Chapter 7

Sound recording and reproduction

T
he vast majority of the music we listen to today is recorded, and recording
technology as a result plays a substantial role in the ecosystem of musical
sound and performance. With the exception of the very earliest recording
technologies over a century ago, all modern recording works in essentially the same
way: sound is converted into an electrical signal, which is stored in some manner,
then recovered at a later time and converted back into a replica of the original sound—
see Fig. 7.1.
Thus the process of recording and playback of music has three main elements:
initial capture of sounds, usually using a microphone, then storage, for instance on
tape, CD, or on a computer, and finally reconstruction of the sound from the stored
signal using a loudspeaker, headphones, or earphones. At various stages along the
chain we may also employ amplifiers that increase the strength of electrical signals—
microphones, for instance, typically produce signals too small for storage, so the
signals must be amplified first.
Closely related to sound recording are the processes of sound reinforcement and
broadcasting. Sound reinforcement, the amplification of sound for live performance,

Figure 7.1: The recording process. Sound is turned into an electrical signal, usually with
a microphone, then the electrical signal is stored in some manner, such as on magnetic tape
or (more commonly) on a computer. At playback, the signal is recovered from storage and
played through a loudspeaker, headphones, or earphones to recreate the original sound.

214
 7.1 | Electricity and sound

involves the same steps as recording except that the signal is never stored but goes
straight to an amplifier and loudspeaker for immediate playback. Broadcasting, over
radio or the Internet, involves an additional step of transmitting the signal to a remote
receiver before the sound is reconstructed for the listener.
In this chapter we study the science behind the recording process, including the
workings of microphones, loudspeakers, and various types of recording media, but
sticking for the moment to traditional analog technologies. Digital recording, which
works in a completely different manner, is discussed separately in the following chap-
ter. We do not discuss amplifiers in detail in this chapter: their design and workings
are largely the domain of electrical engineering and are beyond the scope of the book,
though we will discuss the musical use of amplifiers, particularly in the context of
the electric guitar, in Chapter 14.
First, however, in preparation for our discussion—and because it will also be use-
ful in later chapters—we need to learn about circuits and the use of electrical signals
to represent sound.

7.1 Electricity and sound

Modern music recording and amplification technology operates by turning sound
into an equivalent electrical signal. We have already studied in some depth the sci-
ence behind sound waves. To understand recording and amplification we need to
understand electrical signals too.
Electricity consists of a flow of particles—electrons—each of which carries a cer-
tain amount of electric charge, the same on every electron. Electrons are too small
to be visible to the eye, and the amount of charge they carry is also small, but a large
enough number of them—and there are many trillions in a typical electrical cable or
wire—can add up to a large current.
Electricity flows easily through metals and a few other materials such as carbon,
silicon, and water, but not through most other substances. This is convenient because
it allows us to funnel electricity in any desired direction by channeling it along metal
wires. The number of electrons that can fit in such a wire is large but not infinite and, An electrical plug has
because there is a limit on this number, electricity behaves in some ways like water two prongs and electric-
flowing in a pipe: you can put electricity in one end of a wire only if you take a ity flows into one and
back out of the other,
matching amount out of the other end. For this reason, electricity always flows in a
completing a circuit.
loop or circuit, with an entrance and an exit. This is why a battery has two terminals, Some plugs have three
for instance: one terminal feeds electricity into the circuit and the other takes it out prongs, but the third
again. It’s also why an electrical wall plug has two prongs. (ground) prong is mainly
a safety device—it
Talking of batteries, every circuit needs a source of electricity. Electrons will not doesn’t normally play a
flow on their own but need to be pushed. The push might come from a battery, a role in the circuit.
generator, or a wall socket (which in turn gets electricity from a far-away genera-

215
Chapter 7 | Sound recording and reproduction

tor). A power source such as a battery has an electric potential, also called a voltage
and measured in volts, which is a measure of the potential for that source to drive
electricity around a circuit. Electric potential is always measured between two dif-
ferent points, such as the two terminals of a battery, and a source with a high electric
potential, like a 110 volt wall socket, can produce more electric flow than one with
low potential, like a 1.5 volt battery. The electric potential does not measure how
much electricity is actually flowing in a circuit. Rather, it is a measure of how much
flow one could potentially get. Nonetheless, it plays an important role. If you want
to know how big an electric shock you will get from an outlet, for instance, then it is
the voltage you need to know about: you need to know whether the outlet has the
potential to create a large flow of electricity when you touch it. If it doesn’t then you
will not get a shock.
The actual amount of electricity flowing
around a circuit is called the current. Cur-
rent is measured at a single point, which can
be anywhere in the circuit, and answers the
question “How much electricity is flowing
past this point per second?” In principle, one
could measure current in terms of number
of electrons per second, but following long-
established tradition (going back to well be-
fore we even knew what electrons were) cur-
rent is actually measured in amps. Techni-
cally an amp is equal to a flow of 6.2 × 1018
electrons per second, or six billion billion,
though this will not be important for our
purposes. An amp is merely a convenient
unit of current that everyone agrees on.

7.1.1 Sound as an electrical waveform

Recording technology captures sound wave-
forms as electrical signals by varying the
Figure 7.2: Representation of a sound waveform as an elec- voltage to represent the shape of the wave-
trical signal. At the most fundamental level, all sound technology form. Take a look at Fig. 7.2. The top graph
represents sound waveforms as varying voltages of size from a few shows the sound pressure for a particular
millivolts up to a few volts. The shape of the sound waveform (top) sound. The bottom one shows the corre-
is exactly mirrored in the variation of the voltage (bottom).
sponding voltage varying over time with the
exact same pattern. When one goes up or
down, so does the other. This is how sound is converted into an electrical signal.
Given the right equipment we can also turn it back into sound again and this is the

216
 7.2 | Electrical circuits

principle behind all recording, amplification, and broadcast of sound and music.

7.2 Electrical circuits

Electrical circuits of the kind used in music recording and amplification typically
consist of a set of components, arranged around one or more loops, along with a A resistor
power source to drive the electricity, such as a battery. The three main types of
components found in circuits are resistors, capacitors, and transistors. Let us look at
what each of these does, starting with resistors.

7.2.1 Resistors
A resistor is an electrical component that, as its name suggests, impedes or resists the
flow of electricity, reducing the amount that flows around a circuit. At first glance
this might seem like a strange thing to do, but in most cases we do not want large Two alternative symbols
electric currents flowing around our circuits. Normally it’s more convenient, not to use to represent resistors
mention safer, to work with small ones, and resistors play a central role in almost all in circuit diagrams.
circuits by allowing us to reduce and control the size of the flow.
Physically, a resistor is a small cylinder a few millimeters long with
two wires coming out of its two ends. When we draw pictures of elec- I
trical circuits, however, we conventionally represent resistors and other
components by stylized symbols. The symbol for a resistor is either a
zigzag line or a rectangular box, with straight lines coming out the ends +
V R
to represent the two connecting wires. Both symbols are widely used. −
Here we will use the zigzag line.
Figure 7.3 shows perhaps the simplest possible circuit we could
have, which consists of just two components, a battery and a resistor.
The battery is represented by the symbol on the left, consisting of two Figure 7.3: A simple circuit. This
bars, one long and one short, and the resistor is on the right. circuit consists of a battery with volt-
The amount by which a resistor resists the flow of electricity is age 𝑉 on the left and a single re-
called its resistance and is measured in ohms, denoted by the Greek sym- sistor with resistance 𝑅 on the right.
The battery produces a current 𝐼 that
bol Ω (“omega”). Thus we might say a resistor had resistance 250 ohms
flows around the circuit in the direc-
or 250 Ω. The higher the resistance of a resistor, the less current will tion shown and through the resistor.
flow through it, with the current obeying Ohm’s law of resistance.1 Re-
ferring to Fig. 7.3, the voltage across the resistor, from one side to the
other, is in this case just equal to the voltage 𝑉 of the battery, and if the resistor has
resistance 𝑅 ohms then Ohm’s law says that the current in amps flowing around the

1 Not to be confused with Ohm’s law of acoustics, discussed in Section 4.5, which refers to the ear’s

inability to hear changes in the phase of a sound.

217
Chapter 7 | Sound recording and reproduction

circuit and through the resistor is

𝑉
𝐼= , (7.1)
𝑅
with the current flowing from the positive side of the battery around the circuit to the
negative side, as shown by the arrow in the figure. This law applies to all resistors:
the current in a resistor is always equal to the voltage across the resistor divided by
the resistance.
Suppose for example that we have a 1.5 volt battery and a resistor of 250 Ω. Then
Note that current is con- the current flowing around the circuit and through the resistor will be
ventionally denoted by
the letter 𝐼 and not, for 1.5
𝐼= = 0.006 amps = 6 milliamps, (7.2)
instance, by 𝐶. The use of 250
an 𝐼 has its origin in the which is typical of the small currents used in audio circuits. One amp is a very large
French word for current,
intensité. current. Currents of a few milliamps are more normal.
The equation for Ohm’s law, Eq. (7.1), tells us that the current increases with the
voltage 𝑉 —more volts mean more current—but decreases as the resistance gets larger.
Thus we can use the resistor to regulate the current flowing in our circuit. Resistors
are manufactured with resistance values everywhere from one ohm to billions of
ohms, allowing us wide latitude in our choice of current.

7.2.2 The potential divider or volume control

Now consider the circuit shown in Fig. 7.4. This circuit is more com-
plicated in that it has two resistors with resistances 𝑅1 and 𝑅2 and the
current from the battery has to pass through both resistors in order to
complete the circuit. We say the resistors are wired in series. Resis-
R1 tors in series obey a simple rule: each one resists the flow according its
own resistance and the combined effect is that of a single resistor with
+
V resistance 𝑅1 +𝑅2 . That is, we simply add the resistance values together.
− This allows us to calculate the current flowing around this circuit
using Ohm’s law, Eq. (7.1). We set 𝑅 = 𝑅1 + 𝑅2 and get
R2
𝑉
𝐼= . (7.3)
𝑅1 + 𝑅2
Suppose, for instance, that the battery again has voltage 1.5 volts and
the two resistors have the same resistance 𝑅1 = 𝑅2 = 250 Ω. Then the
Figure 7.4: A circuit with two resis- current around the circuit is
tors. This circuit consists of a battery
with voltage 𝑉 and two resistors in se-
1.5
𝐼= = 0.003 amps = 3 milliamps. (7.4)
ries with resistances 𝑅1 and 𝑅2 . 250 + 250
But now let us ask a different question: what is the voltage 𝑉1 across the
resistor 𝑅1 if we measure the electrical potential between its two terminals? (Recall

218
 7.2 | Electrical circuits

that voltages are always measured between two points.) We can answer this question
by again using Ohm’s law, which links the current flowing in the resistor to the
voltage across it. The same current 𝐼 flows through both resistors and for resistor 𝑅1
the current is related to the voltage 𝑉1 according to 𝐼 = 𝑉1 /𝑅1 . Rearranging, we find
that
𝑉1 = 𝐼𝑅1 = 0.003 × 250 = 0.75 volts. (7.5)
We can do the same calculation for the voltage across the second resistor and we find
that
𝑉2 = 𝐼𝑅2 = 0.003 × 250 = 0.75 volts. (7.6)
Thus each of the resistors has a smaller voltage across it in this case, both of size
0.75 volts, which is half the size of the original 1.5 volts from the battery. This means
that the two voltages on the resistors add up to the original battery voltage: 0.75 +
0.75 = 1.5 volts. This is not a coincidence: it is always true in any circuit, no matter
how it is wired or what components it contains. The voltages always add up in this
way. This will be a useful rule of thumb. We will see that we can use it to quickly
calculate unknown voltages.
If we change the voltage on the battery, the same prin-
ciples still apply. Whatever voltage comes out of the bat- R1
tery in Fig. 7.4, we will get a voltage across each of the
resistors that is a half the size. More generally, for any
choice of the resistances 𝑅1 and 𝑅2 , the current 𝐼 is given
by Eq. (7.3) and the voltage across 𝑅2 is Vin R2 Vout
𝑅2
𝑉2 = 𝐼𝑅2 = 𝑉. (7.7)
𝑅1 + 𝑅2
Thus this circuit acts as a potential divider: it takes what-
ever voltage you start with and cuts it down by a factor Figure 7.5: A volume control. The potential di-
of 𝑅2 /(𝑅1 + 𝑅2 ). By choosing the values of the two resis- vider circuit of Fig. 7.4 can be rearranged to act as
tors we can arrange this factor to be any number we like a volume control. It takes an input waveform in the
between zero and one. This is useful for instance if you form of a time varying voltage 𝑉in and cuts it down
by a factor of 𝑅2 /(𝑅1 + 𝑅2 ), which can be varied by
need a smaller voltage than the one the battery gives you.
changing the resistance 𝑅2 .
If you have a 1.5 volt battery but you want 0.75 volts then
you can achieve this by using a potential divider.
Taking things a step further, suppose the voltage on the left of the circuit is not a
steady voltage from a battery at all, but a time-varying voltage representing a musical
waveform, as discussed in Section 7.1. The divider will take that voltage and cut it
down by precisely the fraction 𝑅2 /(𝑅1 + 𝑅2 ). The shape of the waveform will stay the
same but everything about it will be reduced in size by this same factor. This smaller
waveform will then produce quieter music when we turn the signal back into sound
again, so the circuit is acting as a volume control.

219
Chapter 7 | Sound recording and reproduction

When used in this way, the potential divider circuit is usually drawn slightly
differently, as shown in Fig. 7.5. It’s still the same circuit, just rearranged. The input
waveform arrives on the left as shown and the output waveform is produced on the
right, with
𝑅2
𝑉out = 𝑉in . (7.8)
𝑅1 + 𝑅2
We have also made one other change to the circuit: the zigzag line with an arrow
through it is the symbol for a variable resistor, a standard component that acts as a
resistor but has a knob that you can turn to change the amount of resistance. By
turning the knob we can vary 𝑅2 and hence vary the amount the circuit cuts the
signal down by. This allows us to vary the volume of the sound all the way from
full volume—the entire original signal—when 𝑅2 is very large, to zero when it is very
small. This is precisely how a volume control operates.

Example 7.1: Volume control

A volume control like the one in Fig. 7.5 has 𝑅1 = 1200 Ω and a variable resistor for 𝑅2 . If
the input voltage is 𝑉in = 1 volt and 𝑅2 is set to 400 Ω, what is the output voltage?
To answer this question we apply Eq. (7.8), which gives
𝑅2 400
𝑉out = 𝑉in = × 1 = 0.25 volts. (7.9)
𝑅1 + 𝑅2 1200 + 400
Now what value would we have to set 𝑅2 to if we wanted the output voltage to be
0.6 volts? To answer this question, we rearrange Eq. (7.8) to give 𝑅2 thus:
𝑉out
𝑅2 = 𝑅1 . (7.10)
𝑉in − 𝑉out
Plugging in the values we then have
0.6
𝑅2 = × 1200 = 1800 Ω. (7.11)
1 − 0.6

7.2.3 Capacitors
A capacitor is an electronic component that stores a small amount of electrical charge.
More correctly, it stores two charges, one positive and one negative, of equal size.
(Electrical charge can be both positive and negative. Like savings and debt, positive
and negative charges cancel out when they meet.) Physically a capacitor consists of
two metal plates a short distance apart inside a casing, often with a fluid or other
insulating medium in between. In practice you can’t see the plates, just the casing,
A capacitor and there are two wires, one attached to each plate, that are used to connect the
capacitor to a circuit.
When you connect a capacitor to a battery or other power source, a current flows,
pushing electrons onto one plate of the capacitor and pulling an equal number off

220
 7.2 | Electrical circuits

the other. There is a limit to how much the plates will take, however, and when that
limit is reached no more current will flow, so in practice what we see is an initial
burst of current, then nothing after that.
The result is an electric charge on the two plates in the form of excess electrons
on one and a corresponding deficit on the other. The charges on the plates are equal
and opposite, so the net amount of charge on a capacitor is actually always zero, but
the amount on each plate is not.
On the plate connected to the positive side of the battery there is a positive charge.
As with current we could measure the charge by counting electrons, but by long-
standing tradition we instead measure it in units of coulombs (denoted C). A coulomb
is the charge of 6.2×1018 electrons. Let us denote by 𝑄 the charge in coulombs on the
positive plate. The other plate, the one connected to the negative side of the battery,
then has a negative charge of −𝑄 in coulombs.
The specific value of 𝑄 for a capacitor depends on the applied voltage 𝑉 according
to the capacitor law
𝑄 = 𝐶𝑉 , (7.12)
where 𝐶 is the capacitance, a measure of the storage ability of the capacitor. Capaci-
tance is measured in farads (denoted F). One farad is a large amount of capacitance:
real capacitors come nowhere near this capacity. A typical value in practice is one
microfarad, i.e., one millionth of a farad, denoted 𝜇F.
Now take a look at the circuit in Fig. 7.6, which contains a battery,
a resistor, and a capacitor. The symbol for a capacitor is two parallel
lines as shown—a simplified representation of the two plates. Suppose
that initially the battery is not connected but then we connect it up to
power the circuit. What will happen? R
+
Before the battery is connected there will be no charge on the capac- V
itor, so 𝑄 = 0. Rearranging Eq. (7.12), we find that 𝑉 = 𝑄/𝐶, so the volt- -
age across the capacitor will be zero as well, which makes sense since
C
there is no power source to produce any voltage yet. Now we connect
the battery. When we first connect the battery the resistor must have
the entire voltage 𝑉 across it because, as we have said, voltages always
add up—the voltages across the resistor and the capacitor have to add Figure 7.6: A circuit using a capac-
up to 𝑉 and the voltage across the capacitor is zero. But if there is volt- itor. This circuit consists of a battery
age 𝑉 across the resistor then we will have a current flowing through with voltage 𝑉 and a resistor 𝑅 and ca-
the resistor as given by Ohm’s law 𝐼 = 𝑉 /𝑅 (Eq. (7.1)). pacitor 𝐶.
As this current flows out of the resistor it flows into the capacitor
and charges it up as described above, and as the capacitor charges up the voltage
across it will start to grow: the equation 𝑉 = 𝑄/𝐶 tells us that the more charge
we have the larger the voltage. This means that the voltage across the resistor will
decrease (because the voltages have to add up to 𝑉 ) and hence the current flowing

221
Chapter 7 | Sound recording and reproduction

1.5

Voltage
0.5

Input voltage
Output voltage
0

Time

Figure 7.7: The effect of the circuit in Fig. 7.6. When the input voltage 𝑉 changes to a new
value, the output voltage—the voltage across the capacitor—follows, but with a time lag.

will also decrease.

Overall, therefore, we get an initial rapid flow of current, charging the capacitor,
but over time the current will die off. Eventually, when the charge on the capacitor
becomes large enough, the current will stop altogether. This happens when 𝑄 = 𝐶𝑉 ,
so that the entire voltage 𝑉 is across the capacitor and the voltage across the resistor
is zero, producing zero current.
Another way to say this is that the voltage across the capacitor copies the voltage
of the battery, but with a lag. It starts at zero when there is no battery connected,
then after the battery is connected it rises, until it reaches the battery voltage and
then it stops there. Figure 7.7 shows what this looks like in graph form.
Now suppose the voltage of the battery changes. This does not normally happen
with actual batteries, but the voltages used to represent sound signals vary a lot, so
let’s take a look. If the voltage changes, then again the voltage on the capacitor will
copy that change but with a lag—see Fig. 7.7 again. If the voltage changes repeatedly,
as shown in the figure, then the capacitor will copy every change, but with a lag.
The size of the lag is determined by how long it takes the capacitor to charge
up. If the capacitor has a very large capacitance 𝐶 it will take longer. Twice the
capacitance means it takes twice as long, so the lag is proportional to 𝐶. At the same
time it also depends on the resistance 𝑅, since more resistance means less current
and hence slower charging. Twice the resistance means twice as slow, so the lag is
also proportional to 𝑅. In Section 7.2.5 we show that in fact the time lag, also called
the time constant of the circuit, is equal to 𝑅𝐶. So with a 1 𝜇F capacitor, for instance,

222
 7.2 | Electrical circuits

and a 2000 Ω resistor (also called 2 kilohms or 2 kΩ) the time constant would be

𝑅𝐶 = 2000 × 10−6 = 0.002 seconds. (7.13)

7.2.4 Filters
The circuit of Fig. 7.6 turns out to be musically useful. To
see why, let’s start by rearranging it into the form shown R
in Fig. 7.8. This is just the same circuit as before, but re-
organized to accept an input voltage 𝑉in and produce an
output voltage 𝑉out , similar to the way we drew the poten-
tial divider circuit in Fig. 7.5. Typically the input voltage Vin C Vout
would be a signal representing a sound waveform. And
what would the output voltage look like? Figure 7.9 shows
two examples. In both, the input voltage is a sine wave,
but with different frequencies. In the top panel the fre-
quency is a slow 20 Hz, meaning we have one cycle every Figure 7.8: A low-pass filter. The circuit of Fig. 7.6
0.05 seconds, at the very low end of the musical frequency reorganized as a low-pass filter.
range. Meanwhile, we have made the time constant of
the circuit—the size of the lag between output and input—
equal to 𝑅𝐶 = 0.002 seconds as above. Thus, in this case, the lag is much shorter than
the period of the wave and hence has very little effect. The output voltage lags just
a short distance behind the input and ends up following the input voltage closely, so
the difference between the two is barely visible in the figure.
In the lower panel of the figure the input sine wave has a much higher frequency
of 2000 Hz, near the highest of musical notes, while the time constant of the circuit is
still 0.002 seconds. Now we have one cycle of the wave every 0.0005 seconds, much
less than the time constant. This means that the output voltage does not have nearly
enough time to catch up with the input. It needs around 0.002 seconds to catch up,
but the input signal changes long before this as it goes through its cycle. The result
is that the output never manages to catch up with the input, so the output signal is
much smaller than the input.
Overall then, the effect of this circuit is that when we feed in a low-frequency
signal it comes out almost unchanged, but when we feed in a high-frequency one
very little gets through to the output. We call such a circuit a filter. Specifically,
this is a low-pass filter, a circuit that lets low-frequency signals though but cuts out
high-frequency ones.
Now recall that, as discussed in Section 4.3, any sound can be represented as a
combination of sine waves. Each sine wave in a sound will be affected by the filter
according to its own particular frequency, so for a complex waveform made of many
sine waves the filter will cut down on the high-frequency ones and leave the low-

223
Chapter 7 | Sound recording and reproduction

Voltage
0 100 200 300

0 2 4 6 8 10
Time (milliseconds)

Figure 7.9: The effect of the low-pass filter. Top: when a low-frequency sine-wave signal
is fed into the circuit of Fig. 7.8 it emerges from the output essentially unchanged—there is a
slight lag between the two, but the effect is small. Bottom: when a high-frequency signal is
fed in only a small signal comes from the output. Note that the horizontal scale is different in
the two panels.

frequency ones unchanged. In musical terms, this has the effect of cutting down the
amount of treble while leaving the bass as it is. This is extremely useful for many
purposes. Filters like this are used for instance as tone controls on amplifiers, for
sound shaping in electronic synthesizers, and for many things in recording studios.
With a little mathematics we can calculate the exact effect of our low-pass filter
on an incoming electrical signal. The details are given in Section 7.2.5 but the crucial
result is that for a sine-wave input the intensity of the corresponding sound obeys
the equation
𝐼 in
𝐼 out = . (7.14)
1 + (2𝜋𝑅𝐶 𝑓 ) 2
Whatever the intensity 𝐼 in of the input signal, the output signal has the same intensity
but divided by 1 + (2𝜋𝑅𝐶 𝑓 ) 2 . Figure 7.10 shows what this looks like as the frequency
varies. Look first at the middle curve in the figure. At low frequencies 𝐼 out is equal
to 𝐼 in —the filter has no effect and whatever goes in comes out unchanged. But as the

224
 7.2 | Electrical circuits

0.8
Intensity Iout/Iin

0.6
fc = 125 Hz fc = 500 Hz fc = 2000 Hz

0.4

0.2

0
30 100 300 1000 3000 10000
Frequency (Hz)

Figure 7.10: Frequency response of a low-pass filter. The intensity of the output signal
from the low-pass filter of Fig. 7.8 as a function of frequency for three different values of the
cutoff frequency: 𝑓𝑐 = 125 Hz, 500 Hz, and 2000 Hz.

frequency gets higher the intensity starts to fall off and for the highest frequencies
it is virtually zero.
We can work out the frequency where the intensity falls off by looking at Eq. (7.14).
If 𝑓 = 0 we see that 𝐼 out = 𝐼 in , so the filter has no effect, while if 𝑓 is very large then
𝐼 out will be almost zero. In between these two lies a point where 𝐼 out = 12 𝐼 in , so that
the filter cuts the intensity down by exactly a half. This half-way point is called the
cutoff frequency of the filter, denoted 𝑓𝑐 . It’s straightforward to see that the cutoff
frequency corresponds to the point where (2𝜋𝑅𝐶 𝑓 ) 2 = 1 and, rearranging for 𝑓 , we
find that
1
𝑓𝑐 = . (7.15)
2𝜋𝑅𝐶
Roughly speaking, sound below this frequency gets through the filter and sound
above it gets blocked.
We can vary the cutoff frequency of the filter by varying either 𝑅 or 𝐶 (or both).
Figure 7.10 shows curves for three different choices, 𝑓𝑐 = 125 Hz, 500 Hz, and 2000 Hz,
as indicated by the vertical dashed lines. As we can see, the curve simply gets shifted
from side-to-side as the cutoff frequency varies.
Another way to look at the low-pass filter is in terms of the number of decibels

225
Chapter 7 | Sound recording and reproduction

Sound level decrease (dB)

6d
−5

Bp
er
oct
ave
−10
fc = 125 Hz fc = 500 Hz fc = 2000 Hz

−15

−20
30 100 300 1000 3000 10000
Frequency (Hz)

Figure 7.11: Sound level decrease upon applying the low-pass filter. The amount in
decibels by which the low-pass filter of Fig. 7.8 decreases the sound level of a signal, as a
function of frequency for three different values of the cutoff frequency: 𝑓𝑐 = 125 Hz, 500 Hz,
and 2000 Hz.

it cuts out. Combining Eqs. (7.14) and (7.15) we can write

𝐼 out 1
= , (7.16)
𝐼 in 1 + 𝑓 2 /𝑓𝑐2
and substituting into Eq. (3.16) on page 60, we find the change in the decibel level of
the sound to be  
𝑓2
𝐿out − 𝐿in = −10 log 1 + 2 . (7.17)
𝑓𝑐
Figure 7.11 shows a plot of this formula for the same values of 𝑓𝑐 as in Fig. 7.10, with
the cutoff frequency corresponding roughly to the point where the filter transitions
from allowing the signal to pass unchanged (zero decibels reduction—the flat part of
the curve on the left) to cutting it off (the falling curve on the right).
Above the cutoff frequency, where 𝑓 /𝑓𝑐 is much greater than one, we can ignore
the 1 in the denominator of Eq. (7.16) and to a good approximation

𝐼 out 𝑓𝑐2
= 2, (7.18)
𝐼 in 𝑓

so the output intensity well above 𝑓𝑐 is inversely proportional to 𝑓 2 . Thus if the

frequency doubles the intensity goes down by a factor of four, or equivalently by 6 dB.

226
 7.2 | Electrical circuits

R R

Vin C C Vout

Figure 7.12: A 12 dB-per-octave low-pass filter. This filter consists of two low-pass filters,
one after another, each cutting out 6 dB per octave, for a total of 12 dB per octave.

In other words, the sound level falls by 6 dB per octave. This behavior is shown as
the diagonal dashed line on the right of Fig. 7.11 and, as we can see, the curves do
indeed have roughly this slope.
There is a wide range of uses for low-pass filters, with all kinds of values of the
cutoff frequency. Filters with very high cutoff around 𝑓𝑐 = 20 000 Hz are used as anti-
aliasing filters and reconstruction filters to remove unwanted ultrasonic frequencies
from music. Filters with intermediate values of 𝑓𝑐 are used as tone controls on ampli-
fiers and electric guitars, for equalization or “EQ” in studio recording and mixing, as
sound shaping filters in synthesizers, and as crossover filters in loudspeaker systems.
Filters with an ultra-low cutoff of 10 Hz or less are used as compensation filters for
tape recorders, record players, or dynamic microphones, that otherwise produce too
much high-frequency sound.
Sometimes we may want a filter that cuts out more of the high frequencies than
the 6 dB-per-octave filter described here. In that case, we can string two or more
filters together one after another, as shown in Fig. 7.12. If we have two filters and
each one cuts out 6 dB per octave, then together they will cut 12 dB per octave. Three
filters will cut 18 dB per octave and four will cut 24 dB per octave. You can find each
of these filter types in use in certain situations. For instance, the distinctive sounds
of different brands of electronic synthesizers are in part due to the different filters
they use. Moog brand synthesizers have long made use of 24 dB-per-octave filters,
but other successful brands such as Oberheim and Sequential Circuits use 12 dB-per-
octave filters. This gives the former a warmer sound and the latter a brighter sound.
(We study the workings of synthesizers and their use of filters in Chapter 15.)
Another variant of our filter circuit is shown in Fig. 7.13. The only difference
between this and the low-pass filter is that the resistor and the capacitor are the
opposite way around, but it turns out that this makes all the difference in the world.
When we feed a sine-wave signal into the input of this circuit, the voltage across
the capacitor decreases as frequency goes up just as before. But this means that the

227
Chapter 7 | Sound recording and reproduction

voltage across the resistor—which plays the role of output signal in this circuit—must
go down, since the two voltages always to have to add up to 𝑉in . Thus this circuit is
a high-pass filter. It produces a small signal at low frequencies but a large signal at
high frequencies.
In Section 7.2.5 it is shown that the intensity of the out-
put from the high-pass filter circuit is

(2𝜋𝑅𝐶 𝑓 ) 2 𝐼 in
𝐼 out = 𝐼 in = , (7.19)
C 1 + (2𝜋𝑅𝐶 𝑓 ) 2 1 + 𝑓𝑐2 /𝑓 2
Vin R Vout where the cutoff frequency 𝑓𝑐 is again given by
1
𝑓𝑐 = . (7.20)
2𝜋𝑅𝐶
Figure 7.14 shows a plot of this output intensity against
Figure 7.13: A high-pass filter. This circuit acts as frequency—the equivalent for the high-pass filter of
a high-pass filter, cutting out 6 dB per octave at low Fig. 7.10—and the high-pass behavior is clearly visible. At
frequencies but leaving high frequencies unchanged. low frequencies the filter cuts 6 dB per octave, then it flat-
tens out at frequencies above 𝑓𝑐 . If a stronger effect is
needed one can string two or more high-pass filters together to get 12 dB per oc-
tave or more.

0.8
Intensity Iout/Iin

0.6
fc = 125 Hz fc = 500 Hz fc = 2000 Hz
0.4

0.2

0
30 100 300 1000 3000 10000
Frequency (Hz)

Figure 7.14: Frequency response of a high-pass filter. The intensity of the output signal
from the high-pass filter of Fig. 7.13 as a function of frequency for cutoff frequencies 𝑓𝑐 =
125 Hz, 500 Hz, and 2000 Hz.

228
 7.2 | Electrical circuits

C2
R1

Vin C1 R2 Vout

| {z } | {z }
Low-pass filter High-pass filter

Figure 7.15: A band-pass filter. A band-pass filter consists of a low-pass filter and a high-
pass filter, one after another.

High-pass filters are used as bass tone controls in amplifiers, for studio equaliza-
tion, in crossover circuits for loudspeakers, and, more rarely, for sound shaping in
synthesizers. A more specialized use of a high-pass filter, which will be important to
us occasionally, is for filtering out constant voltages. If we reduce the frequency of
an electrical signal all the way to zero the signal no longer changes and we have a
constant voltage, so a constant voltage can be thought of as a signal with frequency
zero. Sometimes we want a circuit that blocks constant signals while letting varying
ones pass, and we can do this with a high-pass filter. Putting 𝑓 = 0 in Eq. (7.19) we
see that a high-pass filter will produce zero output from a constant input signal, no
matter what the cutoff frequency 𝑓𝑐 . So a high-pass filter with a very low cutoff—
say 1 Hz—will block a constant signal but leave sound waveforms (with frequencies
of 20 Hz and more) untouched. This trick is used, for example, in the condenser
microphone—see Section 7.3.5.
A third type of filter is the band-pass filter, which consists of a low-pass filter
and a high-pass filter, one after another, as shown in Fig. 7.15. The low-pass filter
cuts out high frequencies and the high-pass filter cuts out low frequencies, leaving
only the frequencies in the middle. By choosing the cutoffs of the low- and high-pass
stages, 𝑓 low and 𝑓 high , one can tune the bandwidth of filter—the range of frequencies
it allows to pass. Figure 7.16 shows the frequency response of the filter in Fig. 7.15
when the cutoffs are set at 𝑓 low = 125 Hz and 𝑓 high = 2000 Hz. Band-pass filters
find use in graphic equalizers and other studio tools for shaping and tuning recorded
sound, in crossover circuits for loudspeaker systems, and for creating narrowband
noise as described in Section 4.4.1.

229
Chapter 7 | Sound recording and reproduction

0.8
Intensity Iout/Iin

0.6
fhigh = 2000 Hz
flow = 125 Hz
0.4

0.2

0
30 100 300 1000 3000 10000
Frequency (Hz)

Figure 7.16: Frequency response of a band-pass filter. The intensity of the output signal
from the band-pass filter of Fig. 7.15 as a function of frequency for cutoff frequencies 𝑓low =
125 Hz and 𝑓high = 2000 Hz.

Example 7.2: A low-pass filter

A 6 dB-per-octave low-pass filter like the one in Fig. 7.8 has 𝑅 = 5100 Ω and 𝐶 = 0.1 𝜇F.
What is the cutoff frequency, and how many decibels does the filter cut from a signal at
1000 Hz? To answer this question, we first use Eq. (7.15) to calculate the cutoff frequency:
1 1
𝑓𝑐 = = = 312 Hz. (7.21)
2𝜋𝑅𝐶 2𝜋 × 5100 × 10 −7
Now the change in decibels for a tone with frequency 𝑓 = 1000 Hz is given by Eq. (7.17)
to be    
𝑓2 10002
𝐿out − 𝐿in = −10 log 1 + 2 = −10 log 1 + = −10.5 dB, (7.22)
𝑓𝑐 3122
so the sound level will drop by just a little over 10 dB.

Advanced material

7.2.5 Freqency response of filters circuits of resistors and capacitors. Take the case of the
low-pass filter of Fig. 7.8. Suppose the current flowing
around the circuit at a certain moment is 𝐼 , so that the volt-
We can derive the frequency response of low-pass, high-
age across the resistor is 𝐼𝑅, following Ohm’s law, Eq. (7.1).
pass, and band-pass filters from the equations governing

230
 7.2 | Electrical circuits

The voltages across the resistor and capacitor have to add Differentiating with respect to time and recalling that
up to the input voltage 𝑉in , so d𝑄/d𝑡 = 𝐼 we then get
𝐼𝑅 + 𝑉out = 𝑉in . (7.23) 𝐼 d𝑉out d𝑉in
+ = . (7.30)
𝐶 d𝑡 d𝑡
At the same time the charge 𝑄 on the capacitor obeys the
capacitor equation, Eq. (7.12), so 𝑄 = 𝐶𝑉out . The rate At the same time, Ohm’s law tells us that 𝐼 = 𝑉out /𝑅, so we
at which charge flows onto the capacitor is equal to the end up with the differential equation
current 𝐼 (since current is by definition the rate at which
d𝑉out d𝑉
charge flows) and hence 𝑅𝐶 + 𝑉out = 𝑅𝐶 in . (7.31)
d𝑡 d𝑡
d𝑄 d𝑉out
𝐼= =𝐶 . (7.24) Again taking a sine-wave input in the form 𝑉in =
d𝑡 d𝑡
𝐴in ei2𝜋 𝑓 𝑡 and assuming a corresponding sine-wave output
Combining Eqs. (7.23) and (7.24) we then find that the out- 𝑉out = 𝐴out ei2𝜋 𝑓 𝑡 , we get
put voltage 𝑉out obeys the differential equation
i2𝜋𝑅𝐶 𝑓 𝐴out + 𝐴out = i2𝜋𝑅𝐶 𝑓 𝐴in, (7.32)
d𝑉out
𝑅𝐶 + 𝑉out = 𝑉in . (7.25)
d𝑡 or
i2𝜋𝑅𝐶 𝑓
Now suppose the input to the circuit is a sine wave 𝐴out = 𝐴in . (7.33)
with frequency 𝑓 , which we represent in conventional 1 + i2𝜋𝑅𝐶 𝑓
manner as the real part of the complex number 𝑉in = The real amplitude is then
𝐴in ei2𝜋 𝑓 𝑡 , where 𝐴in is the amplitude of the signal. Then
2𝜋𝑅𝐶 𝑓
the (steady-state) output of the circuit is another sine wave 𝐴out = p 𝐴in , (7.34)
𝑉out = 𝐴out ei2𝜋 𝑓 𝑡 and, substituting into Eq. (7.25) and can- 1 + (2𝜋𝑅𝐶 𝑓 ) 2
celling some factors, we find that i2𝜋𝑅𝐶 𝑓 𝐴out +𝐴out = 𝐴in ,
and the intensity of the resulting sound obeys
or equivalently
𝐴in (2𝜋𝑅𝐶 𝑓 ) 2
𝐴out = . (7.26) 𝐼 out = 𝐼 in , (7.35)
1 + i2𝜋𝑅𝐶 𝑓 1 + (2𝜋𝑅𝐶 𝑓 ) 2
In general this is a complex number. The actual real ampli- as stated previously in Eq. (7.19).
tude of the output signal is given by the magnitude Other filters can be built from combinations of the ba-
sic low- and high-pass filters. For instance, if we string two
𝐴in
𝐴out = p . (7.27) low-pass filters together one after another as in Fig. 7.12,
1 + (2𝜋𝑅𝐶 𝑓 ) 2 the output of the first becomes the input to the second.
Each filter then multiplies the intensity by the factor in
When the electrical signal is transformed back into sound,
Eq. (7.28) and we get
it will produce an intensity proportional to the square of
the amplitude (see Section 3.1.3), so 𝐼 = 𝑘 |𝐴| 2 for some 𝐼 in
constant 𝑘. This means that 𝐼 out = , (7.36)
[1 + (2𝜋𝑅𝐶 𝑓 ) 2 ] 2
2
2 𝑘 𝐴in 𝐼 in which gives us a 12 dB-per-octave low-pass filter by con-
𝐼 out = 𝑘 𝐴out = = , (7.28)
1 + (2𝜋𝑅𝐶 𝑓 ) 2 1 + (2𝜋𝑅𝐶 𝑓 ) 2 trast with the 6 dB per octave we get out of a single-stage
filter. If we string a low-pass and a high-pass filter together
as stated in Eq. (7.14).
as in Fig. 7.15 we get
For the high-pass filter of Fig. 7.13 the analysis is sim-
ilar. The voltages across capacitor and resistor must still 
1

(2𝜋𝑅2𝐶 2 𝑓 ) 2

add up to 𝑉in and Eq. (7.12) tells us that the voltage across 𝐼 out = 𝐼 in, (7.37)
1 + (2𝜋𝑅1𝐶 1 𝑓 ) 2 1 + (2𝜋𝑅2𝐶 2 𝑓 ) 2
the capacitor is equal to 𝑄/𝐶, so
𝑄 which gives the band-pass frequency response shown in
+ 𝑉out = 𝑉in . (7.29) Fig. 7.16.
𝐶

231
Chapter 7 | Sound recording and reproduction

V+

IC = β IB
Collector

Base

IB
Emier Vout
Vin

IE = IB + IC

0 volts

(a) (b)

Figure 7.17: Transistor circuits. (a) When a small current 𝐼𝐵 flows through the base of a
transistor, it causes a much larger current 𝐼𝐶 = 𝛽𝐼𝐵 to flow through the collector. (b) A typical
circuit diagram for a single-stage transistor amplifier. In addition to the transistor itself, shown
in the center, the two resistors on the left act as a potential divider to center the voltage at the
base of the transistor, while the two on the right act to convert currents into voltages via
Ohm’s law. The two capacitors act as high-pass filters to remove constant voltages from the
input and output signals. The top line or “rail” is attached to a battery or other power supply
with voltage 𝑉+ , which provides power to the circuit.

7.2.6 Transistors
The third basic component of electronic circuits is the transistor. Invented at Bell
Labs in the 1940s, the bipolar junction transistor is a device with three connections—
not just two like the resistor and capacitor—whose function is to amplify electrical
signals, taking small signals and making them bigger. This is a crucial step in sound
recording and live audio and the invention of the transistor revolutionized how music
is recorded and performed.2
A transistor A transistor is represented in electrical circuits by the circular symbol shown

2 Before the invention of the transistor, electronic devices made use of a different technology, the

vacuum tube or valve, which can also be used as an amplifier but is significantly inferior to the transistor
in many ways and substantially limits the practical amount of power that amplifiers can generate. Vacuum
tubes have been mostly replaced by transistors in recent decades but do still have some specialized uses,
in part because of their imperfections. For instance, we discuss their use in distorting guitar amplifiers in
Section 14.2.

232
 7.2 | Electrical circuits

in Fig. 7.17a. The three connections to the device each play different roles and are
called the base, collector, and emitter. In normal use, one applies a voltage between
the collector and emitter, but initially no current flows because the transistor has a
high resistance. If, however, one sends a small current into the base and out through
the emitter, then the resistance of the device changes and current starts to flow from
collector to emitter. The crucial observation is that the current 𝐼𝐶 flowing through
the collector is much larger than the current 𝐼𝐵 flowing through the base, so that the
transistor acts as an amplifier. Specifically,
𝐼𝐶 = 𝛽𝐼𝐵 , (7.38)
where the quantity 𝛽 is known as the gain of the transistor. A typical value of the gain
is 𝛽 = 100 or so, so the effect is a large one: you can feed in a current of one milliamp
and get out a current of 100 milliamps. By combining several transistors, one after
another, one can create circuits with a huge overall gain, a factor of thousands or
even millions, that can take the tiniest signals, such as those from microphones or
electric guitars, and make them large enough to fill entire arenas.
Transistors are not normally used on their own, but rather in combination with
a selection of resistors and capacitors to make a complete circuit. Figure 7.17b shows
a typical circuit for a one-transistor amplifier. We will not discuss the design of
transistor circuits in detail here—it is a complex topic, the subject of entire books
just on its own—but we will discuss the use of musical amplifiers of various kinds in
Chapter 14.

7.2.7 Electromagnets and induction

In addition to resistors, capacitors, and transistors, recording technology
uses a range of more specialized components specifically for processing Magnetic field
sound, such as microphones and loudspeakers, which make use of elec-
tromagnetic effects for their operation. An electromagnet is a magnet
powered by electricity. If one takes a loop of wire, as shown in Fig. 7.18,
and passes a current around it (for instance by connecting it to a battery), I
the current creates a magnetic field through the loop.
The magnetic field from a single loop of wire, however, is very small—
barely large enough to be detectable and not of much use. But if you put
two loops together you get twice the field, three loops gives three times,
and so forth. By putting many loops together to form a coil one can create Figure 7.18: An electric current in
a powerful electromagnet with a field strong enough to attract metal and a loop of wire produces a magnetic
magnetize other objects. The field can be made even stronger by placing field.
a metal core inside the coil. The core becomes magnetized by the field
and amplifies the effect of the electromagnet. Powerful electromagnets are used in
hospital MRI machines to scan patients’ insides, in electromagnetic door locks, and

233
Chapter 7 | Sound recording and reproduction

to pick up scrap metal in junk yards. They also play a key role in music technology
as a central component of tape recorders, computer hard drives, and loudspeakers.
A practical electromagnetic might consist of hundreds or thousands of loops
of wire around a metal core, creating the coil called a solenoid—see Fig. 7.19. The
strength 𝐵 of the magnetic field produced when current 𝐼 flows around a solenoid is
The tesla unit (like the measured in units of teslas and given by the formula
automobile of the same
name) is named after 𝜇𝑁 𝐼
𝐵= . (7.39)
Serbian engineer Nikola 𝐿
Tesla, who was responsi-
ble for numerous inven- Here 𝐿 is the length of the solenoid from end to end, 𝑁 is the number of turns in
the coil (how many times it loops around the core), and 𝜇 is a quantity called the
tions in the early days of
the commercial electric- magnetic permeability, which measures how much the core amplifies the field. If the
ity industry.
core is made of a material like iron, which has a large permeability, then the field is
greatly amplified and we will get a strong electromagnet. The core can be extended
past the ends of the coil as shown in Fig. 7.20 or bent around to carry the magnetic
field where it is needed, a useful trick in technological applications where space is
cramped. (When we do this the length 𝐿 appearing in Eq. (7.39) is the length of the
coil, not the length of the core.)
Equation (7.39) tells us that the strength of the mag-
I netic field produced by a solenoid is proportional to the
Magnetic current 𝐼 . This is crucial in audio applications of electro-
field magnets. In a loudspeaker, for instance, as we will see in
Section 7.5.1, an electrical signal is sent through an elec-
tromagnet, producing a magnetic field that moves a dia-
phragm, which in turn pushes on the air and makes a
sound. Because the strength of the electromagnet is pro-
L portional to the current in the electrical signal, the move-
ment of the diaphragm is also proportional to the current,
Figure 7.19: A solenoid. A coil with many turns and hence the resulting sound mimics the waveform of the
produces a strong magnetic field, in proportion
electrical signal—exactly what we need to turn electricity
to the number of turns, when a current 𝐼 passes
through it. into sound.
Related to electromagnets is the phenomenon of elec-
tromagnetic induction, which is in some ways the reverse of what an electromag-
net does. In an electromagnet a current passed through a coil generates a magnetic
field; in electromagnetic induction a magnetic field passes through an otherwise in-
ert coil which induces an electric voltage in the coil. The two processes are not quite
analogous though, for two reasons. First, it is current that produces the field in an
electromagnet but induction generates a voltage. Second, with an electromagnet the
strength of the magnetic field depends on the size of the current, but with induction
the voltage depends not directly on the strength of the field but on how fast the field
is changing. A rapidly changing magnetic field produces a large voltage, a slowly

234
 7.3 | Microphones

changing field produces a small voltage, and a stationary magnetic field, one that is Coil Core
not changing, will produce no voltage at all, no matter how strong it is. This turns
out to be a crucial point for musical applications, as we will shortly see.
For those familiar with calculus, the mathematical formulation of electromag-
netic induction says that if the size of the field in teslas inside a coil with 𝑁 turns
is 𝐵, then the voltage 𝑉 generated is

d𝐵
𝑉 = −𝑁 𝐴 , (7.40)
d𝑡
where 𝐴 is the cross-sectional area of the coil. The quantity Φ = 𝐵𝐴 is called the
magnetic flux in the coil, so you may also see the equation written as

dΦ
𝑉 = −𝑁 . (7.41)
d𝑡
Electromagnetic induction is of central importance in sound recording and repro-
duction. For instance, a dynamic microphone works like the reverse of a loudspeaker, Figure 7.20: The core
of a solenoid can be ex-
with a coil attached to a diaphragm and a permanent magnet mounted nearby. When tended beyond the ends
sound strikes the diaphragm it causes it to vibrate, moving the coil, which changes of the coil or bent around
the magnetic field inside the coil and hence generates an electrical signal. We study to carry the magnetic
microphones starting in the following section, and dynamic microphones in Sec- field in particular direc-
tions. This trick is used
tion 7.3.6. for instance in the mag-
netic “head” of a tape
recorder.
7.3 Microphones
Armed with our knowledge of electrical signals and circuits, let us now turn our at-
tention to how music can be captured electrically, recorded, and then reconstituted
as sound again. The first step of the process, turning sound into electrical signals, is
accomplished with a microphone, a device that detects sound and produces a varying
electrical voltage that mirrors the sound waveform.3 The basic scientific principle be-
hind all microphones is the same: a thin membrane or diaphragm moves in response
to the incoming sound in a manner similar to the movement of the eardrum in the ear
(see Section 5.2.2), then this movement is translated into a corresponding electrical
voltage via one of several mechanisms that we will describe shortly.

3 In Chapter 15 we discuss electronic instruments, such as synthesizers, which directly generate

an electrical signal without first producing any sound. For such instruments there is no need to use a
microphone—we already have our electrical signal. For traditional musical instruments, however, includ-
ing string and wind instruments, percussion, and voice, sound must be first captured with a microphone.
A few instruments, most notably the electric guitar, are intermediate cases. The electric guitar directly
produces an electric signal and so in principle no microphone is needed, but in practice the instrument
is usually played through a guitar amplifier that produces sound and then the sound is captured with a
microphone.

235
Chapter 7 | Sound recording and reproduction

7.3.1 Pressure microphones

The most fundamental type of microphone is the pressure microphone, which directly
captures sound pressure. It works by combining a diaphragm with a sealed capsule,
a small container of air. The arrangement is shown in Fig. 7.21.
The diaphragm is a thin, flexible membrane, normally
P0 + p made of mylar, a durable plastic. The diaphragm, which
is usually circular, forms one wall of the capsule, and the
remaining walls are made of rigid plastic or metal. The
diaphragm is stretched tightly across the open end of the
Diaphragm
capsule, sealing it off from the outside air.
Capsule When sound arrives at the microphone it causes a
change in air pressure. Recall that air pressure 𝑃 is equal
to the prevailing atmospheric pressure 𝑃0 plus the sound
Figure 7.21: A pressure microphone. In a pres-
sure microphone the capsule is sealed so that sound pressure 𝑝. Inside the capsule, however, the pressure is
only reaches the outside of the diaphragm (the top just 𝑃 0 , since the capsule is sealed off. The result is that
in this figure) making the outside pressure 𝑃0 +𝑝 dif- the pressure differs on the outside and inside of the dia-
ferent from the inside pressure 𝑃 0 and causing the phragm. If 𝑝 is positive, meaning the pressure is greater
diaphragm to bow in or out. outside the capsule, it pushes on the diaphragm causing it
to bow inward as shown in Fig. 7.21. If 𝑝 is negative the
pressure is greater on the inside of the capsule and causes the diaphragm to bow out-
ward. In other words, the movement of the diaphragm reflects the net force acting
on it.
If the pressure on the front of the diaphragm is 𝑃 𝑓 and its area is 𝑆, then the force
on the front is 𝐹 𝑓 = 𝑃 𝑓 𝑆. Similarly the force on the back is 𝐹𝑏 = 𝑃𝑏 𝑆, and the net
force 𝐹 is the difference of the two:

𝐹 = 𝐹 𝑓 − 𝐹𝑏 = 𝑃 𝑓 𝑆 − 𝑃𝑏 𝑆 = (𝑃 𝑓 − 𝑃𝑏 )𝑆. (7.42)

In this case we have 𝑃 𝑓 = 𝑃0 + 𝑝 and 𝑃𝑏 = 𝑃0 , so

𝑃 𝑓 − 𝑃𝑏 = 𝑃0 + 𝑝 − 𝑃0 = 𝑝, (7.43)

and the force is simply

𝐹 = 𝑝𝑆, (7.44)
proportional to the sound pressure. The force is also proportional to the area 𝑆 of the
diaphragm, meaning that a larger diaphragm will feel a larger force. High-quality
microphones typically have a large diaphragm for this reason, to make them more
sensitive to quiet sounds. There are limits to how large the diaphragm can be how-
ever, for several reasons. Larger diaphragms are more expensive and they are also
fragile and therefore may not be suitable for rougher conditions such as live stage

236
 7.3 | Microphones

performance. The main limit on the size of the diaphragm, however, is that the cap-
sule needs to be smaller than the wavelength of the sound if the microphone is to
operate properly. This point is discussed further below.
The diaphragm is light enough that it responds very rapidly to changes in pres-
sure. This means in practice that the position of the diaphragm, whether it is bowed
inward or outward, just depends on the net force, i.e., on the sound pressure. If we
can measure this position then we have in effect measured the sound pressure. There
are two commonly used methods for measuring the position, using electrical capac-
itance and electromagnetic induction. We discuss these in Sections 7.3.5 and 7.3.6.
An important point about the pressure microphone is that, since it measures
pressure, it is equally sensitive to sound coming from all directions. Even though
sound travels in a particular direction, pressure does not have a direction. It just
asks on whatever surface is presented to it, in this case the surface of the diaphragm
(see the discussion on page 2). Thus a pressure microphone is omnidirectional. In
practice a pressure microphone may be slightly more sensitive to sound coming from
in front than from behind, because sound from behind is blocked in part by the body
of the capsule itself. This, however, is a small effect. As discussed in Section 3.3,
sound naturally flows around objects smaller than the sound wavelength and does
not feel the obstruction. So as long as the microphone capsule is smaller than the For more discussion of
wavelengths of the sounds it is capturing, sound will therefore flow around it. The this point see Exam-
ple 3.6 on page 70.
shortest sound wavelengths, those of the highest frequency sounds, are about 2 cm,
so the capsules of pressure microphones are deliberately designed to be smaller than
this so that they pick up sound from all directions without obstruction.
The ability of a pressure microphone to pick up sound from any direction can
be useful. Consider for instance the microphone in a mobile phone. It is important
for your phone to pick up your voice clearly no matter how you hold it, or even
if it is lying on the table or on the other side of the room. Phones use pressure
microphones to achieve this. Pressure microphones are also used for headset and
lapel microphones and for live recording, particularly when we want to capture many
instruments at once (such as with an orchestra) or when we want to capture the
acoustics of the performance space as well as the performance itself.
Another use of pressure microphones is in measuring the loudness of sounds
using a sound level meter, as described in Section 3.7. Sound meters measure sound
pressure level as defined in Eq. (3.22) on page 63, which requires us to measure the
mean-square pressure over a period of time. To do this a sound meter uses a pressure
microphone, directly measuring the pressure, then squares the results and averages
them, usually over a period of one second, to calculate the mean-square pressure.
This also means that sound meters are omnidirectional, measuring sound from all
directions and not just the direction they are pointing in, which in most cases is a
desirable feature.

237
Chapter 7 | Sound recording and reproduction

7.3.2 Pressure-gradient microphones

A pressure-gradient The alternative to a pressure microphone is a pressure-gradient microphone. The cru-
microphone is also cial difference between the two is that the capsule of a pressure-gradient microphone
sometimes called a
is not sealed, meaning that sound reaches both sides of the diaphragm. The arrange-
velocity microphone.
ment is shown in Fig. 7.22. The capsule is now a cylindrical tube, closed at one end
by the diaphragm but open at the other end. Like the pressure microphone, the dia-
phragm of a pressure-gradient microphone will bow one way or the other, depending
on which side has the higher pressure.
A defining feature of the pressure-gradient microphone is that, unlike the omni-
directional pressure microphone, its response depends on the direction the sound
comes from. Consider a sound wave arriving at the microphone not straight on but
at an angle 𝜃 , as shown in Fig. 7.22. While sound can reach both top and bottom
of the diaphragm, it has further to travel to get to the bottom. Sound going to the
bottom has to travel the extra distance marked 𝑥 in the figure to reach the far end of
the capsule and then back up the length of capsule to the diaphragm, which is the
distance marked 𝑑. From the geometry we can see that 𝑥 = 𝑑 cos 𝜃 , so the total extra
distance traveled by the sound going to the back of the diaphragm is
𝑥 + 𝑑 = 𝑑 cos 𝜃 + 𝑑 = 𝑑 (cos 𝜃 + 1). (7.45)
Suppose the incoming sound is a pure sine wave at frequency 𝑓 . Then, from
Eq. (4.1), the sound pressure at the front of the diaphragm can be written as
𝑝 𝑓 (𝑡) = 𝐴 sin(2𝜋 𝑓 𝑡), (7.46)
where 𝐴 is the amplitude of the waveform, and the sound pressure at the back of the
diaphragm is the same, but delayed by the amount of time Δ𝑡 it takes to traverse the

Sound

Front

Diaphragm d
x
θ
Capsule
Back

Figure 7.22: A pressure-gradient microphone. In a pressure gradient microphone the

capsule is open at the back so that sound reaches both sides of the diaphragm, but the sound
has to travel a longer distance, as marked, to reach bottom of the diaphragm.

238
 7.3 | Microphones

extra distance, which is

𝑑 (cos 𝜃 + 1)
Δ𝑡 = , (7.47)
𝑐
where 𝑐 is the speed of sound as usual. Thus the sound pressure at the back of the
diaphragm is
𝑝𝑏 (𝑡) = 𝐴 sin[2𝜋 𝑓 (𝑡 − Δ𝑡)]. (7.48)
The total pressure on the front of the diaphragm is thus 𝑃 𝑓 = 𝑃0 + 𝑝 𝑓 and the
pressure on the back is 𝑃𝑏 = 𝑃0 + 𝑝𝑏 and the difference of the two is

𝑃 𝑓 − 𝑃𝑏 = (𝑃 0 + 𝑝 𝑓 ) − (𝑃 0 + 𝑝𝑏 ) = 𝑝 𝑓 − 𝑝𝑏
= 𝐴 sin(2𝜋 𝑓 𝑡) − 𝐴 sin[2𝜋 𝑓 (𝑡 − Δ𝑡)] = 2𝐴 sin(𝜋 𝑓 Δ𝑡) cos[𝜋 𝑓 (2𝑡 − Δ𝑡)],
(7.49)

where we have used a trigonometric identity to get the last formula. Comparing with
Eq. (7.43), this result tells us that the pressure-gradient microphone will behave as if
it were a pressure microphone with the effective sound pressure

𝑝 (𝑡) = 2𝐴 sin(𝜋 𝑓 Δ𝑡) cos[𝜋 𝑓 (2𝑡 − Δ𝑡)], (7.50)

where Δ𝑡 is given by Eq. (7.47).

Now note that the quantity
𝑑 (cos 𝜃 + 1) 𝑑
𝑓 Δ𝑡 = 𝑓 = (cos 𝜃 + 1), (7.51)
𝑐 𝜆
where 𝜆 = 𝑐/𝑓 is the wavelength of the sound from Eq. (2.4). A typical size for a
microphone capsule is about 𝑑 = 1 or 2 cm, much smaller than the wavelength of all
but the highest-frequency audible sounds, so 𝑑/𝜆 is a normally small number, much
less than one. Note also that sin 𝜙 ≃ 𝜙 when 𝜙 is a small number, so sin(𝜋 𝑓 Δ𝑡) ≃
(𝑓 𝑑/𝑐)(cos 𝜃 + 1) and to a good approximation we can write the effective sound
pressure of Eq. (7.50) as
2𝜋𝐴𝑓 𝑑
𝑝 (𝑡) = (cos 𝜃 + 1) cos(2𝜋 𝑓 𝑡), (7.52)
𝑐
| {z } | {z }
Amplitude Waveform

where we have approximated the sine term and dropped the Δ𝑡 term from the second
cosine, since it’s small and doesn’t make much difference.
Recalling that the cosine wave cos(2𝜋 𝑓 𝑡) looks just like a sine wave with fre-
quency 𝑓 except that it is shifted in time, this equation tells us that the pressure-
gradient microphone will feel an effective sound pressure 𝑝 (𝑡) in the shape of a sine
wave with frequency 𝑓 , but with amplitude (2𝜋𝐴𝑓 𝑑/𝑐)(cos 𝜃 + 1). Because this am-
plitude is proportional to cos 𝜃 +1, the effective pressure varies with the direction the

239
Chapter 7 | Sound recording and reproduction

sound is coming from. By contrast, the pressure felt by a pressure microphone does
not depend on direction—a pressure microphone is omnidirectional, as discussed in
Section 7.3.1. Figure 7.23 shows a plot of cos 𝜃 + 1 and, as we can see, the value peaks
at 0◦ , directly in front of the microphone, and falls all the way to zero at 180◦ . This
is a directional microphone. It picks up sound from the direction you point it in and
no sound at all from behind. A microphone with this particular directional pattern
is called a cardioid microphone.
This directionality is very convenient for many mu-
sical applications, for instance in live performance sit-
2 uations. You can point the microphone towards a per-
former or instrument you want to capture and it will
cos θ + 1
pick up mostly their sound and relatively little sound
1
from other directions, such as from other performers,
on-stage speakers, or audience members. In practice,
real directional microphones may not exactly match the
theory—they might pick up a little sound from behind
0
−180 −90 0 90 180 for instance—but they can definitely cut down a lot on
Angle θ unwanted sounds.
A variant on the cardioid microphone is one in which
the diaphragm is placed not at the top of the capsule but
Figure 7.23: The amplitude of the effective sound
pressure felt by a cardioid microphone. The pres- in the middle, as shown in Fig. 7.24. Now the sound at
sure felt by a cardioid microphone is greatest when the front of the capsule has to travel a distance 12 𝑑 down
the sound comes from immediately in front of the the length of the capsule to reach the diaphragm, while
microphone (0◦ ) and zero when it comes from behind the sound at the back has to travel 𝑥 + 21 𝑑, with 𝑥 as in
(180◦ or −180◦ ). The equivalent polar plot is shown in Fig. 7.22. Thus the sound going to the back of the micro-
Fig. 7.25.
phone needs to travel further than the sound going to the
front by a distance (𝑥 + 21 𝑑) − 12 𝑑 = 𝑥 = 𝑑 cos 𝜃 . Modify-
Front ing Eqs. (7.50) and (7.52) appropriately, we find that the
Diaphragm effective pressure felt by the microphone is now

2𝜋𝐴𝑓 𝑑
𝑝 (𝑡) = cos 𝜃 cos(2𝜋 𝑓 𝑡) . (7.53)
𝑐
| {z } | {z }
Capsule Amplitude Waveform
Back
So the amplitude is proportional to cos 𝜃 , which takes its
Figure 7.24: A bidirectional pressure-gradient largest values at 𝜃 = 0◦ and 180◦ and is zero at ±90◦ .
microphone. If the diaphragm of a pressure-gradient In other words, the microphone now picks up sound
microphone is mounted in the center of the capsule it strongly from directly in front and behind, but not at all
produces a bidirectional response that is strongest at from the sides. This is a bidirectional microphone.
the back and front of the microphone and weakest at
A third common choice is one in which the dia-
the sides—see Fig. 7.25.
phragm is placed 23 of the way up the capsule, producing

240
 7.3 | Microphones

a hypercardioid microphone, which is a compromise between cardioid and bidirec-

tional. It has a narrower pickup pattern at the front than a true cardioid, which
is good if one wants a highly directional microphone, but it also picks up a small
amount of sound from behind.

7.3.3 Directional response

The directionality of microphones is often depicted using a polar plot, a circular type
of graph as shown in Fig. 7.25. The direction the sound is coming from is represented
by position around the circle, starting at the top which represents sound coming from
straight in front of the microphone. The line on the plot represents how sensitive
the microphone is in the corresponding direction in terms of the amplitude of the
signal, with values further from the center denoting higher amplitude. In the first
plot of Fig. 7.25, for example, the line is equally far from the center the whole way
around, indicating an omnidirectional microphone—the classic pressure microphone.
The second plot shows a cardioid microphone, whose sensitivity is greatest at the
front and goes to zero at the back, at 180◦ . The shape of this second plot is said by
some people to look like an upside-down heart and this is the origin of the name
“cardioid,” which means “like a heart.” The third plot shows the response pattern
of a hypercardioid microphone and the fourth shows a bidirectional microphone,
sometimes also called a “figure 8” microphone because the polar plot looks like an 8.

0° 0° 0° 0°

90° 90° 90° 90° 90°

180° 180° 180° 180°

Omnidirectional Cardioid Hypercardioid Bidirectional

Figure 7.25: Polar plots of the directional sensitivity of microphones. In a polar plot the direction of the incoming
sound is measured around the circle, with the top of the plot representing the front of the microphone. Distance of the line
from the center of the plot represents the amplitude of the response of the microphone to sound from the corresponding
direction. Thus, in the omnidirectional case the line is equally far from the center all the way around, making a circle on
the plot, so the microphone is equally sensitive in all directions. The cardioid microphone is sensitive mainly to sound from
the front, but not the sides and back. In the bidirectional case the microphone is sensitive to sound from the front and the
back, but not the sides, while the hypercardioid falls in between the cardioid and bidirectional, being sensitive mostly to
sound from the front and somewhat less from the back.

241
Chapter 7 | Sound recording and reproduction

Example 7.3: Directional microphones

How much smaller is the intensity recorded by a cardioid microphone if a sound hits it
90◦ off-center versus straight-on?
To answer this question, we look at Eq. (7.52). The dependence of the effective sound
pressure on the angle 𝜃 that the sound is coming from is given by the factor cos 𝜃 + 1. For
sound hitting the microphone from the front, at 𝜃 = 0, this factor is cos 0 + 1 = 2, but for
sound at 90◦ it is cos 90◦ + 1 = 1. So the sound pressure at the side is a half what it is at
the front. Recall, however, that intensity is proportional to sound pressure squared (see
Eq. (3.6) on page 57), so the intensity at 90◦ is a quarter what it is at 0◦ , or 6 dB lower.

7.3.4 Freqency response

A point to notice about Eq. (7.52) for the pressure-gradient microphone is that the
amplitude of the effective sound pressure 𝑝 (𝑡) is proportional to the frequency 𝑓 .
This means that the pressure-gradient microphone will register a stronger signal at
high frequencies than it does at low ones, in contrast with a pressure microphone,
for which 𝑝 (𝑡) has no variation with frequency. Recall that the intensity of a sound
is proportional to pressure squared (see Eq. (3.7) on page 57), so the effective in-
tensity felt by a pressure-gradient microphone will be proportional to 𝑓 2 . All other
things being the same, this means the intensity will go up by a factor of four, or 6 dB,
when the frequency goes up by a factor of two, i.e., an octave. This makes the sound
captured by a pressure-gradient microphone brighter, with more treble, than a com-
parable pressure microphone.
At the highest frequencies, the approximation we made in Eq. (7.52), that sin 𝜙 ≃
𝜙, is not accurate and we need to calculate the amplitude of the sound pressure from
the full expression in Eq. (7.50). Putting this expression into the formula for intensity,
Eq. (3.9), we get

⟨𝑝 2 ⟩ 4𝐴2 2𝐴2 𝜋𝑓𝑑 

𝐼= = sin2 (𝜋 𝑓 Δ𝑡) cos2 [𝜋 𝑓 (2𝑡 −Δ𝑡)] = sin2 (cos 𝜃 +1) , (7.54)
𝜌𝑐 𝜌𝑐 𝜌𝑐 𝑐

where we have made use of the fact that the average of cos2 𝜙 is 12 .
Converting to decibels, we show in Fig. 7.26 a plot of the resulting sound level as
a function of frequency for sound coming from directly in front of the microphone
at 𝜃 = 0◦ and a capsule of depth 𝑑 = 2 cm. The 6 dB-per-octave increase in sensitivity
is clearly visible in the lower frequencies of the plot, but at higher frequencies the
response levels off and has some pronounced dips. These dips correspond to points
where the sine function in (7.54) is zero. The first zero of the sine function, for in-
stance, occurs when the argument of the function is equal to 𝜋, or in other words
when
𝑐
𝑓 = . (7.55)
𝑑 (cos 𝜃 + 1)

242
 7.3 | Microphones

40
Signal level (dB)

30
ve
octa
er
20
6d Bp

10

0
30 100 300 1000 3000 10000
Frequency (Hz)

Figure 7.26: Theoretical frequency response of a pressure-gradient microphone.

A pressure-gradient microphone is more sensitive to higher frequency sounds. The effective
intensity felt by the microphone is given by Eq. (7.54), which tells us that the sound level goes
up by 6 dB per octave over most of the frequency range, but with marked dips at high frequen-
cies. Real microphones use a variety of techniques to smooth out this behavior, as discussed
in the text.

For the angle 𝜃 = 0◦ used in the figure (for which this frequency is lowest), and
putting 𝑑 = 0.02 m and 𝑐 = 343 m/s, the dip falls at 𝑓 = 8575 Hz, well within the
human hearing range.
The frequency response shown in Fig. 7.26 is less than ideal, for two reasons.
First, while an increased sensitivity to high frequencies is desirable in some cases,
the increase of 6 dB per octave is too large. There is a variation of over 40 dB between
the lowest and highest frequencies shown in the figure, a factor of 10 000 in intensity,
which will be very noticeable. Second, the behavior at high frequencies, with dips in
the response, is definitely undesirable: it means that the microphone will be uneven
in the upper range and will fail to pick up some frequencies altogether.
A number of techniques are used to mitigate these problems. The overall increase
in response can be reduced or eliminated by passing the signal from the microphone
through a low-pass filter of the kind described in Section 7.2.4. Alternatively, one can
reduce the high-frequency response mechanically either by increasing the mass of
the diaphragm, which makes it slower to respond to high frequencies, or by increas-
ing the “damping.” For the interested reader, the details are explained in Section 7.3.7.
To compensate for the high-frequency dips in Fig. 7.26 one can reduce the depth 𝑑
of the microphone capsule, which doesn’t eliminate the dips but pushes their fre-
quency up, via Eq. (7.55), ideally to the point where their effect can no longer be

243
Chapter 7 | Sound recording and reproduction

10
5

Response (dB)
0

−5

−10

−15
100 1000 10000
Frequency (Hz)

Figure 7.27: Frequency response of a typical cardioid microphone. This pressure-

gradient microphone shows an increase in sensitivity with frequency at the lowest frequen-
cies, but the response levels off as frequency gets higher because the microphone has a heavier
diaphragm that limits the high-frequency response. At the highest frequencies the sensitivity
drops off sharply, which may be because of dips in the response of the kind shown in Fig. 7.26,
or because of the mass of the diaphragm or limitations of the electronics.

heard. Decreasing 𝑑, however, also decreases the overall amplitude of the signal via
Eq. (7.52), so the choice of 𝑑 is necessarily a compromise. The smallest capsules are
about 1 cm deep, which raises the frequency of the first dip to 17 150 Hz, right at the
upper frequency limit of audibility, but some microphones opt for a deeper capsule
that produces a stronger signal at the expense of a less satisfactory response at the
highest frequencies.
Figure 7.27 shows the real-world measured frequency response of a pressure-
gradient microphone. As the figure shows, the sensitivity of the microphone does
increase with frequency, particularly at lower frequencies, but then levels off as fre-
quency gets higher. In this case this is achieved by using a heavier diaphragm: the
mylar film of the circular diaphragm of this microphone is thicker in the middle to
increase its mass but thinner around the edge to allow it to move freely with the
incoming sound waves.
A further issue with pressure-gradient microphones is that the directional re-
sponse at the highest frequencies, given by Eq. (7.54), is no longer simply propor-
tional to cos 𝜃 + 1, and hence does not follow the standard cardioid shape. In prac-
tice, this means that pressure-gradient microphones tend to become less directional
at high frequency. Figure 7.28 shows the directional response of the microphone from
Fig. 7.27 at three different frequencies, and we can see that the microphone has a true
cardioid response only at the lowest of the three and becomes more omnidirectional
at higher frequencies.

244
 7.3 | Microphones

0° 0° 0°

90° 90° 90° 90°

180° 180° 180°

500 Hz 2000 Hz 4000 Hz

Figure 7.28: Directionality of a typical cardioid microphone. These polar plots show the measured directional re-
sponse of the same microphone as in Fig. 7.27 at three different frequencies as indicated. In principle, this is a cardioid
microphone, and it does show the classic cardioid at 500 Hz, but it becomes progressively less directional as frequency
increases.

7.3.5 Condenser microphones

So far we have discussed how the diaphragm of a microphone moves in response to
incoming sound. To make a functioning microphone, however, we also need some
way of detecting this movement and turning it into an electrical signal. There are
two primary means by which this is accomplished. The first makes use of the phe-
nomenon of electrical capacitance. The second makes use of electromagnetic induc-
tion. We will look at each of these in turn, starting in this section with capacitance.
As discussed in Section 7.2.3, an electrical capacitor is a device made of two metal
plates separated by a small gap. When a voltage is applied between the plates a
small amount of electrical charge—electrons—flows onto them and stays there. The
amount of charge 𝑄 is given by the capacitor equation 𝑄 = 𝐶𝑉 where 𝑉 is the
voltage across the capacitor and 𝐶 is the capacitance (see Eq. (7.12)). The value of
the capacitance depends on the construction of the capacitor and particularly on the A condenser microphone
distance between the plates. If the distance is 𝑧 then the capacitance is inversely
proportional to 𝑧 thus:
𝑘
𝐶= , (7.56)
𝑧
where 𝑘 is a constant. So when the distance is smaller the capacitance is bigger.
We can use this effect to detect the movement of the diaphragm in a microphone.
A microphone of this kind is called a condenser microphone.
Figure 7.29 shows how such a microphone works. The diaphragm acts as one
plate of a capacitor and we add another fixed plate next to it called the backplate. As

245
Chapter 7 | Sound recording and reproduction

Diaphragm

Perforated
backplate

Figure 7.29: A condenser microphone. A condenser microphone has a diaphragm coated

in metal to make it conduct and a metal backplate, perforated to allow sound and air to pass
through it. Together the diaphragm and backplate form a capacitor whose capacitance depends
on the distance between the two and hence varies as the diaphragm moves.

the diaphragm moves, the distance between the two plates varies and hence so does
the capacitance. We use a power supply to place a fixed charge on the capacitor and
as the capacitance varies this charge produces a varying voltage. All we have to do
is measure this voltage and we have our electrical signal.
This is the basic principle, but there are some important details. First, the dia-
phragm is normally made of mylar plastic, which doesn’t conduct electricity, so it
cannot be directly used as a capacitor plate. To get around this problem, the dia-
phragm is coated with a thin, flexible layer of metal, most often gold, to make it
conduct. If you look closely at a condenser microphone you can sometimes see the
gold color of the diaphragm behind the protective grille on the outside of the micro-
phone.
The backplate is made of rigid metal and is normally placed immediately behind
the diaphragm, out of sight and just a small distance from the diaphragm, typically
only a few micrometers. If the backplate were completely solid it would create a
back-pressure that would hinder the movement of the diaphragm, and for pressure-
gradient microphones it would also block sound from reaching the back of the dia-
phragm, which is important for the way these microphones operate. To get around
these issues, the backplate is normally perforated with a set of holes to let air and
sound through.
The basic circuit diagram for a condenser microphone is shown in Fig. 7.30a.
The microphone is wired in series with a resistor and a constant voltage 𝑉 from a
battery or power supply is applied across both microphone and resistor. As described
This is the circuit symbol in Section 7.2.3, this causes current to flow through the resistor and charge up the
for a microphone capacitor (i.e., the microphone) until the voltage across it reaches 𝑉 , at which point
the current stops and the charge holds steady. If there is no sound arriving, so that
the diaphragm is not moving, and if the capacitance of the microphone is 𝐶 0 when
the diaphragm is at rest, then the capacitor law of Eq. (7.12) tells us that the charge

246
 7.3 | Microphones

R High-pass filter
R | {z }
+ +
V V
- -

Vout
Microphone Vout

(a) (b)

Figure 7.30: Circuit diagram for a condenser microphone. (a) A condenser microphone (conventionally represented
by the symbol with a circle and a line) is wired in series with a large resistor 𝑅 and a power supply with voltage 𝑉 , producing
a constant charge on the microphone. The varying capacitance of the microphone then causes the voltage 𝑉out across it
to vary according to Eq. (7.60). (b) By adding a high-pass filter to the output of the circuit, consisting of a capacitor and
a second resistor (see Section 7.2.4), we can remove the constant voltage 𝑉 , leaving us with the output voltage given in
Eq. (7.61).

on the microphone will be

𝑄 = 𝐶 0𝑉 . (7.57)
The resistance of the resistor is chosen to be very large—typically billions of
ohms—so even the largest current 𝐼 = 𝑉 /𝑅 that can flow through it is very small.
This means that it takes a while for the capacitor to charge up, but once it does the
charge remains essentially constant thereafter, with the value 𝑄 from Eq. (7.57). Even
when sound falls on the diaphragm and pushes it one way or the other, at this slow See Exercise 7.9 on
charging rate there simply isn’t enough time for the charge to change much before page 279 for more details
on how this works.
the diaphragm moves back again.
What does change when sound arrives however is the capacitance 𝐶, which
varies with the separation between the plates of the capacitor, i.e., between the dia-
phragm and backplate. With the charge remaining constant, this means that the
output voltage 𝑉out across the microphone will be
𝑄 𝐶 0𝑉
𝑉out = = 𝑧, (7.58)
𝐶 𝑘
where we have used Eqs. (7.56) and (7.57). Let us write the distance between the
diaphragm and the backplate as 𝑧 = 𝑧 0 + Δ𝑧, where 𝑧 0 is the distance when the
diaphragm is at rest and Δ𝑧 is how much it moves due to the sound pressure. Then
𝐶 0𝑉
𝑉out = (𝑧 0 + Δ𝑧). (7.59)
𝑘

247
Chapter 7 | Sound recording and reproduction

But by definition the capacitance at rest is 𝐶 0 = 𝑘/𝑧 0 and so

𝑉 𝑉
𝑉out = (𝑧 0 + Δ𝑧) = 𝑉 + Δ𝑧. (7.60)
𝑧0 𝑧0
The constant voltage 𝑉 can be filtered out using a high-pass filter with a low cutoff
frequency, as described in Section 7.2.4. This gives us the circuit shown in Fig. 7.30b,
with output voltage
𝑉
𝑉out = Δ𝑧. (7.61)
𝑧0
So the output voltage simply mirrors the displacement Δ𝑧 of the diaphragm.
Equation (7.61) is inversely proportional to 𝑧 0 , the separation at rest between the
diaphragm and the backplate, so by making this separation small we can make the
output signal large. We should, however, avoid making the separation so small that
diaphragm and backplate might touch when the diaphragm moves, which would
cause a short-circuit and destroy the charge on the capacitor. 20 to 40 micrometers
is a typical separation distance for commercial condenser microphones.
Equation (7.61) is also proportional to the power-supply voltage 𝑉 , which means
we can increase the signal by increasing 𝑉 , though we want to keep the voltage small
enough to not risk shocking the user of the microphone if something goes wrong.
By longstanding convention, condenser microphones use a voltage of 48 volts. This
is too large to be conveniently supplied by a battery, so condenser microphones nor-
mally use an external power supply, known in the music industry as a “phantom”
power supply.
Even when we take these steps to make the signal from the microphone as large
as possible, it is still very small—typically just a fraction of a volt—so it is ampli-
fied with a special preamplifier designed for use with microphones. Commonly the
preamplifier and the phantom power supply are combined into one device that both
generates the voltage for the microphone and amplifies the resulting signal.
Condenser microphones can be made with large diaphragms that are highly sen-
sitive to incoming sound and give very high quality recordings. They are widely
used in recording studios. However, they can also be fragile and easily damaged, and
the fact that they require a separate power supply makes them less convenient than
other types of microphones. In live music situations, where ruggedness is a bonus,
or in situations such as amplification of speech, where convenience is at a premium
and sound quality is not, another type of microphone, the dynamic microphone, is
more widely used.

Example 7.4: Condenser microphone design

A certain condenser microphone runs on the standard 48 volt power. An incoming sound
causes the diaphragm to move back and forth by 0.05 𝜇m each way. If we want the micro-

248
 7.3 | Microphones

phone to produce a corresponding 0.1 volts in the output voltage what should we make
spacing between the diaphragm and the backplate?
We can answer this question using Eq. (7.61). Rearranging the equation for the spac-
ing 𝑧 0 , we have
𝑉 48
𝑧0 = Δ𝑧 = × 0.05 𝜇m = 24 𝜇m, (7.62)
𝑉out 0.1
which is typical for today’s condenser microphones.

7.3.6 Dynamic microphones

The primary alternative to a condenser microphone is a dynamic microphone, which
uses electromagnetic induction to measure the movement of the diaphragm. Dy-
namic microphones can give results of good quality, but cannot match the sound
from the best condenser microphones. They are popular, however, for their rugged-
ness, which makes them ideal for live performance situations.
As discussed in Section 7.2.7, electromagnetic induction occurs when the mag- A dynamic microphone
netic field passing through a coil of wire changes, getting either stronger or weaker,
which produces an electric voltage between the ends of the coil. We can use this
effect to build a microphone. Perhaps the most obvious way to proceed would be
to attach a magnet to a diaphragm and place a coil nearby. The movement of the
diaphragm would then make the magnet move and hence change the magnetic field
in the coil, producing a current. In practice, however, this approach doesn’t work
well because magnets are too heavy. It is important that the diaphragm be light, so
it can respond rapidly to changes in pressure, so having a heavy magnet attached is
not ideal.
Instead, therefore, dynamic microphones work the other way around, by attach-
ing the coil (which is relatively light) to the diaphragm and placing a stationary mag-
net nearby—often actually inside the coil or partially inside it. Figure 7.31 shows a

Diaphragm

Coil
Output
Magnet

Figure 7.31: A dynamic microphone. In a dynamic microphone a coil is attached to the

back of the diaphragm and moves as the diaphragm vibrates. A fixed permanent magnet is
mounted so that it extends inside the coil. When the coil moves, the magnetic field running
through it changes, producing a current in the coil which can then be amplified or recorded.

249
Chapter 7 | Sound recording and reproduction

typical arrangement, with the diaphragm forming the top of the microphone capsule
and the coil attached to its underside. The coil moves with the diaphragm, and hence
the magnetic field inside it, produced by the stationary magnet, changes, inducing a
current. The magnetic field itself doesn’t move, but the amount of it inside the coil
changes as the coil goes back and forth.
In addition to their ruggedness, dynamic microphones have an advantage over
condenser microphones in not requiring an external power supply—the process of
electromagnetic induction creates its own electricity. Dynamic microphones come
in both pressure and pressure-gradient varieties. The most common type for musical
applications are cardioid pressure-gradient microphones.

Advanced material

7.3.7 Mechanics of microphones Putting together the forces on the diaphragm, the to-
tal force is 𝐹 = 𝐹𝑝 + 𝐹𝑟 + 𝐹𝑑 . Applying Newton’s second
The signal produced by a microphone depends on a com- law in the form 𝐹 = 𝑚𝑎 and writing the acceleration as
bination of the movement of the diaphragm in response 𝑎 = d2 𝜉/d𝑡 2 , we then find that the motion of the diaphragm
to the sound pressure acting on it and the method used to obeys the differential equation
turn the movement into electricity. The movement of the d2 𝜉 d𝜉
diaphragm is driven by the force due to sound pressure, 𝑚 +𝛾 + 𝑘𝜉 = 𝑆𝑝 (𝑡). (7.63)
d𝑡 2 d𝑡
Eq. (7.44), which we here denote by 𝐹𝑝 = 𝑆𝑝 (𝑡), where 𝑆
is the area of the diaphragm and 𝑝 (𝑡) is either the actual Suppose now that sound strikes the microphone in the
sound pressure (for a pressure microphone) or the effec- form of a sine wave with frequency 𝑓 , which we write
tive pressure, Eq. (7.52), from the difference between the in the standard manner as the real part of a complex
front and back (for a pressure-gradient microphone). wave 𝑝 (𝑡) = 𝑝 0 ei𝜔𝑡 , where 𝜔 = 2𝜋 𝑓 and 𝑝 0 is either
This is not the only force acting on the diaphragm how- equal to the amplitude 𝐴 of the wave for a pressure micro-
ever. There are also two other forces. The first is a restoring phone or, for a pressure-gradient microphone, is given by
force 𝐹𝑟 due to the elastic nature of the diaphragm. When Eq. (7.52):
𝜔𝐴𝑑
the diaphragm is pushed to one side or the other its springi- 𝑝0 = (cos 𝜃 + 1). (7.64)
ness naturally pulls it back. To a good approximation, the 𝑐
The sine-wave sound pressure produces a corresponding
restoring force obeys Hooke’s spring law. If the displace-
sine-wave motion in the diaphragm 𝜉 (𝑡) = 𝜉 0 ei𝜔𝑡 and, sub-
ment of the diaphragm from its rest position is 𝜉, then the
stituting into Eq. (7.63) and cancelling a factor of ei𝜔𝑡 , we
restoring force is 𝐹𝑟 = −𝑘𝜉, where 𝑘 is the spring constant
find that −𝜔 2𝑚𝜉 0 + i𝜔𝛾𝜉 0 + 𝑘𝜉 0 = 𝑆𝑝 0 , or
and the minus sign indicates that the restoring force is in
the opposite direction to the displacement. 𝑆𝑝 0
𝜉0 = . (7.65)
The other force on the diaphragm is a damping force 𝐹𝑑 𝑘 − 𝜔 2𝑚 + i𝜔𝛾
that acts to slow down the diaphragm when it is moving.
How this motion gets turned into an electrical signal
This force comes primarily from air resistance and is pro-
depends on whether we have a condenser microphone or
portional to the velocity of the diaphragm d𝜉/d𝑡, so that
a dynamic microphone. Let us look at each in turn.
𝐹𝑑 = −𝛾 d𝜉/d𝑡, where the minus sign indicates that the
Condenser microphone: For a condenser microphone,
force is in the opposite direction to the velocity and 𝛾 is a
the output voltage 𝑉out is given by Eq. (7.61) with Δ𝑧 =
constant which controls the strength of the damping and
𝜉 (𝑡):
which depends on the size of the diaphragm and on the 𝑉
shape and design of the microphone capsule. 𝑉out = 𝜉 (𝑡), (7.66)
𝑧

250
 7.3 | Microphones

with 𝑉 being the power supply voltage and 𝑧 being the proportional to the displacement 𝜉 of the diaphragm, so
spacing between the diaphragm and the backplate. The
d𝜉
amplitude of the output is then given by the complex mag- 𝑉out = −𝑁𝑏 = −𝑁𝑏 × i𝜔𝜉 0 ei𝜔𝑡 , (7.70)
d𝑡
nitude of 𝑉out , which is
where 𝑏 is a proportionality constant. With 𝜉 0 as in
𝑉 𝑉 𝑆𝑝 0 /𝑧 Eq. (7.65), the amplitude of the output voltage is then again
|𝑉out | = |𝜉 0 | = p . (7.67)
𝑧 (𝑘 − 𝜔 2𝑚) 2 + 𝜔 2𝛾 2 given by the complex magnitude of 𝑉out , which is
𝑁𝑏𝑆𝑝 0𝜔
If our microphone is to have a uniform frequency re- |𝑉out | = p . (7.71)
sponse, we want this magnitude to be simply proportional (𝑘 − 𝜔 2𝑚) 2 + 𝜔 2𝛾 2
to the amplitude 𝐴 of the incoming sound wave and in- A crucial difference from the condenser microphone
dependent of 𝜔. In a pressure microphone, for which is the appearance of 𝜔 in the numerator of this expres-
𝑝 0 = 𝐴, this requires that the denominator of (7.67) itself sion, which means the dynamic microphone will produce
be independent of 𝜔,pwhich we can achieve by making 𝑚 a stronger signal at high frequencies. This is a result of the
and 𝛾 small, so that (𝑘 − 𝜔 2𝑚) 2 + 𝜔 2𝛾 2 ≃ 𝑘 and hence fact that electromagnetic induction responds to the rate of
|𝑉out | = 𝑉 𝑆𝐴/𝑧𝑘. This means we want a very thin, light change of the magnetic field, and not just the size of the
diaphragm that responds rapidly to changes in sound pres- field, and it is separate from, and in addition to, the effect
sure. discussed in Section 7.3.4 where pressure-gradient micro-
On the other hand, if we have a pressure-gradient phones register a stronger signal at higher frequencies. We
microphone, then 𝑝 0 ∝ 𝜔 as in Eq. (7.64). In this case, need to compensate for this effect if we want our micro-
we can do one of two things. Either we can again use a phone to have a uniform frequency response. How we do
thin, light diaphragm so that the denominator in (7.67) is this depends again on whether we have a pressure micro-
constant, which produces a |𝑉out | that is proportional to phone or a pressure-gradient microphone.
frequency 𝜔, but then filter the resulting signal using a For a pressure microphone 𝑝 0 is simply equal to the
6 dB-per-octave low-pass filter to remove this frequency amplitude 𝐴 of the incoming sound wave, so the numer-
dependence. Or we can design the microphone capsule ator of (7.71) is proportional to 𝜔. In a manner similar to
so that the damping 𝛾 is large compared to the mass 𝑚 the pressure-gradient condenser mic, we can then achieve
and
p spring constant 𝑘 of the diaphragm, which means a uniform frequency response either by using a light dia-
(𝑘 − 𝜔 2𝑚) 2 + 𝜔 2𝛾 2 ≃ 𝜔𝛾 and phragm with small mass 𝑚 and damping 𝛾, which gives
|𝑉out | = 𝑁𝑏𝑆𝑝 0𝜔/𝑘, and then passing the output signal
(𝜔𝑆𝐴𝑑/𝑐)(cos 𝜃 + 1) 𝑆𝐴𝑑 through a 6 dB-per-octave low-pass filter to remove the fre-
|𝑉out | = = (cos 𝜃 + 1). (7.68)
𝜔𝛾 𝛾𝑐 quency dependence, or we can make the damping 𝛾 large,
which gives |𝑉out | = 𝑁𝑏𝑆𝑝 0 /𝛾, which is independent of
We cannot make 𝛾 very large, however, because a large 𝛾 frequency from the outset.
will reduce the overall value of |𝑉out | and hence reduce the Conversely, if we have a pressure-gradient dynamic
output of the microphone. This typically means we have to microphone, with 𝑝 0 as in Eq. (7.64), then
make a compromise: limiting the value of 𝛾 may mean that
for very low frequencies the denominator of (7.67) is not (𝑁𝑏𝜔 2𝑆𝐴𝑑/𝑐)(cos 𝜃 + 1)
|𝑉out | = p , (7.72)
exactly proportional to 𝜔 and we will have some residual (𝑘 − 𝜔 2𝑚) 2 + 𝜔 2𝛾 2
frequency dependence in 𝑉out .
Dynamic microphone: For a dynamic microphone the with a numerator that now goes as 𝜔 2 . We can also combat
output voltage is produced by induction in the microphone this frequency dependence in a number of ways. The tradi-
coil and Eq. (7.41) tells us that the voltage across a coil with tional approach is to make a “mass-controlled” microphone
𝑁 turns is with a heavier diaphragm, as described in Section 7.3.4,
dΦ for
p which 𝑚 is large compared to 𝑘 and 𝛾, which means
𝑉out = −𝑁 , (7.69)
d𝑡 (𝑘 − 𝜔 2𝑚) 2 + 𝜔 2𝛾 2 ≃ 𝜔 2𝑚 and
where Φ is the magnetic flux, i.e., the amount of magnetic 𝑁𝑏𝑆𝐴𝑑
field that passes through the coil. In this case the flux is |𝑉out | = (cos 𝜃 + 1). (7.73)
𝑚𝑐

251
Chapter 7 | Sound recording and reproduction

We should not make the mass too large, however, because a lighter diaphragm and then pass the output through a
it will reduce the overall value of |𝑉out |, and again this usu- low-pass filter, either cutting 6 dB per octave if the motion
ally means we have to make a compromise: limiting the of the diaphragm is dominated by the damping 𝛾 or 12 dB
value of 𝑚 may mean that for very low frequencies the de- per octave if both the mass and the damping are small.
nominator of (7.72) is not exactly proportional to 𝜔 2 and Finally, note that for all variants of the dynamic micro-
we will have some frequency dependence in 𝑉out . This kind phone the output voltage is proportional to the number of
of behavior can be seen for instance in Fig. 7.27 on page 244, turns 𝑁 in the coil, so it pays to have as many turns as pos-
where the response of the microphone does vary at low fre- sible. There is a limit to how many we can have if we want
quencies but then levels off. to keep the mass of the coil reasonably low, but a typical
Alternatively for this type of microphone we can use microphone coil has a few thousand turns.

7.3.8 Other types of microphone

Condenser and dynamic microphones are the primary types of microphones used in
modern music recording and amplification, but there are other types that see occa-
sional use.
An electret microphone is a variant of a condenser microphone that uses a back-
plate with a permanent electric charge. Recall that a condenser microphone works
by forming a capacitor between the diaphragm and the backplate, then applying a
voltage across this capacitor using a power supply in order to place a charge on it.
In an electret microphone one dispenses with the applied voltage and instead makes
the backplate out of a special material—an electret—that carries a permanent electric
charge embedded in it. This charge behaves in the same way as the applied charge
in a conventional condenser microphone, producing a voltage across the capacitor
that varies with the distance between diaphragm and backplate.
Compared to condenser microphones, electret microphones have the advantage
of not requiring an external power supply, but their sound quality is often not as
good as true condenser microphones, making them less popular for musical appli-
cations. They do find wide use however for recording speech, in dictation machines
for example, where sound quality is not so important but convenience is.
A ribbon microphone is a variant of a dynamic microphone in which the coil of
Fig. 7.31 is replaced by a thin ribbon of metal foil that acts both as diaphragm and
as coil. The ribbon is mounted in the magnetic field of a permanent magnet and
when sound hits the ribbon and causes it to move, it induces an electric current that
can then be fed into an amplifier. Ribbon microphones are simple and durable, and
hence attractive for live performance, but they produce only a small electric signal
compared to a conventional dynamic microphone and normally have a bidirectional
pickup pattern (see Fig. 7.25), which is less desirable than the cardioid pattern favored
for most live applications.
A carbon microphone uses a completely different mechanism to detect sound. As
shown in Fig. 7.32, the capsule is filled with granules of carbon, which have the

252
 7.4 | Recording

V Diaphragm

Carbon granules

Figure 7.32: A carbon microphone. The capsule of a carbon microphone is filled with car-
bon granules whose resistance to the flow of electricity decreases when they are compressed.
A voltage is applied as shown, producing an electric current through the carbon, and the mo-
tion of diaphragm squeezes the granules causing their resistance to vary and hence varying
the current.

property that they conduct electricity better when compressed. The diaphragm of
the microphone, which is typically made of thin metal foil, rests on the carbon gran-
ules, and a voltage is applied between the diaphragm and the bottom of the capsule,
causing a current to flow through the carbon. When the diaphragm moves with in-
coming sound it compresses the carbon granules, causing them to conduct electricity
better and so increasing the current. The result is a flowing current that varies with
the sound waveform.
Carbon microphones are cheap, simple, robust, and do not require high voltages
to operate, but they produce poor quality sound, not suitable for musical applica-
tions. The best known example of a carbon microphone for many people will be the
mouthpiece in an old-fashioned landline telephone. Most telephones used carbon
microphones until the 1980s, although modern landline phones and mobile phones A vintage telephone
typically use condenser or electret microphones. with a carbon micro-
phone.

7.4 Recording
Once sound has been converted from a sound wave into an electrical signal it can
be recorded using a variety of methods. For most of its history, sound recording was
dominated by two main technologies, magnetic tape and vinyl records. Some of the
most important recordings of all time have been made using these technologies, but
both are now considered obsolete, having been superseded from the 1980s onward by
digital recording techniques that convert sound waves into numbers which are then
stored on hard disks, optical disks, or in the memory of computing devices. Digital
music is a large topic that we deal with separately in the following chapter. In this
section, we discuss tape and records. Although both technologies are obsolete it is
worth taking a look at them in order to understand how they work, why they are

253
Chapter 7 | Sound recording and reproduction

obsolete, and why modern digital recording technologies are superior.

7.4.1 Magnetic tape

Throughout most of the twentieth century the principal medium for recording sound
was magnetic tape, in the form of reel-to-reel tape recorders, studio multitrack record-
ers, cassettes, and 8-track tape. All of these operate on the same principle, making use
of reels of flexible plastic tape coated with powdered ferric oxide, a naturally occur-
ring mineral that contains high quantities of iron. When it comes into close contact
with a magnetic field, the iron becomes magnetized—it becomes a weak magnet in
its own right. The basic setup is shown in Fig. 7.33. A sound signal is fed into the
coil of an electromagnet, called a tape head, with a specially shaped core designed
A traditional reel-to-reel to concentrate the magnetic field in a small region between its ends or poles. The
tape recorder tape passes over this region and becomes magnetized with a strength that mirrors
the signal.
Later, when we want to play back the sound, the same tape is passed once again
over the tape head, which now works in reverse, by electromagnetic induction. The
varying magnetic field produced by the magnetized tape induces a corresponding
voltage across the coil of the tape head, which can be amplified and fed to a loud-
speaker to recreate the original sound. (In more elaborate tape recorders there are
two different heads, one for recording and one for playing back, but the principle is
still the same.)
There are a couple of points to notice about this pro-
cess. First, as discussed in Section 7.2.7, the strength of the
magnetic field produced by an electromagnet depends on
the size of the current flowing in the coil, which means
in turn that the amount of magnetization of the tape is
roughly proportional to the sound signal. Upon playback,
however, the voltage generated by the coil is proportional
not to the magnetic field of the tape but to the rate at which
the field is varying. Because the field varies faster for high
frequency sounds than for low, this means that higher
Figure 7.33: A tape recorder head. A tape frequencies generate a larger voltage, other things being
recorder head consists of a coil wound around a spe- equal. As a result, the frequency balance of the sound gets
cially shaped core. A sound signal is passed through distorted in the playback process, with too much treble and
the coil, creating a magnetic field that concentrates not enough bass.
in the small gap at the bottom between the poles of
the core. The moving tape passes over this gap and
In mathematical terms, the amplitude of the voltage,
becomes magnetized. and hence the amplitude of the sound we ultimately pro-
duce, is proportional to frequency 𝑓 and, bearing in mind
that sound intensity is proportional to the square of amplitude (see Eq. (3.7)), this
means that the intensity will increase by a factor of four, or 6 dB, when the frequency

254
 7.4 | Recording

goes up by one octave. In order for tape recordings to sound correct when we play
them back, we have to correct for this issue. Luckily this is straightforward. As we
saw in Section 7.2.4, a standard low-pass filter cuts out exactly 6 dB per octave above
its cutoff frequency 𝑓𝑐 = 1/2𝜋𝑅𝐶. All we have to do to fix our tape recordings there-
fore is pass the output signal from the tape recorder through a low-pass filter with a
low cutoff frequency—low enough to be below the bottom end of the hearing range
at 20 Hz—so that all audible frequencies get filtered by 6 dB per octave. This process
is known as compensation.
Second, the ability of tape to capture high-frequency sounds depends on the size
of the gap between the poles of the tape head where the magnetic field is concen-
trated and on the speed at which the tape moves. If the tape is moving at velocity 𝑣
and the distance between the poles is 𝑑, then it will take the tape an amount of time
Δ𝑡 = 𝑑/𝑣 to cover this distance. Any change in the sound signal over a time shorter
than this will get blurred out on the tape—the signals at two moments less than Δ𝑡
apart will be recorded on overlapping portions of the tape and will not be distinct.
This means that a sound has to have period greater than Δ𝑡 to be recorded clearly, or
equivalently that the frequency has to be less 1/Δ𝑡. In other words, sounds are only
captured faithfully up to a highest frequency of
1 𝑣
𝑓max = = . (7.74)
Δ𝑡 𝑑
Rearranging this equation, we find that to achieve a particular 𝑓max we need the
speed of the tape to be greater than 𝑣 = 𝑑 𝑓max . A typical size for the gap in a tape
head is 10 micrometers, or 10−5 meters, so to record all audible frequencies up to
20 000 Hz the tape would need to move at a speed of
𝑣 = 10−5 × 20 000 = 0.2 m/s, (7.75)
or 20 centimeters per second. Standard reel-to-reel tape recorders use a speed of
7 21 inches per second, or about 19.1 cm/s, which is close enough to 20 cm/s that we
should expect good reproduction across the full frequency range. Lower speeds are
also sometimes used, at the expense of poorer high-frequency fidelity, and higher
speeds are used in professional recording studios. Cassette recorders for consumer
use employ a substantially lower speed of 4.75 cm/s, in order to fit the maximum
amount of recording time onto a small tape, but they also use a tape head with a A cassette tape
narrower gap to preserve high-frequency response. Nonetheless, cassette tapes have
generally poor performance at the upper end of the frequency range, with the re-
sponse dropping off significantly above about 15 000 Hz.

7.4.2 Vinyl records

The very first sound recordings, in the late nineteenth century, used a groove cut in
wax or foil to represent sound waves, and essentially the same technology is used

255
Chapter 7 | Sound recording and reproduction

in today’s vinyl records. For much of the twentieth century the long-playing record,
or LP, was the premier consumer medium for music reproduction, combining high
sound quality with long running times and relatively sturdy construction. Although
LPs are not competitive in any of these respects with today’s digital recordings, they
still enjoy popularity with collectors (for interesting reasons—see Section 8.4).
Sound on a vinyl record is captured in the shape of the groove that runs in a
spiral around the surface of the record. The earliest recording media took the shape
of a cylinder with grooves on the outside, but since the early twentieth century the
preferred shape has been a flat disk, with grooves usually cut on both sides, so that
one disk can hold two different recordings (such as two songs, or two halves of a
longer musical work like a symphony). The groove wiggles from side to side in im-
A long-playing record, itation of the waveform of the recorded sound, as shown in Fig. 7.34, and the sound
or LP is played back by placing a fine needle in the groove and then rotating the record,
usually with an electric motor, making the needle vibrate back and forth in time with
the wiggles in the groove. Just this on its own is enough to make a tiny sound—if you
listen carefully you can hear sound coming from the needle of a record player as it
vibrates. For proper listening, however, the sound must be amplified. To do this, the
vibration of the needle is converted into an electrical signal using a mechanism sim-
ilar to the dynamic microphone of Section 7.3.6: a coil is attached to the needle and
a magnet placed nearby. The variation of the magnetic field through the coil as the
needle vibrates generates a voltage by electromagnetic induction, which is amplified
and then turned into sound using a loudspeaker.
The process of recording music onto vinyl records is laborious. The sound is
normally captured on tape first, then turned into a record by cutting a groove into
the surface of a metal master disk using a specialized lathe. The master disk is then
used to cast a “negative” of the groove, with ridges sticking out where the grooves
would normally go in. One such negative is cast for each side of the record. Then the
actual LPs are manufactured by pressing molten plastic between the two negatives
to create a disk with grooves on both sides. Although early records were made of
wax, modern LPs are made of vinyl acetate, a form of plastic hard enough to resist
damage, although it can still be scratched or broken by careless treatment, or warped
by extreme heat. The possibility of damage is one of the reasons to prefer digital
recordings over LPs.
Even when they are in perfect condition, however, LPs cannot match the sound
quality of modern digital audio. A primary limitation is their dynamic range. The dy-
namic range of a sound recording, or of a musical performance generally, is the dif-
ference in decibels between the loudest and quietest parts. For example, a symphonic
performance might range from 20 dB in moments of near silence to 90 dB when the
entire orchestra is playing at full volume, giving a dynamic range of 90 − 20 = 70 dB
overall. In order to capture such a performance faithfully a recording medium needs

256
 7.4 | Recording

Figure 7.34: A vinyl record. Each side of a vinyl record is inscribed with a single groove
that starts at the outside of the disk and spirals inward, making several hundred turns before
ending near the center. If you examine the groove carefully (inset) you can see the wiggles
that encode the waveform of the recorded sound.

to be able to record sounds over at least this range.

The quietest sound that can be recorded on a vinyl record corresponds to the
smallest wiggle that can be reliably picked up by the needle and associated electron-
ics in a record player. In practice this smallest wiggle is about 50 nanometers across.
Similarly, the loudest sound corresponds to the largest wiggle, whose width is set
by the need to fit a certain number of grooves on the record. For a typical LP that
plays for 30 minutes a side at 33 revolutions per minute, the groove spirals around
the record a total of 30 × 33 = 990 times in the space of about 10 cm, from its outer-
most point on the disk to its innermost. Thus the spacing of the grooves is around
0.1 m/990 = 100 micrometers, and the wiggles in the grooves are about half this size,
or about 50 micrometers, to allow some room for error. Thus the ratio between the
amplitude of the largest and smallest wiggles is

50 micrometers 50 × 10−6 m
= = 1000. (7.76)
50 nanometers 50 × 10−9 m
The wiggles, as we have said, encode the shape of the sound waveform and hence
their amplitude corresponds to the amplitude of the sound. Equation (7.76) tells us
that the amplitude of the loudest sound that can be recorded is 1000 times that of
the quietest. But recall that the intensity of a sound is proportional to the square
of the amplitude (see Eq. (3.7)), so the loudest sound has 10002 = 1 million times the
intensity of the quietest. A factor of a million in intensity corresponds to a difference
of 60 dB in sound intensity level (see Eq. (3.16)), so the dynamic range of our vinyl

257
Chapter 7 | Sound recording and reproduction

record is about 60 dB—not enough to fully capture the 70 dB range of the orchestra
in our example above.
It is possible to increase the spacing between the grooves on a record, also called
the pitch, to allow larger wiggles and hence louder sounds to be recorded, at the
expense of having fewer grooves overall and hence shorter recording time. It is also
common to temporarily increase the pitch in louder portions of a piece of music to
achieve a better dynamic range and then reduce it again in the quieter portions to
save space. In this way the dynamic range can be pushed to about 70 dB, but that is
the limit of the LP technology.
70 dB is enough to provide a good listening experience, but not enough to cap-
ture the true range of highs and lows in the most dramatic pieces of music. As we
will see in Section 8.1.5, the dynamic range of a standard digital recording is 96 dB,
significantly better than an LP, and enough to faithfully capture essentially any per-
formance audible to the human ear.

Example 7.5: Playing time of a 7-inch single

From the 1950s to the 1990s individual songs or other short recordings were commonly
released on singles, vinyl records 7 inches in diameter that rotated at 45 revolutions per
minute. The distance from the outer edge of the playing area to the inner edge on such a
single is about 1 inch. If we want the record to have a dynamic range of 60 dB, how long
in minutes can the recording be?
We have seen that the smallest wiggles in the groove of a record are about 50 nano-
meters across, which means the largest must be 50 micrometers in order to achieve a dy-
namic range of 60 dB (see Eq. (7.76) and the following discussion). Allowing 100 micro-
meters of separation between one turn of the groove and the next to give ourselves a safety
margin, and noting that 1 inch is about 25 mm, we find that the maximum number of times
the groove can wind around the record is

25 × 10 −3
= 250. (7.77)
100 × 10 −6
Given that the record spins at 45 revs per minute, the maximum playing time is then
250
= 5.56 minutes, (7.78)
45
or about 5 minutes and 33 seconds.
This basic physical limit had a substantial effect on the development of 20th century
popular music. The 7-inch single was for decades the primary format for the sale and
promotion of popular songs, which meant songs had to be written to fit within about
5 minutes. It is possible to stretch the playing time a little longer, at the expense of lower
dynamic range. Hey Jude by the Beatles, which was released as a 7-inch single, clocked
in at 7 minutes and 11 seconds, which is around the maximum that can be achieved with
the format. Don McLean’s American Pie was 8 minutes and 42 seconds long and had to be
split into two halves, one on each side of the record, when it was released as a single.

258
 7.5 | Loudspeakers

7.5 Loudspeakers
The final stage of the music reproduction process requires us to turn our electrical
signals back into sound again so we can hear them, which is achieved using loud-
speakers or headphones. Headphones are essentially just two loudspeakers mounted
on either side of the listener’s head, and the scientific principles of loudspeakers and
headphones are basically the same. We focus here on loudspeakers.
Loudspeakers are the weak link in the music reproduction chain. Modern micro-
phones can capture sound extremely faithfully, and today’s recording and amplifi-
cation technologies are virtually perfect. But even the best loudspeakers are prob-
lematic. Because of fundamental physical limitations, loudspeakers suffer both from
uneven frequency response and haphazard directional output. In the following sec-
tions we will see why this is, and what can be done to improve the situation.

7.5.1 Dynamic loudspeakers

Loudspeakers come in a variety of different types but the most common by far, and
the only one we will discuss at length, is the dynamic loudspeaker, which is the loud-
speaker equivalent of a dynamic microphone: it uses a diaphragm, a coil, and a per-
manent magnet to turn electrical signals into sound.
Figure 7.35 shows the layout of a dynamic loudspeaker. The basic idea behind any
speaker is the reverse of a microphone: we make a diaphragm vibrate and it pushes
on the nearby air, causing it also to vibrate and hence making sound. The diaphragm
of a dynamic loudspeaker, which is called a cone because it is cone-shaped rather A dynamic loudspeaker
than flat, is typically made of paper or sometimes thin plastic. It is fixed around its

Cone

Magnet Coil

Figure 7.35: A dynamic loudspeaker. In a dynamic loudspeaker a diaphragm or cone,

suspended at its edges, is driven back and forth by the attached coil. When electric current
is passed through the coil it creates an electromagnet that is attracted to or repelled from the
large permanent magnet mounted on the frame of the speaker. The motion of the cone causes
the air to vibrate and produces sound.

259
Chapter 7 | Sound recording and reproduction

outside edge to a sturdy frame and the edges are corrugated and flexible to allow
the cone to move in and out. The rest of the cone is relatively rigid and maintains
its shape as it moves. One can think of the cone as a circular piston, pushing the
adjacent air and producing sound.
Mounted on the back of the cone in the center is a coil of wire similar to that in
a dynamic microphone, and inside the coil is the pole-piece of a permanent magnet
that is fixed to the frame of the loudspeaker and held stationary. When we pass
current through the coil it creates an electromagnet with a strength that depends on
the size of the current as described in Section 7.2.7, and this electromagnet is either
attracted to or repelled from the magnet (depending on the direction of the magnetic
field), which pushes or pulls the coil and moves the attached loudspeaker cone.

7.5.2 Sound production

Consider a loudspeaker driven by a pure sine-wave tone with frequency 𝑓 . This
causes the cone to move with a velocity 𝑢 that also follows a sine wave 𝑢 (𝑡) =
𝑢 0 sin(2𝜋 𝑓 𝑡), where the quantity 𝑢 0 represents the maximum velocity that the cone
achieves. This motion in turn causes the air immediately against the front of the
speaker cone to move with the same velocity.
Now recall that, as discussed in Sections 1.3 and 1.6.1, the sound pressure 𝑝 in a
sound wave is proportional to the velocity according to 𝑝 = 𝑧𝑢 (Eq. (1.2)), where the
constant of proportionality 𝑧 is the so-called acoustic impedance of air. This means
that the pressure in this case will be 𝑝 (𝑡) = 𝑧𝑢 0 sin(2𝜋 𝑓 𝑡) and Eq. (3.7) then tells us
that the instantaneous sound intensity is

𝑝2
= 𝑧𝑢 02 sin2 (2𝜋 𝑓 𝑡). (7.79)
𝑧
Then the average intensity is

𝐼 = 𝑧𝑢 02 sin2 (2𝜋 𝑓 𝑡) = 12 𝑧𝑢 02, (7.80)

where we have made use of the fact that the average of sin2 (2𝜋 𝑓 𝑡) is 12 .
Equation (7.80) tells us the intensity of the sound immediately in front of the
speaker cone, i.e., the amount of sound energy per unit area per second. To get the
total amount of energy 𝑃 emitted by the speaker per second we multiply by the area 𝐴
of the cone, which gives
𝑃 = 𝐴𝐼 = 12 𝐴𝑧𝑢 02 . (7.81)
This is the power output by the speaker.
Because the power is proportional to cone area, we will get more sound from a
large speaker than a small one. In simple terms, a large speaker pushes more air.
Usually this means that we would like to make our speakers as large as possible, but

260
 7.5 | Loudspeakers

there are also disadvantages to large speakers as we will shortly see. Real speaker
design, like so many other things, is a compromise between a number of competing
factors.

Example 7.6: Sound output of a loudspeaker

Consider a typical small loudspeaker, circular with a diameter of 10 cm, playing a sine-
wave tone at 500 Hz. About how much sound power will the speaker produce and how
loud will the sound intensity be at 1 meter from the loudspeaker?
We can answer this question using Eq. (7.81), but first we need to know the maximum
velocity 𝑢 0 . Suppose the speaker cone is moving back and forth a distance of 0.1 mm, for
a total of 0.2 mm traveled on each vibration (0.1 mm one way and 0.1 mm back again). It
is doing this 500 times a second, meaning it is moving 500 × 0.2 = 100 mm or 0.1 meter
each second, so its velocity is about 0.1 m/s. Meanwhile the radius of the circular speaker
cone is 𝑟 = 5 cm, so its area is 𝐴 = 𝜋𝑟 2 = 0.0078 m2 . The acoustic impedance of air
is 𝑧 = 413 Pa s/m and, putting everything together and using Eq. (7.81), we get a sound
power output of 𝑃 = 0.004 watts or 4 milliwatts.
This is the total amount of energy put out by the loudspeaker per second. To calculate
the intensity at 1 meter we use Eq. (6.52), which tells us that the intensity of direct sound
from the loudspeaker at distance 𝑟 is
𝑃 0.004
𝐼 dir = = = 0.0003 W/m2 . (7.82)
4𝜋𝑟 2 4𝜋 × 12
Converted to a sound intensity level using Eq. (3.11), this is equivalent to 85 dB. So even a
small loudspeaker such as this can produce an impressive volume level. Moreover, this is
probably an underestimate, because Eq. (7.82) is based on the assumption that the sound
goes equally in all directions. As we will see in Section 7.5.6, in most cases loudspeakers
project the bulk of their sound in the forward direction, making the sound level higher for
a listener in front of the speaker.

7.5.3 Freqency response

Equation (7.81) tells us that the power output of a loudspeaker depends on the maxi-
mum velocity 𝑢 0 of the speaker cone. In Section 7.5.4 we investigate in detail how this
velocity depends on the sound waveform, but even without delving into the mathe-
matics we can understand roughly what will happen. First, the velocity of the cone
is going to be smaller for low frequencies than for high, all other things being equal,
because the cone is vibrating back and forth more slowly. This means that the power
output of a speaker will be lower at low frequencies. However, the velocity of the
cone is also lower at very high frequencies because the cone doesn’t have enough
time to respond to rapidly varying signals. At high frequencies the cone only just
manages to start moving in response to an electrical input before that input reverses
and it has to stop and move in the opposite direction. The cone’s own inertia gets in
its way and prevents it from responding.

261
Chapter 7 | Sound recording and reproduction

Output power (arbitrary units)

γ = 2000

γ = 3000

γ = 5000

20 50 100 200 500 1000 2000 5000 10000 20000

Frequency (Hz)

Figure 7.36: Output power of a loudspeaker. The output power of a loudspeaker, given by
Eq. (7.91), peaks at the resonant frequency 𝑓0 denoted by the dashed line, which is 500 Hz in
this case. The height of the peak depends on the value of the damping parameter 𝛾.

The combined result of these effects is that the output power of a speaker looks
like Fig. 7.36. It is poor both at low frequencies and at high frequencies, and best
in the middle, with a peak at a characteristic frequency 𝑓0 , indicated by the dashed
line in the figure. The characteristic frequency can be raised or lowered by changing
details of the speaker design, but one cannot escape the fact that any real speaker
only works well within a certain band of frequencies. If we want to cover the entire
frequency range, therefore, we normally have to combine two or more speakers with
different characteristic frequencies.
Even if we do this, however, we still have to contend with the fact that speakers
used for low frequencies will have lower output power than similar speakers for high
frequencies, because of the slower movement of the cone. Suppose we play a tone of
frequency 𝑓 through a speaker and suppose the distance the speaker cone moves as it
vibrates—known as the excursion—is 𝜉. Then with each vibration in and out the cone
A speaker cabinet with is moving a total distance 2𝜉 and, since it is vibrating 𝑓 times a second, the distance
separate speakers for
different frequencies of moved in one second, which is the average velocity of the cone, is 2𝜉 𝑓 . The average
sound. velocity is not the same as the maximum velocity 𝑢 0 that appears in Eq. (7.81), but
the two are proportional to one another, so the output power is

𝑃 = 12 𝐴𝑧𝑢 02 ∝ 𝐴𝑧𝜉 2 𝑓 2 . (7.83)

Thus, for given area 𝐴 and excursion 𝜉, the output power is proportional to 𝑓 2 , mean-
ing it will be higher for speakers that operate at high frequencies than for ones that

262
 7.5 | Loudspeakers

operate at low frequencies.4

We can compensate for the lower power output of a low-frequency speaker in two
ways. Either we can make the speaker larger to increase the area 𝐴 or we can increase
the excursion 𝜉, which we do by driving the speaker with a stronger electrical signal.
Most often we make the speaker larger, and this is one of the reasons why low-
frequency speakers (woofers, subwoofers, bass amplifier speakers) are often large.
Increasing the excursion is more difficult, since it can result in a distorted sound, but
in recent years manufacturers have learned how to build high-excursion speakers
that produce a clean bass sound with minimal distortion in a small package, leading
to the current generation of small portable speakers, which sound remarkably good
for their size.
At high frequencies Eq. (7.83) tells us that speakers should have ample power
even with modest size and excursion, so high-frequency speakers can be small, and
indeed work better if they are, since it reduces inertia which further improves high-
frequency performance.
The bottom line is that no speaker can be good at all frequencies, but by varying
their design one can make speakers that work well at different frequencies and then
combine them to cover the whole range.

Advanced material

7.5.4 Movement of a speaker cone where it is attached to the speaker frame. The flexible edges
of the paper cone act like a spring, allowing the cone to
The movement of a speaker cone results from a combi-
move in and out but pulling it back into place when the
nation of forces acting on it. The mathematics is quite
electrical input is turned off. To a good approximation the
similar to that for the movement of the microphone dia-
spring obeys the standard Hooke’s law, which says that
phragm in Section 7.3.7. The driving force 𝐹 𝑣 felt by the
the restoring force 𝐹𝑟 on the cone is proportional to its dis-
cone comes from the magnetic coil and is proportional to
placement 𝜉 (𝑡) from its rest position, so
the voltage 𝑉 (𝑡) sent to the speaker:
𝐹𝑟 = −𝑘𝜉 (𝑡), (7.85)
𝐹 𝑣 = 𝑏𝑉 (𝑡). (7.84)
where 𝑘 is the spring constant and the minus sign indicates
The constant of proportionality 𝑏 depends on the shape of that the restoring force is in the opposite direction to the
the coil, the number of windings, the strength and place- displacement.
ment of the permanent magnet in the speaker, and other The other force felt by the cone is a damping force,
design attributes. which acts to slow down the cone when it is moving. This
There are also two other forces felt by the speaker. One force is proportional to the velocity of the cone 𝑢 = d𝜉/d𝑡
comes from the suspension around the edges of the cone and arises from a combination of several processes. There

4 You might be concerned that Eq. (7.83) does not seem to square up with Fig. 7.36, since the equation

appears to say that power always goes up with frequency while the figure shows it going up then down.
The solution to this paradox is that the excursion 𝜉 itself goes down at high frequencies, because of the
inertia effects mentioned above, and this causes the value in Eq. (7.83) to fall off when the frequency gets
too high.

263
Chapter 7 | Sound recording and reproduction

is friction in the paper of the cone itself, as well as air resis- would hope. The amplitude 𝑢 0 of the velocity is given by
tance from the air around it, and there is also an electrical the complex magnitude of (7.89), which is
phenomenon called back electromotive force or back EMF,
in which the motion of the coil of the loudspeaker induces 𝜔𝑏𝑉0
𝑢 0 = 𝜔𝜉 0 = p . (7.90)
a current in the coil, similar to that in a dynamic micro- (𝑘 − 𝜔 2𝑚) 2 + 𝜔 2𝛾 2
phone, thereby turning some of the cone’s kinetic energy
into electrical energy and so reducing the kinetic energy. Substituting into Eq. (7.81), the output power of the
Together these effects combine to produce a damping force speaker is then given by
that can be written as 1 2 2 2
2 𝜔 𝐴𝑧𝑏 𝑉0
d𝜉 𝑃= . (7.91)
𝐹𝑑 = −𝛾 , (7.86) (𝑘 − 𝜔 2𝑚) 2 + 𝜔 2𝛾 2
d𝑡
where 𝛾 measures the magnitude of the damping and the Figure 7.36 shows how this power varies with fre-
minus sign represents the fact that the damping force is in quency for typical choices of the parameters and, as we
the opposite direction to the velocity. can see, the output is low for both high and low frequen-
Combining the three forces acting on the cone and ap- cies and peaks in the middle. There is a characteristic res-
plying Newton’s second law, we get 𝐹 𝑣 + 𝐹𝑟 + 𝐹𝑑 = 𝑚𝑎, onant frequency 𝑓0 of the speaker at which it produces its
where 𝑚 is the mass and 𝑎 the acceleration of the mov- greatest output, which we can calculate by differentiating
ing cone and its attached coil. Using Eqs. (7.84), (7.85), Eq. (7.91) with respect to 𝜔 and setting the result to zero.
and (7.86) and writing 𝑎 = d2 𝜉/d𝑡 2 , we then find that After some algebra, we find that
d2 𝜉 d𝜉 r
𝑚 +𝛾 + 𝑘𝜉 = 𝑏𝑉 (𝑡). (7.87) 1 𝑘
d𝑡 2 d𝑡 𝑓0 = . (7.92)
2𝜋 𝑚
Now suppose that the input to the speaker is a sine
wave with frequency 𝑓 , which we will represent as the real For frequencies around 𝑓0 the speaker will work well, but
part of a complex voltage 𝑉 (𝑡) = 𝑉0 ei𝜔𝑡 with 𝜔 = 2𝜋 𝑓 . if we get too far away from this frequency the speaker will
Then the steady-state solution to Eq. (7.87) takes the form have poor output power. By varying the spring constant 𝑘
𝜉 (𝑡) = 𝜉 0 ei𝜔𝑡 and, substituting into the equation and can- of the speaker cone and the mass 𝑚, we can make the reso-
celling some factors, we find that −𝜔 2𝑚𝜉 0 + i𝜔𝛾𝜉 0 + 𝑘𝜉 0 = nant frequency higher or lower and hence create speakers
𝑏𝑉0 , or that are suited for high- or low-frequency use.
𝑏𝑉0 The height of the peak in Fig. 7.36 depends on the
𝜉0 = . (7.88)
𝑘 − 𝜔 2𝑚 + i𝜔𝛾 damping parameter 𝛾, going down as the damping is in-
The velocity 𝑢 (𝑡) of the cone is equal to the derivative creased. Curves for three different choices of 𝛾 are shown
of the displacement in the figure. Having a lower peak is a good thing in some
ways—it makes the speaker’s output more uniform over
d𝜉 the frequency range—but it also reduces the overall power
= i𝜔𝜉 0 ei𝜔𝑡 ,
𝑢 (𝑡) = (7.89)
d𝑡 level, so there is a trade-off between having a speaker with
which means the velocity also follows a sine wave and a more uniform response and having one with good output
therefore produces a sine-wave sound pressure, as we power.

7.5.5 Enclosures
A practical problem with loudspeakers is that sound emanates from both the front
and the back of the speaker cone, but the waveform coming from the back is upside-
down relative to the one coming from the front. When the cone pushes on the air
at the front it pulls on the air at the back and vice versa. This gives rise to phase

264
 7.5 | Loudspeakers

cancellation as discussed in Section 4.5. Two sound waves, identical except that one
is inverted relative to the other, will cancel each other out and produce no sound. The
cancellation of sound from front and back of a speaker is not perfect: assuming the
listener is in front of the speaker the sound from the front will normally be louder
than the sound from the back, and there is typically also a time delay for the sound
to reach the listener from the back which means that the two waveforms are not
exact opposites of one another. Nonetheless, a significant amount of cancellation
can occur, reducing the apparent volume of the sound. In technical language, we say
the loudspeaker is a “dipole radiator” of sound, meaning it is a combination of high
pressure on one side of the cone and low pressure on the other.
The solution to this problem is a simple one: we mount the loudspeaker in the
wall of a closed box or enclosure. The front of the speaker faces outward from
the box and projects sound toward the listener but the back is inside where sound
cannot escape. The enclosures that loudspeakers come in are not just for looks or
convenience—they play an important role in the sound.
As discussed in Section 7.5.3, we often use multiple separate loudspeakers of dif-
ferent sizes to cover the whole frequency range, in which case all of the speakers
can be mounted in the same enclosure, which is the typical design for consumer
loudspeakers.

7.5.6 Directionality
Loudspeakers are directional, meaning that they project sound most strongly to the
front and less to the sides. This can be detrimental to sound quality for listeners who An exception to the
are not directly in front of the speaker. Moreover, the effect is more pronounced issues discussed here is
the speakers in head-
for high frequencies than for low, so that a listener situated to the side may hear a
phones, also called
distorted balance of frequencies with too much bass and not enough treble, resulting “drivers.” Because they
in a muffled sound. To see why this happens, consider Fig. 7.37, which shows a sketch are always pointed
of the sound waves emanating from a loudspeaker. directly at our ears,
off-axis sound quality is
Suppose once again that our loudspeaker is playing a sine-wave tone with fre-
normally not an issue
quency 𝑓 and consider sound traveling not directly away from the front of the speaker with headphone drivers.
but off-axis, at an angle 𝜃 as shown. For an off-axis listener hearing this sound, the
sound coming from different parts of the speaker cone has to travel different dis-
tances to reach their ears. If the speaker has diameter 𝑑 then sound from the left
edge of the speaker in the figure has to travel a distance 𝑑 sin 𝜃 further than sound
from the right edge. This means that sound waves from different parts of the speaker
are not perfectly in phase with one another and can cancel each other out, either
partially or completely, decreasing the apparent volume of the sound.
In the situation shown in the figure, for instance, the two highlighted waves, the
first and fourth ones from the left, have exactly opposite phases. Everywhere that
one wave goes up the other one goes down and vice versa. (Use the dashed lines to

265
Chapter 7 | Sound recording and reproduction

guide your eye.) Thus these two waves cancel out exactly and produce no sound. The
same is also true of the second and fifth waves, and the third and sixth—each wave
has a partner that exactly cancels it. The result is that in this case all of the sound
traveling at angle 𝜃 cancels out and a listener at this angle will hear nothing at all.
Mathematically speaking, this total cancellation hap-
pens when the phase of the waves goes through one com-
plete cycle from one side of the speaker to the other. Look
again at Fig. 7.37 and you will see that from the left side
of the speaker to the right the phase slowly varies through
an entire cycle until it comes back into phase with itself.
This means that the extra distance 𝑑 sin 𝜃 traveled by the
leftmost wave in the picture relative to the rightmost one
is exactly one wavelength 𝜆 of the sound, so 𝑑 sin 𝜃 = 𝜆,
or equivalently
𝜆
sin 𝜃 = . (7.93)
𝑑
θ This equation tells us the critical angle 𝜃 at which the
nθ

sound will completely cancel out. From the value of sin 𝜃

d si

d we can calculate 𝜃 itself by taking the arcsine.

If the listener is off axis by less than this critical angle
then the sound will not completely cancel and they will
hear the tone from the loudspeaker, although there will
be partial cancellation as they approach the cutoff, so the
Figure 7.37: Sound leaving a speaker at an an- sound will get quieter. In practice this means that there is a
gle. Sound waves emanating from a loudspeaker at region, or cone, in front of the speaker, of angular width 𝜃
an angle 𝜃 from the central axis can be partially or given by Eq. (7.93), and for clear sound we want to be inside
completely out of phase with one another and so this cone and not too near its edges. Figure 7.38 shows
cancel out. The two highlighted waves, for exam-
ple, have exactly opposite phases and hence cancel
a computer calculation of the actual sound level from a
completely. loudspeaker, with the edges of the cone marked by dashed
lines.
In practice, we would like our speakers to work no matter where the listener
is situated, which means we want to make the cone of projection of the speaker at
least 90◦ wide, so that it occupies the entire space in front of the speaker. Putting
𝜃 = 90◦ in Eq. (7.93) and noting that sin 90◦ = 1, this tells us that to be able to hear
the sound from all directions we need 𝜆/𝑑 to be significantly larger than 1, which
means the diameter 𝑑 should be significantly less than the wavelength 𝜆. This places
a limit on the size that we can make a loudspeaker if we want it to sound good in
practice. (This is in addition to considerations about the inertia of the cone discussed
in Section 7.5.3.)
From Eq. (2.4) we know that wavelength is related to frequency by 𝜆 = 𝑐/𝑓
(where 𝑐 is the speed of sound), so the size limit on a loudspeaker depends on the fre-

266
 7.5 | Loudspeakers

Figure 7.38: The sound projected by a speaker. This plot shows a computer calculation
of the sound pressure level produced by a circular speaker 25 cm in diameter (shown at the
bottom) playing a sine-wave tone at 2000 Hz. Darker shades indicate higher SPL and lighter
shades lower. The dashed lines are the theoretical edges of the cone of projection, although
most of the sound is projected well inside this cone (and a small amount outside it).

quency of the sound we want to reproduce. Higher frequencies correspond to shorter

wavelengths and hence require smaller speakers. This is another reason why high-
frequency speakers (also called tweeters) are generally smaller than low-frequency
ones (woofers). They have to be smaller to project the sound widely. Smaller speaker
area does mean less output power, as discussed in Section 7.5.2, but this is offset by
the fact that the cone moves faster at higher frequencies and produces higher sound
pressure, so usually the smaller size is not a problem.
The calculation here is not exactly correct because it does not take account of the
fact that the speaker is circular. A full calculation, which you can find in Section 7.5.8,
gives a slightly different answer, that in fact a loudspeaker should be smaller than
1.22𝜆 in order to project sound in all directions. In practice, the difference is minor.
It’s fine to use the full formula, but the rough rule of thumb that 𝑑 should be less
than the wavelength is easier to remember and is good enough for most purposes.

267
Chapter 7 | Sound recording and reproduction

Example 7.7: Sizes of loudspeakers

Consider sound in the bass frequency range, at say 250 Hz. How large can a loudspeaker
be if it is to work well in this frequency range?
The corresponding wavelength is 𝜆 = 343/250 = 1.4 meters and the speaker should
be significantly smaller than 1.22𝜆 = 1.7 meters, so a safe choice in this case would be to
make the speaker well under meter in diameter. In practice, the largest size one sees for
bass speakers is about half a meter.
And what about the other end of the scale, at high frequencies? At the top of the human
hearing range around 17 000 Hz, the wavelength is 𝜆 = 343/17 000 = 2 cm and the speaker
should be smaller than 1.22𝜆 = 2.5 cm. In practice, tweeters are right around this size
and not much smaller, which means we are sacrificing some off-axis performance at the
high end of the frequency range. The musical contribution of the very highest frequencies,
above say 15 000 Hz, is only modest however—and for many people actually inaudible—so
this is perhaps a reasonable compromise.

7.5.7 Design of loudspeaker systems

Putting together what we have learned, we can now say something about the design
of complete speaker systems. There are three main issues at play. First, as we have
seen, the mechanics of the speaker cone favors a band of frequencies centered on
the characteristic frequency 𝑓0 , whose value can be varied to create a speaker that
favors low-, mid-, or high-frequency bands, but not all of them at once. If we want to
cover the entire frequency range we normally have to combine two or more speakers
tuned to different bands.
Second, speakers generally produce stronger sound at high frequencies than low,
all other things being equal, because the speaker cone moves faster at high frequen-
cies and generates higher sound pressure. We can compensate for the lower output
power of low-frequency speakers either by increasing the area of the speaker or
by increasing its excursion—the range of movement of the cone—which we do by
driving the speaker with a stronger electrical signal. Historically, speaker design-
ers usually took for the former approach and built larger speakers, since increas-
ing the excursion posed some technical challenges. These difficulties have how-
ever mostly been overcome in recent years, leading to the creation of small, high-
excursion speakers that offer good bass output despite their diminutive size.
The third issue is directionality. Speakers tend to project sound most effectively
to the front and less to the sides, and this effect is more pronounced for large speak-
ers and at high frequencies. In order to project well over a wide angle, the diameter
of a speaker should be less than the wavelength of the sound, which means in prac-
tice that speakers intended for high frequencies must be small—a few centimeters
across at most. Speakers intended for low frequencies on the other hand can be large
without compromising directionality.

268
 7.5 | Loudspeakers

Combining these observations we can now gain an understanding of how to build

an effective speaker system. First, there are some situations where we need to re-
produce only a limited frequency range. Amplifiers for bass guitar, for instance,
typically aim to reproduce sounds only up to about 1000 Hz, which can be done with
a single speaker, usually large since directionality is not an issue and higher output
power is desirable—30 or 40 cm speakers are not unusual. In the more common case
where we want to reproduce the entire audible frequency range, we combine mul-
tiple speakers, small tweeters for the high frequencies and woofers and subwoofers
for the low frequencies which may or may not be larger. Traditional speaker designs,
including most high-quality consumer models produced until the 1990s, used large
low-frequency speakers for their superior output power, resulting in good sound but
bulky construction. Commercial amplification and public address systems, for which
space and weight are not an issue, still use this approach. Recent years, however, have
seen the rise of small portable speakers for home use that produce surprisingly good
sound even in the bass range by using larger excursions instead of larger size.
Finally, in order to prevent sound from the back of a speaker cancelling sound
from the front, it is important as we have seen that all speakers be mounted in a box
or enclosure, the front of the speaker facing outward while the back is sealed inside
the box to prevent sound escaping.

Example 7.8: Designing a two-speaker system

Suppose we want to build a full-frequency-range loudspeaker system with two individual
speakers in a single enclosure, a woofer and a tweeter. The woofer is to have a characteris-
tic frequency of 𝑓𝐿 = 400 Hz and a maximum frequency of 2000 Hz, while the tweeter is to
have characteristic frequency 𝑓𝐻 = 7000 Hz and span the range from 2000 Hz to 18 000 Hz.
Also the tweeter will be limited to a maximum excursion of 1 mm, which is typical for
high-frequency speakers. Suggest suitable sizes for the two speakers that will give good
directionality for both. What will the maximum excursion of the woofer need to be if we
want the maximum output power to be the same for both speakers at their characteristic
frequencies?
A number of factors come into play here. The maximum size of the speakers is dictated
by directionality, as discussed in Section 7.5.6. For best performance we need them to be
smaller than 1.22 times the wavelengths of their highest frequencies, which are
𝑐 343
𝜆= = = 17 cm (7.94)
𝑓 2000
for the woofer and
343
𝜆= = 1.9 cm (7.95)
18 000
for the tweeter. This gives us upper limits of 21 cm and 2.3 cm on our speakers, so let us
say the diameters are 15 cm and 1.5 cm, to give us some margin for error.
At the same time, Eq. (7.83) tells us that output power at frequency 𝑓 is proportional
to 𝐴𝜉 2 𝑓 2 , where 𝐴 is the area and 𝜉 is the excursion. Writing the area in terms of the

269
Chapter 7 | Sound recording and reproduction

diameter 𝑑 as 𝐴 = 14 𝜋𝑑 2 , we then have output power 𝑃 ∝ 𝑑 2 𝑓 2 𝜉 2 . If we want our two

speakers to have the same power we then need

𝑑𝐿2 𝑓𝐿2 𝜉 𝐿2 = 𝑑𝐻
2 2 2
𝑓𝐻 𝜉𝐻 . (7.96)

Rearranging for the excursion 𝜉 𝐿 of the low-frequency speaker, we get

𝑑𝐻 𝑓𝐻 1.5 × 7000
𝜉𝐿 = 𝜉𝐻 = × 1 mm = 1.75 mm. (7.97)
𝑑𝐿 𝑓𝐿 15 × 400
So the woofer needs to have a maximum excursion of about 2 mm, twice that of the tweeter.

Advanced material

7.5.8 Directionality of a circular speaker proportional to sin(2𝜋 𝑓 𝑡), then the delayed sound from
our slice will have pressure proportional to sin[2𝜋 𝑓 (𝑡 −
Calculation of the directionality of loudspeakers is made (𝑥/𝑐) sin 𝜃 )], which can be conveniently represented as the
more complicated by the fact that most speakers are circu- imaginary part of the complex expression
lar. Consider a slice of a circular speaker cone a distance 𝑥
from the center line and of width d𝑥, thus: ei2𝜋 𝑓 (𝑡 − (𝑥/𝑐 ) sin 𝜃 ) = ei𝜔𝑡 e −i𝑘𝑥 sin 𝜃 , (7.98)
where 𝜔 = 2𝜋 𝑓 and 𝑘 = 2𝜋 𝑓 /𝑐 = 2𝜋/𝜆.
√Pythagoras’ theorem tells us that the length of the slice
is 2 𝑟 2 − 𝑥 2 where 𝑟 is the radius of the speaker and, as-
suming an equal amount of sound from each part of the
θ slice, we can add up all of the incoming sound to calculate
dx
the sound from the whole speaker. (There will be a decay
of intensity with distance to the listener, due to the inverse
square law, as described in Section 3.5, but in the far field
this will be approximately the same for sound from all parts
of the speaker, so we can ignore it.) The total sound pres-
sure felt by the listener will then be
For a listener situated at an angle 𝜃 off axis, the extra dis- ∫ 𝑟 p
tance traveled to the listener by sound from this slice, com- 𝑝 (𝑡) ∝ ei𝜔𝑡 e −i𝑘𝑥 sin 𝜃 𝑟 2 − 𝑥 2 d𝑥
−𝑟
pared to sound coming from the center line, is 𝑥 sin 𝜃 . We
i𝜔𝑡𝐽1 (𝑘𝑟 sin 𝜃 )
ignore the slight variation in 𝜃 for different 𝑥 due to the = 𝜋e 𝑟 , (7.99)
changing direction to the listener, as well as the small dif- 𝑘 sin 𝜃
ference in distance from different points along the slice. where 𝐽1 (𝑥) is a Bessel function, a special function that
This approximation, known as the “far field,” is a good one appears commonly in calculations with circular symme-
as long as the listener is sufficiently far away, which in try such as this one. The intensity 𝐼 of the sound heard
practice means that the distance to the listener should be by the listener is proportional to the magnitude squared of
much greater than the speaker diameter, and this is true in the sound pressure, meaning that
most settings. (We briefly discuss two situations where it
𝐽1 (𝑘𝑟 sin 𝜃 ) 2
 
is not at the end of this section.) 𝐼 = 𝐴𝜋 2𝑟 2 , (7.100)
𝑘 sin 𝜃
Given that sound travels at speed 𝑐, it will take a
time (𝑥/𝑐) sin 𝜃 to travel the extra distance. If the loud- where 𝐴 is a proportionality constant. Making use of the
speaker is producing a sine-wave tone with sound pressure fact that 𝐽1 (𝑥)/𝑥 → 21 as 𝑥 → 0, the value of this intensity

270
 7.5 | Loudspeakers

0.8
Intensity I/I0

0.6

0.4

0.2

−60 −30 0 30 60
Angle θ (degrees)

Figure 7.39: Angular output of a circular loudspeaker. The intensity of the sound from a
circular loudspeaker 25 cm in diameter playing a 2000 Hz sine-wave tone, as heard by a listener
at an angle 𝜃 off-axis and measured as a fraction of the on-axis intensity 𝐼 0 .

on-axis at 𝜃 = 0 is 𝐼 0 = 41 𝐴𝜋 2𝑟 4 , and hence to 𝜃 = 90◦ , which means sin 𝜃 = 1. Making this choice
in the equation above tells us that we would need the di-
𝐽1 (𝑘𝑟 sin 𝜃 ) 2
 
𝐼 = 4𝐼 0 . (7.101) ameter of the speaker to be significantly less than 1.22𝜆.
𝑘𝑟 sin 𝜃 This is close to the answer we got in Section 7.5.6, where
Figure 7.39 shows a plot of this quantity for the case our rough calculation indicated that the speaker should be
of a 25 cm loudspeaker playing a tone at 2000 Hz. As we smaller than 𝜆. The calculation here is more correct, taking
can see, the loudspeaker only produces significant inten- proper account of the circular shape, but the simple rule of
sity within an angle of about 30◦ . This would not be a thumb that the speaker should be significantly smaller than
satisfactory speaker in practice, at least for sound at this the wavelength is good enough for most purposes.
frequency. The sound would be too attenuated for off-axis The far-field approximation used here is a good one for
listeners. most loudspeakers. Consumer speakers in people’s homes,
The first zero of the Bessel function 𝐽1 (𝑥) occurs at musical instrument amplifiers, and public address systems
𝑥 = 3.8317, which means Eq. (7.101) falls to zero when are all normally heard by listeners at some distance from
𝑘𝑟 sin 𝜃 = 3.8317. Rearranging and setting 𝑘 = 2𝜋/𝜆 and the speaker. Two situations where this is not the case are
𝑟 = 21 𝑑 where 𝑑 is the diameter of the speaker, we find that headphone speakers, which are mounted close to the lis-
the angular width 𝜃 of the cone of projection of the speaker tener’s ears, and studio monitor speakers used by mixing
satisfies engineers, which are normally placed on the desk imme-
𝜆 diately in front of the engineer in order to minimize the
sin 𝜃 = 1.220 . (7.102)
𝑑 effect of room acoustics. Calculations for these speakers
For the case in Fig. 7.39, for instance, with 𝑑 = 0.25 m and are different and the speakers themselves must be designed
𝑓 = 2000 Hz, this gives 𝜃 = 56.8◦ , and we need to be well to allow for the fact that they will be used in a different
within this angle for good sound from the speaker. (Fig- way. Studio monitors are sometimes labeled as “near-field”
ure 7.39 suggests that about 30◦ or less would be desirable, monitors, to emphasize the fact that they are designed to
as we have said.) be heard close up.
As discussed in Section 7.5.6, we would ideally like the
sound projected by our loudspeaker to extend all the way

271
Chapter 7 | Sound recording and reproduction

7.5.9 Phased arrays

Directionality is not always a bad thing in loudspeakers. Particularly for large public
address systems, such as are used at live music concerts, being able to project sound
in a particular direction can be good. For instance, in an open-air performance sit-
uation there is little point projecting sound from the stage up into the sky when all
the listeners are on the ground. Large-diameter loudspeakers can be useful in such
situations for their directionality, but a more common approach is to use a phased
array.
A phased array, also called a line array in the music industry, is a line of speakers,
usually stacked vertically, all playing the same sound. In effect the phased array acts
as a single tall narrow speaker. Because of its great height—often several meters—the
array projects into a narrow angle vertically, but because its width is comparatively
small it projects into a wide angle horizontally. The result is that the phased array
projects a horizontal sheet of sound that is perfect for broadcasting to an audience
on the ground or floor of a performance space. All of the power generated by the
speakers is concentrated on the audience and none is wasted on the sky.
A phased array consist-
ing of eight speaker en- Example 7.9: Phased arrays
closures stacked verti-
cally. A phased array speaker is 3 meters tall and 50 cm wide. Into approximately what angle
does it project sound horizontally and vertically at 1000 Hz?
For the purposes of this question we can treat the phased array as a single rectangular
speaker, which therefore obeys Eq. (7.93) on page 266. The wavelength of sound at 1000 Hz
is 𝜆 = 𝑐/𝑓 = 343/1000 = 34 cm and, setting the diameter to 50 cm, the horizontal angle
is sin 𝜃 = 34/50 = 0.69. Taking the arcsine then gives us 𝜃 = 43◦ , which is an acceptable
projection angle for a concert stage. On the other hand the vertical angle has sin 𝜃 =
34/300 = 0.115, which gives 𝜃 = 6.6◦ , a much narrower angle. So this speaker will project
a sheet of sound that is very narrow vertically but covers a wide swath of the audience—
exactly what a phased array is designed to do.

7.6 Stereo recording and surround sound

When we hear a live musical performance, sound often comes at us from many dif-
ferent directions. The musicians in a band or orchestra may be scattered about a
stage, for instance, and the ear makes use of the different directions of sounds to
tease apart the music and interpret what we are hearing, as discussed in Section 5.5.
Sound may also be reflected off the walls, floor, and ceiling of a room as discussed
in Section 6.7, coming at us from all directions and creating the sensation known as
“envelopment,” which contributes substantially to the impact of live sound.
Conversely, recorded sound, played over a single loudspeaker—so-called mono-
phonic sound—comes at us, by and large, from just one direction. Sound may bounce

272
 7.6 | Stereo recording and surround sound

off the walls to create a certain degree of envelopment, but most recorded music is
heard in acoustically dry environments such as people’s houses and cars, which re-
flect little sound. The result is that sound from a single speaker lacks some of the
elements that lend live music both its grandeur and its clarity. Stereophonic sound
(or stereo for short) and its sibling surround sound are an attempt to recreate these
lost elements, at least in part, by simultaneously playing sound over two or more
loudspeakers spread out around a room.
Developed from the 1930s onward and finding
widespread use starting in the 1960s, stereo sound is Center Rig
L e ht
played over two loudspeakers. Two slightly different
recordings of the same music are made and played back
over each of the speakers, creating a stereo image—the
acoustic impression of sound coming from a wide spatial
field. Ideally the listener sits (or stands) with the speak-
ers in front of them and separated by a certain angle—
60◦ is often recommended. The sound of an instrument
played only on the left speaker will appear to be coming Listener
◦
directly from that speaker, i.e., from 30 to the left, or the
11 o’clock position. Similarly sound played only over the Rear le Rear right
right speaker will appear to be coming from 30◦ to the
right, or the 1 o’clock position. And sound that comes
equally from both will appear to be coming from a cen- Subwoofer
tral position, 12 o’clock, directly in front of the listener.
By careful adjustment of the relative signal strength in the Figure 7.40: A typical speaker arrangement for
surround sound. In a typical six-speaker surround
two speakers, one can make sound appear to emerge from
◦ sound arrangement there are left and right speakers
any direction in the 60 stereo field, and by giving differ- in front of the listener at about 30◦ to either side, a
ent instruments a different balance of left and right we can center speaker, and two rear speakers about 110◦ to
make them each appear to be located in their own spatial either side. The optional subwoofer can be placed
position and hence create the impression of a band or or- almost anywhere, since the ear is insensitive to the
chestra arrayed in space in front of the listener. direction of the very low-frequency sounds it pro-
duces.
Stereo can capture the sounds of different instruments
in different locations, but it cannot capture the envelop-
ment produced by reverberated sound reaching the listener from all directions in a
concert hall or performance space. For that we need surround sound, an extension of
stereophonic sound that makes use of more than two speakers, some placed behind
the listener, so that he or she is surrounded by sound on all sides. The most common
version of this idea uses six speakers as shown in Fig. 7.40. There are left and right
stereo speakers as normal, a center speaker in between those two, and another stereo
pair behind the listener to the left and right. The sixth speaker, which is optional, is
a subwoofer, devoted to the production of the lowest sound frequencies. Because

273
Chapter 7 | Sound recording and reproduction

the ear is poor at telling which direction low-frequency sound comes from (see Sec-
tion 5.5.2), the subwoofer can be placed almost anywhere within hearing range. Its
location is not important for the surround-sound effect.
You might wonder what the point is of the center speaker if we already have left
and right speakers on either side of it. We can produce the effect of sound coming
from the center of the stereo field by playing it equally through the left and right
speakers, so why do we need the center speaker? In fact, for musical applications we
do not really need it. Surround-sound systems without a center speaker are fine for
listening to music. The main role of the center speaker is in audiovisual applications,
such as watching TV or movies. For the sake of realism, it is important when watch-
ing a movie that sound from events seen on the screen, such as actors talking, appear
to come from the screen, i.e., from the center of the stereo field. This can, as we say,
be achieved by playing the sound equally over the left and right speakers, but this
only works if the listener is situated in the perfect listening position exactly in front
of the speakers. As we saw in Section 5.5, perception of the direction of sound is
based on differences in loudness and timing between a listener’s two ears, and both
loudness and timing can be affected when the listener is off-center. A listener sitting
to the left of center, for instance, will hear sound from the left speaker that is both
louder than the right and arrives sooner, both of which will make the sound seem to
come from the left side of the stereo field and not the center. For a movie watcher
this can have the disorienting effect of making actors’ voices appear to be coming
from a different place from where their mouths are. The center speaker rectifies this
problem by providing a direct sound source that is always in the same place as the
screen. In surround-sound film and video most dialog and other on-screen sound
is played through the center speaker. As an additional benefit, this approach also
allows one to separately adjust the volume of the dialog relative to other sounds by
changing the volume of the center speaker.
In movies, surround sound is used to generate special effects such as footsteps
coming from behind the audience or a plane flying overhead. When used for music,
however, its primary virtue lies in its ability to create envelopment. By playing direct
sound from the musicians over the front speakers and reverberated sound, as if re-
flected off the walls, over the rear speakers, one can create an approximate facsimile
of the experience of hearing live music in a large concert venue.
This at least is the ideal. The practice, unfortunately, often falls short, for a variety
of reasons. First, music has to be recorded in stereo or surround sound to get a stereo
or surround effect. Since the late 1960s, most music has been recorded in stereo, so
stereo recordings are now very common, but surround-sound recordings are not.
Music for film soundtracks, intended to be heard over the multispeaker systems in
movie theaters, has long been recorded with surround sound in mind, but surround-
sound music for home listeners has only started to become widely available in the

274
 7.6 | Stereo recording and surround sound

streaming music era since the mid 2010s.

Second, to get the full effect of stereo or surround sound, a listener has to be posi-
tioned exactly in the middle of the speakers, equidistant from left and right channels.
While this may be possible when watching a movie, it is not common when listen-
ing to music. Busy people listen to music while doing other things and are rarely
sitting still in the ideal position. And a lot of music is heard in the car, where it may
not even be physically possible to sit half way between the speakers. Moreover, al-
though stereo hifi systems were the dominant form of consumer audio equipment
until the early 21st century, the increasing popularity of smaller portable and wire-
less speakers has made true stereo sound rarer in the home in recent years. Some
portable speakers do make an attempt to project stereo sound using clever internal
machinery, but the effect is inferior to a true stereo speaker pair with wide separa-
tion and the resulting sound is essentially only monophonic, especially when heard
from some distance away.
A more subtle issue arises when one listens to music over headphones or ear-
phones. In that case one normally hears the sound intended for the left stereo speaker
in the left ear and the right speaker in the right ear. This, however, produces a com-
pletely different sensation from listening over loudspeakers. When you hear sound
from a loudspeaker, that sound will, to some extent at least, reach both of your ears.
Even if the speaker is located to your left, for instance, you will hear some sound
from it in your right ear. This allows you to tell what direction the sound is coming
from—as discussed in Section 5.5, your brain uses differences in loudness and tim-
ing between your ears to determine direction. When you listen to recordings over
headphones, on the other hand, your brain can no longer do this, since the sound for
each speaker is now heard by only one ear. The result is that, while your two ears
do hear different things when listening on headphones, the stereo image is severely
compromised. It is possible to rectify the situation using special “binaural” record-
ings designed to be heard on headphones, that deliberately introduce differences in
loudness and timing between the two ears to mimic the effect of sounds coming from
different directions and trick the brain into perceiving a stereo field. Indeed binaural
recordings arguably give the best illusion of space and envelopment in music, better
than standard stereo recordings played over loudspeakers, because of the ability to
control very tightly what each ear hears. Unfortunately, true binaural recordings are
rare, although in recent years a number of manufacturers have introduced hardware
that can mimic binaural audio electronically and create an illusion of surround sound
over headphones or earphones.
Technologies were developed in the first half of the 20th century to allow stereo
and surround sound to be captured on magnetic tape and vinyl records. On tape the
principle is simple: two or more separate “tracks” of sound are captured side by side
on the same tape by magnetizing different strips of the tape with different recording

275
Chapter 7 | Sound recording and reproduction

heads. Playing back the tape similarly involves using two or more heads to recover
the separate recordings, along with two or more amplifiers to amplify them and two
or more speakers to produce the sound.
For vinyl records things are a little more complicated. Instead of us-
Le Right ing a groove that moves from side to side only, the groove on a stereo
record is also allowed to move up and down. The two channels of the
Needle stereo are encoded in diagonal movements of the groove up-and-to-
the-left or down-and-to-the-left, as shown in Fig. 7.41. The record is
Record
played back with a needle that picks up motion along the two diagonal
directions separately, creating two separate signals that are amplified
for playback over separate speakers. This design allows the record to
hold two independent channels while still ensuring that a monophonic
Figure 7.41: The groove in a stereo
record, in which the groove moves from side to side only, will play cor-
LP. Left and right channels of a stereo
recording are encoded in the groove of rectly on a stereo record player, with sound coming from both channels.
a record as movements of the groove With digital recordings, there are various ways to capture stereo
diagonally, up and to the left for the sound, the simplest being just to store digital versions of the right and
left channel, down and to the left for left stereo channels separately. A similar approach can be used for sur-
the right channel.
round sound, allowing us to store six separate channels or even more.
We discuss digital sound in more detail in Chapter 8.

Chapter summary:

• The recording and reproduction of sound is achieved by turning sound into

analogous electrical signals and then back into sound again.
• The electrical signals can be manipulated in a variety of useful ways using
electronic circuits, for instance to increase or decrease the volume or filter out
certain frequencies from the sound.
• Microphones capture sound and turn it into electrical signals. A microphone
consists of a diaphragm, usually made of mylar, that senses pressure changes, a
capsule to contain the diaphragm, and some means of detecting the movement
of the diaphragm.
• Microphones come in two main varieties. Pressure microphones directly
measure pressure on the diaphragm using a capsule closed at the back, and
are omnidirectional. Pressure-gradient microphones measure the pressure
difference between the two sides of the diaphragm using an open capsule, and
have a cardioid or bidirectional pickup pattern.

276
 7.6 | Stereo recording and surround sound

• There are also two main techniques for measuring the movement of the dia-
phragm. Condenser microphones use electrical capacitance between the
diaphragm and a metal backplate to convert movement into electrical voltage.
Dynamic microphones use electromagnetic induction, with a coil mounted
on the back of the diaphragm and a permanent magnet inside the capsule.
• Recording of the electrical signal representing a sound can be achieved by a
variety of means. The two dominant recording technologies of the twentieth
century were magnetic tape, which magnetizes a moving strip of tape using
an electromagnet, and vinyl records, which capture the sound waveform in
the pattern of wiggles in a groove cut on the surface of a plastic disk. Both of
these technologies are now obsolete, having been replaced by digital recording
techniques, which are discussed in the next chapter.
• Electrical signals can be turned back into sound using a loudspeaker. Loud-
speakers are the weak link in the sound reproduction chain: it is almost impos-
sible to make a single loudspeaker that can faithfully project a strong sound at
any frequency in any direction, and the design of speaker systems involves a
number of compromises.
• By far the most common type of loudspeaker is the dynamic loudspeaker,
which operates like a dynamic microphone in reverse: an electrical signal
passed through a coil produces a magnetic field that is attracted and repelled
by a nearby permanent magnet, causing the coil to vibrate. The coil is attached
to a diaphragm or cone which thus vibrates as well, producing sound.
• Depending on their size and design, different loudspeakers work well in dif-
ferent frequency bands, but no speaker works well at every frequency, so
complete speakers systems normally combine two or more speakers to cover
all the bases.
• There are also additional issues that must be addressed for both high- and low-
frequency speakers. At high frequencies the sound produced by a speaker
is highly directional, projecting mainly to the front of the speaker and not to
the sides. This issue can be addressed by making the speaker smaller. Direc-
tionality is less of a problem at low frequencies, but low-frequency speakers
have their own issues, suffering from lower output power because the cone
moves with lower velocity and hence produces less sound pressure. This issue
can be addressed either by making the speaker larger, since output power is
proportional to the area of the cone, or by increasing the range of movement
of the cone—also called the excursion—which requires careful engineering to
avoid distortion.

277
Chapter 7 | Sound recording and reproduction

• Complete speaker systems normally combine separate high-frequency

speakers (called tweeters), which are small, and low-frequency speakers (called
woofers) which may be either large or small. Historically, woofers were large,
to give them higher output power, but recent years have seen the introduction
of smaller high-excursion ones, leading to today’s portable speaker systems.
• Many modern speaker systems also combine speakers in different locations to
create stereophonic or surround sound—the illusion of sound coming at the
listener from different directions.

Exercises

7.1 In this circuit 𝑉in = 5 volts, 𝑅1 = 600Ω, and 𝑅2 = 400Ω:

R1

Vin R2 Vout

a) What is 𝑉out ?
b) If we want 𝑉out = 1 volt with 𝑅1 still being 600Ω, what should 𝑅2 be?

7.2 We send a pure sine-wave signal into a 6 dB-per-octave filter like the one in Fig. 7.8.
a) If the frequency of the sine wave is exactly equal to the cutoff frequency 𝑓𝑐 of the filter,
by how many decibels is the intensity of the corresponding sound decreased?
b) What about if the frequency is 2𝑓𝑐 ?
c) If the filter is a 12 dB-per-octave one, as in Fig. 7.12, how much is the intensity cut by
at 𝑓𝑐 ?

7.3 Advanced Consider a band-pass filter like the one in Fig. 7.15 on page 229, built out of a
low-pass filter and a high-pass filter one after the other, and let us denote the cutoff frequencies
of the two filters by
1 1
𝑓𝐿 = , 𝑓𝐻 = .
2𝜋𝑅1𝐶 1 2𝜋𝑅2𝐶 2
Show that the center frequency 𝑓𝐶 of the band-pass filter, which is the frequency at which
the filter
p passes the largest fraction of the input signal—the center of the peak in Fig. 7.16—is
𝑓𝐶 = 𝑓𝐿 𝑓𝐻 .

278
 Exercises

7.4 Microphones come in two main types, pressure microphones and pressure-gradient micro-
phones.
a) How do the two types differ in terms of the direction in which they pick up sound?
If you have a choice between these two microphone types, which would be better in each of
the following applications?
b) The microphone on a laptop used for videoconference calls.
c) A microphone used on a guitar amplifier for live performance.
d) A microphone used for recording vocals in the studio.
e) A headset microphone for an operator in a call center.
f) A microphone used for recording a full symphony orchestra in a concert hall.

7.5 A cardioid microphone picks up sound most strongly from the direction directly in front
of the diaphragm—the 0◦ point in a polar plot like Fig. 7.25 on page 241. By now many decibels
will the sound level drop off at (a) 45◦ , (b) 90◦ , and (c) 135◦ ?
7.6 A hypercardioid microphone has an open-backed capsule like a cardioid microphone,
but the diaphragm is not at the front of the capsule. Instead it is 31 of the way back.
a) Show that the net sound pressure felt by a hypercardioid mic for sound coming from
an angle 𝜃 off-axis is proportional to cos 𝜃 + 31 .
b) Hence show that the microphone picks up no sound at all from an angle of approxi-
mately 109.5◦ off-axis.

7.7 Advanced Either by hand or using a plotting program, make polar plots of the direc-
tional response of a cardioid microphone in the high-frequency region where Eq. (7.54) applies
and (a) 𝑓 𝑑/𝑐 = 21 , (b) 𝑓 𝑑/𝑐 = 1.
7.8 Consider a condenser microphone of the kind discussed in Section 7.3.5.
a) What is the normal voltage of the electricity that such a microphone runs on?
b) How large is the typical separation between the diaphragm and backplate of such a
microphone?
c) The typical voltage output by a condenser microphone (before amplification) is about
a tenth of a volt. Hence, about how far does the diaphragm move back and forth as
sound hits it?

7.9 For a condenser microphone it is important, as discussed in Section 7.3.5, that the ca-
pacitor formed by the diaphragm and backplate charge up only slowly—more slowly than the
rate at which the diaphragm vibrates. In practice, this means that the time constant 𝑅𝐶 of
the resistor-microphone combination in the circuit of Fig. 7.30a on page 247 must be longer
than the period of the lowest-frequency sound waves. A typical capacitance for a condenser
microphone is 50 picofarads. Hence what value should one use for the resistor?
7.10 A professional analog tape recorder, of the kind widely used in recording studios in the
mid 20th century, has a gap of 5 𝜇m between the poles of the tape head. At what speed should
the tape move in order to fully capture the entire frequency range of audible sound?
7.11 The earliest commercial vinyl records were 10 inches in diameter and rotated at 78 rev-
olutions per minute. The playing area extended from the rim of the record inward about 2.5

279
Chapter 7 | Sound recording and reproduction

inches, which is about 6.4 cm. If the recording is to have a dynamic range of 50 dB, approxi-
mately what would be the maximum playing time of such a record?
7.12 A portable radio contains only one loudspeaker, which is 10 cm across. A listener listens
to music on the radio from a point not directly in front of the speaker but 45◦ off axis.
a) Up to about what frequency will they hear the sound clearly?
b) Will this matter from a musical point of view?

7.13 As we have seen, loudspeakers come in a range of sizes.

a) Explain briefly the advantages and disadvantages of large loudspeakers versus small
ones, paying particular attention to the relative merits of each for reproducing high
and low frequencies.
With these issues in mind, approximately what size speakers should one use for the following?
b) A “tweeter” intended for frequencies above 5000 Hz.
c) A speaker for a bass amplifier that needs to capture frequencies up to 500 Hz.
d) The “driver” speaker in a set of over-the-ear headphones.

7.14 A line array speaker, like the one pictured on page 272, projects most of its sound into
a narrow horizontal sheet, with only a small vertical angle and little sound going up or down.
If a particular line array is 2.5 meters high, about how wide will that angle be in degrees, at
1000 Hz?

280