Signals: Discrete-Time Fourier Transform

The Discrete-Time Fourier Transform (DTFT) is the cornerstone tool for analyzing the frequency content of a sequence of numbers. Unlike its cousin, the DFT, which operates on a finite block of data, the DTFT handles sequences that are theoretically infinite in length, providing a continuous view of frequency. Mastering the DTFT is essential because it forms the theoretical bridge between the sampled world of digital signal processing and the continuous world of frequency analysis, enabling you to design filters, understand system behavior, and lay the groundwork for more practical algorithms.

From Discrete Time to Continuous Frequency

The Discrete-Time Fourier Transform (DTFT) takes a discrete-time signal, $x[n]$, and maps it to a continuous, complex-valued function of frequency, $X(e^{j\omega })$. The forward transform is defined by the summation:

$$X(e^{j\omega })=\sum _{n=-\infty }^{\infty }x[n]e^{-j\omega n}$$

Here, $n$ is the discrete time index, and $\omega$ is the digital frequency in radians per sample. This is the key distinction: while $n$ is discrete, $\omega$ is a continuous variable. The inverse transform, which recovers the time-domain signal from its frequency spectrum, is given by the integral:

$$x[n]=\frac{1}{2\pi }{\int }_{-\pi }^{\pi }X(e^{j\omega })e^{j\omega n}d\omega$$

The limits of integration are $-\pi$ to $\pi$. This is not an arbitrary choice. The function $X(e^{j\omega })$ is always periodic with a period of $2\pi$. This periodicity in frequency is a fundamental consequence of sampling a continuous-time signal. In essence, the DTFT tells you that all unique frequency information for a discrete-time signal is contained within any $2\pi$ interval, typically $[-\pi ,\pi )$ or $[0,2\pi )$. Frequencies outside this range are aliases, a concept you'll encounter when relating digital and analog frequencies.

Computing DTFT Pairs and Key Examples

A DTFT pair is the unique association between a sequence $x[n]$ and its transform $X(e^{j\omega })$. You derive these pairs by applying the summation formula. Let's work through two fundamental examples that form the building blocks for more complex signals.

First, consider the impulse sequence, $\delta [n]$, which is 1 at $n=0$ and 0 elsewhere. Its DTFT is remarkably simple: $$X(e^{j\omega })=\sum _{n=-\infty }^{\infty }\delta [n]e^{-j\omega n}=e^{-j\omega \cdot 0}=1$$ This tells us an impulse in time contains all frequencies equally, a crucial property for system testing.

Second, consider a causal exponential sequence, $x[n]=a^{n}u[n]$, where $|a|<1$ for convergence and $u[n]$ is the unit step. Computing its DTFT involves an infinite geometric series: $$X(e^{j\omega })=\sum _{n=0}^{\infty }{a}^n{e}^{-j\omega n}=\sum _{n=0}^{\infty }(a{e}^{-j\omega }{)}^n=\frac{1}{1-a{e}^{-j\omega }},|a{e}^{-j\omega }|=|a|<1$$ The magnitude of this spectrum, $|X(e^{j\omega })|=1/\sqrt{1+a^2-2a\cos \omega }$, shows a low-pass shape when $0<a<1$. These pairs, along with those for a rectangular pulse or a sinusoidal sequence, are essential to memorize and understand, as they are frequently used in analysis and design.

Leveraging DTFT Properties for Manipulation

The true power of the DTFT lies in its properties, which allow you to manipulate signals in the frequency domain without recomputing the full transform. These properties are derived directly from the definition. Here are the most critical ones:

• Linearity: The DTFT of a sum is the sum of the DTFTs. If $y[n]=a{x}_1[n]+b{x}_2[n]$, then $Y(e^{j\omega })=a{X}_1(e^{j\omega })+b{X}_2(e^{j\omega })$. This lets you build complex transforms from simpler ones.
• Time Shifting: Shifting a signal in time corresponds to a phase shift in frequency. If $y[n]=x[n-n_0]$, then $Y(e^{j\omega })=e^{-j\omega n_0}X(e^{j\omega })$. The magnitude spectrum is unchanged—only the phase is altered.
• Frequency Shifting (Modulation): The dual property. Multiplying by a complex exponential in time shifts the frequency. If $y[n]=e^{j{\omega }_{0}n}x[n]$, then $Y(e^{j\omega })=X(e^{j(\omega -{\omega }_0)})$. This is the basis for amplitude modulation schemes.
• Convolution: This is arguably the most important property for system analysis. The convolution of two sequences in the time domain corresponds to multiplication of their DTFTs in the frequency domain. If $y[n]=x[n]\ast h[n]$, then $Y(e^{j\omega })=X(e^{j\omega })H(e^{j\omega })$. This transforms complex convolution operations into simple multiplication.

Periodicity, Symmetry, and the Link to the DFT

As noted, $X(e^{j\omega })$ is always periodic: $X(e^{j(\omega +2\pi )})=X(e^{j\omega })$. This is because the complex exponential $e^{-j\omega n}$ is itself periodic in $\omega$ with period $2\pi$. When $x[n]$ is a real-valued sequence, this periodicity leads to important conjugate symmetry in the DTFT: $X(e^{j\omega })=X^{\ast }(e^{-j\omega })$. This implies the magnitude spectrum is an even function ($|X(e^{j\omega })|=|X(e^{-j\omega })|$) and the phase spectrum is an odd function.

This continuous, periodic nature of the DTFT leads directly to the Discrete Fourier Transform (DFT). The DFT is essentially a sampled version of the DTFT. If you take a finite-length sequence $x[n]$ for $n=0,...,N-1$ and evaluate its DTFT at $N$ equally spaced frequencies ${\omega }_k=2\pi k/N$, you get the DFT coefficients:

$$X[k]=X(e^{j\omega }){|}_{\omega =2\pi k/N}=\sum _{n=0}^{N-1}x[n]e^{-j(2\pi /N)kn}$$

This process is called frequency sampling. The DFT provides a practical, computable snapshot of the continuous DTFT. Understanding this relationship is vital: the DFT assumes the finite sequence you provide is one period of a periodic signal, which aligns perfectly with the inherent periodicity the DTFT reveals in the frequency domain.

Common Pitfalls

1. Confusing Periodicity Domains: A frequent error is misstating what is periodic. The sequence $x[n]$ in the time domain is generally not periodic. The DTFT, $X(e^{j\omega })$, in the continuous frequency domain, is always periodic. Remember: discrete time leads to a continuous, periodic frequency spectrum.

1. Ignoring Convergence Conditions: The DTFT summation ${\sum }_{n=-\infty }^{\infty }x[n]e^{-j\omega n}$ does not converge for all sequences. It requires the sequence to be absolutely summable ($\sum |x[n]|<\infty$) or have finite energy. Sequences like the unit step $u[n]$ or a complex exponential $e^{j{\omega }_{0}n}$ do not have a conventional DTFT; they require the use of impulse functions in frequency, which is an advanced extension of the theory.

1. Misapplying Symmetry Properties: When a sequence is real-valued, its DTFT has conjugate symmetry. A common mistake is to assume this means the DTFT itself is real-valued. It does not. It means the real part is even and the imaginary part is odd. The DTFT is only purely real if the time-domain sequence is also even (symmetric about $n=0$).

1. Forgetting the $2\pi$ Scale Factor in the Inverse DTFT: When performing the inverse transform via integration, omitting the $\frac{1}{2\pi }$ factor is a critical algebraic error. This factor is essential for the correct normalization of energy/power between the time and frequency domains.

Summary

• The Discrete-Time Fourier Transform (DTFT) provides a continuous, periodic frequency representation $X(e^{j\omega })$ for a discrete-time sequence $x[n]$, defined by $X(e^{j\omega })={\sum }_{n=-\infty }^{\infty }x[n]e^{-j\omega n}$.
• Its most critical properties—linearity, time/frequency shifting, and especially convolution—enable powerful frequency-domain analysis and manipulation of signals and systems.
• The DTFT is inherently periodic in frequency with a period of $2\pi$, a direct result of sampling, and for real-valued sequences, it exhibits conjugate symmetry.
• The Discrete Fourier Transform (DFT) is derived from the DTFT via frequency sampling, evaluating the continuous $X(e^{j\omega })$ at discrete points ${\omega }_k=2\pi k/N$.
• Success with the DTFT requires careful attention to its convergence conditions and the correct application of its inverse transform, $x[n]=\frac{1}{2\pi }{\int }_{-\pi }^{\pi }X(e^{j\omega })e^{j\omega n}d\omega$.