Classification of Signals

Introduction

Signals come in many flavors, and being able to classify them is essential for choosing the right analysis tools. Just as a biologist classifies organisms to understand their properties, engineers classify signals to determine which mathematical techniques apply. A periodic signal gets Fourier Series treatment; a causal signal might need the unilateral Laplace transform. The classification tells you what tools to reach for.

Continuous-Time vs. Discrete-Time

The most fundamental classification divides signals by their independent variable:

Continuous-time signals $x(t)$ are defined for all real values of $t$. The signal exists at every instant — there are no gaps.

Discrete-time signals $x[n]$ are defined only at integer values of $n$. Between these integer points, the signal simply doesn't exist.

Important: a discrete-time signal is NOT the same as a "staircase" — it's a sequence of numbers, not a continuous function that happens to look stepwise.

Analog vs. Digital

This classification concerns the amplitude (value) of the signal:

• Analog signal: Both time and amplitude are continuous. Example: voltage from a thermocouple.
• Digital signal: Both time and amplitude are discrete (quantized). Example: data stored in a computer.

A sampled signal (discrete time, continuous amplitude) is neither purely analog nor digital — it's the intermediate step in analog-to-digital conversion.

Periodic vs. Aperiodic

A continuous-time signal is periodic if there exists a positive constant $T$ such that:

$$x(t) = x(t + T) \quad \text{for all } t$$

The smallest such $T$ is the fundamental period, and $f_0 = 1/T$ is the fundamental frequency.

For discrete-time signals, $x[n]$ is periodic with period $N$ if:

$$x[n] = x[n + N] \quad \text{for all } n$$

Example: $x(t) = \cos(2\pi \cdot 5t)$ is periodic with $T = 0.2$ seconds.

Aperiodic signals never repeat. A single rectangular pulse, a decaying exponential, and speech signals are all aperiodic.

Key fact: The sum of two periodic signals is periodic only if the ratio of their periods $T_1/T_2$ is a rational number. If $x_1(t)$ has period $T_1 = 1$ and $x_2(t)$ has period $T_2 = \sqrt{2}$, their sum is NOT periodic because $1/\sqrt{2}$ is irrational.

Even and Odd Signals

Even signal: Symmetric about $t = 0$: $$x(-t) = x(t)$$ Example: $\cos(t)$, $|t|$, constant signals.

Odd signal: Anti-symmetric about $t = 0$: $$x(-t) = -x(t)$$ Example: $\sin(t)$, $t$, $t^3$.

Decomposition theorem: ANY signal can be decomposed into the sum of an even part and an odd part:

$$x(t) = x_e(t) + x_o(t)$$

where: $$x_e(t) = \frac{x(t) + x(-t)}{2} \qquad x_o(t) = \frac{x(t) - x(-t)}{2}$$

This decomposition is unique and extremely useful in Fourier analysis, where even signals produce only cosine terms and odd signals produce only sine terms.

Energy Signals vs. Power Signals

This classification is based on how the signal's "size" is measured:

Energy of a signal: $$E = \int_{-\infty}^{\infty} |x(t)|^2 \, dt \quad \text{(continuous-time)}$$ $$E = \sum_{n=-\infty}^{\infty} |x[n]|^2 \quad \text{(discrete-time)}$$

Average power of a signal: $$P = \lim_{T \to \infty} \frac{1}{2T} \int_{-T}^{T} |x(t)|^2 \, dt$$

Classification rules:

• Energy signal: $0 < E < \infty$ and $P = 0$. These signals have finite total energy. Example: $x(t) = e^{-|t|}$
• Power signal: $E = \infty$ and $0 < P < \infty$. These signals have infinite energy but finite average power. Example: $x(t) = \cos(\omega_0 t)$, constant signal
• Neither: $E = \infty$ and $P = \infty$. Example: $x(t) = t$ (the ramp goes to infinity)

Practical insight: Signals that die out (transients) are energy signals. Signals that persist forever at a steady level are power signals.

Deterministic vs. Random

Deterministic signals can be described by an explicit mathematical expression. You can predict their exact value at any future time. Example: $x(t) = 3e^{-2t}u(t)$.

Random (stochastic) signals require probabilistic description. You can only specify statistical properties like mean, variance, and autocorrelation. Example: thermal noise, stock prices, speech.

In practice, most real-world signals are random, but we often model them as deterministic for analysis purposes (e.g., treating a known audio file as deterministic).

Causal, Anti-Causal, and Non-Causal

Causal signal: Zero for all $t < 0$: $$x(t) = 0 \quad \text{for } t < 0$$ Example: $x(t) = e^{-2t}u(t)$ — the signal "starts" at $t = 0$.

Anti-causal signal: Zero for all $t > 0$: $$x(t) = 0 \quad \text{for } t > 0$$ Example: $x(t) = e^{3t}u(-t)$

Non-causal signal: Exists on both sides of $t = 0$: Example: $x(t) = e^{-|t|}$

This classification matters when signals represent physical inputs — in real-time processing, you can only use present and past values (causal signals).

Bounded vs. Unbounded

A signal is bounded if there exists a finite constant $M$ such that: $$|x(t)| \leq M \quad \text{for all } t$$

Bounded signals: $\sin(t)$, $u(t)$, any constant Unbounded signals: $t$, $e^{2t}$, $\tan(t)$

Boundedness is critical for stability analysis — BIBO (Bounded Input, Bounded Output) stability requires that bounded inputs always produce bounded outputs.

Real vs. Complex Signals

Real signals: $x(t) \in \mathbb{R}$ for all $t$. Most physical signals are real.

Complex signals: $x(t) = x_r(t) + jx_i(t)$ where $j = \sqrt{-1}$. Complex exponentials like $e^{j\omega t}$ are technically complex signals, but they're indispensable in analysis because of Euler's formula:

$$e^{j\omega t} = \cos(\omega t) + j\sin(\omega t)$$

Singularity (Generalized) Functions

Some important signals don't behave like normal functions — they're generalized functions or distributions:

• Unit impulse $\delta(t)$: Infinite at $t=0$, zero everywhere else, integrates to 1
• Unit doublet $\delta'(t)$: Derivative of the impulse

These are classified separately because they require distribution theory rather than ordinary calculus.

Summary Table

Key Takeaways

• Classification determines which analytical tools apply to a signal
• Every signal can be uniquely decomposed into even + odd components
• Energy signals decay to zero; power signals persist indefinitely
• The periodic/aperiodic distinction determines whether to use Fourier Series or Fourier Transform
• Real physical signals are causal (they start at some definite time)
• Understanding these categories helps you quickly identify the right analysis approach for any signal processing problem