Transmission Lines with Exponential Taper*

HAROLD A. WHEELERt, FELLOW, I.R.E.

Summary-Exponential lines, like exponential horns, are over a wide range of frequencies, and which has
useful as impedance-matching devices. They are best inserted in a
high-pass filter having the same cutoff frequency determined by the nearly ideal properties in other respects.
rate of taper. Unusual rules are derived for inserting the line in the
filter with exact matching of iterative and image impedances. Design
formulas are derived for the construction of exponential lines.
ICIO.I IC,I
I. INTRODUCTION
I
A TRANSMISSION line whose impedance js,(l b
IdbI

varies exponentially along its length, has iR k, I I ,iki

electrical impedance-matching capabilities Fig. 1-Electrical dimensions of the exponential line.
similar to those of an exponential horn in acoustical
impedance matching. Likewise, it behaves as a high- The properties of the ideal exponential line will be
pass filter whose cutoff frequency depends only on derived with reference to Fig. 1. It is assumed that
the rate of exponential taper. Such a tapered line there is no dissipation and that the line is tapered in
has apparent utility in matching an antenna (such accordance with the formulas
as a horizontal doublet) with a uniform (nontapered) L1 Cia 47rz
line of much lower impedance. -=-= exp (1)
This paper includes a collection of formulas which Lia Cl xc
have been derived to express the electrical properties in which L1 and C1 are the inductance and capac-
of the exponential line, and their relation to the itance per unit of length, z is the distance along the
mechanical dimensions. Increasing the rate of taper wires from the narrow end (the end denoted by sub-
increases the cutoff frequency, reducing the useful script a), and X, is the "cutoff wavelength" to be
range of frequencies. identified below. It is also assumed that the separa-
The iterative impedance above the cutoff fre- tion is much less than the wavelength in space, at all
quency has a constant magnitude, but its resistive frequencies of interest.
component is equal to the corresponding image im- The line equations are
pedance (resistance) of a constant-k high-pass filter.
Therefore the exponential line can be inserted in such dE dl
z= -IjcoL; - - EjwC, (2)
a filter as a device for matching different impedances. dz dz
Rules and circuits are given for this use of the ex-
ponential line. in which E and I are vectors of alternating voltage
Particular attention is given to the design of an and current along the line. There is assumed, sub-
exponential line comprising a pair of wires whose ject to justification below, a progressive wave, free
separation varies along their length. The method of of terminal reflections starting at the narrow end
design is described with reference to an example in (subscript a)
which a straight-wire doublet is matched with a uni- E = Ea, exp 2grz (-
I
)
form transmission line.
(3)
II. THE EXPONENTIAL LINE I = la, exp 27rz - --
I

The exponentially tapered transmission line, like xc X

the exponential horn, behaves at once like a trans- in which z is the distance along the wires and X is the
former and a high-pass filter. In practice, a trans- wavelength along the wires at the operating fre-
former acts as a band-pass filter, and an exponential quency. From (3)
horn departs from the ideal in its behavior toward
frequencies much higher than the cutoff frequency. dE /1 j\ dI 1 j
_= 2ir --- = 21ri- ---)1. (4)
An exponential line is susceptible of much greater dz XI, X/ dz \ c X
refinement than either a transformer or a horn, be-
cause a line can be made which has small attenuation Substituting in (2), and expressing C1 and L1 in
terms of Cia and Lia from (1), the exponential terms
*
Decimal classification: R142. Original manuscript received cancel out, thereby justifying the assumed progres-
by the Institute, April 28, 1938.
t Hazeltine Service Corporation, New York, N. Y. sive wave of (3). There remain the relations

January, 1939 Proceedings of the I.R.E. 65

66 Proceedings of the I.R.E. January
In further expressions, it is convenient to use the
27r - +)Ea = - laj&'L,a; image resistance Rk of any small section of the line,
/ 1 j (5) so short that its taper can be neglected:
2r(- 1- )Ia = - EaiwC,a.
Rk = = Rka exp -
(11)
Eliminating Ea and Ia, C, ~~~xe
The line is comparable with a transformer whose
47r2 - + 1 ) = co2CiaLia. (6) turns ratio is
/Rkb 2rl
Let q = A! = exp (12)
27r co, 2ir co Rka xc
V0 (7) in which I is the length of the line. It appears that
XI x v
a nominal ratio can be secured with a length only a
in which WC is the cutoff frequency of the line and v is small fraction of X,.
the steady-state wave velocity along the wires, hav- In the transmission band, above the cutoff fre-
ing a limiting value of vo at high frequencies. There- quency wo., the exponential line has a complex itera-
fore, from (6) and (7), the limiting velocity (at in- tive impedance of constant magnitude. The real
finite frequency) is (resistive) component of the latter is equal to the
1* image impedance (resistance) of a constant-k high-
V0 V= aLla (8) pass filter having the same cutoff frequency. The
imaginary (reactive) component of the complex itera-
and the actual velocity is tive impedance is such that the exponential line can
VO
still be matched exactly with a constant-k high-pass
v= (9) filter. The imaginary component always can be rep-
resented by an actual inductance or capacitance. The
real component varies with frequency, and therefore
cannot be represented by an actual resistance.
If the wires have negligible dissipation and are The iterative impedance of the exponential line,
located in air, vo is nearly the velocity of light. The looking respectively into the low-impedance (narrow)
steady-state wave velocity along the wires is greater, and high-impedance (wide) ends, is given by the fol -
becoming infinite at the cutoff frequency WC. The lowing expressions, in which the subscripts a and b
"cutoff wavelength" X, is, in this case, approximately denote the respective ends:
the wavelength in air at the cutoff frequency. This is
a convenient parameter for design, whereas the actual Rk = / )/ RLi, Rkb == / Llb = Rka exp 4rl (13)
wavelength in the line is infinite at the cutoff fre- Cla Clb Xc
quency. At lower frequencies, the velocity is imag-
inary, which denotes attenuation as distinguished Za
= 4/1-
W,2 Xc
-i- =
1
from wave transmission. Rka
In the case mentioned, the steady-state velocity is w2
variable and is greater than the limiting velocity (14)
which is nearly the velocity of light. This is found also Zb co,2 co, 1
in optical filters; it is a form of distortion which is
absent in the ideal uniform line but is found in some Rkb V Cd2 Co CoC
filters. The "group velocity" or "impulse velocity"
/ V 1- coc2
C
,2 _ji Cd
does not really exceed that of light, although the
"phase velocity" or "steady-state velocity" are The latter is obtained in the same manner as the
greater, as affected by the taper. former, the opposite sign of cs, resulting from the
The iterative impedance of the line at the narrow sign of z or 1, denoting the opposite taper effective
end is the driving-point impedance of the progres- toward the backward wave. The square root in each
sive wave assumed in (3); it is obtained by solving of these expressions is equal to the relative mid-series
(5) and substituting from (7): image impedance or mid-shunt image admittance of
the corresponding constant-k high-pass filter, having
Ea (VO o\/LiG the same cutoff frequency. In the transmission band
Za (10) above the cutoff frequency, the square root is real
Ia\V W/co Cia
1939 Wheeler: Exponential Transmission Lines 67

and positive. In the attenuation band below the cut- These relations are valid only in the transmission
off frequency, the square root is imaginary and nega- band, where Ra' and Rb' are equal to the mid-shunt
tive. The remaining term in the above expressions is image resistance of the corresponding filter.
imaginary and its sign depends on which end of the
line is represented.
0- -1--i
) r- --iq--o
,,2 C.'- = -2Cb
Za{ Rct Rb 5

__ __J
c,-_-_
Fig. 2 The resolution of the iterative impedance
into series components.

Fig. 2 shows the resolution of Za and Zb into series

components, as follows:
1
Za = R. +jXa = Ra + (15)
jw(2CG0) U 1 OC/Ca (Trequency

Ra = Rka 1/1 - y
Fig. 4 The iterative impedance and its components.
2

-1 1 Fig. 4 shows graphically the iterative impedance

Xa = 2Ca = and its components in either end of the exponential
(A)(2Ca) line. The abscissas are inversely proportional to fre-
1 quency, to show concisely the limiting conditions at
Zb = Rb + jXb = Rb + jco(- 2Gb) high frequency.
In Figs. 2 and 3, the quantities CG, La, and Cb, Lb
Rb = Rkb |/1 -
are the full-series and full-shunt elements of the cor-
2 1
responding filter. These quantities doubled are the
1 respective mid-series and mid-shunt elements. This
2Cb = relation forms the basis for matching the correspond-
A (2Cb) WcRkb ing line and filter terminations.
These relations are valid only in the transmission III. IMPEDANCE MATCHING
band, where Ra and Rb are equal to the mid-series
image resistance of the corresponding filters. Matching the exponential line with adjacent cir-
cuits is a problem involving the nature of the line
0o.-2Lo; {Ra =-----t4-2L impedance. The line impedance approximately
matches a constant resistance (Rka or Rkb) at fre-
o _z__JL_
_j 0
quencies much higher than the cutoff frequency.
Fig. 3 The resolution of the iterative impedance Closer matching is secured by "power-factor correc-
into parallel components. tion" which converts the line impedance to a pure
resistance which varies with frequency. Either end of
Fig. 3 shows the resolution of Za and Zb into paral- the line can be matched exactly with a high-pass con-
lel components, as follows: stant-k filter. An m-derived termination of such a
1 1 1 1 I filter can be used to match closely a constant resist-
-= j + =
ance.
Za Ra' jXa' Ra' jw(- 2La)
Rka Rka
Rat = Xa' = co (- 2La); 2La = (20)
''C
1-
ijSection *4 4 4 * section
1 1 1 1 1 Fig. 5-Power-factor correction of the line impedance, and
- + = _+ (21) matching the line with a high-pass filter.
Zb Rb' jXb' Rb' jw(2Lb)
Fig. 5 shows the principle of power-factor correction
Rkb Rkb
=
Xb' = Cd (2Lb); 2Lb = (22) of the line impedance. The narrow end of the line is
1 --
c, represented as in Fig. 2, by Ra in series with 2Ca.
The reactance of 2Ca is canceled by connecting 2La
68 Proceedings of the I.R.E. January
in parallel, and the resistance is transformed to Ra'. Since the "constant-k" image resistances Ra, Ra',
(This is the effect of the half-section high-pass filter Rb, and Rb' vary with frequency, they are not very
formed by 2Ca and 2La.) The wide end is also repre- nearly matched with a constant-resistance circuit.
sented as in Fig. 2, by Rb in series with -2Cb. The The matching is greatly improved by the use of an
reactance is canceled by connecting 2Cb in series, m-type half section. The complete circuit at either
leaving purely R&. (It is noted that the elements end then becomes that of Fig. 7. The circuits of Fig. 8
are equivalent to those of Fig. 7, but are in a balanced
4Ca 4b arrangement. Figs. 7 and 8 provide nearly exact
RL Rb matching, at frequencies above the cutoff frequency,
|0C4 4cj between the ends of the exponential line and the
,section 4section constant resistances Rka and Rkb. The choice of m
determines the residual variation of the image re-
Fig. 6-The insertion of the line in a whole-section sistance at the termination in Figs. 7 and 8. A value
high-pass filter.
m =0.6 is suggested, which causes this image re-
added for power-factor correction, 2La and 2Cb, com- sistance to be within about 5 per cent of the terminat-
prise together a half-section filter, divided by the ing resistance at all frequencies more than 15 per cent
line.) above the cutoff frequency. This value places the
Fig. 5 shows also the matching of the line with half- "frequency infinite attenuation" 20 per cent below
of
sections of a constant-k high-pass filter. At the nar- the cutoff frequency.
row end, Ra' is the mid-shunt image impedance of
the half-section shown on the left, whose mid-series
image impedance is Ra. At the wide end, Rb is the
mid-series image impedance of the half-section shown
on the right, whose mid-shunt image impedance is
Rb'. Any number of half sections may be added, ac- 'tLIO.
m
'ZLo
I+m 4Cb, 2mnCb1
[Cb ji/
cording to the ordinary rules of filter design.
The two essential requirements are met by the Fig. 8 Balanced networks for matching either end of
the line with a constant resistance.
matching of line and filters according to Fig. 5. First,
each terminal circuit presents to the line a modified The circuits of Figs. 7 and 8 are about the simplest
image impedance equal to the iterative impedance of that can be used to match closely either end of an
a continuation of the line with the same exponential exponential line with a constant-resistance circuit
taper. For example, at the wide end, 2Cb and Rb in (such as a uniform line). If such close matching is not
series represent the iterative impedance of the narrow required, either of two simpler connections may be
end of a continuation of the line, as if the line were used. The simplest is the direct connection of Rka or
continued without interruption, but actually it is ob- Rkb to the corresponding end of the line. This ex-
tained by the filter elements. Second, each end of the pedient is not nearly ideal at frequencies less than 40
line with its power-factor correction presents to the per cent above the cutoff frequency, because the
respective filter a modified iterative impedance equal power factor of the line impedance is there less than
to the image impedance of the adjoining filter ter- 70 per cent. The power-factor correction of Fig. 5 is
mination. nearly simple,
as and greatly improves the matching.
If the latter is used, the terminating resistance should
mc,m~~~~~~~~ 2Cb have a value which is the geometric mean of the
minimum and maximum values of the image re-
I-rn Rb ~~~~~~~~~~~2L1 sistance of the line, over the required frequency
I X
range. For example, if all frequencies more than 40
per cent above the cutoff frequency are to be used,
Fig. 7-Matching either end of the line with a constant resistance the terminating resistance at the low-impedance end
by means of m-derived half
an section.
of the line should be about 0.84 Rka to match Ra or
Fig. 6 shows an application of Fig. 5 to secure the about 1.19 Rka to match Ra'.
same form of image impedance (mid-series) at both If a definite cutoff frequency is not required, the
ends by the insertion of the line in a whole-section rate of taper may be varied along the line. There is
high-pass filter. The section is divided with three some advantage in approximate matching, if the rate
quarters at one end and one quarter at the other of taper is least at the ends and greatest in the middle
end of the line. part of the line.
1939 Wheeler: Exponential Transmission Lines 69

The usual problem in the design of an exponential z = 0 at the point of maximum separation. This
line is to secure a given impedance ratio over a given greatly simplifies (26) and its simultaneous solution
frequency range, with a minimum length of line, in with (25). This solution involves the wire diameter d,
order to minimize the space requirement and the line the cutoff wavelength XA, and the maximum half
attenuation. The matching at the terminals can be separation Yin. It is expressible in any of several
improved, in general, by using special terminal cir- forms, but an explicit solution for yin is impossible.
cuits, or by increasing the length of line and lowering
its cutoff frequency without changing the impedance Xe 4lrym log d (27)
ratio, or both.
IV. WIRE SEPARATION Xc 4ym 4ym
XI - log- (28)
A two-wire transmission line has an exponential 7,d d d
taper only when the separation of the wires varies xc 4ym
correctly with distance along the wires. If the de- = 47r log (29)
YM d
sired variation of Rk is given, in accordance with (1)
and (11), the actual values of Rk determine the ratio 4Ym
-exp
XC
-- (30)
of separation to wire diameter at all distances. An d 47ry,,
explicit formula for the separation in terms of the d=
4ym
distance, is derived on the assumption of bare wires d= (31)
of zero resistance. exp
XI
4Tym
cI The separation of the wires along the length of the
2y, 2y ZY, exponential line is best expressed in terms of the
ratio y/ym, which is always less than unity. In order
to make this expression as simple as possible, z is
Fig. 9 Space dimensions of an exponential line made measured from the point of maximum separation. At
of a pair of wires.
all points of less separation, and therefore at all other
Referring to Fig. 9 and equation (11), possible points, z is negative. The following formula
gives an explicit solution for y:
Rk
log 4yd 47rz y / d \l-exp (-pn
R-ca 4Y- = exp- (23) (32)
'YM 4ym/
log
d in which
-z
in which d is the wire diameter, y is half the separa-
tion, and z is the distance measured along the wire. Ym
(33)
This equation yields the solution, 4ym
log -
4y 4ya exp 4rz/Xc d
(24)
d d Conversely, the relative distance for a given half
separation y is
There is a maximum separation, beyond which this
.1

formula cannot be satisfied, because the half separa- log 4yin

tion would have to increase more rapidly than the
- z 4y"t d
=
log-log- (34)
distance along each wire. The maximum separation is YM d 4y
log
defined by the relation d
dy the point of
- 1. The shape of an exponential line near
dz maximum separation (z 0), as computed by (32),
=

From (24) is nearly independent of the wire diameter d in prac-

dy wrd 4ya 47rZ 4Ya\exP 4rz/),c tical cases where the latter is very small as compared
--- log exp
-) . (26) with the maximum half separation Yin. A convenient
d: Xc d \ d/
criterion for the shape is the ratio y,/d. Fig. 10 shows
Since z is measured from an arbitrary datum point, the shape of the line for this ratio equal to 1000,
it may be assumed without loss of generality that which is a practical value. (The circles merely indi-
70 Proceedings of the I.R.E. January
cate the plotted points, where p=O, 0.1, 0.2, 0.3, 0.4, the wire diameter d = 0.2 centimeters, and the uni-
0.5 from left to right.) form-line separation D = 10 centimeters are assumed.
The ratio X,/100d is 500. The curve of Fig. 12
gives the corresponding ratio ym/d = 520. From these
two ratios, it appears that X,/yi=96, which value
is verified by Fig. 13. In view of the value of X, the
maximum half separation yin is 104 centimeters.
1000,

1-

L.

.3t amI S.

;d ._-

o 3
Fig. 10 The shape of an exponential line in which the maximum
half separation is 1000 times the wire diameter. 11
u-~ 9

Fig. 11 shows the variation of separation with

distance along the wires from the point of maximum
separation. The ratio ym/d is used as a parameter. -g
18
Ym maximum ~ ~
hal separation (cm)
~~I
2
Interpolation is possible for values of this ratio be- d wire diameter
tween 100 and 10,000. These curves are most useful Fig. 12-The relation between the cutoff wavelength and the
maximum half separation, involving the wire diameter.
in the design of exponential lines.
The minimum half separation is D/2 =5 centi-
meters = 0.048 yin. Reading opposite y/yin = 0.048 in
Fig. 11, it is found that the exponential taper will
y
-1 \\ require a length of wire equal to 3.9 yin, which is 4.0
Y.
meters. Also the variation of separation along the
Y. McLximw-n half separation wires is determined by interpolation from the curves
0.1 .\d. wire diarneter
of Fig. 11. The shape is substantially like that shown
in Fig. 10.

\q K 20 I

-2
- -3 -4- -5 -6 -7 c:V
z wire distance 324
-. macximxm half sepxration tl

Fig. 11 The relative separation along the wires for several values Ds' t00
X :
of the ratio of maximum half separation to the wire diameter.
41
Fig. 12 shows the relation between cutoff wave- -T

length and maximum half separation, computed by

(28). Fig. 13 shows the slight variation of the ratio Y- =maximum half separation °
Xc/yin with the criterion ym/d, computed by (29). d wire diametcr
Fig. 13-The ratio of the cutoff wavelength to the maximum half
V. METHOD OF DESIGN separation, which is not critically dependent on the wire
diameter.
The following procedure is useful in designing an
exponential line according to the above principles Tension cords or separators are connected between
and relations. The example chosen is the problem of the tapered wires, at intervals sufficiently close to
matching a uniform transmission line with a straight- maintain substantially the correct shape. The
wire doublet. The cutoff wavelength X, = 100 meters, straight wire of the doublet is cut out in the center
1939 Wheeler: Exponential Transmission Lines 71
for a space of about 2 meters, and the wide end of (4) H. A. Wheeler and F. D. Murnaghan, "Theory of wave
filters containing a finite number of sections," Phil. Mag., vol. 6,
the tapered line is connected in this space. pp. 146-174; July, (1928). (Lumped line with exponential taper,
If rough matching is sufficient, the narrow end of pp. 162-164.)
(5) "Coupling devices for use in high-frequency circuits,"
the tapered line may be connected directly to the Australian Patent 18,994/29; March 19, 1929. (Lumped and
uniform line. Otherwise a matching filter may be in- smooth lines with exponential taper, applications to impedance
matching.)
serted, such as that on the left-hand side of Fig. 8. (6) "Coupling devices for use in high-frequency circuits,"
No matching filter is required at the doublet end of Australian Patent 18,994/29; March 19, 1929. (Exponentially
tapered smooth lines, also low-pass and high-pass filters.)
the tapered line, in this example, because the rate of (7) A. T. Starr, "The nonuniform transmission line," PROC.
taper decreases gradually in the doublet itself. Like- I.R.E., vol. 20, pp. 1052-1063; June, (1932). (Theory of lines with
algebraic taper.)
wise, improved matching at the junction of the two (8) J. W. Arnold and R. C. Taylor, "Linearly tapered loaded
lines may be secured by gradually reducing the rate transmission lines," PROC. I.R.E., vol. 20, pp. 1811-1817; No-
vember, (1932).
of taper and extending the length of the tapered wires. (9) N. E. Lindenblad, "Antenna for radio communication,"
The above design procedure easily may be modi- U. S. Patent 1,927,522; September, 1933. (The exponential line,
its use as an antenna, its application to matching unequal
fied to fit any given set of conditions. impedances.)
(10) W. van B. Roberts, "Radio frequency transmission line,"
Bibliography U. S. Patent 2,018,320; October, 1935. (Theory of the exponential
line, applications to matching unequal impedances.)
(1) Crandall, "Theory of Vibrating Systems and Sound," pp. (11) R. B. Dome, "Antenna and coupler system, " U. S.
158-174. (1926). (Theory of the exponential horn, bibliography.) Patent 2,104,610; January 4, 1938. (Tapered junction between
(2) E. Dietze, "Tapered filter for alternating currents of doublet antenna and transmission line.)
varying frequency," U. S. Patent 1,603,329; October, (1926.) (12) C. R. Burrows, "The exponential transmission line,"
(Filter with exponential taper for matching unequal impedances.) Bell. Sys. Tech. Jour., vol. 17, pp. 555-573; October, (1938); also
(3) S. Ballantine, "Non-uniform lumped electric lines," Jour. Communications, vol. 18, pp. 7-9, 26-28; October, (1938); and
Frank. Inst., vol. 203, pp. 561-582; April, (1927). (Theory of vol. 18, pp. 11-13, 17-18; November, (1938). (Theory of the ex-
lumped and smooth lines with algebraic taper.) ponential line, experimental results.)