Beruflich Dokumente
Kultur Dokumente
OF ELECTRIC
CIRCUITS AND MACHINES
L. V. BEWLEY
PROFESSOR OF ELECTRICAL ENGINEERING
AND
DEAN OF ENGINEERING
LEHIGH UNIVERSITY
singled out and crowned as his cardinal achievement, this is it. He devised
innumerable clever procedures and artifices for simplifying results, or t o
render them more intelligible, or for saving labor. Sometimes he became so CONTENTS
immersed in his search for a tensor or matrix which would bring about a
desired operation that he failed to notice that i t could be done by inspection
PART I
or by some simple device, but these things are t o be expected in pioneer work
and are easily forgiven. While his intuition was uncanny, i t was too often Matrices, Tensors, and Circuits
substituted for a rigorous and rational development, and this can prove CHAPTER PAGE
disconcerting to the reader.
For twenty years I have taught a course in the applications of tensor
analysis, essentially as covered in this book, to first year graduate students
a t Lehigh University. I am convinced that a course, based on this book,
could be given to electrical engineering seniors, and that in the not too
distant future such a course will become a standard part of undergraduate
curricula, so that this powerful tool may become a part of the equipment
of every electrical engineer.
I wish to express my thanks to the students of my classes, especially
Joseph Teno, William Hollabaugh, and Donald Talhelm, who searched so
diligently for errors, and to Mrs. Marion Stempkowski, who typed the
manuscript.
L. V. BEWLEY
Bethlehem, Pennsylvania
January, 1961
P A R T I1
Machine Analysis
BASICCONCEPTS MACHINES.
IN ELECTRIC
THE QUASI-HOLONOMIC
GENERALIZED
MACHINE .
EQUATION.
USE O F THE LAGRANGE
USE OF MAXWELL'SEQUATION .
USE OF THE MAXWELL-LORBNTZ
EQUATION .
COMPARISON
OF THE GENERAL
SOLUTIONS
OF INDICES
THE RAISINQAND LOWERING AND GENERALIZED
PER-UNITCONCEPTS.
SMALLOSCILLATIONS
AND HUNTING
v
vi CONTENTS
20 SYNCHBONOUS
MACHINES . . 222
21 MACHINES .
INDUCTION . 246
MACHINES
22 COMMUTATOR . 268
MACHINES.
23 INTERCONNECTED . 283
Part I
with the understanding that it is to be summed on k. As so often happens among the currents (or voltages) only, although hybrid relations may some-
when some simplification in nomenclature is adopted, new concepts are born. times have to be established. Numerous cases will be dekeloped in this book
Certainly, in the way Eq. 1 4 was arrived at, ej is a particular voltage depend- by way of illustration.
ing on the value assigned to j, and ik is a particular current depending on k, Usually, but not always, interconnection of the branches results in the
and Z j k is a particular impedance depending on the choice of j and k. But need for fewer currents to describe the performance of the network. Thus
may we not now regard Eq. 1 4 as the embodiment of all the n voltages, all mesh currents may replace branch currents, or currents flowing in one direc-
the n currents, and all the n2 impedances ? If so, Eq. 1 4 vastly extends our tion may be substituted for currents flowing in opposite directions, or one
concept of a voltage, a current, and a n impedance as not single quantities current a t a junction may be replaced by the algebraic sum of other currents
but entireties associated with the whole n-branch system or network. Thus
ea is only a component of ej, ic is only a component of ik, and Z,, is only a
component of Zj,. On this basis Eq. 1 4 may be written in matrix notation
(for convenience showing only three coordinates) :
The arrows on the matrices indicate that the rows of Z j , are to be multiplied
by the column of ik. (The rules and procedures for matrix manipulation are
given in the next chapter.) Notice that voltages are written as rows and
currents as columns, corresponding to the subscript and superscript conven- Fig. 1-2. Examples of substitutions of variables ("transformations of co-
tion. Later on, these will be referred to as covariant and contravariant ordinates").
tensors of rank one.
a t that junction in accordance with Kirchhoff's law, and so on. I n any event,
Much of the engineering literature is already saturated with the dyadic
a set of linear relationships can be formulated to express the relationship
notation between the original (or "old") branch currents and the "new" coordina.tes.
e=Z*i
in which e, Z, and i, written in boldface type, have the same significance as I n setting up these relationships, or equations of constraint, i t is sufficient
the corresponding matrices in Eq. 1-5, and the dot product indicates the to show on the connection diagram only simple links, each consisting of a line
same arrow multiplication as in Eq. 1-5. with terminals a t its ends, since the equations of constraint are independent
of the composition of the branches. Some examples of different kinds of
1-4. The Equations of Constraint. Equations of constraint express the constraints are shown in Fig. 1-2.
conditions imposed by the interconnections of the network or the relation- I n Fig. 1-2a a bra,nch current ia is replaced by a branch current im'
ships which exist between the various currents and voltages. I n circuit theory flowing in the opposite direction, and the equation for this constraint is
these conditions are usually given by Kirchhoff's first and second laws :
ia = -im'
X (currents a t a junction or node) =0
pig. 1-2b shows a junction of three currents ia, ib,ic. Obviously, two currents
X (voltages around a circuit or mesh) = 0 (1-7)
ib' and ic' are sufficient to describe this situation, since
But as a by-product of the idea of invariancy of power, under a transforma-
tion of coordinates in a n electrical network, i t will be shown in Chapter 3
+
ia = ib' ic'
that one of the two laws of Kirchhoff (either one) is redundant. From the ib = jb'
point of view of tensor analysis it is usually sufficient t o set up relationships ic = p'
8 MATRICES, TENSORS, A N D CIRCUITS [Ch. I Art. 1-51 SOME PRELIMINARY CONSIDERATIONS 9
and, in general, the number of k' terms is usually less than the number of k symmetrical component substitution, or sines and cosines as in the synchro-
terms. Fig. 1-2c shows a mesh composed of four links in a network. Replac- nous machine example. Or they may be much more complicated. It is clearly
ing the branch currents ia, ib, ic, id by the loop current i", iP, t',is, i', there evident from Eq. 1-8 or 1-9 that
results
i" = i&- ii"
i b Ii& - iLJ
This is the transformation tensor. I t expresses the relationship between the
ic = j& - Y
i old coordinates or variables and the new coordinates or variables. The
id Ii d - je transformation tensor is the mathematical model of the connection diagram.
The study of its characteristics and behavior might be said to be the essential
Fig. 1-2d shows an ideal transformer in which the sum of the ampere-turns feature of tensor analysis itself. Much of this book will be devoted to its
is balanced (negligible magnetizing current). Accordingly, the secondary consideration. I n order properly to develop the study of the transformation
current may be expressed in terms of the primary current by the turn ratio: tensor, it is necessary first to explore and make available certain mathe-
matical tools. Accordingly, the next chapter provides the essential elements
on matrices. A separate chapter on the definitions and manipulative pro-
cedures of tensor analysis is also included, and the necessary ideas of this
field are further developed along with their application to circuits and
Another transformation is the substitution of symmetrical components machines. I t will be found that they evolve quite naturally and painlessly
(9,il, 3) for phase quantities (ia, ib, iC)
by the relationships and that no great amount of mathematical knowledge is necessary.
1-5. A Few Topological C o n ~ i d e r a t i o n s . ~ ?I~n* the analysis of
networks and electric machines i t is sometimes advantageous to use mesh
networks and a t other times to use node networks, depending on which
method of attack involves the least number of variables, the ease with which
And a final example is the substitution of direct, quadrature, and zero constraints may be recognized, and the known characteristics of certain
components ( i d ,iQ,z?) for phase quantities (ia,ib,ic)in the two-reaction theory elements. For example, it is usually better to handle vacuum tube circuits
of synchronous machines : as node networks. I t is therefore desirable to know a t the onset how many
branches, nodes, meshes, and junction pairs there are in a network. Let
ia = i0+ id cos 6 + iQsin 0 number of branches (which can be counted)
ib= i0+ id cos (I3 - 120") + ia sin (I3 120")
-
number of junctions or nodes (countable)
iC= i0+ idcos (0 + 120") + iQsin ( 8 + 120") number of subnetworks, which are not physically connected to one
another but may be inductively coupled (countable)
There are, of course, innumerable other ways in which new coordinates
minimum number of independent non-redundant meshes or closed
may be substituted for old, but they all reduce to a set of linear equations
circuits in the network for which each branch will be traced a t least
of the form
once
+ Ci,ib' + . . + CE,~"'
ia = ~',a,ia'
minimum number of junction pairs (any two junctions in a sub-
network constitute a junction pair) for which each junction is
included a t least once
Then these concepts are related by the two equations
These equations may be expressed in tensor notation as
B=M+P=M+J-S (1-12)
The C$ coefficients may be numericallj. equal to -1 as in Fig. 1-2a, +1 as * Superior numbers refer to publications in which ideas discussed in this book first
in Fig. 1-2b, f1 as in Fig. 1-2c, or they may be complex numbers as in the appeared. See the Bibliography at the end of the book.
10 MATRICES, TENSORS, A N D CIRCUITS [Ch. I Art. 1-61 SOME PRELIMINARY CONSIDERATIONS
Hence, given any network comprising S independent subnetworks, with a Then, by Eqs. 1-11 and 1-13,
total number of branches B and junctions J, the minimum number of inde-
P=12-3=9 junction pairs
pendent non-redundant meshes M is
M = B + S - J (1-13)
M = 14 + 3 - 12 = 5 meshes (a, p, y, 6, a)
If M < P, the network can generally be more easily analyzed as a mesh One choice of the fire meshes is shown as (a, p, y , 6, a), but any other five
network; while if M > P, i t can generally be more easily analyzed as a node independent meshes may be selected. Similarly, any nine independent pairs
network. Eqs. 1-11 and 1-13 thus furnish simple criteria for choosing the of the junctions may be taken as the nine junction pairs. At each junction
in a network a relationship (Kirchhoff's first law) may be written between
the currents a t that junction. (J - S) such non-redundant relationships, or
TABLE 1-1
Third generalization postulate. A tensor equation true in one reference TABLE 1-2
frame remains invariant in form upon transformation to a new reference The Three Types of Networks
frame of the same or a different type.
The procedure to be followed in passing from simple to more complex Mesh I Junction P a i r Orthogonal Network
systems and interconnections is outlined in Table 1-1.
junction (Pig. l 4 c ) .
MATRICES, TENSORS, AND CIRCUITS A r t 1-81 SOME PRELIMINARY CONSIDERATIONS 15
PROBLEMS
1-1. How many branches, junctions, junction pairs, subnetworks, and meshes
does the delta-quadruple zigzag transformer of Fig. 5-8 have? Would you analyze
this transformer as a mesh or junction-pair network?
1-2. Show that Eq. 1-10 follows from Eq. 1-8 or Eq. 1-9 if the CF, are inde-
pendent of the currents.
1-3. Con~pilea set of duals for mechanical, optical, and thermodynamical
systems.
1-4. Much of this book is concerned with developing methods and procedures
for analyzing a complex system by first breaking it up into its component parts,
Fig. 1-4. finding the solution for each part in terms of its own terminal conditions, and
then interconnecting the parts into a composite whole and obtaining the over-all
solution. Consider a single-phase power system comprising a generator, sending-
end transformer, transmission line, receiving-end transformer, and motor. Sup-
pose that the A, B, C, D constants for each separate part of this system are known
TABLE 1-4 and thus that the relationships between the terminal voltages and currents for
each part are given in the form
Covariance Contravariance
Resistance Conductance G ~ B
Magnetic flux Dielectric flux
(or charge) Qa
Permeance Reluctance pap I, = CE, + DI,
Magnetic flux density Magnetizing force (rnrnf) ha
Inductance Reciprocal inductance Obviously, once the terminal voltages and currents for any part are known, the
Elastance Capacitance Cap
Voltage Dielectric flux density internal behavior of that part can be determined. For example, the voltage and
(displacement) Da current at any distance z from the receiving end of the transmission line may be
Kinetic energy Potential energy V found in terms of its terminal voltages and currents by the well-known formulas
Series Shunt
+ I, dm sinh d
Short circuit Open circuit
E = E, cosh Z/%z E z
2-1. n-Way Matrices. Tensor analysis, in its grand concepts and broad
generalizations, may be regarded as a strategic mathematical tool. But, for
handling the routine operations involved in its applications-addition,
subtraction, multiplication, and inverses-matrix algebra is indispensable.
This chapter is concerned with the matrix algebra necessary for our purposes.
Matrix algebra was developed for the systematic solution of sets of simul-
taneous equations and for effecting transformations of variables.
A matrix is a rectangular (not necessarily square) array of numbers (real,
complex, constant, variable, or operational) with which are associated certain
rules of procedure or operation. In books on mathematics such an array is
usually enclosed by brackets or parentheses:
These I-way matrices might just as well be shown as column matrices rather
than as row matrices, depending on the use t o which they are to be put.
17
18 MATRICES, TENSORS, A N D CIRCUITS [Ch. 2 Art. 2-21 MATRICES 19
Impedances and admittances are represented by 2-way matrices : which higher-rank matrices are handled will be demonstrated as they are
encountered.
A 0-way matrix is a number which has no reference to any coordinate
system-a simple constant of multiplication, for example.
An n-way matrix has kn cells or constituents, although in practice many of
these may be unfilled or zero. The way in which a matrix is filled generally
offers considerable information about the type of physical system involved
and leads to systematic classifications.
Since a matrix expressed by a single symbol like e or i or Z or A represents
not one quantity but a whole set of quantities, there follows the conception
of set. A 2-way matrix represents a 2-way set. The idea of set is of great
importance in modern mathematics.
A 2-way matrix has all its diagonal elements equal to unity and all other
elements equal t o zero; it is called the unit matrix or idemjactor I or
A 3-way matrix may be shown as a cube: Kronecker delta 6 :
a b c
a 1 0 0
1 ifj=k
% = I = b
c
0
o
1
0
0
1
=(
0 ifj#k
(2-7)
a
- b
Y
c
. -
7
I t has the unique property, t o be proved later, that any matrix multiplied
by it remains unchanged.
For manipulation purposes a 3-way matrix can be used only by breaking i t A matrix having all its elements equal to zero is called a null matrix,
up into 2-way matrices. This can be done in many different ways, for example : and it has the property that any matrix multiplied by i t is reduced to zero.
1. By taking horizontal slices along the x-axis so that 2-2. Addition and Subtraction. Only matrices of the same dimensions
and with the same indices may be added. Addition or subtraction then
Ax,, = Aauz + Abuz + Acvz consists in adding or subtracting corresponding (same indices) components.
Thus
2. By taking vertical slices along the y-axis so that
a b c a b c a b c
Ax,, = A,, + Ax,, + Axcz L=JI i2=mA&D
l B ~ CE & F
For 4-way and higher-rank matrices no pictorial symbols are feasible (or
necessary), and they are specified in terms of their 2-way slices. The way in
I It often happens, however, that two matrices with different indices need
to be combined into a single matrix. It is then imagined that both matrices
20 MATRICES, TENSORS, A N D CIRCUITS [Ch. 2 Art. 2-31 MATRICES 21
possess all indices, but that the elements of the non-present indices in each since, no matter how they are designated, they mean the same thing-sum-
matrix are zero. For example, rnation. Matrices of any rank are multiplied according to the rule
where the dummy index y vanishes in the result, and therefore the h a 1
matrix always has two less indices than the sum of the number of indices in
the factor matrices of the product. I n actual practice, only 1- and 2-way
a b c a b c a b c matrices need be multiplied:
the dots showing the positions vacated by the dummy indices and the arrows
Notice that in the tensor notation this is obvious. showing that the rows of Ajk are to be multiplied by the columns of Bk"
Multiplication of two 1-way matrices corresponds exactly t o the dot and added:
product of vector analysis; that is, corresponding elements of the two
matrices are multiplied and added. The result is a scalar or 0-way matrix.
Thus
I' =s .i = l l mI= +
I n tensor notation, remembering the summation convention,
eiia + ebib ecic (2-1 1)
Of course, this is easily seen from the tensor notation Notice that only a square matrix can have an inverse, for a rectangular
(singular) matrix is equivalent to a square matrix having a row (or column)
+
Ajk6k = A j a S k Ajb& + . Ainz6: . + + - - + A j n S Z = Aj, (2-16)
of zero elements, whose determinant would then be zero. This would corre-
since all 6: = 0 for k # nz, while 6::;= 1. spond t o trying t o solve a set of simultaneous equations having either more
variables or less variables than the number of equations.
mjR?j
2-4. Inverse of a 2-Way Matrix. The inverse of a 2-way matrix is
analogous to division. Only 2-way square ma,trices have inverses. Consider As an example let
a system of simultaneous equations :
for which D = 30
If this system of equations is solved for the i k by determinants, then, accord-
ing to Cramer's rule, Z= 4
1. Interchanging rows and columns gives the transpose:
in which, for any k and any j,
= cofactor of Z j ,
ykj (2-1 8) z, = 2
determinant of all the Z j k
For example, if the impedance matrix is
A symmetrical matrix is symmetrical about the main diagonal: ~ u int tensor notation the commutative law with respect t o products is
obeyed :
AjkBkm= BkmAjk
since the indices suffice to indicate the products which are t o be taken.
The conjugate of a matrix A which is a function of complex variables is a
matrix A* whose elements are the conjugate complex variables; that is,
A skew symmetrical matrix has equal elements of opposite sign about the
main diagonal of zero elements: if A = f(x + jy) then A* = f(x - jy)
a B 7 -- -sin xp
Dyadics and matrices do not obey all the laws of algebra. I n these 2-7. Compound Matrices6 I n simple problems involving relatively
notations, few variables, the corresponding matrices have few components and generally
can be handled as they are. But, in more complicated problems involving
Distributive law: A(B f C ) = AB & AC
many variables, the matrices are of large dimensions, and their routine
Associative law: ABC = A(BC) = (AB)C multiplication, taking of inverses, and the like, become a formidable task.
Commutative law: A+B =B+A To alleviate this situation matrices are subdivided, or partitioned, into
AB # BA smaller matrices, and rules of manipulation are establishsd for dealing with
26 MATRICES, TENSORS, A N D CIRCUITS [Ch. 2 Art. 2-81 MATRICES 27
such compound matrices. For example, the matrix equation e =Z i may the order of multiplication of the components in proper sequence. For
be partitioned as follows. example,
a b c d e a b c d e P 4
Note that along the arrows the compound indices ( r , s) must be the same and
.
that R D cannot be substituted for D R in the final product.
;[ml;
2-8. Solution of Matrix I n the matrix representation of
a system of simultaneous equations, let the matrices be split in the following
-
fashion.
k 4 v
/-
in which ( p ,q ) are compound indices each of which embraces several individual 2"k Z", ZU"
axes of the original matrices. Thus p = a, b and q = c, d, e. The partitioning
is quite arbitrary and the resulting submatrices may be either square (Z1 or
Z,) or rectangular ( Z , and Z, or i, and i,).
The rules for manipulating compound matrices follow.
Hereby the set of equations
Transpose. If A, B, C, . . . are matrices of a compound matrix, the
transpose is
has been broken up into sets of equations :
+ +
ea = e, es e ,
iS = ik + +a + iv
where
+
e, = Zjkik Zjaia + Z,,iv
For example, the transpose of the previous impedance matrix is e, = Z,,ik +
Z,ia + Z,iv
+
e, = ZUkik Z,,ia + Z,,iv
in which (j,k = 1, . . . , n,), ( p , q = n, + 1 , . . . , n,), and ( u , v = n, + 1,
... , n,). Obviously, the splitting up into parts could be carried as far as
desired rather than for only three splits, as in the illustration above. Eqs. 2-28
may be solved for any set of currents, say iv from the last set:
i v = Yvue, - YZIUZUkik
- YvUZUaiq (2-29)
in which YvUis the inverse of Z,,. Then, inserting this value of iVin the first
two sets of Eqs. 2-28, there results
Products. The product of two compound matrices is taken in the same +
(e, - Zi,Y"ue,) = ( Z j k- Z j v Y " V u k ) i k ( Z j a- Z,,Y ""Z,,)ia
way as the product of simple matrices, particular care being taken t o keep ""Z,,)ik + (2, - Z,,Y
i (e, - Z,,Y ""e,) = (Z,, - Z,,Y ""Z,,)iq
28 MATRICES, TENSORS, A N D CIRCUITS [Ch. 2 MATRICES 29
These equations may be rewritten Eliminate i3 and reduce to two matrix equations. Then eliminatei2and reduce to
one matrix equation. Check this first result by Eq. 2-31 in the text by eliminat-
ej = Zjkik + Z~,iQ ing i2 and i3 in one step.
ei = Z&ik + Zkiq 2-3. Show that the inverse of a diagonal matrix is a matrix each of whose
elements is the reciprocal of the corresponding element in the original matrix.
Solving for iq from the last equation, Remember from this that, if a matrix can be reduced to diagonal form before an
inverse must be taken, the work is enormously simplified.
2-4. Consider the five simultaneous equations represented by the numerical
and substituting in the other equation, there results matrices given in Art. 2-7. ( a )Eliminate the c-, d-, and e-axes one at a time.
( b ) Eliminate the c-, d-, and e-axes in one step. Compare the amount of work
ey = (ej - Zj,Y'Qpe;) = (Zjk - Z ~ q Y ' a p Z ~ , =
) i kZykik (2-31)
involved in the two processes of elimination.
Thus the original set of equations (2-28) has been reduced to a single equation 2-5. Given the compound matrices
in terms of the current set ik, where
Then iQfollows from Eq. 2-30, and iv from Eq. 2-29. This procedure is the
matrix or tensor equivalent of the ordinary step-at-a-time solution of a set
of simultaneous equations.
If a set of equations is reduced one a t a time, there is no need to take the
inverses of matrices. Thus, if the last of Eqs. 2-28 represents a single equa- .
Find A . B, B A, A t . B, A t . A.
tion, then in Eq. 2-29 Y V U= l / Z u v since Zuv would then be a number and 2-6. Prove that multiplication of a matrix by a scalar is equivalent to multiply-
not a matrix. ing each of its elements by that scalar. (Use the rule for the addition of matrices.)
2-7. Prove that the transpose of a product of two matrices is equal to the
PROBLEMS product of the transposes of these matrices. Extend the proof to the case of n
matrices.
2-1. Given the three matrices 2-8. Prove that the commutation law holds for diagonal matrices.
a b c 2-9. Prove that the inverse of the transpose is equal to the transpose of the
inverse of a matrix.
2-10. Prove that the inverse of the product is equal to the product of the
i = b inverses of two matrices. Extend the proof to the case of n matrices.
Find :
Note carefully how these various operations check and cross-check one another.
2-2. Given the matrix equations (in which each member is a matrix)
Art. 3-21 TRANSFORMATION O F COORDINATES 31
A quadratic form can always be expressed in terms of the symmetric part e3. = Z 3k. ik = ZZjkC;,
yk'i'ej, (3-9)
(Eq. 2-22) of its 2-way matrix, for Thus all quantities have been found in a routine fashion by making use
of the concepts of "sets," "group of transformations," and "invariancy of
form," and by using the rules for matrix multiplication and for taking
inverses. The seven steps outlined above are always sufficient.
This particular relationship is frequently used in machine theory. Since the currents are transformed (Eq. 3-3) directly by C,k,,whereas the
I n a tensor equation certain forms remain invariant, that is, unchanged, voltages are transformed (Eq. 3-5) by the inverse tensor CT, currents are
after a transformation of coordinates. called contravariant vectors and are designated by upper indices, while
voltages are called covariant vectors and are designated by lower indices.
3-2. Transformati~n.~-~ The concepts employed and t h e steps in the
An impedance transforms directly as a doubly covariant tensor and therefore
formulation and solution of tensor equations are illustrated by the analysis
has two lower indices. An admittance transforms inversely as a doubly
for a mesh network :
contravariant tensor and therefore has two upper indices. I n general, a mixed
1. Set up the canonical equations or equation of performance (Eq. 1 4 ) tensor of any rank, having both upper and lower indices, transforms with
(a "set" of simultaneous equations). respect t o its upper indices as a multiple contravariant tensor, and with
e, = Zjkik respect to its lower indices as a multiple covariant tensor. It is t o be noted
30
32 MATRICES, TENSORS, A N D CIRCUITS [Ch. 3 Art. 3-21 TRANSFORMATION O F COORDINATES 33
that the transformation tensor CE. is itself a contravariant tensor in one of Now suppose
-- that the coils are connected as shown in Fig. - 3-lb, and let
its reference frames and a covariant tensor in its other reference frame. new currents ia and iP be selected arbitrarily. I n terms of these new currents
It is worth noting here that we did not employ Kirchhoff's laws explicitly. the old currents are
a B
Instead a n equation of transformation, or transformation tensor, was
established which related the old and the new currents. I n effect, this
corresponds t o an application of Kirchhoff's first law only. The second
Fig. 3-1.
i=c.i'=
-1
o=g -8
It will suffice at this point to give a simple example. I n Fig. 3-2 first let
currents ia, iBbe selected as the new currents so that
u B
e =Z.i = mj[=B
and the voltages across the coils are
-8 - 43
Thus all voltages and currents, in both the old and new reference frames,
Then suppose it is decided to use mesh currents iU,iv. The transformation
between the ( a , /3) and the (u, v) sets is
Fig. 3-4.
Now let these subnetworks be interconnected, having a totality of
coordinates
+ +
(a, B ) = an (j, k) all (p, q) all (u, v) + ..
.
+
(a', B') = all (j', kt) all (p', q') +
all (u', v') + - ..
Then the impedance matrix of the whole network before interconnection of
the subnetworks is The transformation matrix for the interconnection of the two subnetworks,
+ +
Z,.,? = ZjPk' Z,.q! Z,',' ... + in terms of new currents ia", ib", ig", is
= c;,Zjkc,",+ + +.
c;,z,,c,., C ~ ? Z , , C ~ ? . .
Next set up the transformation tensor C: representing the external inter-
connections. The impedance matrix of the entire network is
Za"f = C"Z,.~.C~":
= c$[z,.*'+ z,.,. + Z ,.,, + . - qcj: (3-11)
+ C$Z,C: + G'$Z,,C~, + . - .]G'g;
= c,~'[c,!z,~c$ (3-12)
38 MATRICES. TENSORS, A N D CIRCUITS [Ch. 3 Art. 3-51 TRANSFORMATION OF COORDINATES 39
and the new impedance becomes Now select new coordinates, but among them include all the active external
axes. The remaining new axes may be selected as convenient. I n terms of
these new axes set up a transformation matrix, remembering that the ( r , s)
axes are the same in the new as in the old reference frames; thus
and
and then
which may be condensed to A transformation matrix (Eq. 3-14), which preserves the (a, b) axes is
a' b' c'
Now, if all voltages except e,.. = ea are put equal to zero, that is, short-
circuited, and the iQ" currents are eliminated, there results
The new impedance matrix (Eq. 3-15) then is
a' b' c'
automatically brings about the birth of a covariant variable, and conversely. On the other hand, if the equations of transformation are given and i t is
Thus, when a current is reduced to zero by the opening of a branch, a voltage desired to change over to corresponding equations of constraint, i t is neces-
appears across the opening. Or, when a. velocity component in a given direc- sary only to rearrange the transformation tensor in the form
tion is reduced to zero by a barrier, a reaction force appears in that direction.
Kirchhoff's first law, stating that the sum of the currents a t a junction
must be zero, is a n equation of constraint, since it sets up a n enforced relation-
ship between the currents a t the junction. If there are J junctions in a
network, a total of (J- 1)non-redundant relationships between the currents
may be written, each of the form zik
= ia ib+ + - . = 0.
a
* This precludes 3-phase windings on different legs of s, shell type and 5-leg core-type
design.
44 MATRICES, TENSORS, A N D CIRCUITS [Ch. 3 Art. 3-71 TRANSFORMATION OF COORDINATES 45
Then id and ie, the remaining currents, are the independent variables and the made and, as might be expected, a rational approach will involve both of
transformation matrix for the currents is Kirchhoff's laws.
d' e' Consider only the network as actually connected and make no reference
to the primitive or any other network. Let
ei = actual voltage impressed in series with coil j
ik = actual coil currents
Fig. 3-9.
series impressed coil voltages e,. For this closed circuit Kirchhoff's second
law requires that C (impressed voltages) = X (consumed voltages) or
where the C includes all of the terms yielded by Eq. 3-31. But this is
identically the mesh voltage ej, of Eq. 3-31 and therefore, since in the sum-
mation around the mesh the ej occur in exactly the same way as the v,,
3-1. Verify all operations given in this chapter. Solve the circuit of Fig. 3-6
for all currents and voltages (both old and new) if e, = 4 and .eb = 3.
3-2. Eq. 3-4 does not follow from the previous equation simply by "canceling"
ij' on each side of the equation. Show in detail the proper steps and justify them.
3-3. If the two networks in Fig. 3-10 are connected, find the total impedance
Fig. 3-1 0.
4
SYMMETRICAL COMPONENTS
41. Conjugate Tensors. When the elements- of a matrix A consist of Or, expressed in terms of the impedance,
+
complex numbers such as (a jb), where j = 4 - 1 , the corLjugate of the
matrix is denoted by A* and is found by replacing each complex number
+
(a jb) in A by its conjugate (a - jb), that is, by substituting -j for j 4 2 . Transformation Formula. The transformation formulas for
throughout : currents, voltages, and impedances have t o be modified slightly when con-
jugates are being used. As before,
a+jb o+jd a-jb c-jd
in which
= (E14 + E212)+ j(E112- E A ) (4-5) + +
ia = (iO i1 i2)/ J3
ib= (iO+ a2i1+ ai2)/J3 (4-14)
P = EI cos (8 - 4) = real power (4-6)
Q = EI sin (8 - 4) = reactive power, defined as
+
ic= (iO ail + a2i2)/J3
positive for a leading current (4-7) in which ia, ib, iCare the phase currents; iO,il, i2 are the zero, positive, and
negative phase-sequence currents; and
A resistance power loss is defined as
50 MATRICES, TENSORS, A N D CIRCUITS [Ch. 4 Art. 4-31 SYMMETRICAL C O M P O N E N T S
i = .
C i' where
which gives the phase voltages e in terms of sequence currents it. Writing
in which S O , S1, and S2are the so-called zero, positive, and negative phase- this impedance as
sequence operators. This non-singular tensor, together with its conjugates, 0 1 2
inverses, and transposes, provides a group of transforn~ationswhich are (as
is easily verified)
Among these ten different tensors there are only two different matrices.
Currents and voltages transform by Eqs. 4-1 1 and 4-12 as
a b c
Z + aaZ' + aZ"
Z + aZ' + a2Z"
(4-24)
a b c Hence a grounding coil has only a zero sequence impedance, and its trans-
formation tensor is n
Other mixed networks involving both actual and sequence reference frames
in different combinations may be encountered or set up for special purposes;
for example, a network which will give phase voltages in terms of sequence
currents.
54 MATRICES, TENSORS, A N D CIRCUITS [Ch. 4 Art. 4-51 SYMMETRICAL COMPONENTS 55
Transformation tensors can be established between any two of the four Now, expressing the dependent currents iv"in terms of the independent
principal networks, and formulas developed for transforming from one net- currents in"= 8:: is-, there results
work to another in accordance with the following scheme (in which double-
headed arrows represent reversible transformations, and single-headed
arrows represent generally irreversible transformations). Finally, changing is"to is", the transformation tensor becomes
_
Primitive Primitive
Actual n
i+-+ C,"-ia Sequence ia" sy
(NetworJ " (NetworJ 1 (4428)
= } independent variables
+ ( Actual )
C;,iaP ia* C$, iajn
-(D:c;. - p . ) - l ( D : C ~- C:) } dependent variables
Network Xetwork
The tensor C,$ is the Fortescue symmetrical component transformation,
and i t expresses the relationship between the sequence currents in the primi-
tive sequence network and the phase currents in the primitive actual network,
or, inversely, by CE". This tensor is always known.
The tensor CE, is the ordinary transformation between the phase currents
in the primitive actual network and the actual network. If the actual net-
work has been established in terms of phase quantities, this tensor is easily
found. However, it will generally have a singular matrix and, therefore,
no inverse.
Usually, the sequence network is unknown, and tensor C$, the trans-
formation between the primitive sequence and actual sequence networks,
will have t o be found by a special process to be shown. I n general, it will
have a singular matrix and, therefore, no inverse.
The tensor C$ is the transformation between the actual and sequence
networks.
4-5. Changing C,"$$:"" to C ~ ~ ~If :the ~ currents
~ ~ . of. the actual
network are coil currents, then the transformation
in which the (r, s ) independent variables and the (u, v) dependent variables
may be replaced by an equation of constraint,
Fig. 4-2.
0 = B,U.ia' or 0 = B,Uia since S' = s (4-30)
where Ba is found from C:, by subtracting the Kronecker delta from the The procedure then is:
independent variables (in accordance with Eq. 3-26). Converting to sequence
currents, 1. Arrange C,$ as a compound tensor separating the independent (r, S)
0 = Bija = B:c,"-~~" = B:iaX = (~21;'" + &$'") (473l) and dependent (u, v) variables.
where 2 . Arrange CEn also as a compound tensor separating the independent
S v 8" 2)" 8" v" and dependent variables t o agree with the first step.
3. Find C$ from Eq. 4-34. The sequence network may be found there-
from by inspection.
4. Find em- and ZaYIfrom eaeand Za-,,".
56 MATRICES, TENSORS, A N D CIRCUITS [Ch. 4 Art. 4-51 SYMMETRICAL COMPONENTS 57
5. Use (7,"-= C;"C,"-to establish the sequence network from the primitive The corresponding compound tensor for the Fortescue transformation is
actual network e, and Zap.
6. Determine actual phase currents from i"' = c > ~ .
EXAMPLE. In Fig. 4-2 is shown a %phase grounded neutral generator
supplying an unbalanced load grounded through an impedance. The gener-
ator impedances are specified in terms of sequence quantities, but the load
and grounding impedances are given in terms of phase values. The given
data are
1; 2, + Z1 + 2 2 + 3 2 , 2, + + 32,
=c $ . z ~
zafmPm 8", =
~~"
2; Zo+ 2 + 32, 1 2, + 2, + 2 2 + 3 2 ,
The transformation tensor between the actual network and its primitive, The sequence network currents are then
and its rearrangement separating the independent and dependent variables is found in the usual manner by
The sequence voltages are 4-4. In Fig. 4 4 a generator with a neutral impedance supplies a load. The
impedances and voltages are
Calculate the sequence and actual short-circuit currents and the voltage differ-
ences across each phase of the generator and of the load for the following condi-
tions: (a)line-to-ground fault (switch S1 closed); (b) line-to-linefault (switch S4
which may be checked by ia = C:-C$iaW. closed); (c) double line-to-ground fault (switches S1 and Sp closed); (d) three-
The voltages across each coil are phase fault (switches S1, S2,S3 closed).
PROBLEMS
Fig. 4-4.
4-1. Write the conjugate of
4-5. Repeat the preceding problem when the load is a balanced delta-connected
impedance with z in each side of the delta.
z +jy 3-j4 cos 6 + j sin 6
It is evident from the construction of Eq. 5 4 that the voltage and current
terms involved are of the form
- - % No = (volts per turn)%
ea -
na Na
. N ia
nuta = Q- =
(ampere turns in winding)
MULTIWINDING TRANSFORMER No No
that is, it is as though both windings had No turns.
CIRCUITS21-4e Now consider a multiwinding transformer with (n 1)windings numbered +
from 0 to n inclusive, and let winding 0 be selected as the reference winding.
Let indices ( j , k) range from 1 to n inclusive. Then, ignoring the magnetizing
currents,
5-1. Basic D e f i n i t i o n ~ . ~ l -Multiwinding
~~ transformer circuits are
unique among static networks in that ignoring the magnetizing currents
+
O = i" nkik or i0= -nkz .k (5-5)
and
reduces the number of simultaneous equations by one for each complete
magnetic circuit, and this in turn leads to the concepts of "leakage imped-
+
e.3 = z.30 i0 zikik= (zjk- nkzj0)ik (5-6)
ances" and of voltage differences rather than absolute voltages in the eo = zooiO+ zokik= (zOk- nkzm)ik (5-7)
canonical equations. Subtracting n j times Eq. 5-7 from Eq. 5-6 and rearranging,
Consider two windings a and b linked by a common magnetic circuit, and
having turns Nu and Nb, respectively. Select a winding of No turns as the
"reference winding." Let
Nu turns of winding a
n =-= n, = Nb
-
a No turns of winding 0 No
zaa = Raa + jwLaa = self-impedance of winding a Comparing the terms in parentheses in Eq. 5-8 with the definition of a
leakage impedance as given by Eq. 5 4 , i t is seen that these terms are leakage
zab= zba= jwMab = mutual impedance between windings a and b
impedances, and thus Eq. 5-8 may be rewritten
Then, ignoring the magnetizing current, and writing the voltage equations
e, = za8iS (5-11)
in which
Multiplying Eq. 5-2 by (lln,) and Eq. 5-3 by (lln,), and subtracting, and
dividing by naia,
0
. 0
zoo
1
0 1 ! 0
2
I
...
...
I
n
20.
(5-1 2)
=
60
62 MATRICES, TENSORS, A N D CIRCUITS [Ch.5 Art. 5-21 MULTlWlNDlNG TRANSFORMER CIRCUITS
The magnetizing current is eliminated by the transformation The balance of ampere-turns requires that
Eqs. 5-15 and 5-16 agree with the f i s t line of Eq. 5-8. The advantage of
using the final specification of C: in Eq. 5-14 rather than the entire matrix
expression is evident.
- -
;a1I
:6'
I-
5-1
5-2. Windings in Series. I n Fig.
is shown a transformer having
primary windings numbered from 0 to
m, of total turns N in series carrying
where E = (eo
(em+,+ ..
.
+ el + + em) is the total primary voltage and E' =
+ en) is the total secondary voltage.
The impedance matrix before transformation is
= zjk f Z j s Zrk + + (5-19)
current iO,and secondary windings num-
and i t transforms t o
+
bered ( m 1) to n of total turns N'
z,.,. = C,",Cj.Z,, = C ; . C p j k + c ; * c ~ , Z +j s c;,c;,z,,+ c:,c:,zrs
I 1' '1 I - current in. Let indices
in series carrying
+ Z1, + . . + Zlrn + . . + Zmm)
.
(j,k) range from 1 t o m, and indices = (Z,,
(r, s) range from ( m 1) t o n. + - n(Z1(,+,)+ . . . + + Zmn)
21, '''
Select new currents ia'and ib' flowing
through the primary and secondary - n(Z(m+l)l+ + Z(rn+~)mf
' '' ''' Znm)
I I
I I
I
windings respectively, and put + n2(Z,m+l)(m+l)+ . . . + Z(rn+l)n+ . . . + Znn) (5-20)
I n the case of a single primary winding No and two secondary windings Nl
N=N0+N1+.-.+Nm and N , in series ( m = 0 , n = 2), the impedance reduces to
= total primary turns
N' = Nm+, + ..+N,
Za,B,= n2(Zll + 2Zl2 + Z2,)
N N' = total secondary turns
Fig. 5-1. n = NIN'
This formula is much used in transformer design for the calculation of
= turn ratio reactances.
64 MATRICES, TENSORS, A N D CIRCUITS [Ch. 5
Art. 5-41 M U L T l W l N D l N G TRANSFORMER CIRCUITS 65
5-3. Windings in Parallel. Pig. 5-2 shows two independent windings
For example, if there are four windings ( n = 3 ) and the turns of the second
0 and 1, and ( n - 1) windings 2 t o n in parallel with a common load imped-
and third windings are equal, n2 = n3, the equations become
k ) and ( a , p) both range from 1 t o n inclusive.
ance Z(,+l,(n+l,. Let indices (j,
Then the current in+' flowing in the load impedance Z(n+l,(n+l,is eliminated el - nleo = + Z12i2+ Z13i3
Zllil
by the connection tensor whose components are
-n2eo = Z2,i1 + ( Z 2 2+ Z ) i 2 + ( Z 2 3+ Z ) i 3
c; = 6; ct+l = (6: - g,"+ l ) (5422) -n2eo = Z3,i1 + ( Z S 2+ Z ) i 2 + ( Z 3 3+ Z ) i 3 (5-26)
where g;+l = 1 for all a. The "new" voltages are 5-4. Coupling Windings. Coupling windings are used to reduce
reactances and stray flux, and to transfer ampere-turns. A multiwinding
transformer with a number of coupling windings is shown in Fig. 5-3. It
The transformed impedance becomes
If the turns are equal for each pair of coupling windings, the only surviving
The new equation is then voltage term is
e, = Zapip
(el - nleo) (5-29)
The "new" impedance is
which can be solved for the current iB upon taking the inverse YPa of the
impedance
= YBae,
66 MATRICES, TENSORS, A N D CIRCUITS [Ch. 5 Art. 5-61 M U L T l W l N D l N G TRANSFORMER CIRCUITS
For example, a transformer with two pairs of coupling windings, and The admittance matrix is the inverse of Zapor
equal turns on each winding of a pair, would be represented by the equations
(541)
3 1 0 0 ~ 1 Fig. 5-5. Forke-d autotransformer.
68 MATRICES, TENSORS, A N D CIRCUITS [Ch. 5 Art. 5-81 M U L T l W l N D l N G TRANSFORMER CIRCUITS 69
in Fig. 5-7. There are four windings on each magnetic circuit, but each "elec-
trical" leg involves, through the zigzag, two "magnetic" phases (legs). As
Fig. 5-6. Group o f transformers. this is typical of all phase-shifting connections, some care is required in
setting up the transformation tensor. Let (1, a , a2) be the three roots of
transformer, 11, and a series transformer, 111, is shown in Fig. 5-6. The unity, so that
transformation tensor and primitive impedance matrix are i l = -i2" - -a2i2 or i 2 = - a i l = & 4
The transformation matrix and its conjugate for the "magnetic" phase,
therefore, are
3 4 3 4
*
and C; =
'm
70 MATRICES, TENSORS, A N D CIRCUITS [Ch. 5 Art. E-91 M U L T l W l N D l N G TRANSFORMER CIRCUITS 71
The new voltage is Let the windings be numbered as shown in Fig. 5-8. For simplicity
3 4 assume perfect interleaving of the windings so that there are only three
branch impedances :
2 3 4
*
Ch(e, - njeo) =
3
1 el -lnleo 1 e2 - n2eo 1 e3 - n3eo 1 el - a2e2
3 4
-
- e3 - n3e0 (nl - a2n2)e, (548)
The primitive impedance is I n terms of these impedances the primitive impedance matrix is
1 2 3 4
(5-59)
Hence
since nl = n2= n, = n4 = n, n5 = n , = n , = n8 = n', and from Fig. 5-8
it is clear that el - ae, = 0, e2 - a2e8 = 0, e4 - ae, = 0, e3 - a2e5 = 0.
The new impedance matrix is
(5-61)
The ratio of currents is
and since the new voltages are el = e4 and e2 = e,, while in the impedance
matrix the coefficients of i1and i2are the same, respectively, as those for i4
i1-
- + +
zn' - z"(an' n) z'na = AEje
and i3,only axes 1 and 2 need be retained, so that i2 +
-znr + z"a2(an n') - z'na2
Thus il and i2 are of unequal magnitude and out of phase by a n angle 0 if
both resistance and reactance are present. If the resistances are zero, then
A = 1, but 6 may have any value.
It is interesting to check the preceding equations by the use of compound
matrices. Let the various matrices be written in terms of the following
indices: (j,k) = 1 , . . . , 8; (r, s) = 1 , . . . , 4; ( p , q) = 5 , . . . , 8. Then
where the 2 is required to compensate for the ignored currents i3 and i4.
74 MATRICES, TENSORS, A N D CIRCUITS [Ch. 5
and, carrying out these operations, Zapis confirmed. Similarly, for the
voltage,
PROBLEMS
MULTIVELOCITY TRAVELING WAVES22945.47
5-1. Verify and carry out in detail the calculations for each case given in this
chapter. In this regard it is highly instructive to show the various transformation
tensors, such as Eqs. 5-22 and 5-27, in matrix form.
5-2. Set up the transformation n~atricesand the performance equations for 61. Differential Equations of a Line. A multiconductor transmission
the stub delta connection of Fig. 5-9. system having n incoming lines and n outgoing lines connected through a
transition network is shown in Fig. 6-1. Incident voltage and current waves
(e,, i') approach the transition point and give rise to reflected waves (ei, i'3
and transmitted waves (ef, inT).It is required to set up the equations for
these waves.
5-3. Set up the transformation matrices and the performance equations for
the extended delta connection of Fig. 5-10.
5-4. Set up the transformation matrices and the performance equations for
the inscribed delta connection of Fig. 5-11. where KT8is the inverse of the matrix psT.These coefficients are calculated
for parallel cylindrical conductors in the presence of ground by including the
images of the conductors in the ground surface.
The magnetic flux linkages are given in terms of the inductance coefficients
LTS by
q$ = LTsiS (6-3)
1 The currents flowing to ground and between conductors, including the effects
of both leakage and corona, are given by Grse,. And, h a l l y , the resistance
Fig. 5-1 1. Inscribed delta. drops in the conductors due to the currents are RTsiS.
I 75
76 MATRICES, TENSORS, A N D CIRCUITS [Ch.6 Art. 6-21 MULTIVELOCITY TRAVELING WAVES 77
Hence the differential equations for the multiconductor transmission The complete solution then becomes
system become (putting p = a/at in the Heaviside sense)
in which
Yra = KrSa;vca,
Eqs. 6-13 and 6-14 show that there are on each of the n conductors n pairs
Eliminating iq, there results of forward- and backward-traveling waves and each of these has its own
particular velocity v(,, .
I n the event that the zero-potential plane for both the voltage and
current images is the ground surface, and, if there is no corona, all the L,,
in which 6: is the Kronecker delta and and KTSare related so that in Eq. 6-11
Then First the impedance matrix for the transition network and the outgoing
lines is set up :
+ z,,)
ZaB= ( Z j k = impedance of network and outgoing lines before
i k u '19
interconnection
Za,B,= C$C$,ZaB = impedance after interconnection of the network
and outgoing lines
Now ZaTB, may include branches other than those connected to the incoming
lines. The open-circuit branches will have been eliminated by Ci,, but the
grounded branches other than those connected to the incoming lines will have
to be eliminated by the substitutions
S !l
agp
so that
ir + i f ,= (6-24) Then the new impedance matrix is
The total voltage E , and the total current I+ follow from Eq. 6-23, the
remaining network currents from Eq. 6-21, the network voltages from Eq. '
6-20, and so on; thus all system quantities may be determined.
Za,s. = Cg,ZasC$,= ( b'
zva=;;cl
Since the equations are becoming unwieldy in algebraic form, i t will serve
afml
the matrix as specified by Eq. 6-19 and then using Eq. 6-22. Splitting our purpose just as well t o pass to numerical values for illustrating the
according t o Eq. 6-19, procedure. Let
27, = b'
E, B =(R1-B m ) d
and Eq. 6-22 becomes
R, + 222
yap = -1
500
The total voltages on the incoming side of the transition point are
7
JUNCTION-PAIR NETWORKS6
Hence the currents are
Stipulating the invariancy of power between the primitive and new net-
works, there is
P = E,IU = E,,IU'
Substituting Eq. 7-2 in Eq. 7-3,
C : ' E , , I ~= E,P'
from which
I,' = c;'p
Ed'
Substifuting Eqs. 7-1 and 7-2 in succession in Eq. 7-4,
( a ) Actual network ( b ) Coil voltages ( c ) Junction-pair voltages
I,' = c;'p " ~ c;'y U v c ; ' ~ ,=, ~ " ' v ' E , ,
=,
in which Fig. 7-2.
p ' v ' c;'y,vc;'
junctions, as shown in Fig. 7-2c. A direct comparison between the coil
is the transformed admittance. voltages (Fig. 7-2b) and the junction-pair voltages (Fig. 7-2c) defines the
Eq. 7-5 may be solved for the "new" voltages in terms of the junction-pair transformation tensor.
currents by taking the inverse of the new admittance tensor:
E,, = ( p ' " ' ) - q * ' = ,,,,I"'
The voltages across the individual coils follow from Eq. 7-2, and the currents
flowing in these coils from Eq. 7-1. Thus all currents and voltages may be
determined in routine fashion. E, = Ed' - E,? - E,,' .
7-1. The Selection of Junction Pairs. I n the analysis of a junction-pair Ed = - E d' d
network of J junctions i t is mandatory t o select a minimum of (J - 1) E, = E,, + E,? e
independent junction pairs, even though currents may be impressed acros E, = E,. - Ed.
a fewer number of junction pairs, because unknown response voltages occ f 1~ -1
across all junction pairs and must be recognized in setting up the trans
formation tensor. This is the counterpart of selecting as many independen Let the primitive coil admittance tensor be
meshes as there are unknown currents in a mesh network.
After the (J - 1) junction pairs have been selected, the transformatio
tensor is established by assuming new voltages E,, for these junction
and then equating the old voltages E , across each coil t o the sum of jun
pair voltages around closed circuits for which the coil junctions are termin
Alternatively, equations of constraint may be set up for short-cir
junction pairs in terms of the new junction-pair voltages. The proc
will be clarified in the examples to follow.
EXAMPLE.^ I n Fig. 7-2 is shown a network having six coils and fo
junctions, and therefore (J - 1) = 3 junction pairs. Currents Id and I f ar
impressed across two of the coils.
86 MATRICES, TENSORS, A N D CIRCUITS [Ch. 7 Art. 7-11 JUNCTION-PAIR NETWORKS
87
The new admittance matrix is furnish two equations of constraint, permitting En. and E,. to be eliminated;
thus
YU'V' = c:' Y U V ~ " '
0 En,= 0
-E,. + + +
Ea, E,, Em. = 0 :.
E,. = E,'+ Er# Em, +
When these values of I"' are inserted on the diagram of connections there Expressing the "old" coil voltages in terms of the remaining variables
results the distribution shown in Fig. 7-3, from which i t is seen that the Egt,Erj,Em,, there results
original currents Id and P are correctly q' r' m'
given by the superposition of the new
junction-pair currents. If the inverse of
t,he admittance tensor i s
Es = Ear + Em*
then the junction-pair voltages follow E , = E,, = Ea, + Em, + E,.
from Eq. 7-7, the coil voltages from
Eq. 7-2, and the coil currents from But the same tensor may be written directly from Fig. 7 4 by observing
Eq. 7-1. from the connection diagram that En. = 0 and - E , E,.+ +
Em, E , = +
EXAMPLE OF CONSTRAINTS.~A net- 0 ; hence
Fig. 7-3. Superposition of junction-
pair voltages. work in which certain junction pairs ED = Em#
are short-circuited is shown in Fig. E , = E,. + Em, + E,.
7-4a. The voltages across individual coils are shown in Fig. 7 4 b . Since, Let the primitive admittance tensor be
there are four junctions, there are three junction pairs, and these may be
selected arbitrarily so as to include each junction a t least once, as shown
in Fig. 7 4 c , where new junction-pair voltages Em,,E,,, and E,, are shown.
There are two ways in which the transformation tensor may be set up:
either by introducing equations of constraint specifying the short circuits, or
by writing the tensor from inspection of the diagram of connections.
Using first the method of constraints, let as many new junction-pair
voltages E,., E,,, E,., Em,, En, be selected as there are coils. I n terms of
these voltages, as seen from Fig. 7-4c, the short-circuited junction pairs
MATRICES. TENSORS, A N D CIRCUITS [Ch. 7 Art. 7-31 JUNCTION-PAIR NETWORKS 89
The impressed currents are I g and I t , and therefore the new junction-pair I,' = 1;' + I,"' + 1;' + . . .
currents are Each tensor is assumed to poseess all the
a' r' m' nates for the combined network but has zero ele-
ments except for the specific coordinates of its own Eel
subnetwork. Then a transformation tensor is set Fig. 7-5. Substitution
up to express the interconnection, and the com- of variables.
0---
bined admittance is calculated from
7-2. Change of Variables.6 After the equationsof performance for a net-
..
work h n v- ~heen established
...-
- - in terms of a particular set of junction-pair EXAMPLE.^ Consider the interconnection of the networks of Fig. 7-2 and
coordinates, it may be desirable for some reason or other to restate them in - 7 4 , as shown in Fig.
Fig. - 7-6. The compound admittance is the sum of the
terms of a different set of junction pairs. Of course, there must be as many subnetwork admittances, as previously found.
junction pairs in the new set as there were in the old, and each set of junction
pairs must include each junction in the network a t least once. The procedure
in setting up the appropriate transformation tensor to change from one set of
variables to another consists simply in writing the equations which express
the old junction-pair voltages in terms of the new:
E," = C$Eut
For example, in Fig. 7-5 are shown two sets of junction-pair voltages,
(Ea,,Eb,, Ec,, Ed,, Ee,) and (Ea-,E,", R,, Ed.., E,"), for a network with six
junctions. It is clear from the diagram that
a" b" c" d" e"
!
E,, = Ea- - E,"
--- 4
e'
Ee' -- - E d - - E,"
7-3. Interconnection of Networks.6 When two or more networks a d
t o be interconnected t o form a combined network, two procedures are
possible.
1. The primitive admittance and current tensors for the individual coib
of the entire combined network may be set up, the junction pairs selected,
a transformation tensor established, and the analysis completed. This
JUNCTION-PAIR NETWORKS 91
90 MATRICES, TENSORS, A N D CIRCUITS [Ch. 7
The current tensor for the combined network is applied between junctions 1 and 2 and between junctions 3 and 5. Make a
complete junction-pair analysis of this network, finding all response voltages.
7-6. Solve the network of Fig. 7-7 and indicate on the diagram all coil voltages
and currents. The mutual admittances are all zero.
Fig. 7-7.
I I
Fig. 7-6. Interconnection of t w o subnetworks.
PROBLEMS
7-1. The use of the principle of the invariancy of power under a transforma-
tion of coordinates, or for a change in the connections of a network, was justified
for a mesh network in Art. 3-7. In a similar fashion prove that the use of the
principle is justified for a junction-pair network.
7-2. Superimpose the new current IU' on Fig. 7-4 and show that they combine
to give the actual currents at the original junctions.
7-3. Confirm all calculations indicated in the chapter.
7-4. Prepare a tabular form of three columns to illustrate the correspondence
(or duality) between the analysis of mesh and junction-pair networks. In the
first column list the items: impressed variables, response variables, circuit
parameters, equation of performance, equation of transformation and trans-
formation matrix, invariancy of power, transformation of the response quantities,
invariancy of form, transformation of the circuit parameters, new response
variables, old response variables, old coil variables. I n the second column show
the corresponding steps in the analysis of mesh networks. In the third column
show the corresponding steps in the analysis of junction-pair networks.
7-5. Consider a network in the form of a hexagon with six junctions having a
coil between each pair of junctions (15 in all). Let junction-pair currents be
Art. 8-21 COMPLETE NETWORKS 93
In pure mesh networks there are n known impressed series coil voltages
e,, the shunt currents Iuare zero, the junction-pair voltages Eu, are of no
concern, and the m mesh currents ik' are related to the coil currents ikby a Fig. 8-1. Replacement of open and closed meshes by apparent coils.
singular transformation tensor. I t is sufficient to solve for m unknown mesh
currents ik'in terms of m impressed mesh voltages ek, from the equation of
voltage Ea is short-circuited, that voltage is reduced to zero and in its place
performance,
appears a mesh current ia. Thus the total number of variables is not changed
e.,
3
= z.,
3 k
,ik' by the imposition of constraints; there is only a substitution of one type of
A voltage in an impedanceless branch (z = 0) is treated as a voltage in series variable for another. If, however, one type of variable assumes a known
with a zero impedance coil, and each such voltage adds anadditionalreference value along an axis, there will appear in that same axis the dual type of
axis, or mesh, to the network. variable as a n unknown quantity.
I n pure junction-pair networks there are n known impressed shunt coil An open mesh across which appears an unknown voltage Ea may be
currents I U ,the series voltages ek are zero, the mesh currents ik'are of no. regarded as closed through an impedanceless apparent coil through which
concern, and the (n - m ) junction-pair voltages E,. are related to the coil flows a zero current ia. Similarly, a closed mesh through which flows a n
shunt voltages Eu by a singular transformation tensor. It is sufficient to unknown current ia may be regarded as opened across an admittanceless
solve for (n - m ) unknown junction-pair voltages E,, in terms of (n - m) apparent coil in shunt with which is a zero voltage Ea. These possibilities
impressed junction-pair currents I,' from the equation of performance, are indicated in Fig. 8-1.
impressed shunt current Ikand a voltage E,. Such a general element is shown 1. Close all ( n - m) junction pairs through impedanceless (z = 0)
in Fig. 8-2a. It is to be noted that the current through the terminals is apparent coils. The (n - m) arbitrarily selected junction pairs must include
all admittanceless branches across which known currents are impressed and
also all impedanceless branches having known impressed voltages; that is,
the selected junction pairs in the network must include all the apparent coils.
From the point of view of the remainder of the network, this is the current In order to avoid the inclusion of apparent coils in the primitive impedance,
contributed by the series-shunt coil element, and it is the current which is
all known impressed quantities associated with these apparent coils must be
to be transformed. Evidently, for each such local circuit, considered known response quantities ivand E,. The following nomenclature
+
ej E j = zjkifc,= zjk(ik Ik) + (8-2)
is adhered to:
2. Terminals between which there is an admittanceless branch (that is,
known voltages in series with coils are denoted by ek
no coil but an open circuit), and across which there is impressed a known
junction-pair current iv and a response voltage E, as shown in Fig. 8-2b. known voltages in series with impedanceless branches
(apparent coils, z = 0) are denoted by Eu
known currents impressed across real coils, and not assumed
as junction-pair currents, are denoted by Ik
4. Set up the transformation tensor in the usual fashion, selecting arbi- 10. The total currents in the n new axes are
trarily n new currents represented by coordinates (j', k') but including the
currents in the added meshes. It takes the following form :
ik' Ik' + (8-11)
11. The equation of performance in the new coordinates is (by Eqs. 8-7
\ at kg and 8-9 and expression 8-1 1)
eat + Eat= ~ , , ~ . (+i ~I B ' ) (s12)
I n this equation there is a total of n unknowns. I n order to solve for the
When the junction-pair axes are not required, the corresponding k' columns
may be ignored and Ci, reduces to the singular transformation matrix of
an ordinary mesh network. I n this compound matrix, the n original branch
currents a s given by Eq. 8-1 are related to the n new currents by
I n Fig. 8-3b, the ( n - nz = 3 ) junction pairs p', q', r' have been chosen Note that when the junction-pair axes (p', q', r') are ncit active, the corre-
to include the admittanceless branch q and the impedanceless branch r, and sponding columns of C;. are eliminated, and the remaining singular part of
these junction pairs are shown as closed through impedanceless apparent the transformation matrix C;, is identical with that which would be set up
coils so as to convert the original network to an "all-mesh'' network of five for a simple mesh network of two meshes (a', b').
meshes and mesh currents (a', b', p', q', r'). Between these five new mesh The junction-pair currents transform into themselves :
currents and the primitive terminal currents a non-singular transformation
can be established. There are then the following matrices.
ek = I I e, eb 1 0
-
I I I
0 ef or n = 5 known series impressed voltages
a b c (1 f
0 I0 0 or n = 5 known shunt impressed currents
The new impedance matrix is
known
if or n = 5 new mesh currents, of which
two are known (p', q') and three
unknown (a', b', r')
Finally, there are the voltage equations along the mesh (af and b') and Rearranging the equations so that only one reference axis appears on the
junction-pair (p', q', r f ) axes: left side defines the following transformation matrix :
+ 0 = z,,,,ia' + z,.,,ib' + zatQ.(iQ'+ IQ')+ ~ ~ . ~ ,+( iI?')
e,. "
+ 0 = zb.,.ia' + zb.,ib' + zbrqr(iq'+ Ia')+ zbJir' + IT')
e,,
e,, + E,. = z,,,,za ' + ~ , , ~+, ~
' i~, . '~ .+ ~ ' + z,,,,(ir' + I")
( ila')
eq. + E,. = zQparia' + zQ,,ib' + + IQ') + zQ,,.,(i7'+ IT')
e,, + E,, = z,,,,ia' + zrTb,ib'+ z,,Jia' + IQ') + zr,,,(ir' + I")
In these five equations containing 14 variables there are
9 knowns (eat,e,,, E,,, iQ',lQ',I"", e,,, eQt,eT.)
5 unknowns (E,,, E,,, in', ib',iT')
and therefore there is a sufficient number of equations t o solve for the
unknowns.
8-4. The Conversion of Networks6 Any n-coil network may be used
as a reference frame for the same n coils interconnected into a different
network. The procedure follows.
1. Convert each network to an "all-mesh" network by closing its junction-
pairs through apparent coils. The mesh currents of the reference network
are designated by ik', and those of the new network by ik".
1 2. Express each coil current in terms of the mesh currents in both the
reference network and the new network:
ik= c,k,ik' = ~ , k , , j k " (8-14)
I
3. Rearrange these equations so that only one reference current appears
I
( a Reference network ( b ) Derived network
on the left side of each equation (thereby defining the required transformn tion
tensor) Fig. 8-4. Conversion of "all-mesh" networks.
II i k f -- c k r c k jk" = c k T j l c "
I
k k" P" (8-15)
Thus all n-coil networks made up of the same impedances are related The new impedance and impressed voltage matrices then are
through appropriate transformation tensors. Therefore, when the equations
of performance are known in one reference frame, they may be converted t o
any other reference frame within the group.
EXAMFLE.~ The following example, adapted from Kron's Tensor Analysis
of Networks, illustrates this procedure. I n Fig. 8 4 , five coils are shown
interconnected in two different ways, having three and two junction pairs, '
respectively.
Equating the coil currents in terms of the currents in both the reference
and derived networks, there results
p = ia' - ia" ib" - i d - if"
ib = i n ' + i ~+' i ~+' jr' - - ic"
ic = i b' - p' -
- ia"
id = - ib' -
- ib" + ic" + id'' These matrices could have been found also by setting up the transforma-
if = i a ' + ib' + i"' - - ib" tion tensor Ci. between the new network and the primitive coil currents.
Art. 841 COMPLETE NETWORKS
.m
'4
II IIII II IIII II I1 I1 11
104 MATRICES, TENSORS, A N D CIRCUITS [Ch. 8
The procedure in setting up this equation parallels that for an "all-mesh" THE ELEMENTS OF TENSOR A N A L Y S I P 4
network upon the interchange of dual quantities and the opening of each
mesh for the insertion of an apparent ( Y = 0) coil.
8-6. Comparison of the Three Types of Networks. Every network
comprises n coils, m meshes, and (n - m) junction pairs and has (n) known The previous chapters have been devoted to underlying principles in the
impressed voltages ek in series with the coils, (n) known impressed currents application of matrices and tensors to electric circuit analysis, but actually
Ikin shunt with the coils, and a number of apparent (zero or infinite imped- only the barest notions of tensor algebra have been introduced. However,
ance) coils. I n general (m) response currents ik' flow in the meshes and electric machine theory will require the most advanced concepts of tensor
(n - m) response voltages E,. appear across the junction pairs. Then analysis, and, while many of these concepts can be examined as they occur
in the development of machine theory, i t is, nevertheless, desirable to present
1. If m < (n - m ) , I u = 0, and Eut is of no concern, t,he network is now the elements of tensor analysis as briefly as is consistent with an under-
preferably solved as a simple mesh network. standing and appreciation of them.
2. If m > (n - m), ek = 0, and ik' is of no concern, the network is It has been mentioned that the essence of tensor analysis is bound up in
preferably solved as a simple junction-pair network. the ideas of: set, form, invariancy, group, and traqformation. And there are,
3. If both e, and Iu are impressed, the network may be solved either as of course, the associated notations, conventions, definitions, and rules.
an "all-mesh" or an "all junction-pair" complete (orthogonal) network. 9-1. Sets. The totality of the variables required for the description of a
geometrical or physical system comprises a set and in tensor notation is
A comparison of the several steps involved is given in Table 8-1.
represented by a single symbol:
PROBLEMS
xr E?E (x", xb, . . . , xn) (y-1)
8-1. The use of the principle of the invariancy of power under a transforma-
tion of coordinates, or for a change in the connections of a network, was justified Here x is called the base variable and refers to the type of variable. For
for a mesh network in Art. 3-7. Justification for the use of the principle for example, geometric coordinates, currents, voltages, velocities, and accelera-
junction-pair networks was assigned as Problem 7-1. What is the justification tions are all different types of variables and need to be designated by different
for the use of the principle in the case of complete networks? base letters such as x, i, e, v, a.
8-2. A thorough understanding of the underlying concepts and philosophy The superscript index (not exponent) r may be interpreted either as indi-
of complete networks is absolutely necessary to the analysis of such networks. cating any (but not a particular) one of the several variables xa, xb, etc., or
I t is believed that these matters have been put on a sound basis in the text and alternatively (and this is the preferred point of view) as the whole set of values.
that by Eqs. 8-13 all the essential ideas have been presented in a logical order. When it becomes desirable, or necessary, t o specify a particular variable,
On a large piece of paper show a closed contour within which are representative
a capital letter is used for the index, for example xR, or else a specific ele-
types of adjacent-terminal pairs as shown in Fig. 8-2. Alongside the diagram
write the definitions and procedures to be followed. And, finally, show the ment like xbis written.
impedance and transformation matrices, and equations. This "bird's-eye view" Indices may appear either as superscripts, as in x" or as subscripts, as
is complete when it clearly defines and explains every necessary concept of in y,, depending on the nature of the variables. Variables which transform
complete networks. in a direct fashion are called wntravariant variables, and their indices are
8-3. Check every step in the numerical examples in the text. attached as superscripts. Variables which transform in a n inverse fashion
84. Reduce a complete network to an "all junction-pair" network and are called covariant, and their indices are attached as subscripts. I n a given
establish Eq. 8-16 as its equation of performance. system one type of variable may be selected arbitrarily either as contravariant
or covariant .
105
106 MATRICES, TENSORS, A N D CIRCUITS [Ch. 9 Art. 9 4 THE ELEMENTS O F TENSOR ANALYSIS 107
The idea of set is not restricted to single variables like xT or y,, but can be When several pairs of dummy indices occur in a product, summation is
extended t o tensors of any rank or valence; see Table 9-1. implied with respect t o each pair of dummy indices. Examples are
TABLE 9-1
I
Valence (or Rank)
Tensors
--
Contravariant
I Covariant Total
1 Current, velocity,
coordinate
1 Voltage, force, gradient
2 Impedance
2 Admittance
2 Rotation
2 Kronecker delta
2 Transformation tensor
It is evident that the letter used to represent the repeated or dummy index
3 Affine connection
is immaterial. I n practice, in order to assemble terms, i t is often necessary
4 Riemann curvature
t o substitute letters for the dummy indices occurring in different terms.
For example,
A tensor of valence 2, such as a,, where ( r , s ) range from 1t o n , constitutes
a set of n2 elements. A tensor of valence 3, such as A:4, constitutes a set of 9-3. Invariance. Invariance refers to the constancy of some entity, or
n3 elements. form, which does not change under a transformation of coordinates. There
Furthermore, functions of variables, or equations, comprise sets. For are two types of invariancy of special importance in tensor analysis :
example (using the summation convention),
1 . An entity, like energy W or power P, or the distance d between two points,
which remains unchanged under all permissible transformations of
coordinates.
2 . A mathematical f o r m , like the formula for stored magnetic energy (a
is a linear set of n terms, while quadratic form) L T S i T i s ,or the equation of voltage (a linear form) e, =
zTSiS,which preserves the same form in all reference frames.
It often happens that the tensor equation representing the most general
case will contain terms not present in simpler reference frames. The invariant
form is thus considered to be that corresponding to the most general case,
but some of its terms are zero or vanish for certain reference frames.
9-4. Group Theory. A set of elements, A , B, C, . . . . having a rule of
.
combination (denoted by , but not restricted to multiplication) is defined
as a group if the following four conditions apply:
is a bilinear set of n2 terms. 1. The combinations A . A and A . B are also members (the "group
9-2. The Summation Convention. The summation convention has property").
already been defined and used in previous chapters. Whenever in a tensor 2. The associative l a w holds: ( A . B ) . C = A . ( B . C ) .
product the same index occurs twice--once as a contravariant and once as a 3. There exists a u n i t element I such that combination between it and any
covariant index-summation with respect thereto is implied. Such an index member of the group leaves the member unchanged.
4. Every member has an inverse A-l, also a member of the group, such that
is called a dummy index. An index which is not repeatedis called a free index.
A . A-1 = I .
108 MATRICES, TENSORS, A N D CIRCUITS [Ch. 9 Art. 9-71 THE ELEMENTS OF TENSOR ANALYSIS 109
If only the f i s t three conditions are satisfied, the set is called a semigroup. There are two possible interpretations of a transformation of this type:
If all elements of the group belong to a second group, the set is called a
1. The passive interpretation regards an identical physical configuration
subgroup of the second group.
of a n entity alternatively represented in two different reference frames. It
If a group repeats itself, it is called a cyclic group.
is simply a case of expressing the same entity in two related systems of
If a group has a finite number of members, it is called a Jinite group; if i t
coordinates by a substitution of variables. (For example, the temperature
has an infinite number of members, it is called an inJinite group.
distribution in a flat plate may be expressed in terms of rectangular or polar
For examples see Table 9-2.
TABLE 9-2 coordinates.)
2 . The active interpretation regards two different physical configurations
Group All positive and negative integers lotation by multiples of 90' of a n entity as expressed in the same reference frame. (For example, the
Set . . . , - 2 , - 1 , 0 , 1 , 2 , 3 ,... y, = 90°, y, = 180°, temperature distribution on two identical flat plates may be linearly related
y3 = 270°, y, = 360" in terms of the same coordinate system.)
Rule of combination Addition (+) Addition (+)
Unit element 0 y, = 360" The passive interpretation is used in cases where a mere substitution of
Inverse (2) + (-4 = 0 + +
Yl Y3 = Yz Yz = Ya variables occurs, such as substitution of mesh currents for branch currents
Condition 1 (2) + (&Y) = (2 * Y) +
Y I 7, = y3, etc. or of one set of branch currents for another, a permutation of variables, or
Condition 2 ( x + y ) + z = X + ( ~ + d (71+ Y,) + Y, = yl a n imposition of con~t~raints.
+
(Y,+ Y,) The active interpretation is adopted when a given network is to be recon-
Condition 3 x + O = x +
Y E y4 = YZ.etc. nected into an entirely different network.
Condition 4 (2) + (-4 = 0 +
Y1 Y3 = Yz + Yz = Ya The transformation (Eq. 9 4 ) is reversible if the equations can be solved
for the independent variables in terms of the dependent variables.
The rotation group is cyclic if it is agreed that any angle plus 360' cannot Solving Eq. 9 4 by Cramer's rule,
be distinguished from the angle. (This could not be tolerated in some
investigations; for example, in determining the work done in taking e
magnetic pole around a current many times.) The 90" rotation group is also
a subgroup of a rotation group having any angle 8 as its modulus. in which ICfI is the determinant of the transformation matrix and A: is the
The transformation matrices which substitute one set of variables for cofactor of C: in the determinant. Obviously, this solution does not exist if
another set in a network form a Jinite group, since only a finite number of the determinant vanishes, that is, if its matrix is singular. I n other words,
substitutions is possible (i.e., replacing branch currents with mesh currents). a singular matrix has no inverse and the transformation is irreversible.
The transformation matrices which change one network into another also 9-6. Successive Transformation and Group Property. A set of
form a jinite group (i.e., reconnecting the elements of a network). variables x7 may be transformed to x7', and this in turn to xr", and so on, thus
The transformation matrices which change one network into another
with a different number of coils form a semigroup, since such matrices do not
have an inverse (i.e., reconnecting with a different number of coils).
The transformation matrices which introduce hypothetical currents as
variables form a n inJinite group (i.e., substitution of symmetrical components).
The transformation matrices which substitute variables are a subgroup
of the transformation matrices changing the interconnections of a network, etc. (9-6)
which in turn are a subgroup of the group of connection matrices for the The successive products of transformation matrices possess the group
generalized machine. property.
9-5. Transformation. A set of coordinates xSrepresenting a description 9-7. Definition of a Tensor. A set of numbers or functions which
of some configuration may be transformed into another set xa by a linear remain invariant under a group of linear transformations is called a tensor.
transformation, If the CL are independent of the variables, then
xa = c;xs (94)
in which the coefficients Cf are independent of the variables.
l I0 MATRICES, TENSORS, A N D CIRCUITS [Ch. 9 Art. 9-1 I] T H E ELEMENTS OF TENSOR ANALYSIS I II
and therefore the definition of a tensor requires that But e, and i T cannot be added because they have base letters of different
types (a voltage and a current) and one has a covariant and the other a
contravariant index.
The sum of several tensors transforms in the same way as its constituent
9-8. lnvariancy of Form. Suppose that . the variables transform members :
linearly according to the equation aaB baB = C ~ C B ~ T S C ~ C B ~ T S
7 s 7 s
xT = c;xa (9-9)
= C:Cf(ar" fT" = c y ! d r s = dab (9-19)
What is the law of transformation for the coefficients in the equation
9-10. Contraction. If a pair of contravariant and covariant free
a,xT = b (9-10) indices in a tensor are made identical, they become dummy indices implying
if its form is to remain invariant? Substituting Eq. 9-9 in Eq. 9-10, summation, and the operation is called contraction. For example,
That is,
ar(Cixa)= (aTC;)xa= aaxa = b (9-1 1) st ." a~nt"'- rlt-'
(amno... ) s = n = r n n p . . . - a m l p+ r2t".
am + ~ 3 t . . .
Z p... a m 3 p... +
. . = bms.
r t...
.. (9-20)
Contraction is seen t o reduce the valence of a tensor by 2, since the sum of the
aa = arc: or a, =C;a, (9-12)
terms is a tensor with 2 less indices than the original tensor.
Thus the coefficients transform in a n inverse fashion from the variables, and
this leads to the following distinctions : contravariant tensors (like xT)trans- 9-1 1. Multiplication. I n multiplication of a tensor by a constant, each
form directly according to Eq. 9-9; covariant tensors (like a,) transform element of the tensor is multiplied by this constant (just as in multiplication
inversely according to Eq. 9-12; hence the use of upper indices for contra- of a matrix by a constant), and the resulting tensor has the same valence and
variant and lower indices for covariant tensors. is the same type as the original tensor:
Quadratic forms transform as follows : kars = bn (9-21)
I n the general (or outer) product of two tensors, each element of one tensor
b 7 s =
~ b~T S (~C ; ~~a ) ( C ! ~ p ) = ( b T S c ~ C ! ) ~ a ~ / 3 = b a B ~ a ~ , 9 (9-14) is multiplied in turn by each element of the other tensor (as in the product of
two matrices), yielding a tensor of valence equal to the sum of t,he valences
A bilinear form transforms as follows : of the factor tensors :
a y y , = a : ( C y ) ( C ! y B )= ( a : c ~ c f ) x a y=
B a!xayB (9-15) arb: = (9-22)
I n general, any multilinear form transforms as follows: Transforming the left side of Eq. 9-22 with form invariant,
1 :: xrxsxt . . . ymyny, . - = am"""'
rst-.
CmaC n~ C Y , - - - C T C ~ t x P z u x r y a y .p .y .y
P a r a;b,P = C " , ~ a ~ C ~ C
=~Ca
bm~
~CnGPCSdmr
r u n s = diP, (9-23)
that is showing that the indices in the product are the same and in the same position
BY. .. = Ca CBCY . . . C T C S C.~. . amn"' ' '
ape,... m n p P a r ,st... (9-1 6 ) as the indices in the factor terms.
Therefore, if it is known that a quantity is a mixed tensor of multiple When multiplication and a double contraction are combined on like terms
valence, i t is transformed by as many direct transformation tensors as i t has of the same valence, the resulting operation is called the inner product or
upper indices, and by as many inverse transformation tensors as i t has lower composition. It is defined by (putting r = n and s = m )
indices, and the transformation tensors are prescribed automatically by the a : . b: = arb; = d (9-24)
position and type of the indices.
A constant is a tensor of valence 0,since it has no indices, and it trans- For tensors of higher valence, repeated contraction of the product results
forms into itself. finally in a n invariant, as shown below, using the dot for inner product and
the ( ) for contraction:
9-9. Addition of Tensors. Tensors may be added or subtracted only if
their base letters are of the same type, and each has the same free indices
similarly placed (equal valence). Thus
-
((a:: b',:)) = ((amn ." = ( ( d 3 = ( a 3 = ( f 3= g
,bm,)) (9-25)
A chain product consists of multiplication and a single contraction, such as
+ + e: = E,
e, e; (9-17)
z,, 4 .is + z:s = zrs (9-1 8 )
112 MATRICES, TENSORS, A N D CIRCUITS [Ch. 9 Art. 9-14] T H E ELEMENTS OF TENSOR ANALYSIS 113
The expansion of Eq. 9-26 is the same as for the product of two matrices, 9-13. The Quotient Law. I n tensor analysis, an equation as a whole
each of rank 2. may be a tensor, but i t does not follow that every factor among its terms
is a tensor. The quotient law provides a criterion for testing such factors.
9-12. Multiplication of aTensor by I t s Inverse. Ifthe set of equations
Suppose that
a(rst)ba = dT (9-39)
is solved for the variable xa, there is The right side is a tensor dr of valence 1, and the left side contains the tensor
bst of valence 2. What, then, is the nature of the object a(&) ? Assuming
that the form of the equation remains invariant under transformation,
in which C: is the inverse of CL. Substituting Eq. 9-28 in Eq. 9-29, there there results
must be
2" = c a c T8x 8 = p Bx S = xa (9-30) a(a,5'y)bsY= da = Czar = Czbsta(rst)= C,"CjC:bsYa(rst) (940)
Then
:. c;c; = s; (9-31) a(a,!ly) = C;C;C:a(rst) (941)
It is also possible to prove this relationship from the definition of the element and therefore a(&) is a tensor of valence 3 , contravariant in r and covariant
of an inverse as in s and t :
CT = cofactor of Cz a(&) = a:,
(9-32) (9-42)
" determinant of the Cz and Eq. 9-39 can be rewritten
Then, from the theory of determinants, aT
st
bst = d7 (943)
The indices now balance on both sides of the equation. This r e q u i r e m e n t
that in every product in a tensor equation the indices mnst balance with the
indices in all the other terms of the equation-furnishes a quick and infallible
This is true for the product of any tensor of valence 2 by its inverse. means of identifying an unknown factor by inspection.
By virtue of Eq. 9-31, the equivalent valence of the indices on the two
9-14. General Functional Transformation. Up t o now only linear
sides of a tensor equation is easily demonstrated. Consider the equation
transformations have been considered, and tensors have been defined only
with respect to such transformations.
A general functional transformation is one which relates the variables in
and contract by putting y = a on both sides:
two different reference frames by a n arbitrary set of functions as
a;, = CkC,"C;arD= bgCiarD = CFaFm (9-35)
showing that a , m, and n are dummy indices and there is only one free index,
b, on each side of the equation. Suppose that the transformation is reversible, so that
Or consider the double contraction of Eq. 9-23 by writing (putting p = fi
and a = a )
and that the functions possess derivatives to any order required. Then
((azb:))= a;b{ = ((CkCFCyzarb:))
= CkCzC;~;arbb:= bb",b,"arb:
= aFb: (9-36)
Or consider the single contraction in the chain product ( p = /?) :
(sib:) = ( C k C ~ C ~ C ~=a ~~&b C~z)~ ; C ! a z b=: CkCzG:arb: = CkCzaFb'
in which the CF and Cr, are now no longer constants, as in the case of linear
(9-37) transformations, but functions of the variables. Thus a linear transformation
The Kronecker delta is itself a tensor, as a n examination of its trans- is merely a special case of the general functional transformation. But the
formation shows : differentials transform linearly and are integrable. Therefore the necessary
CzCibl= C;Ci = 6; (9-38) and sufficient condition that the transformation between the differentials
114 MATRICES, TENSORS, A N D CIRCUITS [Ch. 9 Art. 9-17] THE ELEMENTS OF TENSOR ANALYSIS 115
be reversible simply means that the determinant of the coefficients (the 9-16. Derivatives of a Tensor. The derivative of a tensor is not a
Jacobian) does not vanish, as was demonstrated in the case of linear trans- tensor. Let
formations. That is,
It is easily seen from the rule for the product of two determinants that
Icj . Ic;l = lc:c;l = la$ = 1 (9-50)
If xn is an electric charge and t is time, then in = dxn/dt,and
The extended definition of a tensor now becomes :
A function of any order (valence) is a tensor with respect to a general
functional transformation if it transform in accordance with a linear
transformation between the old and new differentials.
Thus, if a& is a third-order function of the variables x*, it is a tensor of This is not a tensor transformation because an additional term, other than
valence 3 if its law of transformation is Ck, appears on the right side of the equation.
9-17. Rectilinear Reference Frames. A point in a plane is uniquely
determined by two numbers, a point in ordinary space by three numbers,
A tensor of valence 1 transforms as and a point in n-dimensional space by n numbers. These numbers are the
coordinates of the space in question. The coordinates may be rectangular
.
%a =
axa i'
- (orthogonal or not) or general curvilinear. Familiar examples of different
axr coordinate systems (or reference frames) are: rectangular, polar, spherical,
If a linear form remains invariant under transformation, and cylindrical. I n every case, the point is determined by the intersection
of a set of parametric surfaces. It is even possible to use different units of
eaia= eTiT
measurement along each axis ; for example, in a rectangular coordinate system
and, upon substituting for ia,
the units of measurement along the x-, y-, z-axes, respectively, may be inches,
feet, and centimeters. As will be shown, there is a remarkable entity called the
metric tensor which automatically takes care of the different yardsticks as
Hence
well as providing the correlation between coordinate systems.
Let yT = (ya, yb,ye) be the Cartesian orthogonal coordinates of a point in
space. Then the equat.ion
If a quadratic form remains invariant under transformation, then
then the square of the distance between these two points is given by for the different axes. For example, suppose the units of measurement along
the axes are in the ratio 1 :h: k. Then, in terms of the unit measurement, the
s2= (=I2 = +
(ylf - Y ; ) ~ (Y; - Y;12 + (Y; - ~3~ distance of a point from the origin would be
= +
(Y; - Y;)(Y; - Y;) (Y; - Y;)(Y; - Y;) (Y: - Y%YE + - ~3
= C;(x; - X;)C;(X; - x[) +
C:(x: - x;)c;(x{ - 2;)
+ C:(xq - x;)c;(xf - xi)
=(Cp; + c:c; + C;C;)(x; - xi)(xlB - x2)
B
When the unit measurements are the same for all axes, gTs is 1 for r = s and
= c;C;(x; - x;)(x! - 2;) 0 for r # s.
= g&; - x;)(g - 4 (9-58)
in which
g,, = ctc; + cp; + cy; (9-59)
I n particular, if one point is the origin,
s2 = g xax, (9-60)
If the two points are infinitely close, (xy - x i ) = dxa and
Now s or ds is an invariant, and therefore the right side of Eq. 9-60 or 9-61
Fig. 9-1. Cartesian and rectilinear axes. Fig. 9-2.
is an invariant and, since xu is a tensor, it follows from the quotient law of
tensors and Eq. 9-60 that
9-18. The Magnitude of a Vector. Similarly to Eq. 9-60, the magni-
gmp= gBa is a symmetric double covariant tensor of valence 2. It tude of a contravariant vector A T is defined as
is called the metric tensor and is of fundamental importance.
A = (gT,~r~y)" (9-63)
As an example of these concepts, consider a point P in a plane whose
Cartesian coordinates are (ya,yb) and whose rectilinear coordinates referred and the magnitude of a covariant vector is defined as
to axes a t fixed angles 8 and 4 from ya are x1 and x2. From Fig. 9-1 it is clear B = (~'T~B,))"
that
ya = x1 cos 8 x2 COS 4 + in which gT8is the inverse of grs; that is,
A comparison of Eqs. 9-68 and 9-69 shows that The interchange of the dummy indices may be made automatically without
the necessity of showing the metric tensors explicitly; for example,
cos I9 = g,,aTbs
If a' and b7 are perpendicular, 0 = 90' and therefore the condition for
two directions being orthogonal is
Similarly, the free index may be raised or lowered simultaneously on both
0 = g,,a'bs (9-7 1 ) sides of an equation to yield a n equivalent equation. For example, if
I n general, for two vectors ATand BTof magnitudes A and B, respectively,
the unit vectors are a' = A T / Aand bT = BT/Band there is
then
A2 = grsA'AS
B2 = gTsBTBS (9-72) and so
cos 19 = gTSaTb" gT,ATBS/AB
The dot product familiar in vector analysis follows as 9-21. Tensor Fields. The general tensor transformation of Eq. 9-52 is
AB cos I9 = gTsATB" true in general for one point only, because the coefficients axa/axkare func-
tions of Z and therefore of xu, and consequently, if the relationship holds
9-20. Associated Tensors: Raising and Lowering of Indices. If the
for one point, there is no assurance that it will hold for any other point.
contravariant vector An is multiplied by g,, and contracted, there results
Thus the tensor is localized at a point, and the various algebraic operations
the covariant vector
which were developed for linear transformations apply in general a t a point.
A, = grmArn (9-73) If, however, A: ::: and A: ::: are functions of xk and satisfy Eq. 9-52 a t
This process is called lowering the index. every point in space where the functions are defined, the aggregate of all
If, now, An is multiplied by grn and contracted, these tensors is called a tensor ,field.
gTmAm= gTmgmnAn
= 6',An = AT (9-74)
PROBLEMS
and this is called raising the index.
Thus, lowering an index and then raising i t restores the original vector. 9-1. Give examples of the following: h i t e group, infinite group, cyclic group,
Two vectors with the same base letter, but with covariant and contravariant semigroup, subgroup.
indices, are called associated vectors and are said to represent the covariant 9-2. Write the set of equations (9-4)for three variables x8 = ( x a , xb, x C )and
and contravariant components of the same vector. xa = (xl,x2, z3) and solve for the x S in terms of the xa in accordance with
Any number of indices may be raised or lowered by repeated applications Eq. 9-5.
of the metric tensor. When a n index is thus changed, a dot is left to show 9-3. Give an example of Eq. 9-6 showing all the C.
9 4 . Give the proof of Eq. 9-33.
the position from which it was moved. Examples are
9-5. In the equation e, = z(r,s)iS,if e, and is are known to be tensors; prove
ATs = gT"Ams (9-75) that z(r, s ) must be a tensor.
9-6. Give the proof of Eq. 9-50.
A,! = gSmATm (9-76)
9-7. The relationships between certain curvilinear ( x l , x 2 , x3) and rectangular
A 7 = gSmA'"t (9-77) ( y l , y2, y3) coordinates for some typical systems are given in the following table.
Indices may be raised and lowered simultaneously by employing a suffici- Tabulate the metric tensors for each case.
ent number of metric tensors; for example,
..I: - gku~mn"
&.I - r m sn
' PI (9-78)
Spherical xl sin x2 cos x3 xl sin x2 sin x3 x1 cos x*
I n a product of tensors, wherever a dummy index occurs it can be raised Cylindrical X1 C 0 8 XZ x1 sin xB
in one tensor and lowered in the other without changing the value of the Parabolic xlx2 cos x3 ~1x2 sin x3 f
product :
Elliptic
ATsB, = gTmAmSgTnBn = 6,mtlmsBn= AmsBm= A,BT (9-79)
Art. 10-21 TENSOR DIFFERENTIATION 121
Adding Eqs. 10-6 and 10-7, subtracting Eq. 10-5, and dividing by 2 there
results
The left-hand side of this equation is called the Christoflel symbol of the
TENSOR first kind and is written
10-1. The Christoffel Symbols. The ordinary and partial derivatives The Christofel symbol of the second kind is defined as
of conventional calculus are only special cases of much more general processes
encountered in tensor analysis; for this reason tensor analysis is sometimes
called the absolute calculus.
Consider a vector-field of parallel vectors of equal magnitude whose
positions are functions of a parameter t, and in terms of Cartesian, yr, and These three-index symbols are not tensors, as will be seen when their equa-
curvilinear, xr, coordinates let this vector-field be represented by Y r ( t )and tions of transformation are derived. They are symmetrical in m , n , as is
X r ( t ) ,respectively. Since the components of these parallel vectors are equal easily seen since the metric tensor gmn is itself symmetric. They vanish in a
in Cartesian coordinates, they are constant with respect t o t and therefore rectilinear reference system. They are of enormous importance in tensor
analysis, and they possess numerous interesting properties.
10-2. Intrinsic and Covariant Derivatives. Return now t o Eq. 10-3
and multiply it by grP(aya/ax"):
But
But
and its derivative therefore is (using Eq. 10-l),
aye aye - gm, by ~ q 1 0. 4 and g m d w = 6;
axm axD
also
Now, according to Eq. 9-59 the metric tensor is qr* 7aye = rnn
ax ax ax
( '] by Eqs. 10-8,10-9,lO-10
Since the left side of this equation is a n invariant, the right side must also
be a n invariant. Interchanging dummy indices r, m, there is the'invariant
The intrinsic derivatives, identity (10-18) or (10-19), for each of the arbitrary
parallel vectors is equal to zero; hence by the quotient law Eq. 10-23 gives
which is the intrinsic derivative for Xi,. I n a similar fashion the intrinsic
is a covariant vector called the intrinsic derivative of X, with respect to t and
derivatives for tensors of any valence are easily derived. The construction
is denoted by 6XT/6t.
is clear from Eq. 10-24, and the intrinsic derivative for higher valence
I n exactly the same way the intrinsic derivative of a contravariant vector
tensors may be written by inspection. Substituting
Xr is found to be
ax:,
--
ax:, axk
Putting
at axL at
in Eq. 10-24, and separating out dxk/dt, there remains a set of terms which
by the quotient law must be a t,ensor and is called the covariant derivative,
in identity (10-19),
and, since dxs/dt is a tensor, by the quotient law the expression in square
I n a Cartesian coordinate system, the metric tensor components are all
brackets is also a tensor, contravariant in r and covariant in s. It is called. constant; the Christoffel symbols are therefore zero; and the intrinsic
the covariant derivative of XTand is denoted by
derivative reduces t o the ordinary derivative, the covariant derivative
reduces to the partial derivative.
The same rules which apply to the ordinary and partial derivatives of
sums and products apply to intrinsic and covariant derivatives, for any
I n a similar way i t may be shown that tensor relationship which is true in a particular reference system must be
true in every reference system. It is therefore only necessary to substitute
intrinsic or covariant differentials for ordinary or partial differentials to pass
from rectangular to curvilinear coordinate systems. Some examples are
The commas in identities (10-21) and (10-22) denote differentiation.
given in Table 10-1.
MATRICES, TENSORS, AND CIRCUITS [Ch. 10 125
Art. 10-81 TENSOR DIFFERENTIATION
TABLE 10-1
I n rectangular coordinates it reduces to
Rectangular coordinates Curvilinear coordinates
Let C: = A: + B:
dC,' dA: dB,'
:.dt=dt+dt 10-7. The Curl of a Vector. I n curvilinear coordinates the curl of a
covariant vector A , is defined as
curl A , = PtAt,, (10-32)
Let C;, = AkB;
dc;, - d A & in which there appears an operator defined by
- - -B,"1+ dB;
at dt
"
erst = qdlgmnl
where
p t =
1{ +10 when
where any two of the indices are equal
r s t is an even permutation of 1 2 3 (10-33)
1 -1when r s t is an odd permutation of 1 2 3
Kronecker delta 6; = constant I n rectangular coordinates this reduces to
as:
--
ax. - O
10-5. Gradient. If a covariant vector A , is derivable from an invariant 10-8. The Curvature Tensor. The curvature tensor is of importance in
4 (a potential function), it is called the gradient of 4, and in rectangular and electric machine theory when oscillating reference axes are encountered.
curvilinear coordinates i t is It is a generalized second partial derivative of a vector.
The covariant derivative of a vector X , is given by identity (10-22).
grad 4 = (a4 a4 , -
-, - 8 6 ) = 4,, = A , Regarding X,,, as a covariant tensor of valence 2: there follows from Eq.
ayl ay2 ay3 10-26 (leaving out the contravariant r index and then substituting r for s,
in which, for a scalar, the covariant derivative is equal to the partial deriva- s for t, and t for k)
tive
By exchanging dummy indices this equation may be rearranged: From this expression i t is evident that
Since the left side of Eq. 10-38 is zero for rectangular coordinates, i t must 10-10. Transformation of the Christoffel Symbol. The Christoffel
be zero for all coordinate systems, and therefore, in all reference frames, symbol of the first kind is defined by Eq. 10-9. I t s transformation is now
sought. Let the metric tensor transform as
R",,, = 0 (10-41
ga,fi,= gafiC;,C$ (10 4 9 )
The left side of Eq. 10-38 is a tensor and therefore the right side is also
a tensor, and, by the quotient law, since X , is a tensor, R!,, is also a tensor. in which the transformation tensor is a function of the coordinates
It is called the Riemann-Christoffel tensor. I t depends only on the metric
tensors gTs and their derivatives of the first and second orders.
The associated curvature tensor is Also, for brevity put
agyTa*- -c;,cj,c;.
ab'
agya
ad
-
ax@
ac:.c;,
+ gya acy,c:,cj.+ gyac:* (10-S4)
1 28 MATRICES, TENSORS, A N D CIRCUITS [Ch.10 Art. 10-101 TENSOR DIFFERENTIATION
Add Eqs. 10-53 and 10-54 and subtract Eq. 10-52, divide each side by 2, Then by Eqs. 10-57 and 10-58, Eq. 10-56 may be expressed as
and then to the right side add and subtract
The last three terms are called the non-holonomic objects; they are written
na.8.,y,
= -(act
"*Y,
2ad
a:) c:.c;.
axa
The term on the left is recognized as the holonomic Christoffel symbol of the It is skew-symmetric in its first two indices, and it is not a tensor.
first kind in the new coordinate system. The first term on the right is It is evident from Eq. 10-59 that the holonomic Christoffel symbol is not
recognized as the holonomic Christoffel symbol in the old coordinates a tensor, since its transformation involves terms other than the transforma-
multiplied by the three transformation tensors. The remaining terms may tion tensors.
now be combined by vertical pairs, and a t the same time the old g4 trans- If the non-holonomic objects are moved to the left sid; of the equation,
formed to new gatr. There results there results an identity called the afine connection:
But, in these combinations, I From Eq. 10-61 it is seen that the affine connection in holonomic space
is the same as the Christoffel symbol whose law of transformation is
I
Also, since the derivative of any of the expressions in (10-57) is zero, there
is, for example,
-
But, in non-holonomic space, the non-holonomic objects appear on the left.
0 = zRRiR + z&iV
from which
SPECIAL A N D USEFUL ARTIFICES
ZRR
and, substituting in Eq. 11-2,
A number of helpful artifices are used in circuit and machine analysis.
Rather than introduce them as they occur in the development of machine
theory, it seems better to assemble them here in a single chapter so that they
This expression follows also from Eq. 11-4 upon substituting e j = 0, j =
can be more conveniently referred to.
k = R, and ykj = (zRR)-l = l/zRn
-
I I I. T h e "Short-Ci rcuit T h e ~ r e r n . " e - ~This is simply the procedure Thus a new matrix, with row R and column R deleted, and having elements
which has frequently been used in this book for the simultaneous elimination
of several rows and columns in a matrix or tensor equation. It consists of an zuRZR,
= ZUV - - (11-10)
appropriate partitioning of the matrix equation so that the variables not
ZRR
needed can be eliminated. Thus, if there are n variables and those up to m defines the retained axes.
are to be eliminated, the partitioned equations are The advantage of this method over the previous method is that the inverse
of a matrix need not be calculated. The disadvantage is that the procedure
must be repeated for every axis eliminated.
11-3. Lauder's Rule.45 A. H. Lauder has given an interesting and
in which indices (j,k) range from 1 to m and indices (u.v) from (nz + 1) to n . practical rule for the elimination of reference axes. To establish his rule, let
A7:3 be the cofactor of z,, in the determinant D = IzIkl, so that
Solving Eq. 11-1 for the currents ik,
Substituting Eq. 11-3 in Eq. 11-2 and rearranging, Eq. 11-4 may then be rewritten
and substituting Eq. 11-5 in Eq. 11-3, But the numerators in this equation may Be recognized as the expansion of
the following "bordered determinants,"* so that the equation becomes
By Eq. 11-5 any set of currents may be solved for, and then if the remain-
ing set of currents is required, Eq. 11-6 may be used.
It is often preferable, in eliminating variables by the procedure above,
to do so one a t a time, for then the inverse matrices are much easier to com- As long as the determinants do not become unwieldy, this is a fast and simple
pute. procedure.
If, as often happens, the set of voltages e, = 0 (as in short-circuit prob- When the voltages (el . . e,,) are all zero, as for amortisseur windings or
lems), then the corresponding terms drop out of Eqs. 11-5 and 11-6 and the ground wires, the first term, e l , drops out.
calculations are greatly simplified.
130
* See NcConnel1.l p. 19.
MATRICES. TENSORS, A N D CIRCUITS [Ch. I I Art. I I d ] SPECIAL A N D USEFUL ARTIFICES 133
Fig. 11-1
As a simple example take the case of Fig. 11-1 in which there is a generator Applying Eq. 11-10 again to eliminate axis 2 :
on winding 3 and loads on 1 and 2. The impedance matrix, partitioned t o
permit elimination of the short-circuited axes, is
Arranging in matrix form, with the first row or column referring to real
quantities and the second row or column to imaginary values, gives
in which (1) and (2)now refer to real and imaginary components, respectively.
For example, the complex matrices of a polyphase induction motor are
= (z33 -
213222231 - 213232221 + 223211232 - 223212231
211222 - '12221
Or the same result may be obtained from Lauder's rule (Eq. 11-13) :
134 MATRICES, TENSORS, A N D CIRCUITS [Ch. I I Art. I 1-81 SPECIAL A N D USEFUL ARTIFICES 135
11-5. T h e Polyphase Matrix. I n balanced polyphase circuits having velocities may be fractions v, v,, v2 of synchronous speed. Then (dispensing
symmetrical polyphase windings and balanced voltages and currents, only with Re)
one phase needs to be dealt with explicitly, since the variables in any other pi = jhWIEjhWt (11-20)
phase are the same except for a phase displacement. Let there be P phases
and N coordinates per phase. The time displacement between adjacent -1 . = -
1 lEihot (11-21)
phases then is 2.rrlP or 2(x - 1)n-/Pbetween phase xand thereference phase 1. P jho
Or, in general,
Then, if indices (r, s) apply to the reference phase and indices ( m , n ) to all
= EEj(hmt+a)
= Z(P, PO, ~ 0 1P02)i
phases, any current is ,
in = E-32;r(z-l)/P 8 (*-1pfsis
n = PsniS (11-15) = Z(jho, v o , vlo, v 2 0 ) I ~ j h w t ( 11-22)
in which P," is the polyphase matrix ~ - ~ ~ " ( ~ - ' ) ~ ~The
6 ~voltage
~ - , , ~ + ~ . and, canceling the exponentials, the a-c relationship
equation then is
(11-16)
E =- Z(jhw, v o , v,o, v2w)I ( 11-23)
em = Zmnin= ZmnP:iS
results. For example, if
Z,,
For the reference phase (x = l ) ,
= Z,.
= 6; and in = i s , em = e,, and
I n rotating machine analysis often P = 4 and N = 2; thens
Z(p)= R + LP + Mp0
and i = Idhmtand p0 = vw, then
in the Heaviside operator, or Laplace transform, sense in the usual way. Y 1 Y 2= [(AC - BD) - j ( A D + BC)1!m1+m2, (11-27)
I n steady-state a-c cases the instantaneous value of a current may be Y,Y,*= [(AC + BD) - j(BC - AD)1(,1-m2) ( 11-28)
expressed
11-8. C r i t e r i a of Stability. When a disturbance superimposed on a
i= I cos hmt = Re Iejhmt (11-19) dynamic system induces an oscillation, the oscillation may die out as a result
in which w = 2n-f is the synchronous (or base) value, and h is the fraction of of the damping effects of the losses in the system, or may be sustained
the synchronous value a t which the current is varying (as, for example, indefinitely, or may cumulatively increase in amplitude. I n the latter case
during hunting or a t subsynchronous speed). Likewise, the mechanical the system reaches its limit of stability and falls out of synchronism.
136 MATRICES, TENSORS, A N D CIRCUITS [Ch. I I Art. 11-81 SPECIAL A N D USEFUL ARTIFICES
I n an electrodynamic system, such as an electric machine, the equation of Hurwitz's criterion for stability requires that
small oscillations takes the form
1. All the coefficients a be positive.
Ae, = Z,, Ais ( 1 1-29) 2. The n determinants
in which the determinant of is a function of the time derivative p of the
form
IT,^( = anpn + an-lpn-l + . . + alp + a0 (11-30)
When equated to zero, this is the characteristic equation of the system, and
its roots define the characteristic values which determine the nature of the
transient. I n general, roots will occur which are pairs of complex numbers
of the type ( a fj o ) , and they give rise t o terms in the solution of the form formed by letting k be 1 , 2, . . . , n in succession be positive.
MACHINE ANALYSIS
BASIC CONCEPTS I N
ELECTRIC MACHINES
general solutions may be applied to special cases. To this end, certain The total flux linked with the coil a t any instant then is
simplifying assumptions need to be made in the interest of economy of time
and space. I n particular, the analysis will be idealized t o the extent that i t
will be assumed that:
in which K , = sin (pn/2) is the pitch coefficient. The average value of a
1. There is no saturation in the magnetic paths. sinusoid is 2/3r times its amplitude, and the area in this case is 71, so that
2. Hysteresis and eddy current losses in the iron may be ignored or else Eq. 12-2 may be expressed in terms of the flux @ = (2/3r)dB:
included in fictitious circuits.
3. Stray losses in the adjoining structures or frame may be ignored.
4. Space harmonics in the distribution of flux do not exist or may be
ignored. in which @ = @ ( t ) , xo = xo(t),and y = y(t) are all functions of time.
5. Time harmonics in the voltages, currents, and fluxes do not exist
under steady-state conditions or may be ignored.
6. The effects of armature reaction and of field flux distribution are
sinusoidally distributed in the air gap space. (This, of course, is equivalent
to assumption 4, but it is restated in this form for emphasis. This assumption
is essentially true in a properly designed machine in which the windings and
magnetic paths are properly distributed.)
7. The winding reduction factors, such as the skew, pitch, distribution,
and connection coefficients, are incorporated in the "effective" turns of a
winding and need not be shown explicitly.
Fig. 12-1.
8. The windings on a magnetic structure (the stator or rotor) are sym-
metrically placed with respect to a reference axis, and are balanced (that is, By Faraday's law the voltage induced in the coil is
each group of windings or phase is like the others).
12-1. T h e Fundamental Voltage E q u a t i ~ n . ~ ~The . ~ l form of the
equation of voltage and its physical interpretation depends on the choice of
the arbitrary reference axis. I n order t o obtain as general and versatile an
expression as possible, and one from which the process of induction in any
machine follows as a special case, i t will be assumed that with respect to the
reference axis :
1. The flux density, sinusoidally distributed in space, is varying in time This equation shows that a voltage may be induced in the coil in three
in any arbitrary manner. different ways :
2. The flux density distribution is moving. 1. By variation of the flux (the so-called transformer component), with
3. The winding is moving. both the coil and flux stationary in position, d@/dt.
Fig. 12-1 shows a coil of N turns, axial length I, and coil pitch p centered 2. By movement of the conductors (the so-called "cutting action"), with
a t xo(t),a function of time t, with respect to a n arbitrary reference axis, and the flux momentarily fixed and unvarying, dxo/dt.
moving with respect to an arbitrary reference axis with a velocity dxo/dt. 3. By movement of the flux, with its amplitude and the position of the
This coil is linked with a sinusoidal flux density distribution of wavelength coil momentarily fixed, dyldt.
27 given by It is obvious from Eq. 1 2 4 that, if the reference axis is taken as
/I = B sin (z
- y)
T
(12-1) fixed to the flux wave, then dy/dt = 0, and there is no component of
voltage induced by movement of the flux;
and whose amplitude B(t), and displacement y(t) from the reference axis are, fixed on the coil, then dxo/dt = 0, and there is no component of voltage
in general, functions of time t. due to cutting action.
Art. 12-21 BASIC CONCEPTS I N ELECTRIC MACHINES 145
144 M A C H I N E ANALYSIS [Ch. I 2
Thus the appearance or vanishing of a component of voltage can often be axes. I t is clear that, if a current flows in a winding through the brushes in
made t o depend only on the arbitrary selection of the reference axi~,~Os~l
and the direct axis, the resulting axis of magnetization is also in the direct axis.
consequently the physical interpretation must change with the selection of For a single stator layer and a single rotor layer let the currents be id', ids,
the axis. iQr,iqsand the self- and mutual inductances be L,,, Lds, La,, Lqs, Md, Ma.
Put 8 = r x 0 / r and p = dldt, and take the flux density distribution as This simplifying variation of tensor notation avoids a complicated ranking of
stationary with respect to the reference axis so that dyldt = 0. Then Eq. subscripts and superscripts. There is, of course, no mutual inductance between
1 2 4 reduces t o windings in quadrature (90" apart in space), and for simplicity i t is assumed
that the mutual inductance in an axis is the same for all windings. Then the
resultant fluxes linked with the rotor in the two axes are
in which the effective turns N are now assumed t o include the effects of coil
pitch, skew, distribution, and connection of coils in a group.
12-2. Direct and Quadrature Axes. I n machine analysis i t is often The positions of these two fluxes with respect to the reference axis are
advisable, and sometimes necessary (as in the case of salient pole machines), indicated in Fig. 12-3, and by comparison with Fig. 12-1 the following
identifications are easily made :
-
- +
- i r ) ( ~ ~ ~ i Mdids)
~' sin (0 + 90) + (Ld,idT+ Mjds)p8cos (0 + 90)]
u
= + lWqiqs)sin (0 + 0) + (L,,iaT+ 1M,iaJ)p8cos (0 + O)]
-[p(Lq,iqT
Fig. 12-2. Direct and quadrature axes and coil designation. .'. ed, = -Li)(LdTidT+ Mdids)+ (LqTiqT + MqiqS)p8]
t o select a pair of axes centered on the field poles (or main axis of magnetiza-
= -Li)4dT + +gT~el (12-9)
tion in the machine) and in quadrature (half a pole pitch) therefrom. These e,, (due to 4dT)
two orthogonal axes are called the direct (d) and quadrature (q) axes, respec- = -@(Ld,idr + Mdids)sin (90 + 90) + (Ld,idT+ Mdids)p8cos (90 + go)]
tively.
Fig. 12-2 shows a symmetrical machine t o which are attached direct and
quadrature axes, labeled d and q, respectively. The stator (or stationary)
coils are designated by s l , 82, etc., counting from the air gap outward. The
= +
--[p(Lq,iQT MaiaS)sin (90 + 0) + (LaTiar+ M,iaS)p8cos (90 + o)]
rotor (or moving) coils are designated by r l , r2, etc., counting from the air .'. ear = -[p(Lq,iqT + MaiaS)- (LdridT+ Mdids)p]0
gap inward. To fix ideas, brushes are shown on the rotor windings in both = -[PA' - 4dTp81 (12-10)
146 M A C H I N E ANALYSIS [Ch. 12 Art. 12-31 BASIC CONCEPTS I N ELECTRIC MACHINES
147
The first term in Eq. 12-9 is a n induced voltage due t o the variation of the Hence the torque on the coil of Fig. 12-1, if the air gap radius is R,is
flux. The second term is a generated voltage due to the movement (PO) of the
rotating conductors cutting the flux. The corresponding (space) vector
diagram shown in Fig. 1 2 4 follows directly from Eqs. 12-6, 7, 9, and 10.
But 27rR/r = P, the number of poles, and, as before, (2/7r)71Bis the flux per
pole. Therefore Eq. 12-13 becomes, putting 8 = (7rx0/r)and absorbing the
pitch factor K, in t,he "effective" turns I?,
Fig. 12-3. Fig. 12-4. Components of induced and The machine torques on the rotor for the several fluxes (Eqs. 12-6 and
generated voltages. 12-7) and rotor currents (per pair of poles and in terms of flux linkages)
then are: for &,, y = -90°, and iqT,8 = 90°,
These equations may be arranged as matrices (taking for simplicity a single
stator layer and a single rotor layer of windings), and including the resistances f - z'" ( L id'
-- + Maids)cos (90 + 90) = --iW(LdTidr+ 1%faids)(12-15)
of the windings : for #,,, y = 0, and idT,0 = 0,
ds dr Pr P
+
fi = idT(LgTiqrMgigs)cos +~
(0 - 0) = idT(Lqriqr & f ~ (12-16)
j ~ ~ )
for y = -90, and id', 8 = 0,
-Tds - Ld# -"d~ 0 0
but now the impedance Zaphas been generalized to include not only resistances
and inductances but also the effects of rotation. Although the torque is an invariant (tensor of zero valence), nevertheless
12-3. The Fundamental Torque Equation. The force on a current- i t may be considered a vector in the direction of the machine shaft if a
carrying conductor of length 1in a magnetic field of density and a t a n angle suitable index is attached. The same index may be used with angular
1therewith, by the law of Biot-Savart is velocity vectors. This artifice will permit the consolidation of the torque
and voltage equations into a single equation of motion, as will be shown in
P = Bli sin 1 (12-12) subsequent chapters.
1 48 M A C H I N E ANALYSIS [Ch. 12
I Voltage ing generalized (or primitive) machine has multilayer windings on both stator
Machine .Reference Axes on Desired and rotor, as indicated in Fig. 12-2. The resulting equation of voltage is of the
type given by Eq. 12-11, and the equation of torque is of the type given by
Transformer Core Winding Eq. 12-19. These developments will now be described in greater detail.
D-c generator Field Armature
Field Armature 13-1. The Generalized Machine (Primitive Machine of the First
Syn. generator
Syn. generator Armature Armature Kind).7@923~~~ The generalized machine with quasi-holonomic reference axes
Induction motor Stator Rotor consists of a cylindrical stator with salient poles and a cylindrical rotor, each
Induction motor Rotor Rotor with any number of concentric layers of windings. I n each winding layer
Induction motor Rotating field Rotor on the stator there are separate coils centered on the direct and quadrature
axes. I n each winding layer on the rotor there is a pair of brushes (hypotheti-
cal perhaps) centered on the direct axis and another pair of brushes centered
on the quadrature axis. Currents idflow in the direct axis coils or brushes,
and currents iqflow in the quadrature axis coils or brushes.
It is assumed that the windings are so designed (coil skew, pitch, and
distribution) that the mmf produced by a current through a brush axis is
sinusoidally distributed in space and centered on the brush axis. The signi-
ficance of this assumption is that the mmf along any brush axis may then be
resolved into components along orthogonal directions upon multiplying by
cos a: and sin a:, where a: is the angle of brush shift. Likewise, a voltage
existing along a brush axis may be resolved into components. Only in this
way may voltages, currents, and mmf's be conceived of as space vectors
resolvable into direct and quadrature components; see Pig. 13-1.
149
M A C H I N E ANALYSIS [Ch. 13 Art. 13-21 THE GENERALIZED MACHINE 151
This equation was set up for a generator with rotating armature, C.W.
jdrl
rotation, and q-axis ahead of the d-axis. It is now necessary to see what
changes need t o be made for a motor instead of a generator, if the direction
(a) (b)
of rotation be reversed, if the field instead of the armature is the rotating
Fig. 13-3. The generalized machine with t w o stator and t w o rotor winding element, or if the q-axis is behind the d-axis.
layers.
152 M A C H I N E ANALYSIS [Ch. 13 Art. 13-21 THE GENERALIZED M A C H I N E 153
I n the case of a motor, either the voltages are reversed (that is, impressed The modifications in the basic impedance tensor for a generator (Eq. 13-l),
voltages are substituted for output voltages) or the currents are reversed occasioned by changing the direction of rotation p8, the position of the d-
(for given terminal voltages the currents flow in the opposite direction). and q-axes, from rotating armature to rotating field, or from generator to
Thus the equation of voltage for a motor becomes
TABLE 13-1
Changes Effected by Reversal of Direction of Rotation,
Axes, or Currents
and, therefore, all that is necessary is to change the signs of all terms in the
Member Axes
- Change in Zap
Field / Armature
-
2'
of Eq. 13-1
-
If the direction of rotation is reversed, i t is only necessary to substitute Stationary Rotating C.C.W. Right Change
terms.signs of pi
-p8 for p8 in the impedance matrix.
--
If the q-axis is reversed, then the direction of rotation is reversed and i t
is necessary to change the sign of p8 in the impedance matrix. Thus Eq. Stationary Rotating C.W. Left Change signs o f p ~
terms.
13-3 would become
--
de dr qr 'P
Change signs of ;o(
ds rd8 + L d 8 ~ M d ~ 0 0 Stationary Rotating C.W. Down Right
terms.
-
d+
.
MdP r, + LdVP -L;d -M;pe
Zmotor - (134)
qr M;Pe Lkpe r, + LavP Map Change signs of p(
terms; see Eq
Rotating Stationary C.W. Right
q* 0 0 Map + L,,P 13-5 for a
-I
T,. motor.
If the field is the rotating element and is rotating c.w., then relative to
the field the armature is rotating counterclockwise (c.c.w.), so that in Eq. Change signs of
Any of above all terms; see
13-1 for a generator, or in Eq. 13-3 for a motor, i t is necessary to substitute Eq. 13-3.
-p8 for p8. For example,
motor, are given in Table 13-1. For example, what changes are necessary
in Eq. 13-1 to adapt i t to a revolving field motor, C.C.W.
rotation, and q-axis
t o the left ?
The impedance matrices for machines with more than one winding layer in which the $us-linkage vector is
R +
on the stator and on the rotor, as well as for the case of zero sequence currents, ds LdsidS Mdid7
will be introduced under appropriate machine headings. For example, a
synchronous generator may have several amortisseur windings in both the dr Mdida+ Ld7idT
+a = Lapis =
direct and the quadrature axes, and zero sequence currents may flow in
addition to the d- and q-component currents.
qr Lqliqp+ Mqigb
For certain reasons it is sometimes advantageous to rearrange the rows
and columns of the impedance ma,trix. Where this is indicated in the analysis
qsI Mqiqr + LqsigE 1
and the cross-flux vector is
of particular machines, the changes will be made without apology.
13-3. Components of the Impedance T e n ~ o r . ~ A 2The ~ J ~impedance
matrix of the generalized machine may be broken down into three component
matrices :
However, since ya has zero elements in its ds- and qs-axes, i t must be multi-
plied only by rotor currents in Eq. 13-12, so that if indices (r, m ) refer to the
rotor and (s)to the stator,
T = iaG
aS
ip
= ir(Grm + Grs)(im+ is)
= irGr,jm + irGrsis (13-1 3)
in which
Qap =
d
ds
/
qr - M :
r
--
dr
-hir
m
-
-
dr
qr
l
qr - M &
qs
rlrl
ds qs
dr
9.
dr
--Ad,
qr
Comparison of Eqs. 13-22 and 13-23 clearly shows that the q component
of ia has become the negative d component of ia, and the d component of ia
has become the q component of iu; that is, the vector ia has been rotated in
the positive direction of 8, by 90" (Fig. 1 3 4 ) .
If, now, Eq. 13-21 is multiplied by the inverse of Eq. 13-20, there results
Similarly,
which is the matrix of Eq. 13-18 and rotates a vector with which it is multi-
CI = (13-20) plied in the negative direction by 90".
q sin 0 , cos 0 ,
The matrices (13-24) and (13-25) are elements of the group
.:,
equation of performance so that The transformation tensor therefore is
ia = cz,ia' (13-27) b
and the transformation matrix is 13-1 1. Successive Transformations. Any number of transformations
m "n may be made in succession-turn ratios, brush shifts, coil displacements,
- sin
(13-37)
sin a, sin cc, Fig. 13-8. Displaced stator coil. Fig. 13-9.
,C
: = (13-38)
13-10. Interconnection of Coils. Stator and rotor coils may be If the a-c currents are given as rms valnes, the torque, in synchronous
interconnected in any fashion. New currents may be selected and the trans- watts, is
formation matrix set up in the usual way. For example, in Fig. 13-9
T,., = ~ . e (~LGJ) (1345)
PBG,BiaiB= (speed) (torque) = mechanical power output ds2 ds 1 drl dr2 qr2
1 1
would demonstrate nothing new. I n general, however, a complete analysis
I
of any machine would include the following steps (but not necessarily in
this order) : L
I?. +~ P M.11~ M.1.P
3. The R,,, L,,, G,, Z,,, em, and inof its primitive machine. -I
4. The transformed L,,,,, G,,,,, Z,,,,, em,, in' corresponding to its diagram
of connections and transformation matrix.
5. The transient analysis based on stationary axes and the derivation of the
equations of performance (voltages, currents, torques).
6. The steady-state Z,,,,, G,,,,, em,, in'. The inductance matrix, L,,, comprises all terms in the impedance matrix
7. The equivalent circuit of the cross-jield theory. having p as a coefficient, and the torque matrix, G,, comprises all terms in
164 M A C H I N E ANALYSIS [Ch. 13 Art. 13-14] T H E GENERALIZED M A C H I N E 165
the impedance matrix having p% as a coefficient; thus other combinations of stator and rotor windings, brush and slip ring arrange-
ments, salient poles, and polyphase coils are possible and have been employed.
ds2 dsl drl dr2 qr2 qrl qsl qs2
Rotating brushes open up additional possibilities, and rotating magnetic
poles or teeth, with or without windings, still others.20
TABLE 13-2
drl Winding Stationary
layers axes Special cam
or
Machine Stator combinatioi
Stator of
Generalized
Salient pole
synchronous
Round-rotor
synchronous
Reactor motor
Salient pole
synchronous
Single-phase
ds 1 drl dr2 or2 arl as1 us2 alternator
Slip ring
induction motor
ds2 Asymmetrical
-- induction motor
cis 1
I I Squirrel-cage
induction motor
Single-phase
induction motor
Shaded pole motor
Compound-wound
d-c motor
Polyphase
commutator motor
Leblanc advancer
scherbius advancer
DBri motor
Repulsion motor
squirrel-cage
repulsion motor
Fynn-Wcichsel motor
lchrage motor
lynchronous
converter
Frequency converter
I n general, voltages and currents may be assumed to exist in all axes,
although in any particular case many of them are zero.
Any winding (axis) may be considered non-existent either by deleting PROBLEMS
the entire row and column in the general matrices given above, or by putting
the corresponding current equal to zero in all equations and ignoring the 13-1. Confirm all equations in the chapter and verify in detail every step
corresponding voltage. in the derivations.
Some machines may be regarded as special cases (or simplified editions) 13-2. If the q-axis is reversed, the text states that the direction of rotation
of other machines, and some are combinations of others. I n any event they is reversed and it is then necessary to change the sign of p0 in the impedance
may be classified by groups as shown in Table 13-2 (in which S, I , and C matrix. Show that this is consistent with changing the signs of the q-axis
refer to synchronous, induction, and commutator, respectively). Numerous voltage8 and currents.
166 M A C H I N E ANALYSIS [Ch.13
13-3. A commutating machine is equipped with two pairs of brushes rotating
at constant speeds in opposite directions. The pairs of brushes are connected in
series. What is the transformation tensor for this arrangement?
13-4. Verify, from f i s t principles, the voltage in the qr2-axis in the matrix 14
of Eq. 13-51.
13-5. Why is the conjugate of iu taken in Eq. 13-45, and why the conjugate
of GBorin Eq. 13-46? U S E OF THE LAGRANGE EQUATION
Each of these points of view has certain advantages and each has a different The transformations may proceed from 1 to 4 and 1 to 7 with 5 and 6 as
physical interpretation, but all of them give the same numerical results. special cases of 4, and 8 and 9 as special cases of 7 ; or alternatively from
By each method there will be derived: 2 to 5 and 2 to 8, and 3 to 6 and 3 to 9.
1. The equation of voltage. 2. Maxwell's equation for stationary circuits (Faraday's law) is used to
2. The equation of torque. establish the equation of voltage in a holonomic reference frame. An indepen-
3. The equation of motion (combination of voltage and torque). dent equation of torque is established and combined tensorily with the equation
of voltage to establish the equation of motion. These equations may be
I n each method of attack these equations will be derived and transformed transformed t o non-holonomic and quasi-holonomic reference frames in
to three reference systems : accordance with the following scheme :
1. Holonomic reference system, in which the axes are fixed t o the con-
ductors, and there are Holonomic Non-holonomic Quasi-holonomic
n unknown variables xa 4
Voltage Maxwell 1
n equations of motion --
Independent 2 5
Lagrange's equation of motion and Maxwell's equation of voltage for station-
ary circuits apply to a holonomic reference frame. Motion Combine 1 and 2 3 9
2. Non-holonomic reference system, in which the axes are free and moving,
and there are The transformations may proceed from 1 to 4 and 1 to 7, from 2 to 5 and
n old variables xa 2 to 8 ; then 4 and 5 may be combined to give 6 , and 7 and 8 may be combined
2n unknowns
n new differentials ia' = dxd/dt to give 9 ; or alternatively from 3 to 6 with 4 and 5 as special cases, and from
n equations of motion 3 to 9 with 7 and 8 as special cases.
2n equations 3. The Maxwell-Lorentz equation (Faraday's law extended to moving
n equations of transformation
circuits) is used to establish the equation of voltage in a quasi-holonomic
3. Quasi-holonomic system, which is a special non-holonomic system reference frame. An independent equation of torque is established and com-
which can be treated as a holonomic system, and there are bined tensorily with the equation of voltage to establish the eqwation of
motion. These equations may be transformed to non-holonomic and holo-
k variables identical in old and new axes
n unknowns nomic reference frames in accordance with the following scheme
(n - k ) new differentials
n equations of motion Holonomic Non-holonomic Quasi-holonomic
1
- -- --
-1 I 5 Independent
--
2
and the equation of torque are then special cases of the equation of motion.
These equations may be transformed to non-holonomic and quasi-holonomic
Motion
1 6 Combined 3
reference frames in accordance with the following scheme: The transformation may proceed from 1 to 4 and 1 to 7, from 2 t o 5 and 2 to
8; then 4 and 5 may be combined to give 6, and 7 and 8 may be combined
Holonomic Non-holonomic Quasi-holonomic to give 9; or alternatively from 3 to 6 with 4 and 5 as special cases, and
from 3 to 9 with 7 and 8 as special cases.
Motion Lagrange 1 Boltzmann-Hamel 4 7
Each method of procedure yields various interesting by-products, such
Voltage Special case 2 5 8 as the Christoffel symbols, the affine connection, the torsion tensors, the
--
Torque 1 Special case 3 6 9
field tensor, the torque tensor, the rotation matrices, and the non-holonomic
objects.
170 M A C H I N E ANALYSIS [Ch. 14 Art. 14-41 USE O F T H E L A G R A N G E E Q U A T I O N 171
14-2. The Idealized Model. I n Fig. 14-1 is shown a machine having in which indices (a, /?) include the electric coordinates (m, n) and the mechan-
stationary (d, q) axes, a rotor rotating a t speed p8 = is, a magnetic pole ical coordinates (u, v), and where now ia represents an electric current if
(without any winding) rotating a t speed p8, = iP, and a moving axis a u = m, n, or a n angular velocity if u = u, v. Likewise xu will represent a n
rotating a t speed pel = i t The indices (p, T , s) represent mechanical coordin- electric charge, xm, or an angular displacement, xu, and ea will represent a
ates, so that xP, xT,xSare particular values of xu, and the velocities ip, iT,is voltage, em, or a torque, e,. With this understanding, Lagrange's equation is
are particular values of iu. I n such a machine the inductances L,, are not
constant but are functions of 8, = xv. If the moving reference axis is fixed
Fig. 14-1. Generalized machine with moving conductors, pole, and axis.
t o the rotor, then 8, = 8; or, if the moving pole is fixed to the rotor, then Substituting these values in Lagrange's equations yields the equation of
8, = 8; but, in general, 8, 8,, 8, are all different and are assumed t o be motion :
orthogonal to each other and t o all electric variables xm.
14-3. The Lagrangian Equation of Moti0n.~498 The energy in a
magnetic field and the energy in a rotating body are given by the quadratic
forms
Wm = iL,,imP (12-20)
Wo =~ J u u ~ U ~ u (12-22)
The energy losses in an electric circuit and in a rotating mechanical system
are also given by quadratic forms :
I n this equation the first term includes the electric resistance drops and
We = Rmnimin (12-21) mechanical friction torque; the second term includes the flux linkage
W, = R U u ~ U ~ u (12-23) induced (variational or transformer) voltage drops and the torques of
angular acceleration; and the final term includes the generated (motional
Since Eqs. 12-20 and 12-22 have identical forms, they may be combined
or cutting action) voltage drops.
to give the total kinetic energy in the electromechanical system, and similarly
the energy loss equations may be combined; thus 14-4. Significance of the Holonomic Christoffel Symb01.~~?8 The
Christoffel symbol appearing in Eq. 1 4 4 includes all the electric and mechan-
T = 1,aa8iai8 = kinetic energy function (14-1)
ical coordinates. However, as will be shown, most of its elements are zero,
F = :R,,+"ia = Rayleigh dissipation function (14-2) and only three orthogonal "slices" of its representative cube have any
172 M A C H I N E ANALYSIS [Ch. 14 Art. 14-51 USE OF T H E L A G R A N G E E Q U A T I O N 173
significance. The inductance coefficients a,, = a,, = Lmn=f (xu) are The equation of torque is
functions only of the mechanical angle, xu = e2. The mechanical moments of di "
inertia a,, = J,,, are constants, and valid only for u = v (there is no such e, = RuViv a,, -
at
+ + [mn, u]imin (14-9a)
thing as a "mutual" moment of inertia). Also, the Christoffel symbol is
div 1 aamnimin
symmetric, so that = R,,iv + a,,, -
at
- --
2 axTL
[UP,Y] = [ h 71 (14-5)
and
must include a t least one mechanical index, u. Furthermore, since there can
machine, u -
The mechanical coordinate may be taken along the shaft of the rotating
s. Then it is seen that the Christoffel symbol has zero elements
when P=
and when
din
d q
-
B
a b s
+
em = Rmnin a,, - + [un, m]i"in + [ku, m]iki" Fig. 14-2.
,at
except for the three parts given in Eqs. 14-8a and 14-9a and shown as the
shaded parts of the cube in Fig. 14-2, in which (d, q) are the direct and
quadrature axes of the stator, (a, b) the rotating axes of the rotor, and s the
mechanical axis of the shaft. Moreover, many of the elements in each of
these slices are zero, and consequently the equations are not so complicated
as might be supposed.
14-5. T h e Non-holonomic Equations.24,8 It is required to transform
from a holonomic system in which there are n unknown variables xu and n
equations of motion t o a non-holonomic system in which there are 2n
Expression of Eq. 14-8c follows upon substituting the scalar mechanical unknowns (the n old variables xa and n new differentials ia') and 2n equations
angles 8 and e2 for the mechanical coordinate xu. (the n equations of motion and the n equations of transformation). The
174 M A C H I N E ANALYSIS [Ch. 14 Art. 14-63 USE OF THE LAGRANGE E Q U A T I O N 175
reference axes in the non-holonomic system are no longer fixed to the con- The equation of torque in a non-holonomic system follows from Eq. 14-12
ductors but are in any position and may move a t any velocity. Let the upon substituting y' = u' for the free index; then
transformation tensor be C:,--a function of the old coordinates. Then div' 8~:'
Q, = + a,,,,
RutV,iv' - + [m'n', u']im'in' - C=,GE.- .n, .mp
ak,,,% z
ia= cz,ia' at axu
and, transforming Eq. 14-4,
This equation could have been found directly upon transforming Eq. 14-9.
14-6. The Quasi-holonornic equation^.^^.^ I t is required to transform
from the holonomic system in which there are n unknown variables x" and n
and, by identity Eq. 10-61, equations of motion to a quasi-holonomic system in which there are n
unknowns [the k mechanical variables xu" = xu which are identical in the
old and new coordinates, and the ( n - k) new electric variables xm"] and
n equations of motion which are functions of the old mechanical coordinates
= R ,.,.ia'+ did'
- + ([af/?',y']
at
- Qp.,.,,. + Q,,a.,fl. - Qatfl,,,,)ia'is'
xu and the new electric variables. The transformation tensor is a function
only of the mechanical coordinates xu. The quasi-holonomic system is thus
a special case of a non-holonomic system in which certain coordinates do not
transform; for this reason it may be treated, in a sense, as though it were
holonomic.
By comparison with Eqs. 1 4 4 , and with Eq. 14-3 from which it came, the Let the transformation tensor be
last expression may be written C:,, = f ( x U ) and C;,. = a:,, (14-18)
Then
im= CK,,~"" (14-19)
i u = c;,,iu"= ,j;,,iu" = iu"
in which By virtue of Eq. 14-18,
Eq. 14-13 is therefore the generalization of Lagrange's equation to non- and therefore the non-holonomic objects in Eq. 14-12 may be considered as
holonomic reference systems. It is called the Roltzmann-Hamel equation tensors :
and is of great importance in analytical dynam'ics.
The equation of voltage i n a non-holonomic system follows from Eq. 14-12 This tensor is skew-symmetric in its first two indices.
upon substituting y' = m' for the free index and using the same arguments The quasi-holonomic equation of motion follows from Eq. 14-12 as
which preceded Eq. 14-8. Then
din'
em, = Rm,,.in' + a,,,, - + [u'n', m']in'iu' + [k'u', m']ik'i"'
dt
the voltage equation becomes Motion (14-11, (14-21, ( 1 1 2 ) (14-13) (14-18), (14-19), (14-21),
(14-3), (14-4) (14-14), (14-15) (14-23), (14-24)
+ dtd +
em,, = ~ , , , ~ , , i " " - (am,,n,.in") Gm.,n,,in'~e
If the scalars B and 8, are substituted for xu., and pe and p8, for iv', t h e
-torque (now a scalar) becomes
1 aLrn,,,,,,.mrp
+ ~ "- -2 2
T = Rpe JP2e - ~ , , , , , , i ~ ' ;- .,,,,
(1629)
2 ae,
in which the first term on the right is the frictional torque, the second term
the acceleration torque, the third term the electric torque, and the last
term the torque due to moving magnetic paths.
Art. 1 5-41 USE O F M A X W E L L ' S E Q U A T I O N 179
and the rate a t which this is varying with respect to 13 and 8, is the electro-
magnetic torque :
15
U S E O F MAXWELL'S EQUATION This torque, plus the mechanical torque T applied to the shaft, is consumed
by friction (assumed proportional to the velocity) and the inertial torque,
so that
15-1. Equation of V0ltage.7~~ Maxwell's equation of voltage (Faraday's T = R p e + J p 2 8 - - - 1 aLmn imp 1 aLmn 2.m.n
2 (15-6)
2 ae 2 ae,
law) applies to stationary circuits, that is, to a holonomic reference frame
in which the axes are fixed to the conductors. If the inductances are functions Let the torque be considered as a vector, e,, along the shaft with a mechanical
only of the mechanical coordinates xu and not of the currents (no saturation), index u, and let the angular displacements 0 and 8, be considered as a vector,
the equution of voltage in a holonomic system is xu, along the shaft, and their derivatives p8 and p8, as a velocity vector, iu.
Then Eq. 15-6 may be rewritten
Since the only mechanical coordinates are the angular displacements of the Obviously, when the free index is electric, y = m, this equation reduces t o
rotating armature and magnetic paths, xu = 8, 8, and i u = p8, PO,, so that Eq. 15-1, since a,, is not a function of the electric coordinates. And, clearly,
when the free index is mechanical, y = u, this equation reduces t o Eq. 15-7,
since a,, = Juv is a constant. Eq. 15-8 may be rewritten
dim
+
e, = RYaia a,, - +
dt
dia
= R,,ia + a,, - + [ap, y]i"is
at
Equations 15-1 and 15-2 are seen to be identical with Eqs. 1 P 8 a to 14&,
This equation is identical with Eq. 1 4 4 . Thus Maxwell's equation (really,
derived Gom other considerations, and the definition of Eq. 15-3 is the same
Faraday's law) yields equations identical with those given by Lagrange's
as for Eq. 14-10.
equation. Transformation to other systems-non-holonomic and quasi-
15-2. Equation of Torque.7~8 The electromagnetic energy stored in the' holonomic-would therefore be the same.
field is
We = QLmnimin (15-4) 15-4. Transformation of the Equation of V ~ l t a g e . ~Let ? ~ the
equation of voltage be transformed directly to the quasi-holonomic system.
I78
180 M A C H I N E ANALYSIS [Ch. 15 Art. 15-51 USE OF M A X W E L L ' S E Q U A T I O N 181
I n the new system the mechanical coordinates are the same as in the holo- 15-5. The Transformation of the Equation of Torque. The equation
nomic system, xu" = xu(8, O,), and the transformation tensor C$ = f(xu) is of torque (15-7) may be transformed directly from rotating (holonomic) to
a function only of the mechanical coordinates. Then stationary (quasi-holonomic) axes upon substituting Eq. 15-10.
Substituting in Eq. 15-lb, there results the equation of voltage in the quasi-
holonomic system :
= + L,,,,.,
Rm,,,.,in"
din" a
-+ Cz,, -( L k n X ~ ) i u " i n "
at axu"
=
.d
dt
+ +
Rm,,n.in" - (Lmttn,,in") ym,,pO= Rm,.,,in'' + d$m"
-+ ym,,p6
dt
(15-11)
Substituting f3 and 6, for xu", and p6 and p6, for iv"(and since Lmmn.is a
The PO, term vanishes if the magnetic paths do not move. In these equations function of 8, only), then by Eq. 15-12 the scalar form of the torque equation
becomes
., ,,, 1 aLm..,..
+
T = Rp6 Jp20 - Gm,rnr,.t.t - - -i ""in" (15-15)
2 ae,
is recognized as the rotation matrix y ~ of
k Eq. 13-18, and consequently, by
There is no need to show the transformation of the equation of motion
the definitions of the torque matrix Gmngiven in Eq. 13-17, and of the cross- because it would be exactly like that carried out in Chapter 14.
flux y m given in Eq. 13-19, there have been substituted The equations developed in this chapter, starting with Maxwell's equation,
can be further discussed in terms of their physical significance and of the
construction of certain matrices and tensors, but, as these same ideas are
presented more explicitly in the next chapter, consideration of them is
deferred.
182 M A C H I N E ANALYSIS [Ch. 15
PROBLEMS
15-1. Confirm all equations in the chapter and verify in detail every step in
the derivations.
15-2. What is the torque equation for a, machine having a salient pole rotor
without windings? USE O F T H E MAXWELL-LORENTZ
15-3. What is the electric torque on a salient pole machine whose inductance
varies as EQUATION
L = +(Ld + L,) + 4(Ld - L,) cos 28 + AL cos so2
15-4. Show a side-by-side comparison, including all essential steps, of the
derivation of Eqs. 14-4 and 15-9. In your opinion, which method of approach
is shorter and simpler? Is one approach more "general" than the other? 16-1. The Equation of Voltage.7.8 The generalized machine described
15-5. What cardinal assumption was made with respect to the inductances in Chapter 13 has stationary d- and q-axes with respect to which the rotor
in writing Eq. 15-4? (Recall that all subsequent equations are valid only to the windings are moving. As shown in Chapter 13, there are two components
extent that this cardinal assumption holds.) of voltage : the induced voltage due to the variation of the flux in each axis,
and the generated voltage due to the cutting of these fluxes by the rotating
conductors. A coordinate frame of this type is called quasi-holonomic.
There are n unknowns : the k mechanical variables xu = (8, 13,) representing
the angular displacements of the rotor and the magnetic paths, and the
( n - k) electric currents im. There are (n - 1 ) equations of voltage and one
equation of torque.
The equation of voltage for a motor (Eq. 13-10) is given by the Maxwell-
Lorentz equivalent of Faraday's law:
,
= Rmnin + L,, din
at
aLm,
- + -i p6,
ae, + T,,,,% w'u
% (16-1)
I n this equation the term
is due to the movement of magnetic paths (Fig. 14-l), such as might be caused
by the rotation of a magnetic mass other than the field poles. For the sake
of generality, the angular posit~on8, and velocity p0, of this mass have been
taken as different from the rotor position O and velocity p0. I n most com-
mercial machines the magnetic paths, except for the effect of the rotor teeth,
I83
184 M A C H I N E ANALYSIS [Ch. 16 Art. 16-31 THE MAXWELL-LORENTZ EQUATION 185
are stationary, and L,,, # f(f3,) but is a constant, and the term (16-2) path, O,, be considered as components of a mechanical vector xu. Then the
vanishes. corresponding velocity is pxu = i u , and Eq. 16-7 may be written
The term T,,,,i21 = GmnpH (16-3)
div 1 aamn ;mjn
e,, = R,,,iU + J,, - + Tm,,,i'"in - - -- (16-8)
defines one component of the torsion tensor T,!,,, which was encountered in at 2 a ~ ' ~
Eq. 14-24. The torque matrix G,, was given in Eqs. 13-6 and 13-14 for a in which
-
machine with one stator and one rotor winding layer.
Eq. 16-1 exhibits four distinct voltage drops in the quasi-holonomic a,, -- L,,,, the inductance considered as a component of the metric tensor
machine : J,, J , the moment of inertia considered as a tensor of valence 2
R,, r R,the frictional resistance considered as a tensor of valence 2
1. The resistance drops Rmnin
= --Gm,, the negative torque matrix considered as a tensor of valence 3
TnLn,,
2. The induced (transformer action) voltages Lmnpin
3. The generated (cutting action) voltages GmninpB Thus each term of Eq. 16-7 has been endowed with an extra index.
4. The generated (moving flux) voltages
aLmn .n
-2p 2
16-3. The Equation of M o t i ~ n . ~If,. ~now, indices (a, 8, y) are con-
802 sidered to include all electric indices ( m , n , k) and all mechanical indices
The last voltage is usually zero, although it is possible t o build a machine (u, v, w), the equation of voltage (16-1) and the equation of torque (16-8)
(a revolving iron pole without windings) which will develop such a voltage. may be combined into a single tensor equation:
16-2. The Equation of The stored magnetic energy is
e, = R,,ia + a,,
dia an,,
- + -iais- - -- 2
1 asap ., L.B + Tasyiais
at ax8 2 ax?
and, if the inductances Lmn are functions of 0, only, the rate a t which the
stored magnetic energy is changing due to the movement of the magnetic
path is
dia
= R,,ia + a,, -
dt
+ [ x p , y]iais + F,,ia ( 16-9)
and this results in equal and opposite torques on the moving magnetic mass in which a new entity, the field tensor, is defined:
and the rotor.
I n addition, there is the torque due to the interaction between the rotor
currents and the stator cross-fluxes, as given in Eq. 13-12 by Upon substitution of y = m for the free index, the term aa,g/axm vanishes
because the a,l, are not functions of the electric coordinates, and
The sum of the applied torque T, the torque (Eq. 16-5) due to varying
magnetic paths, and the torque (Eq. 16-6) due to the interaction between
the rotor currents and the cross-fluxes is consumed by the frictional torque
and the inertial torque, so that the equation of torque is since the inductances Lmn are functions only of 8,. Ordinarily this term
vanishes because the L m n are constant. Finally, if the torsion tensor is
T = R p e + J p 2 8 - y n i n - - - 1 a'mn 2 2
.m.n restricted to only two components in the quasi-holonomic reference frame,
2 ae, then
Tasy = Tunnt, Tmnu
= Rp8 + Jp28 - G,,,i"in -
1 aLmn
-- .m.n
2 r
2 ae, f Gmn,-Gmn (16-12)
Let the torque be considered as a vector e,, along the shaft with a mechanical and Eq. 16-9 reduces to Eq. 16-1 for y = m. And, upon substituting y = u
index u, and let the angular displacen~entsof the rotor, 13, and of the magnetic for the free index, Eq. 16-9 reduces to Eq. 16-8.
186 M A C H I N E ANALYSIS [Ch. 16 Art. 16-41 THE MAXWELL-LORENTZ E Q U A T I O N 187
The Christoffel symbol has only three components in the quasi-holonomic in which the afine connection is defined :
reference frame :
r a . S . , y=
t c:,c;q.[~rp,
ac:, + T ~ , ~ , , ,
+ c ; , u , ~- (16-16)
[up, yl = [un, ml, [nu, ml, Cmn, ul axst
The first two terms of Eq. 16-16 are the transformation of the holonomic
Christoffel symbol.
The pictorial representations of Eqs. 16-12 and 16-13 are shown in Fig. Eq. 16-15 may be regarded as the most general equation of motion in
16-1. The bulk of these cubes have zero elements, and even the slabs which non-holonomic reference frames. I t gives the same numerical results as
are present have many zero elements, as will be seen when particular machines Eq. 14-12. I n a holonomic reference frame the affine connection becomes
are analyzed. the Christoffel symbol.
The equation of voltage in non-holonomic systems may be found as a special
case of Eq. 16-15 upon substituting y = m for the free index and properly
identifying the resulting terms. It will prove more instructive, however,
to transform directly the quasi-holonomic equation of voltage (Eq. 16-1) :
Fig. 16-1. Components of the torsion tensor and of the Christoffel symbol
i n the quasi-holonomic reference frame.
Then the Christoffel matrix is defined; it is also reducible from the second
term on the right of Eq. 16-16:
= +
~ , . , , i ~ ' a,.,. -
dt
+ I'a.S,,,.ia'iS'
dim'
(16-15)
1 88 M A C H I N E ANALYSIS [Ch. 16 Art. 16-61 THE MAXWELL-LORENTZ E Q U A T I O N I89
where Vmn = 0 in the quasi-holonomic system. However, V,, is inserted 16-5. Equation of Voltage in Holonomic I n a holonornic
on the right side of Eq. 16-18 in order to show its law of transformation. frame the reference axes are attached to the moving conductors so that
This is easily verified by transforming Eq. 16-17 with C$ and associating pel = PO, and Eq. 16-22 becomes
with C $ V ~ . ~ , ~ : ~the
~ "term
~ B ~
contributed from
This agrees with the definition arrived a t in Eq. 15-3 by a different approach.
Eq. 16-23 may then be further condensed to
Comparing Eqs. 16-20 and 16-18, it is seen that
vmtn.
= G,.,.
Eq. 16-17 therefore becomes
where p8' = p(8 - 8,) is now the velocity of the moving conductors with PROBLEMS
respect to the reference axes. If the reference system is holonomic, 8' = 0 ,
the final term vanishes. But in all non-holonomic systems the voltage 16-1. Confirm all equations in the chapter and verify in detail every step
equation is now the same as in the quasi-holonomic system. in the derivations.
16-2. Why is there no term involving O1 in Eq. 16-l?
16-7. Transformation of the Equation of Torque t o a Holonomic
16-3. In a quasi-holonomic reference frame why are the L,, functions of
Transforming the quasi-holonomic equation of torque Eq. 16-8, O2 only?
and remembering that = 6:,. 164. Why are there only two components of the torsion tensor Tab?and
only three components of the Christoffel symbol [ap, y ] in the quasi-holonomic
frame?
16-5. Justify the equivalence indicated in Fig. 16-lb.
16-6. Obtain Eq. 16-17 as a special case of Eq. 1616.
16-7. In this chapter the equations of motion, voltage, and torque were
developed for holonomic, non-holonomic, and quasi-holonomic systems, starting
with the Maxwell-Lorentz equation in the quasi-holonomic system. Prepare a
comprehensive chart on a large piece of paper showing the appropriate equations
in each box of the chart in accordance with the following scheme:
But a,, is a function of 8, only, and c,
is a function of 8 , only and not of
I /
Eq. of Aolonomic 1 Non-holonomio 1 Quasi-holonomic I
8,. Therefore
I Mobion 1 1
I
( 1 - 1 ) ( 1 1 5 )( 1 - 1 )
Also, by Eqs. 16-12 and 16-20 and by making use of the fact that am, = a , , ; I
Voltage / (16-231, (16-24),
(16-25) i (16-171, (16-IS),
(16-22), (16-26) (lo-1), (16-2), (163)
is a symmetric tensor,
T,,,,.,,,.z
.
.7nll ac:,,
= -Gmvn,.z.m,,z.,,,,= -c;, - .mrr
Torque 1
,
(16-27), (16-30),
i
zn" am,' ' (16-31)
I
881
16-8. In the chart of Problem 16-7 fill in the blank spaces ( a )by combining
the voltage and torque equations to give the equation of motion for the holonomic
frame, and ( b ) by finding the equation of torque as a special case of the equation
of motion for the non-holonomic frame.
-
- 1 aam-,.. im"in"
2 ae,
Substituting Eq. 16-28 and identity (16-29) in Eq. 16-27,
did' 1 aamrTn.. 1 aam-,.. .,.,.
+
e , , = Ru..v,.iv" J,,,,., -- - -
imr,in,,
z an" (16-30)
at 2 ae, 2 ae,
Eq. 16-31 agrees with Eq. 14-9c, since 19, = 8 for holonomic frames.
Art. 17-51 COMPARISON OF SOLUTIONS 193
Thus all three methods of attack yield the same equation of voltage, and there
Three general methods for establishing the general equations of rotating
is no distinction. This is Faraday's law. The inductances may be functions
machines were given in Chapters 14, 15, and 16. Although each of these
of the rotor displacement 8 and of the magnetic path displacement 8,.
methods yields the complete solutions, they are in somewhat different forms
and subject to different physical interpretations. I n this chapter they will 17-3. Equation of Torque. The equations of torque as developed in
be compared and their implications discussed. This comparison involves Eqs. 14-9, 15-7, and 16-30 have the identical form :
three equations (motion, voltage, and torque), three reference frames (holo-
nomic, non-holonomic, and quasi-holonomic) and three methods of attack
-
Motion +Lagrange- Holonomic
/ Nan)-h!lonomic
2 ae 2 ae,
Thus all three methods yield the same equation of torque and there is no
-
Voltage Maxwell distinction. The inductances may be functions of the rotor displacement 8
Maxwell-Lorsntr
t. 4
4-Quasi - holonomic
and of the magnetic path displacement 8,.
I n the holonomic frame, Eqs. 17-1, 17-2, 17-3, and 1 7 4 are identical for
Torque Energy all three methods.
Fig. 17-1.
Quasi-holonomic Reference System
(Lagrange, Maxwell, and Maxwell-Lorentz). It will probably be most
illuminating to examine each equation and each method for each reference 17-4. The Equation of Motion. Both the Lagrange equation and the
frame in succession, and to number the equations as in the previous chapters. Maxwell-Lorentz equation lead to an identical form (Eqs. 14-23 and 16-9),
The interrelationships are shown in Fig. 17-1. I n this chapter, unprimed, for the quasi-holonomic equation of motion. Also, since the Maxwell equation
primed, and double-primed quantities refer to holonomic, non-holonomic, in the holonomic frame is the same as that from the Lagrange method, all
and quasi-holonomic reference frames, respectively. three methods may be said to give the same quasi-holonomic equation of
motion : dia"
Holonomic Reference Frame + +
e,,, = R,.,a,.ia" a,..,, - [ ~ " b "y"]ia"ifl"
, + Ta,.s,.y,.i""ifl"
dt
17-1. Equation of Motion. The equations as developed in Eqs. 1 4 4 , &a"
15-9, and 16-15 all reduce to +
= R,,,,,,.ia" a,,,,., - + [ul',Y"'r"]i""ifl"
+ Fy,,,,,iu'' (17-5)
dt
dia
e, = R,,iU + at
+
a,, - [up, y]i"is (17-1) 17-5. The Equation of Voltage. The Lagrangian method led t o the
equation
for a holonomic system, since the affine connection as given by (10-61) and
Eq. 16-16 for a non-holonomic system becomes the Christoffel symbol in e
holonomic system. Thus all three methods of attack yield the same equation
in the holonomic frame, and there is no distinction.
192
1 94 M A C H I N E ANALYSIS [Ch. 17 Art. 17-81 COMPARISON OF SOLUTIONS
The Maxwell method led to the equation Also, for the same reason, and upon interchanging dummy indices,
The Maxwell-Lorentz derivation started from as in the final expression of Eq. 14-12.
But the CE, included in the non-holonomic objects of Eq. 14-12 refer to
em. = Rm.,,.,in"+ d+m.
-+ y m 4 d = . .
dt
- (16-1) the transformation from the holonomic frame and are not the same as the
C: included in Eq. 16-16, which refer to the transformation from the quasi-
and in its development embraced Eqs. 14-25 and 15-11. holonomic frame. Consequently, it is not possible to make a direct compari-
Therefore all three methods yield identical equations of voltage, which son between Eqs. 14-12 and 16-16, and the proof that they give identical
take on a variety of forms as shown in Eq. 16-1. results must rest on the argument:
17-6. The Equation of Torque. The Lagrangian method led t o the 1. The holonomic equation (14-4) transforms to the quasi-holonomic
equation equation (14-23), which is identical with Eq. 16-9.
1 aL,.,. .,-.,st
+ - - -2 z
T = RpO Jp20 - Gm,,n,.im"in" (14-29)
2 ae,
The Maxwell method led to an identical equation (15-15).
The Maxwell-Lorentz method established an independent equation of
Holonomic
Eq. 14-4
Non- holonomic
-- 14-213
Quasi- holonomic
Eq
torque (Eq. 16-7) of identical form.
Thus all three methods yield the same equation of torque. \ Eq. 16-15
Eq. 14- 12
Fig. 17-2.
Eq. 16-9
17-7. Equation of Motion. Each method of attack gives the solution 2. The holonomic equation ( 1 4 4 ) transforms to the non-holonomic
in the form equation (14-12), and the quasi-holonomic equation (16-9) transforms to
the non-holonomic equation (16-15).
3. Therefore Eqs. 14-12 and 16-15, although different in form, must
give identical results.
but the affine connection has a different definition for one of the methods
than for the other two: This can be diagrammed as illustrated in Fig. 17-2.
Lagrange : ra,p,,y, + Qy,a,,p,- Ra,p,,yt
= [ a l l l ,yl] - Rpry.,a. 17-8. Equation of Voltage. The equations of voltage developed in
Eqs. 1 6 1 6 and 16-17 from the holonomic and quasi-holonomic frames,
respectively, have the forms
+-aamTnP
802
in$& + Gm,,.inpO (16-17)
However, since the non-holonomic object is skew-symmetric in its first two
indices, Here, again, the E,is different from the E:and a direct comparison be-
!&p.,yia'iB' = 0 (17-7) tween Eqs. 1 6 1 6 and 16-17 is not possible. However, by letting zu' in
196 M A C H I N E ANALYSIS [Ch. 17 Art. 17-10] COMPARISON OF SOLUTIONS 197
Eq. 14-16 take on the particular values 8, €4, 8, and defining G,.,. in Eq. Each observer measures exactly the same total voltage, but the compon-
16-17 as in Eq. 16-20, the two equations become ents into which i t is divided are quite different for the two observers. Indeed,
the moving observer measures a component of voltage V m t n . i n ~ Owhich 1
din' aa,,,,
em, = Rm,,.in' + a,.,. - + -i n'p8 the stationary observer does not see a t all. If the magnetic paths (iron
at ae masses) move a t the same speed as the rotor, p8, = p0, the two voltage
components involving p02 and p8 cannot be separated by either observer.
TABLE 17-1
Lrnnpin
II)
d Q
Thus both equations present the same type of terms. By the same argument 3 2 Axes in which Stator and
Rotor axes
measured: rotor axes rotor axes
as for the equations of motion, the two equations must give identical results. p
8 2 Consider:
Stationary Stationary
17-9. Equation of Torque. The equations developed in Eqs. 14-17 and Ref. axes Stationary
Conductors Stationary Moving Stationary
16-27 have the forms )
Stationary Moving
Magnetic path1 Stationary
aL,.,.
pi"' a,.,.in'pO -
1 aarn,,,,,. ., a,*
e,, = R,,,,iU' + a,,,, dig'
-
at
- --
2 axu'
2
.n,
z + T,,,.,,z,,. 2
Stator and Stator and
Rotor axes
Stator and
measured : rotor axes rotor axes rotor axes
But, by the definition of the torsion tensor (Eq. 14-24), and in terms of the Stationary Moving Stationary Stationary
non-holonomic objects (Eq. 14-22),
1 Magnetic pathi
Currents
Stationary
Stationary
Varying
Stationary
Stationary
Constant
Moving
Stationary
Constant
Stationary
Moving
Corstant
And, if the magnetic paths, rotor, and reference axes all move a t the same
Thus Eqs. 1 6 1 7 and 16-27 are identical in form and give the same results.
speed, pB2 = pel = p8, both observers measure a single generated voltage,
Or, as shown in Chapter 16, but neither is able t o separate i t into its parts. The four voltage components
of the moving observer and the three voltage components of the stationary
observer add up to exactly the same value, but the way in which this total
voltage is distributed among the components is quite different for the two
t7-10. The Components of Voltage. A comparison of the equations observers, and i t can be changed a t will merely by changing the arbitrary
of voltage for stationary axes (Eq. 16-1) and moving axes (Eq. 16-17) brings speed, pel, of the reference axes. Whether a voltage is induced by a varying
t o light two different physical interpretations, although both equations give flux, or generated by a moving flux, or generated by moving conductors,
identical numerical results. The different components of voltage, other than or generated by moving reference axes is to a certain extent a matter of
resistance drops, as measured by an observer on stationary (quasi-holonomic) the selection of the reference frame, and one type of voltage when viewed
axes and an observer on moving (non-holonomic) axes, and the corresponding from one frame is an entirely different type of voltage when viewed from a
physical interpretations, are given in Ta.ble 17-1. different frame.
200 M A C H I N E ANALYSIS [Ch. 17 Art. 17-1 21 COMPARISON O F SOLUTIONS
17-1 1. Recapitulation of the Basic Concepts. The different concepts The inductance of the holonomic machine therefore is
introduced by the three methods of analysis used in deriving the equations
of voltage, torque, and motion are recapitulated here in Table 17-2 for ready
reference, and classified with respect to their origin, equation of definition,
and law of transformation. The = H Q means that the relationship is true
or =
only in a holonomic or quasi-holonomic system, respectively.
17-12. The Two Preferred Generalized Machines. In the great
b -Mdsint9 -LDsin28 L , - LD cos 28 M , cos 8
majority of practical cases, the analysis of a machine is the simplest and
most clear-cut if referred to either a holonomic or quasi-holonomic reference
qs 0 M , sin 8 M, cos 8 1
I
Lo,
where Ls = (L,, +
L,,)/2 and LD = (L,, - L,,)/2.
The impedance matrix is
ds n b PS
+
r , ~ pLd, ~ , cos
p o I
I - ~ , psin o 1 o
a M d p cos 8 r. + p ( L , + LD cos 2 8 )- 1/ - LDp sin 28 filQpsin 0
-
( a ) Quasi holonomic ( b ) Holonomic !la 0 ~ M Q psin 8
--
&fopcos 8
1
+L ~ P
.Fig. 17-3. The two preferred generalized machines. (17-13)
where p = dldt refers to both the 8 terms and to the currents. For example,
frame. There are a few cases, however, in which more general non-holonomic
frames are of advantage, either because of construction details or because Z,,,,ib' = - LDp(sin 28 ib') = -LD(2ib'cos 26pO + sin 20pib')
of special operating conditions in a machine. But, for most cases, one or the
Eq. 17-13 may be verified by the general transformation formula (in
other of the two generalized machines shown in Fig. 17-3 can be adapted.
which p refers only to the currents),
The equations of either may be converted to those of the other by the
transformation tensor, and i t *is immaterial which machine is used as a
starting point. The conditions affecting a choice will be made clear in
Chapters 20 through 23. That Eqs. 17-13 and 17-14 give identical results is easily shown by expanding
The transformation matrix between the two machines and the inductance as follows (see Eqs. 16-23, 16-24, 16-25) :
matrix (Eq. 13-6) of the quasi-holonomic machine are (for 8, = 8):
= + ~ , , , , p + Gm.,,p8 + vrn.,,pe)in'
(R,,,.
= Rm,,,in' + p(L,,,tin') (17-15)
L,, =
Therefore
Z m t n= [ C,m.C~.Zmn LrnnC,m. +
p refers only to in'
m +P
p refers to 8 and in'
M A C H I N E ANALYSIS [Ch. 17
PROBLEMS
17-1. Confirm all equations in the chapter and verify in detail every step in
the derivations.
17-2. Check every entry in Table 17-2 against the chapter and equation
in which it first appeared.
17-3. Prove that the a f i e connection as given in Eq. 1 6 1 6 reduces to the THE RAISING A N D LOWERING OF INDICES
Christoffel symbol in a holonomic system.
174. The voltage in the brush axis of any machine contains a "generated A N D GENERALIZED PER-UNIT CONCEPTS
voltage" term which depends upon the velocity p6 of conductors$xed in position.
Discuss the philosophical implications of endowing a stationary conductor with
a velocity. There is more to this than meets the eye.20
18-1. The Raising and Lowering of in dice^.^ The metric tensor g,,
was defined in Eq. 9-59, and its inverse gap in Eq. 9-65, such that the invari-
ant infinitesimal distance between two points in general curvilinear space is
given by
= gas dxb dxB = gas dx, dxs = dxs dxs (9-6 1)
It was also shown that the metric tensor has the property of raising or
lowering indices; thus
g,As = A, (9-73)
I n an electric machine the stored magnetic and kinetic energy is given by
and by analogy with Eq. 9-61 the asp, which include both inductances L,,
and moments of inertia J,,, may be adopted as the metric tensor with the
property of raising and lowering indices.
The definition of the flux linkage (Eq. 13-11) may be extended to include
mechanical as well as electric variables (that is, 4, represents both flux linkage
and momentum) by
4, = aasis = i , (18-1)
This relationship shows that the covariant form of the current ia is the $uz
linkagf 4,. The energy may then be written in the following equivalent
forms :
2T = a,,ibifi = i,ia = aaB4,4, = 4,y = 4,ia (18-2)
in which
--
The terms of Eq. l P 1 2 represent voltages and torques, but the terms of in which appear
Eq. 18-3 represent rates of change of currents and accelerations. The latter
equation greatly simplifies calculations, since the inductance and inertia
L: = (LsLp- = short-circuit inductance (18-8)
coefficients have disappeared, leaving the coefficient of piY simply unity. LT
Moreover, as will be shown, the resistance coefficients have been replaced by
reciprocal time constants, and the aEne connection is made up of inductance mutual inductance
il = = leakage coeflcient
ratios only, self-inductance
18-3. Conversion of the Equation of V ~ l t a g e . ~ The equation of
voltage for the quasi-holonomic generalized machine, if there are no moving
magnetic paths, is
The mixed tensors R: and G.2 are
Substituting
i n = ank$k
and remembering that the a,, are constant, and that amnank= d:,
in which G: is seen to be the same as the rotation matrix y: upon comparison in which appear the coupling coeficients in place of the leakage coefficients:
with Eq. 13-18. Consequently, by Eqs. 18-12 and 13-19,
A comparison of the impedance matrices Z,,, 22, and 22; discloses the
The mixed impedance tensor, for pB = constant, follows from Eq. 18-14 as following :
ds dr qr Pa
,Z
, has 9 design constants: r,,, r,,, r,,, L,,, L,,, L,,, L,,, Md, M,, and
ds +P -&,6,* 4 zero components, and the coefficients of p and p8 are not unity.
2 2 has 8 design constants: A,,, Ad,, A,,, A,,, d,,, d,,, d,,, d,,, and 6 zero
- s,, +p PO
components, and the coefficients of p and p8 are unity.
z: = R: + + aipe=
!IT -PO +P -&Jar 22: has 6 design constants: l;ld, q,, dds, Bar, d,,, 6,,, and 6 zero components,
and the coefficients of p and p8 are unity.
- 6,. +P
!
Moreover, the design constants in 22 and 22: are all ratios which have
(18-15) the following considerable advantages over actual resistances and inductances :
This impedance can be expressed in terms of other design constants, and
further simplified, if it is transformed by 1. These ratios are of the same order of magnitude for machines of all
sizes, and consequently in design or performance calculations there is less
ds' dr' qr' qs' chance of making a decimal error.
2. These ratios are not nearly so sensitive to the effects of saturation as
are the inductances themselves.
3. Within reasonable design limits, the decrement factors depend only
on the copper in the windings, and the coupling factors only on the iron in
the magnetic circuits, so that the effects of design changes are more easily
anticipated and determined.
4. Graphical performance curves are more naturally and simply plotted
whereupon the new voltage, flux, and impedance matrices become for a whole line of machines in terms of 6 and as parameters.
5. A unified graphical analysis (vector and circle diagrams) can be
established corresponding to the concepts of Eq. 18-19 and using only the
6 and q as constants.
PROBLEMS
18-1. Confirm all equations in the chapter and verify in detail every step
in the derivations.
18-2. Are the Christoffel symbols shown in Eq. 18-4 non-holonomic or
holonomic?
18-3. I n the matrix of Eq. 18-7 show that the following mutual terms are
equal.
- A d ~ / L A ~= - A d s / L A ~
-'w/Lis = -'qs/L&
208 M A C H I N E ANALYSIS [Ch. 18
I n general, the design constants Rap and a@, as well as the transformation
matrices C:. included in the affine connection rSr,a, are functions of the
mechanical coordinates xu.
Suppose, now, that for a machine or a group of machines there suddenly
is superimposed on a condition of steady motion or of uniform acceleration
abrupt changes in the applied torques and voltages, Ae,. As a consequence,
corresponding changes hia occur in the currents and velocities of the rotor.
If the reference axes are attached to the rotor, or move a t a speed depending
on the applied frequency, they too will oscillate about their instantaneous
mean position.
The first-order variation of any quantity (ia, e,, Rap,aa8, rS,,,) in Eq.
19-1 is of the form
The first-order terms in the Taylor series expansion of Eq. 19-1, that is, the
first dqrivatives, become
Ae,
sea AxS = RaSAiS+ a,, -
+-
axs dt + r,,,,(iflAiY + hi%,)
d Aib
= RaSAib + a,, d-
Aib
dt
+ rsY,,(isAiY+ iYAiS)
210 M A C H I N E ANALYSIS [Ch. 19 Art. 19-31 SMALL OSCILLATIONS A N D H U N T I N G 21 1
I n this equation, Ae, represents a suddenly applied external voltage or torque ; The equation of small torques follows from Eq. 19-5 upon substituting
(aea/ax@)Ax@represents the variation in voltage or torque brought about by mechanical indices for the free index, u = u,so that
the oscillation of the reference axes. The f i t four terms on the right repre-
d hiv
sent the changes due t o the oscillating currents and velocities. The last Ae, = RuVAiv'+a,, -
dt
+ l?,,,,Ainim + rmn,,inAim
three terms on the right represent the changes due to the oscillating reference
axes. (or, if the magnetic paths do not move)
In a quasi-holonomic reference frame (stationary axes), the design par-
ameters remain constant and the last three terms of Eq. 19-3 vanish.
I n a holonomic reference frame, the affine connection becomes the Christof-
= R,,AiV + a,, -dAiv
-
dt
+ AFd,im + F,,Aim
fel symbol, which is symmetrical in its &st two indices, so that (and, if there is only one machine)
+
rSy,a(iDAiY AiW) +
[By, a](i@AiY AiSiY) = 2[&, a]iYAiS
The equation of small voltages follows upon substituting electric indices for
the free indices, u = m, so that
he, = +
RmnAin a,,
dAin
-
dt
+ l',,,,AiniU + l',,,,inAiU
(or, if the magnetic paths do not move) The changes in the currents and velocities are
dAin
= +
RmnAin a,, -+ ARp,iu
dt
+ FIm,,Aiu
The motional impedance transforms from the quasi-holonomic system in
(and, if there is only one machine) accordance with the law
d hin
= RmnAin + a,, -+ GmnAinptl+ Gmnin(ApO)
dt
19-3. Classification of Comp~nents.~For the affine connection
dA4m + Aymy,pB+ V,(AP~)
= RmnAin +-at
substitute (Eq. 1 P 2 3 )
rpy,a '
[ P Y ~a] + Tpya (19-13)
212 M A C H I N E ANALYSIS [Ch. 19 Art. 19-41 SMALL OSCILLATIONS A N D H U N T I N G
in Eq. 19-3, which then becomes Substituting Eq. 19-18 in Eq. 19-17,
he, +-
8%
axS AxS = Rap hiS+ a,, - + ([By, + TSy,)(iSAiY
d Aip
at
+ hi%?)LX]
oscillate, the equations of small voltage and torque result: i t follows that
C,",AC,Y'e,,= -C,Y'AC:,e,, = -AC:.e, (19-20)
Ae, + axu
aem Ax" =
-
~ , , , i ~ ~ i ~
+oscillating conductors+ 1 Since the mechanical coordinates are not transformed,
RmnAin a,, + -
dt
+ [nu, m]iYAin+ TunmiYAin
oscillating currents-
["-.
I axtL
i n Ax" + --
aamndin Axu + [nu, m]inAiu
axu at 1 and therefore Eq. 19-19 becomes
refers only to hia, this equation takes the form (since only the a,, terms of 19-5. The Generalized Machine (Stationary Axes).7*8 The motional
Tap have p coefficients and C;,Z,,iy' = e,) impedance for a single machine with stationary (quasi-holonomic) reference
axes was given in Eq. 19-10. Evidently, this matrix consists of the transient
impedance matrix Zmn bordered by the matrices Gmnin,-(Gmn +G,,)i*,
and Jp. For the simple generalized machine (by Eqs. 13-3 and 13-14),
Substituting this in Eq. 19-25 and rearranging the terms, there results the
classification
(
cI oscill. currents currents varying moving ref. axes moving conductors The steady-state currents in the bordering row and column are those
E forced oscillations of reference axes f existing a t the instant of disturbance and are found from the solution of the
+ C~,R,,AC:,i'" + Cz,LmnAC,",pin'+ ~ , 7 , L , , ( p A q , ) i " ' + C ~ , O m n p 0 A C ~ , i d
3g ' resistance drop currents varying moving ref. axes moving conductom
transient impedance equation,
em = Zmnin or in = (Zmn)-le, = Ynmem (19-29)
m + AC:.RmnC:,in' + AC:,L,,,,C;,pin' + AC:.L,,(pC;,)in' + A C ~ , O , , , , p O ~ , i d
resistance drop currents varying moving ref. axes moving conductors The velocitypfl is the constant speed vo existing at the instant of disturbance.
I
216 M A C H I N E ANALYSIS [Ch. 19 Art. 19-61 SMALL OSCILLATIONS A N D H U N T I N G 217
For any machine with stationary reference axes, The equation of small orjcillations (19-3) is then equivalent to
de, 3% d 2 = d(RaBip)+ d
+- (19-36)
a2
For any machine with holonomic reference axes, But, as was pointed out, this is not an invariant equation. However, by
replacing the ordinary differentials with absolute (intrinsic) differentials,
there results a n invariant form for the equation of small oscillations :
I n either case, the steady-state currents in the bordering row and column
must be replaced by
i m = C;,im' (19-32) Now, remembering that the intrinsic differential of a product follows the
same rules as ordinary differentials,
so that only the new currents im' for the particular machine will appear in
its motional impedance.
19-6. Invariant Form of the Equation of Small oscillation^.^^ The
equation of small oscillations (19-3) was obtained by taking the ordinary
derivative of the equation of motion (19-1). But in this form no single term
of Eq. 19-3 is a tensor, nor can any combination of the terms be found which
is a tensor. Moreover, the transformation to other reference frames becomes
complicated and involved. It is therefore desirable to set up the equation since 6aap = 0. The (6/6t)6isterms were added and subtracted arbitrarily
of small oscillations in an invariant form such that all its terms are tensors for reasons which will soon become apparent.
and transformations may be effected by simple tensor transformations. Substituting Eq. 19-34 in Eq. 19-38, th2re results
The intrinsic derivative for a holonomic reference frame was defined in
Eq. 10-19 in terms of the Christoffel symbols. This definition is extended
to nonholonomic reference frames upon replacing each Christoffel symbol
d!( with the corresponding affine connection FtY. The two are equal in a
in which the quantity in parentheses defines a new "resistance tensor of rank
holonomic frame. I n general then, from Eqs. 10-19 and 10-24, the absolute 3" designated Rvpa.
or intrinsic derivatives and differentials are defined as Applying the definition (Eq. 19-33) t o the terms of Eq. 19-39, there
results
dip -
-dip
-+ r:,i" dxy dip
-
6t
-
dt
- = -+
dt dt
rS imiy
by
= R,$ + a,, -
6ip
6t
218 M A C H I N E ANALYSIS [Ch. 19 Art. 19-71 SMALL OSCILLATIONS A N D H U N T I N G 219
The difference between Eqs. 1 9 4 1 and 1 9 4 2 is 19-7. Sinusoidal disturbance^.^.^ Suppose that the disturbance is a
sinusoidal variation ho, where h is its ratio to normal frequency. Such
a disturbance may be due to a variation of shaft torque in a motor driving a
reciprocating compressor, or to the variation of terminal voltage on a motor
= ~,;;&'i&
dxY (1943) connected to a power source containing harmonics.
in which there has been defined a new tensor, KGifl, called the "generalized Where the currents in the bordering row and column are constant (this
Riemann-Christoffel curvature tensor." Its covariant form is case includes all d-c machines, synchronous machines with stationary axes,
aU S K &YA
" . @ = K&Y .la (1944)
and induction or commutator machines whose reference axes move with the
Substituting Eq. 1 9 4 3 in Eq. 19-39 and then substituting Eqs. 19-39 TABLE 19-1
and 19-40 in Eq. 19-37, there finally results as the invariant form of the
Fundamental, w Hunting, o h Double, w ( l f h)
equation of small oscillations
em, im
Oscillating
rotating air gap field), the current and velocity changes Aia are all of hunting
Now in rotating electric machines the affine connection is a function
frequency h o corresponding to the torque and voltage changes Ae, of the
only of the mechanical variables, and therefore i t can be differentiated only same frequency. Then in the motional impedance matrix the following
with respect to the mechanical variables, xu = 8. changes are made :
The curvature tensor (Eq. 1 9 4 3 ) is skew-symmetric in its f i s t two indices
as can be seen by interchanging E and y in its definition; that is, 1. p = jho where
oscillation frequency
h=
Moreover, must always be a mechanical index, u, as can be seen by writing fundamental frequency
out the affine connections in full, according to identity 10-61, remembering 2. Multiply all terms in the bordering column by w and compensatingly
that the metric tensor ampis a function only of the mechanical variables. divide the velocity change Ape by o so that
Furthermore, either E or y must be a mechanical index for the same reason.
From the observations above i t is found that the only curvature tensors
A @ = ~ 2 ) 0 =v =
~ speed variation
which need t o be calculated are w o synchronous (or average) speed
3. Multiply all terms in the bordering row by w so that the torque change
AT is measured in synchronous watts :
w AT = A ( o T ) = AT,,,,.,
The equation of small voltages then is (putting u = m) I n those cases where the steady-state torques and velocities are constant
and the voltages and currents are of fundamental frequency, the torque and
velocity changes will be of two different frequencies, m(1 f h) (see Table
19-1). Therefore the motional jmpedance equation must be solved by separat-
The equa,tion of small torques then is (putting u = u) ing the electrical and mechanical quantities:
de, 6% dxu = Ruv 6iV+ a,,
+-
6x"
-
6t
+
6 6iv RyuuiSdxY f Keyu,iEiu
dzy (1948)
PROBLEMS
19-1. Confirm all equations in the chapter and verify in detail each step
in the derivations.
Art. 20-31 SYNCHRONOUS MACHINES 223
SYNCHRONOUS MACHINES7*' in Eq. 20-1, i t is seen that the only armature voltage is
20-1. The Transient Impedance (Stationary Axes). The analysis for along the q'-axis of the infinite bus, and i t is constant.
a synchronous machine may start from either the first (quasi-holonomic) or If this voltage lags the generator q-axis by an angle 6 , the voltage matrix
second (holonomic) kind of generalized machine, but it is generally simpler of the generator is
to start with the first kind. The impedance for a generator with C.W.rotation,
direct axis salient pole field (df), amortisseur windings on both axes (dk, qk),
and stationary armature (da, qa) is found from Eq. 13-5 by changing all
signs (to convert from motor to generator) and adding additional terms for
the amortisseur windings, thus : Thus the voltages along the d- and q-axes are constant, or d-c voltages, if 6
df dk qk da '?a is constant.
f
L
I-
L P ~ ~ f k p --"fd~
f -
I- where v is a fraction less than unity,
dk -"fk~ -'kd -Lkd~ -"kd~
qa -MfdpB -MkdpB - M ~ P I
- L , ~ B - r - Lap df dk qk da !?a
The matrix has been arranged differently from that of Eq. 13-5 in order to E 0 0 - e sinsot e cos sot (204)
facilitate a future elimination of axes. If there is no amortisseur winding in
an axis, the corresponding row and column are deleted. If a current is Under sudden short-circuit conditions, the cancellation of the armature
known t o be zero (for example, an open-circuited winding), the corresponding voltages by the short circuit is equivalent to the application of unit functions
column is deleted. (in the Heaviside sense), and the voltage matrix is
/d'
df 20-2. The Steady-StateVoltages. Under
steady-state conditions, the voltage on the field
winding is d-c, and the voltages on the short-
circuited amortisseur windings are, of course,
zero. It is customary to refer the armature 20-3. Subsynchronous Speed. If a machine is connected to a n infinite
voltages t o an injnite bus, assuming the field bus but is running below synchronous speed, its speed is p0 = vw, where
of the alternator to be ahead of the bus field v is a fraction (v < 1.0). The fundamental frequency currents applied t o the
$q'
U by the torque angle 8. The angle 6 is positive armature develop an mmf rotating in the forward direction a t synchronous
for a generator and negative for a motor, and speed with respect t o the stationary armature. But, since the field poles
its magnitude is proportional t o the load on (and therefore the d-, q-axes) are rotating forward a t speed vw,the armature
Fig. 20-1. Machine on an
infinite bus. the machine.
[Ch. 20 Art. 20-51 S Y N C H R O N O U S MACHINES 225
224 M A C H I N E ANALYSIS
That is, there are no currents or voltages in the amortisseur windings, and the
mmf is rotating forward a t speed (w - vo) = s o , or slip frequency (s < 1.0),
corresponding rows and columns of Eq. 20-1 may be deleted. The equation
with respect to the d - , q-axes. Therefore the voltages induced in these axes
of voltage is simply
by this armature mmf are a t slip frequency, and in all hp terms of the
da
transient impedance matrix i t is necessary to substitute
and for the speed voltages, represented by the Lp6 terms, i t is necessary t o
substitute
Lpe = Lvo = v x
1
df dk qk da other axis.
df - r f - jaX, -jaXfk -jsXfd The admittance matrix is
-- -- (20-12)
from which i t is seen that
idk - %.ak = 0
-
226 M A C H I N E ANALYSIS [Ch. 20 Art. 20-61 SYNCHRONOUS MACHINES 227
Solving Eq. 20-19 for the currents by multiplying both sides by the
inverse of the impedance matrix,
When the resistances are neglected, the operational impedances are called
the direct and quadrature subtransient reactances. 204. Equivalent Circuits of the Operational Impedances. If all
The terminal voltages, by Eq. 20-12, are mutual inductances in an axis are equal, and self-inductances are expressed
ed =G ~ LrE
( ~ )- + f Lg(~)~8ig
L d ( ~ ) ~ l i d
(20-14)
eg = G(pIp8E - Ld(p)p0id- [r L9(p)plia +
The torque exerted on the stationary armature is then the sum of the
products of the p8 voltage terms and the corresponding currents ; thus
( a ) Direct axis
When t,here are no amortisseur windings, the inductance expressions
reduce to
( b Quadrature axis
Fig. 20-2.
in which is defined the Jield time constant,
as the sum of leakage and mutual inductances, then Eqs. 20-13 may be
To=- bf expressed in the form
rf
and the short-circuit inductance,
'J f
in which ef = (e, +
je,) and eb-= (e, - je,). The torque matrix follows from
Eq. 20-19 upon transforming the terms having v as a coefficient; i t is
This matrix can be converted t o a symmetrical form by the following
substitutionls of forward and backward current components (if, ib)
=B
= +(;fa 3
iqa - ba
i )I- Q" fa +j(Xod+Xw)ifk+j(Xod-Xw)ibk+
2jXadidf j ( X d + X u ) i f a +j ( X d - X u ) i k
4
ba -2jX,idf - X u ) i f a - j ( X d + X , ) i b a-
-j(X,-X,)ifk-fiX,,+X,,)P-j(Xd
This is the rotating field, or 2-phase symmetrical component, transformation
of synchronous machine theory. * (20-29)
If the impedance (Eq. 20-23) is transformed by Eq. 20-24 as c , Z m n C , . The torque, per phase, is
and a t the same time the voltage equations (rows of Zmtn,)are divided by s
+
for t,he df, fk, and bk, by (s v) for the fa, and by (s - v) for the ba rows,
Since the impedance matrix of Eq. 20-27 is symmetrical, i t may be
represented by the equivalent circuit of Fig. 20-3a. This may be verified by
tracing through the meshes in accordance with Kirchhoff's second law (after
putting Xd = Xad
and Xf = Xu, +
xf).
+ +
X,, Xq = Xuq Xt, Xkd = Xad f Xkd, Xk, = Xaq Xkq, +
df fk bk fa ba
The cross-flux (Eq. 20-29) is represented in this equivalent circuit by the
idf I jfk jbk j jfa jba (20-26) voltages E; and Ei, as can be verified by tracing out on the diagram.
I
230 M A C H I N E ANALYSIS [Ch. 20 Art. 20-a S Y N C H R O N O U S MACHINES 231
In addition to the currents flowing in response to the applied voltages ef
and e,, there are the currents flowing in response to the d-c excitation E. The
amortisseur currents ifkand ibkare zero, and the other currents are constant
under this condition. Therefore, putting 8 = 0 in Eq. 20-23, deleting the
k-axes, and transforming by Eq. 20-24,
df fa ba
(20-34)
The equivalent circuit for the d-c excitation is shown in Fig. 20-3b.
The torques (in synchronous watts) which would be measured by watt-
meters in the equivalent circuits with armature applied voltages (W, and
W,), and with field voltages (W, and W,) are given in Pig. 20-3, and their
sum is the total torque. When both field and armature are excited, the two
sets of fluxes and currents produce slip-frequency torques, given by the
( b ) Equivalent circuit for field voltage watts and vars W, + jQ5, etc. The crest of the oscillating torque per phase
Fig. 20-3. Equivalent circuit of the salient pole synchronous machine at is To.
non-synchronous speed. In machines with smooth air gaps, X,, = X,,, the capacitance element of
Fig. 20-3 becomes a short circuit.
232 M A C H I N E ANALYSIS [Ch. 20 Art. 20-81 SYNCHRONOUS MACHINES 233
At synchronous speed, s = 0, three of the branches in Fig. 20-3 become By the general transformation formula for the impedance, and for 8, = 8,
open circuits, and the two circuits may be combined as shown in Fig.. 2 0 4 .
The presence of a negative resistance in this network does not permit a
direct representation on a calculating board.
The elimination of the field axis in Eq. 20-31 and the elimination of the
amortisseur windings lead to more simplified equivalent circuits, but the ds a b q8
matter will not be pursued further here. rd, + Ld,p M,(cos 8 p - sin 8 p8)
1 -Md(sin8p+cos8p8)
0 I
Hebce, with the understanding that p refers to both the currents and 8 terms,
The transformation matrix to rotating holonomic axes fixed to the slip
ds a b 'P
rings, and its derivative are ,
czs rds f L d s ~ pMd cos 8 - p M d sin 8 1 0
ds a b US cls a
a p M d cos 8 r + p ( A + B cos 28) - p B sin 28 p M , sin 8
z,,,, = .
b -pMdsin 8 - p B sin 28 + P ( A - B cos 2 0 ) P M , cos 8
A - B cos 20 M , cos 0
a,.,. = c:,c;,a,* =
b - M d cos 0 - A - B cos 20 B sin 20 -M , sin 0
ds a b
~ f , ,cos (6 + 1
240')
a1 Mds c o s 0 1 Ll +
4 cos 26
Ll +
L, cos 2(0 + -MI +
M , cos 26
/1 A&, sin (6 +
120") I
.I;II '
where
Jfd, cos (0 + -MI + Ll + I M,,, sin (0 +
L , cos 2(0 + ! 240') The holonomic inductance matrix checks (Eq. 2 0 4 6 ) exactly when
M 2 cos);:2: 240') computed by
DlQ,sin (6 + I M,. sin (6240')
120°) I
+ 1 L,, L,.,, = C,".CEfLrnn (20-48)
Comparing the two methods, it is evident that the transformation from
the quasi-holonomic frame required little thinking beyond establishing the
Now let this be confirmed by a transformation from the quasi-holonomic transformation matrix and leads to the final result by a routine process.
frame. I n that frame the three phases must be represented by three separate Moreover, the second harmonic terms came in automatically, But setting
rotor windings, as shown in Fig. 20-6a. The inductances are all constants, up the inductance matrix directly in the holonomic frame required consider-
and the transformation matrix follows upon comparison of Figs. 20-6a able appeal to the physics of the situation and, in particular, careful deter-
and 20-6b. Thus mination of the stator and rotor mutual inductances and recognition of the
second harmonic variation of the rotor self-inductances. I n all machines
ds drl dr2 dr3 qr3 qr2 qr1 qs whose analysis permits a quasi-holonomic reference frame, it is generally
simpler to start from that frame.
I n treatises on synchronous machines,14 it is customary to replace the
self- and mutual inductances by "3-phase" values, or some other combina-
tions, but no purpose is served here by making these substitutions.
I
?he holonomic torque matrix is
-- I _ _ _-- --
I
The holonomic impedance matrix and voltage equation are
z,, = +
(Rrnpn, PL,,,,) (20-49)
e m , = R,,,,in' + p(Lrn,,.in') (20-50)
From here on, the analysis of a synchronous machine follows the pattern
first formulated by R. H. Park and given in detail in Concordia's book.14
20-1 1. Interconnection of Synchronous Machines.'y8 When two or
more synchronous machines are interconnected, a transformation matrix
238 M A C H I N E ANALYSIS [Ch. 20 Art. 20-1 I] S Y N C H R O N O U S MACHINES 239
may be set up from the diagram of connections representing the intercon- It is to be noted that this final result could have been written directly
nections in terms of the slip ring currents, that is, referred t o holonomic axes: from Fig. 20-7 by placing the (dr,, qr,) axes of the first machine on the sketch
im= C'",jm' (20-51) of the second machine an angle 6 ahead and reversing the currents.
The same matrix results for rotating field machines if the poles of the
Both the old currents im(comprising all currents in the group of machines)
first machine are an angle 6 ahead of those of the second machine.
and the new currents im'are with respect to moving axes. Each of them may
be referred t o stationary axes by appropriate transformations :
c;,,jm"
jm = (20-52)
cr'p
im3= (20-53)
Then
jm" = c;"jm = c;"c;,jm' = c,rn"c~,cr';a
= c,mP'ja (20-54)
,4s an example, consider the two simple synchronous machines, say a
generator and a motor, of Fig. 20-7. The second machine is running a t a n
angle 6 = 0, - 0, ahead of the first machine. The transformation matrices
of Eqs. 20-51, 20-52, and 20-53 are then (only the rotor axes need to be
considered in these transformations, since the stator currents are independent
in the two machines and do not change to new values)
Fig. 20-7. T w o synchronous machines.
machine axes,
c;, = The voltage and impedance matrices for two machines, after the field and
stationary axes amortisseur axes have been eliminated, follows from Eq. 20-12 :
\ mf \ m/!
cr,] cos 0, sin 13, I I
bl 1- sin 0, cos B,
8
1
I I
(1, I cos 0 , 1 sin 6 , I
C,"' i'"'paxes,i
moving to
stationary
= ;:t cos
--
-sin 0,
el sin
cos 0,
1 d l edl - G,(p)pEl - ed2cos 6
q~ egl - G,(p)pO,E,
+ G,(p)pE,cos 6
+ ed, sin 6 - G,(p)pE2sin 6
-
-
eq2sin 6
e,, cos 8
+ G Z ( P ) P ~sin~ E6Z
+ Gz(P)P&Ezcos 6
The transformation matrix of the interconnection in terms of stationary (20-58)
zap= c;"c;"z,," =
axes then is (by Eq. 20-54) dl u1
-,7
-cos 6 sin 6
240 M A C H I N E ANALYSIS [Ch. 20
Art. 20-131 SYNCHRONOUS MACHINES 24 1
However, in Eq. 20-58 e,, cos 6 = e,, and e,, sin 6 = -e,, so that some with stationary reference axes the transformation matrix has constant
terms are canceled. elements Cz, = 6:, so that Eq. 19-23 reduces to
If the machines are similar and running a t the same speed, so that Gl(p) =
G,(p), L,, = L,,, r, = r,, and pel = PO,, the equations simplify considerably.
20-12. Ignoring the Quadrature Axes.8 For balanced voltages and During hunting, the angle 6 varies, and, since 8 = ot + 6,
smooth air gap machines the armature voltages and currents in the q-axis are
identical with those in the d-axis, but occur 90' later in time. The q-axis
may then be eliminated by the polyphase matrix of Eq. 11-17:
The transient impedance matrix for a synchronous machine with rotating
field and amortisseur windings is given in Eq. 20-1. The torque matrix
comprises the terms in Eq. 20-1 having p 8 as a coefficient:
da
Gmn = --- (20-65)
The voltage, current, and impedance matrices for a simple synchronous qa -"df -"dk -L d
motor then become
elf da The bordering column and row of the motional impedance matrix (Eq.
19-28) then are, respectively,
The motional impedance (Eqs. 19-10 and 20-1) then is (using s for the he
.
coordinate)
The same result may be obtained by simply noticing that the currents in the
q-axes are the same as those in the corresponding d-axes but lagging by 90".
Hence -j times a q-column is subtracted from a d-axis.
20-13. Hunting and oscillation^.^^^ The motional impedance for the
simple generalized machine is given by Eq. 19-28'. The steady-state voltage
a t synchronous speed is (from Eq. 20-3),
Aid° A84 A6
i
.!
in which G(p), L,(p), and L,(p) are as defined in Eq. 20-13. For sustained
hunting, substitute p = jhw in Eq. 20-70.
20-14. Interconnected Synchronous Machines Os~illating.'.~If
the rotors of both machines in Fig. 20-7 are oscillating, then AO, - AO, =
Ad. With respect to stationary axes, the reference axes of the first machine
may be considered fixed while the axes of the second machine oscillate by A6
The first term of Eq. 20-72 is
about their mean position. Then AC:, is not zero. Alsopd = p(02 - 8,) = 0
for the average velocity, and therefore c:,qpcg.=
dr1
Lblpel
and p Act, = 0. Since the steady-state currents along the stationary axes 6)p + (L;,, sin8 6 + L:,, cos8 6)p& ydlp
are constant, piy' = 0, and as there are no impressed voltages on the slip
+ (L;,, - L;,,) sin 6 cos 6 p02 + (L;,, Li,,) sin 6 cos 6 p
-
rings, ae,/aO = 0.
The transformation matrix is the same as in Eq. 20-56 with sl and s,
rows and columns added:
Then, in the second term of Eq. 20-72, 20-4. The transformation matrix (Eq. 20-24), defines the old currents in
-
~1 82 terms of the new currents (if, ib). By reading the columns, does this matrix also
define the (if,ib)currents in terms of the (id,i q ) currents?
20-5. From Eq. 20-29 identify E; and Ei and from Eq. 20-34 identify E;
and El in Fig. 20-3.
20-6. Prove that T oin Fig. 20-3 gives the crest of the oscillating torque.
This result becomes evident without performing the matrix multiplications 20-7. Show that Eq. 20-38 follows directly from Eqs. 20-35 and 20-36 if
~ , the p in Z,, refers to both the currents and 6 terms.
Z,,,, = C ~ , Z m n Cand
if it is remembered that C$(aC$/aO) is the rotation matrix of Eq. 13-24.
20-8. If Ldr = Lqr, then B = 0 in Eq. 20-40 and the torque terms due to
Multiplying Eq. 20-76 by (C:,2YaB), the second term of Eq. 20-72 becomes
salient poles vanish. is cuss this reluctance torque.
s1 82 20-9. Check Eq. 20-46 by Eq. 20-47.
20-10. Verify Eq. 20-56 by following the suggestion given directly below
k 8 PO, - cos 6(r,, + L & r z ~ )-iq*[L&,,
i q T 2 [ Lsin l sin 6 p 8 , cos 6(r,, + L;,zp)]
-
that equation. Show the appropriate diagrams of connection.
+--ide[L;,2 cos 6 PO, + sin 6(rr2+ L ~ , , p ) l -id*[L& cos 6 p 8 , + sin 6(r,, + Li,,p)] 20-1 1. Simplify Eqs. 20-57, 20-58, and 20-59 for the case where the two
-
machines are similar and running at the same speed.
iqT2[L;,2~ 06sP ~ +Z sin 8(r,, + L&,ap)l -iqT2[L~,, cos 6 p 6 , + sin 6(r,, + L ; , , ~ ) ] 20-12. Obtain Eq. 20-61 without employing the polyphase matriu, but
-idr2[L;,2 sin 6 PO, - cos 4 r , , + L;,,p)l +ide[L,,, sin 6 p 8 , cos 6(r,, + L,,,p)]
-
justify the procedure.
--
ion ( iVTZ Lip, - yd2)+ idrZ(idr2L;,r+ yip,) -iq*(iq*L' dre - +
Y ~ Z-) id*(idr2L~12% a )
20-13. Is Eq. 20-64 true if ot is not constant?
20-14. Write out the complete motional impedance matrix of Eq. 20-80.
(20-77)
The last term of Eq. 20-72 contains the pre-oscillating steady-state
voltage,
drl qrl sl dr2 qr2 s2
e, = ~8,iS=
' O ed,2 em 0 (20-78)
aB B earl %I
Multiplying Eq. 20-78 by Eq. 20-75, the last term of Eq. 20-72 becomes
The motional impedance for the two machines is then the sum of the
equations
TZ.,,
= (20-74) + (20-77) + (20-79) (20-80)
PROBLEMS
20-1. Confirm all equations in the chapter and verify in detail each step in
the derivations.
20-2. If the machine in Fig. 20-1 is running above synchronous speed and is
oscillating a t an angular velocity hw about its mean speed, what is the voltage
matrix corresponding to Eq. 20-3?
20-3. The equivalent circuits of Fig. 20-2 define the subtransient impedances
of a synchronous machine. What are the equivalent circuits for the transient
impedances of Eq. 20-169
Art. 21-21 INDUCTION MACHINES 247
INDUCTION MACHINES
lr - j x m -jx,
warn,,, = -
I 2r
If the rotor is short-circuited (as in a squirrel-cage induction motor), or
closed through external impedances (as in a wound rotor induction motor
with external resistors), the applied rotor voltages vanish, e,, = e,, = 0.
If the motor is operating with a-c applied voltages and a t slip s = (1 - v),
the steady-state matrices are found by substituting p = jm and p8 = vw. Let the l r row of Z,.,, and em, be divided by (1 - v) = s and the 2r row
Then Lp = jwL = jX and Mp8 = vwM = v X m and
248 M A C H I N E ANALYSIS [Ch. 21 Art. 21-41 I N D U C T I O N MACHINES 249
Expressing the self-reactances as sums of leakage and mutual reactances of performance greatly simplified. Multiplying the impedance matrix by the
( X , = x,+ X , and X , = x, +X,), doubling the impedances, and halving polyphase matrix,
the currents, these equations correspond to the equivalent circuit of
Fig. 21-2.
The torque, in synchronous watts, is
and is given as the sum of the wattmeter readings for the rotor resistance
elements.
4 ixs ix,
-
in which 2, and 2, are, respectively, the positive and negative sequence This motor is a special case of the double squirrel-cage induction motor
impedances. discussed later. I t s equivalent circuit is the same as that in Fig. 21-7 with
the r, and x, elements deleted. I t may also be regarded as a special case of
21-3. Balanced Polyphase Voltages. When the applied stator voltages the split-phase motor of Fig. 21-10, where Z = 0, n = 1, and its equivalent
are balanced (equal magnitudes and the q-axis voltage lagging the d-axis circuit is that of Fig. 21-11d.
voltage by 90' in time), the phenomena occurring in the q-axis are exactly
the same as those which occurred in the d-axis a quarter cycle earlier. Then 21-4. Three-Phase Induction Motor.8 The stator phase windings
the polyphase matrix given by Eq. 11-17 may be employed and the equations a, b, and c of a 3-phase induction motor are spaced 120' apart, and each phase
M A C H I N E ANALYSIS [Ch. 21 Art. 21-41 I N D U C T I O N MACHINES
GJ ern.
eqr 9
( a ) Moving axes
qsl q s 2
( b ) Stationary axes
Fig. 21-3. Three-phase induction motor.
is assumed to belong to its own winding layer (Fig. 21-3), so that the trans- where
formation matrix between stationary and moving axes is A = ( -M p +& ~ p 8 ) / 2 B = ( -M p - A ~p8)/2
Z,"," = e
(
in which
eos 30' 1 ! I
The impedance matrix, referred to stationary axes, is (putting Z =r + Lp)
which is identical with Eq. 21-7, except for the zero sequence component Zmn= dr (21-23)
impedance.
The torque tensor comprises the terms involving s in Eq. 21-17 and is
For a smooth air gap, L, = L,, only the - M term of G,, contributes to the
toraue.
and the torque, in synchronous watts per phase, is The steady-state impedance follows upon substituting p = jo and pe =
* ," ," vw; or is the same as that in Eq. 21-2 with the qs row and column deleted.
T = Re (wGm,,,,-i i n ) (21-19) The applied voltage is e,,.
254 M A C H I N E ANALYSIS [Ch. 21 An2141 I N D U C T I O N MACHINES
If 2-phase symmetrical component's are introduced, then for a smooth The motional impedance Tap of the single-
air gap machine, as in Fig. 21-1, the impedance matrix becomes (after phase induction motor is the same as that of the
dividing the l r row by s and the 2r row by 2s) generalized machine (Eq. 19-29), after putting
Ld,=L,,=L,,Ldr=L,,=Lr,Md=M*=M,
deleting the qs column and row, and putting
iQs= 0.
or1
21-6. Double Squirrel-Cage Induction
Motor.8 The connection diagram of this motor,
referred to stationary axes, is shown in Pig. 21-6.
I t s impedance matrix is the same as that of the Fig. 21-6.
generalized machine with one stator and two rotor
layers of windings :
and
e,, = e,, = ed/2
Hence
induction motor.
Z, = drl
+j
jsX,, r,
---
m
+
i jxmz
--
1
jsX, jsX,, (21-29)
I
The torque matrix is
+
Putting X, = X,, x , and X, = X , +
x,, where x, and x , are leakage I
impedances, the equivalent circuit is as shown in Fig. 21-5.
The torque is
* *
256 M A C H I N E ANALYSIS [Ch. 21 Art. 21-71 I N D U C T I O N MACHINES
If the rotor rows of Eq. 21-29 are divided through by the slip s, and if Then the impedance Z,, of (21-1) becomes
+ L8p
a l M p c o s 8 ~ r + ~ , 0
Pa
- M p sin 0
0
a
M p cos 8
M p sin 8
r
b
- M p sin 8
+ L,p
M p cos 0
1 1 '2s
M p sin 8
M p cos 8
r, + L,p
(21-34)
Under balanced conditions, when i b = -kia and i'JS= +ids, the matrix
reduces to
ds a
21-7. Moving Axes.8 Let the 2-phase induction motor (Fig. 21-8) be
referred to moving axes by the transformation
matrix
However, in a smooth air gap machine the a and b columns do not contribute
to the torque and may be ignored. Thus, under balanced conditions, all
that remains is
The steady-state impedance follows from Eq. 21-35 upon writing id* =
Fig. 21-8. I d 8 P t and ia = I a & O t , putting the dWtand cksWtinside the impedance
258 M A C H I N E ANALYSIS [Ch. 21 Art. 21-91 I N D U C T I O N MACHINES 259
mat,rix, and operating on the products with p. The impedance becomes The law of transformation of the impedance to rotating axes gives
act.
z,,,, = c;,c,",z,, + C;,Lmn - pel =
,301
Also
ds dr qr P bs X,
em = -e, sin wt -e, sin (wt + 6) e, cos (at + 6) e, cos wt
21-9. The Capacitor and Split-Phase Motor.8 The capacitor motor
has a two-phase stator, one winding of which is supplied with single-phase
Select new axes ( a , 6) revolving a t synchronous speed and let 8, = o t
a-c voltage, and the other winding, having n times
be the instantaneous displacement of these axes. The transformation matrix
then is (Fig. 21-9)
as ar br bs
as many turns, has an impedance Z (usually a
capacitor) in series. The rotor has a closed winding
!sI'; &
!
or squirrel cage. Although the second stator phase
as cos e I
I
, -sin 0 , is usually closed through its impedance, for the sake
of generality it will be considered here as having iqs+
dr 1 cos 0 , -sin 8,
I3
c;, = (21-41) a n applied voltage. The diagram of connections is
1
-I--
qr sin 8, cos 9 ,
shown in Fig. 21-10. The impedance, torque, and
qs sin el I
i
I
I '
I
cos 0,
voltage matrices of the generalized machine are
given in Eq. 21-2, but this needs to be modified
by multiplying the qs row and column by n to
The new voltage matrix becomes account for the difference in turns between the Fig. 21-10.
qs- and ds-axes (which is all that a turn-ratio
transformation matrix would accomplish). Then in addition the series
impedance Z must be added to the stator impedance in the qs-axis.
262 M A C H I N E ANALYSIS [Ch. 21 Art. 21-1 I ] I N D U C T I O N MACHINES 263
If the rotor columns of Z,,,, Under steady-state conditions, substituting p = jo and p0 = vw, the
... .. are eliminated, the impedance matrix of
Eq. 21-49 reduces to impedance is
Is 2s
zm+,. =
jX,, cos u
rs + jXdscos2 a I
jXmdcos a jX,, sin a
+ jX,, sinZu I
in which the sequence impedances are
I jX,, sin a
I
I I
The equivalent circuits corresponding to Eqs. 2 1 4 9 and to Eq. 21-50 vxmdl 1 - flXmd cos a v X d I
' + jxgr
are shown in Fig. 21-11, together with two special cases. The E,, and E,,
of Eq. 21-50 are easily verified on the equivalent circuit.
21-10. Shaded-Pole Motor.8 The shaded-pole motor of Fig. 21-12 has The torque tensor is
a single-phase stator winding, ds2, and a short-circuited stator winding, a,
considered to be in a different layer. The squirrel-cage rotor is represented as2 dsl
by a pair of short-circuited brushes:
'
21-1 1. Motional Impedance of the Polyphase Induction M o t ~ r . ~
The motional impedance matrix of the genemlized machine, referred to
stationary axes (d, q), was given in Eq. 19-28. For the polyphase induction
motor with smooth rotor, L,, = La, = L,, M, = M, = M, L,, = La, = L,.
Fig. 21-12.
I sin a I
I
I
I Let moving axes (a,b ) , rotating a t synchronous speed,pe, = o,be introduced,
as in Fig. 21-9 and Eq. 21-41 :
The Gansient impedance matrix of this machine, which is the same as that
111,
of the generalized machine, is
dr 1 cos 0, -sin 0,
qs sin el
8 1
264 M A C H I N E ANALYSIS [Ch. 21 Art. 21-12] I N D U C T I O N MACHINES 265
The new motional impedance then is, putting 8, - 8 = 6, The rate of change, if 6 is varying, is
I n accordance with Eq. 19-3, the term appears on the left (or voltage) side
of the equation of oscillations. However, it may be transferred to the
right-hand side,
ae
Ae, = 3?,,Ai@ - -
' A6
as
and placed in the s column of the motional impedance matrix. Thus matrices
(21-58) and (21-59) are augmented in the s column by -e cos 6 in the a, row
and by -e sin 6 in the b, row.
When the applied voltages or torques are of hunting frequency, hw, the
currents and velocities are also of hunting frequency. Substituting p = jhw, 21-12. Indhction Motor with Oscillating Load.8 Let the steady-
pel = w , p 6 = p ( 8 , - 8 ) = sw and multiplying the last row and column state veloci$y, v w , and torque, T , be constant and the applied voltages and
by w , the motional impedance becomes currents be of fundamental frequency, w . If the superimposed oscillatory
torque, AT, is of hunting frequency, hw, the corresponding velocity changes
Aiu will also be of this frequency, but the current changes Aim will have the
two product frequencies ( h f 1)w. That is, the frequencies involved between
the variables are as shown in Table 21-1.
TABLE 21-1
I 1
I
1
Steady State Oscillating
bs X, Xm + jhX. 0
1
T,
1
1
Mechanical vw, T Constant AN, A? Hunting ( h w )
s ibrXm -ia'Xm jhw3J
c t i c e, i Fundamental ( w ) Ae, Ai F'roduct ( h * 1)-
The currents in the bordering row and column are those corresponding to
the steady-state conditions prevailing before hunting. The transient imped- The motional impedance of the unbalanced induction motor is the same
ance matrix was given in Eq. 21-43, and the steady-state impedance in as that of the generalized machine (Eq. 19-28).
Eq. 2 1 4 5 , from which the prehunting steady-state currents may be found as Assuming the current changes to have frequencies ( h f l ) w , and the
in' = (Zmtn,)-lemf (21-59) average speed p8 = v w , the impedance matrix (Eq. 19-28), excluding the
and substituted in the bordering row and column. bordering row and column, upon substituting p = j(h f 1)w becomes
When fundamental frequency voltages are applied to the stator and slip
frequency voltages to the rotor, then both the stator and rotor air gap fields
revoke a t synchronous speed, but the stator field leads the rotor field by a n
angle
6=8,-8,-e
+
in which 8, is the position of the stator flux, ( 8 8,) the position of the rotor
flux, and 0 the displacement of the rotor. The applied voltages then are
B cos (w,t + B) = Re ~ $ ( " a ~ + ~ =' Re Bsi"zt the current changes hinfor each set of frequencies may be solved for sepa-
rately in terms of the voltage changes and then substituted in the torque
yields two frequencies ( w , f w,) : change :
AB Ain = (Zmn)-'(he, - Zm,Aiv) (21-69)
A B cos (w,t + a ) cos (w,t + B) = -2
+ co,)t + a + B]
cos [(o,
Ae, = Z,,Ain +Jo3Aiv (21-70)
PROBLEMS
21-1. Confirm all equations in the chapter and verify in detail every step
in the derivations.
21-2. Obtain Eq. 21-17 by using Eqs. 21-20 and 21-21.
21-3. Derive the motional impedance matrix for the single-phase induction
motor.
Hence the product having the sum of the two frequencies is found by multiply- 2 1 4 . Derive the motional impedance matrix for the double squirrel-cage
ing the two complex numbers together, while the product having the difference induction motor.
of the'two frequencies is found by multiplying one complex number by the 21-5. Derive the motional impedance matrix for the capacitor and split-phase
conjugate of the other. Accordingly, the bordering column of Eq. 19-28 is motor.
21-6. Derive the motional impedance matrix for the shaded-pole motor.
taken as
21-7. From its impedance matrix develop the steady-state equivalent circuit
ds dr P ' PS for the double-fed induction motor.
column: s 1
I
0
I
iarXqp+ iaaXmq - i d d
I
-
I
i d d1
I
0 II (21-64) 21-8. From its impedance matrix develop the steady-state equivalent circuit
of the polyphase induction motor.
for ( h + l ) w frequency of Ae, 21-9. Solve Eqs. 21-69 and 21-70 formally for the change in velocity.
* *
,I -idsXmd
* - idrXdr i
column: s 0 iqrXq, + iaSXma
* 0 (21-65)
for ( h - l ) w frequency of Ae,
Art. 22-11 C O M M U T A T O R MACHINES 269
-
ds dr
a
,, = c:,q.amn = c:.~,
COMMUTATOR MACHINES
8
1
The admittance tensor is the inverse of Eq. 22-3, or
4- L ~ P - M p cos u
There is a great variety of commutating machines of both the a-c and ym'n' = (k,)-l d8
7,
m1 (22-5)
d-c type. Since such machines have stationary brushes, their performance
equations are most easily found by starting with the quasi-holonomic
a - M ( p cos a - p0 sin a ) r. + Lbp
generalized machine (the primitive machine of the f i s t kind). The analysis where
for a number of such machines will be given in this chapter, but primarily D = (L,L, - M2cos2u)p2+ (r,L, + r,L, + M2sin u cos up8)p r,rr (22-6) +
from the point of view of using tensor methods for establishing the machine The only applied voltage is e,, = e, since e, = 0 for the short-circuited
equations and not with any intention to explore detailed performance brush axis, and therefore
characteristics.
id, - (rr L,p)e +
22-1. The Repulsion M o t ~ r . ~ * ~repulsion motor (Pig. 22-1) has
The
an a-c field on the d-axis, a smooth air gap, and a short-
ia
-M cos u pe + M sin a pee
ld circuited pair of brushes on the commutator shifted an D
angle u from the d-axis. The corresponding primitive
ids!
machine is that of Fig. 1 3 3 , having an impedance and for constant speed, p0, the explicit solutions are given by the rules of
(Eq. 1 3 3 ) with the qs column and row deleted: thus operational calculus (or Laplace transforms).
I
The instantaneous torque is
T = Q,,,,i""i'" = -M sin a id%" (22-8)
Under steady-state conditions, with an a-c voltage e applied, substituting
p = jo and p8 = v o in the previous equations gives the corresponding a-c
values. Then Eqs. 22-7 become
Fig. 22-1. -MP~ -&PO T, + L,P ids - (rr + jXJe
D
Its transformation tensor is
ia =
(-jXm cos cr + vX, sin u)e
(22-9)
ds a
D
D = (-XJr + X z cos2u + r,r,) + j(rJs + rJ, + vX; sin a cos u)
and the synchronous torque by Eq. 1 3 4 5 is
The impedance of the motor then is = Re [eX,(v sin u + j cos a) (-X, sin a)
ds a
D* D
e2Xk sin u(X, cos u - r,v sin a)
Z, =C:,C:,Z, =
ds ?a + Lap M p cos u -
(r,r, + X i cos u - X,X,) (rJS rJT + Xzv sin u + +
(22-3)
a M ( p cos a - p0 sin a ) T, + L,p (22-10)
268
270 M A C H I N E ANALYSIS [Ch. 22 Art. 22-31 C O M M U T A T O R MACHINES
22-2. Repulsion Motor with Two Sets of Brushes8 Fig. 22-2 shows
a repulsion motor with two sets of brushes connected in series. The trans- 'P
formation matrix is if-
C$, = dr (22-1 1)
sin cc + sin 3j
The steady-state impedance then is
Fig. 22-2. Fig. 22-3.
'
in which I t s torque matrix comprises the terms in Eq. 22-14 having p6 as a coefficient,
d = - -a + - mean brush axis
2
2y =?!t - a = brush separation
If the brush separation is zero, y = 0, a = 8, and 6 = E, and Eq. 22-12 Under steady-state conditions, p = 0, the impedance matrix reduces to
reduces t o Eq. 22-3 if e, and i" are divided by 2 t o take into consideration
the fact that the rotor circuit is entered twice in Fig. 22-2 and only once in
w:- 00 1
From a combination of Eqs. 22-14 and 22-17, the acceleration matrix 224. Schrage M o t ~ r .The ~ connection diagram of the 4-phase Schrage
may be set up, using s for the p0 coordinate and transferring the p0 terms motor is shown in Fig. 2 2 4 . This motor actually has a n inner rotor connected
to that column; thus t o slip rings, but the slip rings have been replaced by stationary brushes on
which slip frequency voltages are assumed to be impressed. The outer rotor
has a commutator on which bear two pairs o l brushes shifted angles a and
,fl from the d-axis and two pairs of brushes shifted angles a and ,fl from the
If i t is assumed that the currents in the s row and column are known and
remain constant during acceleration, this equation is readily solved. If the
actual variation of the current is to be taken into account, Eq. 22-18 can be
solved step by step.
The motional impedance of the d-c machine, by Eq. 19-10, is
Fig. 22-4.
q-axis. These brushes are connected to the stator coils as shown. Assuming
that each stator coil has n turns for each rotor winding turn, and projecting
all currents on the d- and q-axes, respectively, the transformation matrix
is seen to be
7
=
f
dr2
k n'
ds
1
f
--
2n
f
q;l
dr2
;,
(22-22)
am,,,, =
dr2
f
El
j4nM sin y
2jnM'
in which only those terms which actually contribute to the torque have been
retained.
(22-26)
+ E-@)
j
(22-23)
ds
ds
1
a dr2 qr2
E
@i
qr1 j(&ja -- dr2
dr 1 cos a
Ps j2n --
CE, = dr2 1 (22-27)
The impedance matrix for a generalized machine with two rotor layers and --
a smooth air gap is 1
qr2
ds dr 1 dr2 P.2 qr 1 qs --
ds r, + L,P Mp
--
M'P 1 0 0 0 qr 1 sin a
Fig. 22-5.
MP 1 r.1 + 4 1 ~ M.P M?PO
I
I L.,pO MpB The impedance matrix (Eq. 22-24), with the qs column
dr2 P 1 M.P r,, +
L,P 1 L,PO 1 M,~O
1 wo
and row deleted, is used as the starting point for this motor. In that
matrix, putting p = jco and p0 = vw for steady-state conditions, and
z,, =
~2
qrl
---Me
M P
I -M&
~ - L ~ ~ P -1wee
~
1
/
--
-Lr2p8
/
r., + Lap
M,P
1
rr1
M,p
+4,p /
1 M3p
~p
transforming by Eq. 22-27, there results
ds a dr2 qr.2
I
0 0 0 M'P
1 MP
T8 + &P
(22-24)
ds r, +jX, j X , cos a 9;
qr2 -vx;
- v cos a )
-vx,2 rr2 + 5%
2jXm, sin ys-j8 (22-28)
Z, = (22-25)
dr2 2njXml + ZjX,, sin y d B rr2 +jx12 The only applied voltage is e,, = e, the other axes being short-circuited.
276 M A C H I N E ANALYSIS [Ch. 22 Art. 22-71 C O M M U T A T O R MACHINES
The torque matrix (of which only the first column contributes to the but the polyphase matrix (Eq. 22-22) is not applicable.
torque) is
22-7. Frequency C o n ~ e r t e r . A
~ rotary converter armature inside a
smooth stator with no windings may be driven mechanically, and i t will
convert the frequency applied t o its slip rings t o some other frequency a t its
brushes. Its connection diagram is shown in Fig. 22-7, and its transformation
matrix is
d a b q
r, + Lap M cos u p
M,(cos u p M,(sin u p
iq*M, cos u c:, =
d 1
0
cOse -
-
sin 8
--sine
cos 8
I 0
1
(22-31)
-ideM, sin a q
- sin a p e ) + cos u p e )
-id8M sin a
The transient impedance
- and inductance matrices
M,(cos u p for the generalized machine with windings on the rotor
+
dGl
sin u p0) only and a smooth air gap are Fig. 22-7.
d d q
Fig. 22-9, and its impedance and torque matrices, referred t o stationary
axes, are
(22-35) dr 99.
ds n 0
dr cos a -sin a
qr sin a cos u
P 0 n
Fig. 22-10. Fig. 22-1 1.
Applying Eq. 2242 to the impedance and torque matrices of the simpler
generalized machine (Eq. 13-3), and substituting p = jw and p0 = vw for Under balanced polyphase conditions, the voltages and currents in the
steady-state conditions, there results b- and q-axes are lagging those in the a- and d-axes, respectively, by 90' in
time, and therefore, substituting ib= -jia and iQa = -jid* and combining
a b
columns with the same currents, the transformation matrix becomes
+ + +
nVa jXJ (rv j X r )
+ n x m ( 2 j cos u - v sin u )
vxv + nvxm cos
I
+
n V , jXA + +
(rr jXT)
+nXm(2j cos u - v sin a ) cos a + j sin a = efu
-
a b
ds 1
/ If the ds-axis is connected in series with the a-axis, and the qs-axis is
connected in series with the b-axis, this machine becomes identical with the
Leblanc advancer.
dr cos u -sin a
PROBLEMS
P7 sin u cos u
22-1. Confirm all equations in t h e chapter and v e r i f y in detail every step
q.3 1 in t h e derivations.
282 M A C H I N E ANALYSIS [Ch. 22
22-2. Carry out the steady-state analysis for the repulsion motor with two
sets of brushes (Fig. 22-2) in the same way as was done for the simple repulsion
motor (Fig. 22-1).
22-3. Carry out the analysis for a shunt-wound d-c motor, and check the
results a t each stage of the development as a special case of the analysis for the
compound-wound motor of Fig. 22-3.
2 2 4 . Carry out the same analysis for a series-wound d-c motor, and check
the results at each stage of the analysis as a special case of the analysis for the INTERCONNECTED MACHINES
compound-wound motor of Fig. 22-3.
22-5. Derive Eq. 22-23 without using the polyphase matrix in Eq. 22-22.
22-6. Why does only the Grst column of Eq. 22-29 contribute to the torque?
22-7. Show that the Scherbius advancer is a special case of the Lebhnc Machines of all types may be electrically interconnected, or operated in
advancer. combination, for such purposes as energy conversion, speed control, changing
22-8. Develop the impedance and torque matrices for the Fynn-Weichsel phase or frequency, and power factor correction. The interconnection be-
motor of Fig. 22-6. tween machines may also involve static networks as part of a combined
22-9. Will the L, term of Eq. 22-50 contribute to the torque? system.
22-10. At what angle of brush shift will the torque be a maximum for the If the transient impedance, inductance, and torque tensors of several
shunt polyphase commutator motor?
individual machines and networks are represented by lZmn, ,Z,,, . .. ,,L,,,
.. .
BLmn,. . . , and ,arnn, ,Gmn, , respectively, and are interconnected in such
fashion that the transformation matrix of the connections is q., then the
transient impedance and torque tensors of the entire group are
d82 1
j
I
T-
- --
-
S
--- I l1
ds'2
f'
---
1
I
---
sf 1
for the first machine and using primes for the corresponding quantities of the d
second machine, the impedances of the two interconnected machines may
be combined in a single matrix and transformed by matrix (23-3) to give the
impedance for the pair of machines. (In this simple connection the trans-
formation matrix multiplication may be dispensed with and columns
Fig. 23-1.
M' id'
a(
+ 2n;ih)
M;ih
U R' + J'p
I n matrices (23-5) and (23-6) the steady-state currents id, id', and ih must be
determined from the solution of
combined after putting if = if' = i h ) . The motional impedance transforms Eq. 2 3 4 . The applied voltages e,,,,
directly as 3a,p,= C ~ 2 2 ' ~ ~since
C j ~the transformation matrix contains only
constants. The transient, acceleration, and motional impedance matrices
e,&,,
--- and eh = e , + e,. are all -
constant.
for the interconnected machines then are L
P
23-2. Kapp V i b r a t ~ r . ' . ~The
Kapp vibrator consists of a poly-
phase slip ring induction motor with
------
each of its rotor phases connected
t o a d-c motor, as shown in Fig. 23-2
for one phase only. The rotor of the
d-c motor oscillates a t slip fre-
quency, but it does not rotate. The Fig. 23-2.
rotor of the induction motor rotates
at a uniform speed, but i t does not oscillate. The connection matrix is
The motional impedance of the d-c machine, as a special case of Eq. 22-19, If the s-axis is eliminated (the mechanical torque is zero), the motional
that is, without series or commutating field, is impedance reduces t o
Since the induction motor does not oscillate, its motional impedance is the
same as its transient impedance (Eq. 21-35), (23-12)
ds2 a
The oscillating d-c rotor introduces a capacitance in the rotor impedance
proportional t o the square of field flux idslMdl linked with the rotor, and'
inversely proportional t o the slip s and to the inertia J1.
23-3. Phase Advancer.73 Fig. 23-3 shows (for one phase only) a
Combining Eqs. 23-8 and 23-9 and transforming by Eq. 23-7, there results
polyphasefrequency converter fed from
the slip rings of a polyphase induction
motor. The connection matrix is dr2
dsl a1 a2
I
The pre-oscillating current of the d-c motor, appearing in the bordering
Fig. 23-3.
row and column, is idS1 = edsl/rdsland is constant. The current ids2 in the
stator of the induction motor is of fundamental frequency. The current if
is of slip frequency. Since the d-c motor does not rotate, the average velocity The impedance of the induction motor is given by Eq. 21-35, and that of
is zero, pOl = 0. Therefore the frequency converter by Eq. 22-35. The impedance of the group is
Jf2p(Ekez s ' " ~ = kmJf2 AIfskut
Aif) = M ~ ( ~ " ~ - AIfEkSut)
. M,p(E-iez Aids2) = ~ ~ ~ ( ~ - j ( l A- ~s d) sE~tjutt ) = jsmMZ AIds2&jsot
dsl a1 a2
Then, multiplying the bordering row and column by w , the steady-state
motional impedance is seen t o be
dsl f (792 s
The current in axes a1 and dr2 is of slip frequency, sw. Hence, putting
O2 = O1 +
6, ial = Ialdsot, ia2 = Ia2dwt,and idsl = Idsldot, there results for
288 M A C H I N E ANALYSIS [Ch. 23 Art. 23-51 I N T E R C O N N E C T E D MACHINES 289
the steady-state impedance of the group The currents in axes a1 and dr2 are of slip frequency, s = 1 - Plv, where v
is the ratio of the actual rpm to the 2-pole synchronous speed (3600 for 60
cycles). Since the two machines are on the same shaft, pOl/Pl = pO,/P, =
vw. Substituting idsl = IdS1dwt and ia = the steady-state impedance
becomes
Z,,,, =
(23-18)
The transient and steady-state torque matrices of the group are
L 1
The transient impedance of the first machine lZm, is given by Eq. 21-35, Fig. 23-5.
and that of the second machine ,Zm, by Eq. 21-8. Then, for the group,
polyphase induction motor and on t h e same shaft. The connection matrix
is
dsl a
(23-20)
290 M A C H I N E ANALYSIS [Ch. 23 Art. 23-61 I N T E R C O N N E C T E D MACHINES 29 1
The transient impedance ,Zmn of the induction motor is given by Eq. The sum of the individual impedances of the two machines, referred t o
21-35, and the impedance ,Zmn of the commutator motor is their own axes, from Eq. 21-35, is
f
.
dsl
Fig. 23-6.
there results
292 M A C H I N E ANALYSIS [Ch.23 Art. 23-71 I N T E R C O N N E C T E D MACHINES 293
Since p(~-jei)
=~ - j - ( ~ and 8, - 8, = 6, this matrix becomes
~jp8)i The equivalent circuit corresponding to (23-33) is shown in Fig. 23-7. The
torques a3.e
f~",,,. 2 c,~c,B.~Z",@
act, .,, + C:*Lap-ac;.P' .,. f ac:,
+ C:*SaP- 1
ax, ax,' ab
Under steady-state conditions all currents are of fundamental frequency
o,and, upon putting p = jw and p - jp8 = jo - j ( 1 - s)o = jso, there
The Zapfor two generalized machines with stationary axes is given in
results
Eq. 20-73. The transformation matrix for two oscillating induction motors
Hence the.motiona1 impedance for two oscillating power selsyns is the sum
of the four matrices :
~ , p ~ j ~ l 0 0 1
1
dsl ds2 ds3
I I
a - ~ x , ~ E ' ~jX,, 0
COG,,,, = (2345)
ds2 0 0 -jXrn3
296 M A C H I N E ANALYSIS [Ch. 23 Art. 23-81 I N T E R C O N N E C T E D MACHINES
23-8. Arnplidyne Voltage Regulator.' This regulator comprises an exciter, the impedance matrices for the three devices are
amplidyne, a stabilizing transformer, and an exciter, interconnected as
1 2 3 d Q
Fig. 23-9.
Neglecting the induced voltage (Lp and Mp terms) in both the stator and The voltage equation may be rewritten by employing the identity
+
rotor windings of the amplidyne, as well as the (r, Lg)term for the
298 MACHINE ANALYSIS [Ch. 23 Art. 23-81 INTERCONNECTED MACHINES
in which Kkj is the inverse of R,,. Then in which p, is the over-all amplification factor and
in which
Ti, = KkiLmk= time constant matrix Eq. 23-53, a cubic (p3) in the time operator, gives the change in exciter
voltage Aef corresponding to a change hea impressed on the control field.
= KkiGm,= amplification matrix
The stability of the regulator may be determined by applying Routh's
Eq. 23-51 substitutes resistance drops for currents, and this form of the criterion, coefficients (11-32), or simply by solving for the roots of Eq. 23-53.
voltage equation is equivalent t o multiplying each current in each axis by
the resistance in the corresponding diagonal term of Eq. 2 3 4 8 , and then PROBLEMS
compensating by dividing the column of Eq. 2 3 4 8 by the diagonal resistances.
Then Eq. 2 3 4 8 becomes 23-1. Cordinn all equations in the chapter and verify in detail every step
in the derivations.
23-2. Develop the motional impedance matrix for the cascade induction
motors of Fig. 23-4.
23-3. Develop the motional impedance matrix for the cascade induction and
commutator motors of Fig. 23-5.
23-4. Develop the motional impedance matrix for the differential selsyns
of Fig. 23-8.
23-5. Set up the transformation matrix and the transient impedance matrix
for a synchronous generator connected to a polyphase induction motor.
23-6. Set up the transformation matrix and the transient impedance matrix
for a synchronous generator connected to a series-wounda-c commutating motor.
23-7. Set up the transformation matrix and the transient impedance matrix
for a shunt-wound exciter and a synchronous generator.
Upon elimination of the 3-, q-, c-, and p-axes (there are no voltages
impressed on these axes), there results
BIBLIOGRAPHY