Fourier Series and Transforms in Depth With easy examples.

The Fourier series is the discovery that any periodic signal — a sound wave, a heartbeat, an alternating current — can be rebuilt by adding simple sine and cosine waves of the right frequencies and amplitudes. The Fourier transform extends the idea to non-periodic signals, decomposing any signal into its frequency ingredients: the time domain (what happens when) versus the frequency domain (which frequencies are present).

It matters because the frequency view makes hard problems easy. Filtering noise from audio, compressing images, analysing bridge vibrations and reconstructing MRI scans are all frequency-domain operations — Fourier's 1807 insight now sits inside nearly every digital device.

Key Concepts

• Periodic function: a signal repeating every T seconds, with fundamental frequency f₀ = 1/T in hertz.
• Harmonics: components at integer multiples of the fundamental (f₀, 2f₀, 3f₀, …).
• Fourier coefficients (aₙ, bₙ): the amplitude of each cosine/sine harmonic; computing them is analysis, rebuilding the signal is synthesis.
• Spectrum: amplitude versus frequency — the signal's frequency-domain fingerprint.
• Fourier transform: the generalisation to non-periodic signals, giving a continuous spectrum F(ω).
• DFT / FFT: the Discrete Fourier Transform handles sampled data; the Fast Fourier Transform computes it in n·log(n) steps instead of n², enabling real-time processing.

Formulas

• Fourier series: f(x) = a₀/2 + Σ (aₙ·cos(nx) + bₙ·sin(nx)), for n = 1, 2, 3, ….
• Coefficients: aₙ = (1/π)·∫₋π^π f(x)·cos(nx) dx, bₙ = (1/π)·∫₋π^π f(x)·sin(nx) dx, for period 2π.
• Fourier transform: F(ω) = ∫₋∞^∞ f(t)·e^(−jωt) dt, with angular frequency ω in rad/s and j the imaginary unit.
• Converges for the piecewise-smooth signals met in engineering; frequencies in Hz, angular frequencies in rad/s.

Worked Example

Approximate a square wave of amplitude 1 and period 2π (odd function, so all aₙ = 0) with its first two sine terms. Theory gives bₙ = 4/(nπ) for odd n, so f(x) ≈ (4/π)·(sin x + sin 3x / 3). Evaluate at x = π/2, where the wave equals 1.

1. First term: (4/π)·sin(π/2) = 4/π ≈ 1.2732.
2. Second term: (4/π)·sin(3π/2)/3 = (4/π)·(−1)/3 ≈ −0.4244.
3. Sum ≈ 0.8488 — already approaching 1 with two terms. Adding the 5th harmonic gives ≈ 1.1036, overshooting slightly; the series converges to 1 as more terms join.

Each harmonic refines the approximation: the fundamental sets the shape, higher harmonics sharpen the edges. The small overshoot near the jump is the famous Gibbs phenomenon.

Common Misconceptions

• "Fourier analysis is only for sound." Images are two-dimensional signals — JPEG compression is essentially a Fourier-like transform discarding high frequencies your eyes miss.
• "An infinite series means infinite computation." A finite number of terms (or a finite DFT) captures everything the application needs; the FFT made it fast.
• "The transform throws away timing." The plain transform averages over all time, but windowed versions like the Short-Time Fourier Transform recover when each frequency occurs.

Real-World Applications

• Audio: MP3/AAC compression, equalisers and noise removal operate on spectra.
• Imaging: JPEG compression and MRI reconstruction are Fourier-domain techniques.
• Communications: OFDM (Wi-Fi, 4G, 5G) packs data onto orthogonal frequency harmonics.

The Fourier series builds any periodic signal from harmonics, and the Fourier transform reveals any signal's frequency content — two views of the same reality. Learn to read a spectrum, and signals stop being squiggles and start being recipes.