and upper triangular decoupling is implemen- decoupling problem.
table, based on the stable coprime factorisation criteria
proposed in Gomez and Goodwin (2000).
By checking the TDD condition, we can easily References
obtain the same conclusion:
Anderson, B., and Hung, N. (1978), Multivariable Design
(a) For the lower triangular case: we have
Problem Reduction to Scalar Design Problems, in
G~ s 0  G s 1. This shows that Proceedings of Alternatives for Linear Multivariable
G(s) is not TDD. Thus, the lower triangular Control, National Engineering Consortium, Chicago,
decoupling problem is unsolvable. pp. 8895.
(b) For the upper triangular case: we have Chughtai, S., and Munro, N. (2004), LMI Based Gain
P~ s P s 1. This shows that P(s) Scheduled Control, in Proceedings of the United Kingdom
is TDD. Thus, the upper triangular decoupling Automatic Control Council (UKACC) Control,
problem is solvable. pp. 389396.
Chughtai, S., and Wang, H. (2007), A High Integrity
Now, we apply MFDs to test the solvability of Multivariable Robust Control with Application to
triangular decoupling: a Process Control Rig, IEEE Transactions on Control
Systems Technology, 15, 775785.
(a) For the lower triangular case: Giindeq, A. (1990), Parameterization of all Decoupling
Gs NsD1 s Compensators and all Achievable Diagonal Maps for the
Unity-feedback System, in Proceedings of the 29th IEEE
" #" #1
s1 0 s1 0 Conference on Decision and Control, pp. 24922493.
: Gomez, G., and Goodwin, G. (2000), An Algebraic
s1 s2 0 s 12 Approach to Decoupling in Linear Multivariable
Systems, International Journal of Control, 73, 582599.
Because the first diagonal entries of N(s) and Hung, N., and Anderson, B. (1979), New Multivariable
D(s) have common non-minimum phase zero 1, Frequency-domain-design Techniques, in Conference on
the lower triangular decoupling problem is Control Engineering, Melbourne, pp. 144148.
unsolvable. Hung, N., and Anderson, B. (1979), Triangularization
(b) For the upper triangular case: Technique for the Design of Multivariable Control
Systems, IEEE Transactions on Automatic Control, 24,
Pup s Nup sD1
up s 455460.
 " #1 Koumboulis, F., and Skarpetis, M. (2000), Robust
s  1s  2 s1 s 13 0 Triangular Decoupling with Application to 4WS Cars,
:
0 s1 0 s1 IEEE Transactions on Automatic Control, 45, 344352.
Koumboulis, F. (1996), Input-output Triangular
Because the diagonal entries of Nup(s) and Decoupling and Data Sensitivity, Automatica, 32,
Dup(s) have no common non-minimum phase 569573.
zero, the upper triangular decoupling problem Lin, C., and Hsieh, T. (1991), Decoupling Controller Design
is solvable. for Linear Multivariable Plants, IEEE Transactions on
Automatic Control, 36, 485489.
It is also interesting to see that Lin, C. (1997), Necessary and Sufficient Conditions for
Existence of Decoupling Controllers, IEEE Transactions
jGsj  0 5 G s 1: 16 on Automatic Control, 42, 11571160.
 International Journal of Control
Rosenbrock, H. (1969), Design of Multivariable Control
Systems using the Inverse Nyquist Array, in Proceedings
of the Institute of Electrical Engineers, 116, 19261936.
Vardulakis, A. (1991), Linear Multivariable Control:
Algebraic Analysis and Synthesis Methods, Chichester:
Wiley.
Vidyasagar, M. (1985), Control System Synthesis: A
factorisation Approach, London: The MIT Press.
Wang, Q., and Yang, Y. (2002), Transfer Function Matrix
Approach to Decoupling Problem with Stability, Systems
& Control Letters, 47, 103110.
Wang, Q. (1992), Decoupling with Internal Stability for
Unity Output Feedback Systems, Automatica, 28,
411415.
Appendix A. Proof of Lemma 2.5
Proof: Intuitively, as P(s) is proper, it has no poles at
infinity. Because the zeros at infinity of each diagonal entry
of L Dp (over S) are the poles of P(s), we can draw the
conclusion that there is no infinity zero in each diagonal
entry of L Dp .
Now, we give a more explicit proof.
We can always find a description
Ps NL s1 s, one of its diagonal entries.
1581
such that NL(s) 2 R[s]mm is a lower triangular matrix, and
(s) 2 R[s]mm is a diagonal matrix. It should be noted that
the matrices NL(s) and (s) may not be right coprime.
Let kj, j 2 {1, . . . , m} be the j-th column degrees of (s).
Let
di s ski ai,ki 1 ski 1    ai,1 s ai,0 2 R i 2 f1, . . . , mg
be arbitrary polynomials with no zeros in the closed right-
half plane C and define
Dd s diagd1 s, . . . , dm s 2 Rsmm ,
As NL sDd s1 2 S mm ,
Bs sDd s1 2 S mm :
Then, we have P(s) A(s)B1(s), where A(s) 2 S mm is a
proper rational lower triangular matrix and B(s) 2 S mm is a
proper rational triangular matrix. Furthermore, all diagonal
entries of B(s) are non-strictly proper.
As Ps LNp sL Dp s AsB1 s, we should be able
to find a proper rational lower triangular matrix
C(s) 2 S mm, such that
As LNp sCs,
Bs L Dp sCs: A1
Now, we assume L Dp s has an infinity zero in one of its
diagonal entries. Then according to Equation (A1), C(s)
should be non-proper, as B(s) is non-strictly proper. This is
incorrect. Thus, L Dp s should not have an infinity zero in