Professional Documents
Culture Documents
Mtthntn's ffijjnBQtnyhs an
General Editor:
INTRODUCTION TO THE
LAPLACE TRANSFORMATION
WORSNOP,
B.So., Ph.D.
CHAPMAN. S. CHILDS, W. H.
DINGLE, H.
J.
EMELEUS, K,
G.
THE SPECIAL THEORY OF RELATIVITY THE CONDUCTION OF ELECTRICITY THROUGH GASES DIPOLE MOMENTS
HAGUE,
B.
E.
HIRSCHLAFF,
JAEGER,
J. C.
WAVE MECHANICS MOLECULAR BEAMS ,,~TCS AN INTRODUCTION TO VECTOR ANALYSIS FOR PHYSICISTS AND ENGINEERS FLUORESCENCE AND PHOSPHORESCENCE LOW TEMPERATURE PHYSICS WAVE FILTERS ._HIGH FREQUENCY TRANSMISSION LINES _ HIGH VOLTAGE PHYSICS A INTRODUCTION TO THE LAPLACE TRANS-
JAMES, R.
W.
C.
J.
JOHNSON. R.
JONES, T. J. JOSEPHS, H.
MoCREA, W. H. McVITTIE, G. C.
FORMATION X-RAY CRYSTALLOGRAPHY ATOMIC SPECTRA THERMIONIC EMISSION ___. HEAVISIDE'S ELECTRIC CIRCUIT THEORY THE KINETIC THEORY OF GASES WAVE GUIDES RELATIVITY PHYSICS COSMOLOGICAL THEORY THE CYCLOTRON WIRELESS RECEIVERS _.., ALTERNATING CURRENT MEASUREMENTS THE METHOD OF DIMENSIONS
THERMODYNAMICS
,,_
SOMMER,
A.
C.
STYLE, D. W. G.
STONER, E.
SUTHERLAND,
TEAGO, F.
J.,
G. B. B. M,
and
GILL, J. F.
TEAGO, F. J. TEMPLE, G.
QUANTUM
SPECTRA.
TOLANSKY,
F.
S.
APPLICATIONS OF INTERFEROMETRY
An
Introduction to the
LAPLACE TRANSFORMATION
With Engineering Applications
by
J.
C.
JAEGER
WITH 31 DIAGRAMS
METHUEN
LONDON
IIA Lib.,
Illill
Illl
nil
First published in
1949
PREFACE
THIS book
contains the substance of a course of lectures
engineers and physicists at the National Standards Laboratory, Sydney, in 1944, when the author to the Radiophysics Laboratory of the Council for Scientific and Industrial Research. A mimeowas produced by the C.S.I.R. in 1945 edition graphed
delivered to
was seconded
There
is
an increasing
demand
for
operational
methods by engineers
but whose mathematical equipment is restricted to that of an ordinary engineering course. Of the many books available, the older ones use " the rather suspect and certainly obscure operational methods ", while the newer ones require a knowledge of the theory of functions of a complex variable, and some comparatively advanced mathematics. Here I have atencounter transient problems,
tempted to go as far as possible using no mathematics beyond ordinary calculus, but at the same time catering for
who wishes to acquire the manipulative skill necessary to solve his own problems to this end, the book it is, in fact, contains as little theory as possible largely a collection of worked examples illustrating the methods of solution of the various types of problem commonly arising a handful of problems, with answers, in circuit theory has been added to enable the reader to test his skill. For problems on ordinary linear differential equations, and thus on electric circuits with lumped constants, the use of the theory of functions of a complex variable is
the reader
: ; ;
but it is needed for a complete study quite unnecessary of partial differential equations, and in the higher parts The engineer who wishes to go deeply of the subject.
;
must study
this theory,
sophisticated points in it do arise, he must study it fully and carefully, and so before he starts he is entitled to
VI
know what types of problem he will be able to solve, and how useful the solutions will be. I have therefore in Chapter IV discussed the Equations of the transmission
^
line
earlier chapters,
all
using
calculus
this
treatment covers
the
common,
problems, and will enable the reader to follow published work on the subject. It is to be understood that solutions obtained in this way are correct, and that the algebra is the same as it would be if the complex variable were used the latter merely changes the point of view, and allows ^
of the
reasons for the notation used here for the transform roo
e-**y(t)dt
.
The
Laplace
v
.
(*)
Jo
of y(t) may be given briefly. Any notation should three conditions it must be short, and easy to (i)
:
satisfy
write
must be applicable to any symbols not be used in any other connection. notation of all'would be to write a capital
(ii)
it
(iii)
it
should
letter
; Laplace transform of the corresponding small letter but this may be most inconvenient since it is not applicable to many symbols, in particular to capital and many A " star ", y*, is not much used in other conletters.
Greek
nections
notation, y,
bar but is rather clumsy to write. The used here, was originally used by Le"vy can reintrodueed by Carslaw it is easy to write, and be applied to any symbols, but it has the disadvantage of
"
and
being employed in other senses, these, however, rarely occur in those fields where the Laplace transformation
is
most
useful,
is
most important:
desideratum.
can be replaced by
is
y used
by Millman.
A much
more
relative
form
v(t}dt
PREFACE
Both notations have
their
Vll
advantages, but the essential point is that form (2) gives an exact correspondence with the large body of work done by the older operational methods, while (i) corresponds with the newer books, with most of the theoretical work, and with corresponding work on the Fourier transform. Finally, it is a pleasure to acknowledge the assistance given to me by many officers of the C.S.I.R., in particular that of the Secretary in making available the drawings for the text-figures.
own minor
REFERENCES
following are recommended for further reading and are occasionally referred to below. They all use the complex variable. The first four use the form (i) of the Laplace transform, and the last three the form (a).
The
Carslaw and Jaeger, Operational Methods in Applied Mathematics (Oxford, 1941). This is referred to in the text as C. and J. Churchill, Modern Operational Mathematics in Engineering
(McGraw
Doetsch,
Hill, 1944).
Theorie
Linear
Systems,
Vol.
Methods
Mathematical Physics
(Cam-
CONTENTS
CHAPTER
I.
PAGE
.
.
FUNDAMENTAL THEORY
II.
.....
.
.
26
III.
82
.100
131
INDEX
CHAPTER
FUNDAMENTAL THEORY
i.
In this book we shall primarily be concerned with problems on the behaviour of 'dynamical or electrical systems which are started in a given way at some instant, o. Mathematiwhich we take to be the zero of time, t " initial value problems ", and may be cally, these are called a of as the solution of differential system equaexpressed o. tions with given initial values at the instant t Negative values of the time do not arise in either the specification of
The Laplace transthe problems or in their solution. formation is a mathematical device which is useful for a if is specified for all function such solving problems positive values of the time, we can (in the simple cases needed here) write down a related quantity called its conversely, if the Laplace transform Laplace transform is known, the value of the function can be found from it. The importance of the method arises from the fact that, for the types of problem we are interested in, it is easy to write down the Laplace transform of the solution, and from this the solution itself is then easily found. We suppose y(f) to be a known function of t for values Then the Laplace transform y(p) of y(t] is of f> o.
:
defined as
y(p)
f*rf\
poo
.
.
(i.i)
Jo
where p
(i.i)
is
number
sufficiently large to
if
make
i
the integral
convergent.
For .example,
a
y(t)
any positive
Of course p may be
sufficiently large to
make
a
of the theory,
positive
it is
little
Laplace transform
formed
as in (i.i)
number, and " " is one of a number of integral transforms by multiplying a function of t by a chosen
real part is (i.i) convergent, but, in the early parts more definite to think of p as a real nothing is lost by this restriction. The
of
in
We
>
them
but
y for the Laplace transform of y, it and t being understood that these. are functions of p
need first a collection of the Laplace transforms few common functions. For example, if y where a may be real or complex,
of a
We
= eS
p
where p must be greater than the Again, if a is real and
real part
(1.2)
of a.
y
using (1.2)
= sin at =
iat
.(e
~e- iai ),
we have
2i\p
ia
+ iaj
+ a*
JX
Proceeding in this way, or quoting the results from known definite integrals, we can build up the Table I of Laplace transforms which contains all the -transforms needed in the first three chapters. This Table, and Theorems I to III below, provide the whole of the theory
needed
in
Chapters I and
II.
In (1.6), as remarked above, p has to be greater than the real part of a, while in (1.7) to (1.12) a is real, and in
(1.9) and (i.io) p must be greater than* results of Table I are elementary integrals
\a\.
All
the
except (i.il)
and
in
(1.12), which have been included because they arise cases of resonance. The simplest way of deriving
function of t and p, and integrating with respect to t. The Fourier transform (26.1), is another example. Each integral transform has characteristic properties, but those of the Laplace transform which are proved below make it the moat suitable for the solution of initial value problems. * If a is = a, if a o. o real, \a\ a, if a |a|
>
<
If
we
differentiate *
as if a
* This technique of differentiating under the integral sign with respect to a parameter is a powerful weapon of pure mathematics naturally its use needs justification, and this can usually be supplied when, as here, it is employed in deducing new Laplace transforms from known ones.
;
which
Treating (1.7) in the equivalent to (i.n). way gives (1.12). The collection of transforms in Table I can be greatly extended by the use of the following Theorem.
same
THEOREM I. If y(p) is the Laplace transform of and a is any number, real or complex, then y(p a) is at Laplace transform of e~ y(t). This follows immediately, since the Laplace transform
of e~ aty(t)
f
is
V*
e~ aty(t}dt
=
Jo
\
er<*+*)ty (t)dt
= y(p + a).
Jo
As an example
(-
+ fl)'
(1.8)
it
follows that
the
(p
+ b) + a
.
N2 z
s, ' 2
I
,
and
(1.7) that
.
the transform of
5. 2
e~u sin at
2.
is
O + *) +
...
To j^c?
Laplace
transform.
It will
be seen
method
with constant
of solving an ordinary linear differential equation coefficients and given initial conditions down an algebraic equation for the transform y(p] of the solution y(t}. This transform will to found have the form be usually
consists of writing
'
2 (ail ( -*>
p which have
no
of
common
factor,
THEORY 5 of the labour * of the solution consists of finding the y(t) corresponding to a y(p) given in this form. This is done by looking up y(p) in the Table I of transforms, It is shown at the possibly combined with Theorem I. end of this secton that the y(t) found in this way is the has which function for transform. only y(p]
Most
For example,
if
FUNDAMENTAL
=
p
+2
;
it
= e~
2t
.
Similarly,
from
(1.7)
-f-
and
(1.8), if
y
If the
= 2*
p*
6
A>
z cos 2t
+3
sm
2t.
denominator of y is a general quadratic in p, we complete the square, and use Theorem I and the Table. For example, if
#
it
P+7
follows
P+7
I
j>
from Theorem
e~* cos it
3e~* sin
2Jf.
Notice that both numerator and denominator of y have to be expressed in terms of p -f- a (in the case above, p -\- i) before Theorem I is used. If the denominator of y is of higher degree than the second, we have first to find its factors,f then we express
partial fractions by the usual algebraic methods, finally write down the function of t corresponding each fraction using Table I. Alternatively, Theorem II below may be used when it is applicable.
in
and
to
The student is particularly advised to make himself skilful Labour can be saved by the in the processes of this section. use of a more extensive table of transforms than that given here, but transforms not included in the tables at present available a case solid background of manipulative frequently arise, and in any skill is extremely useful. t For the method of doing this in problems where they are
not obvious see
12.
Assume *
Then
Equating
P+-IS* A(p
coefficients
+ 4^ + 8) + p(Bp + C)
A+
(2.2)
of p^, p,
respectively,
we have
= o, 8A =
B
i!
we
find
-
P-4
~^j
I
(2-3)
find Then from Theorem e~ 2i{cos 22 3 sin 2t}. y= The process of equating coefficients used above
is
always available, but the results can often be obtained more o in the quickly otherwise. For example, if we put p Then, putting 1/8 identity (2.2) we find A 1/8.
in (2.2), this
becomes Bp
= +C
+
A=
(4
p),
=
(p
+ 4)
Then
m
coefficients
Equating
we
find
* It is shown in text-books on algebra that if we assume a fraction with a constant numerator corresponding to each linear factor of the denominator, and a fraction with a linear numerator corresponding to each quadratic factor (or squared linear factor) of the denominator, and so on, we shall always have just the right number of equations to determine the unknown constants in these fractions.
FUNDAMENTAL THEORY
Then, using the Table and Theorem
y
There
I,
we
is
$ cos 2t
Theorem
applicable to
many
important cases
where f(p) and g(p) are polynomials in p, the degree off(p) being less than that of
II.
THEOREM
and
if
If y(p}
= f(p)/g(p),
%(p),
(P) ($ ^(P a z) - - (P where a^ a z an are constants, which ., complex but must all be different, then
, .
an \
(24)
may
be real or
(2.7)
is
y(p)
is
S-5?f\ eV
(2-8)
(2-9)
(2.8)
is
Theorem
follow
in
Heaviside's
Expansion
.
Since (2.8)
(2.7),
(2.10)
respectively, by " * In Heaviside's form of notation, and in the "^-multiplied the Laplace transformation, in which (in order to have an exact correspondence with the Heaviside notation) the Laplace transform is denned as p times the integral in (i.i), an additional term appears arising from this extra factor, p, and the simplicity
directly
(2.5)
.
. .
from
of (2.8)
is lost.
8
1 appears that the essential content of Theorei"* (27) for putting? simply the formulae (3.5) f(p}(g(p) into partial fractions in the case in which g(p) no repeated factors. This is a well-known result,* fon~t for completeness we give a proof below.
using
II
(1.6), it
is
has
All of (2.5) to (2.10) are useful (2.6) is easily renaen*ar ) is bered by noticing that the coefficient of i/(p ar in f(p)lg(p), omitting the facto** result of putting p
;
tn^
ar ) of g(p] (p (2.7) is simply another way of writing? this statement. Similar remarks apply to (2.9) and (3.10);
partial fraction formulae (2.5) and (2.6) are themselves very useful, cf. Example 4, below. With regard to (^.8) and (2.9), it is usually better to use (2.9) when g(p) has
The
&
few simple factors, as in the examples of this section It sometimes happens that g(p) has complicated factors, which need not even be set out explicitly, as in the problems of 1 6, and then (2.8) is the more useful.
To
derive (2.6)
we
methods discussed
earlier,
by assuming
Then
Putting
ar
in this gives
.
. .
(^-^-O^-flH-i)
r
(r
i, 2,
. .
n)
., -,
and substituting
(2.5),
To deduce
p by
we notice that differentiating (2.4) with respect the ordinary rule for differentiating a product gives
to
Putting p
=a
in this, gives
.
.
g'(ar}=(ar -a^
(ar
-0r-i)K-r+i)
(r-^)-
(2.
Using
*
FUNDAMENTAL THEORY
Before proceeding further it is desirable to settle some questions of terminology. Firstly, suppose y(p) is given in the form (2.1). The simplest case is that of (24) in which g(p) is a product of n different factors,
an are the roots of the equation be spoken of as the zeros of g(p), since they are the values of p for which g(p) is zero. are the values of p for which the function Finally, they y(p] is infinite, and in this sense are described as the poles of y(p). In the more general case in which g(p] has n factors such as (p a^ a: is called a multiple (or more
,
g(P}
.,
also
precisely an
g(p\ or
of g(p)
o,
an
n-ple zero of
n of
tion
z
is
=x
-f-
iy
complex
x the
real
number,
we
<
FIG.
i.
\X(#
SC
the angle
sn
V
cos
= -'
<j>
tan
*V
shall call this angle the argument of x, written other names for it are the amplitude of s, (amp #), arg z or the phase of z.
;
We
*'This definition still leaves undetermined by 'an integral this does not matter in the applications made multiple of z-rr If desired, here. can be restricted to the range it IT, when it is called the Principal Value of arg x.
<j)
<f>
>
<
>
10
-\-
may be
where
arg (x iy). importance of this in the present connection is when y(t) is found from y(p) by the use of (2.9), it frequently involve products or quotients of co numbers, and the most rapid way of reducing suclrx expressions .to a real form is by using (2.14) for each
<
ar+^V^+yy*, =
+
(2.
1 4-)
The
(cf.
Example
5,
below).
One
mentioned
,-
thus, if
is
= arg( + iy) = tanr ^/*), x> /. T 5 \ ^ ^ = + tan-^/a), * < o\ oj we had written = tan~ (y/w) in (2-14) (and this
if
if
TT
,
<j>
very frequently done), there might be an error of argument of as.-}- iy, leading to an error in sign the final result. This point is particularly important Tables of transforms, and in problems such as Example below, where numerical values of some parameters not specified and thus may be taken with either sign. We now give some examples on the use of Theorem IIin the
Ex AMPLE
By
(2.6),
3. 6
V y
Xp + iXP+ *)(P +
;
;
TT
TT
-.
3)
<
and
3;
-t:..
* tan tangent
is
Jir
and
TT
whose
is x.
II
EXAMPLE
Using
(2.6)
4.
y '
a,
a
_|_
2(^2
!)(2
+ 4)
*.
Thus
= -t
4
3
sin
t -\
sin zt.
24
Problems on resistanceless circuits often lead to transforms which can be conveniently treated, as above, as functions
ofp*.
EXAMPLE
5.
y
(P
~\~
-\1
a
2 ~l~
1
j
fy[(P H~ *)
^ 2]
(*'
^
(P
l6)
This is a generalization of Example i, and could be dealt with either by resolving into partial fractions with
real
denominators, as was done in Example i, or by resolving into partial fractions with linear complex denominators, that is, by using one of the forms of Theorem II. Thus applying (2.9) to (2. 16) gives *
yy
- by +
-|j & Conjugate) j
(2. \
17) /y
Then
writing
* o give in (2.9) the sum of a pair of conjugate roots of g(p} pair of conjugate terms, of which only one need be written down. The method used below for reducing such a pair of terms to real
form
will be
12
the complex
numbers
in
(2.17),
we
(a *
wV
B r 2
re*(*+* l-#.)
e
i
=
J
7 .. 01
--2z
-k
^ Conjugate J
V J
- b)
/>
Theorem II covers the case * in which g(p} has no repeated factors ; for example, it is not applicable to Examples 2 or 4 above. It is easy to generalize the theorem to the case of repeated factors. Theorem II may be stated in
the
form that, to each linear factor (p ator ofy(p) there corresponds a term
<z
r)
of the denomin"
fgf*
in the solution.
(MS)
The
2
(p
(
Z>)
to each squared factor generalization is that, of the denominator ofy(p) there corresponds a term
L
i^e solution.
*
In
(3. i ) it
:
factors
was stated that/() and g(p) were to have no common this is desirable but not essential. With complicated
expressions (such as those of 16) it would greatly increase the labour to separate out such a factor, and, in fact, if f(p) and g(p) at both _have the same factor (p a), the term in e corresponding to this in Theorem II has zero coefficient, and so the application of Theorem II gives the correct result. If g(p) contained a
factor (p
2
a)
and
f(p} a factor (p
a),
Theorem
II could
be applied after cancelling the factor (p a), whereas a cursory inspection would suggest that it is not applicable and that (2.18) would have to be used (this, also, would give the correct result).
FUNDAMENTAL THEORY
each m-ple factor (p ofy(p) there corresponds a term
to
13
And
c)
of the denominator
in #ze solution.
__ ~
(P
__ __
consider the problem of
i
Example 2
*)*(P
_ + 4)~(P +
(2.18)
*Y(P
- 2i)(p + aty
Then by
(2.19)
and
+
leading to the same result as before. " " Finally, the question of uniqueness
must be
dis-
corresponds Given y(p) we have sometimes only a single function y(t). been able to find a function y(t) which has y(p) for transform by looking up y(p) in the Table of transforms but there is so far no reason why the function of t found this way should be the only function which has y(p) for transform. The same question arises in the integral calculus when we seek a function which has a given function for differential coefficient, and there it is found that there are infinitely 'many such functions (differing by additive constants). In the present case the matter is settled definitely by Lerch's Theorem* which states that, if two continuous
;
To a given function y(t) there corresponds a cussed. It is not clear, single transform y(p) defined by (i.i). however, that to a given transform y(p) there
* For a proof see C. and J., Appendix I. The actual theorem a great deal wider than that stated above, and includes also the useful result that if two functions y^t") and y z (f) are continuous except at a number of ordinary discontinuities (i.e. jumps in value), and have the same Laplace transform, they must have the same points of discontinuity and can differ only at these points.
is
14
functions have the same Laplace transform they must be identical. Thus, if we find a continuous function y(t) from a given y(p), for example by the processes described above which use the Table of transforms, this is the
continuous function which has y(p) for transform. It follows also from Lerch's Theorem that, if k
constant,
only
is
and
2r ) is not continuous provided y(t] is it can only differ from zero at isolated points.
> .
y(P)
>
k,
t>
2 (
3.
Some
additional theorems
III.
THEOREM
y(f),
\
// y(p)
is
the Laplace
and y(t] is continuous f and tends to y as t-> o, then the Laplace transform of dyjdt is py(p) y . (3.1) in If, addition, dyfdt is continuous and tends to y: as t-+ o>
then
the Laplace transform of
is
transform
of
-^
p y(p\ -py
- ^.
3>2)
Andifd y/dt
r
3
. .
~ z>3>
-,
as t-^- o, then
P -y
These
parts.
f
o
(3.3)
results
follow
To
we have
immediately by integration
by
dt
t
10
esu
not likS to
fs us 1
,
t c01
Ve
-
T L
t6
pla e transform s.
',
^'
FUNDAMENTAL THEORY
since ye~ Fi ->y
as
15
as
t-^-o,
;-
oo,
provided p
(3.3).
THEOREM
y(t), then
IV.
If y(p)
is
the
f
Laplace transform
i
of
the transform of
Jo
y(t'}dt' is -y(p).
(3-4)
For
roo
ft
Jo"
ft
\
-too
Jo
Jo
y(t')dt'
Jo
+-
PJO
<r**y(t)dt
since the integrated part vanishes at both limits. The results of Theorems III and IV can often be used as a simple example it follows to obtain new transforms
:
from
XP +*)
and
a) is
13
a
f
(1.7)
57 is
sin at dt
= ~(i
i
Jo
// y(p)
is
cos at).
THEOREM V.
is
)
(3.5)
"
For
e-**y(ti>t)dt
=
is
Jo
useful when it is desired to arrange For an the algebra of a problem in dimensionless form. of its use see 14. example
4.
linear
differential
equations
l6
D ny
for
is
to solve the
differential equation
(4.1)
if,
are constants, /() is a given function 1 ., D"" ^ are to take the given values
t
yw y
t
-,
We
yn ~i when
= o.
from o to
J
multiply (4.1) by e~^ and integrate with respect to oo. This gives
. .
.
V*{Dy + cj)-*y +
/(j>)
is
+ cny} dt = /(#),
(4.2)
where
written
down* from Table I and Theorem I for common types of function which usually occur. On the left-hand side of (4.2) we use Theorem
'
which can be
the
III,
which gives
Or
i}-
(4-3)
equation (4.3) is called the subsidiary equation corresponding to the differential equation (4.1) and the given initial conditions. It is formed from the differential r equation (4.1) by replacing o, ...,, on the y, r left-hand side, by fry- on the right-hand side, f(t) is
The
replaced by its transform f(p), and terms containing the initial conditions are added according to the rule
:
for for
for
Dy we
D*y
add
y
py
+ yi
foTWy
D^
ptyo+pfr+yt
.
p'-iy,
*The solution for a function whose Laplace transform is not known, or even which has no Laplace transform, can be obtained by the use of Theorem IX, 24.
FUNDAMENTAL THEORY
The
17
subsidiary equation (4.3) gives y(p), the transform To find y(t) from of the solution y(t) of the problem. methods of 2 are used. Since, by the uniquethe y(p), ness theorem of 2, only one continuous function y(t) corresponds to a given y(p), it follows that this will be the
EXAMPLE
with
i.
(D
= o, > o + = y Dy = y when =
2
1)3;
Q,
ly
o.
The
subsidiary equation
is
Thus
And
EXAMPLE
with
2.
(D
3>
i,
i,
is
equation
is
The
2, 2,
function
Ex. Ex.
i,
y corresponding to this has been found in and for a similar but more general case in
(D
5.
EXAMPLE
with
3.
i)
y
zero
sin zt,
y and Dy
subsidiary equation
when
> o, = o.
The
is
The
2,
function
3;
corresponding to this
y
i
Ex.
2.
EXAMPLE
with
2
4.
(D
.y
n z )y
D^v
=
==
cos
/,
>
t
o,
o.
==
3/0,
^ 1; when
l8
The
Thus
=
=
5.
__+ ^+
t
2
+7Z
2 2
)
Using (i.n),
(1.7), (1.8), it
t
follows that
y
EXAMPLE
with
sm
(D
;rf
3>
cos nf
+
t *
v,
sin
+
y
i)y
=
0>
= y when
.
t*e-*,
> =
o,
o.
is
Using (1,5) and Theorem I, the transform of t ze-t 3 found to be 2/(p Thus the subsidiary equation is i)
(7TT?
2
,
.
y*
Then, using
(1.5)
and Theorem
I,
we
find
y
5.
= $t*e-*+ yop-4.
as in 4, using the transformation of the equations in turn, gives a equations, the subsidiary equations, the solutions. for
Suppose,
example,
where the
T
'
'
i, 2, ., n, t> o, (5.1) given functions of t, and y r and r, take the values ur and w when o. r 7/1 Multiplying the equations (5.1) by -*, integrating with respect to t from o to oo, and using (3.1) and (7.2) gives the system of n subsidiary equations
. .
r=
/,(*) are
''
W> are to
Dy =
FUNDAMENTAL THEORY
in
+
j-i
K(M + w + M
)
S],
=
.
i,
...,.
<y n)
(5.2)
These have
ylt
.,
.,
an d then
EXAMPLE.
(D
+ z)x
tf
,
D^-h(D 2
to be solved with * =
= = D = y = Dy =
Dj/
2)3;
when
o.
The
px
Solving
+ (p 2 + z)y =
we
find
2X
4) 3
i)
Thus
V+
+ ^(2
+ 4/
cos zt
4- &(2tf
cos
;?
i)
6.
Comparison with the ordinary method of solving differential equations, and with Heaviside's operational calculus.
linear differential
(D
cjD""
+ cn )y =/(*)>
.
.
(^
consists of ^finding
4**',
r-l
I
(6.2)
where Aj,
.
.
are arbitrary constants, and the <%, an are the roots of <j>(p) o (provided these are all ., function different, if they are not the complementary takes a slightly different form provided for by the theory)
. .
.,
An
20
and adding to this a particular integral determined from This gives the general </>(D) and /() by certain rules.
solution of
initial
(6.1).
To
satisfying given
algebraic
(6.3)
To
an
determine^ from
(6.3)
we have to put
the right-band
.
.
side into partial fractions,* and for this the roots a^ ., of <j)(p) o are needed thus these must be calculated in either method, but when they have been found, the
work by
shorter
;
the
Laplace transformation
. .
method
is
much
for example, if a^ an are all different, ., y can be -written down by (2.8), while with the older method of solution the n algebraic equations for the constants A n of (6.2) still have to be solved. Thus the ., Aj, advantage of the Laplace transform method over the ordinary one increases enormously as the order of the equation to be solved increases. The same remark applies to simultaneous linear differential equations the greater the number and order of
.
.
is
formation method.
The
(6.3)
is
_
(6.1),
part of y(t] derived from the term /(p) /</>(/>) of a particular integral of but it may differ from
the particular integral found by the rules of the older theory by terms of the complementary function. To see the connection between the present method and Heaviside's operational method, we again discuss Heaviside usually considered only the case in which (6.1). " the initial conditions were zero, and unit f(t) was his function ", which is zero for t o and unity for t o,
<
>
* If /CO is not one of the simple functions dealt with hitherto, it may not be possible to put /()/<() into partial fractions, but the solution can always be found as in 24.
FUNDAMENTAL THEORY
and
is
21
represented by the symbol 1. On the left-hand side is replaced by p ; this is regarded as an of (6.1), operator as a number as in the Laplace transformation method). (not
Thus
(6.1)
leads to
'
'
'
(6<4)
the operational expression for the solution y. The solution itself is then derived from (6.4) by Heaviside's expansion theorem which is similar in type to (2.8). For this problem, the Laplace transform of the solution is
which
is
(6 s)
'
and
this differs
from
In
(6.4)
all
the denominator.
expression transform.
is
just
in
times
i,
corresponding
Laplace
the Laplace transformation procedure is designed for, and applies immediately to, the " " of dynamics and electric circuit initial value problems theory, in which differential equations have to be solved o with given initial conditions at t o. for times t " In other fields, such as the theory of elasticity, boundary " in a differential which value problems arise, equation x and the has to be solved in a region, say o /, solution has to satisfy some conditions at the boundary at x the L The some older x boundary o, and method of solution, as the sum of a complementary
As remarked
>
< <
a particular integral, applies equally (6.2) and both types, since, in a boundary value problem, still be n conditions, some at x o, and some A n The Laplace to determine the A 15 at x ., I, transformation can still be used, since the initial values be regarded as arbitrary constants some ., y n may ylt of these may be fixed by the conditions at x o, and the from at x the conditions can be determined / remainder function
well to
there will
22
at the Z, freely hinged L horizontally at the other end, x== for differentiation with respect to x, the differWriting ential equation for the deflection is
and clamped
where E, with x
This has to be solved D# = o at x I. o at x o, and x We solve (6.6) with values o, y^, o, y3 of y, Dy, D zy, D 3;y, as unknown respectively, at x=.o, regarding y^ and y 3 Iconstants to be determined from the conditions at x
= DV=
I,
and
< =
I,
(6.6)
Writing
to
the
y for the Laplace transform of y with respect present independent variable x, the subsidiary
is
equation
Thus
mi
The
conditions
= Dy = o at x =
'
I,
give
24EI
mr
Thus
and
7.
TAe justification of the solutions of 4, 5 /row pure-mathematical point of view. In deriving the subsidiary equations of 4 and 5 some assumptions were made, which, from the point of view
method is well suited to problems of this type, parwith concentrated or variable loads. For other examples Jaeger, Math. Gazette, 23 (1939), 62 Pipes, Journ App. Phys., 14 (1943), 486.
; ;
* This
ticularly see 21
FUNDAMENTAL THEORY
of
strict
33
pure mathematics, would have to be defended. For example, it was assumed that y and its derivatives had Laplace transforms. To make the solution quite rigorous further discussion would be needed, but this can be avoided by checking the results obtained by the methods above, that is to say verifying that they do satisfy the differential equations and initial conditions of their problems. This can be done quite generally, and since the process itself is purely algebraical and adds nothing to the theory of the subject it is sufficient to state the
results here.
For the problem of 4, the ordinary linear differential equation of order n with constant coefficients, it can be that the solution obtained by the Laplace transformation method does satisfy the differential equation
verified *
verification
5, simultaneous ordinary linear it is possible to give a similar general that the solutions obtained by the Laplace
transformation method satisfy the differential equations, and that they satisfy the initial conditions, provided that the determinant of the coefficients of the highest powers of in the terms does not vanish. This restriction is required, not because of a flaw in the method of solution, but because of a fault in the specification of the problem. To see what it implies, consider the system of equations in this case the determinant in question is (5.1)
it is possible, by adding together appropriate constant multiples of, the equations (5.1), to form at least one equation containing no second derivatives. Thus, instead of having a system of n second order equations, which could be solved with 2n initial conditions, we have i (or less) second order equations, in fact a system of n
If this vanishes
* C. and
J.,
34, 35
Doetsch,
loc. cit.,
Kap.
18.
24
and some
first order or algebraic equations, and for such. a system only zn i, or less, arbitrary initial conditions can be prescribed. If the problem is stated with initial conditions, there must be relations between these or there will not be a solution. Problems of this type occasionally arise in practical also similar difficulties occur in switching
applications,
problems
in
20.
in circuit theory.
EXAMPLES ON CHAPTER
i.
z.
3>o>
yi,
initial
values
(i)
(D
M)y
i.
fy =
L,
-+
f*
fyo\
A1
(ii)
(D 2
+ 5D +
6)3;
i.
(iii)
(D
n*)y
(iv)
(D
i)
25
i
."1
'
'
ttn
f.
3-
If
3;
d*yldx*
x,
with
dy/dw
2
=
3
o at
/,
I,
show that
zl
~ sZ
).
28
series
with a leaky condenser of capacity C and leakage conductance i/G, the subsidiary equations are
initial
Returning to (8.6), we remark that it involves the current in the inductance and the initial charge on the condenser (these are related to the initial kinetic and potential energies, respectively, of the current).
Further, since the terms involving them appear in the right-hand side added to the transform V of the applied on voltage, it follows that the effect of initial charge the condenser is the same as that of an applied voltage whose transform is ( Q/Cp), that is of a constant voltage in the ( similarly, the effect of initial current 1 Q/C) inductance is the same as that of an applied voltage whose transform is Ll, this will be found in 19 to be an im-
pulsive voltage LlS(f). The same remark is true in. general that initial conditions may be treated as voltage sources of the appropriate types.
9.
Examples on L, R,
circuits.
EXAMPLE
o Constant voltage, E, applied at t R, C in series,* with zero initial current and charge. Here the subsidiary equation (8.6) is
i.
to
Thus,
T
= + (R/L)p + (i/LC)) R _ _
2
f '
'
~LC
^~LC
later.
The
*
L=R=
see
29
(1.5),
using
Theorem
find
if
and
(1.7),
and
(1.9), respectively,
we
n2 n
n
2
> = =
=r
flr' *,
if
o kz
JL
if
< o.
C
EXAMPLE
voltage
2.
is
condenser discharged at t
.4
sistance L,
R.
Here
Q=
CE,
o,
V=
o,
and
(8.6)
becomes
which
is the same as (9.1) except for the negative sign. This implies that the current starts flowing away from the positive side of the condenser. Using (8.5) and (9.4), the transform of the charge on the condenser is found to be
'
CE
CE
Then,
in the
where the notation (9.3) has been used. case n z > o, we have
Q.= CEfi-Wcos
EXAMPLE
nt 4-
sin nt\.
3. Steady current E/R is flowing from a battery of voltage, Ej through a parallel combination of condenser and inductive resistance L, o the (Fig. 3), when at t switch S is opened. It is required to find the potential across the condenser terminals.
After the switch S has been opened, the circuit consists of L, R, C in series. The initial charge on the condenser
30
is
is CE, and the initial current 1 in the inductance the minus sign being chosen, since, from the conditions the from P os1 of the problem, this is flowing away With these initial values (8.0) side of the condenser.
becomes
=
(9.3)
Hence, con-
>
nt
+-
sin nt\
FIG.
3.
If v
= Q/C
it
is
minals,
__Q_E_ v
c~p
R)
i)
(p
cos nt
-( -=
ju)
n\zL
-rr^
RC
sn
3!
the simplest type of circuit breaker restriking problem in which it is required to find the voltage across the contacts of a switch after they have been opened. v. In this case the voltage across the switch S is E Anothe-r method of solving this problem is given in Ex. 2
at the
end of
this
Chapter.
EXAMPLE
at
t
=o
4.
to
an L, R,
(wt
sin (o>t -f- <) applied Alternating voltage circuit with %ero initial current
and
charge.
sin
is
Since
transform
-f cos
wt
sin
<^,
its
,
<j>
*
'
o>
'
'
'
(9 5)
(9.3),
_E
L
2
j>
'
sin
-{-
<ft
*
,.
'
(9
2
We
Here
consider
>
o.
to find I
we may
put
denominators,
or
notice
that
its
+ iot)(p +
latter
2
t
M-
in)(P
and use
, I
(2.9).
o)
a
The
sin<^
i
method
.
gives
.
*
.
LI E ""LI
f
=H
EC
+ ia> cos<j> ^}V 1 r~fv +n & Conjugate + W)\8 + w sr - + m sin + a>- + cos ut+tni
ico
*
3zo)[(ju,
<>
Conjugatej
(9.7)
* Notice that if only the steady state part of I is required, only the terms of (2.9) arising from the zeros i<a of the denominator of (9.6) need be written down. The other zeros in give transient terms. See also 13.
/u,
32
as
iti
2,
Ex.
5.
Thus, writing
-[a/to)
O)
2
(/i
)]
(9- 8 )
2 [(/x
2
)
sin
|u,aj
cos <j
-+-
i[no)
cos
<
2^n
sin ^>]
2>2^
*|
(9.7)
becomes
S 2 ),
(9.9)
where Z and 8 may be written out explicitly as the mo d.xilus and argument of the complex number on the left of C9*^)i and similarly Zlt B lt Z 2 8 2
, .
work
as built
up of
netth<
of
M rs
*^OT^
W
FIG. 4.
,
type studied in 8, with their terminals connected in som< Let Lr R r Cr be the inductance specified manner. resistance, and capacity in the r-th of these elements thf (Fig. 4), let Vr be the voltage drop over it, o let I r current in the element, with initial value I r and let be the charge on the condenser, with initial value <!$,. To he quite general mutual inductance must be included
,
be
33
* between the r-th so let Ta be the mutual inductance and s-th elements. Then the equations for the r-ih circuit element are JT
rk
L. _> r ^
M
iVJ.
j-r
rs
=V
v r,
do i.} ^IU.A;
(10.2)
dt
=I
r.
8,
is
found to be
4-
#1 4-2-s
where
="V
r
Li
4-
5!
1-
=L +R + = M p *rt
zr
r
--,
^"rP
(10.4) (10.5)
ra
the form of the network, further equations can be written down connecting the 1 T and r in different circuit elements. Firstly we have Kirchhoffs First
From
V
.
Law
f
]^]
= o,
= o,
at a junction.
(10.6)
This leads to
^
Kirchhoff's Second in the network, the
at a junction.
(10.7)
externally applied voltages, V, in the closed circuit, equals the sum of the voltage drops, Vr over the circuit elements included in the closed
,
that, for
circuit.
That
is,
writing
^
o
to
* r , will have a sign depending on the way in which the coils are wound, and on the convention of sign we set up for the Currents in the different circuit elements.
t
either: directly or
Usually labour can be saved by specifying the currents, by the 'use of mesh currents, so that they satisfy
3
(10.6) automatically.
34
a closed circuit
Vy
and thus
V CO ZV
r
y,
'
'
= ]T?.
Using
(10.8) in conjunction
with (10.3)
we have
By choosing suitable closed circuits, sufficient equations of types (10.9) and (10.7) can be found to give all the I r When voltage V, applied' to the terminals of any network with zero initial conditions, produces at the terminals current I given by
.
V,
(10.10)
we
of the network s the generalized impedance viewed from these terminals * for example, %r defined in (10.4) is the generalized impedance of an iir R r> C r circuit. Now for a single circuit element of a network with no mutual inductance and with zero initial conditions, (10.3) becomes
call
; ,
"
"
*r I r
=V
f.
(io.ii)
That
and zr are connected in the same way as current, voltage, and resistance in direct current theory. Thus in this case the generalized impedance of the network can be written down by combining the zr of the circuit elements which comprise it according to the ordinary laws for combining resistances in direct current theory. For example, the generalized impedance of a combination of n generalized impedances z1} z2 zn in parallel is ., given by
is,
Ir ,
Vr
1
*
*
2(10))
is
= 1 + 1 + ...+JL.
impedance
of
^ (10.13)
alternating
35
Again, the generalized impedance of the ladder network, Fig. 5, can, by repeated application of the result for two resistances in parallel, be written down as the continued
fraction
*==*!+-
-1
,
(10.13)
FIG.
It
5.
the same way from the remarks at the end of 8, that the current in any branch of a network, due to initial current 1 in an inductance and initial charge on a condenser in that branch, is given by
follows in
= Ll - (QfCp)
(10.14)
where z(p) is the generalized impedance of the network viewed from a pair of terminals in the branch.
Fig. 6 are coupled by Voltage V(t) is applied to the o with zero initial conditions. It is reprimary at t quired to find the secondary current (the primary resistance
mutual inductance
M.
is
neglected).
(10.3) give
Mpl,
4-
L^ + R + JLl a =
2
o.
36
(LX L 2
M
FIG. 6.
If
V = E,
constant, so that
V = E/p,
this
becomes
ME
and
where
ME
a
sin
a".
The
case
LA L 2 =
2.
is
discussed in
20.
Z/ze
EXAMPLE
Ira Z/ze
circuit of Fig.
t
= o,
the switch
is
closed.
We
Kirchhoff's
(10.9) respectively,
for
the
law. closed
circuits
ABD
and
BGHD
are,
CE
CE
37
T= ~
'
we find Zjr 2n *
f
E
where
2
+ VS
n 2 /'
i/LC.
Therefore
EXAMPLE
the
current
current
In the circuit of Fig. 8 it is required to find 3. I in the inductance L caused by prescribed o I supplied at the terminals AB, starting at t
with zero
initial conditions.
FIG.
8.
From
^ R ^ Lp
RLp
RLC
+ Lp + R.
L
i? Also the generalized impedance of the inductance hav Lp. Since the same potential is applied to both we
38
For example
i,
for
i>
o,
so that I
i/p,
and,
if
^>
-
p(RLCp
o,
+ Lp + R)'
(/ft
er** sin
+^
1
where a
4.
JL
_a
2
,
= tan"
(]8/a).
Bridge networks. We consider a general bridge network represented by to be the scheme of Fig. 9, and suppose known voltage
EXAMPLE
-Is
H
FIG.
9.
o with zero initial conditions. It is required applied at i to find I 5 which is the galvanometer current in most Choosing the currents as shown in Fig. 9 applications. to satisfy KirchhofFs first law, the subsidiary equations for the closed circuits ABDHA, ABFA, BDFB, re,
spectively, are
39
we
find *
1
+#2
*!
-*a
%A
*a
("4)
where
z^
#
# #
the bridge can be used for measurein (11.4) ments is determined by the form of zz s3 si-jjs^ as a function of p. If, by suitable choice of the circuit constants, this can be made to vanish for all p, we have
15
o,
As an example
x
of this
which
consists of inductance
and
resistance
Rx
resistance
parallel.
#3 of
R4
in
*!
+R
15
#2
=R %=R
2,
3,
#4
R4 Q>"
TTniQ JL J..LU.O
***2
y y3
y14 y
Then
and
if
we have
voltage
I5
o for
>
o,
may
be.
is
* It is hardly necessary to point out the advantage of using determinants, particularly when, as here, only one of the currents needed.
40
,As
an example of a bridge whiqh cannot be balanced in this way, consider the case in which z1 K-2) #3 RS> ltp -\- RI, #2 and #4 consists of resistance ^ in series with a parallel
combination of resistance rz
and capacity C,
so that
and 2 2#3
r"
Q>+1
(i i. 6)
all values of p Clearly (11.6) cannot be made to vanish for by choice of the circuit constants. However, we rnay use the circuit as an alternating current bridge by requiring that the steady state part of I 5 is to vanish when the applied Anticis sinusoidal and of frequency o>/27r. voltage
and 13, this requires that the real ipating the results of imaginary parts of the numerator of (n.6), with j& replaced is that shall both vanish, by ioi,
and
Lfa + ra ) 2 LCr^co
= (R R - Vi) = Rj^ + r - R R \ J
2
2
'
" x~\ (T x/ v
.
a)
Alternatively the bridge may be used as a ballistic at bridge ; in this case a battery of voltage E is applied o with zero initial conditions, so that t E/p in and the that total is the charge requirement passed (11.4), through the galvanometer is to be zero, that is
V=
= o.
'0
By Theorem
VIII,
lim 1 5
p->0
^
= o.
(n.8)
V=
rzY
does not
E/p,
it
we must have
R 2 R3 =
R^ +
41
In practice, instead of applying a battery to an initially dead circuit, it is usual to open the battery circuit when steady current is flowing from the battery. In that case the impedances %, #4 of Fig. 9 must be precisely ., specified, and the circuit of Fig. 9 (with the arm removed so that I 2 Ij) has to be studied with known initial values of the currents in the inductances and the on the condensers. charges
.
.
AHD
12.
factors of a circuit
.
.
I n in the transforms of the currents, say I l5 ., various branches of an electrical network are found by the solution of equations of types (10.9) and (10.7), and always take the form
,
The
t
r
_ ~~'
fr(P)
_
,
T
'
'
'
"
'
'
'
(,~ ,\
(
where
the same for all l r it is in fact the determinant of the coefficients of the l r in the equations and depends on the circuit only. The numerators fr(J>] depend on r, and involve the initial conditions and the applied voltages. " Before we can find the I r we need the roots of the period
<3>(p) is
"
equation
o,
. .
(12.2)
of (12.2) there corresponds a term e~^ sin (vt -j- 8) in the I r , that is, a term 1 of period 2w/v and damping factor er **. The essential point is that the period equation (12.2),
and
we know
that to a root
ft
*"
whose roots
give the natural frequencies and damping factors of the circuit, is always found in the course of the calculation -of any of the l r but the I r contain in addition the complete solution of the transient problem. Thus in it follows from the first expression for 1 that Ex.
,
ir,
2,
the natural frequencies of the circuit are n[(3 Vs)]*/ 377 In demonstration problems, such as most of the examples linear obvious have to solved here, 3>(p) is usually arranged In practice $(p) may be a polyor quadratic factors. nomial in p of degree n with coefficients given numerically,
-
42
we
Here an example of such a problem is given in 14. give some references on the methods available for the
solution of algebraic equations.
methods are
1
For the available, described in text books on algebra.* cubic, some tables and charts are given in Jahnke-Emde.'j
Numerical methods can be used for cubic and quartic equations, and must be used for equations of higher degree. Real roots can be found from first principles, using a graph and improving the rough values obtained in this way,
or
14) by interpolation (an example of this is given in cubic has by Newton's method (W. & R., 44). always one real root which can be found in this way dividing by the corresponding factor leaves a quadratic. The more sophisticated methods for finding real roots are
either
Homer's -method (C. S., 475, W..and R., 53) and the root squaring method (W. & R., 54), the latter is preferin
Problems on resistanceless circuits lead to equations z p with real (negative) roots which are easily solved by the above methods. For equations having several pairs of complex roots, the modified root squaring method seems the most satisfactory J another method, and also a method of successive approximation to a complex root, is given, by Frazer and Duncan. If any of the roots of the period equation have positive real parts, the solution will contain terms which increase in this case the system is exponentially with the tune said to be unstable. If all the roots have negative real parts the system is stable. Clearly it will be useful to
able.
; ;
will indicate
whether a system
is
stable
* Charles Smith, Treatise on Algebra, Edn. 5 (1921), referred to as C. S. ; Turnbull, Theory of .Equations (Oliver and Boyd, 1939)- Whittaker and Robinson, Calculus of Observations (Blackie, 1924), referred to as W. & R. t Tables of Functions (Edn. 2, Teubner, 1933). I Brodetskyand Smeal.Proc. Cambridge Phil. Soc., 22 (1934), 83. Note that when there are several pairs of complex roots the method given in W. & R. is unsatisfactory. Proc. Roy. Soc., A 125 (1929), 68.
43
or unstable, without the labour of solving the period equaOne such criterion is Hurwitz's * which states that tion. the roots of the equation
all
if all
the determinants
are positive,
ar
is
down
>
the determinants,
Steady state alternating currents 12 that if voltage V is applied to was remarked in o, the transform of the current I r in any any circuit at t branch of the circuit will always take the* form
13.
It
(13-1)
circuit
only,
and
Now
d>
suppose
so that
d> -\-
sin
and
(13.1)
+ p sin
<f>)fr ( p) _j_
gr(p)
'
JJ
o If the system is a dissipative one, the roots of <(p) have negative real parts, and thus will give rise to If only the steady state altern ating current transient teams.
will all
* Frank- v. Mises, Differentialgleichungen der Physik (Edn. 2, Vol. i, p. i6a. This is equivalent to Routh's Rule, 4, 1884), Chap. vi.
1930),
44
is
we need only evaluate the two terms of f P* Ctj2 *<*> These corresponding to the zeros
(^ - 9)
Conjugate!
2icu$(i(a) '
J
where
^Thus,
if
circuit due to applied voltage V and any initial conditions has been found, the steady state alternating current be deduced immediately. An example has been given in As another we find the steady state alternating (11.7). current in the secondary of Fig. 6 due to the primary
can
voltage
(p*
_
sin (cot
a
+ 0).
(13.3) give
Mj>E
The
+ o^KI^L,
+
(a)
cos
My
<f>
+ p sin + R L^ +
2
,
,
steady state alternating part of I 2 is icof *'a> sin Mi'a>E(o) cos ft </>)g
'
^
.
\ J
.
COS
(ait
where and
If
^
+
,
S),
2 2
]
SN
)^
}*
)]
the generalized impedance of a dissipative 60 * circuit looking in at a pair of terminals, and voltage Ee* is applied at these terminals with zero initial conditions, the transform of the input current will be
5f(p)
is
E
y
'
(p-io>}z(p)
be
45
treatment of more complicated circuits 14. Numerical In the problems previously considered, the denominators of the transforms of the solutions have had obvious linear or quadratic factors, so that explicit expressions for the In more solutions could be written down immediately. the denominators may contain excomplicated circuits, of the third or higher degree in p which have
pressions
no simple
If the circuit constants are given. factors. the denominator can be found numerically, the zeros of mentioned in ra, and the solution methods the of one by constants are unspecified, families completed if the circuit the behaviour of curves can be obtained which will show of the circuit under any circumstances.
;
+L
FIG. 10.
from both these points In this section we study, due to unit I in the circuit of Fig 10, view, the current In o with zero mitial conditions. at t volTage applied the usual way we find
of
(14-1)
;*Xi
becomes
As an example
in
which
th
^ c=
10
)
,
-
10
*='.oO +i
4
2
ohms
Then
.1)
(14.2)
To
evalulte l!
tCoots
^t^^^.
46
are needed.
io 5 # in
this,
gives the
more
(14-4)
+ x + 0-4 = o.
This cubic can be solved by any of the methods of jj 12, but it is very simple to work entirely from first principles. From a graph it is found that there is a real root at This root can be found approximately x 0-77. more precisely as follows we find /( 0-00353, 0-77) between /( 0-78) 0-00279, so tnat tne root ^ es between ran if and further, linearly 0-77 /(#) 0-78 at its values at these two 0-776points it would vanish
:
By
as repeating this process, as many places of decimals desired can be found. Working with three places of
decimals only,
f(x)
we
divide f(x)
2
by
(x -f 0-776),
and obtain
a ) ].
(i4'5)
Returning to (14-2) we find, u'sing Theorem II as in 2, Ex. 5, that the function whose Laplace transform is
where
V)
8= ar g (P+a-7j +2^)+arg(a-^+^). (14.7) 5 Using the values a = 0-776 X io 5 ^ = 0-312 X io 5 a = io 10 found above, and writing 0-647 X io
2
,
,
io-6 T, so that
T is
and
6 47
T- 1-485)
(14.8) =
to
77/8
To
when
T=
plot I
correspond
47
increases of 22^ in the argument of the sine, and only the sine need to be looked up. The caleight values of and B being the first culation may be set out as follows,
(14.8),
the It very frequently happens in practice that it is not transient behaviour of a circuit for given numerical values
of the circuit constants
which
is
the transient behaviour for any values of description of The circuit of Fig. 10 will the circuit constants. The first thing to this point of view. from be studied do is to express (14.1) in a form involving only simple ratios, dimensionless and fundamental quantities, such as time constants of parts of the circuit, etc. Thus, writing
now
(IA. 1} V
becomes
^-pt+dp*..
terms of
of nit by
(i4-io)
.
so that the solution A = p/i, V. Thus writing Theorem k = nln = we have jRI +A ^
i?T
obtained
i,
(14-")
The
right
sionless,
parameters
the two dimenhand side of (14.12) contains to study the general behaviour b and k
;
48
of the circuit
arbitrary value, and see how the solution changes 'as the other takes all possible values. shall give b fixed values, and for each of these allow k to run from o to oo, that is,
We
say,
decrease from oo to zero. First we have to study the position of the roots of the
R to
period equation
x
for constant b
+k=
(14-13)
could give k a number of values and solve (14.13) for each of these, but this would be very laborious, and also we could not be sure that the values of k chosen would fairly represent all types of behaviour in the circuit. Both these difficulties can be avoided by the following procedure the cubic (14.13) has always one negative real root, call this a, and
and k >o.
We
db if\ suppose that it has also a pair of complex roots of modulus r. Then we know that * the sum of the roots at two of (14.13) equals kb, the sum of their products k, a time equals i, and the product of the roots equals
Or, written in
full,
ocr
where
for
r2
= k, = +
.
(i4 l6 )
(14-1?)
2
r)
. .
If we regard a as a parameter, and solve (14.14) to (14.16) r z and k in terms of it, we find
,
'
That and
is,
3 the values of r a, i^/(r ), with given by (14.18), are the roots of (14.13) with the
2 2
If
au
an
...,
oc n
are
the
xn
then
Zoc!
+ a^- + + = a Zoc^ = a
1
. . .
_!*
z,
i}
ajoc,
an
Sa^ocs = (-)*.
a a>
.,
49
value of k given by (14-18). If the value of given by turns out to be greater than r z the same (14.18) argument three are real roots, shows that there
<x,
and
-j-
^/ (2
_r
of k.
Naif
5um
of roots
FIG. ii.
$ -\- ii) In Fig. 1 1 the position of the complex root of (14.13) is plotted as a function of the parameter a, When <x o this root is for the values 2, 4, 8, 9, 10 of &. curves shown i for all values of b, and as <x increases the The positions of the roots for the values are traced out. of a are marked by successive crosses 0-2, 0-4, 0-6, i-o, 2-0 on the curves, k is given by (14.18), and, when there is
50
a complex root, increases as a increases, that is, as we move round the curves. If b 9 there is always a pair of complex roots of If b (14.13). 9 there is a real triple root for k == i/(3V3) at a i/Vs for all other values of k there is a If b pair of complex roots. 9 there are three real roots for some values of k these are indicated
<
>
by
plotting (dotted in Fig. n) half the sum of the real roots other than a as abscissa, and half their difference as ordinate, measured below the axis. T^us a diagram su ch as Fig. gives an immediate indication of the way in which the roots of (14.13) "vary
.
FIG. 12.
with k and
study the way in which the current I derived from (14.12) varies with k and b, we select typical values of b from Fig. 11, for example, the transition value b than 9. 9, a value of b less than 9, and a value On each of these curves we choose a few greater
b.
To
points and calculate I for these using (14.6). In Fig. 12 the resulting set of curves for the case b
is
representative
shown.
Here RI
0-2,
is
0-05,
o%
i/V3,
2,
values
which
correspond the values 0-049, '93> 'i52, 0-192, 0-270, 0-469 0-901, 1-784, 3-558 of k. Clearly these curves, together with two or three such sets for other values oi 1 give a fairly comP lete indication of the way in V^J the which transient behaviour of the circuit on
k and
depends
b.
51
vacuum
tubes
Transient problems in valve circuits are easily treated of course, by the Laplace transformation method, provided,
that the valves are operated in linear parts of their characteristics. Only triodes are considered here, but similar considerations apply to other valves. The characteristic equation of the triode supplies a linear relation between the anode current, I a , and the
namely
(15.1) c ), tfV, pi (V a ) where p is the anode slope resistance, and fj, the amplificaIn transient problems, the quantities a tion factor.
.
ffl
-V +
Va
V,,
-V
respectively,
V V
,
FIG. 13.
Vc
the variable parts of the appearing in (15.1) are the transient behaviour voltages and current caused by the constant volthey are superposed on being studied caused current and anode by the anode, and grid bias,
,
tages
batteries.
The grid current is negligible in normal cirIn to be zero. cumstances, and here is always assumed whether the interelectrode any problem it must be decided if not, itself can be neglected capacities in the valve below. By virtue of they are easily included as shown an valve the to equivalent by replace (15.1) it is possible circuit element consisting of resistance p and a voltage V ) inserted between anode and cathode ; ^(V but it is a little more fundathis is the common
mental
to
As
first
work directly from (15.1) as will be done here. that example we consider the type of problem
:
practice,
in the circuit occurs in the amplification of transients of Fig. 13 interelectrode capacities are neglected, and
52
steady conditions prevail due to grid bias and plate batteries t o a known addi(not shown in the figure), when at it is required to is applied- to the grid tional voltage Since and find the change, l a in the anode current. the changes in currents and voltages it causes are superto know the need not do we the on steady values, posed latter, and need only write down equations for the changes, o. All the currents and voltages all of which vanish at t
ff
ff
13-16 are to be understood in this sense. We notice that (15.1) implies a convention of sign, namely, I is flowing in the direction of the arrow in if a the same current I a also flows from the cathode. Fig. 13
marked
in Figs.
V >V
;
FIG. 14.
The
are
the signs being determined by the convention stated above. Also the transform of (15.1) is
pi.
= VB + /*?. - (x + /OV
we
find
8
(15-4)
+ * + (i + /*>
= /*?,,
ta-S)
and from this l a can be found if V g is known. As a more definite problem we take the case of resistance-
53
value of 2 (which capacity coupling, Fig. 14, and find the would be the grid voltage of a second valve) corresponding o to the grid of the t to constant voltage E applied at Here we have, in the notation of Fig. 13, first valve.
E
*
i
+ C (R
2
Also
R 1Ra C aj)I a
(iS-6)
la
FIG. 15.
where (15.5) has been used in the last line of (15.6). Since the denominator of (15.6) is a quadratic in p, it is easy to
write
down
of inter-electrode capacity and more complicated system, we consider the circuit Here grid-plate capacity is included in s 3 of Fig. 15. anode-cathode capacity grid-cathode capacity in z, and We seek the anode voltage V a due to voltage V in *2
To
a rather
applied at
at
the terminals
AB when
steady con-
ditions prevail.
54
*& +
(15-7)
(15-8)
(15-9)
.* a l a
=-V
is,
a,
(15-10)
(15.1),
which
in this case,
FIG. 16.
Solving
^^ V =
*
n.
we
find
(I5-I2)
whence
can be found if the impedances z are known. Finally we treat an oscillator circuit (Fig. 16) from the in present point of view. Here there is a closed system which the grid voltage is not prescribed from outside, but currents As always, there are steady is free to adjust itself. and voltages due to plate and grid bias batteries, and the
I,l a
is
,
Va
Va V
,
ff
to
have
all
this solution
are departures from these clearly one solution of these zero, and the problem is to find how behaves if the system is disturbed. To fix
:
55
suppose there is a change of current I in the inductance L at t o, then the subsidiary equations are
Rl = ? +
Li
(15.15)
The equations (15.13) to (15.16) can be solved for All I, I a) a g in terms of 1 and the circuit constants. the solutions will have for denominator
V V
,
o
i
Cp
i
Cp
o
,
1,
(15.17)
2, 3,
ef***,
where a r r=i,
,
have negative real are the zeros of (15.17). parts the disturbance dies out gradually and the system is stable if any of the oc r has a positive real part the system
If all the
oc r
;
is
theoretically unstable, actually it executes an oscillation of amplitude determined by the limits to tshe linearity of the characteristic. The condition for this, which is the condition for the circuit to act as an oscillation generator, is obtained by applying Hurwitz's criterion, (12.3), to
(15.17);
it is
C[p(L
1 6.
+ LO + (i + ^(RLt + LRa)]
Filter circuits
essential feature of such circuits is that a number of precisely similar circuit elements are arranged to form a
The
* This
is
the determinant
done solely as an illustration, all that is needed is A which does not involve the initial disturbance.
S^
we repeating pattern. Mathematically this implies that can write down a general relation, called a difference equation, connecting the transforms of the currents in successive elements, and by solving this we get a general expression for the current at any point. As a definite example which includes many cases of Here practical interest we consider the circuit of Fig. 17. %', there are i impedances impedances z, and m
arranged as shown. Yoltage V is applied at the beginning through impedance ss{> and the whole is terminated by We require the current Ir at any point due impedance # to the voltage o with zero initial conditions.* applied at t
.
t
I
_f~*
vi;
VUf
T
FIG. 17.
m"
T
"
The
meshes
of
Fig. 17 are
The equations (16.2) which connect the transforms of any three consecutive currents are a system of difference
equations.
where
gives
We
z
seek a solution of
r.
is
independent of
Ae^,
(2$ -f #')0
+ #e
/
= o,
.
that
is
cosh
Q= i +
-.
(16.4)
not zero, as in the discharge of a. filter circuit, there -will be terms on the right hand sides of the subsidiary equations which depend on the particular
initial conditions are
* If the charged
impedances z and
57
Thus,
if
is
defined
by
and
* Of
it, is
It
both signs of
positive sign
occur in
= #74*,
and, since
may be taken,
sinh
with the
(16.6)
-*
40
(l\*.
22
2z
FIG. 18 (a),
FIG. 18 (6).
To
and B in (16.5), this must be and (16.3), which give two equations
(*,
A(#
-f
=
)
1)0
a
i)ff+2
sinh (/w
a)0
(16.7)
This general expression simplifies greatly when_ and % i have values corresponding to the common terminaThe network of Fig. 17 may be regarded as built up tions. of either T elements (Fig. i&z) or ff elements (Fig. 180) e ta (16.3) corresponds to the c differential equations. jT functions of r on the right hand function, and, if there had been to be includetj, had have would a integral of side (16.3), particular in Differential Piaggio, brief reference to the theory is given be The Laplace transformation method can Equations, p. 315. ^ ctt. cf. Gardner and Barnes, loc. applied to difference equations,
*The
is
analogous to that of
mPl^
58
connected
way
in. tandem, and the general expression (16.7) takes specially simple forms when circuits built up in this are open or short circuited. For
example
series
impedance,
V
ss
cosh
(ii)
zz sinh (m
t
'
Ol0 '
at
oo,
o.
r?-
The
i.
I r ~~
(iii)
#o
=z
0'
9'
sections,
with
at the
V cosh
*4
Jar',
(m
'
(l6 IO)
'
(iii),
but with
open
jr
(
'
'
) %i
8 cosh mff
/
'
,
'
I2)
If
"9",
cosh cosh
a"' 1 f cosh 5
\
.
.
- cos
(cosh \
\ fcosh an/ \
an
cos
- cos (2n ~
cos
an
l)?rN
)
(J6.I7) *'
v
sinh w0
= 2"- 1 sinh
.
.
(cosh
\
.
-\
nj\
("cosh
- cos
0- cos (cosh
^^,
\ n J
(16.14)
59
appears that (16.7) can always be expressed as a quotient ^or of polynomials in p. Both (16.13) and (16.14) no any positive integer n, even or odd. It is often advisable to use them to see the exact form of the denominators of the transforms of the currents, though actual evaluations of the currents I r are best carried out by working in terms of 6 as in the examples below.
it
applied to a circuit R, % i/Cp, with no additional series impedance, and open circuit at the far end.
i.
EXAMPLE
Constant voltage
of m,
sections in
which
%'
Here
(16.4) gives
cosh 6
= +
i
|RQ>.
(16.15)
And, using
(16.11),
CE
T
sinh (m
- r)9
^
sinh
cosh mQ
To evaluate I r we need to know the zeros of the denominator of (16.16). These can be found from (16.13) with the value (16.15) of cosh 6, or by noticing that cosh mQ
is
zero
when
By
(16.14),
nominator of (16.16) so
The
by
values of
>
i,
.,
OT-I.
(16.18)
To
II,
evaluate
I r it is
Theorem
and
for this
we need
60
rd
I
j- (sinh
-^
\_dd
(sinh
v
cosh m0)
-5-
cosh
cosh
md
i)tn-
-\-
dpjp^a,, sinh
sinh
L*
sinh
)
sinh
......
using (16.19),
OT
(16.19)
Then, applying
,
lr
(2.8) to (16.16),
sE^=J -
>
e*s*
( sin ^
-- r^ )(2s
&
EXAMPLE
to a
filter
2.
of m,
with open
Voltage
V=
sin tat is
applied at
L#>,
i/Cp,
Here
(16.4)
gives
cosh
And from
= +
i
LQ>
2
..
.
(16.21)
(16.11) T
r
'
l6
'
s2
The
are
+ ico,
(16.23)
^-
'
'
'
*>
'
'
m~
P
where
and
it is
= =
to
ip,,
....
be
to.
(16.34)
\* f i
- cos c &Ltl ^r
assumed
that
6l
Example
0)T
we
find
a+1
( )
(sinh 6 cosh
LCwz/33
(16.26)
J*P
And
Ir
mL *~ t~
o
-=- "S*
o> C
B zy a
==-
-2m
.
+ U sinh 0(to>)
,
Ceo sinh (m
.
fl/
_,
,.
"I
-f-
Conjugate J te
i0(ia>)
V J
(16.27)
where
period
0(zo>)
is
= uo.
.
=
1
JLCoj
:
2
,
(16.28)
and
(i)
two
where and in
possibilities
z'S,
2
),
of period 27r/co
^ o>t.
.
is
,
~
Ceo
(ii)
sin (wz
\,
sin 8
if a;
2
~ cos cos mo
r) S
.,
(16.29)
where and in
,
A=
( v Nr ) '
>
4/LC,
0(co)
2
ZTT -(-
A,
cosh- 1
(LCa>
i),
is
/ u \ (16.30) J J \
-1 1r Coj
,
As
on the other hand (16.29) oscillates. That is, if to 2 > 4/LC the term of period 27r/co dies out exponentially as we move 2 along the circuit, while if to 4/LC this term oscillates The circuit is, of course, a low pass trigonometrically. filter, and we have found both its transient and steady state
while
<
cof.
Vibrations of mechanical systems Problems on the motion of bodies of concentrated inertia, connected by shafts or springs of negligible inertia, lead
62
very readily to systems of ordinary differential equations which have to be solved with given initial conditions. In order tp make these differential equations linear, the dissipative forces are restricted to the case of resistance to motion proportional to the velocity, though in some
problems Coulomb friction (constant and opposite to the direction of motion) can be treated. Rotating systems only will be considered here, since 'they are of perhaps a little more interest than the corresponding problems in linear motion, and lead to precisely similar equations.
We
inertia of a
ment from
write throughout this section, J for the moment of wheel about its axis, 9 for its angular displacea standard position, for differentiation with
respect to time,
D9
>
.^
(l?>l)
for angular velocity, kca for resistance to motion proportional to angular velocity, for torque, and A for the stiffness of a shaft, that is the torque required to produce unit relative
angular displacement between its ends. The equation of motion of a wheel of moment of inertia J, acted on by torque T, and with resistance to motion
ka) '
is
(JD-M)o,
T.
(17.2)
Also the torque T required to produce a relative displacement 6 between the ends of a shaft of stiffness A is
T=
A0.
..
(17.3)
Using these equations we can write down differential equations for the motion of any combination of wheels
connected by shafts of known stiffness. A number of examples of the differential equations which arise in such problems is given below clearly these can be solved just as in 5 with given initial values of the o> and 9.
:
EXAMPLE i. A shaft of stiffness A has one end fixed, and on the other a wheel of moment of inertia J (Fig. 19(0)). Torque T, a given function of time, is applied to the wheel, and there is resistance ku> to its motion. If 9 is the displacement of the wheel from its equilibrium
63
due to twist of the shaft is A0, and the differential equation for the motion of the wheel is T A0. (174) *) (JD
ir
FIG. 19
(a).
T
-
FIG. 19
(&).
Equations (17.1) and (17.4) can be solved with given If preferred, initial values of 6 and co in the usual way. this gives the second order equation CD can be eliminated
;
(JD
+ 6D + A)0 - T.
(17-5)
a)
B*
EXAMPLE
stiffness
J\
Bi
FIG. 20
(6).
B2
B,
2. Three wheels are connected by shafts of and Ax as shown in Fig. 20(0), in which the
64
symbols used all have the same meanings as before, 6, #a a being displacements of the wheels from a reference position in which the shafts are unstrained. Torque T is applied to the wheel J. Here the differential equations for the motions of the
wheels
J, J15 J 2 , respectively, are
CJiD
(J^
and
a)
(17.7)
(17.8)
a> z
(17.9)
in
the usual
.
values of 6, 8^ 2> u>, o^, o 2 Clearly the problem of any number of wheels connected by shafts of given stiffnesses can be solved in the same way if all the wheels have the same moment of inertia and the same resistance to motion, and all the shafts the same stiffness, the sub16 sidiary equations form a chain of the type met hi and are treated in the same way.
;
EXAMPLE
First
it is
3.
^4
geared system.
necessary to study the type of complication introduced into the system by gearing. Suppose a gear of radius r drives a gear of radius
r a> tne i r angular velocities being wj and cu 2 in the directions of Then the arrows in Fig. 21. if #! and disa are the angular placements of the two gears
from
p*
FIG. 21.
(17-")
65
an equal, oppositely directed, reaction force P on the wheel rv Then if T 2 is the driving torque on the wheel rz and T! is the reaction torque on the wheel ^
,
and thus
(17.12)
FIG. 32
FIG. 22
(6).
form a pair of linear connections between the angular velocities and torques in the two
(17.11) and (17.12)
shafts.
More complicated
systems,
such as
differential
same way.
of Fig. 22 (a) in which torque is applied to the wheel J which drives a gear of radius rx and moment of inertia J^ through a
66
The gear Jj meshes with a gear of radius rz and moment of inertia J 2 which drives a wheel of moment of inertia J s through a shaft of stiffness Ar With T! and 2 defined as in (17.12) the equations of motion are
shaft of stiffness A.
T + A^-fl)
(J 2
(17.13)
(J 3
D+ D+
> >
2 3
= Ai(0 - + T = -A^ 8
--A(0 l
-.0)--T1
a)
2 2 ).
(17.14)
(17.15)
(17.16)
(17.12) are sufficient to give the motion for any conditions consistent * with (17.10) and (17.11).
In many cases it is possible to set up an electric circuit which corresponds f to a given mechanical vibration problem, in the sense that both lead to precisely the same systems of differential equations. This does not make the problem any easier to solve, but it gives an interesting physical insight into it, also the corresponding
problem
electrical
system may be
set up
and studied
in
the laboratory,
finally, since the theory of electric circuits is more highly developed than that of mechanical systems, it may be useful in suggesting further developments in that
and
field.
To see the
mechanical
sort of correspondence J
which
exists
between
and
circuit
problems,
compare
equations
(17.2), (17.3), (17.1), respectively, from which the differential equations for mechanical problems were built up,
differ-
* Initial conditions inconsistent with these correspond to o a pair of gears inconsistently moving forcing into mesh at t or displaced. t For a large number of such equivalent problems, cf. Olson,
The
correspondence described here is not the only one ; might have used the equation for
(CD
G)V
;
I,
and
in this
way
cf.
67'
:
and
.
.
DQ =
I,
(17.19)
for the connection between charge and current. It appears that (17.2), (17.3), a nd (17.17) to (17.19) are the (17.1). same sets of equations in terms of different quantities, the correspondence between these being set out in the
Table below,
Electrical.
in
linear motion of a
for
Linear Motion.
Moment
of inertia, J.
coefficient, k.
i /A.
Damping
Damping
Reciprocal stiffness,
Torque, T.
Reciprocal Force.
coefficient. stiffness.
correspondence we can now set up electric differential equations as the mechanical examples considered earlier. For example the circuit of Fig. 19 (b) with inductance J, resistance k, and capacity i/A, with current o>, charge 9 on the condenser, and applied voltage T, gives the same differential equation (17.4) as the mechanical problem of Fig. 19 (a). * to Similarly Fig. 20 (b) is the equivalent circuit Fig.
Using
this
circuits
(a), so that the differential equations found by applying Kirchhoff's second law to the closed circuits B 1 B, 1 AjAaBjjB!, and A 2 A 3 B 3 B 2 are (17.6) to (17.8) respectively, where 6, 6^ 6 Z are the amounts of charge which have passed through the inductances J, J 1; J 2 at time t, so that
20
AA
Leaky condensers in Fig. zo(b) correspond to resistance motion proportional to the relative velocities of the wheels Fig. 20 (a).
to in
68
0J
i
on the condensers
and i/Ax respectively. Finally we consider the equivalent of the geared system The linear connections (17.11) and (17.12), Fig. 22 (a).
/A
i.e.
<DI _i
COo,
To = _2 =r
"
-?,
'
(17.20)
become
r ^2
On
^=7> M.
vl
(*?)
which are the connections between currents and voltages in an ideal * transformer of r 2 secondary and rx primary
turns.
this
Fig. 22 (a} is Fig. 22 (b), and it is easy to check that these lead to the same differential equations.
1 8.
Servomechanisms
ideas involved f may be illustrated by considering a rotation control system in wmch a massive body of moment of inertia J has to be driven so that its displacement follows as closely as possible the prescribed displacement 9i(t) of a pointer. To do this, voltage f is fed into an amplifier-motor system which applies torque This torque ) to the body, where k is a constant. k(6t " " tends always to bring the output displacement " " into agreement with the and is input i5 displacement " " error To provide proportional to the " f " feed-back into damping for the system, it is usual to the amplifier a voltage depending on the output velocity D0 or on the error velocity D0. The simplest arrangement is to mount a generator on the output shaft which
The
is
Wave
Hazen,
"
232
218
Minorsky,
69
with negative sign. The torque supplied by the motor is 6 then k(6 t &iD0 ), and the equation of motion of
the body
is
JD 2
is
k(e t
If the motion
started at
- 8 - &,D0 = o from
Q
).
(18.1)
is
.
.
kfa^kei,
(18.2)
where
starting at
t
co
fc/J.
...
(18.4)
is
= Q,
t,
so that
~O
It follows that
feff
0(^42
~"~
+
2
cos ni
+ -(^
oj
2
^> 2
4
.
i) sin
nt\
(18.7)
where
l^ ^
6 t in this 9 It is seen from (18.7) that the error 6 case is composed of a constant lag, k]Q together with transient terms. If instead of feedback ,p0 the feedback is arranged to be proportional to the error velocity, kjDd, the equation
,
,
of motion
is
JD 2 JD
2
or
k(-
k^Dd)
(18.8)
JD 2 ^.
_
z
at zero displacement,
(p
+ k^p +
)0"=
-p
et
(18.9)
70
where w 2
= &/J.
x
If,
as
before,
8t
Q,t,
this
gives
-O
and
0=
--*-**!"*
w
sin
*.
(18.10) N
In this case the error (18.10) consists of transient terms only and there is no permanent lag as in (18.7). In practice the voltage kjl)6 or kj^d is usually not fed back directly to the, amplifier, but applied to some
circuit,
circuit
and voltage is tapped off from some part of this and fed to the amplifier. In this way rather more
complicated transforms arise. Another complication of practical importance is the presence of time lag in some part of the system. To illustrate the effect of this suppose that, in the first case considered above, the voltage 6 t 9Q k^Dd^ instead of being fed directly to the amplifier, is applied to an inductance L and resistance R in series, and that the voltage across the resistance is fed to the amplifier as before. Here there is a time delay determined by the time constant of the L, R system. The subsidiary equation for the current I in the L, R circuit, with zero initial conditions is
6t (Lp (i8.ii) R)I krffa, (i and the voltage V across the resistance R is given by
.
+
.
V=
RI.
(i8.ia)
Also the subsidiary equation for the motion of the body of moment of inertia J, with applied torque KV is
=
kV.
(18.13)
From
(i8.ii) to (18.13)
it:
follows that
where
o>
= kj],
The
error
may
(18.15)
<)
when the
7*
are known,. Hurwitz's criterion (12.3) for the roots of this equation to have negative real parts, that is for the system to be stable, requires
^>iK.
Thus
if
^
(18.17)
there is time delay in the system it is necessary for the amount of feedback to be greater than the minimum amount (18.17) or tn e system will not be stable. The way in which the roots of (18.16) and the values of
may
19. The impulsive function In dynamics the idea of a very large force acting for a very short time and such that the time integral of the force is finite is extremely useful, and the "impulsive " of dynamical systems is extensively studied. motion In circuit theory the ideas of an impulsive voltage (that is, a very great voltage applied for a very short time and such
is
finite)
or an impulsive current,
convenient to have a symbolic representation of such a function and this may be obtained as follows. Consider the function </>() defined f by
=o
(19.1)
t
>e
* For an interesting account of impulsive functions and their applications see van der Pol, jfourn. Inst. Elect. Eng., 81 (1937), 381. t Many other functions have the same properties, e.g. the
except that the latter has its maximum at t ~ o. to choose a function symmetrical about t = o, (c), but there is some advantage in the present applications in being able to think of, say, an applied voltage o and becomes very large immediately after which is zero at t
Fig.
23
(c),
It is
more usual
such as Fig. 23
this instant.
72
where
This is very large in a very small e is very small. region to the right of the origin (Fig. 23 (#)) and zero
O 2
FIG. 23
(a).
FIG. 23
(5).
FIG. 23
is,
(c),
elsewhere, and,
no matter how
f
small e
J-oo
Also,
if
<j)(f)dt
i.
(19-2)
a f(t) is
TOO
continuous function
/-<H a+e j f
J-oo
eja
= /(*) + O(e),
where O(e) is written for a term involving and thus which tends to zero as e -^- o.
(I9-3)
e as a factor,
now define the " impulsive " function 8(i) to possess the properties of (j>(t) as e ^ o that is
We
S(t)
= o for < o,
and for
>
o,
a"|
/'
poo
J-OO
fco
8(t)dt
=i,
(1.9.5)
f(t}^(t
J
if /(i) is
oo
a)dt=-f(a\
(19.6)
a continuous function.
73
from
(19.6) that
i,
o
.
.
.
(19.7)
that
the Laplace transform of the 8 function is unity. may now regard a blow of impulse P as a force an impulsive voltage whose time integral is E, ; PS() and an impulsive current which transfers as a voltage ES(i) as a current QS(), and proceed to solve problems charge the validity of the in the usual way them involving solutions so obtained is discussed later.
is,
We
EXAMPLE
oj
2
is set in
position
attached to a spring of stiffness A mass o from rest in its equilibrium motion at t by a blow of impulse P. The equation of
i.
motion
is
is
Thus
x
2.
(P/Mco) sin
tat.
(19.8)
impulsive voltage ES(i) is applied at o to an L, R, C circuit with zero initial conditions. t Here the subsidiary equation (8.6) is, using (19.7), and the
EXAMPLE
An
notation (9.3),
_
2
Ej>
Thus,
in the case n
I
> E =_ e
Ljfi
o,
-A*<{
w cos nt
i
ju,
sin nt}.
(19.9)
The
solutions of
Examples
obtained by the ordinary Laplace transformation procedure, using (19.7) for the transform of S(), and, since any use of the 8 function must be regarded as symbolical only,*
* The S function was denned as the limit of some function as -> o. Any mathematical operations on this function, such as differentiation or integration, would involve the interchange of order of two limiting processes and be very difficult to justify.
e
74
we must inquire whether these solutions are mathematically The simplest way to do this is to verify that justifiable.
they do in fact satisfy the differential equations and initial conditions of their problems. Consider the solution (19.8) from this point of view: it satisfies the differential
equation
for the
in
o also as t -+ o for t motion of the mass Now P/M is the velocity (19.8), x -> o, and Dx -> P/M.
;
MD # + Mco * = M
2
o,
*>
o,
(19.10)
>
with which, according to the theory of impulsive motion in dynamics, the mass would be started by a blow of impulse P. Thus the solution obtained by the formal Laplace transformation procedure applied to Example i does in fact giye the complete solution of the dynamical problem correctly, and in addition it avoids the necessity of evaluating the initial velocity due to the blow. It can
easily
be
same
is
motion
by blows.
Precisely the
same
as
is
t
Example
2.
Here,
(19.9) tends to E/L, and we need an electrical analogue of the theory of impulsive motion in dynamics to show that an impulsive voltage ES(), applied to an L, R, C circuit in which no current is flowing, will instantaneously produce a current of this magnitude. This can be deduced from the differential equation for current and charge in the circuit, namely
(LD
as
+ R)I +
:
= E8().
we
(I9-")
Proceeding dynamical theory, integrate this over a very small time, T the integral of S(i) over this time is i that of the finite quantities I and Q is very small ; and that of DI is the difference between the final
in
;
*
loc.
By an
cit.,
Jaeger, Phil,
all
Mag. (7), 36 (1945), 644, gives a the types of problem arising in this
and
20.
75
I.
and the
initial
Thus
LI,
= E,
(19.12)
and the theory of impulsive motion gives the current I/ E/L, to which the solution obtained by the Laplace transformation procedure was found to tend as t -> o.
is easy to verify that (19.9) satisfies the differential o, so it does in fact give equation of the problem for t the correct solution of the physical problem. It can be proved in general that for a complicated circuit with various impulsive applied voltages, the solutions obtained by the Laplace transformation procedure, treating B(t) as having (19.7) for transform, do satisfy the equations of circuit theory, and, as t > o, the currents tend to the values which would be found from the equations of impulsive motion.
It
>
20. Justification
of
the
solutions
of
electrical
network
problems
It was remarked in 7 that, for a system of linear differential equations, a general verification could be given that the solutions obtained by the Laplace transformation
satisfied the differential equations and initial conditions, provided a certain determinant did not vanish. of circuit theory often give rise to a system of differential equations to which this verification applies immediately. In general, however, they give a combina-
method
The problems
tion of differential or algebraic equations (10.1) and algeit may happen that the specified braic equations (10.6)
;
initial
some of these
values of the currents and charges do not satisfy equations, and in that case there will be an It is in impulsive readjustment of these initial values. fact true * in all cases that the solutions obtained by the 10, giving Laplace transformation method, used as in each inductance its own initial current, and each condenser its own initial charge, in the subsidiary equations, do in fact give solutions of the equations of the problems
* Jaeger, loc.
cit.
76
for
t
>
o,
which
as
which
exist
impulsive readjustment of initial The general theory will conditions, if this takes place. not be given here, but only examples illustrating the different cases which can arise.
i. Steady current E/R is flowing in of Fig. 24 with the switch S closed. At t the switch is opened.
EXAMPLE
circuit
the o
innf^L,
i,
FIG. 24.
and
Since the initial current is E/R in L, and zero in L 2 the subsidiary equations are
,
(20.1)
Solving we find
Rp
R{(LL!
+ LL +
2
H- L.,)}
(20-2)
And
thus
As
E(LLX
-|-
LL
2)
77
and, since this does not equal E/R, it appears that there has been an impulsive readjustment of current between the three inductances just after the switch was opened.
To investigate what happens in this case, and to verify this result, consider the equations of the system for t o,
>
namely
(LD
+ R)I
L 2 DI 2
4-
L DI - E, - LjDIi = I + l!=I.
2
a
(20.5)
o,
(20.6)
(20.7)
values E/R, o, o, which the currents I, I I 2 had 1} before the switch S was opened, do not satisfy (20.7). Suppose that these change to I<), 1^) and I 2() in a very short time r following the opening of S at } o, then,
The
o to integrating (20.5) and (20.6) from t =r,'and neglecting the integrals of the finite quantities E, I, l lt I 2 over this very small time, we find
L{I<>
(E/R)}
L 2 I 2W
+ L I C> = o, - L^W = o.
2
(20.8) (20.9)
Also
I<), I^o)
and
I 2 () satisfy (20.7),
that
is
m
.
KO)
^(0)
yo).
20 10 )
.
Solving (20.8) to (20.10) gives the new values of the currents, and, in particular, I<) is found to have the value If the voltage drops across the inductances were (20.4). evaluated in the usual way, they would be found to contain S functions, that is, the finite changes of current may be
regarded as being produced by impulsive voltages. EXAMPLE 2. The condenser C in the circuit of Fig. 25 is un(p. 78) is charged to potential E, and the condenser o the switch S is closed. charged, when at t
are
(20.11)
78
we
find
_
R(C
-f'
_ _ ECQ
Thus
EC 2
-
I=-
ECCj
R(C
(30.12
FIG. 25.
-f C^)]8(t) in (20.12) represents impulsive current which instantaneously exchanges charge ECC1 /(C -(- Qi) between the two condensers the second term represents the subsequent discharge of the condensers through the resistance. Whenever a network includes a closed circuit containing condensers only, there will be an instantaneous exchange of charge in, this way between the condensers if these are incompatibly charged. If we had used the subsidiary equations for charge instead of those for current, this finite change of charge would have been found (in the same way as the finite change of current in Example i) but as we have worked with current, the impulsive current which transfers this amount of charge has appeared. The result (20.12) can be verified by writing down the differential equations for the charges on the condensers and proceeding as in
[
The term
ECC^C
an
>
Example
i.
79
Steady current
transformer,
Fig.
E/^
26,
1
current
is
M
FIG. 26.
o the switch S is opened. Writing Ij and I 2 for the primary and secondary currents, the subsidiary equation for the secondary is
zero,
at
it
when
Also
Mpl x + L = o,
(20.13)
(Lap
for
t
+ R )I = ME/Rj > o, so = o.
2 2
(20.13)
Thus- from
Ia
K
EXAMPLE 4. L]L 2
(20.14)
Here as t -> o, I 2 -> ME/RjL^ There was a sudden change in the currents when the switch was opened which could have been studied as in Example i.
case of
The problem
2
of
n, Example
.
i,
for the
(perfect coupling).
Here
.
In
ME
:
(11.2)
'
becomes
,
and thus
It is
ME
.
(20.16)
known from general theory that LjL 2 > 2 is of interest for two limiting case L]L 2 =
firstly
it
secondly,
in 1 2
2 the reasons
;
in
practice
if
denominator
and
is reduced by one, here a quadratic has become linear similarly a cubic becomes a quadratic, and the algebra is greatly simplified.
8o
significance of perfect coupling from the theoretical * is best seen by considering the differential point of view these are equations of the circuit of Fig. 26
:
(LJ>
If LjL.2 the second by
+ ROIi + MDI = E
2
(
7)
=M
2
,
multiplying the
subtracting,
a a
first
M, and
we
of (20.17) obtain
.
by
L2
L a Rj!
Thus
in this case
- MR I = LaE.
(20.18)
algebraic equation by elimination from the given differential equations, and if the given initial currents do not satisfy (20.18) they will change suddenly to hew values which can be calculated
as in
we can form an
Example i. The examples above illustrate the types of problem in which there may be sudden changes of charges or currents on closing switches, and the way in which these could have been found has been sketched briefly. A general discussion along these lines can be given, which shows
that the Laplace transformation procedure, applied in the usual way, does in fact always give the correct solution of
EXAMPLES ON CHAPTER
1.
II
o to Alternating voltage E sin (nt -f- e) is applied at t and capacity C in series. n z inductance i/LC. If the initial current and charge are zero, show, using (i.u) and (1,12), that the current at any time is
e)
sin
sin nt}.
2. Constant current (E/R) is supplied for f > o to a parallel combination of inductive resistance L, R and capacity C with zero initial conditions. Show that the voltage drop across the condenser is
E* This
is
Ear*" {cos nt
[(R/anL)
(i/RC)]
sin nt},
an example of the exceptional case referred to in 2 for 7, since the determinant discussed there is LjL 2 the system of equations (20.17).
8l
and n
Deduce
the solution of
Ex.
3.
n, Ex.
*
2,
show
on the
condenser
is
(CE/aVs){(i
VS)
cos
(i
VS)
cos wi[i(3
4. Regarding an imperfect condenser as capacity C in series with resistance R1} the whole being shunted by conductance G, show that the subsidiary equation for voltage V applied to such a condenser in series with an inductive resistance L, R is
_-r T ? ~ V+ LI
,
In the bridge of Fig. 9, z\ consists of resistance Rj. and capacity C x in parallel, z z consists of resistance R s and capacity C 3 in parallel, and st a and #4 consist of resistances R a and R 4
5.
,
respectively. voltage if
Show
that there
is
CI/CB
R4/Ra
a 6. In the circuit of Fig. 6, LaL 2 steady-state conditions prevail due to alternating voltage of peak value E and frequency W/ZTT in the primary. At the instant of zero secondary current the condenser is instantaneously discharged, show that at time t after this instant the secondary current is
,
= RB/RI> = M and
R Ca
a
In the resistance-capacity coupled amplifier of Figs. 13, = o, show that if the grid voltage is changed by 14, with # E at t o, the change in grid voltage of the next valve is
7.
a At t o, hangs by a spring of stiffness is at rest in its equilibrium position, the other end Show that of the spring is given a downwards motion a sin a>t. the displacement of the mass from its equilibrium position at
.
V A
*=
time
t is
wa
nn Uft
-i
n4
;[w sin nt
is
n sin
at
t
eat].
o to a condenser of 9. Constant|voltage applied capacity Cj in series with a leaky condenser of capacity C and Show that, if the initial charges are leakage conductance G.
zero, the current
is
CHAPTER
III
21. Heaviside's shifting theorem THEOREM VI. // y(p) is the is the then e~ ap
>
o,
y(p]
transform of transform of
a),
.
y(t),
and
y(t
where *
H(t)
= o, < o\ = t> of
t
i,
a)H(*
(ai.i)
(21.2)
y(t-aV H(t-a)
FIG. 27
(a).
a)
is
zero
if
<a
and unity
if t
a, a, and for a) is zero for t o, shifted to the graph exactly that of y(t) for f hence the right through a distance a, as in Fig. 27 (tf) name " shifting theorem ". * This is Heaviside's unit function; the notation (21.2) will be found very useful in many problems. It has not appeared earlier in the present treatment, since problems have been stated " hitherto in the, form of, say, unit voltage applied to a circuit " " at t = o ", instead of it voltage H(f) applied to a circuit must be noticed that these statements are identical only for t > o ; for t o the first does not specify the voltage but the second makes it zero. 82
a)H(i
< >
>
fl,
the
>
<
83
To
poo
we have only
foo
to notice that
(co
JO
erv*y(t
Using
a large
this
theorem
we
number of
useful step
O.
FIG. 37
x-,-
EXAMPLE
i.e.
i.
XO = E{i
= E, o < < = o, >a - H(* - a)}, cf. Fig. 27 (6), = !(* -*-*)
y(f)
t
ff\
'
I J
J"
(21-3)
7W
FIG. 27
(c).
EXAMPLE
2.
A half sine
a>t,
wave, Fig. 27
7T/W
/'
(c),
y
i.e.
= E sin = 0, y = E sin
Eo>
jy
o
*
>
wt
+E
sin cj(
84
y
i.e.
3;
= E, rT < < (r + i)T \ = - E, (r + i)T < < (r + a)TJ = E{H() - 2H(* - T) + 2H(i - 2
t *
3.
"
"
square
wave, Fig. 27
3T
FIG. 27 (d).
y=
p =tanh
T
FIG. 27
2T
(e).
EXAMPLE
4.
A
rT),
"
saw-tooth
"
wave, Fig.
i)T,
a
y=P
k(t
rT
< < (r +
t
EXAMPLE
f(t
f(P)
if r + rT) = fooe~#f(t)dt =
85
that
= /(),
is
any integer.
e-Ptf(t)dt
pT
Jo
Jo
+ JT e
f2T
Jo
f *-**/(*>&
+ <r*TJo eI*
(21-5)
and
be deduced from
this result.
EXAMPLE
3 to 5
all involved hyperbolic functions. A technique of great importance both here and in connection is that for differential with partial equations, finding the function which has a given transform of this type. To do this we expand the transform by the Binomial Theorem for example if in a series of negative exponentials
have
is
which
Ei _
_
E
.2V
1
e~ 2^ T
~P''
X1
~r e
_E
~^'
Then, by Theorem VI, Ei 2E(* y(t)
.,
(21.8)
the function
y
7.
is
shown
27 (/)
(p. 86).
EXAMPLE
to an L, R,
t
= T the battery is
circuit
86
EH(i
T), and
Then, as
in (9.2), the
-
is
and,
I
by Theorem VI,
-X*TlLj
= -{e-# sin nt 2
T)
sin n(t
- T)},
(21.9)
>
o.
2T
The
solution for
4T
FIG. 27 (/).
applied
voltages
such
as
those of
Examples 3 and 4 is found in the same way (cf. Example 5 at the end of this Chapter). An alternative form for the solution in these cases can be obtained by using the expansion formula of
33.
8. To find the deflection y at any point of a uniform beam of length /, which is clamped horio and x l, zontally at the same level at its ends, x and carries a concentrated load at x a.
EXAMPLE
light
Using the S-function defined in 19, a concentrated load at x a may be treated as a distributed load
WS(#
6,
we have
to solve
with
= Dy =
o,
(21,10)
I.
The
&7 THEOREMS AND THEIR APPLICATIONS VI, subsidiary equation is, by (19.7) and Theorem
where
3>2
6EI
^s are
an
the conditions
= Dy =*
unknown
o,
constants to
at
be found from.
Z.
we
get finally,
6/
22. Heaviside's series expansion
EI
This is a useful method for determining the behaviour solutions of the solution for small values of the time the or obtained previously are equally suitable for all values small for are this section useful those of only the time obtained values of the time, but are much more quickly than the general solutions. been have II I and in transforms Most of the Chapters found to take the form
:
where g(p)
f(J>) is a polynomial in p degree n, we divide numerator and polynomial of lower degree. If n and in ascending of expand p by denominator (22.!) find an exof ifp by the Binomial Theorem, we
,
of
and
powers
the pression of
form
.
5+
++
. .
(22.2)
( 2-' 2 \' "^)
and thus
y
and y
= % + oj + a^ +
is
by
(i-S),
obtained as a series of
ascending
powers
oft.
88
EXAMPLE
In
9,
Example
i,
we had
i/c
TV
EXAMPLE
2.
\ ;
In the problem of
Therefore,
If
= in,
we letp ->
oo
(32.2),
we
find
and
-> o, we since, by (22-3), ! is the value of y when have proved, /or the form (22.1) o/y, the following Theorem.
THEOREM VII.
t-*0
Ify(p) is the transform ofy(t), then lim y(t) lim py(p). (22.4)
p-^-ao
true under fairly general conditions. The following theorem, of the same type as (22.4), may often be used to determine the behaviour of a function for large values of the time
:
This result
is in fact
// y(p) =f(p}jg(p\ where f(p) and g(p) are polynomials in p, the degree of f(p) being less than that of g(p), and if the roots of g(p) o are either zero or
THEOREM
VIII.
89
whose
y(p)
limy(t)
t--co
p*Q
limpy(p)
(22.5)
y(t}dt
J
= lim y(p)
p-+Q
(22.6)
The
result follows
immediately from
the formulae of
2 for expressing y(p) in partial fractions. of (22.6) has been given in (n.8).
23. Network theorems
An
application
Many of the general theorems of steady state network theory have immediate generalizations in terms of Laplace transforms. Here we mention only TheVenin's theorem because of its importance in simplifying the algebra in complicated networks. A special case of this theorem, which is ademany purposes, may be stated as follows Suppose AB and CD are two pairs of terminals of a let Z be the generalized impedance of the netnetwork
quate for
:
work looking
in at CD with AB short-circuited also let v be the transform of the open -circuit voltage drop across CD due to voltage V() applied for t o at AB with zero Then the transform of the current I initial conditions. in an impedance # connected across CD, due to voltage t o with zero initial conditions, V(;f) applied at AB for
;
>
>
As an example, suppose
of
in,
the network
is
a filter circuit
Z denned
above
is,
* This Theorern_ is frequently stated in the literature with no conditions on y(p~). Clearly it does not hold, for example,
if y(t)
sin
ait.
90
and
r
=m
is
is,
by
.
(16.11) with
.
= V sech mO
(23.3)
Thus, by
SQ
an output impedance
given by
2 sinh 9 sinh
m6
+#
cosh
tnff
This agrees with (16-7), but the calculation is simpler, since The'venin's theorem allows the result for any terminal impedance to be deduced from the solutions for open and
short-circuited terminations and these are often and in any case are easy to calculate ab initio.
known
* theorem 24. The superposition THEOREM IX. If y^p} is the transform of y^t), and the transy t (p) is the transform of y z (t}, then yi(p)y z (p)
form of
ft
n(
(24.1)
not particularly difficult, but is omitted here as it depends on the transformation of double integrals, and involves only routine pure mathematics.
is
The proof f
Theorem, and sometimes as Borel's or as Duhamel's Theorem. t C. and J., 33 Doetsch, loc. cit., p. 161. The essential steps, which require justification and some conditions on the
;
f-**yjfyi)du\
Jo
=
f
^e-
>
o.
t,
up
to
integral is taken over the quadrant u o, f In this put u v t, v t', then t can take all values and t all values up to ro, and the double integral becomes
>
"-'
Jo
91
equation
to be solved with
==
Dy
=o
when
= o.
The sub-
Clearly, any system of ordinary linear differential equations with constant coefficients and with arbitrary functions on the right hand side can be treated in the same
way.
present point of view, the chief use the following suppose we require current I at some place in a network due to voltage In at with some conditions.* zero initial point applied
or
is
the
where
!/#()> Theorem
I
IX
gives
(24-3)
plicated
Thus for any applied voltage V, which may be so comthat we cannot evaluate its transform, or which
may be we can
in
an oscillogram, given numerically, for example by a definite integral always express the current by which may be evaluated numerically. In the same way, if the current in any part of a circuit is known, the current
on
terms initial conditions there will be additional I. the right of (24.2) which can be dealt with as in Chapter
92
with zero
voltage
V applied to
we have
an L, R,
circuit
initial conditions,
as in (9.2)
and thus,
I
if
>
o,
*'){
cos
*'
/it
sin
*'}*-**'
<fc'
(24.4)
The function #() used above whose transform was the current due to a unit impulsive i/#Q>) is, by (24.2), This is an interesting i. applied voltage, 8(), V(p) transient in its own right, and the present result shows
V=
that if
it is
V, may be evaluated
simple and interesting current Ij due to unit applied voltage, (24.2) has the transform
transient
to any applied voltage, a definite integral. The other in the circuit is the
V=
I,
which, by
If
IJL
may
an integral involving
has been found as a function of t from (24.5), we express the current due to any applied voltage V as I 1} by writing (24.2) in the form
' '
(34 6)
'
Now, by Theorem
is
whose transform
is
&'
(24-7)
Then
it
follows
from
(24.6)
t'}dt' y
(24.8)
93
in the study of difficult or arbitrary funcif tions, or functions given numerically by oscillograms is a simple exponential or trigonometric function, it is to evaluate I directly by the methods used earlier
simpler
than
25. The
This
inversion theorem
is a
general theorem* which gives y(t) from y(p}> subject, of course, to conditions on y(t] or y(p}. Omitting these conditions, it states that if
y(P)
=
f J o
j
o,
(25-1)
ry+tco
y(t)=-.\
Z7TZJ
eKy(X)dX,
>
o.
y_oo
(25-2)
complex variable, 'and the integral is to be interpreted and evaluated by the theory of functions of a complex variable. If this theory, and in particular the calculus of residues, is known, (25.2) provides an important method for finding the function which has a given transform. For the problems of Chapters I and II
In
(25.2),
is
which led to ordinary linear differential equations, this gives an alternative method for finding y(f), which, however, is no shorter than those used earlier, so that for such problems there is no necessity to learn the technique of complex variable. For a thorough study of the applications of the method to partial differential equations, the use of
theorem is necessary. the point of view of the historical development of the subject the correspondence between (25.1) and (25.2) Carson -f regarded (25.1) as an is of great importance. integral equation from which the solution of the problem could be obtained. Bromwich, and to a certain extent
the
in version
From
to
* C. and Doetsch, loc. cit., Kap. 6. The result 29, 30. J., It is related is sometimes called the Fourier-Mellin Theorem. Fourier's and Mellin's inversion formulae, cf. Doetsch, loc. cit. f Electric Circuit Theory and Operational Calculus (McGraw
Hill, 1926).
94
Wagner,
-integral of
the type (25.2), with y(X) found from the usual subsidiary equation, as the solution of the problem. The Laplace transformation and its inversion theorem provides the link between these two ideas this point of view is largely
;
due to Doetsch.
26.
Fourier's
integral
theorem and
Fourier's integral
ditions
theorem* and
e imty(t)dt,
(26.1)
then
y (t)
= -L
277-J_co
e- i(otY(a>)daj.
(26.2)
The
is
that
J-OO
1X01^
be convergent.
(26.3)
dt^
r
ZTTJ _oo
|Y(o>)i
^o>.
(26.4)
For comparison with the problems previously discussed, suppose that y(f) is real, and is zero J when t < o. Then (26.1) and (26.2) become
:
a proof see Carslaw, Fourier Series and Integrals, edn. 3, the result (26.2), (26.1) follows from his on using the 119 exponential form for the cosine. See also Titchmar'sh, Theory i.i, 1.2. of Fourier Integrals, The theorem is often referred 2.1. t Titchmarsh, loc. cit., to as Rayleigh's theorem, and its use in physical applications is due to him [Phil. Mag., 27 (1889), 466]. J It is frequently stated that the Laplace transformation method contemplates only functions of this type. In the method as developed in Chapters I and II the problem was to find the solution of a differential equation for t o, which satisfied certain conditions at t o, and in fact the solutions found would also hold
For
;
>
95
then*
and
-Rfi-**Ya)Xa>=XO>
f Jo
|
(26.6)
iR^'YOXo. = o,
Jo
if *
provided
-
JXOI*
from
is
convergent.
case,
To -deduce
(26.6)
(26.2) in this
we
notice
equal to
CD) is
the conjugate
of
Y(OJ),
and
becomes
of y(t). Y(u) in (26.5) is called the Fourier transform Since (26.5) is precisely the same as the equation defining with the Laplace transform y(p) of y,(t), p replaced by
(
z'o>),
it
follows that, if
Y(w)
exists,
.
.
Y(w)=.y(-*co),
(26.9)
Thus
the degenerate form of the latter exists when p is pure the degenerate form of (25.2) if
is
y
t
o with i)>" o; for example, the solution of (D e~* for either t > o or t = o is found to be y though we are only interested in the former case. In this section and in 25 there is a difference in principle (though not in the results obtained for t > o) since a solution, la sought for example the solution of the differwhich is zero for t < o ential equation above, found from (25.2), would be y(t) ** o, = i, t = o y(t) = e"*, t > o. t < o y(t)
t<
i
when
<
o,
R is
written for
"
96
y can be taken
exists for
many
functions for which the Fourier transform does not (for example, I, e* or sin f) but, if the latter does exist, it can be found from (26.9) and previous results.
For example,*
so that
if
y(t)
e~ at sm
fit
(26.10)
_-_,
.
(26.Il)
the point of view of electric circuit theory the great importance of Fourier's integral theorem lies in its The Laplace transform has no physical interpretation. obvious physical significance it is a purely mathematical device for finding a solution this is not altogether a as it that difficulties can be formdisadvantage implies any ulated and treated mathematically, while it was on points of interpretation that the old operational method was obscure. On the other hand (26.6) expresses y(t] as a combination (that is the integral) of harmonic terms e~ imt , each with its own amplitude Y(o>)Ja), in the same way that a periodic quantity is expressed by Fourier's theorem as a sum of fundamental and harmonics, each with its
;
From
amplitude. Now suppose a voltage y(f) is applied to a circuit with zero initial conditions, and the current I in any branch of it is calculated, then, proceeding as in Chapter II we find
own
* Fourier transforms of many functions of this type, with numerical illustrations and applications, are given in Burch and
Bloemsma,
Phil.
Mag.
(6),
49 (1935), 480.
^
l(t]
the circuit.
be
j7(
(26.9),
'Then z'o>), by
Thus, by
(26-6)
-,
since e- imt ]z(
,>o.
(.6,5)
due
That is, we have an extension of the results ia>)d<a. y( of steady state alternating current theory to non-periodic applied voltages, although, for the solution of specific problems, this is inferior to the Laplace transformation methods described earlier. One application of (26.14), however, is new and of great importance. Applying Parseval's theorem (26.8) to
(26.14) gives
""
amplitude
TTJO
<""'
dot.
(26.16) V '
quantity on the left gives the energy absorbed in the branch of the circuit, alternatively it is the total response of a square law detector, which is of considerable practical interest, and (26-16) gives this as a real infinite integral without having actually to evaluate the current. For example, if voltage e~ a< is applied at t o to an inductive resistance R, L we have
The
o
03
(R
2 2
Lzo>)(a
ia>)
dw
(t0
-f
a
jB
)(w
2
(X
[continued overleaf
~a tan
~ _!
'a Jo
where
j8
= R/L.
In more complicated problems the integral (26.16) may be evaluated numerically, and approximations can be introduced, for example, with a selective circuit, \z(ioj}\ is a certain range of values of cu, and very large except in over this range. (26.16) need be evaluated only
i.
2T <
t If y(f) is i, a, 3,.. .. in the regions o T, " " t function) 3T, and so on (a staircase
< <
T< <
t
aT,
<
2. If y(t)
is
two
< <
tt
y(t)
o,
ta
< <T
t
and
this
(a
3.
If y(f)
=
|
sin
cat
|,
(full
wave
(
y(p}
4.
=
is
,^ ?
1,'
uniform
$1
beam
/.
and zero in
<x<
w (constant)
clamped horizontally at both ends. It per unit length in o < x < JZ Show that the deflection at any point is
24EI
If the square-wave voltage of 21, Ex. 3, is applied at to an L, R, circuit with zero initial current and charge, show that the current in the circuit at time t t' t where asT
f
5. = o
is
an integer and
t'
<
T,
is
z o. Deduce that using the notation (9.3) and assuming n in the steady state due to a square wave voltage, the current at ' a the is t after of time cycle beginning
99
n(T
t')
nL
6.
ze-pT cos
nT + e~W
2,
In the problem of
I ~~
n,
Ex.
show
*^ 60
of the time
_ _ E / _ nW + LV 6
'
'
'
In the circuit of Fig. 7, an unknown voltage is applied at o with zero initial currents and charges. If the voltage S at f drop across the first condenser is *;(), show that the voltage drop across the second condenser is
7.
r*
n
8.
JQ
If y(i)
t'}dt'.
(e~**
e'P),
>
o,
show
that
its
Fourier trans-
form
Y()
is
(a
u)(/J
z'o))'
Show
L/,
R from
RQ3
where
p.
q)'(q
jg
^)
a)'
R/L.
CHAPTER
IV
In the preceding chapters it has been found that problems be leading to ordinary linear differential equations could ^n solved with only the elementary apparatus of 1-3that the study of partial differential equations it will appear the of solution, it is still easy to find the transform Laplace but, since this usually involves more complicated functions of p than those considered in Chapter I, it is rather more The, Table of difficult to derive the solution from it.
Transforms can of course be enlarged, but the most powergeneral method is the application of the inversion theorem, 25, combined with the use of contour integration to evaluate the complex integral which occurs in (z$-2)>
ful
For a complete study of the subject, particularly in its theoretical aspects, a knowledge of this technique is essential but it is possible to go far without it, for example most results on finite transmission lines, and some on infinite lines, can be obtained, in some cases rigorously, and in others with the proviso that for a complete treatment the complex variable would have to be used, although the actual process of writing down the solution would be
;
precisely the
same
as that
used here.
differential equations for a uniform electric transmission line will be found in 28, and, for shortness, only these will be studied, although many of the problems
The
considered are problems in wave motion or conduction of heat in one dimension, with a slightly altered notation.
28.
The
line
transmission
We
suppose the
capacity C,
line to have resistance R, inductance L, and leakage conductance G, per unit length.
IOI
be the current in the direction of increasing x the point x of the line, and let V be the voltage
will be funcdrop across the line at this point. I and to tions of both distance x along the line and time t emphasize this they will sometimes be written I(x, f) and
;
V(x V ^"V,
ft
lj,
To
B
at
and voltage
x
at
-|-
Sx.
by the current
and
they will be I
$x and
The
A
A->.
I
H
.
LSx
T37T
"W^
""-^~
> "---T
K
FIG. 28.
*.fSoc
by the equivalent circuit of Fig. 28. Here the voltage drop from A to B over the series portion of the line is
-dtf
Bx,
it
is
-Sx + d#
neglecting
HK
of
it is
ox
S#,
thus
102
From
and
V are
(2,B\2)
it
These are a pair of simultaneous partial differential o wish to solve them for t equations for I and V. with given initial values of I and V, which we shall write t(x) and V(x) to emphasize that they are functions of x. Explicitly these initial conditions are
We
>
>-
1(#), as t
>
V(#), as
i
(
fixed x,
(28.5)
fixed x.
(38.6)
boundary conditions to be satisfied at the ends of the line; these depend on the particular problem. For example, if the line is short-circuited at * o o for x a and all t a, we must have V = o for if there is open circuit at a, the condition is I x a and all t o if the voltage is prescribed at x a, o V(#, t) must be this prescribed function of t for t
also
There are
= =
= =
>
if
the line
is
> = > x = a,
; ;
and
29.
of partial differential equations by the Laplace transformation We consider first the system of equations (28.3), o with initial con(28.4) which have to be solved for t ditions (28.5) and (28. 6), 'and with boundary conditions
solution
The
>
to
be
specified.
t
Proceeding
to oo.
precisely as in Chapter I,
4,
we multiply
III,
the equations
by e~^ and
respect to
from o
Then,
Theorem
* As
and,
iii
Chapter
strictly,
I are
implied,
103
t)dt
vWx
u=o
=*-!(*)+ pi
(39.1)
where I is now a function of x and jp, to emphasize this, it is written I(x, p).
and,
when we wish
where we
of
differentiation
have assumed that the orders of integration, a'nd with respect to the parameter x, may be
interchanged. Using these results, and the corresponding ones for V, obtain from (28.3) and (28.4) and the initial conditions,
we
the
"
"
subsidiary equations
(Lp
R)I(*. p)
= -^
+ Ll(*)
+ CVW
(29.3)
(Cp
+ G)?(*. p) = I
(29.4)
Eliminating
for?,
where
q*
= (Lp +
R)(C#>
is
+ G)
(29.6)
is,
When V
by
=~
Lp
+ R"^ +
Lp+.R'
It appears that the effect of the Laplace transformation with respect to t is to reduce a partial differential equation in x and t to an ordinary differential equation in x. This
lias to be solved with boundary conditions which are the are Laplace transforms of those originally prescribed they "best written down as they arise in specific problems. The quantity q which appears in (29.6) is, in the general
104
case, the
square root of a quadratic function of p, but in three important cases it takes a simpler form. These are " The lossless line ", R (i) o, for which
q = p]c,
(ii)
= G= where c =
(LC)-*
"
in
(29.8)
Heaviside's
"
distortionless line
which
L
and thus
(iii)
-
R_G_ ~ ~p
C
p)]c,
'
-f-
where
= (LC)~*.
== o, for
. .
(29.9)
The
ideal
submarine cable,
q
L=G
which
(29.10)
= -\/(RCp).
Another
the
diffusivity)
to
be solved with v
= v(x), when
is
o.
The
subsidiary
dv
-j9 Z
dx
p_ -v
K
i., -v(x)
,
,
.
.
'
(29.12) ' ^
and thus
of
L=G
is
= o.
The
where y
is
the displacement.
This has
to
be solved with
y
The
= y(x\
d zv
i)
^
z
= Y(),
is
-b
when
o.
subsidiary equation
2-=-)-),
T
.
(2 9 .14)
105
R=G=
is
When the subsidiary equation (29.5) with its boundary conditions has been solved,* we know the transform For problems on semi*V(x, f) of the solution V(x, t). infinite lines we rely on the Table of Transforms to find the solution from its transform. For problems on finite an extension of the lines two methods are available (i) expansion theorem (2.8), which gives the solution in the or (ii) an expansion in form of a trigonometric series terms of the solutions for the semi-infinite line, which has a physical interpretation in terms of direct and multiply reflected waves. Both these methods will be used, treating the latter first.
:
We
x
t
consider the
First, suppose o, with zero initial current and charge. the end x o to be maintained at constant voltage E for
>
problem of the
semi-infinite
line,
>
o.
___
The boundary
line are
(30.2)
(30.3)
The
(304)
(30.5)
oo.
6, solve it by the Laplace transformation could, as in method, but, for the simple problems arising here, it is just as easy to write down the particular integral and complementary function in the usual way.
(
We
106
The
and
V
(30.5) requires
gives
= E/p.
In the general case the transform involved in (30.6) is quite different from anything previously encountered, and in fact turns out, after a difficult calculation which will
not be given here, to be a complicated integral involving Bessel functions (cf. In the important special cases 35). mentioned in 29, however, V can be evaluated simply. o (30.6) becomes For the lossless line, R
where
(30.7)
(30.8)
Then
it
E, that
.
.
V:=EH[*
that
is
:
-(*/<:)],
(30.9)
is
zero
up
travelling
with velocity
at this
point
we have
.
^=
and'
p
e-(:P-t-P)/fl }
. .
(30.10)
'(30.11)
V = Eg-"/H(*
xjc)
velocity
line.
In this case the disturbance is still propagated with c, but there is attenuation as we move along the
Some problems on
will
= G ==
a
o,
be studied in
35.
Suppose,
now,
that
of applying
constant
107
cot.
at
o,
we apply
is
a sinusoidal voltage
E sin
The
only modification
we have
this
line,
(p
+ p)/c,
and
Still
V=
more
Ee-p*>/
sin to(t
\
- -W* - -)
c/ \
c /
(30.14)
generally, for
have
in place of (30.6)
V =/(>>-.
For the
distortionless line this
17
(30.15)
becomes
)
.
=^)e-(i'+p)/c
(30.16)
(30.17)
This may be interpreted by the statement that the voltage at x is zero up to time x/c, and subsequently follows that at x o with a time lag of x/c, and a reduction in
e~i> a l c .
In the next section the application to problems on found above will be studied.
finite
31,
The
line
travelling
wave
As an example of the method we consider the problem of the finite transmission line o x <l, with zero initial o is to be earthed, and current and charge. The end * o. the end x / maintained at constant voltage E for t
< =
>
IO8
The
^_
The boundary
The
V=
o,
o<x<L
/,
'
(3
conditions are
V = E, V = o,
* *
= = o,
t>o, t>o.
The
solution of (31.1)
which
satisfies (31.4)
and (31.5)
.
is
y=
Here q
is
E
m '
sinh g* sinh ql
'
(31.6)
denned by (29.6) so that V is a complicated One method of proceeding is suggested function of p. by the device of expanding in a series of negative exwas used in 21, Example 6. We write which ponentials and (31.6) in a form involving negative exponentials, the denominator by the binomial theorem. This expand
gives
E J T? _ e-e(z-*)( I
'
_ g-2a)(i
-j-
(29.8),
"^
line,
for which,
'
by
E Jr
e -V(l-x)lc
e -j>(l+as)/c
_|_
e -D(3Z-aO/o
"\
^ <
I.
^1 .8)
109
is
shown
I
in Fig. 29.
(/
x)/c, at
E up point x. The voltage then has the constant value a wave travelling from the to the time (I x)/c, at which x o would end the and from x I reflected back end From this time up to the time of arrival of a arrive. twice reflected wave, it has the value zero, and so on.
FIG. 29.
Returning to the general expression (31.7), it is seen 5 which that the terms of this are all of the form (i/p)efor the .appeared in (30.6) in the corresponding problem for this case is semi-infinite line. Thus, if the solution
,
be expressed
in
As
a second example,
we
charged to
o
is
insulated.
(29.5)
becomes,
110
?=o,
I
o,
= o, =L
(31-")
(31.12)
Also
it
'f. Lpdx
'
'
(3
,r 3 )
is
v
The solution p
cos
g-J"/c[ I
i.
_Ef =
j>V p
{l
e-pa>/C
Thu,
^j-Hfi-^E \ c)
k
.:.
Also by (31.13)
(31.15)
The most
At
this point
interesting quantity
is
the current at
= o.
(3I- 1 ?)
that
is,
III
current
is
I r,
the line
is
^..Q
y.*
...
Al^
Similarly,
if
voltage v
at
is
applied
through
the
line
ZQ
to
impedance x o
;,
writing V for the voltage at o of the line and the point x the boundary condition at * ==
IGl
"
is
that a lossless line of length / o it is charged to voltage E, and that at t o, the end discharged through a resistance R at x x I being insulated. In this case the*subsidiary equation
is
As an example, suppose
initially
(29.5) becomes
where
= =
(LC)~*.
At x
I
/,
we have
x
= o,
I
1,
(32.4)
and
at x
is
. .
R
Also
= -?..
..
(32.5)
we know from
(29.7) that
1=-^^,
and thus
(32.5)
.-
(32-6)
becomes
= 0.
112
The
V = ? + A cosh p(l-x)lc.
Substituting this in (32:7) gives
(32.8)
.R A=4
Lc
Using
sinh
pi
A pi -+A cosh-ic c
,
H-
E T
p
o.
this value of
in (32.8),
we
find
p
The
current
is
sinh (p//c)
+ Lc cosh (p//
+R
'
cosl
'
where
x V(L/C). For shortness we evaluate o. In this case, expressing (3 2 only the current at x I in negative exponentials, and writing a=(Ri
R =
Lc =
^)
we
find
E(r
.
-*
+ RO)(I
E
(a
Therefore
so that the current has a series of constant values, diminishRx -\/(L/C), (the charactering in magnitude. If R istic impedance), there is a E/aR single pulse of current for time zljc, and zero current subsequently.
113
Clearly any terminal impedances can be discussed in the same way, but since the coefficients of the later exare usually rather complicated functions of p, ponentials the method is most suitable for use when the solution is needed for fairly small values of the time.
33. The expansion formula for finite lines In chapters I and II the transforms of the solutions of problems on ordinary differential equations were usually found to be of the form
<p(p)
...
,
r^^.i)
g(p] were polynomials in p. The solution was found from (j>(p] by expressing (33.1) in partial ff>(f) or fractions, by the equivalent technique of Theorem II, for example, by the result (2.8) that if a^ . ., n are the zeros ofg(p), and these are all different, then
finite lines in this chapter, rather complicated transforms have appeared. The new feature of these is the occurrence of quotients of hyperbolic functions of q. For example, the expression
In the problems on
more
'
pcohpl/c
in (31.14). Since cosh % represented by the infinite products *
(33-3)
appeared
>
(33-4)
;nb)
*
Carslaw, Plane Trigonometry, edn.
3,
>
(33-5)
168.
may
still
C33-)
except that nov?f(p) and g(p) may have an infinite number Then of factors, instead of a finite number as in (33.1). if a lt a 2 are the zeros of g(p), and if these are all exbe can it is to that assume different, (33.2) plausible tended to this case that is, if $(p) is given by (33.6),
, .
.
'
'
(33 ' 7)
to the case in which 2 (33.7) Just as in applies only if it has, the corresponding g(p) has no repeated zeros extensions (2.19) or (2.20) must be used. The result (33.7) will be referred to as the expansion
;
formula
it
is
merely suggested
as a possible extension to the case of an infinite number of zeros, of the result (2.8) which was proved for the case of a finite number of zeros. - That is precisely the approach which was, used by Heaviside. The justification for the
use of (33.7)
method
is as follows the rigorous pure mathematical of finding </)(t) from (33.6) consists of the application of the inversion theorem (25.2), followed by contour integration and a certain amount of pure mathematical
:
discussion, and, in the case of finite lines,* the final step in procedure is equivalent to the use of (33.7). Thus it may be taken that there is a solid theoretical foundation for results on finite lines (or on wave motion, or conduction
this
of heat, in a finite region) obtained by the use of (33.7) this remark does not apply to semi-infinite lines,^ for example (33.7) must not be applied to (30.6).
J
Any
of Table
acteristic
applied voltages must be combinations of the functions I, and the lines must not be terminated by their char-
impedance.
t Or to lines terminated by their characteristic impedance in .this case the transforms encountered are of the same type as those for semi-infinite lines.
115
example of the use of (33.7), we apply it to find (j)(t) from (33.3). First the zeros of the denominator must be found. This could be done by putting * z plfc in (33.4), but, from the point of view of the subsequent
calculation,
clearly
=^=
costu
\,
is
better
We (in + ___
i
n==0>
I(
2,..,,
(33:8)
cosh
--~We
21
l\ V CJ
= cos (zn 4v
!
iW
= o.
/
To
we need
,
Thus
__(*cosh UA
i.
(33.10) ** '
(zn rf =-
i>'77
(zn smh.
\_dp\
^/Jj)=(2n+l)Mro/2Z
-+
2
i)zV
Using these
from
(33.3)
Conjugate
I
COS
jc -
cos
(2M+ L I)TTC<
,
-y
aZ
(33. iz)
* (33.4) and (33.5) can be used to show the actual factors of numerator and denominator this is often useful in cases where a factor can be cancelled between them, or where some exceptional
:
factor arises.
Il6
o of the
Using
(33 I3)
solution of the same problem, expressed in terms of The successively reflected waves, was given in (31.15). connection between this and (33.13) is of course that at
any point x, the solution (31.15) is a periodic function of t' with period 4//c, and if this function is expressed as a Fourier series, the result is (33.13). As another example of the application of (33.7) we deterfor the first problem of mine 3 1 from its transform
(31.6),
namely
,
,
'
'
(33 I4)
in
of
g-,
by
?
2
= (Lp + R)(Q> +
is,
(33-15)
The
first
point to notice
that if
we use
get
.
the infinite
gl
Vl
we
'g V
t33
. .
although V, as given in (33.14), appears to be a function of q (which is the square root of a function of p), it in fact involves only 2 and thus is a function of q p which does not involve square roots. The same result holds for all transforms which arise in on finite problems
,
It follows that
lines.
The
below
;
general case of (33.14) will be studied in Example 3 special cases are given in Examples i and 2.
117
EXAMPLE
(LC)-i
i.
The
'"
lossless
"
line,
= pjc,
.
where
V=
.
E sinh pxlc
.
p
One
sinh pi c
is
.,
-.
(33.17)
o,
p=inirc[l,
.
ni,
=
.
.
2...
(33.18)
since
sinh
fimrc
<
{I
1} V c)
sm mr = o.
Also
AI
-j-
sinh
T.
==
C/Jp=iTaa>ll n imT,
LC
n
cosh
i,
Cjj,
2,
iwtc //
|_0p\
= (}
...
(33.19)
o of the denominator is clearly exceptional since the numerator of (33.17) also vanishes for this value. The simplest way to see what happens in such a case, and to evaluate the corresponding term in the expansion formula, is to use the series for sinh z and cosh z, namely
zero
j>
The
=s
+-J+TJ +
%3
z5
-.
(33-20)
(33- 21 )
in (33.17),
(33.17)
common
. .
factor. ctor.
Using
(33.20),
+ (pWI^} +
.]
does not vanish when and, in this form, the numerator of o, and the denominator has only a single zero there.
n8
Using
(33.18),
this
result for
we
formula (33.7)
p o, and (33.19) for the zeros get finally from UJ (33.17) and the expansion F
"
F*
.-:
--
__ sn
difficulty
e nxo
__
Conjugatej
.
(33-24)
The minor
at
o arises in
many
to see precisely what happens at such a point by substituting the series (33.20), (33.21), or the products (33.4), (33.5) for
problems, and
always advisable
the
tactor.
hyperbolic functions, and cancelling any common In calculating the contribution from such a point, instead of using a term of the a type [f(ar ]e ^/g'(ary\ which occurred in (33.7) and (2.8), one of the equivalent type
(2.10),
namely
which is the first term of (33.24). Note also that if, after cancelling
as in (33.22), the denominator of has a multiple zero, a term of the type (2.20) must be used for this zero.
any
common
factors
EXAMPLE
(p/ic)*,
where
'
'
''
an exceptional value at Example both numerator and denominator vanish. Usin we find
i,
As
in
p= o
p*mhlV(PlK)'
is
which
.20
and
The
p
since
= - KM//
l(
a
,
i, 2,
.,
(33-28)
sinh
^-
db sinh imr
o.
Also
\~
.
(33.39)
-KnWIP
(33.26) and
and
v=
+f5y
3.
*w|
(33-30)
EXAMPLE
(33.15).
The
Here p
r^-(psinh?oT
\_dp
[sinh
gZ]j,
=o
=
;
sinh Z(RG)i
(33-30
Jcp=Q
other zeros of the denominator are found from the o and these are q values of q which make sinh ql zero
The
need not be considered, since a factor and denominator in q has cancelled between numerator Using the value (33.15) of q, (33.32) gives (33.16).
The
zero
q = = 9
-j-
//,
i,
.,
(33-3 2 )
(Lp
+ R)(Cp + G) +
f-
= o,
i,
.,
(33-33)
130
This
a quadratic
which gives two values of p correspondn it could also have been deduced
;
from
P
(33.16).
(R/aL)
(LC)-*,
(33-34)
(33-33)
becomes
+ 2pp + p
o-
+ ^
,
=o,
1,2
(33-35)
The
where
of
vn
=
Z
-Pivn
[(wW/Z
a2
;
(33-,3
.
6)
2
)
-a
]*,
(33-3?)
that
is (TTC[I)
>
if this
will
have
of n, and
these
Si
must be
treated separately.
As usual we need
-j-(p sinh ql
Lfy
*J P =- P +. n
= \lp-
cosh
L'dp
ql tf
J,
cos
P4 *
W!s
Using
_T
V = Esinh*(RG)* ^ -n-Ec 2
x
_
sinh
'
(33.38)
(33.31),
we have
'
,
finally
2
.
by
(33.7)
-\
-Z
p-pt
sinh/(RG)i
Conj.j
"
where
tan Sn
= p/vn
(33-39)
121
to problems
on
lines. From the results (33.13), (33.24), (33.30), (33.39) it appears that the expansion formula gives the solution as an infinite series, whose terms are products of
trigonometric terms in x and exponential or trigonometric t. These are quite different in form to those in 31, 32 in terms of multiply reflected waves. Each type of solution has its own advantages, so that it is
terms in obtained
hard
to give any general indication of which method better to use preferably both should be tried in
:
it is
any
However, the position as seen from the specific problem. results of the examples worked out above is as follows
For the lossless and distcfrtionless lines with simple (i) terminations, both methods are available, but the solution in terms of multiply reflected waves gives the actual wave form of the disturbance with a simple physical interpretaion,
(ii)
and thus is preferable. For more general lines the solution by the expansion
is
very little more complicated than that for the while the solutions in terms, of multiply waves involve much more complicated functions (not given here). For lines with terminal impedances, the solution in .(iii) terms of multiply reflected waves is useful in general only for the first few waves, that is, for not too large values of the time, while the solution obtained by the expansion formula (discussed in 34) is available for all values of the it is often difficult to determine the zeros of time,
formula
lossless
line,
reflected
though
the denominator of
34.
its
transform.
to the finite
impedances. Problems of this type are treated precisely as in 33, the only new feature being that the transforms have more to complicated denominators whose zeros may be difficult find. As a specific example we consider a lossless line of o a at t that and at x insulated I, suppose length /,
122
charged to voltage E,
is,
is
discharged
writing c
= (LC)-*,
.
<*</,
=
(34-1)
/I A >2 -9^ \34 -J
dx
If
= O,
'
'
'
'
is
the point o ot the line, the subsidiary equation' for the condenser ^
at
and
and current
_.JLi_ ? c
oP
that
*
is,
_E
p'
,
'
( ^34'S;
using (29.7),
i
dV
^L
E
,
when x
= o.
i)
(34.4) \<J~ TV
This
is
the
= o.
(34.2)
A
Thus
(L(f^
'
V=
The
1^2
Cc
cosh j>(/
sinn (pi/c)
x)/c
+ (pC
n=
P=
where the
arep
.
=
.
o,
and
(34.6)
^ocn //,
i,
.,
+ ka. cos a =
p //C /in *-C
.
o,
where
is
(34.7)
/
n\
(34.8)
to the capacity
of
these line. Approximate values of be found from the intersections of the curves
the whole
roots can
and
as
y y
in Fig. 31,
= tan a
ka.f'
shown
interpolation.
,y^
FIG. 31.
**
To find V we need
from
(34.5)
(33.7),
=
where k
is
(l/c)[k
Hh
A:
a]
cos
a,
(34.9)
given by (34.8).
Also
.
(34.-)
124
Using
and
(34.10), the
cos
35. Further discussion of semi-infinite
lines
In 30 the transforms of the solutions of problems semi-infinite lines were found to contain the factor
.
.
on
J (35- )
where
and
lines
in
q =
31,
*/{(Lp
R)(Cj> +
G)},
(35- 2)
finite
were found
thus
very important to know the functions which correspond to transforms of these types. In the case of the lossless and distortionless lines these can be found by the use of Theorem VI in all other cases, either the inversion
;
theorem, or a table of transforms, must be used. Perhaps the most useful collection of transforms is given in Campbell ' and Foster x* " Fourier Integrals for Practical Applications in the relevant part of which, Part 8, most of the results f can be read as Laplace transforms, p being our p, and g being written for our t. Many results for the general case (35.2) are given in their Nos. 860.5 onwards, and, for the submarine cable case, L o, earlier. Here we consider only the submarine cable case, which exalso to linear flow of heat, and give an corresponds tension of our table of transforms which will cover a
,
=G=
number
of problems of this type. The simplest and most important results follow from the known definite integral
f
V&*u3-<a*/4w) fa
= ?Le- a
20
le
i>,
(35-3)
also symbolique
JO
*Bell
8-584.
calcul
See
2, p.
470.
125
bz
= p,
.
and sub-
*-W^ = j-'*
(354)
a,
x are given
in
most books of
tables) (35.5)
becomes
-e
-ai>*
(35 8)
-
In Table 2 the results (354) and (35.8) are given, together with some others which are useful in more complicated transforms can be deduced from these problems. Other Theorems' I, III, IV, IX, also by differentiating or
by
to a parameter, just as (35.8) was integrating with respect deduced from (354). For example (35.9) follows from
with respect to the parameter a be derived independently from Also (35.11) may be deduced u t~*). (3S-3) by putting from (35.9) by Theorem IV. The results (35.12)^0 a rather more elaborate discussion (35.14) may be proved by
(35.10) by differentiating it may (or alternatively
of the
*
same type.
under the processes of differentiating or integrating most useful, since integral sign with respect to a parameter are as (35.9) to (35.11), they often allow a chain of transforms (such or (35.12) to (35-14)) to be deduced from a single transform. It must always be understood that the use of these processes needs justification, but this is usually readily supplied for integrals of the types considered here.
The
126
mentioned.
This
is
P
TABLE 2
Jo
y(P).
y(t).
(35-9)
erfc (Jflfi)
-^ +
e al>+W erfc
_)_
.
(35.13) (35.I4)
P(b
.
(35-15)
If
v>
i,
by P(v
Its
i)
values of the
most
F function
the integral in (35.16) exists and this gives the result (35.15).
is denoted Numerical
and
Also
r(ri)
= (n
is
so that (1.5)
127
As examples of the use of Table 2 we consider the following problems on the semi-infinite submarine cable.
= o with zero L = G = o,
t
EXAMPLE
i.
Constant voltage
initial
conditions.
V=je-*(Pl"*
where
K
it
(35.19)
(35.20)
i/RC.
Then
follows
from
(35.11) that
Using
(29.7),
T T
the current
,
I is to
be found from
.
dV = EC*
dx
(Rp)*
, .1
g-^WK)*
fac.22)
v
Thus, from
(35.10),
If
we had
becomes
in this case
(35 24)
'
(35.18),
' ' '
(35 2S)
'
EXAMPLE
voltage
is
2.
As
in
an impedance * applied to the line through for the line is 30, the subsidiary equation
The problem
of
Example
i,
= fj_V K ax*
V
o,
*>o,
The
'.
(35.26)
solution of (35.26)
= Ae-V(*M.
(35.27)
128
Also,
The
constant
is
**l
rr'
=p-~>
JVK "
That
is,
P.
Thus
If
V=
is
^
R,,, it
e-V(p/)
(35.29)
a resistance,
.
follows
from
(35.14) that
r
V = E erfc
where k
Again,
-^
-f
k-\/(i<f)
=
if
(35-30)
is
a condenser of capacity
(35.13) gives
v=
1 (35-3 )
For other values of # the expression i/feop^ R*) /n (35.29) must be expressed in partial fractions whose denominators are linear in^>*.
,
EXAMPLES ON CHAPTER IV
sistance
sin (<at o through a rea) is applied at t o, with zero initial RQ to a semi-infinite lossless line x Show that the voltage in the line is current and charge.
1.
Voltage
>
T7"D
r*/
Bfc[
(*
R ^7R
**\
*j
+ajH (i
**\ *
,
where c = (LC)~*, Rj = V(L/C). 2. Constant voltage E is applied at t of a lossless line of length /, the end x
=
/
o
If
I2Q
current and charge are zero, show that the voltage any point of the line is
~
(zn
z)l
At
I to x = a, 3. A lossless transmission line extends from x the end x a being insulated, and the end / being earthed. t o there is no current in the line, the portion o < x < a is charged to voltage E, and the portion I < x < o is uncharged. Show that the current at x == I at time t is
zn(l
a}
+
-]
(2tt
i)(Z
a)
L
where
4. c (LC)~*. Constant voltage
is
resistance
L R
,
applied at
o
to the
end x
o'f
being insulated, and there being zero Show that the current at w = o .is
aERi
= (LC)-i, Rt = (L/C)*, and a = (R + RJLp. submarine cable (L = G = o, and K= i/RC in (29.10)) The end x = o I has zero initial current and charge. is insulated, and a constant voltage E is applied at x = I, show that the voltage at any point is
where
5.
c
of length
7T
6.
A lossless
is
and
for
t
>
line of length Z has zero initial current and charge, o /. insulated at x Voltage E sin wt is applied at x o, show that the voltage at any point is
cos
[tu(Z
8E<0/C
^2.
(zn
13
tor
3nd w
is
any
not to be e 1 ual to
Z
(2
7-
INDEX
Algebraic equations behaviour of roots
:
48
HURWITZ'S
solution
of,
criterion for, 43
42
Amplifier circuits, 52
FOURIER transforms, 95
connection
with
LAPLACE
transforms, 95
FOURIER-MELLIN Theorem, 93
Functions of a complex variable,
uses
of,
93
Generalized impedance, 34
HEAVISIDE
Deflection of beams, 22, 86 Delta (8) function, 72 Difference equations, 57 HURWITZ'S criterion, 43 Differential equations ordinary linear with constant Impulsive function, 71 coefficients, 16, 91 Integral transforms, i partial, 102 simultaneous ordinary linear, Inversion theorem, 93
:
expansion theorem, 7 operational method, 20 series expansion, 87 shifting theorem, 82 unit function, 20, 28
KIRCHOFF'S Laws, 33
DUHAMEL'S Theorem, 90
Electrical circuits with
constants
^35,'
45
notations
for, vi.
theorems on,
88, 90, 93
4, 7, 14,
:
15, 82,
submarine cable,
:
104,
LAPLACE transforms
Tables
of, 3,
124
Electric transmission line
differential equations, 101 line with terminal impedances,
126
13
of the 8 function, 73
LERCH'S Theorem,
Lossless
line,
104
III, 121 series solutions, 113 travelling wave solutions, 105 types of solution, 121
61