Matthew Schwartz

Lecture 7:
Music

1 Why do notes sound good?

In the previous lecture, we saw that if you pluck a string, it will excite various frequencies. The
amplitude of each frequency which is excited will be proportional to the coefficient in the
Fourier decomposition. In this lecture we will start to understand how different frequencies com-
bine to produce music. This lecture is best studied alongside the Mathematica notebook
music.nb on the isite.
In the first section “playing notes” of the notebook, you can listen to a pure frequency (C4 =
middle C = 261 Hz). It sounds pleasant, but not particularly interesting. Now play the “square
wave” version of middle C – you should notice that it sounds somewhat tinny and unpleasant.
These are the same notes, but different sounds. Why do they sound different? One way to
understand the difference is to compare the Fourier decomposition of the sine wave and the
square wave:

1.0 1.2
1.0
0.5 0.8
0.0 0.6
0.4
- 0.5 0.2
-1.0 0.0
- 0.2
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Sine wave – one frequency only Square wave – lots of frequencies, sounds tinny
Figure 1. Comparison of Fourier modes for sin and square wave

For both the sine wave and the square wave, the dominant mode is the fundamental. How-
ever, the square wave has lots of other modes which make the note sound less pure. One way to
understand why the square wave sounds worse is that has many high frequency notes with sig-
nificant amplitude. It is hard for our brains to process all these high frequency notes, so we find
it jarring. In fact, if all frequencies are present at once we get so-called white noise. White
noise is perhaps as unmusical as you can get.
Now consider playing two notes at once. In the “playing pairs of notes” section, play the 250
Hz and 270 Hz notes at the same time. It doesn’t sound great. Why?

1
2 Section 2

The problem is that you hear a rattling around 20 times a second. This 20 Hz rattling is the
beat frequency between the 250 Hz and 270 Hz. Indeed,

cos(270 Hz 2πt) + cos(250 Hz (2πt)) = 2 cos(10 Hz 2πt) cos(260 Hz2πt) (1)

This the sum of those two notes oscillates at 10 Hz and at 260 Hz. The 10 Hz oscillation (which
we hear as a 20 Hz beat frequency) is jarring – your mind tries to process it consciously. Fre-
quencies as high as 260 Hz do not have this effect.
Thus there seem to be two reasons sounds appear unmusical:

• Too many frequencies are present at once

• Beating occurs at frequencies we can consciously process.

This is a physics class, not a biology class, so we will not try to explain why these facts hold.
We merely observe that whenever each criteria is satisfied, sounds appear unmusical. However,
now that we have defined the problem, we can start to study music scientifically.

2 Dissonant and consonant note pairs

Now we’re ready to study music. If we had a pure sine wave at 300 Hz, then it sounds nasty
when played at the same time as a sine wave of 320 Hz. However, if we play it with a sine wave
of 580 Hz it does not sound so bad. That is because

cos(300 Hz 2πt) + cos(580 Hz 2πt) = 2 cos(140 Hz 2πt) cos(440 Hz 2πt) (2)

The beat frequency 2 × 140 Hz = 280Hz is not low enough to be harsh – it is just a note (try the
Mathematica notebook). On the other hand, if we played 300 Hz and 580 Hz on an actual
instrument it would sound horrible.
We can see why from studying our plucked string example. Recall that for a string plucked
1
near the end, the relative Fourier coefficients scale like n . So the dominant frequency n = 1 (the
fundamental) has only twice the amplitude of the first harmonic (n = 2). Thus playing 580 Hz
along side a plucked string would give
X 1
f (t) = cos(580 Hz 2πt) + cos(300n Hz 2πt) (3)
n
n=1

Writing T = Hz 2πt to clean up the equation and expanding the sum

1
f (t) = cos(580 T ) + cos(300 T ) + cos(600 T ) + ··· (4)
2

Let us combine the 580 Hz oscillation with the 600 Hz oscillation using trig sum rules. Using
1 3 1
cos(580T ) + cos(600T ) = cos(10T )cos(590T ) + sin(10T )sin(590T ) (5)
2 2 2
we find
∞
3 1 X 1
f (t) = cos(300T ) + cos(10T )cos(590T ) + sin(10T )sin(590T ) + cos(300nT ) (6)
2 2 n
n=3

Now we see beating at 2 × 10 Hz = 20Hz, which is audible and jarring. There is beating between
the 580 Hz note and the first harmonic of the plucked string. The point is that with pure sine
waves, no harmonics are excited, but with real instruments they are.
Dissonant and consonant note pairs 3

In general, instruments will have significant amplitudes for many harmonics. Here is the
Fourier spectrum of a flute

Figure 2. Spectrum of the note G4 on a flute.

This is a flute playing the note G4 which is 392 Hz. You see that not only is the funda-
mental frequency (the G4) largely excited, but many other modes have significant amplitudes.
These modes are all the higher harmonics. The harmonics determine what an instrument sounds
like – its timbre. Timbre is what a note sounds like when played, as distinguished from its
pitch, which is the fundamental frequency played, and intensity, which is the power going into
sound from the instrument. We will study the timbres of different instruments in lecture, in
section and on psets. The pitch of a note is the frequency with highest intensity, which is usu-
ally the lowest frequency where there is a peak. If the first peak is not the highest, what the
pitch is can be somewhat subjective. The relative heights of the different peaks and their Q
values is the timbre, and the absolute scale is the intensity. All of this can be read off the
Fourier spectrum, as in Figure 2 for a flute. We’ll discuss timbre more later, but for now, the
main point is that most instruments will have integer multiples of the fundamental frequency
excited with significant amplitudes; it is these harmonics which are the key to the scale in
Western music.
Key points are:

• On a real instrument, there will be unmusical beating whenever an integer multiple of

one harmonic is close but not equal to an integer multiple of another harmonic.

• Conversely, the most consonant notes will have some harmonics which exactly agree.

For example, let’s start with middle C. This note is called C4 (the C is the note and 4 is the
octave) and has a frequency of ν0 = 261 Hz. Which notes sound good along side C4? Well, the
261 Hz note has harmonics of ν0, 2ν0, 3ν0, 4ν0, etc.:

261 Hz, 522 Hz, 783 Hz, 1044 Hz, 1305 Hz, ··· (7)

Thus if we play any of those notes along with C4 it will sound harmonic. If the fundamental fre-
quency is ν0 then

2ν0 = 1 octave = 1st harmonic = C5 (8)

4 Section 3

Are there more notes which are harmonious? Yes. Consider the note with ν5 = 391Hz. This
note has 2ν5 = 783Hz. Thus the second harmonic of ν5 matches the 3rd harmonic of ν0. We call
3
the note ν5 = 2 ν0 the perfect fifth

3
ν5 = ν0 = perfect fifth = G4 (9)
2

This is the G above middle C. In the same way, consider ν4 = 348 Hz. The 3rd harmonic of ν4
agrees with the 4rd harmonic of ν0. We call this the perfect fourth

4
ν4 = ν0 = perfect fourth = F 4 (10)
3
And so on.
It is easy to see that any rational number ratio of frequencies will be consonant. Many of
these ratios have names

ν 3 4 5 6 5 8
ν0
1 2 2 3 4 5 3 5
perfect perfect Major Minor Major Minor
name fundamental octave
fifth fourth third third sixth sixth
example C4 C5 G4 F4 E4 E♭ A5 A♭5
Table 1. Notes names and ratios in the just intonation scale

There are an infinite number of rational numbers. So where do we stop? The answer is that
3 8
the lower the numbers in the ratio (that is, the 3 and 2 in 2 are lower than the 8 and 5 in 5 ),
11
the more consonant they will be. That’s because for something like 17 , one would need the 17th
harmonic of one note to match the 11th harmonic of another note. By such high harmonics, the
amplitudes are no longer large, and the spectrum is messy (as you can see in Figure 2). Also, it
11
is more likely for frequencies with a ratio of 17 to give harmonics which are close but not equal,
generating dissonant beating, before generating the harmonic consonance. Thus, for numbers
large than about 5 in the ratio, notes are no longer appreciated as harmonic.

3 Scales

If we have a given note, say C4, we can define all the other notes so that they will be harmonic
with C4.

3.1 Just intonation scale

The most harmonic notes will have the smallest integers in the ratio, as in Table 1. This is a
particular choice of tuning known as the just intonation scale. The just intonation scale is in
a sense the most harmonic choice for the frequencies of notes in a scale (it is default tuning for
some non-Western instruments, such as the Turkish Baglama). But it is just a choice.
Note that if we pick a set of notes that sound harmoic with C4, the same set of notes will
generically not sound harmonic with another note, like D4. Thus if we’re playing a song, the set
of notes we want to use is determined relative to some starting note. This starting note is called
the key. For example, if you are in the key of C, the notes C, G and F sound good. But if you
are in the key of D, the notes C, G and F will generally not sound as good.
Scales 5

On some instruments, such as a violin which has no frets, there are no predefined notes.
Thus on a violin, if you work in the key of C, you can play all the harmonics in exactly the
right place (if you have a skilled enough ear and hand). Thus you can play the just intonation
scale in any key. There are an infinite number of notes you can play on a violin. The same is
true on most instruments actually. For example, even though an oboe has fixed holes corre-
sponding to notes, oboe players can easily move the notes up or down by manipulating the reed.
Controlling the precise frequency of a note with your mouth is critical to playing any woodwind
instrument well.
On other instruments, like a piano or a stringed instrument with frets like a guitar, the notes
are essentially built in to the instrument. You can sometimes tweak the notes if you are skilled,
or tune the instrument to a different key, but there are a finite number of notes which can be
played in a giving tuning. Unfortunately, it’s impossible to have an instrument with a finite
number of notes be capable of playing the most harmonic notes in every key.
To see the problem, suppose you want your piano to be in the just intonation scale in the
key of C. That means that you want all the other notes to be defined so that they are related to
C by rational numbers with low integers. For example, the whole notes can be defined as

note C D E F G A B C
ν 9 5 4 3 5 15
ν0
1 8 4 3 2 3 8
2
decimal 1 1.125 1.25 1.333 1.5 1.666 1.875 2
Table 2. Notes names and ratios in just intonation

9
This defines the note D as having the frequency 8 times the frequency of the C. Now, where
3 9 27
is the 5th of D? This should be at 2 × 8 = 16 = 1.688 times ν0. This note is somewhere between
the A and the B, but it is not exactly a note in the key of C. It is not hard to see that to get
an instrument which could play any note in any key, you would need an enormous number of
available notes. Please make sure you understand this point, as it is key to understanding scales.
So what can we do? There are two options: the first is you can tune your instrument to the
key you want to play in. Stringed instruments can do this. But it is not so appealing of an
option if we want to play music in different keys without retuning every time. The other option
is to compromise. Most of our ears are not sensitive enough to distinguish close but slightly dif-
ferent notes. Thus we can choose scales which are not exactly correct in any key, but close to
correct in all keys.

3.2 Pythagorean scale

3
One way to approximate the scale is by choosing the notes to be related by powers of 2 and
3
octaves to be related by factors of 2, For example, 2 ν0 is the perfect fifth (G4 when C4=ν0).
3
2
Then 2 ν0 which is the fifth of the fifth, or the fifth of G4 which is D5. The next note has
3
3 3
2
ν0 or the fifth of D5 which is A6 and so on. We can bring any power of 2 back to the
9
interval between 0 and 1 by dividing by 2 to some power. For example, since D5 = 4 ν0 then
9
D4 = 8 ν0. We then get

note C D E F G A B C
ν 9 81 4 3 27 243
ν0
1 8 64 3 2 16 128
2
decimal 1 1.125 1.266 1.333 1.5 1.688 1.898 2
Table 3. Notes names and ratios in Pythagorean tuning

This is called the Pythagorean tuning. Note that the octave, perfect fifth (G) and perfect
81
fourth (F) agree with their values in Table 2. Some notes do not agree: for example, E is 64 ν0 =
5
1.266ν0 in this tuning. This ratio is close to 4 but not exactly. Thus if we play C and E it will
be close to a consonant sounding note, but not exactly.
6 Section 3

The advantage of the Pythagorean tuning is that the perfect fifth and perfect fourth of every
note is included in the scale.

 n
3
Figure 3. Pythagorean guitar: frets positions are related by powers of 2
2m

3.3 Equal-tempered scale

An interesting feature of the Pythagorean scale is that the 12th fifth is very close to 8 octaves:
 12
3
= 129.748 ≈ 128 = 27 (11)
2

Thus another compromise is to say that when we go 12 steps around the circle of fifths, we get
back to the note we started at. Then we can devise a scale which does not choose any key. We
simply relate the notes by powers of 21/12 . Each half-step gives another factor of 21/12 . The
result are the frequencies in Table 4.

note C C♯ D D♯ E F F♯ G G♯ A A♯ B C
1 2 3 4 5 6 7 8 9 10 11
ν
ν0
1 2 12 2 12 2 12 2 12 2 12 2 12 2 12 2 12 2 12 2 12 2 12 2
decimal 1 1.059 1.122 1.189 1.260 1.335 1.414 1.498 1.587 1.682 1.782 1.888 2
Table 4. Notes names and ratios in the equal-tempered scale.

This tuning is called the equal-tempered scale. The equal-tempered scale is the standard
tuning for all of Western music.
Because of Eq. (11) if you keep going up by perfect fifths, and normalizing by octaves, you
will get back to the note you started at. We can see this in the circle of fifths
Scales 7

Figure 4. Circle of fifths. Each note going clockwise is a perfect fifth above the previous note. Going
counterclockwise, each note is a perfect 4th above the previous note. The circle only closes in the equal-
tempered scale.

In this circle, each note is 1 fifth above the note clockwise. So G is a fifth above C, D a fifth
above G and so on. Going counterclockwise, the intervals are fourths: C is a fourth above G and
F is a fourth above C. Going up a fifth is the same as going down a fourth and adding an
octave,
If the notes are defined with the Pythagorean scale, the circle doesn’t close: going up by 12
3

steps, and normalizing back down to the original octave leaves you 2 12 2−8 = 1.014 times where
you started. Thus the circle doesn’t close by 1.4%. In the equal-tempered scale, it does exactly
2
close, however, none of the notes have frequency ratios of exactly 3 .

3.4 Summary
We discussed 3 scales. The just intonation scale chooses notes to be related by rational number
ratios with integers as small as possible in the numerator and denominator. The Pythagorean
scale has all notes related by 3n2m for some m and n. Both just intonation and the Pythagorean
scale require a key to start in. nThe third scale is the equal-tempered scale. Notes in the equal-
tempered scale are related by 2 1 2 for some n.
Here is a comparison between the relative frequencies of the 3 scales in the key of C:

note C D E F G A B C
9 5 4 3 5 15
just-intonation 1 8 4 3 2 3 8
2
9 81 4 3 27 243
Pythagorean 1 8 64 3 2 16 128
2
2 4 5 7 9 11
equal-tempered 1 2 12 2 12 2 12 2 12 2 12 2 12 2
Table 5. Comparison of scales: exact ratios.

In decimals

note C D E F G A B C
just-intonation 1 1.125 1.25 1.333 1.5 1.666 1.875 2
Pythagorean 1 1.125 1.266 1.333 1.5 1.688 1.898 2
equal-tempered 1 1.122 1.260 1.335 1.498 1.682 1.888 2
Table 6. Comparison of scales: decimal approximations.

Here is a graphical comparison of how far off the frequency is in the equal-tempered scale
from the frequency in the just intonation scale
8 Section 3

Figure 5. Difference between the equal-tempered frequencies νW T and the just intonation frequencies
for whole notes C , D, E....C labeled as 1 to 8.

One thing we can see is that the perfect 4th and perfect 5th are very close to their optimal
values, while the 6th and 7ths are not so close.