EECE 301

Signals & Systems

Prof. Mark Fowler
Note Set #9
C-T Signals: FS Spectrum

1/7

Trig Form Spectrum Is Single Sided

Best for thinking about real-world ideas
Trig Form: Amplitude & Phase

x(t ) A0 Ak cos(k0t k )
k 1

Need Ak and k
for k = 0, 1, 2
Ak = Amplitude
k = Phase

So to describe a signal via FS we specify:

Amplitude & Phase @ Each Frequency

A good way to see the FS coefficients is by plotting them vs. frequency:

Ak

Amplitude Spectrum

0 20 30 40 50

For this form of FS:

Do not need negative freqs
Single Sided Spectrum

Phase Spectrum

0 20 30 40 50

2/7

Exp Form Spectrum Is Double Sided

Best for doing math (ck are like phasors!!)
Need ck (complex!)
for k = -2, -1, 0, 1, 2

Exponential Form

x(t )

jk0t
c
e
k

|ck| = Magnitude
ck = Phase

ck e jk0t ck e jck e jk0t

So to describe a signal via FS we specify:

Magnitude & Phase @ Each Frequency

|ck|

ck e j k0t ck

Magnitude Spectrum

50 40 30
0
20

ck

0 20 30 40 50

For this form of FS:

Do need negative freqs
Double Sided Spectrum

Phase Spectrum

50 40 30
0
20

0 20 30 40 50

3/7

Spectrum Characteristics
Exponential Form

Trig Form: Amplitude & Phase

x(t )

x(t ) A0 Ak cos(k0t k )
For Trig Form of FS Spectrum:
Single Sided Spectrum
Ak 0 for k > 0
A0: positive or negative
k is in radians 0 = 0
Magnitude Spectrum

50 40 30
0
20

ck

jk0t
c
e
k

k 1

|ck|

0 20 30 40 50

Phase Spectrum

For Exp Form of FS Spectrum:

Double Sided Spectrum
|ck| 0 for all k
Even Symmetry for Magn.
ck is in radians
c0 = 0 or
ck = ck
Odd Symmetry for Phase

ck 12 Ak e jk
k 1, 2, 3,
j k
1
c k 2 Ak e

50 40 30
0
20

0 20 30 40 50

4/7

Parsevals Theorem
We saw earlier how to compute the average power of a periodic signal if we are
given its time-domain model:
T t

1
T

t0

x 2 (t )dt

Q: Can we compute the average power from the frequency domain model
A: Parsevals Theorem says Yes!

ck ,

k 0, 1, 2,...

Parsevals theorem says that the avg. power can be computed this way:

ck

ck are the Exp. Form FS coefficients

1
T

t 0 T

t0

x (t )dt
2

ck

Left side is clearly finite for

real-world signals
Thus, the |ck| must decay fast
enough as k
Tells us something about how the
magnitude spectrum should look!

5/7

Interpreting Parsevals Theorem

1 t T 2
x (t )dt ck

t
T
k
0

sum of squares in time-domain model

x 2 (t ) = power at time t (includes

effects of all frequencies)
We can find the power in the time
domain by adding up all the
powers at each time

sum of squares in freq.-domain model

ck power at frequency k0
2

(includes effects of all times)

We can find the power in the frequency

domain by adding up all the powers at
each frequency

6/7

One Use for Parsevals Theorem

When numerically computing the FS approximation PT allows you to compute
the power of the error term:
First find Avg Power:

1
P
T

T t0

t0

x 2 (t )dt

Do analytically or numerically

Then find power of approximate using PT: Papprox

Then find power of error as Perror = P Papprox
It is easy to show that Perror 2

k K 1

ck

k K

ck c0 2 ck
2

k 1

Since the |ck| decay as k

this shows that we can make
Perror as small as we want by
making K big enough!

|ck|2

50 40 30
0
20

0 20 30 40 50

7/7