Signals and the frequency domain

ENGR 40M lecture notes — July 31, 2017

Chuan-Zheng Lee, Stanford University

A signal is a function, in the mathematical sense, normally a function of time. We often refer to functions
as signals to convey that the function represents some sort of phenomenon—for example, an audio signal,
the electromagnetic signal broadcast in FM radio, a currency exchange rate, or the voltage somewhere in a
circuit.
One of the great advances of the 19th century was courtesy of a French mathematician called Joseph Fourier,
who showed in his work about heat flow that representing a signal as a sum of sinusoids opens the door to
more powerful analytical tools. This idea gave rise to what is now known as the frequency domain, where
we think of signals as a function of frequency, as opposed to a function of time. Our goal today is to define,
in some sense, what that means.

Preliminaries
Periodic signals. At least to begin, we’ll mainly be concerned with signals that are periodic. Informally, a
periodic signal is one that repeats, over and over, forever. To be more precise:

A signal x(t) is said to be periodic if there exists some number T such that, for all t, x(t) = x(t+T ).
The number T is known as the period of the signal.
The smallest T satisfying x(t) = x(t + T ) is known as the fundamental period.

Sinusoids. Precisely what do we mean by a sinusoid ? The term “sinusoid” means a sine wave, but we don’t
just mean the standard sin(t). To enable our analysis, we want to be able to work sine waves of different
heights, widths and phases. So, to us, a single sinusoid means a function of the form
x(t) = A sin(2πf t + φ),
for some A (its amplitude), f (its frequency) and φ (its phase).

x(t) 1/f

A
t

φ
2πf

The Fourier series

Joseph Fourier’s bold idea was to express periodic signals as a sum of sinusoids.
Theorem. If x(t) is a well-behaved periodic signal with period T , then it can be written in the form
∞
X
x(t) = A0 + An sin(2πnf t + φn ). (1)
n=1

where f = 1/T , and for some A0 , A1 , A2 , . . . (which we call magnitudes) and φ1 , φ2 , . . . (which we call
phases). We call this sum the Fourier series of x(t).
That “well-behaved” caveat is worth expanding on, briefly. In engineering, every practical signal we ever
deal with is “well-behaved” and has a Fourier series. If you talk to a mathematician, they’ll say that we’re
sweeping a lot of details under the carpet. They’re of course right, but in real-world applications this tends
not to bother us.
It’s worth reflecting on why Fourier’s claim was so significant. It’s not too hard to believe that some periodic
signals can be represented by a sum of sinusoids. But included in the “well-behaved” category are signals
with sudden jumps (discontinuities) in them. Can a discontinuous signal really be expressed as a sum of
smooth sinusoids? Pretty much, yes. The catch is that you might need an infinite number of sinusoids.

Time-domain and frequency-domain representations

Another fact relating to Fourier series is that the magnitudes A0 , A1 , A2 , . . . and phases φ1 , φ2 , . . . in (1)
uniquely determine x(t). That is, if we can find the A0 , A1 , A2 , . . . and φ1 , φ2 , . . . corresponding to a periodic
signal x(t), then, in effect, we have another way of describing x(t).
When we represent a periodic signal using the magnitudes and phases in its Fourier series, we call that the
frequency-domain representation of the signal. We often plot the magnitudes in the Fourier series using a
stem graph, labeling the frequency axis by frequency.1 In this sense, this representation is a function of
frequency.
To emphasize the equivalence between the two, we call plain old x(t) the time-domain representation, since
it’s a function of time. For example, here are both representations of a square wave:

Time-domain representation Frequency-domain representation

0.8
x(t) A
1
0.6

0.4
0.5
0.2
t f
0.5 1 1.5 2 5 10 15

Here’s some more terminology:

ˆ In electrical engineering, we call the term A0 the DC component, DC offset or simply offset. “DC”
stands for “direct current”, in contrast to the sinusoids, which “alternate”, though we use this term
even if the signal isn’t a current.
In a sense, the DC component is like the “zero frequency component”, since cos(2π · 0 · t) = 1. We often
think of offset in this way, and plot the DC offset at f = 0 in the frequency-domain representation.
The DC component is often easy to eyeball—it’s equal to the average value of the signal over a period.
For example, in the signal above, the DC offset is 0.5.
ˆ The sinusoidal terms are often called harmonics, a term borrowed from music. The harmonics will
have frequencies f , 2f , 3f , 4f and so on.
ˆ We also call each harmonic, An sin(2πnf t + φn ), the frequency component of x(t) at frequency nf .
For example, if f = 10 Hz, we call the harmonic for which n = 3 the “30 Hz component”, reflecting
that in this case sin(2πnf t + φn ) is a sinusoid of frequency 30 Hz.

1 It’s also common to plot the phases on another graph, but we won’t in this course.

2
Getting to know the frequency domain
What happens to our humble square wave if we don’t pick up on all of the infinitely many frequencies? Say,
if everything above 9 Hz disappeared?

Time-domain representation Frequency-domain representation

x(t) 0.8
A
1 0.6

0.4
0.5
0.2
t f
0.5 1 1.5 2 5 10 15

We still have an approximation to the square wave: it’s got the general shape right, it’s just not working so
well at the corners. In fact, as you might have expected, that discontinuity gives the Fourier series—a sum of
continuous functions—a bit of a hard time. Nonetheless, the more frequencies we include, the closer to the
true square wave we get. Here’s an aphorism that encapsulates this observation: It takes high frequencies to
make jump discontinuities.
More generally, the following statements provide some intuition for how to think in the frequency domain.

ˆ High frequencies come in where the signal changes rapidly.

At the extreme, when a signal changes suddenly, infinitely high frequencies come into play.
ˆ Low frequencies come in where the signal changes slowly.
At the extreme, when it doesn’t change at all, we have the DC component (zero frequency).

Now, you might be thinking: This is all very well for periodic signals, but in real life, no signal is truly
periodic, not least because no signal lasts forever. Then what do we do? Fear not—a lot of the same
instincts from periodic signals also apply to signals that are roughly repetitive in the short term, including
audio signals and your heartbeat.2

Circuits with capacitors and inductors

The following is a perhaps surprising fact about circuits involving capacitors and inductors.

Any circuit with only sources, resistors, capacitors and inductors acts on frequencies individually.
That is, the circuit takes each frequency component, multiplies each component by a gain specific
to that frequency, and then outputs the sum of the new frequency components.

The frequency domain thus gives us a richer way to understand how these circuits work. We’ll study this in
the coming lectures.

2 In
fact, there is a close cousin of the Fourier series, known as the Fourier transform, that deals with non-periodic signals
much more rigorously.

3
Computing Fourier coefficients (not examinable)
You might be wondering: It’s all very nice that these magnitudes and phases exist, but can we compute
them? We can, but it turns out that the form in (1) is rather awkward to do these calculations with. Another
way to write a Fourier series is
∞
a0 X
x(t) = + [an cos(2πnf t) + bn sin(2πnf t)] . (2)
2 n=1

Showing that (1) and (2) are equivalent is left as an exercise. The coefficients can then be found using the
following integrals, where T = 1/f :
Z T Z T
2 2
an = x(t) cos(2πnf t) dt, bn = x(t) sin(2πnf t) dt.
T 0 T 0

Then a0 , a1 , a2 , . . . and b1 , b2 , . . . can be converted to magnitudes and phases to fit the form of (1).
We will not ask you to apply these formulae in this course.

4
Exercises
Exercise 1. Which of these could be the time-domain representation of the signal whose frequency-domain
representation is below?

Frequency-domain representation

A
1

0.5

f
2 4 6 8

Candidate time-domain representation 1 Candidate time-domain representation 2

2 2
x(t) x(t)

1 1

t t
−0.5 0.5 1 1.5 2 −0.5 0.5 1 1.5 2

−1 −1

−2 −2

Candidate time-domain representation 3 Candidate time-domain representation 4

2 2
x(t) x(t)

1 1

t t
−1 1 2 3 4 −0.5 0.5 1 1.5 2

−1 −1

−2 −2

Candidate time-domain representation 5 Candidate time-domain representation 6

2 2
x(t) x(t)

1 1

t t
−0.5 0.5 1 1.5 2 −0.5 0.5 1 1.5 2

−1 −1

−2 −2

5
Exercise 2. Which do you think is the correct frequency-domain representation of the signal whose time-
domain representation is below?

Time-domain representation

x(t)
0.5

t
2 4 6

−0.5

Candidate frequency-domain representation 1 Candidate frequency-domain representation 2

0.6 0.6
A A

0.4 0.4

0.2 0.2

f f
5 10 15 2 4 6 8

Candidate frequency-domain representation 3 Candidate frequency-domain representation 4

0.6 0.6
A A

0.4 0.4

0.2 0.2

f f
5 10 15 2 4 6 8