Professional Documents
Culture Documents
Proton Synchrotron
Department
4 February 1970
GENEVA
1970
CERN 70-4
Proton Synchrotron
Department
4 February 1970
GENEVA
1970
SB~ICONDUCTOR
LOSSES
ABSTRACT
The distortion of arbitrary pulses is computed for coaxial transmission cables with
different kinds of losses. Starting from cables with normal skin effect losses, formulae
for attenuation and rise-time of rectangular and nearly rectangular pulses are developed.
The theory of combined conductor and dielectric losses is reviewed. Furthermore, the
contribution of additional conducting or semiconducting layers to the losses in highvoltage pulse cables is investigated. By use of Maxwell's equations, the propagation
constants in the frequency domain are calculated. For coaxial cables with semiconducting
layers also an approximate solution in the time domain i~ presented, by which the distortion of arbitrary pulse shapes can be numerically computed.
SIS/jmr/mta
CONTENTS
Page
1.
INTRODUCTION
2.
3.
4.
REFEREI1CES
11
13
14
15
23
27
29
33
1.
INTRODUCTION
The purpose of tl1is Section is to review briefly the well-known general transmission
line theory [see, for example, Johnson 1 ), Matick 2 ), and Guillemin 3 ) ] and to give some definitions which are used in the following Sections. The time response of a transmission
line to an input signal can be calculated from the transmission line equations for voltage
V and current I as functions of time t and space coordinate z:
(1)
For the spectral amplitudes of voltage and current, V , I , it follows from Eq. (1)
w w
dVw(z)
=-
(R + jwL) Iw (z)
=-
Z5 (w) Iw(z)
(2)
diw(z)
d-;-
-(G + jwC)VJz)
-Yp(w) Vw(z)
These equations are also valid if the series impedance/length, z , and the parallel
s
admittance/length, Yp, are arbitrary functions of w. The general solutions of Eq. (2)
are given by
Vw(z) =
v;
lw(z)
exp (-Yz) +
[v:
v:,
exp (-Yz)-
exp (+Yz)
v:,
(3)
exp (+Yz)] ,
- 2 -
where V+, V- are the voltage amplitudes of the waves moving in the +z or -z direction,
w w
respectively. The characteristic impedance Z0 is
(4)
HIGH-FREQUENCY
APPROXI~lATION
For coaxial
stant and y is a
several hundreds
R << wL and G <<
cables with small losses, Z0 can nearly be regarded as a real conrather simple f1mction of frequency in the frequency range between
of kHz and the cut-off frequency. From Eqs. (2) and (4) we see, if
wC, that
z0 = -y!1.
c
and
Y =
or, if y
1 + R/jwL ,
1 + G/jwC
[1
f1_
(6)
VC
~ (j~L
j~c)]
(7)
a+ jS
"'~"'lR+lG,
- 2 z;; 2 '0
(8)
(9)
The first term of the attenuation constant a represents the losses in the cable conductors, whilst the second term is the loss contribution of the dielectric insulator.
Below a few hundreds of l~z the dielectric losses are negligible. However, at very high
frequencies (more than 1 GHz) they may predominate.
3.1
Let us consider first a coaxial cable with an ideal dielectric insulator, but with
conductors, which have very small, however finite, conductivities a.,
a (see Fig. 1).
l
a
The currents produced by the longitudinal field component in the conductors therefore
lead to losses, which increase with frequency on account of skin effect. The series impedance Zs [Eq. (2)] in this case can be easily calculated with the aid of the Maxwell
equations for cylindrical fields. Here the exact solution [which can be found in Johnson 1 )
and Fidecaro 4 )] is not presented, because already at fairly low frequencies there holds a
very simple, but rather accurate, approximation for Z . One can write
s
(10)
- 3 -
where Ri is the resistance of the conductors per length, and Li and La are the internal
and external inductances per length.
OUTER CONDUCTOR
INNER CONDUCTOR
DIELECTRIC INSULATOR
Fig. 1
!~ell
equations is given by
(11)
with r.,
ra, a.,
aa and~-,~
1
1
1
a being the radii, conductivities and permeabilities of the
inner and outer conductor, respectively (see Fig. 1). From Eqs. (4), (5), (10) and (11)
and with the well-known formulas for La and C
2mo:
c
where
E
~0
~i =~a~ ~ 0 ,
y = Aj;
where
A=l
a=
~A
Bw
Bjw
(12)
(13)
11.
A ,
/E(_l_+_l_) ln(ra/rJ
1
2 V ~ ra
(14)
(15)
ri
These formulae are valid as far as the skin depth 6 12/~wac is small with respect
to the radial thickness of the inner and outer conductors.
(16)
- 4 -
For a matched or infinitely long coaxial cable it is now possible to calculate the
distortion of arbitrary input signals V.(t),
the frequency content of which does not extend
l
considerably beyond the limits of validity of Eqs. (13), (15) and (16). From Eqs. (3) and
(12) one gets the following relation between the Fourier components Vw(i) and Vw(O) of the
output voltage Vout(t,i) and the input voltage Vi(t)
(17)
If V.(t)
= 0 fort< O, we find the output voltage Vou tlt,i) appearing after the length i
l
of cable by use of the convolution theorem 5 ) as
Vout (t,i)
A~
t-B
2v 1T
2
)
~/
dx H(t-Bi),
(18)
where H(t) is the Heaviside function. If one neglects the pure cable delay, Bl, andrestricts oneself to the distortion of the signal and if the "cable rise-time" To = (A) 2
is introduced, one can write Eq. (8) withy= t/T 0 , z = x/To also as
y
Vout(Y)
2~
(19)
The output voltage Vout' according to Eqs. (18) and (191 can be numerically evaluated for
any input signal by a FORTRAN program on the computer. However, in several cases one can
easily find an exact solution.
A very simple response comes out for a delta pulse of weight 1, Vi(y) = o(y)/T 0 , in
the input of the cable. By definition of the delta function o(y) the delta response dc (y)
of a cable with conductor losses is expressed as
(20)
The delta response dc(y) is a good approximation for rectangular pulses of weight 1,
if the pulse length is much smaller than T0 In Fig. 2, dc(y) is plotted for To = 1. The
response uc(y) to a unit step input pulse, Vi(y) = H(y), follows directly from Eq. (19) as
Vout(Y)
uc(Y)
erfc
( 2 ~)
H(y)
[1- erf
(z~)] H(y) ,
(21)
where
erf(x)
1 - erfc (x)
Vout(Y)
sc(Y)
erfc ( 2
(22)
- 5 de (y)
1.0
0.9
0.8
0.7
0.6
0.5
01+
0.3
0.2
0.1
0
0.5
1.0
10.0
5.0
y = t/<o
A plot of square pulses with the same initial length T distorted by different lengths of
cable is given in Fig. 4.
The next example is a trapezoidal pulse (see Appendix I) with a flat top length T
and linear rise and fall T'. With a= T/< 0 , a' = T'/<o the response tr c (y) to such a
pulse can be written
V,,, (yJ
y-r
+ y-rerfc [,);-] dx
H(y- a - 2a')]
(23)
(24)
- 6 -
1,0 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
0.8
0.6
0.4
0.2
o~------------,--------------r------~-----r,-~--~-r--r--r~--,-~--~.-
40
15 20
10
60
80
100
Sc
(tiT)
1 .o+-----------~T~/~T~=~oo~----------~
0.8
0.6
0.4
0.2
0.5
1.0
1.5
2.0
(tiT)
TI1e distortion of a trapezoidal pulse with a'/a = 0.2 can be seen in Fig. 5. Figure 6
shows the response to a parabolic input pulse (see Appendix I), which is computed numerically from Eq. (19). For small rise and fall times, Eq. (24) is also a good approximation
for the parabolic pulse response. Finally we regard the distortion of a pulse without
flat top, namely Vi(y) = sin 2 (rry/a), the response of which is presented in Fig. 7.
- 7 -
For practical purposes it is often not necessary to know the total shape of the distorted signal. Generally it is sufficient to know the attenuation and the rise-time of
the pulse. Since the pulse, after it has passed through a certain length of cable, does
not anymore present a flat top, the most convenient definition of attenuation of square
pulses with and without initial rise-time is att = 1- Vmax/V 0 , where V0 is the flat top
voltage of the input pulse and Vmax the maximum voltage of the distorted pulse.
trc (tiT)
1.0
0.8
a= T/t 0 = 35
a'= T'/T = 0.2
0.6
0.4
0.2
0.2
0.4
0.6
0.8
1.2
1.0
1.4
1.8
1.6
2.0
tiT
T'
rise-time;
Vout
PARASOL~
INPUT PULSE
1.0
o.8
DISTORTED PARABOLIC PULSE
SQUARE PULSE AFTER
LENGTH OF CABLE
0.6
a:T/t 0 :35
a'= T' IT = 0.2
ME
0.4
0.2
tIT= 1
04L----+---~~--~----~-----r----~----~----~----r---~-------..
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
tiT
- 8 -
Yout
0.5
1.0
1.5
tiT
A rough idea of the rise-t:iJne of a distorted pulse is given by the "cable rise-t:iJne"
To = (M) 2 For a unit step To indicates the t:iJne after which the output signal is reaching 48% of the input voltage V0 With this value, however, one does not take into account
the strong bending of the response function between V0 /2 and Vmax Therefore it seems more
convenient to define the rise-t:iJne t
as the t:iJne in which the output signal is increasing
rx
from 0 to a certain ratio x of the max:iJnum voltage Vmax. For a square pulse of length T
the maximum nearly coincides with the starting point of the second term of sc(y) [see
Eq. (22) and Fig. 4] provided T > T0 Therefore it holds approx:iJnately with q = T/T 0
that
(25)
q > 1
and
att
~ erf ( 2 ~)
(26)
q > 1
We can also calculate the relative rise-t:iJne for a rectangular pulse Yrx
Eqs. (22) and (25).
Yrx
~ [(1 - x) .fTici
+ xT
t rx/T, from
(27)
ATTENUATION
ott
i
I
t
:1 i:I I II i
I
I
'
'
I
'
0.1
0.01
!
'<
0.001,_~~~~~~--~-W~~~~~~~~~+4~~~~~~~~~~~~~--. .
0.1
10
100
1000
q = T I '1'0
Fig. 8 Attenuation (att) of rectangular pulses in dependence of pulse length T and of the "cable rise-time"
conductor losses only. To = (A~) 2 ; ~=cable length.
To
NORMALIZED RISE-TIME
Yrx
!. I
t I .
: .i
Ill .
1'j....
I I!
I
I . .
-~
I
i
0.1
0.01
~
Cl
i
'j1:: :i
..... j
I I
0.001
Ii Il
II
.I
I
'
0.0001
I 1
10
!
I
j I
100
!'
1000
10 000
q =T I "t 0
Fig. 9 Normalized rise-time Yx (from 0 to x Vmax) of distorted rectangular pulses for cables with conductor losses only.
y rx = t rx /T; t rx = rise-time from 0 to x Vmax ; T =pulse length; To = (A) 2 ; = cable length.
- 11 -
attenuation factor A of the cable, either theoretically [Eq. (15)] or from a measured
attenuation curve a(w) [Eq. (13)].
Approximative formulae for attenuation and rise-time of distorted pulses with finite,
but, compared with the pulse length T, small initial rise-times T' can be derived from
Eq. (24). With q = T/T 0 and a" = T'/T the attenuation is given by
att
1
erf ( 2/Cl
J [1 - a"4 exp (-
41q)] ~
a"J
1 (1 - 4 ,
;:rrq
(28)
a"
+z
y~x
+X
= t~x/T from
3/2) _
Yrx - Yrx
0 to x
= V(y~x)/Vmax of
+ a"
z '
(29)
jWB
+ AfJW + ! Bwtg 8
(30)
Physically this expression cannot be completely correct, since any transfer function
should contain j and w only in the combination jw. Nevertheless, Eq. (30) is often
used 4 ' 6 ' 7 ) and seems to be a rather good approximation.
The distortion of arbitrary input signals on a cable with conductor and dielectric
losses can again be computed according to the convolution theorem. But before doing this
let us first deal with the pulse distortion by dielectric losses only. Disregarding the
cable delay Bt, it holds for the ratio of output and input spectral amplitudes that
(31)
where b =! B tg e. For an arbitrary input pulse Vi(t), Eq. (31) can be solved by Fourier
transformation. The response to a delta pulse of weight 1, V.(t)
= o(t) follows then
l
- 12 simply as
(32a)
or with Td = o, t/Td =
Vout(u)
dd(u)
2 1
1
TI Td 1 +
=-- - 2
(32b)
(33)
Vout(Y) =
(34)
1vhere fl(y) is the response to a delta pulse of weight 1 of a cable with conductor and
dielectric losses. fl(y) can be computed by the convolution
y
ll(y)
T02
J dd(y' x) dc(Y- x) dx
(35)
tl T
- 13 -
convolution between the square pulse response sd(u) for dielectric losses only and the
delta response dc(y) for conductor losses only:
y
Vout(y) = sciY) =
T0
Jsiy'
(36)
x) dc(Y- x) dx
In Fig. 11 a few distorted square pulses of constant initial length T are presented
for different values of y'.
INPUT PULSE
1.0
---------'--------,
I
I
I
0.8
0.6
0.4
0.2
0.5
1.0
1.5
2.0
tiT
4.
- 14 -
4.1
The losses arising in an additional conducting layer are mainly due to skin effect
as in the metal conductors. The total loss of a combination of a well-conducting (metal)
and a 'vorse-conducting medium (for instance graphite) will now be studied for a plane
case as shown in Fig. 12. We assume having a longitudinal electrical field in z direction,
MEDIUM 0
MEDIUM 1
IDEAL DIELECTRIC OR
FREE SPACE
E, IJ.IJ.o,o'sO
BAD CONDUCTOR
<l'=<l'l,IJ.-IJ.o
MEDIUM 2
GOOD CONDUCTOR
<l'=<l'2,IJ.;IJ.o
E:-Eo
y
ELECTRICAL FIELD
STRENGTH AMPLITUDE
Ez(O)
Ez(Y) exp (jwt), the amplitude of which only depends on they coordinate. Since o 1 << o2 ,
at low frequencies the main current is flowing in medium 2 with a skin depth of
Oz = /Zjlpowo 2
Then the resistance per unit length (in z direction) of a strip of width
6x (in x direction) is given by
(37)
At frequencies high enough, where 82 is going to zero, the current is forced to flow
mainly in medium 1, where again skin effect is taking place resulting in the higher
resistance
(38)
if 81 << d = thickness of the conducting layer 1. The exact solution for the total series
impedance ZSl. of medium 1 and 2 is given in Appendix II. The frequency interval where the
real part of ZSl., the resistance RS (w), is rising from R2 to R1 , is determined by the parameters d, o 1 , o2 Figure 13 shows the exact frequency dependence of Rs(w) for some special
values of these parameters and for 6x = lm.
- 15 Rs(w)
(Q/m)
~l~:~-(V
~
10
.., ll:l-1
~
'\
;I~ ~
.
6.x :1m
d
d
=0.35mm_
=0.35mm
=tl.2mm
=0.3
k! =0.2rr
d
mm
\
'\.
\ l jv
~~
t?j ~
,..
,.. ,.....
,.. ......
'?~;
c,O
,.,.6.,
;,.'
'II l.~
,o,
...
o'1-,
vv ~
'
, ... ,..
,..
;;,'
......
~:
...
1 o-
~
10 6
~,
10 7
10'
10.
10 11
10 12
w = 2ttf ( 1/sec)
Fig. 13 Real part, Rs(w), of the series impedance Zsi calculated for skin
effect in two conducting media (e.g. copper and graphite) according to
Eq. (AII.S). d; thickness of the bad conducting layer (medium 1);
01 ; conductivity of the medium 1.
With Eq. (8) and Eq. (AII.S) the attenuation constant a can be expressed for 6x
2nr as
a
( 40)
In the time domain the computation of the distortion of arbitrary pulse shapes by a
cable with a propagation constant y, as given by Eq. (40), can be done only in a rough
approximation.
- 16 -
OUTER CONDUCTOR
INNER CONDUCTOR
SEMICONDUCTING LAYER
DIELECTRIC INSULATOR
conductors and the dielectric insulator to be ideal (oc = oo, odiel = 0) and the semiconductor tO have the Same permittivity E and permeability ~ as the insulator (~ = ~
c
~diel
= ~s = ~o)
In Appendix III, ys is calculated with
as follows
( 41)
Ys
For 6 1 we can simplify
(42)
Separating this expression into real and imaginary parts it holds, with B
~.
that
(43)
and
wB
(1
1
)
+-b,2 1 + (Ew/os}
2
(44)
These formulae are equivalent to the circuit shown ill Fig. 15.
Equations (41) to (44) were developed under the condition of infinite conductivity
of the conductors. If we assume now having also losses in the metal there will not be
very much change in the field distribution. The longitudinal electrical field components
will not be exactly zero at the interfaces between insulator and metal and between semiconductor and metal. Nevertheless they are so small that one can linearly superpose them
on the semiconductor losses. So we can write the total propagation constant
(45)
- 17 OUTER CONDUCTOR
INNER
CONDUCTOR
Fig. 15
C
s
G
s
-TnP
'-or-'
nonnal cond.
delay losses
(47)
semicond.
losses
The antitransformation of Eq. (47) into the time domain is possible by using the inverse Laplace transform J 0 (alt) of the image function exp (-a 2 /4p)/p [see Ref. 5]. J 0
is the normal Bessel function of order zero. Disregarding the normal cable delay 'n we
find the transfer function FT(t,~) in the time domain as
t
FT(u.)
>
..rr:r;,w
J exp[-T0/4(t0 D 5
r:::
2vTI
t- X
)3/2
x)J
exp [-ws (x
dx
-r;;
(48)
yX
where I 1 is the modified Bessel function of order 1. TI1en the response to an arbitrary
input pulse V.(t)
can be computed from the convolution
1
(49)
ATTENUATION
(db/100m)
o<.
1000...,
'lj'
rj
II
!' ..
II'
1
.. ,1
i
i
I'"'
l.
,NO
MIC)N
. j~-+~~" , ,
ifi
:,
1:
r-.
.. 11tth,
'"'il:i! ~ :,;1i:WiH~
, ::r
~l ~,OR
:!.: '
.., ,... l
,
I
: ;J:i
j' "
11
lj
,,,J
!L
H+
!
~ ~~~,.,
j
11
:!'l'l'i'
~.tjLIJijl,:
t ~i : :,
I I I
'Ill'''I' I"I,
'" 1
: :, ,' ::
11
1
t
lqrlt''t
fl
'I
ltj
~ LL:,..~.J,
11
144;.1).o!"!" .J11I
~.,.,~'".:"
',h
I '
}:IW;~1:1:ir!i.,~~;;::::,i
:'i;:1i '1:,'1::~ ~:~:,1:
~!-;-tt ~~~I~~~~.~::. . . I.
.. i1 ~..:;;.~,.,;;r.~~.,,.
~~
"""'I,,
:! .,, 1,.,,...,.
:,
II
11
11
.. ,
11
1,
'
''I'
iii'
tl
lt,l
'
~.u!
,'.1[;~11
I,
iii
'
'I
'
ijil}:t ,..
I"1
,,,,
'
'
'
'"
,,
'
U n:
it;:n
0 "
"'"~
~ "'I '1 I~.
'1 ,1 '1I ;,.
I
,
ijl "'
I~
li"'
,lfii:ti
1
J1 '"I I
t,
'
,,
l:!ill
1 I
,r,,,,
.::"' '. II
!'
,,
I'
1 J'
:q:nlnr.~!
.. I,
w~: ~-T:"'~tn
... ..
tr;.
~~~~
,:
:;;
:-
~~
+1
'~:.
~~~ ~
II
'
i:'I
cc;;I::
i'i'
,,~. "~.,
11
1 ,,
1:
::. ;~~
,1: :_;~.-? i!';r
.:.
:~FL~~:::::;~.~ITi
':il:l!i'
'H!JIT
,f,,"'~+-:
~~m
I:.:.
:.' ' .. t
'I
I j
'J
....
..
'
,.
":
'
: '
f-'
.:c
. :
', ,
.
"
.l
!i1
:::
;:!
,::.i
I'
::
::
:!r,
:
i
I
:
..
'
,:;
,.
ijl
', " l :;
,Htrr
' 1t ' I '
'
;l)ffi,.rt I!'
:li~Jlli:': : l ~
"
00
...' .'
I"
' . ', . ,.
.
J !
.:.r
""
ll:
IWI
. .. " 1; 1
lt'
"...
ii
!li
r ; i!i,
:j
rt:
IH+:t: :K
... .
:i
1: 1
1:"
!:::
!'....!i
....
:c:
.. ,
: :. :_
I'
c,
'll_;
l'>'"fl':
!'i!]:
'
:t
;-;-TT
..
'
I!
Hl::u
'I
'
,j,,
'
,,.., II
!1'
;it
,,,
[i[F'
' '
,:::u
.
::i , .. ' ' '
'iil;i.;:;:
;:i;
.. ..
"'i't
j!: !i ljiJit1.;;::'
1
100
ij
;j,tti F7"illi:
l ,.. ,~1
.II
lljjil
I"Cfh'"i:ft
1: ;..j; !":.;
1:'' cl' '!.,
I +th mel:: tt+1;
,.,
.
1 ,
'
...1t. 11 .f:"
.
1 1
'
1000
p;;J.
IJ!]'
'+
~l[ I h-''
111
;;;,,
I
I<
BILE
i(j
lii''''"'
'ji.'i!f"''lj
!:l;!~n
~;t!t+f1
11 ''
wl!Jli:i:,w;l::::jr'
j'[[)1i1!!,
:1 " 11 ,,1f+H,,lt ...
l'lil!l
10
Ir . i
I' t"+c
.
u
i''Fi!: :~~!'::~
. ..
::;1t
1!!H
1 li~~~
'!r:::: 'I' :
I' ......
t:~:,", 1,~~~
,,, .... r+qHf*~,
, , ,
~ .:.,.c 1':!: 1,"
'
r:'~iHTt,
liiitijjj" 1 tt
:n nn ,, iti1
'I
1,1,
~~rr,,
" 'II'
' :
IIll
~-I''''""
'
0.1
"i
Iq
I I
'
,.
'""'I'
,, ...
117 ""
I,
- '""
1.. , I
lI
,: , .,.,
.
~ ......
I j I I,
1 .,,1,~.,
,,,
o I
,-t
~~.,1
r8.m1
1ril 1 1
1 1-+
.'
'
-~h
!:, ~~
~'-~~~
r1 ~~
tt' LJJl! I I i.ol ~
1'1
1 "1 ,,~~~ t~r-!LHfi ~ H
::~i
t
'"l'n.l':j
f Tt'rt'
~~ 1JPI!W
lti! 11
11 Ll'ill!i
I '~H'
:i :;l;
''I
' , ,.
:",,;"!
'
,:
':;
~~~
i
11i
il1't rr:l~l+i~~;;HU!'Hir
[]I
' l
~
..
'
'"'
, 1.' I "~~~
. ~
'
,.
r ,,,,.
,.,
't"' ~~'''
1I'T
:~,iT
H.t'"
NL
+-""itW++j+fj:li++:,.]J!frt
1 1
0 1
j,il~l""""
1'1,
11
I!
Ii I
I'
I'
. .
I:~ ~
"'z~
;:
t ~ .[r
1 --~!~"~
!'rl
,...-~
:I
lj[lI
..
-4 . ) . . . . .'" '. '"-
~
~t:'['tlf
H~ 1~ jj~;r t,
...t~~~~.1
I
1"
;---;-
t I
1/l I
i!
I
l
i
i,ll
j" ,. '
;~11
~1,1
,, '
.~
::
11
'I
1-\ilj,;j,,;,l
,1!11 1t 1 'I'
!;:
I'I
,:;1
II
IIi"'
I
J:,
' I i!
~: III II
f
;,
~''1I,H I , r"
l"""l'i!; 1
':::,. ': . ~~
'
I.,
tl
' I
" . ::
'
I:
'!.~,
[I
I
I
: 'I ; I I
''I,/
<
'
I'
It
I II II i' I I"1':
. ''
J
I
.. :,ii!l,,, .. ,i!l~'IJ'
~
;11'
J
tm,11
iii! :
jJtjr
1
1 I L
11 ..;
...
''
'
r'
C).
'
I'.
I I
V:II
,.
'i
.:. :::,1::
"
j' 1
I+ ttdtli~;ffipj
+!+!:"~ ,
, ~i;+H-t
'+ 1
;it '''t.~,
~ ;'!! ~ 'tj '~~--'
1 Iii' 'lf,l I r I: I : w~
n.:l:;'"''lqtlri d ,r '.
1
1
11'1
'
'
' ~t:
iI!
. ;:
~:.1::
:i.,
11i'j
11''11~
CTHEO~Er
..
:.J~;"'
,; :
]:,,
lH,,,~IriJ
t!
~"""
!
:.,..;~..
I,
'l:ll .. Bl
H
, 1 r
'
'-
'';;;1' ,;,'i:c::
: .;:;,; ,:;
! .
IttM
,iitT
i~tHjf!itj~J
tij'i~
j : : +:'
"'''"
''
I /
.:. . ,
...
rJ
LO
..
.......
''.
,,. ... '
,; +1 :!
~"I l:~
~ ' '
...:::J1 ':~.
,:rn_;:.;:~.:::L:~
~ HGf :C
1-~:t'-c'~:J:
_,d
TT -,
\1
-.
1-:r
-~, ;::,, .....
l;::i:1.fl;;;,,..
"; Hie:"'
I 'IH~ -''
I:L
I I
r. . . . ,
~..--j
l..~ ..-!Ti
.. ,nn .,i
'
'
'
1-h: ....
r-c-
I
' I
iI
100-l ..JfoiLEU
,,. .., C.4tf13l
,...... ''
:J_,..J:::.:
...
I, J--,r;;: i: '"
.I
I. . I
10
'
I
:i;
q ..
.. iljf,,
ttd'i
I'
I';
'
I i
..
FREQUENCY ( MHz)
Fig. 16 Real and theoretically calculated attenuation constant a as function of frequency for BICC 40P3/20
conducting layer.
high-voltage pulse
cable, which has a semiconducting layer of d = 0.7 mm nominal thickness between the insulator and the inner conductor. The characteristic parameters of this cable are given
by r.1 = 6.3 mm, r a = 10.6 mm, 0c ~ 2 x 10- 8 (1/~), s/s 0 = 2.3. Since the conductivity 0s
of the semiconducting layer was not exactly known, the frequency response (attenuation)
of the cable was measured (see Fig. 16). A theoretical curve according to Eq. (45) was
fitted to the experimental one by choosing the parameters w = 5 x 10 9 (1/sec) or
s
1
0s = 10- 1 (1/~), respectively,~= 0.23 and A= 2.70 x 10- 7 (sec 2 /m). The effective
values of ~ and A are higher than those calculated from the nominal values of the characteristic cable parameters. This may be mainly due to the braided conductors.
With the parameter values given above the unit step response
t
Vout ( t,)
JFr(x,) dx
(SO)
was computed for a cable length of 80 m (To = 0.46 nsec, TD = 46 nsec). Figure 17 shows
a comparison between the computed response and the response experimentally measured after
the same length of cable.
Figure 18 shows a few distorted pulses computed with different linear input risetimes. As for cables with conductor losses only [Eq. (29)], it holds approximately that
the rise-time of the output pulse (e.g., from 10 to 90%) is the sum of the cable rise-time
for a unit step pulse plus half of the initial ris~-t~e of the input pulse.
From Figs. 17 and 18 it appears that the pulse rise-time is much more deteriorated
by a cable with a semiconducting layer than by a cable with conductor losses only. Besides
the normal delay Tn' a considerable additional delay is produced, which can be quantitatively
UNIT STEP RESPONSE
CALCULATED DISTORTION BY CONDUCTOR LOSSES
1.0 - - - - - - - - - - -
ONLY
------
0.8
0.6
l
0.4
0.2
0~----~----~--~----~-----r-----r----.-----.-----~--------~-----r--~~
10
20
30
40
50
60
70
so
90
100
110
120
TIME ( nsec)
Fig. 17
- 20 described by TD = 6B~/2 [for definition see Figs. 19 and 20; compare also with the
approximation of Eq. (47) for small p]. The attenuation of a long pulse is mainly determined by the conductor losses (Fig. 20). The semiconductivity as influences only the risetime of the pulse (Fig. 21). Figures 19, 20 and 21 present the unit-step response for constant cable length, but different sets of the parameters TD' T0 and ws.
1.0 -Vi-/Vo
-T---.------,.--------------------------1
I
/
(J
(J/
"-'/
PI
:'tl
cl
"/
~--,
,, /
~.;,v
//
I /
I I
0.4
J>
~,
0.6
<-'7'
"''
''/
0.8
//
I I //
/I/
II/
L~/
0.2
0~----T-----r---~~--~--~-r-----r----,-----~----r---~~--~-----r--~~
10
20
30
40
50
60
70
80
90
100
110
120
TIME (nsec)
1.0
..l..o
0.8
.lo
= 42.5 nsec
= 4 7.5 nsc
0.6
'to
0.4
0.2
10
20
30
40
50
60
70
80
90
100
110
120
TIME (nsec)
- 21 -
Increasing 6, which means the thickness of the semiconducting layer, results in an increasing additional delay only. Higher skin effect attenuation (A) leads to more attenuation
and higher rise-time, whereas decreasing the semiconductivity a increases the rise-time
s
only.
1.0
:0.31nsec
:0.52nsec
=1.44nsec
0.8
0.6
to
0.4
0.2
0+---~----~--~_.~r---~---,----~--~----~--~----~--~----~
10
20
30
40
50
60
70
80
90
100
110
120
TIME (n&ec)
1.0
0.8
(IJ
=2.8x10 9 (1/sec)
0.6
0.4
0.2
10
20
30
40
50
60
70
80
90
100
110
120
TIME (nsec)
- 22 -
Acknowledgements
I would like to thank !llr. B. Kuiper for encouraging this work and for his helpful
criticism.
It is also a pleasure to make acknowledgement to Mr. A. Messina for several stimulating discussions and his help with experimental work. I am grateful to !ltr. U. Berger for
the preparation of the figures.
- 23 APPENDIX I
SDr~E
FREQUENCY ANALYSIS OF
The harmonic content of the transmitted pulse gives an indication for the applicability of a certain transmission line theory. Therefore, the frequency spectra A(w) of
some typical pulse shapes V(t) computed by
+oo
A(w)
(AI.l)
are presented.
a)
Rectangular pulse
ll'
v
Vo
a)
- T/2
-12tt/T
Fig. A.l
-8tt/T
+ T/2
-4tt/T
4tt/T
Stt/T
12tt/T W
The rectangular pulse in (a) the time and (b) the frequency domain.
A( w)
V sin w :!:2
0
TTW
(AI. 2)
- 24 -
The main frequencies are concentrated in the first few peaks of the frequency distribution (Fig. A.l). The frequency spectrum of a delta pulse, which is a special case of a
square pulse with V0 + oo, T + 0, extends from 0 to w + oo with the same amplitude,
A(w) = V0 T/2n = const.
b)
Trapezoidal pulse
v
Vo
a)
-(T/2+T")
+ (T/2+T')
b)
-12n/T
Fig. A.2
-8n/T
2n/T' -4n/T
4n/T
2n/T'
8n/T
12n/T W
Trapezoidal pulse in (a) the time and (b) the frequency domain.
The frequency spectrum of a trapezoidal pulse with a linear rise and fall of length T'
can be expressed as
A(w)
2V
= - --0
TIW 2
T'
(T T')
T'
sin w - +
- - sin w2
2
(AI.3)
Compared with the ideal square pulse the finite rise-time T' leads to a concentration of
- 25 -
the harmonic content in the first few peaks of the frequency distribution (Fig. A.Zb).
This is even more true. for a parabolic pulse.
c)
Parabolic pulse
A parabolic pulse (Fig. A.3) is an even better, and rather simple, mathematical
representation of a real "rectangular" pulse.
a)
-(T/2+T')
+<TI2+T')
lA (w)l
Vg(T + T')
2Tt
b)
-12~t1T
2~t!T'
-81t/T
-4Tt/T
Fig. A.3 Parabolic pulse in (a) the time and (b) the frequency domain.
Rise, fall and top of the parabolic pulse are given in the time domain by
v =V-0
2
(t
+ T/2 + T')
V = V0
for
(T'/2)
V =V-0 ( 1+2 t
(t
+ T/2 + T'/2)
2
(T'/2)
for
for
AI.4)
- 26 -
l~ V~2
TIW
+ T')/2]
(AI .5)
- 27 -
APPENDIX II
SKIN EFFECT IN TWO CONDUCTING LAYERS WITH DIFFERENT CONDUCTIVITIES
The series impedance Zsi of the arrangement shown in Fig. 12 can be calculated with
Maxwell's equations
_,.
a _,.
rot E = -
ll
at H
_,.
_,.
rot H = a E
(AILl)
a _,.
+s
at E
(AIL 2)
a2 _,.
2 E
at
(AIL3)
_,.
_,.
liE =
]l<J -
a _,.
E
'dt
+ JlE:
We assume having only a longitudinal electrical field component E (y) exp (jwt) in
z
the conducting layers. For o 1 , o 2 >> sw it holds that
(AIL4)
]1 1
= ]1 2 = ll
+ j) v]
t.' exp [ -(1
ll exp [-(1
L:'
L:'
exp [(1
+ j)
v]
exp [(1 + j) vJ
Eo
DN
+ j) vJ Eo
- 28 -
The total
(Ail. 7)
The total series impedance Zsi = E0 /(J!1x) of a strip of width !1x per unit length can be
expressed as
(1
+ j)[z' exp (v) exp (jv) +11' exp (-v) exp (-jv)J
(AILS)
where Rs(w) is the real part of Zsi and Li the internal series inductance per length.
- 29 APPENDIX III
CALCULATION OF THE CDr1PLEX PROPAGATION CONSTANT y 5 IN THE CASE OF AN
IDEAL TRANSMISSION LINE WITH A SEMICONDUCTING LAYER
We apply the Maxwell equations to the coaxial structure shown in Fig. 15. Conductors
and insulator are assumed to be ideal. Only the longitudinal electrical field component
E , the radial component E and the azimuthal magnetic field component H~ are different
z
r
~
from zero. All derivatives a/a in the Maxwell equations (AII.l) and (AII.2) are equal to
zero.
We can write the field components:
Ez = E2 ( r) exp (- y 5 Z) exp (jwt)
(AIII .1)
(AIII.2)
Hq,
(AIII.3)
where Ez, Er and H are the amplitudes of the waves spreading in z direction with the propagation constant-ys. For the component Ez it follows from Eqs. (AII.3) and (AIII.l) that
(AIII.4)
where
(AIII.S)
(AIII.6)
The solutions E (l), E ( 2) in the two media 1 (semiconductor) and 2 (insulator) can be
z
z
written with the Bessel function J 0 and the Neumann function N0 as
where
A. 1
and
A. 2
~( 1 )(r)
A1 J 0 (jA. 1r)
+B
~(z)(r)
A2 J 0 (jA. 2r)
N0 (jA. 1r)
(AI II. 7)
B2 N0 (jA. 2 r),
(AIII.S)
= Ez (raJ = 0
- 30 (AIII.lO)
(AIII .11)
The boundary condition for the magnetic field component H at the interface of insulator
and semiconductor,
(l)
H (r i
+ d)
(2)(
ri
+ d)
(AIII.l2)
(AIII.l3)
and together with Eqs. (AIII.7) to (AIII.l2) the following equation for
is obtained:
(AIII.l4)
Generally it is difficuLt to solve this complex transcendental equation, which contains also the higher modes
of wave propagation. For our case we try to find an approximate solution. For an ideal
dielectric insulator (2) it holds that
where
y = 1.78107.
(z)
- -TT ln =yz
(AIII.l7)
- 31 with~=
ln (1 + d/ri)/ln (ra/ri).
(AIII.l6) results finally in
(AIII .18)
or for 6. 1
(AIII.l9)
- 33 -
REFERENCES
1)
W.C. Johnson, Transmission lines and networks 04cGraw Hill, New York, 1950).
2)
R.E. Matick, Transmission lines for digital and communicative networks U4cGraw Hill,
New York, 1969).
3)
E.A. Guillemin, Communication networks, Vol. 2 (John Wiley and Sons, New York, 1953).
4)
G. Fidecaro, The high frequency properties of a coaxial cable and the distortion of
fast pulses, Nuovo Cimento Suppl. ser. X, 15, 254 (1960).
5)
6)
G. Brianti, Distortion of fast pulses in coaxial cables, CERN 65-10 (CERN, Geneva, 1965).
7)
M.J. Lorrin, Transmission d'une impulsion tres breve par cable coaxial, Bulletin de
la Societe Fran~aise des Electriciens, 7e serie, tome 8, No. 94, 1958.
8)
(Oldenburg,