Sound & Vibration

2 Sound & Vibration

2.1 Introduction ......................................................................... 2

2.2 What is sound ....................................................................... 3
2.3 An introduction to acoustics .................................................... 6
The one-dimensional wave equation 6
2.4 Definitions............................................................................ 9
2.5 Sound fields ........................................................................13
The far field (plane waves) 13
Reflections 13
The near field 15
Very near field 16
Evanescent sound waves 19
Coincidence frequency 19
Standing waves 20
Diffuse sound field 21
Diffuse field; direct field and reverberation radius 22
Impulse like sounds 22
2.6 Representation of measurement results ...................................22
The Fast Fourier Transform 23
Autospectrum Sxx 23
The noise free cross spectrum Sxy 25
The noisy cross spectrum Sxy 26
Autospectrum versus cross spectrum 27
Transfer function Hxy 28
Coherence 28
An alternative method to calculate the Fourier response 29
Time-frequency representation 30
2.7 Reciprocity ..........................................................................31
Acoustic reciprocity 32
Vibro-acoustic reciprocity 33
2.8 Point sources .......................................................................34
2.9 The Helmholtz Integral Equation .............................................35
2.10 Sound & vibration sensors ...................................................36
Measuring sound pressure 37
Measuring sound pressure with a Microflown sensor 41
The Human ear 41
Measuring sound pressure gradient 42
Measuring particle velocity 45
Measuring sound intensity 51
Measuring vibration 53
2.11 References ........................................................................54

Fig. 2.1 (previous page): The sound absorbing wall of the large anechoic room of the
Danish Technical University. The ‘floor’ is a metal net with below similar absorbers as the
wall.

2.1 Introduction

The acoustic technical world is divided in two disciplines: sound and

vibrations. To put is simple: “sound” is what one can hear and “vibrations”
is what can feel if a structure vibrates. Vibrating objects usually produce
sound so the two disciplines have definitely some overlap. However the

2-2
 Sound & Vibration Sound & Vibration

measurement tools and theory behind the subjects is very different and this Pressure is defined as force per unit area. This force is caused by an
is probably why the acoustic world is divided in two disciplines. amount of air particles at a certain position. So variations in the (sound)
pressure is influenced by variations in the number of air particles. If the
This chapter starts with theory on sound. Three matters are discussed.
number of particles in a certain volume changes during time, there must be
What is sound; how can one detect sound and what acoustic measurements
a particle motion towards that volume and backwards. The velocity of the
can be performed? Definitions of acoustic quantities are given. Relations are
movement of these air particles is called particle velocity. Therefore sound
shown between sound pressure and particle velocity. Operation principles of
consists always of two parts: the compression and decompression of air,
sound pressure, pressure-gradient and particle velocity microphones are
sound pressure, and the movement of the air: the particle velocity.
presented. Measurements that require both sound pressure as particle
velocity like sound intensity, sound energy and acoustic impedance For the complete quantification of a sound field, both sound pressure as
measurements complete this part. particle velocity should be measured.
After this the subject ‘vibrations’ is considered. Vibrations are In an electrical analogy, sound pressure corresponds to an electric
movements in solids and these are usually measured with other type of voltage and particle velocity to electric current. The amount of power is
sensors than the sensors used for measuring sound. The two most common quantified in Watts. Multiplication of voltage and current results in electrical
ones are accelerometers and lasers. Accelerometers are sensors that must power and analogously, sound pressure times particle velocity is associated
be attached to the vibrating object and are sensitive for the acceleration of with sound power in the acoustical domain.
the object rather than the movement. A laser is a non-contact method to
In general the particle velocity is associated with the cause of an acoustic
measure the movement of the object.
event and the sound pressure is associated with the result. If for example a
loudspeaker is generating a sound wave in a room. Then the particle
2.2 What is sound velocity measured close to the loudspeaker quantifies the level of vibration
(the cause of the sound wave). The sound pressure measure in the room
Very simply, sound can be observed as a propagating vibration in any quantifies the amount of sound that is heard. If the walls of the room are
substance. This substance can be air, water, wood, or any other material: in hard reflecting, the measured sound pressure will be higher than if the walls
fact only in vacuum sound can not propagate. Here, sound is observed of the room were fully sound absorbing. The particle velocity measured
where it is perceived most: in the air. close to the loudspeaker is nit affected by the acoustic properties of the
walls.
Sound is produced by a source, for example a firecracker, it travels
through the air with a certain velocity (which is about 340m/s or 1225km/h) Sound pressure, affected by the number of air particles at a certain
and can be reflected perhaps a few times by a wall or the ground. An ear or position, does not have a direction; sound pressure is a scalar quantity.
microphone detects it. After perceiving the sound of the firecracker one can Contrary, the velocity of air particles does have a certain direction: particle
distinguish its origin. Besides that, it can be said to be a loud bang and velocity is a vector quantity.
perhaps to be an annoying experience. To find out the direction of propagation of a sound wave, sound pressure
This annoying experience, or the quantification “loud”, is a subjective and particle velocity must be measured simultaneously at one location. The
interpretation of sound. If large amounts of people are annoyed by a certain product of both provides information about the direction of the sound wave.
distress factor, regulations will follow. In this case sound measurements to The human ear detects sound pressure only. So how is it possible that we
quantify the amount of noise should be performed. Questions like “was it can find out where sound is coming from? This has to do with the shape of
really a loud bang or was it somebody who didn’t like firecrackers?” should the ear and with the fact that we have two ears. Actually this is an example
be answered in a proper way because of the subjectivity of perceiving of the most elementary array technology. A human head can be seen as
sound. two pressure microphones (Pleft&Pright) separated 20cm by a solid (the
The human ear is very sensitive for small pressure variations in the air. head), see Fig. 2.2 Here the sound source localisation will be explained very
Not all frequencies are in a similar way similar. The human ear is most briefly, in [14] theory is presented in more detail.
sensitive for pressure variations of about 3kHz (corresponding to 3000 With the use of both ears one can detect the sound level (difference) and
variations per second). For frequencies of 30Hz the ear is about 200 times one can detect the time difference between perceiving sound in one ear and
less sensitive. The static air pressure is about hundred thousand Pascal the other. If it is assumed that the sound source is located left forward (Fig.
(Pa), a pressure variation of one Pascal is perceived as very loud at 1kHz, 2.2A) and the frequency is not too high, the sound must travel longer to the
but as a soft tone at 20Hz, see also Fig. 2.22. right ear than to the left (Lleft<Lright). Since low frequency sound waves have
Sound can be understood as variations of the static air pressure and as a very long wavelengths, the head is considered a small obstacle. The sound
velocity of air particles: particle velocity. The following may explain this. wave tends to bend around small obstacles and therefore the intensity

2-3 2-4
 Sound & Vibration Sound & Vibration

difference between the two ears is little. By the use of this time delay (or 2.3 An introduction to acoustics
more specific, the phase shift) the source is localised.
Low Frequency High Frequency In this paragraph concepts like sound pressure, particle velocity and their
sound sound
relationship are described. For simple cases like plane waves, solving the
Lleft
Lright one-dimensional wave equation yields the relation between the sound
pressure and particle velocity. If the situation becomes more complex, like
Pleft Pright Pleft Pright Pleft Pright Pleft Pright
for instance sound fields near a sound source, the wave equation must be
solved three-dimensionally. Only the one-dimensional wave equation is
Lleft Lleft solved here.
Lright Lright
A B C D Once the sound pressure and the particle velocity are known, the product
and the ratio of both can be calculated. The product (associated with the
sound intensity) provides information on the amount of acoustic noise and
Fig. 2.2: Schematic head perceiving sound and sound waves, represented very basically. the ratio (the specific acoustic impedance) provides information on the
nature of the sound field and can for instance be used to determine the
Very simply explained, if the sound source is located left backward (Fig. reflection coefficient of acoustic absorbent materials.
2.2B) one detects the same phase shift: the shape of the ears makes the The sound intensity measurement, which is associated with the product
discrimination possible between forward and backward. (Head movements of sound pressure and particle velocity, is a main application of the
do have also something to do with the source localisation). Microflown. Nowadays this measurement is performed with a probe that
Since for higher frequencies the wavelength becomes shorter than the consists of two closely spaced pressure microphones, a p-p probe. The
separation of both ears and the phase shift method becomes invalid. sound pressure is determined with the sum of the two microphones; the
Because the head now is larger than the wavelength it becomes an obstacle particle velocity is calculated from their differential signal. The one-
for sound waves. In situation Fig. 2.2C the left ear will perceive more sound dimensional wave equation should be understood to comprehend this.
than the right ear: the sound source location is localised by sound level
difference. Again, if the sound source is located left backward (Fig. 2.2D) The one-dimensional wave equation
one detects the level difference. The shape of the ears is used again to
discriminate between forward and backward. As sound waves travel through matter, each molecule hits another and
returns to its original position. The result is that regions of the medium
As stated, the ear can detect sound pressure only and it is impossible to become alternately more dense, being condensations, and less dense,
perceive particle velocity. How about acoustic sensors? Well, there are three called rarefactions.
types of sensors. First, sound pressure sensors, like the ear. Second, sound
pressure gradient microphones; this type of acoustic sensor is sensitive for Now, observe a small volume (S·dx) of air, see Fig. 2.3. Inside the
the pressure difference over a distance (like the human head for low volume there is a certain number of air particles, which can be associated
frequencies: two pressure sensors separated by a certain distance). Third, with the air density. At place x particles enter the volume with a certain
there is only one type of particle velocity sensitive microphone: the velocity u(x,t) and particles leave the volume (at place x+dx) with a
Microflown. velocity u(x+dx,t): the sound wave is travelling in the x-direction only.

Sound pressure microphones detect pressure variations by deflection of a

very thin membrane due to the sound pressure difference inside and outside
the sensor. It functions similar to the ear. A pressure gradient microphone
(also called a velocity microphone) measures the difference in the sound
pressure at two spaced places. It has, in a limited bandwidth, a close
resemblance to the particle velocity.
To get a better understanding of the relation between the sound
pressure, sound pressure gradient and the particle velocity, the wave
equation will be examined.
Fig. 2.3: A volume of air through which an acoustic disturbance passes.

2-5 2-6
 Sound & Vibration Sound & Vibration

The amount of mass that enters the volume equals: S·ρ(x,t) u(x,t). The
amount that leaves the volume: S·ρ (x+dx,t) u(x+dx,t). Using ρ as the air
density. The net rate of mass inflow therefore equals:
∂
S ρ ( x, t )u ( x, t ) − S ρ ( x + dx, t )u ( x + dx, t ) = − S ( ρ ( x, t )u ( x, t ) ) dx (2.1)
∂x

The principle of mass conservation requires that this net inflow must be
balanced by an increase in mass of the element:
∂ d Fig. 2.4: A volume of air through which an acoustic disturbance passes.
−S ρ ( x, t )u ( x, t )dx = S ⋅ dx ⋅ ρ ( x, t ) (2.2)
∂x dt For the acceleration (the time derivative of the velocity) of the particles
in the volume element S⋅dx can be written:
The density inside the volume can be described with ρ(x,t)=ρ0+Δρ(x,t), a
constant density ρ0 plus a time varying part Δρ(x,t). The product of the d ∂u ( x, t ) ∂x ∂ ∂u ( x, t ) ∂
u ( x, t ) = + u ( x, t ) = u ( x , t ) + u ( x, t )
density and the particle velocity equals: ρ(x,t)⋅u(x,t)=u(x,t)ρ0+Δρ(x,t) dt ∂x ∂t ∂t ∂x ∂t (2.6)
u(x,t). If the density variation can be considered very small compared to
the static density, the latter equation can be simplified into: For a harmonic particle velocity u(x,t)⋅∂u(x,t)/∂x is much (u/c) smaller
ρ(x,t)⋅u(x,t)≈u(x,t)ρ0. The simplified version of Eq. (2.2) is known as the then ∂u(x,t)/∂t and therefore Eq. (2.6) will simplify in to:
linearised equation of mass conservation and is only valid if the variation in
du ( x, t ) ∂u ( x, t )
density is small compared to the density itself: ≈
dt ∂t (2.7)
∂ ∂
Δρ ( x, t ) + ρ 0 u ( x, t ) = 0
∂t ∂x (2.3) The mass of the volume is given by: S·dx·ρ0. The net force in the x-
direction equals: S⋅p(x,t)-S⋅p(x+dx,t). This net force equals the product of
The linearised equation of mass conservation can be used to calculate the mass times and acceleration of the volume:
sound pressure component of a sound wave if the particle velocity is
∂ ∂
measured at two closely spaced places. This will be clear when Eq. (2.3) is S ⋅ p( x, t ) − S ⋅ p( x + dx, t ) = − p( x, t ) ⋅ S ⋅ dx = S ⋅ dx ⋅ ρ 0 u ( x, t )
integrated with respect to time: ∂x ∂t
m
F a (2.8)
∂ u ( x + Δx , t ) − u ( x , t )
Δρ ( x, t ) = − ρ 0 ∫ ∂x u( x, t )dt ≈ − ρ ∫ 0
Δx
dt
(2.4) The latter can be simplified into:
For “normal” sound levels (less than 135dB) it is assumed that the ∂ ∂
ρ0 u ( x, t ) + p ( x, t ) = 0
increase in pressure and increase in density are linearly related by p(t,x)=c2 ∂t ∂x (2.9)
Δρ(t,x). And thus:
ρ0c 2 This relationship is known as the linearised equation of momentum
p ( x, t ) = −
Δx ∫ u ( x + Δx, t ) − u ( x, t )dt (2.5)
conservation and is only valid if the particle velocity u is small compared to
the velocity of sound c.
With the use of two closely spaced (spacing Δx) particle velocity probes Analogous to the calculation of sound pressure using the u-u method, the
(the separation distance must be much smaller than the wavelength) the linearised equation of momentum conservation can be applied to deduce an
(one-dimensional) sound pressure can be determined. (In fact 6 expression for the particle velocity (by use of two closely spaced pressure
Microflowns are needed to determine the sound pressure in a three microphones, called the p-p method):
dimensional sound field). 1 d 1
ρ 0 ∫ dx ∫ p( x + Δx, t ) − p( x, t )dt
u ( x, t ) = − p ( x, t )dt ≈ −
The same small volume (S⋅dx) of air is observed again, see Fig. 2.3. Δx ⋅ ρ 0 (2.10)
Now, for this volume the Newton’s second law, force equals mass times
acceleration, (F=ma) is solved. If the separation distance of the microphones is much smaller than the
wavelength, the particle velocity can be determined in one direction. This is
called the p-p method and will be dealt with in more detail at the end of this

2-7 2-8
 Sound & Vibration Sound & Vibration

chapter (section “Measuring sound intensity”) and in chapter 5: ‘sound In air acoustic wavelengths extend over a range from 17 millimetres
intensity’. (f=20kHz) to 17 meters (f=20Hz).
Differentiating Eq. (2.3) with respect to time and Eq. (2.9) with respect The specific acoustic impedance at a point in a sound field is defined as
to x yields: the ratio of the complex amplitude of an individual frequency component of
sound pressure at that point, to the complex amplitude of the associated
∂ 2 p( x, t ) ∂ 2 Δρ ( x, t ) component of particle velocity.
− =0
∂x 2 ∂t 2 (2.11) It is a complex number (not a vector), giving the magnitude and phase of
the ratio of the individual frequency components. It can be compared to the
The sound pressure and density variations are linearly related: (p(x,t)=c2 electrical resistance in the electric domain.
Δρ(x,t)). By substituting this in Eq. (2.11), the one-dimensional wave The acoustic impedance can be e.g. be used to determine the reflection
equation is achieved: coefficient of acoustic absorbing materials or to find out if a source is an
effective radiator.
∂ 2 p (t , x ) 1 ∂ 2 p (t , x )
− 2 =0
∂x 2 ∂t 2 The specific acoustic impedance is given by:
c (2.12)
p
Zs = [ Ns / m 3 ]
This wave equation must be satisfied by acoustic waves propagating in u (2.14)
the medium. However, in order to find solutions to this equation, the form
that the sound field must take must be specified in more detail. Boundary Note that the specific acoustic impedance is a complex figure and not a
conditions must be specified in order to find the form of the sound field. vector quantity.

2.4 Definitions The measurement of the specific acoustic impedance is used for example
to determine the sound absorbing characteristics of materials like glass wool
or foams. The reflection coefficient can be calculated from the impedance,
In acoustics a particle is defined as a volume of air that is small see further chapter 6: ‘acoustic impedance’.
compared to the dimensions of our measuring device or the wave length of
the acoustic wave and large compared to the molecular mean free path The characteristic impedance is the ratio of the effective sound pressure
(5·10-8m) [13]. The minimal wavelength in air being about 17 millimetres, at a given point to the effective particle velocity at that point in a free,
the measuring device can be for instance an eardrum (5 millimetre) or a plane, progressive sound wave. It is equal to the product of the density ρ,
Microflown (100 micrometer). A particle may be interpreted as a cube of air of the medium (air) times the speed of sound c in the medium
with dimensions of 1μm×1μm×1μm. (Z=ρc≈435Nsm-3). A plane wave is an acoustic disturbance in which the
pressure is uniform in the direction normal to the direction of propagation.
The sound pressure, p(t), is the incremental change from the static When measured far from the acoustic source in the free field one obtains a
pressure at a given instant caused by the presence of the sound wave. plane wave. In anechoic rooms free field conditions can be found.
Pressure can be associated with the number of air particles. The particle
velocity u(t), is the velocity of a given infinitesimal part of the medium (the The sound intensity in a specified direction at a point is the average rate
above mentioned particle) at a given instant due to a sound wave only. at which sound energy is transmitted through a unit area perpendicular to
Note that sound pressure is a scalar and particle velocity is a vector the specified direction at the point considered. It is thus a vector quantity;
quantity. defined as the time averaged product of the sound pressure (a scalar
quantity, p) and the corresponding particle velocity (a vector quantity, u) at
The human ear is sensitive for sound pressure in the frequency, f, range the same position.
from 20Hz up to 20kHz. Sound waves with frequencies higher than 20kHz
T
are called ultra sound and sound waves with frequencies lower than 20Hz 1
are called infra sound. Low frequency sound waves have frequencies lower
I [W / m 2 ] = lim
T →∞ ∫
T 0
u(t ) ⋅ p (t )dt
than 200Hz. (2.15)

The wavelength λ is given by the speed of propagation of the wave, c, Note: the sound intensity is not depending on time.
divided by the frequency of vibration:
Sound energy density E is defined as the sum of potential energy U and
c the kinetic energy K, both per unit volume. The potential energy U is
λ= [m]
f associated with the sound pressure and is given by:
(2.13)

2-9 2-10
 Sound & Vibration Sound & Vibration

p
2 The reference 20 micro Pascal is chosen because a human can just hear
U= this amount of sound at one kilohertz. In underwater acoustics a sound
2 ρ 0c 2 (2.16) pressure reference of 1μPa is used.

The kinetic energy density K is associated with the particle velocity and is The particle velocity level, also given in decibels, has a reference
given by: effective particle velocity of 50 nano-meter per second. The origin of this
reference is the reference pressure level divided by the characteristic
K = 12 ρ 0 u
2
impedance (20μPa/ρc≈46nm/s). So it is the particle velocity corresponding
(2.17)
to the reference sound pressure in a free, plane, progressive sound wave.
Since nowadays apart from the sound pressure, particle velocity can be It can be argued that the reference should be 20μPa/ρc≈46nm/s in stead
determined instantly, acoustic quantities as sound intensity, acoustic of 50nm/s. However the characteristic impedance (ρc) is (amongst others)
impedance and sound energy density can be measured three dimensional, temperature dependent so if the reference is chosen 20μPa/ρc, the
instantaneous and on one position in space. reference level would also be. This is rather inconvenient but once the
Sound intensity is associated with the product of sound pressure and reference is fixed on 50nm/s, sound pressure levels and particle velocity
particle velocity and quantifies the amount of sound energy that levels may not be the same in a free progressive wave.
propagates. The specific acoustic impedance is related with the ratio of So by definition the particle velocity level (PVL) equals:
both, and is a useful quantity to determine for example the reflection or
u
absorption of matter. Sound energy is associated with the sum of sound PVL = 20 Log10 (re. = 50nm ⋅ s −1 )[dB ]
pressure and particle velocity. u ref
(2.20)
Sound energy density quantifies how much energy is “stored” in an acoustic
wave, sound intensity quantifies how much sound energy is transported and The sound intensity level has a reference intensity of one pico-Watt.
specific acoustic impedance quantifies the possibility for sound energy to be
I
transported. SIL = 10 Log10 (re. = 1 pW )[dB ]
I ref
Sound intensity is the rate of which sound energy is transported. For a (2.21)
one dimensional case the following expression is valid:
The reference of 1pW is because 20μPa×50nm/s=1pW.
∂ ∂
E+ I =0 (2.18) Various values of the magnitudes mentioned quantities are given in Table
∂t ∂x 2.1. In this case it is assumed that the specific acoustic impedance equals
the characteristic impedance so that the values of PVL and SIL are equal to
For most purposes, the absolute numbers obtained from measurements
the SPL.
have little significance themselves; they are almost always compared to a
reference, and are usually quoted as: levels "re. …" that reference. As in
many physical sciences, the units and reference levels are derived from Table 2.1: Some acoustic levels
common observed phenomena. The zero-level of sound pressure, for
example, is not a physical zero (that is the absence of sound pressure);
rather, it is something of an average "threshold of hearing" at about 1kHz Description level [dB] p [Pa] u [m/s] I
for humans. The physical pressure associated with this level is very small [W/m2]
(20 micro Pascal). About one million times greater sound pressure tends to Artillery fire, airplane 140 200 0.45 100
cause considerable discomfort. Since the sound intensity generally varies
with the square of the sound pressure, this (audio) range represents a ratio Rock concert 120 20 0.045 1
of one trillion (1012) to one in energy! Reference level 94 1 2.5×10-3 2.5×10-3
The most known property is the sound pressure level (SPL) which is Vacuum cleaner 80 0.2 4.5×10 -4
1×10-4
given in decibels. It is defined as twenty times the logarithm to the base ten -5
Conversation, face-to-face 60 0.02 4.5×10 1×10-6
of the ratio of the measured sound pressure of this sound to a reference
effective sound pressure of 20 micro Pascal. Whispering 40 2×10 -3
4.5×10 -6
1×10-8
p Inside empty theatre 20 2×10-4 4.5×10-7 1×10-10
SPL = 20 Log10 (re. = 20μPa)[ dB ]
-5 -8
p ref .
(2.19) Threshold of hearing 0 2×10 4.5×10 1×10-12

2-11 2-12
 Sound & Vibration Sound & Vibration

the pressure signal has the same phase as the forward travelling wave and
2.5 Sound fields the reflected sound wave is reduced in amplitude due to damping.

M icro flo w n &

There are some specific sound fields that one should know to make S p ea ker
M icro p h o n e
acoustic problems easier to understand. Practical sound fields can be
considered as a combination of these fields. 7cm
As long as the acoustic phenomena are linear (i.e. the sound levels are 1m
below 135dB) separate sound fields do not interact. 40m

The far field (plane waves)

Fig. 2.6: The 40m long tube measurement set up is used to demonstrate the effect of
The far field is a sound field far from a sound source in an environment reflections.
where no reflections exist. The sound pressure and particle velocity are in
phase or out of phase and their ratio is given by the characteristic
impedance Z=p/u=ρc. Speaker input signal

The amplitude of the sound pressure and particle velocity is depending

inversely on the distance to the source. If the distance to the source
doubles, the amplitude of sound pressure and particle velocity will be half. Pressure signal

Only under free field conditions (far from the source where the sound
waves are plain and no reflections are present), both sound pressure and
particle velocity are completely in (A) or out phase (B) and the measured Reflections
sound levels will be of the same magnitude, see Fig. 2.5.

u u
u u u Microflown signal
u p
0 100 200 300 400
p p p p p C
Delay Time [ms]

u Fig. 2.7: A burst measurement in a 40 meters long tube.

u
sum sum p A short sine wave burst is not the signal that is most common. Normally
p signals are generated continually and the sound field will be a summation of
A B
the original signal with the reflections.
D
Compared to the original signal in general a reflection has: 1) an
Fig. 2.5: output of a microphone and Microflown for plane sound waves. opposite phase between the sound pressure and particle velocity; 2) the
sound pressure has the same phase; 3) the amplitudes of sound pressure
and particle velocity are reduced.
Reflections
As an example of the difference between sound pressure and particle
Reflections occur if in the path of the sound wave the acoustic impedance velocity, a burst of two sound waves are generated in a tube in opposite
alters. Normally reflections are caused due to the presence of a material but directions, see Fig. 2.8.
also temperature variation in the air temperature can cause a reflection.
As an example an acoustic burst is generated in a 40 meters long tube
with a loudspeaker at the left side and the sound is reflected at the right
side. As can be seen, the sound pressure and particle velocity are in phase
in the forward travelling wave. The reflected sound wave is out of phase,

2-13 2-14
 Sound & Vibration Sound & Vibration

Microflown & velocity are linear related, therefore they will both be a function of 1/r. The
Speaker Speaker radiation from the point source is spherically symmetric and the wave
Microphone
equation for spherically symmetric wave propagation is given by (see
appendix 3):
∂ 2 (r ⋅ p) 1 ∂ 2 (r ⋅ p)
− 2 =0
Fig. 2.8: Measurement set-up to demonstrate the difference between sound pressure and
∂r 2 c ∂t 2 (2.22)
particle velocity.
Since the function r⋅p should obey the wave equation Eq. (2.22), of which
One speaker generates a 150Hz burst and the other a 200Hz burst. The the solution is of the form Aei(kx±ωt), (k=ω/c known as the wave number) the
Microflown and microphone are situated in the middle measuring both pressure p should be a harmonic outwards-travelling wave proportional to
sound waves travelling in opposite directions (upper two curves in Fig. 2.5). 1/r:
As can be seen, the output of both microphones is different. If the two {
p (r , t ) = Re Ae i ( wt − kr ) ⋅ r −1 } (2.23)
signals are subtracted the result will be the leftwards-travelling wave and
the sum will result in the rightwards-travelling wave. This is because the
Substitution of Eq. (2.23) in Eq. (2.22) shows that this function of the
particle velocity has a phase shift compared to the sound pressure if the
sound pressure obeys the wave equation. The linearised equation of
direction is altered from forward to backward, see Fig. 2.6. If both
momentum conservation can be used to find the particle velocity that is
Microflown as microphone have the same sensitivity, the sum of both
associated with the sound pressure:
sensors will result in a uni-directional sensor assembly.
i (ωt − kr )
−1 d A ⎛1 ⎞e
ρ 0 ∫ dr
u (r , t ) = p(r , t )dt = ⎜ + ik ⎟
iωρ 0 ⎝ r ⎠ r (2.24)

The specific acoustic impedance as a function of the radial distance:

⎛ ikr ⎞
z (r ) = ρ0 c ⎜ ⎟
⎝ ikr + 1 ⎠ (2.25)

Eq. (2.25) shows that, if the distance to the sound source is small, for
spherical waves the specific acoustic impedance is proportional to k, and
thus for the frequency. In practice this means that relative close to the
source the acoustic impedance is lower than ρc and frequency dependent.
At lower frequencies the particle velocity level will be larger than the sound
pressure level. This is sometimes called the nearfield effect.

Very near field

In this paragraph theory is presented that predicts the region of the very
near field and how sound pressure and particle velocity behave in this
Fig. 2.9: Measured signals of a Microflown and a microphone; their sum and difference.
region [6], [7]. A sound wave of frequency f can be described by the
acoustic potential ϕ(r) obeying the Helmholtz equation:
The near field
Δϕ + k 2ϕ = 0 (2.26)
As an example the wave equation will be solved for a spherical sound
field radiated from a point source. One can observe the sound intensity to Here Δ is the Laplace operator, k=2π/λ=2πf/c is the wave number, λ is
find a suitable function of the sound field. Since the surface of the sphere the wavelength. To describe the sound field from some source, which is a
around the point source increases with the square of the radial distance the vibrating surface, this equation should be solved with the boundary
sound intensity, the sound power per unit area, decreases with 1/r2. The conditions:
sound power is a product of the sound pressure and the particle velocity.
This product is a function of 1/r2 and the sound pressure and the particle

2-15 2-16
 Sound & Vibration Sound & Vibration

∂ϕ harmonic in space function like sin ( 2π x / L ) and similar for y-direction. The
= un on the surface,
∂n first two terms in Eq. (2.33) then can be estimated as:
exp ( ik0 r )
ϕ∝ at infinity r → ∞, ⎛ ∂2 ∂2 ⎞ ⎛ 2π ⎞
2
⎛ 2π ⎞
2

⎜ 2 + 2 ⎟ ϕ0 ∼ ⎜ ⎟ ϕ0 , k ϕ0 = ⎜ ⎟ ϕ0
2
r (2.27)
⎝ ∂x ∂y ⎠ ⎝ L ⎠ ⎝ λ ⎠ (2.34)
where ∂/∂n is the derivative normal to the surface, un is the normal
component of the velocity, in general, a function of the point on the Typical situation in acoustic is that the vibrating body radiates the sound
vibrating surface. The observable acoustic values, sound pressure p and waves with the wavelength:
particle velocity u, are connected with the potential in the following way: λ>>L (2.35)
u = grad ϕ , p = −iωρϕ (2.28)
Then, according to (2.34), the second term in Eq. (2.26) is small in
comparison with the first one and in the vicinity of the body the Helmholtz
where ρ is the density of the background medium (air).
equation (2.26) is reduced to the Laplace equation:
Suppose that the vibrating surface can be considered as flat on some
Δϕ = 0 (2.36)
lateral length scale L. One can choose the x-y plane on the surface then the
z direction will coincide with the normal to the surface. In the direct vicinity
of the surface the potential can be expanded in the power series in z: Equation (2.36) is the equation of uncompressible fluid that means that
in this approximation the pressure level (dB) is negligible compared to the
ϕ ( x, y, z ) ≈ ϕ 0 ( x, y ) + ϕ1 ( x, y ) z + ϕ 2 ( x, y ) z 2 + O ( z 3 )
1 velocity level. The normal velocity in this range coincides with velocity of
2 (2.29) the vibrating surface.
Physically it means that nearby the surface the fluid can be considered as
The functions ϕ0,1,2 are connected with the velocities on the surface. From
uncompressible and for this reason the sound pressure level (dB) is small
first the boundary condition (2.27) it follows that:
compared to the particle velocity level.
ϕ1 ( x, y ) = u z ( x, y, 0) ≡ u z The latter has to be clearly explained. The sound waves are the
(2.30)
compression-decompression waves and if Eq. (2.36) would be exact there
It means that measured particle velocity very close the surface coincides would be no sound pressure at all. In reality this equation is approximate
with the surface velocity (this is verified by measurements, see below). and the pressure still finite but it is suppressed.
Taking the second derivative with z from Eq. (2.29), one finds: It shows that the sound pressure level is suppressed in comparison with
the particle velocity level by the factor L/λ (the ratio of the spatial
⎛ ∂u ⎞
S

ϕ 2 ( x, y ) = ⎜ z ⎟ wavelength L of the source and the wavelength) and the phase between the
⎝ ∂z ⎠ (2.31) particle velocity and sound pressure is shifted to 90 degrees.
Eq. (2.36) is true if the normal distance rn=z to the vibrating surface is
Therefore, the expansion Eq. (2.29) can be written as:
small in comparison with the size L and that the wavelength λ is larger than
1 ⎛ ∂u ⎞
S the vibrating surface L.
ϕ ( x, y, z ) ≈ ϕ0 + u zS z + ⎜ z ⎟ z 2
2 ⎝ ∂z ⎠ (2.32)
L λ c
rn << << =
2π 2π 2π f (2.37)
Let us insert now the potential (2.32) in the Helmholtz equation (2.26)
and take the limit z→0. We will find: The condition (2.37) can be named by the condition of the very near
field. In this range the normal component of the velocity coincides with
⎛ ∂ ∂ ⎞ ⎛ ∂u z ⎞
2 2 S

⎜ 2 + 2 ⎟ ϕ 0 + k ϕ0 + ⎜ that for the vibrating surface but there are no restrictions on the lateral
⎟ =0 (2.33)
2

⎝ ∂x ∂y ⎠ ⎝ ∂z ⎠ components of the velocity, which can be distributed in some way along the
surface. In the table below the properties summarized.
Suppose that the surface vibration can be described by some spatial
wavelength L. It means that the surface vibrations can be represented by a

2-17 2-18
 Sound & Vibration Sound & Vibration

i.e. the sound energy of the sound wave is transferred into the air behind
Region Condition u [m/s] p [Pa] Phase [deg]
the plate.
L λ u(rn) ≈ constant p(rn) ≈ constant As can be seen a plate that is normally reflecting the sound waves will be
Very near field rn << << 80-90
‘acoustic transparent’ at the coincidence frequency.
2π 2π U(f) = constant p(f) ~ f
L λ This effect is noticed best for lateral sound waves (so sound waves not
Near field << rn << u(rn)~r-2 P(rn)~r-1 80-10
2π 2π normal to the plate).
λ
far field rn >> u(rn)~r-1 p(rn)~r-1 0-10 Standing waves
2π
A pure standing wave can be generated in for instance a standing wave
tube. In such tube, a rigidly terminated tube with rigged side walls, the
Where rn is the normal distance to the source, L is the typical size of the sound wave can only travel in one dimension and all the sound is reflected
source and λ is the wavelength of the sound wave. at the end of the tube.
See further chapter 7: ‘Vibration measurements’. If a p-u probe is positioned in the tube at a distance (l-x) to the end of
the rigid end the sound pressure and the particle velocity can be calculated,
Evanescent sound waves see Fig. 2.10.
If the wavelength of the bending waves in the plate is shorter than the u,p probe
acoustic wavelength, then the acoustic waves do not propagate into the
u(x)
fluid. In such cases 'evanescent' waves occur where the pressure amplitude
and the normal particle velocity decreases exponentially with distance from
x=0 x=l
the plate. These sound waves carry no energy.
l-x
The particle velocity induced by a vibrating panel when the travelling
wave speed in the panel is less than the wave speed in the fluid. The Fig. 2.10: A standing wave tube that is rigidly terminated at x=l and in which the fluid is
amplitude of the particle motion decreases exponentially with increasing driven by a loudspeaker at x=0.
distance from the boundary. The particles move in an ellipse.
For harmonic excitation, one solution that will satisfy the wave equation
This effect can be understood in the following way. If the wavelength in
is given by the superposition of the complex pressures associated with a
air is longer than the bending waves in the plate, the waves in the plate
positive travelling and a negative travelling plane wave. In general the
create an infinite number of acoustic shortcuts where the air particles just
pressure is given by:
are ‘pumped around’. If the wavelength in air is shorter than the bending
waves in the plate the air particles that are ‘pushed forward’ can not reach p ( x) = Ae −ikx + Be ikx (2.38)
the part of the plate where the particles are ‘sucked back’. In this case an
acoustic wave can be created that propagates. Where A and B are arbitrary complex numbers which represent the
amplitude and relative phases of waves travelling in the positive and
Coincidence frequency negative x direction respectively, k is known as the wave number and
equals k=ω/c (ω=2πf using f as the frequency).
The speed of sound in air is almost constant but the speed of sound in
(sheet) materials varies with frequency. Because of this, there is always a The particle velocity associated with this pressure wave can be found
frequency where the wavelength in the air is the same as the wavelength in using the linearised momentum equation (∂p(x)/∂x=-iωρu(x) ).
the material. This is called the coincidence frequency.
Ae −ikx Be ikx
If sound waves travel over a plate at the coincidence frequency it is very u ( x) = −
ρc ρc (2.39)
easy for the plate to start vibrating in at the same frequency. If this
happened, the sound wave is ‘coupled in’, i.e. the sound energy of the
The values of the complex constants A and B are solved knowing that the
sound wave in the air is transferred into the plate.
particle velocity is zero at x=l: Ae − ikl = Beikl and at the left hand side of the
Once the sound wave is transferred into the plate and the plate is now tube the particle velocity is given by the movement of the piston u(0)=U:
vibrating at the coincidence frequency, at the other side of the plate the air ρcU = A − B . The sound pressure and particle velocity at any place in the
will vibrate with the same wavelength and the sound wave is ‘coupled out’,
tube are given by:

2-19 2-20
 Sound & Vibration Sound & Vibration

cos(k (l − x)) sin( k (l − x)) frequency is changed it will change in an unpredictable manner. In a diffuse
p ( x) = −iρcU u ( x) = U
sin( kl ) sin( kl ) field the sound pressure level and the particle velocity level do not vary in
(2.40)
space, in a standing wave the sound pressure and particle velocity do vary
The specific acoustic impedance at place x in the tube is given by: much.

p ( x) In a standing wave the reactive intensity is high and in a diffuse sound

Zs = = −i ⋅ ρc ⋅ cot( k (l − x)) field the reactive intensity zero. In a diffuse sound field both active and
u ( x) (2.41) reactive intensity are zero.

As can be seen, the specific acoustic impedance in a closed tube does not A diffuse sound field can be found in a reverberant room; a large room
equal the characteristic impedance but it is a co-tangent function that varies that has hard sound reflecting walls that are not parallel or perpendicular to
from plus to minus infinity and that in a standing wave tube particle velocity each other.
and sound pressure are 90 degrees out of phase.
Diffuse field; direct field and reverberation radius
The sound energy has a value in the tube but the sound intensity is zero.
This makes sense because in the tube no sound is propagating: all the In a reverberant room the sound pressure level is the same at all
sound that is put in the tube is reflected. This follows also from the theory positions in the room, the same applies for the particle velocity level. One
that is presented in the following paragraph. exception is at the position of the loudspeaker and at a certain radius
As shown in Eq. (2.47) the time averaged sound intensity is defined as: around it. Here the sound levels are higher than the sound levels at other
positions in the room. The point from where the sound level is not increased
1
T
anymore is called the reverberation radius. Within the reverberation radius
I = lim
T →∞ 2T ∫ p(t )u(t )dt = 1
2
ˆ ˆ cos(ϕ )
pu the sound field of the loudspeaker is dominant and outside the
−T (2.42) reverberation radius the diffuse sound field is higher than the direct field.
The phase shift in the tube is 90 degrees and therefore the time With a sound intensity probe only the direct field is measured, so the
averaged sound intensity is zero. direct field can be determined outside the reverberation radius. This is
because intensity of a diffuse sound field is zero.
In contrast to the (active) sound intensity, the reactive intensity is not
zero. The reactive intensity is calculated as: Impulse like sounds
ˆ ˆ sin(ϕ )
J = 12 pu (2.43) An impulse like sound source, i.e. a short event that occurs once, is
difficult to interpret in the frequency domain. This is (a.o.) because the Fast
As can be expected, if the phase shift is 90 degrees, the reactive Fourier Transform (FFT) works only for stationary signals. If a sound field is
intensity is large and the active intensity is low. not stationary (so its statistical properties do vary with time) the FFT does
The reactive intensity alone is not of much relevance but when combined not provide a meaningful answer.
with the active intensity, it describes the sound field. A high value of the If a single impulse like signal is generated, like a gunshot, it is possible to
ratio of reactive to active intensity shows that we are in a near field where analyse the signal in the frequency domain. Signals like squeak and rattle
the pressure and the particle velocity are in quadrature. This knowledge is noise in a car are short and distributed randomly in time and therefore
useful because sound measurements are affected by such conditions. impossible to analyse with FFT algorithms.
The ratio of reactive and active intensity is a very important figure to find
out the quality of the pu intensity measurement. 2.6 Representation of measurement results

Diffuse sound field

Although signals are recorded as time signals, the representation of
(stationary signals) in the frequency domain provides more insight for
In a pure diffuse sound field sound is propagating in all directions with
acoustic problems in most cases. The transformation from time domain to
the same magnitude. Similar as in a standing wave tube, in a pure diffuse
the frequency is called Fourier transform. FFT, Fast Fourier Transformation
sound field the sound energy has a value and the intensity is zero: the
is a method that is used in computer programs to calculate the
sound energy is not propagating.
transformation. The terminology is explained here intuitively, without too
The difference between a standing wave and a diffuse sound field is that much mathematics.
in a pure standing wave the phase shift is 90 degrees and in a diffuse sound
field the phase shift is random. It takes all values, and when the position or

2-21 2-22
 Sound & Vibration Sound & Vibration

The Fast Fourier Transform ‘noise’ that was generated in the 5 minutes of music. Now one knows how
much noise was generated at each frequency component but with this
In a lot of cases sound fields are studied in the frequency domain. representation it impossible to ever listen to the music anymore.
Normally the signals are captured in the time domain and transformed with
A sine wave function in the time domain will result in a single point
the Fast Fourier Transform (FFT). The technique is not explained here in
output in the frequency domain. As an example I whistled (with
detail but in very brief steps it goes like this: the time signal is divided in
approximately 1kHz) in a microphone during a flight in an aeroplane. In the
multiple pieces of equal length.
left of Fig. 2.11 the time representation is shown, and right the same signal
Each part (or block) is windowed in a certain way (there are many ways). is shown in the frequency domain.
This windowing means that the block of time data is multiplied with a
function that starts and ends with a zero value so that the beginning and 0,8

Output [V]
-10

10 log(output) [V ]
2
the end of the block will be zero. 0,6
-20

Each block is then transformed to the frequency domain. After the 0,4
-30
transformation of each windowed block the results are averaged. 0,2
-40
Due to the nature of the FFT process (the signal is divided in pieces, 0,0

windowed transformed and then averaged), the method only works for -0,2
-50

stationary signals. -60

-0,4
Stationary signals are constant in their statistical parameters over time. -70
If you look at a stationary signal for a few moments and then wait an hour
-0,6 100 1k 10k
0 5 10 15 20
Time [ms] Frequency [Hz]
and look at it again, it would look essentially the same, i.e. its overall level
would be about the same and its amplitude distribution and standard
deviation would be about the same. Rotating machinery generally produces
stationary vibration signals. Fig. 2.11: Left: a time signal; Right the auto spectrum of that signal.

So in short: acoustic and vibration phenomena are normally studied in The mathematical equation behind this is given by the Fourier
the frequency domain. The FFT algorithm that is normally used dictates that transformation:
the signals are stationary.
∞

∫ x(t )e
− ift
The first question in analysing signals is if the signal is stationary. If so it X(f ) = dt
can be studied in the frequency domain. −∞ (2.44)
An example of data that is represented in the time domain is a music CD.
When looking closer the Fourier transform can be understood. The
Music is stored on a cd and when it is put in a cd-player the data can be
integral is a time integrator and can be seen as a low pass filter in the
retrieved and listened to.
frequency domain. So if the signal x is given by: x(t ) = xˆ cos(ωt + ϕ ) the
For measurement purposes one is often more interested in a Fourier transformation works out as:
representation in the frequency domain. If the time data is transformed in
∞
to the frequency domain one is able to find out how much sound was
∫ xˆ cos(ωt + ϕ )e
− iωa t
recorded in a certain bandwidth (frequency span). So if a music cd is
X (ωa ) = dt
−∞
transferred into the frequency domain one is able to see how much ‘noise’ ∞
was made at lower frequencies (if for example a contrabass was playing) or
at high frequencies (if for example a violin or flute was playing).
= ∫ xˆ cos(ωt + ϕ ) [cos(ω t ) + i sin(ω t )] dt
−∞
a a

∞
(2.45)
Autospectrum Sxx xˆ
= ∫ cos ( (ωa − ω )t + ϕ ) + cos ( (ωa + ω )t + ϕ ) dt +
2 −∞
The autospectrum is the representation of a signal “x” in the frequency ∞
ixˆ
domain. It is often an average value, so it is possible to transform a signal
from the time domain into the frequency domain but not vice versa. 2 −∞∫ sin ( (ωa − ω )t + ϕ ) + sin ( (ωa + ω )t + ϕ ) dt
It is possible to calculate the autospectrum of 5 minutes music. This
results in a frequency representation (the frequency on the horizontal axis
and the signal strength on the vertical axis) that represents the average

2-23 2-24
 Sound & Vibration Sound & Vibration

Only if the center frequency ωa equals the signal frequency ω, the cosine As shown in Eq. (2.47) this time averaged intensity is a DC (non-time
and the sine with the difference term (ωat-ωt) becomes time independent. dependent) value. If both pressure as velocity have a certain frequency, the
All the time dependent cosine and sine functions are averaged to zero. non-averaged output equals a DC value depending on the phase shift, plus
a signal of double frequency. This double frequency part of the equation is
∞
averaged to zero.
∫ x(t )e
− iωt
X (ωa ) = dt
−∞ If the signals are 100% correlated (so they have no noisy, uncorrelated
xˆ
∞
ixˆ
∞
(2.46) components), the cross spectrum can be rewritten in the autospectra of
if ωa =ω 2 ∫
= {cos(ϕ ) + cos(2ωt + ϕ )} dt + ∫ {sin(ϕ ) + sin(2ωt + ϕ )} dt both signals and the phase shift between them:
−∞
2 −∞

=
xˆ ixˆ xˆ
cos(ϕ ) + sin(ϕ ) = eiϕ S xy = S xx S yy (cos ϕ + i sin ϕ )
2 2 2 (2.48)

In the frequency domain the (active) intensity is defined as the real part
So the Fourier transform is able to extract the frequency components of a
of the cross spectrum of sound pressure and particle velocity. If the sound
given time signal.
pressure and the particle velocity are free of noise, the cross spectrum can
be expressed as the product of the autospectra:
The noise free cross spectrum Sxy
I = Re ( S pu ) = Re ( )
S pp Suu (cos ϕ pu + i sin ϕ pu ) = S pp Suu cos ϕ pu
(2.49)
The cross spectrum is the frequency representation of what in the time
domain would be related to the product of two signals “x” and “y”.
If both pressure p (t ) = pˆ cos(ωt ) as velocity u (t ) = uˆ cos(ωt + ϕ pu ) have a
To calculate a cross spectrum one needs two signals. Not only the
certain frequency ω the cross spectrum is represented as a single value
amplitude of the signal is required but also the phase shift between the
signals is necessary. The amplitude of a signal is often well understood but pˆ uˆ at the (single) frequency ω and not at the double frequency.
the phase is more difficult to comprehend.
The noisy cross spectrum Sxy
If the example of the music-cd is taken again; one can calculate the
cross spectrum of a stereo signal. If a singer is recorded with two If the pressure p (t ) = pˆ cos(ωt ) and velocity u (t ) = uˆ cos(ω t + ϕ ) signals are
microphones and the distance to the left microphone is a little less than the
not fully correlated, an additional noise component is to be taken into
right one, the sound will be recorded with the left microphone a little earlier
account. If two noisy signals from a pressure microphone and a velocity
than with the right one. Due to this, the amplitude of both channels will be
microphone are observed:
approximately of the same magnitude but due to the time difference a
phase shift between the left and right signal will occur. A time shift in the p(t ) = pˆ s cos(ω t ) + pˆ n cos(ω t + θ (t )) = PS + PN
time domain will produce a phase shift in the frequency domain.
u (t ) = uˆs cos(ω t ) + uˆn cos(ω t + ϑ (t )) = U S + U N (2.50)
The cross spectrum of a stereo recording is something that has no much
meaning. A more meaningful use is the determination of the sound
The additional noise components PN and UN have constant amplitude and
intensity.
a phase shift θ that is varying randomly over time. If the intensity is
The sound intensity is defined as the time averaged product of sound calculated it will show that in theory the noise components do not affect the
pressure, for example p (t ) = pˆ cos(ωt ) and particle velocity (with a certain outcome:
phase shift ϕ compared to the pressure signal) u (t ) = uˆ cos(ωt + ϕ ) : T T
1 1
1
T
1
T I = lim
T →∞ 2T ∫ p(t )u (t )dt = lim 2T ∫ ( P
T →∞
S + PN )(U S + U N )dt
I = lim
T →∞ 2T ∫ p(t )u (t )dt = lim 2T ∫ pˆ cos(ωt )uˆ cos(ωt + ϕ )dt
−T
T →∞
−T T
−T −T

1
1
T = lim ∫ PU S S + PS U N + PN U S + PN U N dt
ˆ ˆ ( cos(ϕ ) + cos(2ωt + ϕ ) ) dt
T →∞
∫
= lim 1
pu 2T −T
2
T →∞ 2T T
−T 1
1
T = 1
2 pˆ s uˆs cos(ϕ ) + lim
T →∞ 2T ∫ PU S N + PN U S + PN U N dt
= 1
2
ˆ ˆ cos(ϕ ) + lim
pu
T →∞ ∫ 1
2
ˆ ˆ cos(2ωt + ϕ )dt = 12 pu
pu ˆ ˆ cos(ϕ ) ≡ Re{S pu } −T
2T −T (2.47) = 1
2 pˆ s uˆs cos(ϕ ) (2.51)

2-25 2-26
 Sound & Vibration Sound & Vibration

The last simplification is possible because the time average product of The uncorrelated parts of two noisy sources are reduced if the cross
two uncorrelated signals is zero, so e.g.: spectrum is used. When the autospectrum is taken from the noisy pressure
T T signal from the previous example, the noise component does not reduce as
1 1 can be seen in the expressions below:
lim
T →∞ 2T ∫ PU
−T
S N dt = lim
T →∞ 2T ∫ pˆ
−T
s cos(ωt ) × uˆn cos(ωt + ϑ (t ))dt
T T
1 1
S pp = lim ∫ (P + PN )( PS + PN )dt = lim ∫P + PN2 + 2 PS PN dt
T 2

= lim
1
T →∞ 2T ∫ 1
2 pu (
ˆ ˆ cos ϑ (t ) + cos(2ωt + ϑ (t )) dt ) T →∞ 2T −T
S
T →∞ 2T −T
S

−T T
1
= lim ∫P + PN2 dt = S PS PS + S PN PN = pˆ s2 + 12 pˆ n2
T 2 1
1
∫
S
ˆ ˆ cos ϑ (t )dt = 0
2
= lim 1
2 pu T →∞ 2T
−T (2.57)
T →∞ 2T
−T (2.52)
The time averaged product of 2PSPN is again zero but the squared noise
This effect can be understood if one observes the phase: it is varying component PN2 is not. It shows that the noise is will not be reduced when
randomly in time so the product is half of the time positive and half of the the autospectrum is used.
time negative. Therefore the time average is zero. Of course the averaging
time is not infinite and the noise in two signals is often not completely If one wants to measure for example the autospectrum of the pressure
uncorrelated so the noise reduces not to zero. In practical cases the noise is that is generated by a very low level sound source, selfnoise of the
reduced 10dB-40dB. transducer may be a problem. One can use two identical transducers and
measure at the same location the cross spectrum of the two signals. The
Conclusion is that in theory the noise component does not affect the selfnoise will be reduced considerable that way. A reduction of 40dB can be
cross spectrum of two signals. This is not the case for the auto spectrum as achieved this way.
will be shown in the paragraph below.
The formal equations to calculate the cross spectrum are introduced via Transfer function Hxy
the correlation function which indicates the time average relationship
The transfer function of two signals is the frequency representation of
between two signals in the time domain. The cross-correlation function
what in the time domain would be the ratio of two signals “x” and “y”.
between the sound pressure and particle velocity is given by:
Just like the cross spectrum, to calculate the transfer function one needs
1 T
R pu (τ ) = lim
T →∞ T ∫ 0
p (t )u (t + τ )dt
(2.53)
two signals. If the example of the music-cd is taken once more; one can
calculate the transfer function of a stereo signal. If the same singer is
recorded with the two microphones, just like the example before, the
So, the sound intensity is given by: amplitude of both channels will be approximately the same but due to the
1 T time difference a phase shift of the stereo signal will occur. The phase shift
T ∫0
I = lim p (t )u (t )dt = R pu (0) that is calculated with the transfer function is the same as calculated with
T →∞
(2.54) the cross spectrum but when the amplitude of both channels are
approximately the same the magnitude of the transfer function will be
The Fourier transform of the cross-correlation function gives the cross approximately unity.
spectrum:
The transfer function is defined as the cross spectrum divided by the
1 ∞
autospectrum of one of the signals.
S pu (ω ) =
2π ∫
−∞
R pu (τ ) eiωτ dτ
(2.55) S xy
H xy =
S xx (2.58)
Autospectrum versus cross spectrum

The auto spectrum of a signal is in fact nothing more than the cross Coherence
spectrum of the signal with itself. Two identical signals are of course 100%
correlated and the phase shift is zero, so Eq. (2.48) may be used: The coherence is a value that is often used to check if measurements are
valid. The value equals unity if two signals p and u 100% correlated. If so,
S xy = S xx S xx ( cos 0 + i sin 0 ) = S xx
x= y
(2.56) the cross spectrum can be rewritten as the product of two autospectra:

2-27 2-28
 Sound & Vibration Sound & Vibration

S pp S pp ( cos(ϕ pu ) + i s in(ϕ pu ) )
2 2 2
The imaginary part of the transfer function is determined in a similar
S pu S pp S pp
coh = = = =1 manner.
S pp Suu S pp Suu S pp Suu If the input sine wave is swept through the frequency band of interest,
(2.59)
the complex transfer function is known.
If the signals p and u are noisy they can be represented as done in Eq.
(2.50). The cross spectrum will not be affected by the noisy component Time-frequency representation
but both autospectra will be a larger value:
A time-frequency representation (TFR) is a view of a signal (that is a
pˆ s uˆs ( cos(ϕ ) + i sin(ϕ ) )
2 2
S pu 1
pˆ s2 uˆs2 function of time) represented over both time and frequency. A signal, as a
coh = = = <1
2
function of time, may be considered as a representation with perfect
S pp Suu ( 1
2 pˆ s2 + 12 pˆ n2 )( 1
2 uˆs2 + 12 uˆn2 ) ( pˆ 2
s )(
+ pˆ n2 uˆs2 + uˆn2 ) (2.60) temporal resolution. The magnitude of the Fourier transform (FT) of the
signal may be considered as a representation with perfect spectral
resolution but with no temporal information (although the magnitude of the
An alternative method to calculate the Fourier response FT conveys frequency content, it fails to convey where in time different
events occur in the signal). The time-frequency representation provides
The Fourier transformation can be done with an electronic network. The
some temporal information and some spectral information, simultaneously.
block diagram is shown in Fig. 2.12 [18].
They are used for the analysis of signals containing multiple time-varying
frequencies.
t
A B Re{H(i )} As an example, a simulated signal which the time representation is
H(i ) X depicted in Fig. 2.13 left. Clearly, this is an oscillating signal whose
LPF frequency varies with time. However, it is difficult to conclude from this
representation the relationship between frequency vs. time. The spectrum
of this signal does not give more indications, expect the fact that its
frequency goes through the entire bandwidth see Fig. 2.13 right.
Im{H(i )}
X
LPF
sin t

Fig. 2.12: Block diagram of a system that determines the imaginary and real part of a
system by means of a sine signal.

If in Fig. 2.12 a sine wave cos(ωt) is submitted to a system H(iω), the

signal at position ‘A’ of the system will be:

A = cos(ωt ) × H (iω ) cos(ωt + arg{H (iω )})

(2.61)

After multiplying signal ‘A’ with the input signal cos(ωt) the signal at Fig. 2.13: Left: time representation of a signal; right: the frequency representation of the
position ‘B’ of the system is: same signal.

A = cos(ωt ) × H (iω ) cos(ωt + arg{H (iω )}) The objective of time-frequency analysis is to offer a more informative
description of the signal which reveals the temporal variation of its
= H (iω ) cos(arg{H (iω )}) + H (iω ) cos(2ωt + arg{H (iω )})
(2.62) frequency contents. The plot in Fig. 2.14 gives an illustration: it shows that
the signal is a simple sine sweep.
After proper low pass filtering:
A = H (iω ) cos(arg{H (iω )}) = Re{H (iω )}
(2.63)

2-29 2-30
 Sound & Vibration Sound & Vibration

Reciprocal measurements are found in NVH applications for the reason of

the very different space requirements of sound sources and sensors;
measurements of acoustic transfer functions are often much easier done
reciprocally.
A typical example application is the measurement of acoustic transfer
paths of sound radiated from a vehicle's engine to the driver's ear. The
engine compartments of today's cars are almost completely filled with the
engine itself and ifs various subsystems. Therefore a microphone can be
installed much easier than a speaker, whereas in the cabin there should be
enough space for a sound source.
Furthermore, reciprocal measurements can offer better results, and again
size is a reason. Small sensors can be placed much closer to sound-
radiating surfaces to which transfer functions are to be determined, and
measurement positions can be chosen almost without restriction Thus the
local sound field is sampled better than by the direct measurement, which
employs a sound source inside the engine compartments that usually
cannot be placed at the exact point of interest. Hence the reciprocal transfer
function may approximate the real transfer path better than the direct
measurement.
Moreover, when measuring reciprocally, e.g. with a sound source inside
the cabin, all transfer paths are excited - and thus can be measured -
simultaneously. This means a significant reduction in time requirements
Fig. 2.14: time-frequency representation of a sine sweep signal.
compared with the direct method, where each transfer function of interest
has to be measured one after another.
2.7 Reciprocity
Acoustic reciprocity
The following text is a summary from [1], [2], [3], [4] and [9]. Loosely stated, the reciprocity principle states that the transfer function
In a linear elastic system, vibrational excitation at one point causes a p/Q between a monopole sound source (Q) and the resulting sound
response at another point. Generally speaking, if the system is passive and pressure field (p) is unchanged if one interchanges the points where the
time-invariant, the transmission of vibration is invariant with respect to monopole source is placed and where the sound pressure field is measured.
exchange of the points of excitation and observation. Such a system is
reciprocal. In other words, reciprocity states that the transfer path in one
direction equals the one in the opposite direction. The principle is valid in
mechanical, electrical and acoustical systems as well as hybrid systems,
e.g. acoustic transducers.
In its most elementary form the acoustic reciprocity principle states that
an acoustic response remains the same when the source and receiver are
interchanged.
Acoustic reciprocity was first mentioned by Helmholtz 1860 regarding
sound transmission through pipes. A general theory of reciprocity, not
limited to acoustics but valid for arbitrary systems containing friction and
viscous damping is formulated in 1873 by Lord Rayleigh [10]. Lyamshev
published a formal proof of these assumptions in 1959 [11]. He stated that
any vibrating structure can be incorporated in the reciprocal system, thus
paving the way for modern NVH applications. Fig. 2.15: Reciprocity for sound pressure with monopoles and particle velocity with
dipoles, from [1].

2-31 2-32
 Sound & Vibration Sound & Vibration

In a similar way the relationship u/M between a dipole sound source (M) 2.8 Point sources
and the resulting particle velocity field (u) is unchanged if one interchanges
the points where the dipole source is placed and where the particle velocity
The simplest source to describe mathematically is a pulsating sphere with
field is measured, see Fig. 2.15.
radius a, that expands and contracts harmonically with spherical symmetry.
The latter can be deducted by the previous statement since a dipole If the radius of the sphere is much smaller than the wavelength (ka<<1),
sound source can be considered as two equal strength monopoles that are the source becomes a monopole or in other words a point source. Such
out of phase and a small distance, d apart (such that kd<<1) and particle sources are difficult to create in practice but the monopole sound source is a
velocity is linked to sound pressure gradient, see Eq. (2.10). central concept in theoretical acoustics. Also for reciprocal measurements
In the presence of impedance boundaries and if nonlinear sound the point source is an essential tool.
propagation is avoided (sound pressure level less than about 135dB in
air) the reciprocity principle is valid.
The reciprocity relation can be noted in a more general form by writing:

∫∫ u
S
2 ⋅n p1dS = ∫∫ u1 ⋅n p2 dS
S
(2.64)

p1 and u1 and p2 and u2 are two arbitrary sound fields in two arbitrary points
in a certain volume V that is enclosed by the surface S and n is the is a
surface normal vector pointing into the volume V.
So p1 is the sound pressure in point one in sound field one and u1 is the Fig. 2.17: A pulsating sphere can be regarded as a monopole if the radius of the sphere is
particle velocity in point two in sound field one. Subsequently p2 is the small compared to the wavelength.
sound pressure in point one in sound field two and u2 is the particle velocity
in point two in sound field two. The sound pressure generated by the point source is given by:

In chapter 10: ‘source path contribution’, some applications of the ρωQ i (ωt − kr )
p=i e
reciprocity principle are shown. 4π r (2.65)

Vibro-acoustic reciprocity With r the distance to the source and Q the source strength [m3s-1]. The
sound power is given by:
The transfer function between a mechanical vibrational force applied
to an elastic plate or shell and the resulting sound pressure (p/F) equals ρ ck 2 Q 2
the ratio v/Q. Where v is the particle velocity determined close to the P=
surface and Q is a monopole sound source, see Fig. 2.16: p/F=v/Q.
8π (2.66)

It is proportional to the square of the frequency, indicating that a small

pulsating sphere is not an efficient radiator at low frequencies. The acoustic
impedance of a point source is given in Eq. (2.25). With that, the particle
velocity field can be calculated.

Q i (ωt − kr ) p ikr
p(r , t ) = i ρ ck e , Z = = ρc
4π r u ikr + 1
(2.67)
Q ikr + 1 i (ωt − kr )
u (r , t ) = e
4π r 2

As can be seen, relative close the source (kr<1), the particle velocity field
Fig. 2.16: Reciprocity of an applied force on a surface and sound pressure with particle
velocity at the surface and a monopole sound source. is decreasing with the square of the distance and in the far field (kr>1), the
particle velocity field is decreasing proportional with distance.

2-33 2-34
 Sound & Vibration Sound & Vibration

Practical realisations of point sources are found in chapter 9: ‘monopole Physically, G(r) represents the effect observed at point x created by a
sound sources’. unit source located at point y . However only one of the unknown variables
has to be prescribed on the boundary S to solve the problem. In
2.9 The Helmholtz Integral Equation engineering situations, the usual given surface normal velocity on the
surface S is used.

There is much work in many fields of acoustics which relies heavily on It can be proved (see e.g. chapter 8, [12]) that an alternative Green’s can
the Helmholtz Integral Equation. In particular, the field of boundary be found so that:
elements has as its starting point this important equation (paragraph is ∂GN
summary from [8] and [12]). =0
∂n y
Fig. 2.18 illustrates a sound radiating object with an arbitrarily shaped,
but closed, boundary surface S. This is called the Neumann Green function. Now the Helmholz equation
(Eq. (2.68)) alters in:

p( x ) = ∫∫ iωρ GN vn dS (2.70)
s

It can be understood that the Neumann Green function can be found

experimentally by e.g. a reciprocal method. With such measurement the
sound pressure in a field point p ( x ) is related to a normal particle velocity
at a surface.
The previous relation can be rewritten in the discrete form if the surface
S divided in small sub surfaces ΔS so that in each sub surface the normal
velocity and the Neumann Green function can be considered constant:
p ( x ) = iωρ ∑ GN vn ΔS (2.71)
Fig. 2.18:Nomenclature in the exterior acoustic problem.
This relation is used in the source path contribution method that is
An integral equation is known as the Helmholtz integral equation. It explained in chapter 10. The Neumann Green function is measured in a
relates the acoustic pressure and normal velocity on the closed boundary reciprocal way. With that Neumann Green function the pressure contribution
surface S of a vibrating object to the radiated pressure field in the fluid of each part in e.g. a car is determined by measuring the surface particle
domain: velocity.
⎧⎪ ∂G (r ) ⎫⎪
p( x ) = ∫ ⎨
∂
p( y ) + iωρ G (r )vn ( y ) ⎬ dS + pin ( x ) (2.68) 2.10 Sound & vibration sensors
⎩⎪ n ⎭⎪
S y
y

In this chapter the most common sensors that detect sound and vibration
In a field point x , the p ( x ) can be calculated. The point x is also called
are presented.
the evaluation point because the integral has to be evaluated for each point
x . The unit normal to the surface at source point y , denoted as ny , is Transducers are devices that convert one quantity into another.
Membrane microphones are acoustic transducers that are commonly applied
pointed into the fluid domain. Distance r is the length of vector r that is to convert sound pressure into electrical signals. The sound pressure
directed from the source point y to the field point x : r = x − y . The term “pushes” a membrane forward and backwards and the deflection of the
pin represents the incident acoustic wave in the case of a scattering membrane is detected electrically. When this membrane closes a cavity the
analysis. In an unbounded fluid domain without reflecting objects (free microphone detects sound pressure. In a construction in which both sides of
space), Green’s function G(r), also referred to as the kernel of the integral the membrane are subject to the acoustic wave, the microphone is sensitive
equation, is: for sound pressure gradient.

e − ikr Sound, however, consist of two elements: sound pressure and particle
G (r ) = (2.69) velocity. A pressure microphone is constructed to measure sound pressure
4π r

2-35 2-36
 Sound & Vibration Sound & Vibration

whereas the Microflown enables the direct and reliable measurement of deflection of the membrane causes a variation of the capacitance. This type
particle velocity. of microphone needs an additional (bias) voltage for operation and is called
the condenser microphone. To avoid external biasing it is possible to
To determine the amount of noise, sound intensity has to be measured.
provide one of the plates with an electret, an electrically charged layer,
Two sensors are required for the determination of sound intensity. The most
creating a so-called electret microphone. Capacitive microphones have good
obvious way is measuring sound pressure (p) with a microphone and
noise properties (or signal to noise ratio), which is the most important
particle velocity (u) with a Microflown, and then multiply and time integrate
parameter of a microphone. The first condenser (electret) microphone was
both signals. This is called the p-u method. Different methods use closely
built in 1918 [16]. The construction of today’s measurement condenser
spaced microphones (p-p method) or Microflowns (u-u method).
microphone is illustrated in Fig. 2.19.
Measuring sound pressure Although several different ways to detect deflection of a membrane exist,
the previous mentioned are the most common. To get a better
The detection of sound pressure of a microphone is similar to the way the understanding of the sound pressure measurement, the condenser
human ear perceives sound. A membrane closes of a cavity, inside the microphone is investigated further.
cavity the pressure equals the static air-pressure and pressure variations
Two electrical conducting plates separated by an insulator will create a
outside the cavity cause a movement of the membrane. This movement is
condenser. The condenser microphone consists on an electrical conducting,
detected. In this paragraph the functioning of a capacitive microphone is
stretched membrane over an electrical conducting backplate forming
explained.
together, with the air gap (insulator), a condenser. Sound pressure excites
All practical microphones detect the (sound pressure induced) deflection, the membrane, changing the capacitance of the microphone. This
or movement, of a membrane. There are several ways to implement this. capacitance is electrically measured. The performance of a condenser
microphone depends upon the design of its electrical and mechanical
systems.
A high polarisation voltage, large membrane radius, small air gap, and
low membrane tension favour the sensitivity. Sensitivity is a microphone's
electrical output caused by a sound pressure given in Volts per Pascal.
The bandwidth, the range of frequencies within which a microphone
yields a usable output signal, is favoured by a high membrane tension,
small membrane radius and low surface mass density (implying a thin
membrane). The response of a microphone “rolls-off” at low frequencies
(frequencies lower than 1 to 125Hz) due to the presence of a capillary vent,
see Fig. 2.19. The purpose of this vent is to provide a path for pressure
equalisation on both sides of the membrane following a change in ambient
pressure. However, this vent supplies also a path for low-frequency
acoustical waves to reach the backside of the membrane and thus
suppresses the low-frequency response of the microphone.
The selfnoise is the electrical output of a microphone (in an absolute
silent environment) caused by noise, expressed in a quantity corresponding
to a certain sound level. Resistive damping of the membrane by the air
Fig. 2.19: A cross-section of a condenser microphone. layer and the holes in the backplate (see Fig. 2.19) in the air gap causes the
thermal noise of a condenser microphone. For lower frequencies, the
Electromagnetic microphones consist of a coil connected to the preamplifier 1/f-noise dominates. The thermal noise of the microphone is an
membrane. The movement of the membrane causes the coil to move in a inverse function of the membrane radius. If the sensitivity of the
magnetic field resulting in an electrical signal. The first who developed this microphone is low the preamplifier noise becomes more dominant.
type of microphone was the German teacher Phillipp Reis [15] in the year
The preamplifier is needed to convert the very high impedance of the
1860. Since then the development of sound pressure sensors began.
capacitive microphone to a lower output impedance. The voltage gain of
The capacitive microphone detects the acoustic pressure variations by these preamplifiers is usually unity.
means of a condenser [16]. The membrane acts as one plate of the
The non-linear relationship between the capacitance and the membrane
capacitor whereas a perforated-rigid-back-electrode is the other plate. A
displacement is the main reason for the harmonic distortion at high sound

2-37 2-38
 Sound & Vibration Sound & Vibration

pressure levels. A high membrane tension and small membrane radius will with the membrane radius so that particularly useful compromises in
increase the upper sound pressure limit. bandwidth, dynamic range, and sensitivity can be achieved. The upper four
microphones are used for measurement purposes and the Knowles is used
A polarisation voltage causes an attracting force between the membrane
for audio pick-up.
and the backplate. If the polarisation voltage is chosen too high, the
membrane will collapse to the backplate. The cylindrical microphone (Knowles Electronics) for instance has a
diameter of 0.1" (the company Microtronic has a similar microphone). This
A polar pattern (or microphone directivity) expresses the sensitivity to
microphone is used for hearing aid purposes. The selfnoise of this miniature
the angle of incidence of the sound field. A sound pressure microphone
microphone is 27dB(A), the effective* bandwidth however is limited to
should be omni-directional which implies a constant sensitivity for any angle
approximately 20kHz (1/8” B&K has a bandwidth of 160kHz), and the upper
of incidence of the sound field. The larger the membrane radius (and thus
limit of the sound pressure is 100dB SPL. The sensitivity is 30mV/Pa. The
physical size) the more the microphone behaves as an obstacle in the sound
high sensitivity and low sound-pressure upper-limit suggest low membrane
field. Due to this effect the directivity of the microphone increases for higher
tension.
frequencies. Another cause of directivity of sound pressure microphones is
related to the diameter. When the wavelength is becoming smaller than the
diameter, and the angle of incidence sound wave is perpendicular to the
membrane, one part of the membrane is pulled and another part is pushed.
The sensitivity will be reduced due to this phenomenon and due to this the
sound pressure microphone becomes directional.
A microphone with a half-inch membrane radius and 200V polarisation
voltage is chosen for general measuring purposes. This is because the
microphone has a relatively low selfnoise (20dB(A)), a sufficiently large
bandwidth (20kHz) and reasonable omni-directional properties (3dB
variation at 8kHz). The microphone can be used for sound levels up to
160dB.
The half inch pressure microphone is used most commonly. The
bandwidth of such microphone is specified as higher than 20kHz but the
frequency response is not flat for frequencies higher than 4kHz. Fig. 2.20: The effect of mechanical properties on acoustical specs of condenser
microphones.
For higher frequencies the frequency response becomes directional.
Therefore two types of ½” microphones are designed, a pressure The lower membrane tension will result in a different mechanical design
microphone and a free field microphone. i.e. air layer and holes in the backplate; therefore the selfnoise due to the
resistive damping of the membrane by the air layer and the holes in the
backplate is lower than the 1/8” B&K. Due to the low membrane tension the
Table 2.2: General microphone specifications related to the membrane radius sensitivity is high and the preamplifier noise becomes less dominant.
Membrane radius Sensitivity Bandwidth Selfnoise Upper limit
[mV/Pa] [kHz] dB(A) [3% distortion]
1" 50 10 10 150 dB SPL
½" 12,5 24 22 160 dB SPL
¼" 4 100 36 170 dB SPL
1/8" 1 160 55 180 dB SPL
1/10" (Knowles) 30 20 27 100 dB SPL
Fig. 2.20 shows how mechanical properties affect the frequency response
and dynamic range and table 2.2 shows how dynamic range, bandwidth and
sensitivity depend on microphone size. When the membrane diameter
decreases, the maximum sound pressure level a microphone can withstand,
Fig. 2.21: free field corrections for various angles of incidence.
and the noise floor will increase. Sensitivity is proportional to the diaphragm
area. In actual microphone design, other parameters may be changed along

2-39 2-40
 Sound & Vibration Sound & Vibration

A Free-field microphone is designed to measure the sound pressure in a an impulse to the brain that analyses and translates the acoustic impulse
sound field and compensate for the influence of its presence in the sound into a concept or situation.
field. A free-field microphone measures the sound pressure as it existed
In other words the essential difference between the detection of sound
before the microphone was introduced into the sound field; in effect,
with a pressure microphone and hearing is that hearing is the combination
maintaining free-field conditions. A free-field microphone should be pointed
of two pressure sensors and a processing unit. A microphone can only
towards the sound source, i.e. at an angle of 0º incidence to the source.
detect the sound pressure, interpretation must be done afterwards.
A pressure microphone is designed to measure the sound pressure that
Loudness is the physical response to sound. At any given frequency
actually exists in front of its diaphragm. This means that the microphone
between 20Hz and 20kHz the loudness varies directly with the sound
also measures any changes in the sound field brought about by the
pressure, but not in a simple straight-line manner. In Fig. 2.22 the equal
presence of the microphone. A pressure microphone is used, typically, in
loudness contours are shown, they represent the sound pressure level
coupler measurements and, when flush-mounted on a wall for example, for
necessary at each frequency to produce the same loudness response in the
sound-pressure measurements on the surface of the wall. In such
average listener. As can be seen loudness is a very non-linear physical
applications, the effect of the microphone's presence in the sound field is
response. Humans are quite deaf at low frequencies. At 20Hz the level must
intended.
be almost 70dB higher (ten million times as much sound energy) as at 2kHz
The difference in the two designs is the different frequency response. to produce hearing. Another non-linearity appears at low frequencies: if, for
example, the sound level is rising from 80dB to 90dB, humans perceive a
As can be seen in Fig. 2.21, at 20kHz the frequency response of a ½”
20dB increase of the sound pressure.
microphone varies 13dB as function of angle of incidence. Sound pressure is
a scalar value and a sound pressure microphone should not be directional.
It is therefore in general recommendable to use a smaller microphone if
sound pressure is measured above 4kHz.

Measuring sound pressure with a Microflown sensor

There are various ways to be able to measure sound pressure with a

(particle velocity sensitive) Microflown. If a Microflown is positioned very
close behind a membrane the vibrations of the membrane will be sensed.
The membrane is pressure sensitive so the Microflown will be too. In this
case the sensing part of a pressure microphone is replaced by a Microflown.
(See also chapter 7: “vibration measurements”).
It is also possible to put a Microflown in a standing wave tube. If the
sound pressure in increased in front of the tube air will flow in to the tube
and if the sound pressure is reduced, air will flow out. So the tube
transforms sound pressure into particle velocity.
Fig. 2.22: Equal loudness contours of the human hearing.
In stead of a tube it is also possible to use a Helmholtz resonator and put
the Microflown in the throat. Again, sound pressure causes air to flow in and To convert sound measurements performed with a microphone in one
out of the resonator. figure representing loudness, the “A”-weighing curve is introduced:
The microphones that are made by this techniques are not better, see ( 12200 f 2 )2
W A ( f ) = 1.26 (2.72)
further chapter 13: “the pressure Microflown”. (f +20.6 ) (f +12200 2 ) (f 2+107 .7 2 ) (f 2+737 .9 2 )
2 2 2

The Human ear

This “A”-weighing curve is similar to the inverse of the equal loudness
contour of 0dB, as shown in Fig. 2.22.
The human ear is sensitive for sound pressure. Hearing is a subjective
response to sound pressure. At very low frequencies or very high pressure-
Measuring sound pressure gradient
levels, additional sensations are experienced but normally the ear is the
acoustic receiver. It receives the vibrations on the eardrum, multiplies them
The difference between the configuration of a sound pressure microphone
by means of small bones in the middle ear, and transmits the vibrations
and a (sound pressure) gradient microphone is that the gradient
through a fluid to nerve endings within the inner ear. These nerves transmit
microphone has two (in stead of one) acoustic entrances separated with a

2-41 2-42
 Sound & Vibration Sound & Vibration

distance L, see Fig. 2.23A. The membrane will deflect due to a sound For harmonic sound waves (p(t,x)=p0 cos(ωt-ωx/c)), the frequency
pressure difference over the membrane. (In fact a pressure microphone response of a gradient microphone is not constant; the response increases
detects a pressure gradient over the membrane as well, but in the back linear with frequency:
chamber the pressure is constant).
p ( x, t ) − p( x + Δx, t ) d ω
p ( x, t ) − p( x + Δx, t ) = Δx ≈ p ( x , t ) Δx = p ( x , t ) Δ x
Membrane including Δx dx c (2.25)
electronic readout

Acoustic
entrance
Acoustic
entrance
When the frequency increases so much that such that half the
P1 xx xx P2 P1 xx xx P2 P1 xx xx P2
wavelength equals the separation distance, the sensitivity becomes
maximal: two times higher than if the membrane was used in one-input
L
R configuration (regular pressure microphone). After this maximum the
C sensitivity drops dramatically, see Fig. 2.25. To show this effect, the
A B C frequency response due to a separation distance of 5cm is plotted: as can
be seen the sensitivity is at its maximum at 3.3kHz;
Fig. 2.23: A cross-section of a pressure gradient microphone.
(½λ=½c/f=0.5x330/3300=5cm).
To create a useful output signal the response should be integrated (flat
If a sound wave enters in the sensitive direction, see Fig. 2.23B, the frequency response). Simply said, in the frequency domain integrating is
sound wave reaches first input P1 and a while later input P2. In this way the the same as multiplying the sensitivity with 1/f.
pressure gradient is determined; the output of this type of microphone will
be linear with respect to: p(x)-p(x+Δx). Observing Eq. (2.10), it is not Integrating can quite easily be achieved with an electronic circuit. In this
surprising that this type of microphone is also called “velocity microphone” case, for decreasing frequencies the signal to noise ratio will decrease.
since the particle velocity can be calculated by a pressure difference of two It is also possible to use acoustic elements to achieve the integrating.
closely spaced microphones. Very simply explained, a dissipating element R is placed between the
The velocity microphone will not be sensitive in the direction acoustic inputs and the membrane (that is stretched very loose), indicated
perpendicular to the sensitive direction, see Fig. 2.23C. The sound wave with “xxx” in Fig. 2.25 and a large acoustic chamber creating a compliance
reaches both sides of the membrane at the same time and therefore the C are together creating an acoustic RC low pass filter (integrator). The
subtraction p(x)-p(x+Δx) is zero. resulting frequency response will be reasonable flat from 250Hz up to
16kHz. Due to the lesser damping of the pressure gradients’ membrane, the
It can be deduced that the sensitivity as function of the angle entry (the inherent noise of the capsule is much lower than a sound pressure sensitive
polar pattern) is a cosine function, the response is plotted in Fig. 2.24. At microphone. However, the frequency response is filtered by the acoustic
the right side this cosine function is shown. This shape of polar pattern is low-pass filter and therefore the selfnoise of a pressure gradient
often called a “figure of eight”. Normally a logarithm representation is microphone is a few dB worse than a pressure microphone. For audio
preferred; it is shown at the left side. applications where signal to noise values should be as high as possible,
acoustic integration is used whereas for measurement purposes where
linearity and a flat frequency response is the most important issue,
electronic integration is employed.

Fig. 2.24: Response of a sound pressure microphone (dashed line; omni directional) and a
sound pressure gradient microphone (solid line; figure of eight).

2-43 2-44
 Sound & Vibration Sound & Vibration

sensor; a sensor that is designed to measure DC-flow. The fast response

20dB time and improved sensitivity is the only difference between a mass-flow
sensor and a Microflown.

Output [dBV]
e
lop
cs
/de A general behaviour of pressure gradient microphones is the lower
dB
20 1cm separation sensitivity for lower frequencies and ‘comb filtering effect’ at high
frequencies.
Signal to noise Unlike this crucial difference between pressure gradient microphones and
distance
1/f noise Preamplifier
Microflowns there are also similarities. For instance the polar pattern is the
same for a Microflown and a pressure gradient microphone. The near field
Thermal noise
effect (see above “the one-dimensional wave equation”) is also noticed in
10 100 1k 10k the same manner.
Frequency [Hz]
Another type of “velocity” microphone is the so-called ribbon
Fig. 2.25: Response of a pressure gradient microphone with an acoustic input separation microphone [17]. The operation principle of this type of microphone is
of 5cm and 1cm. again based on pressure gradient. In this case an aluminium foil (the
ribbon) is placed under a low tension in a strong magnetic field, see Fig.
Integration of the output of a 5cm-separation configuration results in a 2.26. The difference in sound pressure between the two sides drives the
flat frequency response from low frequencies up to 3.3kHz. For frequencies ribbon and the velocity of the ribbon has the same response as the pressure
higher than 3.4kHz the response will be useless. gradient microphone. The output voltage due to the motion of the ribbon
however is proportional to the product of the flux density, the length and
To achieve a useful output signal for high frequencies the separation
the velocity of the ribbon. The result is a “flat” frequency response. (The
distance should not exceed 1cm. But as can be seen in see Fig. 2.25 for low
ribbon moving in a magnetic field has an integrating character, whereas the
frequencies the signal to noise ratio is very poor compared to a regular
pressure gradient microphone integrating the integration of the signal have
pressure microphone.
to be performed electronically afterwards or acoustically on forehand).
As stated, the output signal is integrated to create a flat frequency
The ribbon velocity microphone is a pressure gradient type of microphone
response. When corrected electronically afterwards, the noise is also
that has (thus) no DC sensitivity. The selfnoise of such microphones is quite
integrated and the signal to noise ratio therefore is not affected by this
low since there is (of course) no thermal noise of the airgap. The main noise
correction. The selfnoise of a sound pressure gradient microphone (pair)
source is thermal noise of the electrical resistance of the ribbon.
that makes use of electrical correction afterwards is therefore always worse
than a sound pressure microphone. When corrected acoustically
beforehand, the frequency response will lack the first decade (20Hz-250Hz)
and the noise figures will be only a few decibels worse than sound pressure
microphones. A good sound pressure gradient microphone (Schoeps MK8)
has a selfnoise of 19dB(A) and a good sound pressure microphone (Schoeps
MK2H) has a selfnoise of 13dB(A) to 15dB(A).

Measuring particle velocity

The direct measurement of particle velocity is quite new. Until the

invention of the Microflown in 1994, the particle velocity could only be
measured indirectly. Here three velocity detecting methods will be
described: the ribbon microphone, an ultrasonic detection method and the
Microflown.
As stated before, sound pressure gradient shows a great resemblance to
the particle velocity. After all, Eq. (2.10) states that the pressure gradient is
proportional to the particle velocity. However in Fig. 2.25 one can see that Fig. 2.26: A ribbon type of microphone. Left: schematic overview; right: the RCA BK11A
the sensitivity of a pressure gradient microphone increases proportional velocity microphone [17].
with the frequency. For zero Hertz (a DC or constant particle flow) the
sensitivity is zero. The Microflown however originates from a mass flow A complete other way of sensing particle velocity is the principle of
ultrasonic transduction: two parallel ultrasonic beams are launched in

2-45 2-46
 Sound & Vibration Sound & Vibration

opposite directions. The travelling time from the transmitter to the receiver seeding particles (tracers) suspended in the fluid. Two laser beams of equal
is linear proportional to the speed of sound in the air. When the air is intensity are focused and crossed at the point under investigation, forming
moving this movement should be added to this speed. The probe consists of an ellipsoidal volume consisting of equidistant dark and bright fringes. The
two transmitter-receiver pairs that are positioned opposite directions, see scattered light is collected on a photomultiplier.
Fig. 2.27. The difference signal of the ultrasonic sound waves is proportional
The Microflown is a particle velocity sensor that is based on a thermal
to the particle velocity; this type of velocity probe is also capable to
principle. It consists of two very closely spaced and thin wires of silicon
measure DC flow. It is a true particle velocity sensor but it is a distributed
nitride with an electrically conducting platinum pattern on top of them. An
sensor: it doesn’t determine the particle velocity in one spot.
example of an older type of Microflown is shown in Fig. 2.28. The sensor is
This probe however is not used very often since it is very sensitive for DC made from silicon bulk material with platinum electrical connections on top
flows as for example wind or movements of the probe. Furthermore it is, of it and two platinum temperature sensors. At the top of the die one can
just like the p-p probe, a distributed sensor with the problems associated see two sensors sticking out. The electrical connecting wirebonds are also
with this type of sensing (for example maximal frequency limited by the visible.
spacing). Reflections of the ultrasonic signals caused problems when the
The size of the two wires is 1mm in length, 5μm in width and 200nm in
probe was used nearby reflecting objects and at last, due to it physical
thickness. The metal pattern is used as temperature sensor and heater. The
dimensions, the probe is difficult to calibrate.
silicon nitride layer is used as a mechanical carrier for the platinum resistor
patterns. The sensors are powered by an electrical current, causing the
sensors to heat up. The temperature difference of the two cantilevers is
linear dependent on the particle velocity level.
The operation principle will be explained briefly here, a more detailed
mathematical model of the Microflown model is presented in chapter 3. The
two squares S1 and S2 in Fig. 2.29 represent the two temperature sensors
of the Microflown. The temperature sensors are implemented as platinum
resistors and are powered by an electrical current dissipating an electrical
power, causing them to heat up. An increase of the temperature of the
sensors leads to an increase of the resistance as well.
The sensors have a typical operational temperature of about 200ºC to
400ºC if no particle velocity is present and all the heat is transferred in the
surrounding air. When particle velocity is present, it alters the temperature
distribution around the resistors. The temperature difference of the two
sensors quantifies the particle velocity.

Fig. 2.27: Type 216 p-u intensity probe (Norwegian electronics).

If the signal from only one ultrasonic beam is observed one notices quite
large a pressure sensitivity. This is caused by the temperature variations
that are generated by the pressure variations of the sound field. The
temperature variations cause a variation in the speed of sound which is
measured to detect particle velocity with this method. To get rid of the
unwanted pressure sensitivity a second ultrasonic beam is used in the
opposite direction. The pressure sensitivity of this second beam is the same
as the first one but the particle velocity sensitivity has a phase shift of 180
degrees. The subtraction of both signals got rid of the pressure sensitivity.
A (stereoscopic) Laser Doppler Velocimeter is measurement device
used for estimating particle velocities, by means of specific signal
processing of scattered light. It is an optical technique allowing direct
measurement of local and instantaneous particle velocity. Its principle is
based on the determination of the Doppler shift of light scattered from

2-47 2-48
 Sound & Vibration Sound & Vibration

Fig. 2.29: Schematic overview of the heat flows around a Microflown.

At higher frequencies the sensitivity of the Microflown is decreasing. This

high frequency roll-off that is caused by the time it takes heat to travel from
one wire to the other can be estimated by a first order low pass frequency
response that has a (diffusion) corner frequency in the order of 1kHz. The
second high frequency roll-off is caused by the heat capacity (or thermal
mass) and shows an exact first order low pass behaviour that has a heat
capacity corner frequency in the order of 10kHz for modern Microflowns.

Fig. 2.28: Scanning electron microscope photo of cantilever type of Microflowns. The two
wires that are sticking out are the actual Microflown, the three squares are the

Particle velocity

Particle velocity
“bondpads” that are used to make electrical connections and the three wires in the lower
part of the picture are wirebonds: Aluminium wires that are 80 micron in diameter. See
further chapter 1 and 3.

Particle velocity causes a convective heat transfer of both sensors which

will create a temperature drop of both sensors. The upstream sensor
however, drops more in temperature than the downstream sensor since the
downstream sensor is heated by the upstream convective heat loss see Fig.
2.29. A temperature difference is the result. This temperature difference is Obstructions Position Microflown
proportional to the particle velocity.
Not all the convective heat loss of the upstream sensor will be Fig. 2.30: A well-chosen package will result in a particle velocity gain (top view).
transferred to the downstream sensor; a certain percentage will be lost.
This percentage will rise if the sensors are positioned further apart from To protect its fragile sensors, the Microflown has to be packaged. An
each other. If, on the other hand, the sensors are brought closer together positive aspect of the packaging is that the particle velocity level increases
another phenomenon will become dominant. The particle velocity induced when a well-chosen obstacle is placed near the sensors, see Fig. 2.30 and
temperature difference will cause a conductive heat flow in the opposite Fig. 2.31. The so-called package gain is caused due to a “channelling” of the
direction. This feedback heat flow will temper the sensitivity. The closer the particle velocity, this can result in an increase of the particle velocity level of
sensors are placed, the more conductive heat flow will take its effect. a factor 3 up to 30 times (+10dB to +30dB).
Since the noise has a thermal electrical origin, it is not be affected by the
package (gain). The selfnoise of a packaged Microflown therefore can be
expected 10dB to 20dB less than a non-packaged Microflown.

2-49 2-50
 Sound & Vibration Sound & Vibration

method involves combining a pressure transducer with a particle velocity

transducer.
The p-p method is based upon Newton's second law that is the equation
of motion:
∂u r ∂p
ρ⋅ =−
∂t ∂r (2.26)

Where ur is the particle velocity in one direction r. Since the pressure

gradient is proportional to the particle acceleration, the particle velocity can
be obtained by integrating the pressure gradient in respect to time.

Fig. 2.31: An example of a package with a gain of +12 to +12dB: the half-inch particle
velocity probe (1998).

Measuring sound intensity

Sound intensity is useful for measurement of sound power, identification

Fig. 2.32: A well-chosen package will result in a particle velocity gain (top view).
and ranking of sources, visualisation of sound fields, measurement of
transmission loss, identification of transmission paths, etc. In the pioneering In practice, the pressure gradient is approximated by measuring the
days in the 1970s and 1980s sound intensity was a hot research topic, but pressure at two points at a close distance Δr.
since then it has become a well-known and established method, described
in several standards. The conventional measurement technique employs 1
ρΔ r ∫
ur = − ( pa − pb )dt
two matched condenser microphones. However, a particle velocity (2.27)
transducer called the ‘Microflown’ and an intensity probe based on this
transducer combined with a small pressure microphone have recently This approximation is valid as long as the separation, Δr, is small
become available. compared with the wavelength. If the frequency increases the spacing
The sound intensity in a specified direction at a point is the average rate should be chosen smaller. In practice three spacings are being used: 5cm,
at which sound energy is transmitted through a unit area perpendicular to 12mm and 6mm. A practical sound intensity probe consist of two closely
the specified direction at the point considered. This measurement allows spaced pressure microphones, as for instance the Brüel & Kjær intensity
obtaining information about noise sources independent on the environment probe fitted with two ½" pressure microphones, see Fig. 2.32.
or other noise sources. It is thus a vector quantity; defined as the time The p-p method is sensitive to large values of the pressure-intensity
averaged product of the sound pressure (a scalar quantity, p) and the index, but not to high values of the reactivity. By contrast the p-u method is
corresponding particle velocity (a vector quantity, u) at the same position. sensitive to high values of the reactivity, but not to large values of the
In the general case measurement of sound intensity requires the use of pressure-intensity index. Thus p-u intensity measurement systems are
at least two transducers. Sound intensity is usually measured with the p-p potentially less affected by extraneous noise but more affected by reactive
method, which combines two pressure micro-phones and makes use of a near fields than p-p intensity measurement systems. Both limitations can be
finite-difference approximation to the pressure gradient. The alternative p-u serious in practical measurements, and one cannot conclude from these
considerations that one method is superior to the other.

2-51 2-52
 Sound & Vibration Sound & Vibration

In measurements with p-p probes the acceptable pressure-intensity of points with a scanning laser vibrometer. Disadvantages of a laser
index depends on how well the two microphones are matched, that is, on vibrometer are its large size, a stability requirement, it high price and that it
the ‘pressure-residual intensity index’ of the probe [1]. With state-of-the-art is sometimes difficult to use.
equipment a pressure-intensity index exceeding 10dB is ‘high’ in most of
An accelerometer is a low cost sensor that is sensitive for the
the frequency range; below 200Hz even an index of 5dB is ‘high’. Such high
acceleration of a structure (so the time derivative of the velocity). The
values of the pressure-intensity index will occur if there is strong
sensor is broad banded and has a large dynamic range. Disadvantage is
background noise from other sources than the one under test, in particular
that the sensor must be fixed to the structure which is time consuming and
if the measurement takes place in reverberant surroundings. If the
may alter the structure to be measured.
pressure-intensity index is unacceptably high possible countermeasures
include moving the measurement surface towards the source under test,
shielding the extraneous sources with temporary screens, and introducing
more absorption in the room.
In measurements with p-u probes the acceptable reactivity depends on
the accuracy of the phase calibration of the device (described in section 3).
A reactivity of more than 5dB is ‘high’, but values of up to 25dB can occur
[8]. Since the phenomenon is associated with near fields of sources the only
remedy is to use a measurement surface further away from the source
under investigation.
It is more complicated to calibrate the Microflown sound intensity probe
than a p-p sound intensity probe. On the other hand the Microflown probe is
much smaller than a p-p intensity probe, which is an advantage in many
applications. With a p-u probe is easily possible to measure the 3D intensity Fig. 2.34: An example of an accelerometer (8mm x 11mm, 2gram in weight, bandwidth
1.5Hz to 18kHz). Right a three dimensional accelerometer.
and the bandwidth of the probes is better than the p-p probes.
It is possible to measure the vibration with Microflowns. See further
Measuring vibration chapter 7: ‘vibration measurements’
In the previous paragraphs methods to measure sound where described.
In this paragraph deals with the laser vibrometer and the accelerometer,
2.11 References
the two most known methods to determine the vibration of a structure.
[1] F.J. Fahy, Some applications of the reciprocity principle in
experimental vibroacoustics; acoustical physics, vol. 48, nr 2, 2003,
pp 217-229).
[2] J.W. Verheij, Experimental procedures for quantifying sound paths to
the interior of road vehicles, 2nd international conference on vehicle
comfort, part 1, Bologna, 1992, pp. 483-491.
[3] J.W. Verheij, Inverse and Reciprocity Methods for Machinary Noise
Source Characterization and Sound Path Quantification. Part 1:
Sources, Int. J. Acoust. Vibr. Vol. 2 , pp. 11-20 (1997).
[4] J.W. Verheij, Inverse and Reciprocity Methods for Machinary Noise
Source Characterization and Sound Path Quantification. Part 2:
transmission paths, Int. J. Acoust. Vibr. Vol. 3, pp. 103-112 (1997).
[5] Gunnar Lundmark, Skating on thin ice - And the acoustics of infinite
plates, internoise 2001
Fig. 2.33: An example of a laser vibrometer (range: 1µm/s to 5mm/s; bandwidth: DC to
25kHz, laser head dimensions: 50x25x20cm; Weight: 12kg). [6] H-E de Bree, V.B. Svetovoy, R. Raangs, R. Visser, The very near
field; theory, simulations and measurements of sound pressure and
A laser vibrometer determines the structural surface velocity with the use particle velocity in the very near field, ICSV11, St. Petersburg 2004
of a laser beam. It is a sensitive, non contact method that operates in a [7] H-E de Bree, V.B. Svetovoy, R. Visser, The very near field II; an
wide frequency bandwidth and large dynamic range. The measurement introduction to very near field holography, SAE traverse city, 2005
distance may be large (up to meters). It is possible to scan a large number

2-53 2-54
 Sound & Vibration

[8] R. Visser, A boundary element approach to acoustic radiation and

source indentification, PhD Thesis 2004, Enschede, The Netherlands
[9] Sottek et al, An Artificial Head which speaks from its ears:
investigations on reciprocal transfer path analysis in vehicles, using a
binaural sound source, SAE 2003.
[10] Rayleigh, J.W., Some general theorems relating to vibrations, proc.
Lond. Math.Soc., 1873.
[11] Lyamshev, L.M., A question in connection with the principle of
reciprocity in acoustics, Soviet Physics Doklady, 1959
[12] Earl G. Williams, Fourier Acoustics: Sound Radiation and Nearfield
Acoustical Holography, ISBN: 0127539603, June 1999
[13] P.A. Nelson & S.J.Elliot, Active Control of Sound, Academic press,
London,1992.
[14] R. Streicher & F. Alton Everest, The new stereo soundbook,
Audio Engineering Associates California, 1998,ISBN 0-9665162-0-6.
[15] O. Brosze, Aufbau un wirkungsweise elektroakustischer Wandler, Der
Fernmelde Ingenieur, 35 1981 Heft 6, 1-36, Heft 7, 1-35.
[16] E.C. Wente, A condenser transmitter as a uniformly sensitive
instrument for the absolute measurement of sound intensity, Phys.
Rev., 10-1917.
[17] Harry F. Olson, Ribbon Velocity Microphones, June 1970.
[18] Massarani, P. Muller, S. “Transfer-Function Measurement with
Sweeps”, J. Audio Eng. Soc, 2000, pp. 443-471, June 2001

2-55