Beruflich Dokumente
Kultur Dokumente
A seriee of nine leotures given at the Combined Research Group, Naval Research Laboratory Washington, D. C.
by
Edwin T. Ja.ynes
August, 1945
CONTENTS
01assical and Heavieide Method of Solution Laplace Transform Method
Transition from Laplace to Fourier Transforms
• Development of Fourier Integral from Fourier Series General Properties of Fourier Transforms
Transients in a Band-Pasa Amplifier
Analysis of Transmission Lines by Laplace Transforms
1 8 14 18
21, 25 21
27
LAPLACE TRjli~SFOru\rJS
Te.x·t b e ok ~
Ot.h.€I' books by Berg, Bush , Cohen~ Carson, Kurt z and OO.rcoran,
Skilling, Stephen.s~, and Churohill :may be consulted.",
Th€! olassical met.hod will be stuu.ied. first, using as an example: a sIB!'les-peaked video amplirie:r stage. Tiris problem will later 'be solved operationa.lly.. We assu.m.e a. lIuni t s.teptr cU!tr'ent applied 'to the lnters.tag€ eonpling c.ircui t , dei'in.edby:
r :. ,0., t<:'"
i ~ to ., t ~(J.
T.heo ire ult and te,rmiuDlc!gy are a,s; shown below::
i{n
'\A.t\I'J \I'VVV ~
Z~(t)
e(t.)
I
The, problem will be solved wnen we have found ,€ (tJ., The
standa.rd e'La.s sLca I procedure would be to se c up a differential equa.td on in which e, {t) ia the unknown and l {t,) 'the d.riving f'U!nC€iiI>I!,~
t:.ion ~ However ~ A in t.hf.s re.l a tively simple problem.~the drL ying rune-
tiOD is not dirrarenti8ble~ and the pTooedure described would rail~
I
Hence, even in this case we must use an artifice in order to arrive at e (t)~which consists off finding 12 (t) first, and then using the solution of the sub-problem to find e (t). The fundamental equations
of the circuit may be written as follows:
f r ~ i - i~ = C ~ (I )
oft
« l L oi 1~ - e. (t)
J. t cl.t -
In order to eliminate i" we differentiate ( 2) :
c/e J..;' . cR. i, J (. -)
- :: L t~ +d\.-= c Z-tJ..
cit dt~ c1t (;;_)
Rearranging this equation, we have for the differential equation
giving i2:
~ -' .
J. ll, R cRl~
+~-
clt~ L cit
l.l +-LC
2: C t) LC.
This may be put into a standard form by substituting
jet. = ~ If :L t: .L:
0'- L ~ -v- I_ C.
The differential equation is then
cJ.'1-,:· D( fJ. :l - _ -(C t)
dt.2. +Ja.. eft: +t L».
The solution of which is:
i,;. = C, e. - (~-+ V~~) t + C.,l. e - (V- - Va.1.-t--) t
(5 )
(t )
It is seen that the condition for oscillations
is:
The
constants Cl and C2 ci i~
=0 C:(t-
may be evaluated from the initial conditions:
,
.cr: t=O
(7)
We have then:
C r -+ C,._ t l' 0 = 0
c. (ct.-t- V 0.. ~-t.- '-) of- C~C a. - Va.. "'"-.t "1,,) = 0 from which we have:
C,=~[~...,-I]
(J? )
C;_ ~ -:: [Y<l~~' 1- I ]
The voltage e (t) may then be found from i2=
e (t) = IR i ~ t L ~ ~::, - ~ 1; [(; - V e - C" f J) t _ Cf T ~ e - C ~ - r) t: J
+ L i!. [_ c e.-Co. +rrJ r :- \ - (._ -If)t]
J- -? + v
(7)
Combining terms, we have:
e, (t) = C( i; [J - <7 (~ - t - t) e - C L + r) t: ~ ()+ T+ f) r:-(~~~)t J
where
•
This is the solution of the problem. There are a few generalizations that m~y be made about the classical method. It is found
that in all networks in which the components are linear and inde-
pendent of time, the differential equations are of the type known
as linear equations with constant coefficients. Their solution
consists of the complementary function, which is the general solution
of the homogeneous differential equation obtained by putting the
driving-force equal to zero, and a particular integral which depffrrlds
on the form of the driving-force function. These are known to the
engineer as the transient and steady-state responses respectively.
The transient part of the solution contains a sufficient number of
arbitrary constants to enable it to meet all the prescribed initial
conditions. The functions appearing in the transient are always
exponential. If these exponentials are complex they always occur
in complex conjugate pairs, so that their sum represents a real, exponentially damped oscillation. The difference between two
real exponential functions occurs often and is called a surge function
It has the form illustrated below:
-itt -t.t f(t) = e - e
The two modes for a simple LRC circuit vary with increasing losses
as shown below:
Vii th no losses, (tXS"bo the modes are at :tft...Vo' and re-
j~o r _ ivJo t
present an undamped sine wave, since t:> L "\ r: 4-
~ \.J -r e. := J. ~ w., \.
As losses increase, the modes move along the arc of a Circle, re-
presenting an exponentially decaying oscillation of lower frequency
than 00• Critical damping occurs when the modes coincide at
, and at this point there is only a single ex-
ponential term c:.. ~ Wo t
As the losses increase further, the modes
s
separate, and move along the real axis, wh6:lre at infinite loss one
". ..
approaches ( ..f2, r;;p ) and the other approaches zero.
The distinction between transient and steady-state is very
great in the classical method of solution, but it is a mathematical rather than physical distinction, which is lost in the operational
method o~ solution.
The Heaviside Operational Method
In the
Heaviside method, we obtain an operational equation from
A.. cR~ It
equation by substituting f-=.. o(t ) a ~ :dt~) ; = ( )Jt,
01-- J- 0_
) of t ) ~. Thi s transforms an
the differential
f~~Itf/
integro-differential equation into
q
an algeb~ic one.
If this is
solved for the unknown treating p as an algebraic quantity, the re-
suIt is a symbolic solution involving p and the driving function.
This can be converted into a real time function if we know how to
interpret the operational equations; we will now consider how this
is done. The fundamental type of driving function is the unit step
which is defined by:
Heaviside found, by working backwards from known solutions of simple
differential equations, that the following interpretations of
operational expressions could be made:
,
( 6)
and, in general, we have:
t n
;;;1
I
(II)
It should be clearly understood what this type of equation means; it is not valid to cancel out the 1 sign, because the left hand side of the equation is not a product, but rather represents an operation on 7. The right hand side, however, is an ordinary product, and simply means that the time function is cut off before
t equals 0, which amounts to assuming zero for all initial conditions. We see that if some function" F(?) is given, the result of operating with this function on the unit step 7 may be found
by expanding F (p) in a power series in ~ , after which each term may be interpreted according to equation (11). Heaviside called
·this process "a.Lge br-Lz Lng "; By means of it, an answer in the form
of a power series in t can be obtained for any F (p) that is analytic in; . Some of these power series are recognized as well-known functions; in this way the following identities have been established~
af 1 . t 1
- .-4.<.- a.
(J'.J it(:;' -
(J J-)
We see that a time function can also be obtained in the form
of a series of exponential terms if we expand }I' (p) into partial
±f_
fractions of the form f+~ and use the appropriate formula above.
The general rule for this process is called the Heaviside expansion
theorem. If F (P) is expressi ble as the ra tio NCr) /ot f)
of two
rational algebraic polynomials, we have:
7
NCP)
o (f) 1 -
NCo) D (0)
7
(73 )
where the f}~ are the- n roots of the equation D (If) : 0
I cr"'\' d'" c( 0(1') 1 (j IF
D K) 1S the e r Lva t Lve eX' F eva uated at :. Vk
, and
•
It is seen that the different "modes" of network response appear
as different solutions of D{ 1') ::; 0
which ensures that all COffi-
plex roots will appear in complex conjugate pairs.
If the time function operated on by F (p) is not the unit
step, it is usually easiest to express that time function in terms of a function G (P) operating on the unit step, so that the product F (p) G (p) then becomes the function to interpret. For
example,
e
(I( )
Second Lecture
The last lecture covered classical and Heaviside methods at
solving ditf.erential equations.
The reason for success of Heaviside method was that it re-
duced the differential equation to an algebraic one which could
be solved easily. This was because we put f' d~ 0-4 f: Jl_tc )<1"(. All this was formalism, and its justification depended upon the possibility of interpreting the result ot the algebraic solution,
which interpretation could usually, but not always be made.
There is, however, another method of solution which has the
same property of reducing the differ.ential equation to an algebraic one, in a very similar way. The Laplace transform is defined by:
1 [f(t)] :: I~t)e-st.lt
o
F (s)
(J)
It defines a functional transformation, and relates a time function f (t) to a function at the complex variable S which fDr
the moment bas no physical
1( I)" J "; _ s r oft
o
significance. Examples are:
I ";',s t/~ _ I
- e __
S o S
(2)
Jet) s~ I (~~d'~)
-st (3)
= t:e cit" - -
- S~
0
In general, ClIrnf:ITe He., 'I; s iJ-< f" r- n.ud;e:
= J ( ~,:) I s:
- ,~ 7 -
S "+1 - YI /
I d' -q t
~ L (e -I{t) iM1 -
- - e
5 + a... (Y)
a.~ 7
't =L (~~t) ~J.-. - ~a_t
- -
5;~~
-=1 [Ju(lLt)] d' 1
- Jo eeL t)
-- V'~+(1..;' -
-ys :+tJ. a- Notice the great similarity - there is one more power of P
than of S in each case. This is true in general, and is in fact
the relation by means of which mathematicians have been able to just-
ify the Heaviside rules for the solution of differential eQuations.
The operations of differentiation and integration are analyzed as
~~f~)
t{=-e-S(
..1? - _st-JI-
~ t( - - Se CO'\ (..
o(V:"c(F(t) v:f(f)
follows: if
r f dF(t) -sf· 1 [f (t)J ~ 7t e dt"
a
=SF(s)-fCo) (%F(t)~~)
J [I:Wt"j : 1 [t'hJ] - ; [res) + {')eO)]
(GJ
())
The initial conditions .... are introduced automatically. It is seen that the correspondence with the Heaviside method is kept
here, differentiation corresponding to multiplication of the time
function by P in the Heaviside system, and to multiplication of
the complex function by S in the Laplace method.
The reduction of the differential equation to an algebraic
10
one is then done as follows:
eli . If (
Lit+ (jZl +c i.Jt =e t)
Taking the;t_ transform of both sides, we have:
(2)
1 [L# +~i +fSiJtJ =j[e(tl] ~ £(5)
Or, since the s: transform is linear:
(1)
L 1 H + «--!iCt) + d 1 rs iolt J = E (5)
Using our formulas (6) and (7), we have:
This is now reduced to an algebraic eauation in S. Note that
the initial conditions give the correct quantities to specify, namely the initial current ib an L, and the initial charge on a C. Quantities such as the initial voltage on L are not per-
missible initial conditions, because they can be changed instantaneously. The Heaviside method automatically assumes zero initial condi tions, because the operational interpretation of / : jj ~t )'p. t: requires the (-0) as the lower limit, in order that P and liP be
exactly inverse operations, whereas the Laplace transforms have
(~) as the lower limit.
If we solve (11) for I (S), we have:
E ( s ) + L i# _ 2'-0
I(S) = cs
(I '--)
LS i-(R_ t <::+
The denominator of (12) looks very much like the standard
expression for the steady-state impedance, if we replace S by j~.
It can be shown that this is true in general; in other words:
The relation between the transforms I (3) Bnd E(S) is
identical with the relation between the A.C. vectors I and E for jw IS -re.,.73c.eJ ),)1' .s.
the steady-state case, if S 'mil .J 10 epQ 8!Ptl;1JQl...rt
This rule is of great convenience, and enables us to by-
pass the differential equation. stage unless there is seme
difficul ty with __ initial. condi tions. If' ini tiel condi tiona
are ignored, the solution assumes them to be zero.
J~
Third Lecture
We may apply this rule to the series-peaked video amplifier
circuit discussed in lecture one, as fbllows:
I (s) [( e. + L s ) c7 ] X(S) [I}:+LS]
res) - -
- _L - +LCS,;2_
O\.+CS + ;+6\cs
c.s
. ~~I(S)
The current applied was a unit step of amplitude 20, I) )
_ io
--
s ,
and;
Q.arJn!r 'i- Bar-ne.s
Consulting theAtab1es, we find pair 1.109 to be closest, but
it requires factoring the denominator of the equation:
[(5 )
(s ~. ~)
(3)
c
This requires solving the following equation for S:
.1 ~ }
S -t L S"t Lc.. ~ 0
The quadratic formula is used, The roots are then:
(I})
We then can identify:
a.. 0 -
I
-
Lc_
(5)
1- ~
8-L
-y ~~ J
YL'- - i.c.
The solution is then:
e(t) = '"0 rQo + ~o- DC... e - eXt + t{" -7 e -yt:J
c. Lo<'j tx.(o{-Y) 'Y{r-IX)
fl._ ~ _ if'
L ,).i._
d- d'6~ +IJ')
This reduces to:
.. \ I ( e.
e(t)~ ~1dL!-ta ~
.f<.
( ) ,.. G,- -l-rY) t:
- (fCRc - I e
This
may be rearranged in the more symmetrical
_ 10 i \" 1- J_ (;;._ - ~ - J L(f) e - G~ +~) t
- \j" 0 L 'I d'-I' ~
e. tt)
J
form!
which agrees with the result arrived at in the classical method
in lecture one.
)3
(7)
Fourth Lecture
Correspondence between Laplace and Fourier transforms may
be seen by writing:
F(s) = f;(t)e-St'oIt
o
,
,
fC'f!OO
_!_ st-
f( t) = J 7Ti Fer) e of.r
<:. - ! 'iP
then put
5 -
FCS) = F(/w) :. l:Ct)e - i.vt.Jr
o
f Ct') -::
(Jc'tj~ . t f~' t
. j W • ~ ! ..1W .. n . ,
d TTl f(_1 w) e of C'J vJ / =:Jrr F (7 w) ~ 0{ ~
C-J~ - Qo:>
last formulae are called Fourler transforms. The ex-
The
tension of the lower limit to0~~n the direct transform is made
eas l Lv if fCt)::: 0) t <. 0
•
This is the main difference between
the Fourier and Laplace transforms, and we will investigate
further to see just how the complex inversion formula leads to
time functions vanishing before
t~ 0
•
As a tool useful for later work, let us derive a preliminary result. Consider the function
as ~ increases without limit:
/
-t:~
'The central peak becomes higher and sharper, and the It cycles t1
IS
crowd together as qJ increases. However, the area under the curve
remains, finite and independent of W, since:
f()lO ~ uJ-r
oe t -:: 7T t'
-boO
Now if this function is multiplied by a function , (~) and in-
tegra ted, the rapidly alternating p'l us and minus va l ues of (sin (a)t.)/t
(9)
will cancel out contributions except in the neighborhood of t equals
0, where one infinite loop is not cancelled out by another one.
This means that it is only values of f (t) near to t equals 0 that
can contribute appreciably to the integral. In the limit, the integral is given simply by the value of f (0) multiplied by the area under the oscillating curve, which isl1:
IfOG( 5/'''' (wt) .
t) 1= d.. t:
-vo
:: 7T f{())
(S)
Generalizing slightly by shifting the origin of the oscillating
curve, we have a formula:
L j_J.O"'Ht) ~w(t-(.{) aft' :: feLt)
W_'"" 17 _~ (t-I{)
O--VJ~:cl ~ t~c.{ : 500
Now we return to the Laplace formulae:
F( $)
(~~~t~
~~/~) (7)
f (t) : ,)11j
jC>i'" s t:
F(J)e 0/1
C-j....J
If we sUbstitute (7) into {S1, we have:
F ( t) = 3:-.. A-j 1:':;.i 'iF ~) est( 011{
c.-,fw 0
The order of integration may be reversed, giving:
(7)
f(t) ~
Evaluating the inner
Jct;'w .
• e s(t-<1)J~ _
C.-/~
integral, we have: s Ct - Ion / C f-jW
e _ I
(t-I.() . -Ct-l.{)
c - .11.<,)
c(t-t.i) e
(0)
'ive then have:
This is seen to be the same as (6a) except for the limits of integration. From the discussion preceding (6a) it is clear that negative values of t must lead to the value f (t) equals 0, since
contri butions to the integral arise only when t and I.{ n cross over".
It is seen that the extra "conve rg i ng power" of the Lap.l.ac e
_ t:r t
transform due to the factor e is obtained at the expense of
cutting off all time functions before t equals O. If we wish to keep f (t) for all values of time, we cannot use the term e - .. t but must put S equals j~, as in equations (2) and (3), which are
standard Fourier transforms. In general, whatever lower limit was
taken in the integral defining the direct transformation is the
time before which the inversion integral gives' (t) equals O.
vVe may also see the above result from the theory of residues.
17
The Cauchy residue theorem is:
I J Fez.)
G\.es (a..) E: )"j 2. -a... of2. =
(J:J)
where the integral encloses the point Z = a._ , and f r z ) has no
pole at z.. ~ o;
•
The loop integral around any region is then
equal to J 77 j • E residues of all poles enclosed in the region.
for t >0 we can define a loop enclosing all poles of the
function F (3) as below:
/
(
/
I
As the integral around the semi-
circle vanishes for rR~ ~ , the
integral along the straight line
is equal to ;)TTi AE~ • For t: c:::: 0
st
, the factor e be-
y \ \
\
\ \ -,
~,-
--
•
comes divergent on the semicircle,
and we have to complete the loop
by a semicircle to the right of the straight line path. Since the new
loop encloses no poles, we have
F (t) = 0
Fifth Lecture
Development of Fourier Integral from Fourier Series
The usual form of a Fourier series is:
n::)
where
,
'f 7T
.t., = 1J f( t) ~ n w" t
-'Tr
J(w"t)
( }S)
This can be put into the complex form by noting that the sine and
cosine terms of any frequency may be combined as follows:
where
c, =j(a..'t-1-fL'1) (-11 : ~ (C( h t /1. '1 )
Co :- l: n if """"0
- c *
11
We then have:
fit) = t c,» i"uo, t
n=-V'
L f 11 ,g - i niJo t d. ( W" t)
c , :: J.7T -71" f(t) '- L.
The similarity between these and the Fourier transform
formulae is evident. VI/e will not go through the limit process
because it is too difficult, but it can be seen that a change of
variable results in the formula
jJto
Wo
_ - ret)
- .,. iT
7J'
w#
If now we let w~ ~ 0 , the limits on variation of t increase without
limit, and we become interested in much larger values ofh, such that the product Y) iN" remains finite, and equal to ttJ. Then we
w Wo
put C~ :: ;;-1T" F(l1 wo) = ~" F (0) , and we have:
JC,rO • t
- J W".
= f(C) ~ " oft"
-0"
The olher formula becomes:
f Ct)
11 = _ q'"
This suggests the definition of an integral, which becomes,
in the limit of W .. -41 dt..cJ ~ 0 :
so that both of the Fourier integral formulae can be deduced from
the Fourier series formulae.
J' f: .""" c:h·,c ,)
The F~,ajQal interpretation of a Fourier series is that of ro-
tating vectors of various frequencies:
as wp ~ 0 , this merges into a oontinuous curve, the projection of the end of which is the voltage. '~~~'f~ II
Vr.:.' , j
( 3 )
In general, Lapl~ce transforrruwill solve definite problems with known initial conditions more easily, while the Fourier transform is best for general arguments about phase and amplitudes, since they are directly measurable.
(J/
Sixth Lecture
There are certain general conclusions that can be d r awn from
the Fourier analysis that are useful in a n1..U'Uber of practical
cases.
For instance, if (('0
and
f (t.0) are transforms of each _/wt~
F (w) e a re also trans-
other,
Vie find tha t f( t- tel)
and
forms, so that an addition of a linear phase term corresponds to
8ltlIi@;8 •
0ir2ilarly,
second for every cycle / :J. jvJo r
f{t)<! and f(w~vJc)
sJ(),<: of tn~ I'hasfl. Ct.(rlle. cycle'!-Moise of "11 Si- el .. ,c.
are transforms.
,HYlf}itllde - n?,uju.)~f!d a c a rr i er of frequency Wo, ~ted uluis04 A
a time delay of one
The time function represents
by f(t)
, end the F(w~uJo) term shows that the spectrum is simply
that of f(t) but centered around til" instead of z.e ro .
Using these theorems, we can find the effect of suddenly im-
pressing a
Fr<,~el'l<:.I~S sine 'N[,1 .. ve of freqnency tiJo on 8. filter which ne.ssesAfrorrl
fore L [ o w"t]
'j'he j);;pressed ~ is renresented as IRI! t< (t) e
Wl to r42.
The t rans r orn of t h e
F (w) :;
HCW):: 0) j'wl L W,
-= I) UJ,"::;'!wL. t,.}.;.......
- 0) (wI :> vJJ.__
(l~o 38 SLLCt need be c one id e r ed , ,1S I t "culd'·~.eJcj_y 19ad te;:.~ t i me
The response is then:
/ r VO jwt- I
P, U) = ~ )71' J_ :(I.V) HCw) e. ~ (..J
J-W' j'..Jt I f v.1J_ e: iw t
_ to _J_ e clw + CR.e - --
- \J'.Q. dii (w -Wei) d1Tj (0 -W",)
-L.).,L wJ
The integral over negative frequencies may be neglected if ~
is not too far from the positive pass bandt so we have:
1 W j'wt
- ~ ;;j ... (wB_we)
WI
E (t)
olw
(5)
Definitions:
l~dt(
C 7.- ex) - - «
IX. /
(t)
The cos. integral may be f'~Jit into four integrals:
(a) .
(¥)
(7)
SEVENTH LECTURE
If xl and x2 are positive, xl :rl XI~' x2 : JX2~' and integrals (I) and (3) vanish, leaving:
sx, ~3. 01" " Ci (h) - C7(>,,)
)..,
( 8)
If xl < 0 < X2 , we ba ve
SfXd
- ~ c/t(
- l{
"XI
(9 )
€--»o
(IO)
the 1st and 3rd integrals cancel, sincecoa'u,) is an odd u
fUnction. We are left vdth-
r ~ ;:-i.I. d. « ':: 1 ~;;, z: IJ t= /J ( •
J _... -::- ~ ~ - -<A(- G) - ~ -1): J 7T ( 11 )
-€ -e:
". fX'-~~d<{ =/7f + c;(x,,-) -CrC)(,) 'X,
If xl < Xz <: 0 J we na ve
J Xa. ~ oft( : j7T -(i(1 XII) - j11 + CiC! xt./)
. I,..(
The ~~tegral ~ Cj(IX~/) -Ci(;X,/)
J (W,1- w.) 1: ~!.(
~ of l{ - 5 i (c.,J.J. - 00) t - S~' (W, - i.AJ,,) t
(...;, - w.,.)t
the total response is then:
1;.(0 , k ~J:1o ([ji( ,,1,- uJ.)t - SiCw.- w.)t] +i[c,·Cw.-"'-J t - Ci( w, - kJ.) t] }
(12)
(13)
(14)
-1-
23
if ~ is outside the pass band. If Il1o is in the band, "We have:
Paradox: When the impressed frequency is not in the passband, the response ultimately dies out. Then if the impressed frequency is suddenly cut off, the principle of superposition requires that another response of reversed Sign shall tak.e place. But the net-
work has physically no driving force at that time, so this response must be a free transient, with modes characteristic of the network alone. How is it then that ~ can appear in the solutionf ..
--
l'a.lia I _sa.,. low does the network know what frequency to OBC illa te at when the drive is removed? How can the network ~remember" the
frequency Wo?
t
') --
1£1 GHTH LE CTlJRE
Ooncluding Remarks on Feurier Integrals
Oertain general features of Fourier integrals can often
be used to get qualitative answers from physical reasoning without
any e-quations. Thus, discontinuities in either f{t) or F(W) produce wiggles in the other function. This can be seen from the
"Snake in the Oomplex p Lane" interpretation of the Fourier integral. (p./ if) If the snake is cut off at a certain frequency, the severed end is
rotating at that frequency, and wiggles of that frequency appear
in the time function. This fact may be used in untenna design
where it is desired to eliminate side lobes. In a broadside array, the angular radiation pattern is the Fourier transform of the ampli-
tude distribution pattern across the array. Since side lobes
represent wiggles on the g pattern, it is seen that the way to avoid side lobes is to make the amplitude X distribution falloff
as smoothly as possible, the smoo t.hea't way being a Gauss error
curve. The Q pattern is then another Gauss error curve. w ..... oo
The way the frequenoy function falls off as ~may be
found by inspection of the funotion f{t} and the following rules:
-1-
·--.-~.-----~--- .-.-.- .--.--------~- -----~~-I~-------~·~-·--~----~----·---·--·--~--~~-~---~-----
!
i
If F(t) has a! 'I F(OO) goes down as:
.. -.~~------.--~~~-----~--------~-___i---~-----.---.- .... -- ---~------------------.--
finite discontinuity 11m = 6 db/octave
J£
discontinuity in <It 1/.,2= 12 db/octave
.,( 1-f
" " dt1- i
tl.'rl,f 'i 1/M3= 18 db/octave
" - l/,,{n+l)= 6(n+l) db/ootave
" d. t~ No discontinuity in any
derivative
infinite discontinuity (discont. in~dt)
~-------;--------~------------
F (.) _,. cons t ,
infinite doublet diso.
.;.-1 .,.
7~ ~ason for this is in the varying quality of cancellation in the integral defining F(CII) J due to the smoothness of l(t). The smoother this function, the more perfect the cancellation ~ixkx
__ po iF -1 w t
will be between adjacent half-cycles of the kernel ~j l!!it e J •
-2-
Lulit ce
Analysis of Transmission Lines by the IRIlu!8 Transform
it Z!f]S1L (;:, t)
1(~~;;)
Fundamental equations are:
I,):.V()C,t) , i = 7:('1-, t)
( 1 )
taking the Laplace transforms of these equations:
(vex, s) := i;("~) < _st<£t-, .i t:«, S) = 1 ~:"'d) e-st~ 0
we have:
"dV
-;
dX
~.:r (
- ': - & V 4- C 5 V - C v .. )
JX
( (2)
c,. + c. s) V r C Va (X)
in most practical cases we have 10= 0, vo = 0, so we may set up a differential equation for V or I as follows:
- yrt.I
where
1~~(~+L5)CG-TCS) = L C:[(S+f')~- ~.;1J
( 4)
,
The solution of (3) is obviously
V -_ C Y7< - r x
I e -I- C .... t::
::r =: 'IZd [ C, "(; 1')( _ C.;. e -17< J
( 5)
-3-
;;"7
where
if we have finite initial valuesj we may put
(6 )
and we then add to the solution for V the terms:
(7)
Equations (5) are recognized as traveling waves, and their similarity with the well-know4,steady-state equations suggests that we can simply use the steady-state equations in all mani-
pulations, and then take the inverse transform of the result to get our transient solution.
NINTH LECTURE
Example: Distortion of a Unit step After Traveling thru a Length of Trang.mission Line.
Initial voltage = -v(o,t) .. u(t) voltage at point x = v(Xlt) Transform of initial voltage III -¥iil, '$)
1
, k- V ( 0) s) ~ S s
r =
where
,_L
YLC
e::1(f+~)
attenuation constant
a: : 1. C!. - .Q_)
;;- L C
distortion constant
- ~
S
Pair 863.1 of C.F. is e - '#- V(s +r)(Hf)
-~[s+~{_rrf)]
-e
(;; 4. t)
-1-
so the pair becomes:
the quantity e-(s+"ho has the inverse transform e-tt6q_/{t_tt,) _ P't6
,So aewe have:
= to with area
a "unit impulse" at t
e
PC
I, [(f" -..Jt~~t~;..] + e- f'&o !{_I (t- -to)
lis which denotes
integration with respect· to t.
Therefore, we have:
it _tt
veX,e-) =!2 y'!:.._ X,[cTVt ....... t;1.]Jt + e-ft't/(t-to)
v, tL.._ toL
to
-f't-c.
This is a unit step of height e at t = to, plus
distortion terms due to the integral. The modified Bessel
Function I, bears somewhat the same relation to an ordinary Bessel Function In(Z) as hyperbolic functions do to trigonometric
ones. Thus:
:r ." L 7.-) • - 11 J"~(2.i) (i= h)
::: Z
which corresponds to
~(L) . - ( ~(zi)
.:. z. this analogy can be seen also from the equations JJ.(z.) = .rs: ~"Z.
~ V 1fi..
Tf_[Z) ': M~"Z.
Infinite series for In(Z) is:
~
::£ " (2- J ~ (f J n .r;
rn « 0
-2-
the series for In{Z) is the same except that signs alternate + and - "J again corresponding to th~ relation between hyperbolic and trigonometric functions. Using the infinite series representation for
Il(2), we bave v Cx/t) ::
_I'f ~ J 2.~ - e c,
I?!.I C h r M ) .' + e u. (1. - to)
(f )~~ rn] (Yl-tYn)!
The result is expressed as an infinite series.oThe accuracy attained depends on the number of terms carried. If we carry the first two, with m.O and m. 1, the result will be valid for ~~-~~ ~ I
If m = 0, we have
1 t:e-f't:cRt- :: I[e-Pta _e-ft]
tf)
": 1.. t:- J.. ]
e t +.;l- +- _
_ - F f>';"
r
- fro < J
+ !. - rt/ + .,l l4 -r __!_] + ~ r - r t- - =.
rr: L C r ('..~ r L e -.e
v (:"j(~ t)
For the voltage .. 1 ) we then have, to this approximation:
;I(Y,t) 0 :? {CJ- +i«> _,,_(t)
-3-
The final value at t _, "" may be round from the following theorem
s res)
_ --x.l/"i&
= e
so that the final distribution is a steady voltage, devreasing exponentially with distance from the generator.
The pulse shape taking only the 1st term of the infinite series is given by:
(J' ;I t 6 [ _ (' to ~ (' tJ
1J( 'X'lt) z _ e - e
.7i
This predicts a final value of
- f t.tI [ ,,1.. t OJ
e 11----j
~F
whereas the exact value has be_en found equal to:
_ x\W~ - t()~ - rt" !~)"L.-
e - e ::. e e :
This is identical with the value given by the first term of the series, showing that the approximate solution is quite good, even for large t. Incidentally, the equations show that the percentage of
distortion tenus increases as we go d own the line, so that eventually
-4-
the initial sharp rise is lost in distortion terms. The rinal value is always greater than the value of the initial sharp jump.
-5-
3..3