Upload
wajdi
View
212
Download
0
Embed Size (px)
Citation preview
This article was downloaded by: [McMaster University]On: 30 October 2014, At: 11:17Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
International Journal of MathematicalEducation in Science and TechnologyPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/tmes20
Circuit analysis as a discrete boundaryvalue problemRaghib Abu-Saris a & Wajdi Ahmad ba Department of Basic Sciences , University of Sharjah , P. O. Box27272, Sharjah, U.A.Eb Department of Electrical and Electronics Engineering ,University of Sharjah , P. O. Box 27272, Sharjah, U.A.EPublished online: 02 Nov 2009.
To cite this article: Raghib Abu-Saris & Wajdi Ahmad (2007) Circuit analysis as a discrete boundaryvalue problem, International Journal of Mathematical Education in Science and Technology, 38:6,832-839, DOI: 10.1080/00207390701240877
To link to this article: http://dx.doi.org/10.1080/00207390701240877
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.
This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Circuit analysis as a discrete boundary value problem
RAGHIB ABU-SARIS*y and WAJDI AHMADz
Department of Basic Sciences, University of Sharjah,P. O. Box 27272, Sharjah, U.A.E.
Department of Electrical and Electronics Engineering,University of Sharjah, P. O. Box 27272, Sharjah, U.A.E.
(Received 30 March 2006; accepted 19 October 2006)
In this note, it is shown that mesh and nodal techniques for electric circuitanalysis can be cast as non-autonomous discrete boundary value problems.An elegant necessary and sufficient condition for the existence of a uniquesolution for such problems is described. The developed condition iscomputationally tractable, as it only involves checking a determinant that isspecially constructed from the coefficients of the underlying difference equation.Illustrative examples are presented to demonstrate the validity of the result.
1. Motivation and problem formulation
The discrete boundary value problem (DBVP) addresses difference equations subjectto boundary conditions. Such a problem may result from discretizing a continuousboundary value problem (CBVP) with an underlying ordinary differential equation.Alternatively, it can be a result of modelling a discrete (dynamical) process that hasto satisfy a set of prescribed conditions or targets. Typically, part of these conditionsis known at the initial or present time, while the rest are described at future orlater time(s). Such problems appear in a wide variety of science and engineeringapplications. The following example, in particular, is drawn from the field ofelectrical circuit analysis, where it is shown that a circuit analysis problem can becast as a non-autonomous DBVP.
Nodal and Mesh equations are popular techniques for analysis of electricalcircuits. Nodal analysis begins by assigning different voltages to the different nodesin the circuit, as shown in figure 1, and writing the Kirchoff Current Law (KCL)equation for each one of the nodes, which essentially states that the sum of allcurrents entering the node must equal the sum of all currents leaving the node. Onthe other hand, mesh analysis begins by assigning currents to the different loops inthe circuit, as shown in figure 2, and writing the Kirchoff Voltage Law (KVL) foreach closed loop or mesh. The KVL dictates that the sum of all voltage drops in aclosed loop be zero. In either technique, a second-order discrete difference equationis obtained, where the index j of the equation represents, respectively, the node andloop numbers as shown below. The classification ‘‘boundary value’’ becomesapparent when we impose conditions on the voltages (currents) at two ends
DOI: 10.1080/00207390701240877
*Corresponding author. Email: [email protected]
832 Classroom notes
Dow
nloa
ded
by [
McM
aste
r U
nive
rsity
] at
11:
17 3
0 O
ctob
er 2
014
(loops) of the circuit, e.g. Vð0Þ ¼ V0ðIð0Þ ¼ I0Þ and VðN Þ ¼ VNðI ðN Þ ¼ INÞ where Nis one of the circuit nodes (loops). The voltages V(0) and V(N ) (currents I(0) andIðN ÞÞ are, respectively, called initial and final (or boundary) conditions. A commonscenario might be VðN Þ ¼ 0ðIðNÞ ¼ 0Þ which signifies a short circuit (open switch, S)condition imposed at node (loop) N, thus bringing the voltage (current) down tozero. This could represent a short-circuit fault in the first case, or a load beingdisconnected in the second case. Hence, the response of the difference equation isrestricted at the two ends, and there is a ‘‘gap’’ between the initial and final indices.Within the gap, the response is not restricted.
The equations representing the two circuits above can be easily derived asfollows. Applying KCL to the circuit in figure 1 yields:
Vðj Þ � Vð jþ 1Þ
R2j¼
Vð jþ 1Þ
R2jþ1þVð jþ 1Þ � Vð jþ 2Þ
R2jþ2, j ¼ 0, 1, . . . ,N� 2
Hence, the following DBVP is obtained:
Vð jþ 2Þ þ a1ð j ÞVð jþ 1Þ þ a0ð j ÞVð j Þ ¼ 0 ð1Þ
Vð0Þ ¼ V0 and VðN Þ ¼ 0 ð2Þ
where
a1ðj Þ ¼ � 1þR2jþ2
R2jþR2jþ2
R2jþ1
� �and a0ð j Þ ¼
R2jþ2
R2j: ð3Þ
Similarly, applying KVL to the circuit in figure 2, we get:
Rjþ1Ið jþ 2Þ � Rj þ Rjþ1
� Ið jþ 1Þ þ RjIð j Þ ¼ 0, j ¼ 0, 1, . . . ,N� 2 ð4Þ
Figure 1. Nodal setup.
Figure 2. Mesh setup.
Classroom notes 833
Dow
nloa
ded
by [
McM
aste
r U
nive
rsity
] at
11:
17 3
0 O
ctob
er 2
014
and so we have the following DBVP:
Ið jþ 2Þ þ a1ð j ÞIð jþ 1Þ þ a0ð j ÞIð j Þ ¼ 0, ð5Þ
Ið0Þ ¼ I0 and IðN Þ ¼ 0 ð6Þ
where
a1ðjÞ ¼ � 1þRj
Rjþ1
� �and a0ðj Þ ¼
Rj
Rjþ1ð7Þ
Notice that the linear difference equations (1) and (5) stated above arenon-autonomous owing to the fact that their coefficients depend on the index j.
In this paper, we are concerned with the fundamental question of existence anduniqueness which is a corner stone for well-posedness of a problem. For definiteness,we shall consider a non-autonomous linear difference equation of order k52 givenin the recursive form:
Xki¼0
aiðtÞyðtþ iÞ ¼ gðtÞ, t ¼ 0, 1, 2, . . . ð8Þ
and subject to separable boundary conditions:
yðni þ jÞ ¼ �ij, i ¼ 1, . . . , r and j ¼ 0, . . . , ki � 1 ð9Þ
or to boundary conditions of the type
�jyðniÞ ¼ �ij, 14i4r, 04j4ki � 1 ð10Þ
where
0 ¼ n14n1 þ k1 � 1 < n24n2 þ k2 � 1 < � � � < nr�14nr�1 þ kr�1 þ 1 < nr
and ki are positive integers such that 14r4k and k1 þ � � � þ kr ¼ k. This type ofboundary conditions is commonly known in the literature as Hermite type. Our goalis to develop a necessary and sufficient condition for the existence of a uniquesolution of DBVP (8), (9) and DBVP (8)–(10).
DBVPs received a great deal of attention recently (see [1, pp. 629–744] and[2, pp. 229–278, 327–341], and the references cited therein). In general, the resultsobtained thus far provide sufficient conditions for the existence and/or uniquenessof solution(s) of a DBVP. Recently, the authors in [3, 4] have developed acomputationally tractable and easy-to-apply necessary and sufficient condition forthe existence of a unique solution of DBVP (8), (9) in which the governing differenceequation is autonomous. However, the proofs presented therein are computationallyinvolved and employ a generalization of what is called the exponential Vandermondedeterminant in which the entries are powers of the roots of the associated character-istic equation. This approach, unfortunately, is not valid for the non-autonomouscase, as it is meaningless to speak of characteristic equations for non-autonomousdifference equations.
This paper is organized as follows: in section 2 we establish the main result.In section 3, we give illustrative numerical examples. Finally, in section 4, we drawconclusions.
834 Classroom notes
Dow
nloa
ded
by [
McM
aste
r U
nive
rsity
] at
11:
17 3
0 O
ctob
er 2
014
2. Existence and uniqueness results
First, it is known (see [2, p. 14]) that
yðtþ jÞ ¼Xj‘¼0
j
‘
� ��‘yðtÞ
where �‘ is the ‘th-order forward difference operator. Therefore, for each 14i4r,equation (10) can be rewritten as
yðni þ jÞ ¼Xj‘¼0
j
‘
� ��‘yðniÞ ¼
Xj‘¼0
j
‘
� ��i‘ ¼ �ij, j ¼ 0, . . . , ki � 1
This means that the difference between Hermite boundary conditions (10) andseparable boundary conditions (9) is rather symbolic, and the two types are,essentially, equivalent.
Now, we are ready to establish our existence and uniqueness theorem.Theorem 2.1: Let s‘ ¼ n‘þ1 � n‘ � k‘, ‘ ¼ 1, . . . , r� 1, and let s ¼
Pr�1‘¼1 s‘ ¼ nrþ
kr � k. Then the existence of a unique solution of DBVP (8), (9) is equivalent to thenonsingularity of the block matrix
A ¼ A1 � � � Ar�1
where Aj is the s sj matrix given by
Aj ¼
anjþkj ð0Þ anjþkjþ1ð0Þ � � � anjþ1�1ð0Þ
anjþkj�1ð1Þ anjþkj ð1Þ � � � anjþ1�2ð1Þ
..
. ... ..
.
anjþkj�sþ1ðs� 1Þ anjþkj�sþ2ðs� 1Þ � � � anjþ1�sðs� 1Þ
0BBBBB@
1CCCCCA
j ¼ 1, . . . , r� 1.
Proof: First, to simplify notation we define ajðtÞ ¼ 0 if j<0 or j> k. Using thisindex extension and equation (8), we have
gðtÞ ¼Xkþtj¼t
aj�tðtÞyð jÞ, t ¼ 0, 1, 2, . . .
¼Xk1�1j¼t
aj�tðtÞyðj Þ
þXr�1‘¼1
Xn‘þ1�1j¼n‘þk‘
aj�tðtÞyðj Þ þXn‘þ1þk‘þ1�1
j¼n‘þ1
aj�tðtÞyðj Þ
" #
þXkþt
j¼nrþkr
aj�tðtÞyðj Þ
Classroom notes 835
Dow
nloa
ded
by [
McM
aste
r U
nive
rsity
] at
11:
17 3
0 O
ctob
er 2
014
It is worth mentioning that some of these sums are zeros due to either beingempty sums (upper index is less than lower index) or due to the index extensionintroduced above. Furthermore, the purpose of the sum-splitting presented above isto facilitate the use of boundary conditions given in equation (9) as will beseen below.
Now, a unique solution of DBVP (8), (9) exists if and only if the missing y-valuesat intermediate time instances, i.e. the ‘gaps’, are uniquely determined from the givenboundary conditions. To this end, rearranging the above equations fort ¼ 0, 1, . . . , nr þ kr � k� 1, we obtain the following system of linear equations:
Xr�1‘¼1
Xn‘þ1�1j¼n‘þk‘
aj�tðtÞyðj Þ ¼ gðtÞ �Xk1�1j¼t
aj�tðtÞ�1j �Xr�1‘¼1
Xn‘þ1þk‘þ1�1
j¼n‘þ1
aj�tðtÞ�ð‘þ1Þj
Observe that the coefficient matrix of this system of linear equations is given by
ak1 ð0Þ � � � an2�1ð0Þ j � � � j anr�1þkr�1 ð0Þ � � � anr�1ð0Þak1�1ð1Þ � � � an2�2ð1Þ j � � � j anr�2þkr�1�1ð1Þ � � � anr�2ð1Þ
..
. ...
j j ... ..
.
ak1�sþ1ðs� 1Þ � � � an2�sðs� 1Þ j � � � j anr�2þkr�1�sþ1ðs� 1Þ � � � anr�sðs� 1Þ
0BBB@
1CCCA
Hence, the result follows. œ
3. Applications
To illustrate the applicability of Theorem 2.1, we present the following examples.
Example 3.1: Consider the DBVP
yðtþ 5Þ þ a4ðtÞyðtþ 4Þ þ a3ðtÞyðtþ 3Þ þ a2ðtÞyðtþ 2Þ
þa1ðtÞyðtþ 1Þ þ a0ðtÞyðtÞ ¼ gðtÞ, t ¼ 0, 1, 2, . . .
yð0Þ ¼ �10, yð2Þ ¼ �20, yð4Þ ¼ �30, yð6Þ ¼ �40, yð8Þ ¼ �50:
8><>:
In this case, we have k¼ 5, n1¼ 0, n2¼ 2, n3¼ 4, n4¼ 6, n5¼ 8, andk1 ¼ � � � ¼ k5 ¼ 1. Thus, s1 ¼ s2 ¼ s3 ¼ s4 ¼ 1, and there are four blocks each ofsize 4 1. Therefore,
AT1 ¼ a1ð0Þ a0ð1Þ a�1ð2Þ a�2ð3Þ
¼ a1ð0Þ a0ð1Þ 0 0
AT2 ¼ a3ð0Þ a2ð1Þ a1ð2Þ a0ð3Þ
AT3 ¼ a5ð0Þ a4ð1Þ a3ð2Þ a2ð3Þ
¼ 1 a4ð1Þ a3ð2Þ a2ð3Þ
AT4 ¼ a7ð0Þ a6ð1Þ a5ð2Þ a4ð3Þ
¼ 0 0 1 a4ð3Þ
836 Classroom notes
Dow
nloa
ded
by [
McM
aste
r U
nive
rsity
] at
11:
17 3
0 O
ctob
er 2
014
and
A ¼ A1 A2 A3 A4
¼
a1ð0Þ j a3ð0Þ j 1 j 0
a0ð1Þ j a2ð1Þ j a4ð1Þ j 0
0 j a1ð2Þ j a3ð2Þ j 1
0 j a0ð3Þ j a2ð3Þ j a4ð3Þ
0BBB@
1CCCA
Hence, the non-autonomous DBVP above has a unique solution if and only if theabove-constructed A is nonsingular. In particular, if a4ðtÞ ¼ a3ðtÞ ¼ a2ðtÞ ¼ a1ðtÞ ¼ 0and a0ðtÞ ¼ cosð�tÞ, then
jAj ¼
0 0 1 0
�1 0 0 0
0 0 0 1
0 �1 0 0
���������
���������¼ �1 6¼ 0
and so a unique solution is guaranteed to exist.
Example 3.2: We turn back to our motivating example in Section 1. In view ofTheorem 2.1, DBVPs (1), (2) and (5), (6) have unique solutions if and only if thefollowing tridiagonal matrix
A ¼
a1ð0Þ 1
a0ð1Þ a1ð1Þ. .
.
. .. . .
.1
a0ðN� 2Þ a1ðN� 2Þ
0BBBBB@
1CCCCCA
is nonsingular.Equation (3) implies that the matrix A is strictly diagonally dominant and so it is
nonsingular [5, Theorem 6.20, pp. 404, 405]. Therefore, DBVP (1), (2) always has aunique solution. On the other hand, as can be seen from equation (7), thecorresponding matrix A is diagonally dominant but not a strict one. However,nonsingularity is still valid. The justification is as follows.
Suppose to the contrary that ¼ 0 is an eigenvalue of A and ~x ¼ ðx1, . . . , xN�1Þ isthe corresponding eigenvector. Then A~x ¼ ~0, or more explicitly
� 1þR0
R1
� �x1 þ x2 ¼ 0
Rj
Rjþ1xj � 1þ
Rj
Rjþ1
� �xjþ1 þ xjþ2 ¼ 0, j ¼ 1, . . . ,N� 3
RN�2
RN�1xN�2 � 1þ
RN�2
RN�1
� �xN�1 ¼ 0
ðÞ
8>>>>>>>>>>>>><>>>>>>>>>>>>>:
Classroom notes 837
Dow
nloa
ded
by [
McM
aste
r U
nive
rsity
] at
11:
17 3
0 O
ctob
er 2
014
Since the resistances R0js are positive, the first equation of () implies thatjx2j > jx1jy. Combining this fact with Triangle Inequality (TI), we conclude:
jx3j5 1þR1
R2
� �jx2j �
R1
R2jx1j > jx1j
This, in turn, leads to the conclusion that jx3j > jx2j. Because otherwise we have,again, by TI
R1
R2jx1j5 1þ
R1
R2
� �jx2j � jx3j5
R1
R2jx3j
which is a contradiction to what we found so far. Following this line of reasoning forj ¼ 1, . . . ,N� 3, we get jx1j < jx2j < � � � < jxN�1j. But, from the last equation of (),we have
1þRN�2
RN�1
� �xN�1 ¼
RN�2
RN�1xN�2
which implies that jxN�1j < jxN�2j which is a contradiction!From a circuit theory standpoint this makes perfect sense, since a singular
matrix A would imply multiple solutions to the circuit, which implies that the meshcurrents are not unique! This, of course, cannot happen in a real circuit wherecurrents are determined uniquely by the values of the resistors and the voltagesimposed on them.
4. Conclusions
We have established a necessary and sufficient condition for the existence of a uniquesolution for linear non-autonomous DBVPs. The boundary conditions examined areof Hermite and separable types. The condition thus established is easy-to-apply andlends itself nicely to computations. It is directly linked to the very structure ofthe difference equations. Calculating a determinant obtained from the equationcoefficients is all that is required. We have demonstrated the validity of the resultvia an example from circuit analysis where the results obtained here supportwell-established electrical circuit theory.
The results obtained in this research degenerate to the authors previouslyestablished results for autonomous DBVPs in [4], with simplified proofs.
As for future research, the authors are currently studying other structures ofboundary conditions.
yExamining the circuit in figure 2 and its corresponding equation (4), we can see from theoutset that this scenario leads to violation of circuit theory. It is straightforward to showthat the ‘mesh’ currents become bigger as we get closer to the current source of the circuit,i.e. as we move to the left. Hence, a lower-index ‘‘mesh’’ current is bigger than a higher-index one, which is contrary to the above implication.
838 Classroom notes
Dow
nloa
ded
by [
McM
aste
r U
nive
rsity
] at
11:
17 3
0 O
ctob
er 2
014
References
[1] Agarwal, R., 2000, Difference Equations and Inequalities: Theory, Methods, andApplications, 2nd edn (New York: Marcel Dekker).
[2] Kelley, W. and Peterson, A., 2001, Difference Equations: An Introduction withApplications (New York: Harcourt Academic Press).
[3] Abu-Saris, R. and Ahmad, W., 2003, A necessary and sufficient condition for existence ofunique solution of discrete boundary value problem, International Journal of Mathematicsand Mathematical Sciences, 39, 2455–2463.
[4] Abu-Saris, R. and Ahmad, W., 2004, Generalized exponential Vandermonde determinantand Hermite multipoint discrete boundary value problem, SIAM Journal of MatrixAnalysis and Applications, 25, 921–929.
[5] Burden, R and Faires, J., 1997, Numerical Analysis, 6th edn (New York: Brooks/Cole).[6] Abu-Saris, R., 2006, On solving non-autonomous linear difference equations with
applications, International Journal of Mathematical Education in Science and Technology,37, 618–623.
[7] Tay, E., Toh, T., Dong, F. and Lee, T., 2004, Convergence of a linear recursive sequence,International Journal of Mathematical Education in Science and Technology, 35, 51–63.
A refined Cauchy–Schwarz inequality
PETER R. MERCER*
Department of Mathematics,Buffalo State College, Buffalo, NY 14222, USA
(Received 5 July 2006)
Much of what follows is inspired by [1], in which refinements of the triangleinequality and its reverse inequality are obtained for nonzero x and y in a normedlinear space. The refinements are given in terms of the angular distance
x
xk k�
y
y�� ��
����������
and they have some interesting consequences.Now upper and lower bounds involving kx==kxk � y==kykk can be just as well
regarded as upper and lower bounds for kx==kxk � y=kykk by way of algebraicmanipulations. And if we are not just in a normed linear space, but a real innerproduct space then a computation reveals that
1�1
2
x
xk k�
y
y�� ��
����������2
¼x, y� �xk k y�� ��
*Email: [email protected]
DOI: 10.1080/00207390701359370
Classroom notes 839
Dow
nloa
ded
by [
McM
aste
r U
nive
rsity
] at
11:
17 3
0 O
ctob
er 2
014