Page MenuHomec4science

20-DTFT.tex
No OneTemporary

File Metadata

Created
Thu, Mar 13, 14:18

20-DTFT.tex

\section{The Discrete-\!Time Fourier Transform}
\label{sec:fa:dtft}
The DFT provides us with a frequency-domain representation of signals that contain a finite amount of data; periodic signals, which are also uniquely described by a finite number of data samples, use a formally identical frequency representation (which, solely for clarity, we label DFS). In both cases, the transform is a simple change of basis in $\mathbb{C}^N$ and, as such, it can be described algorithmically as a matrix-vector multiplication involving a finite number of operations.
We now consider the problem of obtaining a frequency-domain representation for aperiodic, infinite-length sequences. Such signals are an idealized mathematical concept, of course, but they are indispensable in order to obtain fundamental results in signal processing that hold for \textit{any} finite-length, finite-support or periodic signal. We can always think of an infinite-length sequence as the limit of a finite-length signal whose length grows to infinity; or, alternatively, as the limit of a periodic signal whose period grows unbounded. It is therefore reasonable to expect that the frequency representation of infinite-length signals should be consistent with the limit of a DFT or DFS, and that it should retain its intuitive interpretation as a change of basis in an appropriate vector space. Indeed, we will see that this is the case although, when we move from finite to infinite-length signals, something ``breaks down'' in the formalism and we won't be able to obtain a Fourier transform operator that acts as an isomorphism on the underlying signal space. In other words, given a signal in $\ell_2(\mathbb{Z})$, its Fourier transform will not be itself an element of $\ell_2(\mathbb{Z})$. Although the exact reasons for this are quite technical and beyond the scope of this book, we will try to provide some intuition in Section~\ref{sec:fa:dtftlimit}.
The frequency-domain representation for infinite-length, discrete-time signals is called the Discrete-Time Fourier Transform (DTFT). We will first introduce the transform as a formal operator and then show its ``operational'' equivalence to a change of basis in a suitable vector space.
\subsection{Formal Definition}
The Discrete-Time Fourier Transform of a sequence $\mathbf{x}$ is defined as\index{DTFT|mie}
\begin{equation}\label{eq:fa:DTFTsyn}
X(e^{j\omega}) = \sum_{n=-\infty}^{\infty} x[n] \, e^{-j\omega n}
\end{equation}
Formally, the DTFT is an operator that maps discrete-time sequences to a complex-valued functions of the frequency variable $\omega \in \mathbb{R}$. Since the argument $\omega$ only appears as the phase of a complex exponential in the integral, the resulting function is always $2\pi$-periodic. As we already mentioned in Section~\ref{sec:intro:notation}, the somewhat odd notation $X(e^{j\omega})$ is rather standard in the signal processing literature and offers several advantages:
\begin{itemize}
\item it stresses the periodic nature of the transform since, independently of the actual expression for $X$, $X(e^{j(\omega + 2\pi)}) = X(e^{j\omega})$;
\item regardless of context, it immediately identifies the expression as a DTFT;
\item it provides a convenient notational framework which unifies the Fourier transform and the $z$-transform, as we will see in Chapter~\ref{ch:zt}.
\end{itemize}
The DTFT, when it exists, can be inverted via the integration
\begin{equation}\label{eq:fa:DTFTrec}
x[n] = \frac{1}{2\pi} \int_{-\pi}^{\pi} X (e^{j\omega}) \,e^{j\omega n} d\omega;
\end{equation}
this can be easily verified by substituting~(\ref{eq:fa:DTFTsyn}) into~\ref{eq:fa:DTFTrec}) and recalling that
\[
\int_{-\pi}^{\pi} e^{-j\omega (n-k)} = 2\pi\, \delta[n-k].
\]
In fact, due to the $2\pi$-periodicity of the DTFT, the integral in~(\ref{eq:fa:DTFTrec}) can be computed over \emph{any} $2\pi$-wide interval on the real line; by convention, though, the DTFT is generally represented over the $[-\pi, \pi]$ interval. For the DTFT to exist, the sum in~(\ref{eq:fa:DTFTsyn}) must converge: if we define the partial sum
\begin{equation}\label{eq:fa:DTFTpartial}
X_M(e^{j\omega}) = \sum_{n=-M}^{M} x[n]\,e^{-j\omega n}
\end{equation}
existence of the DTFT is equivalent to the convergence of $\lim_{M \rightarrow\infty} X_M(e^{j\omega})$. Convergence is very easy to prove for \emph{absolutely summable} sequences, that is for sequences satisfying
\begin{equation}
\lim_{M \rightarrow \infty} = \sum_{n=-M}^{M} \bigl|x[n] \bigr| < \infty
\end{equation}
since, according to the triangle inequality,
\begin{equation}
\bigl| X_M(e^{j\omega}) \bigr| \leq
\sum_{n=-M}^{M} \bigl| x[n] \, e^{-j\omega n} \bigr| =
\sum_{n=-M}^{M} \bigl|x[n] \bigr|
\end{equation}
For this class of sequences it can also be proved that the convergence of $X_M(e^{j\omega})$ to $X(e^{j\omega})$ is uniform and that $X(e^{j\omega})$ is continuous. While absolute summability is a sufficient condition for the existence of the DTFT, it can be shown that the sum in~(\ref{eq:fa:DTFTpartial}) is convergent also for all \emph{square-summable} sequences, i.e. for sequences whose energy is finite, that is, for all sequences in $\ell_2(\mathbb{Z})$. In the case of square summability only, however, the convergence of~(\ref{eq:fa:DTFTpartial}) is no longer uniform but takes place only in the mean-square sense, i.e.{}
\begin{equation}
\lim_{M\rightarrow\infty} \int_{-\pi}^{\pi} \bigl|X_M(e^{j\omega}) - X(e^{j\omega}) \bigr|^2 \, d\omega = 0
\end{equation}
This type of convergence implies that, while the total energy of the difference between functions goes to zero, the functions may differ in value over a countable set of points.\footnote{
A particular manifestation of this behavior is called the \emph{Gibbs phenomenon}, which has important consequences in the problem of filter design, as we will study in Chapter~\ref{ch:fd}.}
Furthermore, in the case of square-summable sequences, $X(e^{j\omega})$ is no longer guaranteed to be
continuous.
As an example, consider the sequence:
\begin{equation}\label{DTFTexeq}
x[n] = \left\{ \! \begin{array}{ll}
1 & \mbox{ for } -N \leq n \leq N \\
0 & \mbox{ otherwise}
\end{array}
\right.
\end{equation}
Its DTFT can be computed as the sum\footnote{Remember that
$ \displaystyle %\[
\sum_{n=0}^{N}x^n = \frac{1 - x^{N+1}}{1-x} \, $.}
%\]}
\begin{align*}
X(e^{j\omega}) &= \sum_{n=-N}^{N}e^{-j\omega n}
\\ &
= \sum_{n=1}^{N} e^{j\omega n} + \sum_{n=0}^{N} e^{-j\omega n}
\\
&= \frac{1-e^{-j\omega (N+1)}}{1-e^{-j\omega}} + \frac{1-e^{j\omega (N+1)}}{1-e^{j\omega}} - 1
\\
&= e^{j\omega/2}\,\frac{1-e^{-j\omega (N+1)}}{e^{j\omega/2} - e^{-j\omega/2}} + e^{-j\omega/2}\,\frac{1-e^{j\omega (N+1)}}{e^{-j\omega/2} - e^{j\omega/2}} - 1
\\
&= \frac{e^{j\omega (N+\frac{1}{2})} - e^{-j\omega
(N+\frac{1}{2})}}{e^{j\omega/2} - e^{-j\omega/2}}
\\ &
=
\frac{\sin \left(\omega \left(N+\frac{1}{2} \right)\right) }{\sin( \omega/2)}
\end{align*}
\end{document}
%%% 4.10 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\def\specShapeEx{x RadtoDeg %
dup 10.5 mul sin exch %
0.5 mul sin div}
\begin{figure}[h]
\center\small
\begin{psDFplot}[dy=5,pifrac=2]{-7}{23}
\psDFplotFunction[linewidth=1.2pt]{\specShapeEx}
\end{psDFplot}
\vskip-5mm
\caption{The DTFT of the signal in~(\ref{DTFTexeq}).}\label{DTFTexampleFigA}
\end{figure}
The DTFT of this particular signal turns out to be real (we
will see later that this is a consequence of the signal's
symmetry) and it is plotted in Figure~\ref{DTFTexampleFigA}.
When, as is very often the case, the DTFT is complex-valued,
the usual way to represent it graphically takes the
magnitude and the phase separately into account. The DTFT is
always a $2\pi$-periodic\index{spectrum!periodicity|mie}
function and the standard convention is to plot the interval
from $-\pi$ to $\pi$. Larger intervals can be considered if
the periodicity needs to be made explicit;
Figure~\ref{DTFTexampleFigB}, for instance, shows five full
periods of the same function.
%% 4.11
\begin{figure}[h]
\center\small
\begin{psDFplot}[dy=5,reps=5]{-7}{23}
\psDFplotFunction[linewidth=1.2pt,linecolor=gray]{\specShapeEx}
\psDFplotFunction[linewidth=1.2pt,fmin=-3.14,fmax=3.14]{\specShapeEx}
\end{psDFplot}
\vskip-5mm
\caption{The DTFT of the signal in~(\ref{DTFTexeq}), with
explicit periodicity.}\label{DTFTexampleFigB}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{The DTFT as the Limit of a DFS}
A way to gain some intuition about the structure of the DTFT
formulas is to consider the DFS of periodic sequences with
larger and larger periods.
%increasingly longs periods.
Intuitively, as we look at the
structure of the Fourier basis for the DFS, we can see that
the number of basis vectors in~(\ref{DFTsyn}) grows with the
length $N$ of the period and,
\pagebreak%
consequently, the frequencies
of the underlying
complex exponentials become ``denser''
between $0$ and $2\pi$. We want to show that, in the limit,
we end up with the reconstruction formula of the DTFT.
To do so, let us restrict ourselves to the domain of
absolute summable sequences; for these sequences, we know
that the sum in~(\ref{DTFTsyn}) exists. Now, given an
absolutely summable sequence $x[n]$, we can always build an
$N$-periodic sequence $\tilde{x}[n]$ as
\index{periodization}
\begin{equation}\label{periodicx}
\tilde{x}[n] = \sum_{i = -\infty}^{\infty} x[n + iN]
\end{equation}
for any value of $N$ (see
Example~\ref{periodizationExample}); this is guaranteed by
the fact that the above sum converges for all $n \in
\mathbb{Z}$ (because of the absolute summability of $x[n]$)
so that all values of $\tilde{x}[n]$ are finite. Clearly,
there is overlap between successive copies of $x[n]$; the
intuition, however, is the following: since in the end we
will consider very large values for $N$ and since $x[n]$,
because of absolute summability, decays rather fast with
$n$, the resulting overlap of ``tails''{} will be negligible.
%In other words, we have:
This can be expressed as
\[
\lim_{N \rightarrow \infty} \tilde{x}[n] = x[n]
\]
Now consider the DFS of $\tilde{x}[n]$:
\begin{equation}
\tilde{X}[k] = \sum_{n=0}^{N-1} \tilde{x}[n] \,
e^{-j\frac{2\pi}{N}nk} = \sum_{i = -\infty}^{\infty}
\left( \sum_{n=0}^{N-1} x[n +iN] \, e^{-
j\frac{2\pi}{N}(n+iN)k} \right)
\end{equation}
where in the last term we have used~(\ref{periodicx}),
interchanged the order of the summation and exploited the
fact that $e^{-j(2\pi /N)(n+iN)k} = e^{-j(2\pi /N)nk}$. We
can see that, for every value of $i$ in the outer sum, the
argument of the inner sum varies between $iN$ and $iN+N -
1$, i.e.\ non-overlapping intervals, so that the double
summation can be simplified as
\begin{equation}
\tilde{X}[k] = \sum_{m = -\infty}^{\infty} x[m] \, e^{-j\frac{2\pi}{N}mk}
\end{equation}
and therefore
\begin{equation}\label{DFSCoefAsDTFT}
\tilde{X}[k] = X(e^{j\omega}) \bigl|_{\omega = \frac{2\pi}{N}k}
\end{equation}
This already gives us a noteworthy piece of intuition: the
DFS coefficients for the periodized signal are a discrete
set of values of its DTFT (here considered solely as a
formal operator) computed at multiples of $2\pi/N$. As $N$
grows, the spacing between these frequency intervals narrows
more and more so that, in the limit, the DFS converges to
the DTFT.
To check that this assertion is consistent, we can now write
the DFS reconstruction formula using the DFS values given to
us by inserting (\ref{DFSCoefAsDTFT}) in~(\ref{DFTrec}):
\begin{equation}
\tilde{x}[n] =
\frac{1}{N} \sum_{k = 0}^{N-1} X(e^{j\frac{2\pi}{N}k})
\, e^{j\frac{2\pi}{N}nk}
\end{equation}
By defining $\Delta = (2\pi/N)$, we can rewrite the above expression as
\begin{equation}
\tilde{x}[n] = \frac{1}{2\pi} \sum_{k = 0}^{N-1}
X(e^{j(k\Delta)}) \, e^{j(k\Delta) n}\, \Delta
\end{equation}
and the summation is easily recognized as the Riemann sum with step~$\Delta$ approximating the integral of $f(\omega) = X(e^{j\omega})e^{j\omega n}$ between $0$ and $2\pi$. As $N$ goes to infinity (and therefore $\tilde{x}[n] \rightarrow x[n]$), we can therefore write
\begin{equation}\label{finalDFSDTFT}
\tilde{x}[n] \rightarrow \frac{1}{2\pi} \int_{0}^{2\pi} X(e^{j\omega})\,
e^{j\omega n}\, d\omega
\end{equation}
which is indeed the DTFT reconstruction
formula~(\ref{DTFTrec}).\footnote{Clearly~(\ref{finalDFSDTFT})
is equivalent to~(\ref{DTFTrec}) in spite of the different
integration limits since all the quantities under the
integral sign are $2\pi$-periodic and we are integrating
over a period.}
\subsection{The DTFT as a Formal Change of Basis}\label{FourierDirac}
We %will
now show that, if we are willing to sacrifice
mathematical rigor, the DTFT can be cast in the same
conceptual framework we used for the DFT and DFS, namely as
a basis change in a vector space. The following formulas are
to be taken as nothing more than a set of purely symbolic
derivations, since the mathematical hypotheses under which
the results are well defined are far from obvious and are
completely hidden by the formalism. It is only fair to say,
however, that the following expressions represent a very
handy and intuitive toolbox to grasp the essence of the
duality between the discrete-time and the frequency domains
and that they can be put to use very effectively to derive
quick results when manipulating sequences.
One way of interpreting Equation~(\ref{DTFTsyn}) is to see that,
for any given value $\omega_{0}$, the corresponding value of
the DTFT is the inner product in $\ell_2(\mathbb{Z})$ of the
sequence $x[n]$ with the sequence $e^{j\omega_{0}n}$;
formally, at
least, we are still performing a projection in a
vector space akin to $\mathbb{C}^{\infty}$:
\[
X(e^{j\omega}) = \bigl\langle e^{j\omega n}, x[n] \bigr\rangle
\]
Here, however, the set of ``basis vectors''{}
$\{e^{j\omega n}\}_{\omega \in \mathbb{R}}$ is indexed by
the real variable $\omega$ and is therefore uncountable.
This uncountability is mirrored in the inversion
formula~(\ref{DTFTrec}), in which the usual summation is
replaced by an integral; in fact, the DTFT operator maps
$\ell_2(\mathbb{Z})$ onto $L_2 \bigl([-\pi, \pi] \bigr)$ which is a
space of $2\pi$-periodic, square integrable functions. This
interpretation preserves the physical meaning given to the
inner products in~(\ref{DTFTsyn}) as a way to measure the
frequency content of the signal at a given frequency; in
this case the number of oscillators is infinite and their
frequency separation becomes infinitesimally small.
To complete the picture of the DTFT as a change of basis, we
want to show that, at least formally, the set $\{e^{j\omega
n}\}_{\omega \in \mathbb{R}}$ constitutes an orthogonal
``basis''{} for $\ell_2(\mathbb{Z})$.\footnote{You can see here
already why this line of thought is shaky
unsafe: indeed,
$e^{j\omega n} \not\in \ell_2(\mathbb{Z})$!}
In order to do
so, we need to introduce a quirky mathematical entity called
the Dirac delta functional; this is defined in an implicit
way by the following formula\index{Dirac delta|mie}
\begin{equation}\label{deltadef}
\int_{-\infty}^{\infty} \delta(t-\tau) f(t)\, dt = f(\tau)
\end{equation}
where $f(t)$ is an arbitrary integrable function on the real line;
in particular
\begin{equation}
\int_{-\infty}^{\infty} \delta(t) f(t) \,dt = f(0)
\end{equation}
While no ordinary function satisfies the above equation,
$\delta(t)$ can be interpreted as shorthand for a limiting
operation. Consider, for instance, the family of parametric
functions\footnote{The rect function is discussed more
exhaustively in Section~\ref{idealFilters} its definition
is
\[ \mbox{rect}(x) = \left\{\!
\begin{array}{ll}
1 & \mbox{ for } |x| \leq 1/2 \\
0 & \mbox{ otherwise}
\end{array}
\right.
\]
}
\begin{equation}\label{limtodelta}
r_{k}(t) = k\,\mbox{rect}(kt)
\end{equation}
which are plotted in Figure~\ref{fig-delta-limit}. For any
continuous function $f(t)$ we can write
\begin{equation}
\int_{-\infty}^{\infty} r_k(t) f(t) \,dt
=
k \int_{-1/2k}^{1/2k} f(t)\, dt
=
f ( \gamma)\bigr|_{\gamma \in [-1/2k, 1/2k]}
\end{equation}
where we have used the Mean Value theorem. Now, as $k$ goes
to infinity, % we have that
the integral converges to $f(0)$; hence
%and so
we can say that the limit of the series of functions
$r_k(t)$ converges then to the Dirac delta.
As already stated, % we said,
the delta cannot be considered as a proper function, so the
expression $\delta(t)$ outside of an integral sign has no
mathematical meaning; it is customary however to associate
an ``idea''{} of function to the delta and we can think of it
as being undefined for $t \neq 0$ and to have a value of
$\infty$ at $t=0$. This interpretation, together
with~(\ref{deltadef}), defines the so-called \emph{sifting
property} of the Dirac delta; this property allows us to
write (outside of the integral sign):
\begin{equation}\label{sifting}
\delta(t-\tau)f(t) = \delta(t-\tau) \, f(\tau)
\end{equation}
%% 4.12 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[h]
\center\small
\begin{psDFplot}[dy=2,labelx=false]{0}{7}
\selectPlotFont
\definecolor{mygray}{gray}{0.5}
\psset{linecolor=mygray}
\psDFplotData[linewidth=1.5pt](-1,0)(-0.5, 0)(-0.5,1)(0.5,1)(0.5,0)(1,0)
\definecolor{mygray}{gray}{0.4} \psset{linecolor=mygray}
\psDFplotData[linewidth=1.5pt](-1,0)(-0.25, 0)(-0.25,2)(0.25,2)(0.25,0)(1,0)
\definecolor{mygray}{gray}{0.3} \psset{linecolor=mygray}
\psDFplotData[linewidth=1.5pt](-1,0)(-0.1666, 0)(-0.1666,3)(0.1666,3)(0.1666,0)(1,0)
\definecolor{mygray}{gray}{0.2} \psset{linecolor=mygray}
\psDFplotData[linewidth=1.5pt](-1,0)(-0.125, 0)(-0.125,4)(0.125,4)(0.125,0)(1,0)
\definecolor{mygray}{gray}{0.1} \psset{linecolor=mygray}
\psDFplotData[linewidth=1.5pt](-1,0)(-0.1, 0)(-0.1,5)(0.1,5)(0.1,0)(1,0)
\definecolor{mygray}{gray}{0} \psset{linecolor=mygray}
\psDFplotData[linewidth=1.5pt](-1,0)(-0.0833, 0)(-0.0833,6)(0.0833,6)(0.0833,0)(1,0)
\psset{xunit=3.14}
\uput[-90](-1,0){-1} \uput[-90](1,0){1}
\uput[-90](-.5,0){-0.5} \uput[-90](0.5,0){0.5}
\uput[45](0.5,1){$k=1$}
\uput[45](0.25,2){$k=2$}
\uput[45](0.1666,3){$k=3$}
\uput[45](0.125,4){$k=4$}
\uput[45](0.1,5){$k=5$}
\uput[45](0.0833,6){$k=6$}
\end{psDFplot}
\vskip-5mm
\caption{The Dirac delta as the limit of a family of
rectangular functions.}\label{fig-delta-limit}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
The physical interpretation of the Dirac delta is related to
quantities expressed as continuous \emph{distributions} for
which the most familiar example is probably that of a
probability distribution (pdf). These functions represent a
value which makes physical sense only over an interval of
nonzero measure; the punctual value of a distribution is
only an abstraction. The Dirac delta is the operator that
extracts this punctual value from a distribution, in a sense
capturing the essence of considering smaller and smaller
observation intervals.
To see how the Dirac delta applies to our basis expansion,
note that equation~(\ref{deltadef}) is formally identical to
an inner product over the space of functions on the real
line; by using the definition of such an inner product we
can therefore write
\begin{equation}\label{RecFormCont}
f(t) = \int_{-\infty}^{\infty}
\bigl\langle \delta(s-\tau), f(s) \bigr\rangle \, \delta(t-\tau) \, d\tau
\end{equation}
which is, in turn, formally identical to the reconstruction
formula of Section~\ref{InfSeqSpaceDef}. In reality, we are
interested in the space of $2\pi$-periodic functions, since
that is where DTFTs live; this is easily accomplished by
building a $2\pi$-periodic version of the delta
as\index{Dirac delta!pulse train}
\begin{equation}
\tilde{\delta}(\omega) = 2\pi \sum_{k=-\infty}^{\infty} \delta(\omega - 2\pi k)
\end{equation}
where the leading $2\pi$ factor is for later convenience.
The resulting object is called a \emph{pulse train},
similarly to what we built for the case of periodic
sequences in $\tilde{\mathbb{C}}^N$. Using the
%\emph
{pulse train}
and given any $2\pi$-periodic function $ f(\omega)$, the
reconstruction formula~(\ref{RecFormCont}) becomes
\begin{equation}
f(\omega) = \frac{1}{2\pi}
\int_{\sigma}^{\sigma+2\pi}
\bigl\langle \tilde\delta(\theta-\phi), f(\theta) \bigr\rangle
\, \tilde\delta(\omega-\phi) \, d\phi
\end{equation}
for any $\sigma \in \mathbb{R}$.
Now that we have the delta notation in place, we are ready
to start. First of all, we %will
show the formal
orthogonality of the basis functions $\{ e^{j\omega n}
\}_{\omega \in \mathbb{R}}$. We can write
\begin{equation}
\frac{1}{2\pi}\int_{-\pi}^{\pi} \tilde{\delta}
(\omega - \omega_{0})\, e^{j\omega n}d\omega =
e^{j\omega_0 n}
\end{equation}
The left-hand side of this equation has the exact form of
the DTFT reconstruction formula~(\ref{DTFTrec}); %and so
hence we
have found the fundamental relationship\index{Dirac delta!DTFT of} \begin{equation}\label{DTFTofexp}
e^{j\omega_0 n} \;
\stackrel{\scriptstyle\text{DTFT}}{\longleftrightarrow} \;
\tilde{\delta}(\omega - \omega_{0})
\end{equation}
Now, the DTFT of a complex exponential $e^{j\sigma n}$ is,
in our change of basis interpretation, simply the inner
product $\langle e^{j\omega n}, e^{j\sigma n} \rangle$;
because of~(\ref{DTFTofexp}) we can therefore express this as % write:
\begin{equation}
\bigl\langle e^{j\omega n}, e^{j\sigma n} \bigr\rangle =
\tilde{\delta}(\omega - \sigma)
\end{equation}
which is formally equivalent to the orthogonality relation in~(\ref{exportog}).
%\vspace{1em}
We %will
now recall for the last time that the delta notation
subsumes a limiting operation: the DTFT
pair~(\ref{DTFTofexp}) should be interpreted as shorthand
for the limit of the partial sums
\[
s_k(\omega) = \sum_{n=-k}^{k} e^{-j\omega n} \nonumber
\]
(where we have chosen $\omega_0 = 0$ for the sake of
example). Figure~\ref{DTFTconstFig} plots $|s_k(\omega)|$ for
increasing values of $k$ (we show only the $[-\pi,\pi]$
interval, although of course the functions are
$2\pi$-periodic). The family of functions $s_k(\omega)$ is exactly
equivalent to the family of functions $r_k(t)$ we saw
in~(\ref{limtodelta}); they too become
%narrower and narrower
increasingly narrow
while keeping a constant area (which turns out to be
$2\pi$). That is why we can %say
simply state that $s_k(\omega)
\rightarrow \tilde{\delta}(\omega)$.
\eject
%% 4.13 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[H]
\vskip2mm
\center\small
\begin{psDFplot}[dy=20,pifrac=2]{0}{81}
\psDFplotFile[linewidth=1.4pt,plotstyle=line]%,linecolor=gray]%
{data/fdelta5.txt}
\psDFplotFile[linewidth=0.9pt,plotstyle=line]%,linecolor=darkgray]%
{data/fdelta15.txt}
\psDFplotFile[linewidth=0.3pt,plotstyle=line]%,linestyle=dashed]%
{data/fdelta40.txt}
\selectPlotFont
\psset{xunit=3.14}
\uput[45](0.065,8){$k=5$} %0.5
\uput[45](0.02,20){$k=15$} %0.05
\uput[45](0.00,70){$k=40$}
\end{psDFplot}
\vskip-5mm
\caption{The sum
$\bigl|\sum_{n=-k}^{k} \, e^{-j\omega n} \bigr|$ for different values
of $k$.}\label{DTFTconstFig}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\medskip
From~(\ref{DTFTofexp}) we can easily obtain other
interesting results: by setting $\omega_0 = 0$ and by
exploiting the linearity of the DTFT operator, we can derive
the DTFT of a constant sequence:
\begin{equation}
\alpha \; \stackrel{\scriptstyle
\text{DTFT}}{\longleftrightarrow} \;
\alpha \tilde{\delta}(\omega)
\end{equation}
or, using Euler's formulas, the DTFTs of sinusoidal functions:
\begin{align}
&\cos(\omega_0 n + \phi) \; \stackrel{\scriptstyle
\text{DTFT}}{\longleftrightarrow} \;
\frac{1}{2} \bigl[e^{j\phi} \tilde{\delta}(\omega -\omega_0) + e^{-j\phi} \tilde{\delta}(\omega + \omega_0 ) \bigr]
\label{DTFTCosDef} \\
&\sin(\omega_0 n + \phi) \;
\stackrel{\scriptstyle \text{DTFT}}{\longleftrightarrow}
\; \frac{-j}{2} \bigl[e^{j\phi} \tilde{\delta}(\omega -\omega_0)
- e^{-j\phi} \tilde{\delta}(\omega + \omega_0) \bigr] \label{DTFTSinDef}
\end{align}
As we can see from the above examples, we are defining the
DTFT for sequences which are not even square-summable;
again, these transforms are purely a notational formalism
used to capture a behavior, in the limit, as we showed before.\index{DTFT|)}

Event Timeline