Deterministic vs Random Signals

Signals in real engineering systems come in two fundamentally different forms: those whose future values can be exactly predicted from their mathematical description, and those that carry inherent uncertainty. Understanding the difference between deterministic and random signals is essential for designing communication systems, noise analysis, and statistical signal processing, all of which are core topics in both university curricula and GATE examination.

Core Concept Explanation

A deterministic signal is one whose value at every point in time can be computed exactly from a known mathematical expression. Given the formula x(t) = 5 sin(2π × 1000 × t), you can calculate the exact voltage at t = 0.001 seconds without any ambiguity. The signal has no randomness embedded in it. Its past, present, and future values are all determined by the same closed-form expression.

A random signal, also called a stochastic signal, is one whose exact value cannot be predicted in advance. Thermal noise in a resistor, speech recorded from a microphone, or the interference received by an antenna are all random signals. At any given time, you can only state the probability that the signal lies within a certain range. The mathematical tool used to describe such signals is the probability density function (PDF), along with statistical quantities like mean, variance, autocorrelation, and power spectral density.

The distinction is not merely theoretical. In communication system design, the transmitted carrier is a deterministic signal while the channel noise is random. The entire framework of detection theory, matched filtering, and signal-to-noise ratio optimization exists precisely because these two categories of signals must be handled with completely different mathematical tools.

Mathematical Expression

For a deterministic signal, the complete description is the function x(t) itself. Properties like energy, power, and Fourier transform are computed directly by integration or summation. There is no statistical averaging needed because the signal is fully known.

For a random signal, the mathematical description involves a random process X(t), which is a family of sample functions. At any fixed time t₁, X(t₁) is a random variable characterized by its PDF f(x; t₁). The mean is defined as μ(t) = E[X(t)] = integral of x times f(x,t) dx. The autocorrelation function R(t₁, t₂) = E[X(t₁) X(t₂)] describes how values at two time instants are statistically related. For a wide-sense stationary (WSS) process, mean is constant and autocorrelation depends only on the time difference τ = t₂ - t₁, written as R(τ).

Practical Understanding

In practical systems, most signals of interest in communications are modeled as random processes. The information being transmitted, whether voice, data, or video, is inherently unpredictable to the receiver. If it were fully predictable, there would be no need to transmit it. Channel noise modeled as additive white Gaussian noise (AWGN) is a random process with known statistical properties: zero mean and a flat power spectral density of N₀/2 watts per hertz.

Deterministic signals appear in system analysis as test signals: sinusoids for frequency response testing, impulses for impulse response measurement, and step functions for transient analysis. In signal processing labs, a deterministic sinusoid at a known frequency is injected into a circuit and the output is measured to characterize the system's behavior. This clean separation between deterministic test signals and random operating signals is fundamental to experimental system identification.

Given:
A random signal X(t) has PDF: f(x) = (1/2) e^(-|x|) for all x (Laplace distribution)
Find: Mean μ and Variance σ²

Why this formula applies:
For a stationary random process, mean and variance are found from the marginal PDF using expectation integrals.

Formula:
μ = E[X] = ∫ x · f(x) dx from -∞ to +∞
σ² = E[X²] - μ²

Substitution:
μ = ∫ x · (1/2) e^(-|x|) dx from -∞ to +∞
Since f(x) is symmetric and x is odd, integrand is odd:
μ = 0

E[X²] = ∫ x² · (1/2) e^(-|x|) dx
= 2 × ∫₀^∞ x² · (1/2) e^(-x) dx  [using symmetry]
= ∫₀^∞ x² e^(-x) dx = Γ(3) = 2! = 2

Calculation:
σ² = E[X²] - μ² = 2 - 0 = 2

Final Answer:
Mean μ = 0, Variance σ² = 2 (dimensionless for normalized signal)

Exam Tip: In GATE, deterministic vs random is often tested through autocorrelation. A deterministic signal has a definite autocorrelation R(τ) = integral of x(t) x(t+τ) dt. A WSS random process has R(τ) = E[X(t) X(t+τ)]. Remember that R(0) equals the average power, and R(τ) is maximum at τ = 0. Also, WSS process implies ergodic behavior is assumed in many GATE problems unless stated otherwise.

Quick Revision

• Deterministic signals: completely described by a mathematical function. No uncertainty. Future values can be exactly computed.
• Random signals: described by statistical properties such as PDF, mean, variance, autocorrelation, and power spectral density (PSD).
• A random process X(t) is called WSS (wide-sense stationary) if mean is constant and autocorrelation R(τ) depends only on time difference τ, not absolute time.
• Autocorrelation at zero lag R(0) equals the average power of the random signal.
• GATE trap: Do not apply Fourier transform directly to a random signal sample function and call it the spectrum. The correct tool is the Power Spectral Density (PSD), which is the Fourier transform of the autocorrelation function R(τ) via the Wiener-Khinchin theorem.
• Examples: Deterministic: sinusoid, unit step, ramp, exponential pulse. Random: thermal noise, shot noise, speech signal, channel fading.
• Ergodic process: time averages equal ensemble averages. All ergodic processes are stationary, but not all stationary processes are ergodic.