Discretetime Fourier Series and Fourier Transforms

We now start considering discretetime signals. A discretetime signal is a function (real or complex
valued) whose argument runs over the integers, rather than over the real line. We shall use square brackets,
as in x[n], for discretetime signals and round parentheses, as in x(t), for continuoustime signals. This is
the notation used in EECE 359 and EECE 369. Discretetime signals arise in two ways. Firstly, the signal
could really be representing a discrete sequence of values. For example, x[n] could be the n
th
digit in a
string of binary digits being transmitted along some data bus in a computer. Or it could be the maximum
temperature for day number n. Secondly, a discretetime signal could arise from sampling a continuoustime
signal at a discrete sequence of times.
Periodic Signals
Just as in the continuoustime case, discretetime signals may or may not be periodic. We start by
considering the periodic case. Imagine an application in which we have to measure some function x(t), that
is periodic of period 2, and compute its Fourier coecients from the measurements. We can think of x(t)
as the amplitude of some periodic signal at time t. Because we can only make nitely many measurements,
we cannot determine x(t) for all values of t. Suppose that we measure x(t) at N equally spaced values of t
covering the full period 0 t < 2. Say at t = 0,
2
N
, 2
2
N
, , (N 1)
2
N
. Because we do not know x(t)
for all t we cannot compute the complex
(1)
Fourier coecient
c
k
=
1
2
_
2
0
x(t)e
ik

t
dt (1)
exactly. But we can get a Riemann sum approximation to it using only ts for which x(t) is known. All
we need to do is divide the domain of integration up into N intervals each of length
2
N
. For t in the
t
y
y = x(t)e
ik

t
0 2
2
N
2
2
N
3
2
N
interval n
2
N
t < (n + 1)
2
N
, we approximate the integrand x(t)e
ik

t
by its value at t = n
2
N
, which is
x
_
n
2
N
_
e
ik

n
2
N
= x
_
n
2
N
_
e
2i
kn
N
. So we approximate the integral over n
2
N
t < (n + 1)
2
N
by the area of
a rectangle of height x
_
n
2
N
_
e
2i
kn
N
and width
2
N
. This gives
c
k
c
(N)
k
=
1
2
N1

n=0
x
_
n
2
N
_
e
2i
kn
N
2
N
=
1
N
N1

n=0
x
_
n
2
N
_
e
2i
kn
N
(1)
We are using complex Fourier series rather than sin/ cos Fourier series only because the computations are cleaner. It is
perfectly possible to use sin/ cos Fourier series instead. Alternatively, the sin/ cos Fourier series coecients can be easily
computed from the complex ones as we did in the notes Fourier Series.
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 1
To save writing for what follows set x[n] = x
_
n
2
N
_
and x[k] = c
(N)
k
. Then
x[k] =
1
N
N1

n=0
x[n]e
2i
kn
N
Note that x[n] and x[k] are both periodic of period N. That is
x[k + N] =
1
N
N1

n=0
x[n]e
2i
(k+N)n
N
=
1
N
N1

n=0
x[n]e
2i
kn
N
e
2i
Nn
N
=
1
N
N1

n=0
x[n]e
2i
kn
N
because e
2ni
= 1
= x[k]
x[n + N] = x
_
(n + N)
2
N
_
= x
_
n
2
N
+ 2
_
= x
_
n
2
N
_
= x[n]
The vector
_
x[k]
_
k=0,1,2,,N1
, dened by,
x[k] =
1
N
N1

n=0
x[n]e
2i
kn
N
(2)
is called the discrete Fourier series (or by some people the discrete Fourier transform) of the vector
_
x[j]
_
j=0,1,2,,N1
. One of the main facts about discrete Fourier series is that we can recover all of the
(N dierent) x[n]s exactly from x[0], x[1], , x[N 1] (or any other N consecutive x[k]s) using the inverse
formula
x[n] =
N1

k=0
x[k] e
2i
nk
N
(3)
Proof: We need to show that if x[k] is dened by (2), then (3) is true. To verify this we just substitute
the denition of x[k] into the right hand side of (3), taking care to rename the summation variable to ensure
that we dont use n to stand for two dierent quantities in the same formula.
N1

k=0
e
2i
nk
N
x[k] =
N1

k=0
e
2i
nk
N
1
N
N1

=0
e
2i
n

k
N
x[n

] =
1
N
N1

=0
N1

k=0
e
2i
k(nn

)
N
x[n

]
=
1
N
N1

=0
x[n

]
N1

k=0
e
2i
k(nn

)
N
=
1
N
N1

=0
x[n

]
N1

k=0
_
e
2i
nn

N
_
k
=
1
N
N1

=0
x[n

]
N1

k=0
r
k
with r = e
2i
nn

N
For one value of n

, namely n

= n, r = 1 and
N1

k=0
e
2i
k(nn

)
N
=
N1

k=0
1 = N
For all other values of n

, we can use the standard formula

)
N
to get
N1

k=0
e
2i
k(nn

)
N
=
N1

k=0
_
e
2i
(nn

)
N
_
k
=
1
_
e
2i
(nn

)
N
_
N
1 e
2i(nn

)/N
=
1 e
2i(nn

)
1 e
2i(nn

)/N
= 0
because n n

is an integer. Substituting the values we have just found for the k sums gives
N1

k=0
e
2i
nk
N
x[k] =
1
N
N1

=0
x[n

]
N1

k=0
e
2i
k(nn

)
N
=
1
N
N1

=0
x[n

]
_
N if n

= n
0 if n

= n
= x[n]
as desired.
Example 1 In this example, we nd the Fourier series for the discretetime periodic square wave shown
in the gure
1
0 2 2 11 11
n
This signal has period N = 11. In computing its Fourier coecients, we may sum n over any 11 consecutive
values. We choose
x[k] =
1
N
8

n=2
x[n]e
2i
kn
N
=
1
11
2

n=2
e
2i
kn
11
The sum
2

n=2
e
2i
kn
11
is a nite geometric series
a + ar + ar
2
+ + ar
p
=
_
a
1r
p+1
1r
if r = 1
a(p + 1) if r = 1
_
=
_
aar
p+1
1r
if r = 1
a(p + 1) if r = 1
with
the rst term being a = e
2i
kn
11

n=2
= e
4i
k
11
,
the rst omitted term being ar
p+1
= e
2i
kn
11

n=3
= e
6i
k
11
, and
the ratio between successive terms being r = e
2i
k
11
.
Hence
x[k] =
1
11
e
4i
k
11
e
6i
k
11
1 e
2i
k
11
=
1
11
e
4i
k
11
e
6i
k
11
e
i
k
11
_
e
i
k
11
e
i
k
11
_ =
1
11
e
5i
k
11
e
5i
k
11
2i sin
_

k
11
_ =
1
11
sin
_
5
k
11
_
sin
_

k
11
_
provided the ratio r = 1. That is, provided k = 0, 11, 22, . When k = 0, 11, 22, , we have that
a =
1
11
, r = 1 and ve terms, so that x[k] =
5
11
. These Fourier coecients are graphed in the gure
0 11 11 k
Both of the sums in (2) and (3) are nite. So there is no problem of truncation error or Gibbs
phenomenon when computing discrete Fourier series, at least if N is not humongous.
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 3
Discretetime Fourier series have properties very similar to the linearity, time shifting, etc. properties
of the Fourier transform. A table of some of the most important properties is provided at the end of these
notes. Here are derivations of a few of them.
Time Shifting: Let n
0
be any integer. If x[n] is a discretetime signal of period N, then so is y[n] =
x[n n
0
]. The k
th
Fourier coecient of y[n] is
y[k] =
1
N
N1

n=0
y[n]e
2i
kn
N
=
1
N
N1

n=0
x[n n
0
]e
2i
kn
N
Now substitute m = n n
0
in the sum:
y[k] =
1
N
N1n0

m=n0
x[m]e
2i
k(m+n
0
)
N
= e
2i
kn
0
N
_
1
N
N1n0

m=n0
x[m]e
2i
km
N
_
The summand is periodic of period N. That is, replacing m by m+N has no eect on the summand. So all
domains of summation consisting of a single full period give the same sum. Consequently we may replace
the sum

N1n0
m=n0
by the sum

N1
m=0
and the sum in parentheses is exactly x[k]. Thus y[k] = e
2i
kn
0
N
x[k].
Conjugation: Notice that if x[n] is a discretetime signal of period N, then so is y[n] = x[n]. The k
th
Fourier coecient of y[n] is
y[k] =
1
N
N1

n=0
y[n] e
2i
nk
N
=
1
N
N1

n=0
x[n] e
2i
nk
N
=
1
N
N1

n=0
x[n] e
2i
n(k)
N
= x[k]
This tells us that the k
th
Fourier coecient of the periodic discretetime signal x[n] is x[k]. In particular,
x[n] is real valued if and only if x[n] = x[n] for all n, which is true if and only the Fourier coecients of x[n]
and y[n] = x[n] are the same. That is,
x[n] is real for all n x[k] = x[k] for all k
Parsevals relation: We can derive a version of Parsevals relation for discretetime Fourier series just as
we did for the Fourier transform. Subbing (2) into
N1

k=0
| x[k]|
2
=
N1

k=0
_
x[k]
_
x[k] =
N1

k=0
_
1
N
N1

n=0
x[n]e
2i
kn
N
_
x[k]
=
1
N
N1

n=0
N1

k=0
x[n]e
2i
kn
N
x[k] =
1
N
N1

n=0
x[n]
_
N1

k=0
e
2i
kn
N
x[k]
_
=
1
N
N1

n=0
x[n] x[n] =
1
N
N1

n=0

x[n]

2
The Fast Fourier Transform
The Fast Fourier Transform does not refer to a new or dierent type of Fourier transform. It refers
to a very ecient algorithm (made popular by a publication of J. W. Cooley and J. W. Tukey in 1965, but
actually known to Gauss in about 1805) for computing the discretetime Fourier and inverse Fourier sums
(2) and (3). We will not be covering this algorithm in this course, though it is not particularly sophisticated.
The main idea behind it is explained in the supplementary notes The Fast Fourier Transform.
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 4
You have access to the fast Fourier transform through the MATLAB commands fft and ifft. But a little
care must be exercised when using fft and ifft because they implement dierent conventions than ours.
If the input vector x is of length N, then xhat = fft(x) is a vector xhat with the N elements
xhat(k) =
N

n=1
x(n)e
2i
(k1)(n1)
N
, 1 k N
and if the input vector xhat is of length N, then x = ifft(xhat) is a vector x with the N elements
x(n) =
1
N
N

k=1
xhat(k)e
2i
(k1)(n1)
N
, 1 n N.
In contrast to equations (2) and (3), the factor of
1
N
appears in ifft. The reason for the funny looking
exponents is that, in MATLAB, vector indices start with 1 rather than 0.
So given any complex numbers x[0], . . . , x[N 1], the vector x given by equation (2) above can be
computed, in MATLAB, as follows:
x =
_
x[0], x[1], x[2], . . . , x[N 1]

;
xhat = (1/N)*fft(x);
Notice the factor of 1/N in the second line. The resulting vector xhat will have N entries, corresponding to
_
x[0], x[1], . . . , x[N 1]

. But notice that MATLABs subscripting rules require

;
x = N*ifft(xhat);
Here, again, the factor of N in the second line corrects for a dierent system of conventions between these
notes and the MATLAB software system.
Aperiodic Signals
We now develop a frequency expansion for non-periodic discretetime functions using the same strategy
as we did in the continuoustime case.
Again, for simplicity well only develop the expansions for functions x[n] that are zero for all suciently
large |n|. Again, our conclusions will actually apply to a much broader class of functions. Let N be an even
integer that is suciently large that x[n] = 0 for all |n|
1
2
N. We can get a discretetime Fourier series
expansion for the part of x[n] with |n| <
1
2
N by using the periodic extension trick. Dene x
(N)
[n] to be the
unique discretetime function determined by the requirements that
i) x
(N)
[n] = x[n] for
N
2
< n
N
2
ii) x
(N)
[n] is periodic of period N
Then, for
N
2
< n
N
2
,
x[n] = x
(N)
[n] =

N
2
<k
N
2

x
(N)
[k] e
2i
nk
N
where

x
(N)
[k] =
1
N

N
2
<n
N
2
x[n] e
2i
nk
N
(4)
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 5
Here we have exploited the fact that since both x
(N)
[n] and

x
(N)
[k] are periodic of period N, we may choose
the range of summation in (2) and (3) to be any consecutive set of N integers. We have chosen
N
2
< k
N
2
and
N
2
< n
N
2
because they will lead to nice limits when we send N .
To get a representation of x[n] that is is valid for all ns , not just those in a nite interval
N
2
< n
N
2
,
we take the limit N . To evaluate this limit we again interpret the sum over k in (4) as a Riemann
sum approximation to a certain integral. For each integer k, dene the k
th
frequency to be
k
= 2
k
N
.
Also use =
2
N
to denote the spacing,
k+1

k
, between any two successive frequencies and dene
x() =

n=
x[n]e
in
. Since x[n] = 0 for all |n|
N
2
,

x
(N)
[k] =
1
N

N
2
<n
N
2
x[n]e
2i
nk
N
=
1
N

n=
x[n]e
i2
k
N
n
=
1
N

n=
x[n]e
i
k
n
=
1
2
x(
k
)
In this notation,
x[n] = x
(N)
[n] =

N
2
<k
N
2
1
2
x(
k
) e
2i
nk
N
=
1
2

k with
<
k

x(
k
)e
i
k
n

for any
N
2
< n
N
2
. Note that we have multiplied the summation restriction
N
2
< k
N
2
by
2
N
to get
the equivalent restriction <
k
. As we let N , the restriction
N
2
< n
N
2
disappears and
the right hand side, which is exactly a Riemann sum approximation sum to the integral
1
2
_

x()e
in
d,
converges to that integral. We conclude that
x[n] =
1
2
_

x()e
in
d where x() =

n=
x[n]e
in
(5)
The function x() is called the discretetime Fourier transform of x[n] or the spectrum of x[n]. For any
integer m,
x( + 2m) =

n=
x[n]e
i(+2m)n
=

n=
x[n]e
in
e
i2mn
=

n=
x[n]e
in
= x()
So x() is periodic of period 2 and we may choose as the domain of integration in (5) any interval of length
2.
Formulae (5) should look familiar. They are exactly the rst Fourier expansion that we saw, but with
the roles of the time and frequency domains exchanged. In Theorem 1 of the notes Fourier Series, we saw
that, if f(t) is continuous with continuous rst derivative and is also periodic with period 2, then
f(t) =

k=
c
k
e
ikt
where c
k
=
1
2
_

f(t)e
ikt
dt
If we make the substitutions t = and k = n we get
f() =

n=
c
n
e
in
where c
n
=
1
2
_

f()e
in
d
which is exactly (5) with x[n] = c
n
and x() = f().
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 6
Example 2 The discretetime signal
x[n] = a
n
u[n] where u[n] =
_
1 if n 0
0 if n < 0
is graphed in the gure
1
0 1 2 3 4 1 2 3 4
y = x[n]
n
It has the discrete Fourier transform
x() =

n=
x[n]e
in
=

n=0
a
n
e
in
=

n=0
_
ae
i
_
n
=
1
1 ae
i
Example 3 The discretetime signal shown in the gure
1
0 2 2
n
consists of a single pulse from the square wave signal of Example 1. The computation of its Fourier transform
is virtually identical to the computation of the Fourier coecients in Example 1.
x() =

n=
x[n]e
in
=
2

n=2
e
in
This is again a nite geometric series
a + ar + ar
2
+ + ar
p
= a
1r
p+1
1r
if r = 1
this time with rst term a = e
2i
, ratio r = e
i
and p = 4. Hence
x() = e
2i
1 e
5i
1 e
i
=
e
5
2
i
e
1
2
i
1 e
5i
1 e
i
=
e
5
2
i
e

5
2
i
e
1
2
i
e

1
2
i
=
sin
_
5
2

_
sin
_
1
2

_
when e
i
= 1, i.e. when = 2k, with k an integer. When = 2k, we have that a = 1 and r = 1 so that
x() = 5. This Fourier transform is graphed in the gure

As sample derivations of the properties of the transform (5), we now develop the two properties that
involve convolutions. First suppose that we take the product p[n] = x[n]y[n] of the two discretetime signals
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 7
x[n] and y[n]. Then, by (5),
p[] =

n=
x[n]y[n]e
in
=

n=
_
1
2
_

x()e
in
d
_
y[n]e
in
=
1
2
_

d x()

n=
y[n]e
i()n
=
1
2
_

d x() y( )
(6)
which is, aside from the prefactor of
1
2
, the convolution of x and y. Note that the domain of integration is
over one period of x and y.
Second, if we take the convolution
c[n] = (x y)[n] =

m=
x[n m]y[m]
then
c() =

n=
c[n]e
in
=

n=

m=
x[n m]y[m]e
in
=

m=

n=
x[n m]e
i(nm)
y[m]e
im
For each xed m

n=
x[n m]e
i(nm)
k=nm
=

k=
x[k]e
ik
= x()
so
c() =

m=
x()y[m]e
im
= x()

m=
y[m]e
im
= x() y() (7)
Example 4 Suppose that we take the convolution of the impulse signal
[n] =
_
1 if n = 0
0 if n = 0
with some other signal x[n]. Then
(x )[n] =

m=
x[n m][m]
But, by the denition of , every single term in the sum, except that with m = 0 is zero. So
(x )[n] = x[n 0][0] = x[n]
By (7), the discrete Fourier transform of x[n] = (x)[n] is x()

(). So

() ought to be one. By denition,

() =

n=
[n]e
in
Again, by the denition of , every term in the sum, except that with n = 0, is zero. So

() = [0]e
i0
= 1
as expected.
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 8
Example 5 This time, lets convolve an impulse

n0
[n] =
_
1 if n = n
0
0 if n = n
0
at some time n
0
, possibly nonzero, with some other signal x[n]. Then
(x
n0
)[n] =

m=
x[n m]
n0
[m] = x[n n
0
]
n0
[n
0
] = x[n n
0
]
The DFT of
n0
is

n0
() =

n=

n0
[n]e
in
=
n0
[n
0
]e
in0
= e
in0
So (7) tells us that the DFT of the time shifted signal x[n n
0
] is

n0
() x() = e
in0
x()
Example 6 As a less trivial example of a direct evaluation of a convolution consider the boxcar
[n] =
_
1 if |n| 2
0 if |n| > 2
of Example 3. The convolution
( )[n] =

m=
[n m][m]
The rst factor, [n m] is zero unless |n m| 2, i.e. unless m is within distance 2 of n, i.e. unless
n 2 m n +2. The second factor, [m] is zero unless 2 m 2. So the convolution ( )[n] is the
number of integers m that obey both n 2 m n + 2 and 2 m 2.
If n is so negative that n + 2 < 2 (i.e. n < 4) there are no ms that obey both n 2 m n + 2
and 2 m 2. This is illustrated in the gure below. In his case, the convolution is zero.
2 2
n + 2 n 2
m
If we increase n so that 2 n + 2 2 (i.e. 4 n 0) then the inequalities n 2 m n + 2 and
2 m 2 reduce to 2 m n + 2. This is illustrated in the gure below. In general, if k p
are two integers, the number of integers m that obey k m p is p k + 1. (In particular, if p = k,
there is one allowed m and if p = k + 1, there are two allowed ms.) Applying this with p = n + 2 and
k = 2, we have that, in his case, the convolution is (n + 2) (2) + 1 = n + 5.
2 2
n + 2 n 2
m
If we increase n again so that 2 n 2 2 (i.e. 0 n 4) then the inequalities n 2 m n + 2
and 2 m 2 reduce to n 2 m 2. This is illustrated in the gure below. In his case the
convolution is 2 (n 2) + 1 = 5 n.
2 2
n + 2 n 2
m
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 9
Finally, if n is so positive that n2 > 2 (i.e. n > 4) there are no ms that obey both n2 m n+2
and 2 m 2. This is illustrated in the gure below. In his case the convolution is zero.
2 2
n 2 n + 2
m
We conclude that
( )[n] =
_

_
0 if n < 4
n + 5 if 4 n 0
5 n if 0 n 4
0 if n > 4
4 0 4
5
1
n
We have now seen a number of dierent, but closely related, Fourier expansions. They are given in the
following table.
time domain frequency domain
continuous, period 2 discrete x(t) =

k=
x[k] e
ik

t
x[k] =
1
2
_

f(t) e
ik

t
dt
continuous continuous x(t) =
1
2
_

f()e
it
d x() =
_

f(t)e
it
dt
discrete, period N discrete, period N x[n] =
N1

k=0
x[k] e
2i
nk
N
x[k] =
1
N
N1

n=0
x[n]e
2i
kn
N
discrete continuous, period 2 x[n] =
1
2
_

x()e
in
d x() =

n=
x[n]e
in
Observe that
If in either domain, time or frequency, the function is periodic (not periodic) then the argument in the
other domain is discrete (runs continuously).
If in either domain, time or frequency, the function has a discrete argument (continuous argument),
then the transformed function in the other domain is periodic (not periodic).
Not surprisingly, all of four of these transforms have properties very similar to the linearity, time shifting,
etc. properties of the Fourier transform. The detailed versions of these properties are given in the following
tables.
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 10
Continuoustime, period 2
Property Periodic Signal Fourier Coecients
x(t) =

k
x[k]e
ik

t
x[k] =
1
2
_

x(t)e
ik

t
dt
y(t) =

k
y[k]e
ik

t
y[k] =
1
2
_

y(t)e
ik

t
dt
Linearity Ax(t) + By(t) A x[k] + B y[k]
Time Shifting x(t t
0
) e
ik

t0
x[k]
Frequency Shifting e
in

t
x(t) x[k n]
Conjugation x(t) x[k]
Time Reversal x(t) x[k]
Dierentiation x

(t) ik

x[k]
Convolution
_

x()y(t ) d 2 x[k] y[k]

m=
x[m] y[k m]
Parseval
1
2
_

|x(t)|
2
dt =

k=
| x[k]|
2
Continuoustime, aperiodic
Property Aperiodic Signal Fourier Transform
x(t) =
1
2
_

x()e
it
d x() =
_

x(t)e
it
dt
y(t) =
1
2
_

y()e
it
d y() =
_

y(t)e
it
dt
Linearity Ax(t) + By(t) A x() + B y()
Time Shifting x(t t
0
) e
it0
x()
Frequency Shifting e
i0t
x(t) x(
0
)
Scaling x
_
t

_
|| x()
Time Shift & Scaling x
_
tt0

) ||e
it0
x()
Frequency shift & scaling ||e
i0t
x(t) x
_
0

_
Conjugation x(t) x()
Time Reversal x(t) x()
tDierentiation x

(t) i x()
Dierentiation tx(t) i
d
d
x()
Convolution
_

x()y(t ) d x() y()

x() y( ) d
Duality x(t) 2x()
Parseval
_

|x(t)|
2
dt =
1
2
_

| x()|
2
d
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 11
Discretetime, period N
Property Periodic Signal Fourier Coecients
x[n] =
N1

k=0
x[k]e
2i
kn
N
x[k] =
1
N
N1

n=0
x[n]e
2i
kn
N
y[n] =
N1

k=0
y[k]e
2i
kn
N
y[k] =
1
N
N1

n=0
y[n]e
2i
kn
N
Linearity Ax[n] + By[n] A x[k] + B y[k]
Time Shifting x[n n
0
] e
2i
kn
0
N
x[k]
Frequency Shifting e
2i
nk
0
N
x[n] x[k k
0
]
Conjugation x[n] x[k]
Time Reversal x[n] x[k]
Dierence x[n] x[n 1]
_
1 e
2i
k
N
_
x[k]
Convolution
N1

m=0
x[m]y[n m] N x[k] y[k]
Multiplication x[n]y[n]
N1

m=0
x[m] y[k m]
Duality x[n]
1
N
x[k]
Parseval
1
N
N1

n=0
|x[n]|
2
=
N1

k=0
| x[k]|
2
Discretetime, aperiodic
Property Aperiodic Signal Fourier Transform
x[n] =
1
2
_

x()e
in
d x() =

n=
x[n]e
in
y[n] =
1
2
_

y()e
in
d y() =

n=
y[n]e
in
Linearity Ax[n] + By[n] A x() + B y()
Time Shifting x[n n
0
] e
in0
x()
Frequency Shifting e
i0n
x[n] x(
0
)
Conjugation x[n] x()
Time Reversal x[n] x()
nDierence x[n] x[n 1]
_
1 e
i
_
x()
Dierentiation nx[n] i
d
d
x()
Convolution

m=
x[m]y[n m] x() y()
Multiplication x[n]y[n]
1
2
_

x() y( ) d
Parseval

n=
|x[n]|
2
=
1
2
_

| x()|
2
d
March 21, 2007 Discretetime Fourier Series and Fourier Transforms 12