University of California, Berkeley

Department of EECS
EE120: S IGNALS AND S YSTEMS (Spring 2021)
March 9, 2021

Notes on Discrete-Time Frequency and Fourier Series

1 Frequencies of Discrete-Time Signals

DTFT, just like CTFT, is continuous in frequency domain. However, unlike signals in continuous time,
frequency for discrete time signal is 2π-periodic. This means that, given any frequency ω0 , the frequency
ω1 = ω0 + 2π is the same frequency as ω0 . (Be aware that this 2π periodicity is a different issue from 2π
wrapping of the phase.)

We can show that this is the case by considering a complex exponential of frequency ω1 = ω0 + 2π:

ejω1 n = ej(ω0 +2π)n = ejω0 n ej2πn .

Since n is an integer, ej2πn = 1, ∀n, so

ejω1 n = eiω0 n .

One direct consequence of this is how we define high and low frequencies in discrete time. For discrete-
time signals, low frequencies are defined as being close to even multiples of π (0, 2π, etc), and high
frequencies are defined as being close to odd multiples of π (π, 3π, etc).
To see why, plot out x1 [n] = cos[πn] and x2 [n] = cos[2πn]:

1 x1 [n] 1 x2 [n]

n
1 2 3 4
n
−1 1 2 3 4

Notice that x1 [n] oscillates as fast as is possible in discrete time, while x2 [n] is a constant 1.

1
2 Periodicity of DTFS, DFT, and DTFT
Discrete-time frequency is 2π-periodic, which means that the DTFT is also 2π-periodic, while the DTFS
and DFT coefficients are N -periodic.

• DTFT: Examine X(ejω ), which is the DTFT of some signal x[n]. For every frequency ω, ω + 2π =
ω, so X(ej(ω+2π) ) = X(ejω ), ∀ω. X(ejω ) is a continuous function in ω.

• DTFS: Recall the DTFS synthesis equation (complex exponential representation) for N -periodic
signal x[n]:
N −1
2π
X
x[n] = ak ej N kn .
k=0

Each coefficient ak corresponds to frequency ω = 2π

N
k. So, ak+N corresponds to frequency ω =
2π 2π 2π
N
(k+N ) = N k+2π = N k. ak and ak+N correspond to the same frequency, so they are equivalent.
ak+N = ak .

• DFT:As we will show in the next section, DFT coefficient Xk is equivalent to N times the cor-
responding DTFS coefficient ak (Xk = N ak ). So, the periodicity of the DTFS coefficients also
applies to the DFT coefficients. For ease of using the DFT as a matrix-vector operation, we usually
consider the DFT coefficients in the range {0, . . . , N − 1}.
This is the reason why we only sum over a region of length N in the DTFS or DFT synthesis equations
and integrate over a region of length 2π in the DTFT synthesis equation.

Note that this periodicity does not apply to the CTFS or CTFT.

3 Relationship between DTFS and DFT

Equation DTFS DFT
j 2π 2π
P −1 PN −1
Synthesis x[n] = Nk=0 ak e
N
kn
x[n] = N1 k=0 ak ej N kn
−j 2π −j 2π
P −1 P −1
Analysis ak = N1 Nn=0 x[n]e
N
kn
ak = N n=0 x[n]e
N
kn

The main difference between the DTFS and DFT is that the scaling factor of 1/N is in the analysis equa-
tion for DTFS, and in the synthesis equation for DFT. As a result, we have the relationship Xk = N ak ,
where Xk is a DFT coefficient and ak is the corresponding DTFS coefficient.

The DFT is typically used to represent finite-length discrete signals in the frequency domain. Either the
DTFS or the DFT can be used for periodic discrete-time signals. (In case you have heard of fast Fourier
transform (FFT), FFT is a computationally efficient implementation of DFT.)

2
4 Relationship between DTFS and DTFT
Given that the DTFS expansion of x[n] is
N
X −1
x[n] = ak eiω0 kn ,
k=0

the DTFT of x[n] is as follows:

∞
X
jω
X(e ) = 2π ak δ(ω − kω0 ).
k=−∞

Derivation
Recall that the DTFT of a complex exponential is a Dirac comb:

F{eiω0 n } = 2πX2π (ω − ω0 )

For DTFT, since it’s 2π periodic, we only care one 2π frequency span. Therefore, we can write the DTFT
of complex exponential as a shifted delta:

F{eiω0 n } = δ(ω − ω0 )

In the above expression, it is implicitly known that the delta repeats 2π-periodically.
Using this information, you can take the DTFT of the complex exponential expression for x[n]:
N
X −1
jω
X(e ) = F{ ak eiω0 kn }
k=0

(Apply linearity of DTFT)

N
X −1
= ak F{eiω0 kn }
k=0
N
X −1
= 2π ak δ(ω − kω0 ).
k=0

5 Plotting the DTFT

In discrete time, we typically plot the magnitude and phase of the the DTFT in a region of length 2π, since
we know it will repeat outside of that region (though we can also plot the spectrum over a longer range if

3
we want). If the DTFT is purely real or imaginary, we can also plot it directly without creating separate
magnitude and phase plots. Some examples are shown below:

Plotting Magnitude and Phase:

jω
2 |H(e )| ∠H(ejω )
π
1
2
π
ω
ω −2π −π − 1 π 0 π 2π
−2π −π 0 π 2π 2
−π

Plotting Magnitude and Phase with Dirac Deltas:

jω
1 |H(e )| π ∠H(ejω )

0.5 ω
−4π/5 −2π/5 2π/5 4π/5
ω
−4π/5 −2π/5 2π/5 4π/5 −π

Plotting a Purely Imaginary DTFT Representation:

jω
j H(e )

ω
−2π −π 0 π 2π

−j

4
We typically do not plot DTFS or DFT coefficients and instead plot the corresponding DTFT representa-
tion. However, if you are asked to plot the DTFS of a signal, you can make two stem plots, one for the
magnitude of ak and one for the phase. The x-axis will include discrete values of k over a region of length
N , as shown below:

1 |ak | π ∠ak

0.5 k
1 2 3 4
k
1 2 3 4 −π