Introduction to Horn Theory

Bjørn Kolbrek

September 21, 2007

Abstract c The speed of sound, 344m/s at 20°C

This article will deal with the theory of acoustical ρ0 Density of air, 1.205 kg/m3
horns, in the way it applies to loudspeakers. The
basic assumptions behind classical horn theory as f Frequency, Hz
it stands, will be reviewed, the dierent types of
horns will be presented and their properties dis- ω Angular frequency, radians/s, ω = 2πf
cussed. Directivity control, wave-front shapes and
distortion will be discussed. In this article, I will try k Wave number or spatial frequency, ra-
to keep the math simple, and where it is required, ω 2πf
dians/m, k= c = c
the meaning of the equations will be explained or
visualized. The focus is on the understanding of S Area
what is going on in a horn. The practical aspects
of horn design will not be treated in this article. p Pressure

ZA Acoustical impedance
1 Terminology
√
The following terminology will be used:
j Imaginary operator, j= −1

Impedance: Quantity impeding or reducing ow

of energy. Can be electrical, mechanical or 2 The Purpose of a Horn
acoustical.
Before we look at the theory of horns, it can be
Acoustical Impedance: The ratio of sound pres-
useful to look at the purpose of the horn. Where
sure to volume velocity of air. In a horn, the
are horns used, and for what?
acoustical impedance will increase when the
Throughout the history of electroacoustics, there
cross-section of the horn decreases, as a de-
have been two important aspects:
crease in cross section will limit the ow of air
at a certain pressure.
ˆ Loading of the driver
Volume Velocity: Flow of air through a surface
in m3 /s, equals particle velocity times area. ˆ Directivity control

Throat: The small end of the horn, where the

One would also think that increasing the output
driver is attached.
would be one aspect of horns, but this is included
in both. Increasing the loading of the driver over
Mouth: The far end of the horn, that radiates into
that of free air increases eciency and hence the
the air.
output, and concentrating the sound into a certain
Driver: Loudspeaker unit used for driving the solid angle increases the output further. Let us look
horn. at the two points in detail.

1
Loading of the Driver The loudspeaker is
a generator of pressure. It has an internal
source impedance, and it drives an external load
impedance. The air is the ultimate load, and the
impedance of air is low, because of its low density.
The source impedance of any loudspeaker, on the
other hand, is high, so there will be a considerable
mismatch between the source and the load. The
result of this is that most of the energy put into a
direct radiating loudspeaker will not reach the air,
but will be converted to heat in the voice coil and
mechanical resistances in the unit. The problem
is worse at low frequencies, as here, the size of the
source will be small compared to a wavelength, and
it will merely push the medium away. At higher fre-
quencies, the radiation from the source will be in Figure 1: Throat impedance of a typical constant

the form of plane waves that do not spread out. directivity horn (- - -) and a tractrix horn (). [1]

The load, as seen from the driver, is at its highest,

and the system is as ecient as it can be.
heavily dependent on frequency, becoming sharper
If we could make the driver radiate plane waves in
and sharper as frequency increases. This prob-
its entire operating range, ecient operation would
lem can be solved by using multiple driving units
be secured at all frequencies. The driver would
and digital signal processing, but a far simpler and
work into a constant load, and if this load could
cheaper way to achieve predictable directivity con-
be made to match the impedances of the driver,
trol, is to use a horn. The walls of the horn will
resonances would be suppressed. This is because
restrict the spreading of the sound waves, so that
the driver is a mechanical lter, and it needs to be
sound can be focused into the areas where it is
terminated in its characteristic impedance, ideally
needed, and kept out of areas where it is unwanted.
a pure resistance. If the driver is allowed to radiate
Directivity control is most important in sound re-
plane waves, resistive loading is secured.
inforcement systems, where a large audience should
The easiest way to make the driver radiate plane have the same distribution of low and high frequen-
waves, is to connect it to a long, uniform tube. But cies, and where reverberation and reections can be
the end of the tube will still be small compared to a problem. In a studio or home environment, this
a wavelength at low frequencies. To avoid reec- is not as big a problem.
tions, the cross section of the tube must be large As the art and science of electroacoustics has de-
compared to a wavelength, at the same time it must veloped, the focus has changed from loading to di-
be small, to t the driver and present the required rectivity control. Most modern horns oer direc-
load. To solve this dilemma, we need to taper the tivity control at the expense of driver loading, of-
tube. When we do this, we can take radiation from ten presenting the driver with a load full of reso-
the driver in the form of plane waves, and trans- nances and reections. Fig.1 compares the throat
form the high pressure, low velocity vibrations at impedance of a typical constant directivity horn
the throat into low pressure, high velocity vibra- (dashed line) with the throat impedance of a trac-
tions that can eciently be radiated into the air. trix horn (solid line) [1]. The irregularities above
Depending on how the tube ares, it is possible to 8kHz come from higher order modes.
present a load to the driver that is constant over a
large frequency range.
3 Fundamental Theory

Directivity Control The directivity of a cone or Horn theory, as it has been developed, is based on
dome diaphragm is largely uncontrolled. It is dic- a series of assumptions and simplications, but the
tated by the dimensions of the diaphragm, and is resulting equations can still give useful information

2
about the behavior. We will later review the as- 1. Innitesimal amplitude: The sound pressure
sumptions, and discuss how well they hold up in amplitude is insignicant compared to the
practice. steady air pressure. This condition is easily
The problem of sound propagation in horns is satised for most speech and music, but in high
a complicated one, and has not yet been rigor- power sound reinforcement, the sound pressure
ously solved analytically. Initially, it is a three- at the throat of a horn can easily reach 150-
dimensional problem, but solving the wave equa- 170dB SPL. We will take a closer look at dis-
tion in 3D is very complicated in all but the most tortion in horns due to the non-linearity of air
elementary cases. The wave equation for three di- later, at present it is sucient to note that
mensions (in Cartesian coordinates) looks like this the distortion at home reproduction levels is
[2]: insignicant.

∂2φ ∂2φ ∂2φ ∂2φ

 
2. The medium is considered to be a uniform
− c2 + 2 + 2 =0 (1)
∂t2 ∂x2 ∂y ∂z uid. This is not the case with air, but is per-
missible at the levels (see 1) and frequencies
This equation describes how sound waves of very involved.
small (innitesimal) amplitudes behave in a three
dimensional medium. We shall not discuss this 3. Viscosity and friction are neglected. The equa-
equation, only note that it is not easily solved in tions involving these quantities are not easily
the case of horns. solved in the case of horns.
In 1919, Webster [3] presented a solution to
the problem by simplifying eq. 1 from a three- 4. No external forces, like gravity, act on the

dimensional to a one-dimensional problem. He did medium.

this by assuming that the sound energy was uni-

5. The motion is assumed irrotational.
formly distributed over a plane wave-front perpen-
dicular to the horn axis, and considering only mo-
6. The walls of the horn are perfectly rigid and
tion in the axial direction. The result of these sim-
smooth.
plications is the so-called Webster's Horn Equa-
tion, which can be solved for a large number of
7. The pressure is uniform over the wave-front.
cases:
Webster originally considered tubes of in-
nitesimal cross section, and in this case prop-
d2 φ d ln S dφ
+ − k2 φ = 0 (2) agation will be by plane waves. The horn equa-
dx dx dx tion does not require plane waves, as is of-
where ten assumed. But it requires the wave-front

2πf
to be a function of x alone. This in turn
k= c , the wave number or spatial frequency (ra- means that the center of curvature of the wave-
dians per meter),
fronts must not change. If this is the case,
the horn is said to admit one-parameter (1P)
φ is the velocity potential (see appendix for expla-
waves [6], and according to Putland [7], the
nation), and
only horns that admit 1P waves are the uni-

S is the cross-sectional area of the horn as a func- form tube, the parabolic horn with cylindrical

tion of x. wave-fronts, and the conical horn. For other

horns, we require the horn radius to be small
The derivation of eq. (2) is given in the appendix. compared to the wavelength.
This equation can be used to predict what is go-
ing on inside a horn, neglecting higher order ef- Since the horn equation is not able to predict the in-
fects, but it can't say anything about what is going terior and exterior sound eld for horns other than
on outside the horn, so it can't predict directivity. true 1P horns, it has been much criticized. It has
But let us now review the assumptions eq. (2) is however been shown [8, 9] that the approximation
based on [4, 5]: is not as bad as one would perhaps think in the rst

3
instance. Holland [10] has shown that the perfor-
mance of horns of arbitrary shape can be predicted
by considering the wave-front area expansion in-
stead of the physical cross-section of the horn. The
author has also developed software based on the
same principles, and has been able to predict the
throat impedance of horns with good accuracy.

4 Solutions

In this section the solution of eq. (2) for the most

interesting horns will be presented, and we will look
at the values for throat impedance for the dierent
types. This can be calculated by solving the horn
equation, but this will not be done in full mathe-
matical rigor in this article. Figure 2: Throat acoustical resistance, rA and re-

The solution of eq. (2) can, in a general way, be actance xA as a function of frequency for dierent

set up as a sum of two functions u and v: horn types.

φ = Au + Bv (3)
The parabolic horn is a true 1P horn if it is rect-
where A and B represent the outgoing (diverging) angular with two parallel sides, the two other sides
and reected (converging) wave, respectively, and expanding linearly, and the wave-fronts are concen-
u and v depend on the specic type of horn. For tric cylinders. It has an area expansion given as
the case of an innite horn, there is no reected S = St x. The expression for throat impedance is
wave, and B = 0. We will rst consider innite very complicated, and will not be given here.
horns, and present the solutions for the most com- The throat impedances for both the uniform pipe
mon types [11]. The solutions are given in terms and the parabolic horn are given in g. 2. We note
ρ0 c
of absolute acoustical impedance, that the pipe has the best, and the parabolic horn
St , the specic
throat impedance (impedance per unit area) can be the worst, loading performance of all horns shown.
found by multiplying by St , the throat area, and

4.2 Conical Horn

the normalized throat resistance can be found by
St
multiplying by
ρ0 c .
The conical horn is a true 1P-horn in spherical co-
4.1 The Pipe and Parabolic Horns ordinates. If we use spherical coordinates, the cross
sectional area of the spherical wave-front in an ax-
Both these horns are true 1P horns. The innite
isymmetric conical horn is S = Ω(x+x0 )2 where x0
pipe of uniform cross-section acts as a pure resis-
is the distance from the vertex to the throat, and
tance equal to
ρ0 c Ω is the solid angle of the cone. If we know the half
zA = . (4) angle θ (wall tangent angle) of the cone,
St
An innite, uniform pipe does not sound very Ω = 2π(1 − cos θ). (5)
useful. But a suitably damped, long pipe (plane
wave tube) closely approximates the resistive load In the case where we are interested in calculating
impedance of an innite pipe across a wide band the plane cross sectional area at a distance x from
of frequencies, and is very valuable for testing com- the throat,
pression drivers [12, 13]. It presents a constant fre-
 2
quency independent load, and as such acts like the x + x0
perfect horn.
S(x) = St . (6)
x0

4
 The throat impedance of an innite conical horn
is
k 2 x20 + jkx0
 
ρ0 c
zA = . (7)
St 1 + k 2 x20
We should note that eq. (7) is identical to the ex-
pression for the radiation impedance of a pulsating
sphere of radius x0 .
The throat resistance of the conical horn rises
slowly, see g. 2. At what frequency it reaches its
asymptotic value depends on the solid angle Ω, be-
ing lower for smaller solid angle. This means that
for good loading at low frequencies, the horn must
open up slowly. As we will see in section 5, a cer-
tain minimum mouth area is required to minimize
reections at the open end. This area is larger for Figure 3: Joined pipe segments.
horns intended for low frequency use (it depends on
the wavelength), which means that a conical horn
The throat impedance of an innite exponential
would need to be very long to provide satisfying
horn is
performance at low frequencies. As such, the coni-
cal expansion is not very useful in bass horns. In- r !
deed, the conical horn is not very useful at all in ap-
ρ0 c m2 m
zA = 1− 2 +j (9)
plications requiring good loading performance, but St 4k 2k
it has certain virtues in directivity control.
mc
When m = 2k or f =
4π , the throat resistance

4.3 Exponential Horn

becomes zero, and the horn is said to cut o. Be-
low this frequency, the throat impedance is entirely
reactive and is
Imagine we have two pipes of unequal cross sec-
tional areas S0 and S2 , joined by a third segment r !
of cross sectional area S1 , as in g. 3. At each
ρ0 c m m2
zA = j − −1 . (10)
of the discontinuities, there will be reections, and St 2k 4k 2
the total reection of a wave passing from S0 to S2
will be dependent on S1 . It can be shown that the The throat impedance of an exponential horn is

condition for least reection occurs when shown in g. 2. Above the cuto frequency, the
throat resistance rises quickly, and the horn starts
p
S1 = S0 S2 . (8) to load the driver at a much lower frequency than
the corresponding conical horn. In the case shown,
This means that S1 = S0 k and S2 = S1 k , thus the exponential horn throat resistance reaches 80%
S2 = S0 k 2 . Further expansion along this line gives if its nal value at 270Hz, while the conical horn
for the nth segment, Sn = S0 k n , given that each reaches the same value at about 1200Hz.
m
segment has the same length. If k = e , and n is An innite horn will not transmit anything below
replaced by x, we have the exponential horn, where cuto, but the matters are dierent with a nite
the cross sectional area of the wave-fronts is given horn, as we will see in section 4.6.
as S = St emx . If we assume plane wave-fronts, At this point it should be noted that for an ex-
this is also the cross sectional area of the horn at a ponential horn to be a real exponential horn, it is
distance x from the throat. the wave-front areas, not the cross sectional areas,
The exponential horn is not a true 1P-horn, so that should increase exponentially. Since the wave-
its performance can not be exactly predicted. But fronts are curved, as will be shown in section 6,
much information can be gained from the equa- the physical horn contour has to be corrected to
tions. account for this.

5
 The throat impedance of a hypex horn with T = 0.5
is shown in g. 2. The throat impedance of a family
of horns with T ranging from 0 to 5 is shown in g.
5.

Figure 4: A family of representative hypex con-

tours, T=0 (lower curve), 0.5, 1, 2, 5 and innite
(upper curve).

4.4 Hyperbolic Horns

The hyperbolic horns, also called hyperbolic-
exponential or hypex horns, were rst presented by
Salmon [14], and is a general family of horns given
by the wave-front area expansion

x x
S = St (cosh + T sinh )2 . (11)
x0 x0

T is a parameter that sets the shape of the horn,

see g. 4. For T = 1, the horn is an exponen-
tial horn, and for T → ∞ the horn becomes a
conical horn.

c
x0 is the reference distance given as x0 = 2πfc
where fc is the cuto frequency.

Figure 5: Throat impedance of a family of hypex

A representative selection of hypex contours is
horns.
shown in g. 4.
Above cuto, the throat impedance of an innite
hyperbolic horn is Exponential and hyperbolic horns have much

q  slower are close to the throat than the conical

1
ρ0 c  1 − µ2
T horn, and thus have much better low frequency
µ
zA = 2 + j 1−T 2
, (12) loading. When T < 1, the throat resistance of
St 1 − 1−T
µ2 1 − µ2
the hyperbolic horn rises faster to its asymptotic
√
value than the exponential, and for T < 2 it
and below cuto, the throat impedance is entirely rises above this value right above cuto. The range
reactive and is 0.5 < T < 1 is most useful when the purpose is to
 q
1 1
 improve loading. When T = 0, there is no reac-
ρ0 c  µ µ2 − 1  tance component above cuto for an innite horn,
zA = j 2 (13)
St 1 − 1−T but the large peak in the throat resistance may
µ2
cause peaks in the SPL response of a horn speaker.

where Due to the slower aring close to the throat,

horns with low values of T will also have somewhat
f
µ is the normalized frequency, µ= fc . higher distortion than horns with higher T values.

6
4.5 What is Cuto? stant. Here the transition from reactive to resistive
wave propagation happens at the same frequency
Both exponential and hyperbolic horns have a
throughout the entire horn. This frequency is the
property called cuto. Below this frequency, the
cuto frequency. There is no gradual transition, no
horn transmits nothing, and its throat impedance
frequency dependent change in propagation type,
is purely reactive. But what is it that happens at
and that's why the change is so abrupt.
this frequency? What separates the exponential
and hyperbolic horns from the conical horn that
does not have a cuto frequency? 4.6 Finite Horns
To explain this, we rst have to look at the dif-
For a nite horn, both parts of eq. (3) have to be
ference between plane and spherical waves [10]. A
considered. By solving the horn equation this way
plane wave propagating in a uniform tube will not
[3, 15], we get the following results for pressure and
have any expansion of the wave-front. The normal-
volume velocity at the ends of a horn:
ized acoustical impedance is uniform and equal to
unity through the entire tube. pm = apt + bUt (14)
A propagating spherical wave, on the other hand,
has an acoustical impedance that changes with fre- Um = f pt + gUt (15)
quency and distance from the source. At low fre-
quencies and small radii, the acoustical impedance
where p and U denote the pressure and volume ve-
locity respectively, and the subscripts denote the
is dominated by reactance. When kr = 1, i.e.
when
λ throat and mouth of the horn. We can now nd
the distance from the source is
2π , the reactive
the impedance at the throat of a horn, given that
and resistive parts of the impedance are equal, and
above this frequency, resistance dominates.
we know a, b, f and g:
The dierence between the two cases is that the gZm − b
air particles in the spherical wave move apart as Zt = (16)
a − f Zm
the wave propagates; the wave-front gets stretched.
This introduces reactance in the system, because where Zm is the terminating impedance at the
we get two components in the propagating wave: mouth.
the pressure that propagates outward, and the pres- The expressions for a, b, f and g are quite com-
sure that stretches the wave-front. The propagat- plicated, and are given by Stewart [15] for the
ing pressure is the same as in the non-expanding uniform tube, the conical and the exponential horn.
plane wave, and gives the resistive component of We see that the value of mouth impedance will
the impedance. The stretching pressure steals en- dictate what value of throat impedance we get. As
ergy from the propagating wave and stores it, intro- explained in section 2, there will usually be reec-
ducing a reactive component where no power is dis- tions at the mouth, and depending on the phase
sipated. We can say that below kr = 1, there is re- and magnitude of the reected wave, it may in-
actively dominated propagation, and above kr = 1 crease or decrease the throat impedance. A horn
there is resistive dominated propagation. with strong reections will have large variations in
If we apply this concept to the conical and ex- throat impedance.
ponential horns by looking at how the wave-fronts Reections also imply standing waves and res-
expand in these two horns, we will see why the cut- onance. To avoid this, it is important to termi-
o phenomenon occurs in the exponential horn. We nate the horn correctly, so that reections are min-
have to consider the are rate of the horn, which is imized. This will be discussed in the next section.
dened as (rate of change of wave-front area with It can be interesting to see the what eect the
distance) / (wave-front area). length has on the performance of a horn. Fig. 6
In a conical horn, the are rate changes through- shows the throat impedance of 75 Hz exponential
out the horn, and the point where propagation horns of dierent lengths, but the same mouth size.
changes from reactive to resistive, changes with fre- As the horn length increases, the throat resistance
quency throughout the horn. rises faster to a useful value, and the peaks in the
In an exponential horn, the are rate is con- throat impedance become more closely spaced.

7
 Figure 7: Finite exponential horn terminated by an
innite pipe.

Finite horns will transmit sound below their cut-

o frequency. This can be explained as follows: the
horn is an acoustical transformer, transforming the
high impedance at the throat to a low impedance
at the mouth. But this applies only above cuto.
Below cuto there is no transformer action, and the
horn only adds a mass reactance.
An innite exponential horn can be viewed as
a nite exponential horn terminated by an innite
one of with the same cuto. As we have seen, the
throat resistance of an innite exponential horn is
zero below cuto, and the throat resistance of the
nite horn will thus be zero. But if the impedance
present at the mouth has a non-zero resistance be-
low cuto, a resistance will be present at the throat.
This is illustrated in g. 7, where a small expo-
nential horn with a mouth three times larger than
its throat is terminated by an innite pipe (a pure
acoustical resistance). The acoustical resistance
present at the throat below cuto approaches that
of the pipe alone, one third of the value above cut-
o.
The same is true for any mouth termination. As
Figure 6: The eect of increasing the length of a 75 long as there is a resistive impedance present at the
Hz exponential horn with krm = 0.5. The lengths mouth below cuto, power can be drawn from the
are (top to bottom) 50, 100 and 200 cm. horn.

5 Termination of the Horn

We have briey mentioned that there can be reec-

tions from the mouth of a horn. The magnitude
of this reection depends on frequency and mouth

8
size.
Consider a wave of long wavelength [16]. While
it is progressing along a tube, it occupies a constant
volume, but when it leaves the tube, it expands into
an approximate hemispherical shape, see g.8. The
volume thus increases, the pressure falls, increasing
the velocity of air inside the tube, pulling it out.
We have created an impulse that travels backwards
from the end of the tube, a reection.

Figure 9: Waves with short wavelength at the end

of a tube. [16]

throat impedance decreases. When krm is in-

creased beyond 1, however, the ripple increases
again, as shown in g. 11. This led Keele to inves-
tigate what the optimum horn mouth size would be
[19]. For a horn terminated by a piston in an in-
nite bae, he found the optimum krm to be slightly
less than 1, the exact value depending on how close
to cuto the horn is to be operated. His ndings
Figure 8: Waves with long wavelength at the end were based on the plane wave assumption, which
of a tube. [16] we will see in section 6 does not hold in practice.
As a historical side note, Flanders also discovered
The matters are dierent at higher frequencies. increased mouth reections for krm larger than 1
The relative volume of a half-wavelength of sound for a plane wave exponential horn [18], in 1924.
is much smaller, and the resulting volume increase If the same horn is calculated assuming spheri-
is less, creating less reections, see g.9. cal waves, there is no obvious optimum mouth size.
We may ask what the optimum size of the end If we consider the throat impedance of two 100Hz
of the tube is, to minimize reection in a certain horns with krm = 0.93 and 1.4, assuming spher-
frequency range. This has been investigated since ical wave-fronts and the same mouth termination
the early 1920s [2, 5, 16, 17, 18], and has led to the as before, it can be seen that the ripple decreases,
general assumption that if the mouth circumference not increases, for higher values of krm . See g. 12.
of an exponential horn is at least one wavelength at The reason may be that the wave-front expansion
the cuto frequency of the horn, so that krm ≥ 1, of a horn where the cross section is calculated as
the reections will be negligible. rm is the radius S = St emx will have a are rate that decreases to-
of the mouth. wards the mouth. This is because the wave-fronts
The eect of dierent mouth sizes is shown in bulge, see g. 13. The areas of the curved wave-
g. 10, where the throat impedance of a 100Hz fronts are larger than those of the plane ones, and
exponential horn is shown. The throat impedance the distance between them is also larger. But the
is calculated assuming plane wave-fronts, and us- distance between successive curved wave-fronts in-
ing the impedance of a piston in an innite wall creases faster than their areas, so the outer parts
as termination. The mouth sizes correspond to of the horn will have a lower cuto. The required
krm = 0.23, 0.46, 0, 70 and 0, 93. It can be seen krm for optimum termination becomes larger, and
that for higher values of krm , the ripple in the it increases as the horn is made longer.

9
 Figure 11: Increasing ripple for over-sized mouth,
krm = 1.4.

The above considerations are valid for exponen-

tial horns. What then about hyperbolic and conical
horns? Hyperbolic horns with T <1 will approxi-
mate the exponential horn expansion a certain dis-
tance from the throat, and the mouth termination
conditions will be similar to those for an exponen-
tial horn. Conical horns show no sign of having
an optimum mouth size. As length and mouth
size increases, simulations show that the throat
impedance ripple steadily decreases, and the horn
approaches the characteristics of an innite horn.
As a conclusion, we may say that the mouth area
of a horn can hardly be made too large, but it can
easily be made too small. krm in the range 0.7-1
will usually give smooth response for bass horns,
while midrange and tweeter horns will benet from
values ≥ 1.
Another thing that needs to be considered, is
the termination at the throat. If there is a mis-
match between the driver and the horn, the re-
ected waves traveling from the mouth, will again
be reected when they reach the throat, creating
standing waves in the horn.

6 Curved Wave-fronts

By logical reasoning, the assumption that the wave-

fronts in a horn are plane cannot be true. If it was
so, the speed of sound along the horn walls would
have to be greater than the speed of sound along
the axis. This cannot be the case, and the result is
that the wave-front on the axis must gain on that at

10
Figure 10: Throat impedance of nite horns,
krm =(top to bottom) 0.23, 0.46, 0.70 and 0.93.
 Figure 14: Wave-fronts in an exponential horn at
120 Hz. [4]

the horn walls, so the wave-fronts will dene convex

surfaces. In a circular conical horn the wave-fronts
are spherical with centers at the apex. In the uni-
form tube, the wave-fronts are plane. In both cases
the wave-fronts cut the axis and the walls at right
angles. It is therefore logical to assume that this
will happen in other horn types too.

6.1 Wave-fronts in Horns

Figure 12: Throat impedance of horns with krm = In 1928, Hall did a detailed investigation of the

0.93 and 1.4, assuming spherical wave-fronts. sound eld inside horns [4, 20], showing how the
wave-fronts curve in an exponential and a conical
horn. The wave-fronts in a 120 Hz exponential horn
at the cuto frequency are shown in g. 14. We
can see that the wave-fronts are very nearly normal
to the walls.
At 800 Hz, the matters are dierent, g. 15. A
certain distance from the throat, the pressure wave-
fronts gets seriously disturbed. Hall attributes this
to reections at the outer rim of the mouth, that
are more powerful than at the center, since the dis-
continuity is greater. Another explanation [8] is
that higher order modes (see section 9) will distort
the shape of the amplitude wave-fronts. This is also
what is most evident in g. 15. In a aring horn,
higher order modes will not appear at the same fre-
quencies throughout the horn. Close to the throat,
where the radius is small, they will appear at fairly
Figure 13: Comparing plane and curved wave-fonts
high frequencies, but closer to the mouth, they will
in an exponential horn.
appear at lower frequencies.
Conical horns do not look any better than expo-

11
 Figure 17: Dimensions of the tractrix horn

nential horns. Hall also investigated a large coni-

cal horn, 183 cm long, throat diameter 2 cm, and
mouth diameter 76 cm. Simulations show that the
horn does not have signicant mouth reections, as
the throat impedance is close to that of an innite
horn. Still the amplitude wave-fronts are seriously
disturbed, even close to the throat, something that
Figure 15: Wave-fronts in an exponential horn at
does not happen in an exponential horn. See g.
800 Hz. [4]
16. We can see two nodal lines, each about half-
way between the horn wall and the axis. This is a
result of higher order modes, and can be predicted.
The frequency where Ra = X a , i.e. where
the throat acoustical resistance and reactance are
equal, is about 1kHz for this horn. This indicates
that higher order modes is a problem in conical
horns even below the frequency where it has useful
loading properties.

6.2 Tractrix Horn

The tractrix horn is a kind of horn generated by
revolution of the tractrix curve around the x-axis.
The equation for the tractrix curve is given as
p
rm + rm2 − r2 p
x 2 − r2
x = rm ln − rm x (17)
rx
where

rm is the mouth radius, usually taken as

λc c
2π = 2πfc , where fc is the horn cuto
frequency, and

Figure 16: Wave-fronts in a conical horn at 800 Hz.

rx is the radius of the horn at a distance
[4]
x from the horn mouth. See g. 17.

Because the radius (or cross-section) is not a func-

tion of x, as in most other horns, the tractrix con-
tour is not as straightforward to calculate, but it
should not pose any problems.

12
 Figure 19: Throat impedance of a tractrix horn.

Figure 18: Assumed wave-fronts in a tractrix horn.

The tractrix horn expands faster than the expo-

nential horn close to the mouth, as seen in g. 28.

The tractrix curve was rst employed for horn

use by P.G.A.H. Voigt, and patented in 1926
[21]. In more recent times it was popularized by Figure 20: Dimensions of a spherical wave horn.

Dinsdale [22], and most of all by Dr. Bruce

Edgar [23, 24]. The main assumption in the
sphere with radius equal to the mouth radius as
tractrix horn, is that the sound waves propagate
mouth termination, is shown in g. 19.
through the horn as spherical wave-fronts with con-
stant radius, rm , which also is tangent to the walls
at all times, see g. 18. 6.3 Spherical Wave Horn
For this requirement to hold, the wave-front must
be spherical at all frequencies, and the velocity of The spherical wave (or Kugelwellen) horn was in-

the sound must be constant throughout the horn. vented by Klanglm, the motion picture division
of Siemens, in the late 1940s [26, 27]. It is often
A theory of the tractrix horn was worked out by
mistaken for being the same as the tractrix horn.
Lambert [25]. The throat impedance of a horn
It's not. But it is built on similar assumptions;
was calculated using both a hemisphere and a pis-
that the wave-fronts are spherical with a constant
ton as radiation load, and the results compared to
radius. The wave-front area expansion is exponen-
measurements. It appeared that the wave-front at
tial.
the mouth was neither spherical nor plane. Also di-
To calculate the spherical wave horn contour,
rectivity measurements showed increased beaming
at higher frequencies. This means that the tractrix
rst decide a cut-o frequency fc and a throat ra-

horn does not present a hemispherical wavefront at

dius yo . See g. 20. The constant radius r0 is given
as
the mouth at all frequencies. It does come close c
at low frequencies, but so does almost every horn r0 = . (18)
πfc
type.
The height of the wave-front at the throat is
The throat impedance of a 100Hz tractrix horn,
assuming wave-fronts in the form of attened spher- q
ical caps and using the radiation impedance of a h 0 = r0 − r02 − y02 (19)

13
Figure 21: Assumed wave-fronts in spherical wave Figure 22: Spherical wave horn folding back.
horns.

The area of the curved wave-front at the throat is

S0 = 2πr0 h0 ,

and the area of the wave-front with height h is

in the same manner 2πr0 h. Thus for the area to
increase exponentially, h has to increase exponen-
tially:
h = h0 emx (20)

where x is the distance of the top of the wave-front

from the top of the throat wave-front and m =
4πfc
c . Now that we know the area of the wave-
Figure 23: Throat impedance of a spherical wave
front, we can nd the radius and the distance of
horn.
this radius from the origin.

S = 2πr0 h of a sphere with radius equal to the mouth radius

as mouth termination, is shown in g. 23. It can
r
S
y= − h2 (21) be seen that it is not very dierent from the throat
π impedance of a tractrix horn.
xh = x − h + h0 (22)

The assumed wave-fronts in a spherical wave

6.4 Le Cléac'h Horn
horn are shown in g. 21. Notice that the wave-
fronts are not assumed to be 90° on the horn walls. Jean-Michel Le Cléac'h presented a horn that does
Another property of the spherical wave horn is that not rely on an assumed wave-front shape. Rather,
it can fold back on itself, g. 22. This is contrary it follows a natural expansion. The principle is
to the tractrix horn, which is limited to 90° tangent shown in g. 24. The wave-front surface at the
angle. throat (F1) starts out plane (line 01). At the
The throat impedance of a 100Hz spherical wave point it reaches F2, the wave-front area has ex-
horn, assuming wave-fronts in the form of attened panded, and to account for this, a small triangular
spherical caps and using the radiation impedance element (or, really, a sector of a circle) b1 is added.

14
Figure 24: The principle of the Le Cléac'h expan-
sion.
Figure 25: Contour of a Le Cléac'h horn.

The wave-front expansion from b1 (line 34) con-

tinues in element a3, and an element b2 is added investigated several coordinate systems, and found
to account for further wave expansion at F3. The the oblate spheroidal coordinate system to admit
process is repeated, and the wave-front is shown to 1P waves. Putland [7] later showed that this was
become a curved surface, perpendicular to the axis not strictly the case. More work by Geddes [29]
and the walls, but without making any assumptions showed that the oblate spheroidal waveguide will
regarding the shape prior to the calculations. The act like a 1P horn for a restricted frequency range.
wave-fronts are equidistant to each other, and ap- Above a certain frequency dictated by throat radius
pear to take the shape of attened spherical caps. and horn angle, there will be higher order modes
The resulting contour of the horn is shown in g. that invalidate the 1P assumptions.
25. The contour of the oblate spheroidal waveguide

The wave-front expansion is according to the is shown in g. 26. It follows the coordinate sur-

Salmon family of hyperbolic horns. There is faces in the coordinate system used, but in ordinary

no simple expansion equation for the contour of Cartesian coordinates, the radius of the horn as a

the Le Cléac'h horn, but the it can be calcu- function of x is given as

lated with the help of spreadsheets available at q

http://ndaviden.club.fr/pavillon/lecleah.html r = rt2 + tan2 (θ0 )x2 (23)

6.5 Oblate Spheroidal Waveguide where

rt is the throat radius, and

This is a horn rst investigated by Freehafer [28],
later and independently by Geddes [6]. Geddes θ0 is half the coverage angle.
wanted to develop a horn suitable for directivity
control where the sound eld both inside and out- The throat acoustical impedance is not given as an
side the horn could be accurately predicted. To analytical function, it must be found by numerical
do this, the horn must be a true 1P-horn. Geddes integration. The throat impedance for a waveguide

15
Figure 26: Contour of the oblate spheroidal waveg-
uide.

Figure 28: Comparison of the exponential horn

with the tractrix and the Wilson modied expo-
nential horn. [22]

is required. For example, for a 1 driver and 60°

included angle (θ0 = 30), the cuto is about 862Hz.
The advantages of the OS waveguide are that it
oers improved loading over a conical horn of the
same coverage angle, and has about the same direc-
tional properties. It also oers a very smooth tran-
sition from plane to spherical wave-fronts, which is
a good thing, as most drivers produce plane wave-
fronts.
The greatest disadvantage of the OS waveguide
Figure 27: Normalized throat impedance of a 60°
is that it is not suitable for low frequency use. Bass
included angle innite oblate spheroidal waveguide.
and lower mid-range horns based on this horn type
will run into the same problems as conical horns:
the horns become very long and narrow for good
with a throat diameter of 35.7mm and θ0 = 30 is
loading.
shown in g. 27.
To sum up, the OS waveguide provides excellent
The OS waveguide does not have a sharp cuto
directivity control and fairly good loading at fre-
like the exponential or hyperbolic horns, but it will
quencies above about 1kHz.
be useful to be able to predict at what frequency
the throat impedance of the waveguide becomes too
low to be useful. If we set this frequency at the 6.6 Other Horns
point where the throat resistance is 0.2 times its
There are three other horn types assuming curved
asymptotic value [30], so that the meaning of the
wave-fronts that are worth mentioning: the West-
cuto frequency becomes similar to the meaning of
ern Electric horns, the Wilson modied exponen-
the term as used with exponential horns, we get
tial, and the Iwata horn. These horns have in com-

0.2c sin θ0 mon that they do not assume curved wave-fronts of

fc = . (24) constant radius.
π rt
The Western Electric type horn, described in
We see that the cuto of the waveguide depends [17], uses wave-fronts of constantly increasing ra-
on both the angle and the throat radius. For a dius, all being centered around a vertex a certain
low cuto, a larger throat and/or a smaller angle distance from the throat. See g. 29.

16
Figure 29: Wave-fronts in the Western Electric type
exponential horn. [17]

In the Wilson modied exponential horn [31], the

waves start out plane at the throat, and become
more and more spherical. The horn radius is cor-
rected in an iterative process based on the wall tan-
gent angle, and the contour lies inside that of the
plane-wave exponential horn, being a little longer
and with a slightly smaller mouth are tangent an-
gle. See g. 28. Unfortunately, the Wilson method
only corrects the wavefront areas, not the distance
between the successive wave-fronts.
There is not much information available about
the Iwata horn [32, 33]. What is available is a
drawing and dimensions, but no description of the
concept. It looks like a radial horn, and seems to
have cylindrical wave-fronts expanding in area like
√
a hypex-horn with T = 2. The ratio of height to
width increases linearly from throat to mouth.

7 Directivity Control

Control of directivity is an important aspect of horn

Figure 30: Contour of the Iwata horn. [32]
design. An exponential horn can provide the driver
with uniform loading, but at high frequencies, it
starts to beam. It will therefore have a coverage
angle that decreases with frequency, which is unde-
sirable in many circumstances. Often we want the
horn to radiate into a dened area, spilling as little
sound energy as possible in other areas. Many horn
types have been designed to achieve this.
To get the real picture of the directivity perfor-
mance of a horn, the polar plot for a series of fre-
quencies must be given. But sometimes we also
want to get an idea of how the coverage angle of

17
the horn varies with frequency, or we want to just
get a rough idea on how much amplication a horn
gives. This is the purpose of the directivity fac-
tor and the directivity index. They are dened as
follows [34]:

Directivity Factor (Q): The directivity factor

is the ratio of the intensity on a given axis
(usually the axis of maximum radiation) of the
horn (or other radiator) to the intensity that
would be produced at the same position by a
point source radiating the same power as the
horn.

Directivity Index (DI): The directivity index is

dened as: DI(f ) = 10 log10 Q(f ). It indi-
cates the number of dB increase in SPL at the
observation point when the horn is used com-
pared to a point source.

Figure 31: Radiation into a solid cone of space de-

Since intensity is watts per square meter, it is in-
versely proportional to area, and a simple ratio of
ned dened by angles α and β.
areas can be used [35]. Consider a sound source ra-
diating in all directions and observed at a distance
where
r. At this distance, the sound will ll a sphere of
radius r. Its area is 4πr2 . The ratio of the area fI is the intercept frequency in Hz where the horn
to the area covered by a perfect point source is 1, loses directivity control,
and thus Q = 1. If the sound source is radiating
into a hemisphere, the coverage area is cut in half, x is the size of the horn mouth in mm in the plane
but the same sound power is radiated, so the sound of coverage, and
power per square meter is doubled. Thus Q = 2.
θ is the desired coverage angle in degrees in that
If the hemisphere is cut in half, the area is 1/4 the
plane.
area covered by a point source, and Q = 4.
For a horn with coverage angles α and β as shown
A large horn is thus needed to control directivity
in g. 31, Q can be computed as
down to low frequencies.
180 Most methods of directivity control rely on simu-
Q= −1
. (25)
sin (sin α2 sin β2 ) lating a segment of a sphere. The dierent methods
will be listed in historical order.
Most constant directivity horns try to act as a

7.1 Multicellular Horns

segment of a sphere. A sphere will emit sound uni-
formly in all directions, and a segment of a sphere
will emit sound uniformly in the angle it denes, Dividing the horn into many conduits is an old idea.
provided its dimensions are large compared to the Both Hanna [36] and Slepian [37] have patented
wavelength [11]. But when the wavelength is com- multicellular designs, with the conduits extending
parable to the dimensions of the spherical segment, all the way back to the source. The source con-
the beam width narrows down to 4050% of its ini- sists either of multiple drivers, or one driver with
tial value. multiple outlets, so that each horn is driven form a
A spherical segment can control directivity down separate point on the diaphragm.
to a frequency given as The patent for the traditional multicellular horn
6 belongs to Edward C. Wente [38]. It was born
25 · 10
fI = (26) from the need to accurately control directivity, and
x·θ

18
 Figure 33: -6dB beam widths of Electro-Voice
model M253 2 by 5 cell horn. [42]

rowing as the ideal sphere segment, and second,

the polar pattern shows considerably ngering at
high frequencies. This may not be as serious as has
Figure 32: Multicellular horn. [38] been thought, however. The -6dB beam widths of
a typical multicellular horn is shown in g. 33. The
ngering at high frequencies is shown in g. 34.
at the same time providing the driver with proper The beam width of a multicellular horn with dif-
loading, and was created for use in the Bell Labs ex- ferent number of cells is shown in g. 35 [34]. The
periment of transmitting the sound of a symphonic narrowing in beam width where the dimensions of
orchestra from one concert hall to another [39]. the horn is comparable to the wavelength is evident.
A cut view of the multicellular horn, as patented
by Wente, is shown in g. 32. In this rst kind of
multicellular horn, the individual horns started al-
7.2 Radial Horns
most parallel at the throat, but later designs often
used straight horn cells to simplify manufacture of The radial or sectoral horn is a much simpler con-
these complex horns. As can be seen, the multicel- cept than the multicellular horn. Horizontal and
lular horn is a cluster of smaller exponential horns, vertical view of a radial horn is shown in g. 36.
each with a mouth small enough to avoid beaming The horizontal expansion is conical, and denes the
in a large frequency range, but together they form horizontal coverage angle of the horn. The vertical
a sector of a sphere large enough to control direc- expansion is designed to keep an exponential ex-
tivity down to fairly low frequencies. The cluster pansion of the wave-front, which is assumed to be
acts as one big horn at low frequencies. At higher curved in the horizontal plane. Directivity control
frequencies, the individual horns start to beam, but in the horizontal plane is fairly good, but has the
since they are distributed on an arc, coverage will same midrange narrowing as the multicellular horn.
still be quite uniform. In addition, the there is almost no directivity con-
The multicellular horn has two problems, how- trol in the vertical plane, and the beam width is
ever. First, it has the same lower midrange nar- constantly narrowing with increasing frequency.

19
Figure 34: High frequency ngering of EV M253
horn at 10kHz. [42]

Figure 36: Prole of a radial horn. [42]

Figure 35: Beam width of a multicellular horn con-

structed as shown in the insert. [34]

20
7.3 Reversed Flare Horns
The reversed are horns can be considered to be
a soft diraction horn, contrary to Manta-Ray
horns and other modern constant directivity de-
signs that rely on hard diraction for directivity
control. This class of horns was patented for di-
rectivity control by Sidney E. Levy and Abra-
ham B. Cohen at University Loudspeakers in the
early 1950s [40, 41]. The same geometry appeared
in many Western Electric horns back to the early
1920s, and is even mentioned in the claims in [17],
but the purpose does not seem to be that of direc-
tivity control.
The principle is illustrated for a horn with good
horizontal dispersion in g. 37. The wave is al-
lowed to expand in the vertical direction only, then
the direction of expansion is changed. The wave-
front expansion is restricted vertically, and is re-
leased horizontally. The result is, that the hori-
zontal pressure that builds up in the rst part of
the horn causes the wave-front to expand more as
it reaches the second part. That it is restricted
in the vertical plane helps further. Since the wave-
front expansion is to be exponential all the way, the
discontinuity at the are reversal point (where the
expansion changes direction) is small. In addition,
the change of curvature at the are reversal point
is made smoother in practical horns than what is
shown in the gure.

Figure 37: wave-front expansion in reversed are

7.4 CE-Horns horns. [41]

In the early 1970s, Keele, then working for Electro-

where
Voice, gave an answer to the problems associated
with multicellular and radial horns by introducing
r is the radius at the junction point,
a completely new class of horns that provided both
good loading for the driver and excellent directivity θ is the half angle of the cone with solid
Ω, θ = cos−1 1 − Ω


control [42]. angle
2π , and
The principle is based on joining of an exponen-
tial or hyperbolic throat segment for driver load- kc is the wave number at the cuto fre-
2πfc
ing with two conical mouth segments for directivity quency, kc = c .
control. The exponential and conical segments are
joined at a point where the conical horn of the cho- The problem of midrange narrowing was solved by

sen solid angle is an optimum termination for the having a more rapid are close to the mouth of the

exponential horn. This point is given by Keele as conical part of the horn. Good results were ob-

the point where the radius of the exponential horn tained by doubling the included angle in the last

is third of the conical part. This decreases the acous-

0.95 sin θ tical source size in the frequency range of midrange

r= (27)
narrowing, causing the beam width to widen, and
kc

21
 removing the narrowing. The result is a horn with
good directivity control down to the frequency dic-
tated by the mouth size.
For a horn with dierent horizontal and vertical
coverage angles, the width and height of the mouth
will not be equal. The aspect ratio of the mouth
will be given as

XH sin θ2H
R= = , (28)
XV sin θ2H
or, if θH and θV are limited to 120°,

θH
R≈ . (29)
θV
The lower frequency of directivity control will also
be dictated by the mouth aspect ratio. Substituting
eq. (26) into eq. (29) and solving for the ratio of
intercept frequencies, we get

 2
fIH θH
≈ . (30)
fIV θV
For a 40° by 20° (HV) horn, the vertical intercept
frequency will be four times higher than the hori-
zontal intercept frequency.

7.5 Manta-Ray Horns

The Altec Manta-Ray horn sought to solve the
problems of the CE horns, mainly the inability to
independently specify the horizontal and vertical
intercept frequencies [43]. To achieve directivity
down to a lower frequency in the vertical plane,
the vertical dimension of the mouth must be in-
creased. Since the dispersion angle is smaller, the
expansion has to start further back, behind where
Figure 38: Example of the Electro-Voice CE Con-
the horizontal expansion starts. The result is the
stant Directivity horns. This horn covers 40° by
unique geometry shown in g. 39 (although it's not
20°.[42]
so unique anymore).
At the point where the horizontal expansion
starts, the wave is diracted to ll the width of
the horn, and dispersion is controlled by the horn
walls.
The Manta-Ray horn incorporates the same
rapid mouth aring as the CE horns to avoid
midrange narrowing, but does not use radial expan-
sion of the walls. The reason for this was that radial
walls produced a waist-banding eect, where the
horn lost much energy out to the sides in the up-
per midrange. This eect cannot be seen in the

22
 Figure 39: The Manta-Ray geometry. [43]

polar plots for the CE horns, which suggests that

the waist-banding eect can be a result of the
Manta-Ray geometry, and not solely of radial wall
contours. Figure 40: Adiabatic curve for air.

7.6 New Methods 1808 that, generally, sound waves cannot be propa-
gated in air without change in form, resulting in the
Most newer constant directivity designs have been
generation of distortion, like harmonics and inter-
based on either the conical horn, some sort of radial
modulation products. The distortion is caused by
horn (including the JBL Biradial design), or dirac-
the inherent non-linearity of air. If equal positive
tion methods like the Manta-Ray design. The only
and negative increments of pressure are impressed
notable exception is the oblate spheroidal waveg-
on a mass of air, the changes in volume of that mass
uide introduced by Geddes, which has been treated
will not be equal. The volume change for positive
in section 6.5.
pressure will be less than that for the equal negative
The general trend in horns designed for directiv-
pressure [44]. An idea of the nature of the distor-
ity control, has been to regard the control of di-
tion can be had from the adiabatic curve for air, g.
rectivity the most important, as it is always pos-
40. The undisturbed pressure and specic volume
sible to correct the frequency response. A at fre-
1
quency response does not, however, grant a per- ρ P0 V0 . Deviation
of air ( ) is indicated in the point
from the tangent of the curve in this point will re-
fect impulse response, especially not in the pres-
sult in the generation of unwanted frequencies, the
ence of reections. Reected waves in the horn at
peak of the wave being stretched and the trough
the high levels in question, will also cause the re-
compressed.
sulting horn/driver combination to produce higher
The speed of sound is given as
distortion than necessary, because the driver is pre-
sented with a non-linear and resonant load. This is r
p
discussed in the next section. c= γ (31)
ρ

where
8 Distortion
γ is the adiabatic constant of air, γ = 1.403.
As mentioned, the horn equation is derived assum-
ing that the pressure variations are innitesimal. It can be seen that the speed of sound increases
For the intensities appearing at the throat of horns, with increasing pressure. So for the high pressure
this assumption does not hold. Poisson showed in at the peaks of the wave-front, the speed of sound

23
 p2 is the r.m.s. pressure of the second harmonic at
x,

p0 is the static pressure of air, and

m is the are rate of the exponential horn.

It can be seen that distortion increases with fre-

quency relative to the cuto frequency. This is eas-
ier to see in the simplication for an innite expo-
nential horn given by Beranek [34]:

fp
D2 [%] = 1.73 · 10−2 It (33)
Figure 41: Distorted waveform due to non-constant
fc
velocity of sound. [5]
where

It is the intensity at the throat, in watts per square

is higher than at the troughs. The result is that
meter.
as the wave propagates, the peaks will gain on the
troughs, altering the shape of the waveform, and
Holland et al [45] has investigated the distortion
introducing harmonics. This is shown in g. 41.
generated by horns both with the use of a com-
There are thus two kinds of distortion of a sound puter model, and by measurements. The model
wave: one because of the unequal alteration of vol- considered the harmonics required at the throat to
ume, and one because of the propagation itself. generate a pure sine wave at the mouth (backward
This last kind of distortion is most noticeable in modeling), and also took reections from the mouth
a plane wave and in waves that expand slowly, as into account. For a horn with a 400Hz cuto and 4
in horns, and has the property that distortion in- inch throat, and a mouth SPL of 150dB, the distri-
creases with the length propagated. Both kinds of bution of harmonics is shown in g. 42. The peak
distortion generate mainly a second harmonic com- at the cuto frequency is due to the very high level
ponent. required at the throat to generate the required SPL
Fortunately, as the horn expands, the pressure at the mouth. Fig. 43 shows the level of the har-
is reduced, and the propagation distortion reaches monics at the throat at 1kHz for a given SPL at the
an asymptotic value. This value can be found for mouth. Measurements showed that the prediction
the horn in question, considering how it expands. of the second harmonic level was quite accurate,
It will be higher for a horn that expands slowly but measured levels of the higher harmonics were
near the throat than for one that expands rapidly. higher than predicted. This was recognized as be-
For example, a hyperbolic-exponential horn with a ing due to nonlinearities in the driver.
low value for T will have higher distortion than a As can be seen from the results, the level of har-
conical horn. For an exponential horn, the pressure monics is quite low at the levels usually encountered
ratio of second harmonic to fundamental is given as in the home listening environment, but can be quite
[44] considerable in the case of high level public address
p2 γ + 1 p1t ω 1 − e−mx/2 and sound reinforcement systems.
= √ (32)
One point that should be mentioned is the impor-
p1 2 2 γp0 c m/2
tance of reducing the amount of reection to reduce
where distortion. At the high levels involved, the reected
wave from diraction slots or from the mouth, will
p1t is the r.m.s. pressure of the fundamental at the not combine with the forward propagating wave in
throat, a linear manner. The result will be higher distor-
tion, and a nonlinear load for the driver. A driver
p1 is the r.m.s. pressure of the fundamental at x, working into a nonlinear load will not perform at its

24
 Figure 44: Higher order modes in a conical horn.
[47]

best, but will produce higher distortion levels than

it would under optimum loading conditions [45].
Directivity of the horn also plays a role in the
total distortion performance [46]. If the horn does
not have constant directivity, the harmonics, since
they are higher in frequency, will be concentrated
towards the axis, while the fundamental spreads
Figure 42: Level of harmonics at the throat for
out more. This means that distortion will be higher
a sinusoidal mouth sound pressure level of 150dB,
on-axis than o-axis.
frequency sweep. [45]

9 Higher Order Modes

At low frequencies, wave transmission in most

horns can be considered one-dimensional (1P
waves). When the wavelength of sound becomes
comparable to the dimensions of the horn, however,
cross reections can occur. The mode of propaga-
tion changes from the simple fundamental mode to
what is called higher order modes. The behavior of
these modes can be predicted for the uniform pipe
and the conical horn [47, 48, 49], and it is found
that they have cuto frequencies below which they
do not occur. In 1925, Hoersch did a theoretical
study of higher order modes in a conical horn, and
calculated the equipressure contours for two kinds
of modes. The results are shown in g. 44, and
shows the equipressure contours including both the
radial and non-radial vibrations. The left part of
the gure shows a pattern that resembles what Hall
measured in a conical horn, g. 16. For a aring
horn like the exponential, however, the higher order

Figure 43: Level of harmonics at the throat for a modes will occur at dierent frequencies at dierent

1kHz sinusoidal wave at the mouth, level sweep. places in the horn [8].

[45] Higher order modes will also be generated by

rapid changes in are, like discontinuities, so the
slower and smoother the horn curvature changes,
the less the chance for generating higher order
modes.

25
 The eect of the higher order modes is to disturb sound waves. The change in the mass of the uid
the shape of the pressure wave-front, so that direc- in this volume is
tivity will be unpredictable in the range where the
∂(ρS)
modes occur. According to Geddes, they may also − dx
have a substantial impact on the perceived sound
∂t
quality of horns [50]. dx not changing with time.
The particle velocity of the uid moving along
the x-axis through the element is u, and the dif-
10 Closing Remarks ference in the mass of uid entering one plane and
leaving the other, is
In this article, I have tried to present both classical
and modern horn theory in a comprehensive way. A ∂(uS)
short article like this can never cover all aspects of
ρ dx
∂x
horns. But I hope it has provided useful informa-
tion about how horns work, maybe also shedding
this is ρ (the density of the medium) times the

light on some lesser known aspects and research.

change in volume velocity, uS , with x. Because
the uid is continuous, these two quantities must
Finally, I would like to thank Thomas Dunker
be equal, so
and David McBean for proof reading, discussion
and suggestions.
∂(uS) ∂(ρS)
ρ =−
∂x ∂t
Appendix We expand both sides to get

Derivation of the Webster Horn Equa-

   
∂S ∂u ∂S ∂ρ
ρ u +S =− ρ +S
tion ∂x ∂x ∂t ∂t

This derivation is based on the innitesimal ampli- Now we introduce the concept of velocity poten-

tude, one-parameter plane wave assumption from tial. It can be looked upon as similar to electric

the start, as is given in [5, ch. X] and discussed in potential along a resistive conductor. If this con-

section 3. ductor has a resistance R per unit length, the re-

sistance of a small length ∂x is R∂x. If a current
I ows through the conductor, the voltage drop
(electrical potential) across ∂x is

∂U = −RI∂x.

Setting R = 1, we have

∂U
I=− .
∂x
Acoustically, we may say that the velocity poten-
tial replaces U , and the particle velocity replaces I .
Thus u = − ∂φ
∂x . We also have the relation that
Figure 45: Plane wave propagation in a horn. ∂ρ ∂s ρ0 ∂ 2 φ
∂t = ρ0 ∂t = c2 ∂t2 , where s is the condensation
of the medium, and ρ0 is the static density of the
Consider a aring horn as shown in g. 45, where
medium. For innitesimal amplitudes, ρ = ρ0 , and
dx is the short axial length between two plane wave- ∂A
= 0, since the area at a given value of x is
∂t
fronts of area S . The volume of this element is Sdx,
independent of time. After substitution, we have
where S is given as an arbitrary function of x. Fluid
∂2φ 1 ∂2φ
 
(air) will ow into this element from one side, and 1 ∂A ∂φ
out of it on the other side, due to the passage of
+ − = 0.
∂x2 A ∂x ∂x c2 ∂t2

26
 1 ∂S ∂ ln S
Since
S ∂x = ∂x , we can write this as [10] K.R. Holland, F.J. Fahy and C.L. Morfey:
Prediction and Measurement of the One-
∂ 2 φ ∂ ln S ∂φ 1 ∂2φ parameter Behavior of Horns, JAES Vol. 39,
+ − = 0,
∂x2 ∂x ∂x c2 ∂t2 May 1991, pp. 315337

this is the fundamental horn equation for innitesi- [11] H.F. Olson: Elements of Acoustical Engineer-
mal amplitudes. If we have simple harmonic motion ing, 2nd edition, Van Nostrand, 1947
(a sine or cosine wave of a single frequency), we can
[12] AES Information Document: Plane-Wave
write φ = φ1 cos ωt, which gives ∂ 2 φ/∂t2 = −ω 2 φ,
Tubes: Design and Practice, JAES vol. 39,
ω = 2πf . By using this substitution, and remem-
No. 6. June 1991, pp. 474480
bering that k = ω/c, we get

[13] M. Buck: Plane Wave Tubes - Uses and Limi-

d2 φ d ln S dφ
+ − k 2 φ = 0, tations, AES Convention Paper no. 6268, Oct.
dx dx dx 2004

which is the most convenient form of the horn equa-

[14] V. Salmon: A New Family of Horns, JASA
tion.
vol. 17, Jan. 1946, pp. 212218

[15] G.W. Stewart and R.B. Lindsay: "Acoustics",

References Van Nostrand, 1930

[1] G.K. Behler and M. Makarski: Two-Port Rep- [16] C.R. Hanna and J. Slepian: The function and
resentation of the Connection between Horn design of Horns for Loudspeakers, Transac-
Driver and Horn, JAES vol. 51, No. 10, Oc- tions of the AIEE, 1924
tober 2003, pp. 883897
[17] H.C. Harrison: Improvements in Acoustic
[2] I.B. Crandall: Theory of Vibrating Systems Horns, British Patent no. 213 528, 1923/1925
and Sound, Van Nostrand, 1926
[18] P.B. Flanders: Improvements in Acoustic

[3] A.G. Webster: Acoustical Impedance and the Horns, British Patent no. 245 415, 1924/1926

theory of Horns and of the Phonograph, Proc.

[19] D.B. Keele: Optimum Horn Mouth Size,
N. Ac. of Sci., 1919, pp. 275282
AES Preprint no 933, 1973

[4] W.M. Hall: Comments on the Theory of

[20] W.M. Hall: An investigation of Sound Fields
Horns, JASA vol 3, 1932, pp. 552561
in Regions Restricted by Finite Boundaries,
MS Thesis, MIT 1928
[5] N.W. McLachlan: Loud Speakers, Oxford,
1934
[21] P.G.A.H. Voigt: Improvements in Horns for
Acoustic Instruments, British Patent no. 278
[6] E.R. Geddes: Acoustic Waveguide Theory,
098, 1926/1927
JAES Vol. 37, Jul/Aug 1989, pp. 554569

[22] J. Dinsdale: Horn Loudspeaker Design part

[7] G.R. Putland: Every One-parameter Acous-
1-3, Wireless World, March, May and June
tic eld Obeys Webster's Horn Equation,
1974
JAES Vol. 41, No. 6, June 1993, pp. 435451

[23] B. Edgar: The Tractrix Horn Contour,

[8] A.H. Benade and E.V. Jansson: On Plane
Speaker Builder no. 2, 1981, pp. 915
and Spherical Waves in Horns with Nonuni-
form Flare, Part 1, Acoustica vol. 31, 1974, [24] B. Edgar: The Edgar Midrange Horn,
pp. 7998 Speaker Builder no. 1, 1986, pp. 717

[9] J.T. Post and E.L. Hixon: A Modeling and [25] R.F. Lambert: Acoustical studies of the Trac-
Measurement Study of Acoustic Horns, dis- trix Horn part I and II, JAES Nov. 1954, pp.
sertation, 1994 10241033

27
[26] Klanglm GmbH.: Lautsprecher mit Ex- [42] D.B. Keele: What's so Sacred About Expo-
ponentialtrichter, Swiss Patent no. 279947, nential Horns?, AES preprint no. 1038, 1975
1948/1951
[43] C.A. Henricksen and M.S. Ureda: The
[27] H. Schmidt: Über eine neuartige Laut- Manta-Ray Horns, JAES Vol. 26, No. 9, Sept.
sprecherkombination, Funk und Ton no. 5, 1978, pp. 629-634
1950, pp. 226232
[44] A.L. Thuras, R.T. Jenkins, H.T. O'Neil: Ex-

[28] J.E. Freehafer: The Acoustical Impedance of traneous Frequencies Generated in Air Carry-

an Innite Hyperbolic Horn, JASA vol. 11, ing Intense Sound Waves, BSTJ Jan. 1935,

April 1940, pp. 467476 pp. 159172

[29] E.R. Geddes: Acoustic Waveguide Theory [45] K. Holland and C. L. Morfey: A Model of

Revisited, JAES Vol. 41, No. 6, June 1993, Nonlinear Propagation in Horns, JAES Vol.

pp.452461 44, No. 7/8, Jul/Aug 1996, pp. 569580

[30] D. Mapes-Riordan: Horn Modeling with Con- [46] T. Kikkawa, A. Yukiyoshi and N. Sakamoto:

ical and Cylindrical Transmission-Line Ele- A New Horn Loudspeaker Design yields low

ments, JAES Vol. 41, No. 6, June 1993, pp. Distortion and wide Dispersion, AES preprint

471484 no. 1151, 1976

[47] V.A. Hoersch: Non-radial Vibrations Within

[31] P. Wilson and G. Webb: Modern Gramo-
a Conical Horn, Physical Review, Feb. 1925,
phones and Electrical Reproducers, Cassell
pp. 218224
and Company Ltd, 1929

[48] P.M. Morse and K.U. Ingard: "Theoretical

[32] N. Iwata, Jap. patent no 54-59129, 1979
Acoustics", 1968

[33] J. Hiraga: Les Haut-Parleurs, 3rd edition,

[49] W.P. Mason: Electromechanical Transducers
Dunod 2000
and Wave Filters, Van Nostrand, 1942

[34] L. Beranek: Acoustics, McGraw-Hill, 1954

[50] L.W. Lee and E.R. Geddes: Audibility of Lin-

[35] Notes on Loudspeaker Directivity, Altec ear Distortion with Variations in Sound Pres-

Technical Letter no. 211 sure Level and Group Delay, AES Convention
Paper no. 6888, Oct. 2006
[36] C.R. Hanna: Multiple Inlet Horn, US Patent
no. 1 715 706, 1925/1929

[37] J. Slepian: Sound-Generating Device, US

Patent no. 1 684 975, 1926/1928

[38] E.C. Wente: Acoustic Device, US Patent no.

1 992 268, 1933/1935

[39] E.C. Wente and A.L. Thuras: Loud Speakers

and Microphones, BSTJ April 1934, p. 259
277

[40] S.E. Levy and A.B. Cohen: Acoustic Device,

US Patent no. 2 690 231, 1950/1954

[41] A.B. Cohen: Wide Angle Dispersion of High-

Frequency Sound, Audio Engineering Dec.
1952, pp. 2425, 5759

28