Fourier Series vs Fourier Transform: Building Waves vs Decoding Their Spectrum
Both simulators reconstruct waveforms from rotating phasors, but they teach opposite halves of Fourier analysis — one builds a periodic wave from a discrete ladder of harmonics, the other decodes the continuous spectrum hidden inside any signal.
🧪 Try both simulations
⚡ Quick answer
Both simulators share the same rotating-phasor (epicycle) visual, but they zoom in on different halves of one idea. Fourier Series treats a signal as already periodic and rebuilds it from a discrete set of harmonics at integer multiples of one fundamental frequency — the lesson is the Gibbs phenomenon, a ~9% overshoot near discontinuities that refuses to vanish no matter how many terms you add. Fourier Transform generalises this to any signal, periodic or not, and its lesson is the machinery underneath real spectrum analysis: why the FFT beats the DFT, how windowing tames spectral leakage, and why sharper timing means blurrier frequency (and vice versa).
📊 Fourier Series vs Fourier Transform — key differences
| Fourier Series | Fourier Transform | |
|---|---|---|
| What it decomposes | An already-periodic function into a discrete sum of harmonics at n·ω₀ | Any signal, periodic or not, into a continuous spectrum over all frequencies ω |
| Core formula | f(x) = a₀/2 + Σ[aₙ·cos(nω₀x) + bₙ·sin(nω₀x)] — discrete coefficients aₙ, bₙ | F(ω) = ∫f(t)e^(−iωt)dt — a continuous integral, computed in practice via the DFT/FFT |
| Spectrum shape | Discrete line spectrum: energy only at the fundamental and its integer harmonics | Continuous (or densely sampled) spectrum spanning every frequency, including non-harmonic content and noise |
| Algorithm behind it | A simple finite partial sum over N chosen terms — no transform algorithm needed | Naive DFT costs O(N²); the Cooley–Tukey FFT (1965) cuts that to O(N log N) |
| Interactive hook on this site | Pick a target waveform, add harmonics one by one, and watch the epicycle sum converge | Same phasor-building mechanic, extended with frequency-spectrum bars, windowing controls and leakage discussion |
| Signature phenomenon | Gibbs phenomenon — a ~9% overshoot at every jump discontinuity that never disappears | Spectral leakage and the time–frequency uncertainty principle (Δt·Δf ≥ 1/4π) |
| Best for learning | How sines and cosines form an orthogonal basis that rebuilds any periodic wave | How real spectrum-analysis tools (FFT, windowing, resolution trade-offs) work in audio, imaging and radar |
The Fourier Series simulator starts from the assumption that your signal already repeats forever, so the only question is which harmonics — and how many of them — reproduce it. Sines and cosines at n·ω₀ form an orthogonal basis for periodic functions, the same way x, y, z axes form a basis for 3D space, and each coefficient aₙ, bₙ can be found independently by projecting the target wave onto that harmonic. As you drag the harmonic count N upward on this site, the rotating epicycles chain together and the partial sum sharpens toward a square, sawtooth or triangle wave — except right at the corners, where a stubborn ~9% overshoot (the Gibbs phenomenon, explained by J. W. Gibbs in 1899) never goes away no matter how many terms you add.
The Fourier Transform simulator keeps the same rotating-phasor visual but pushes into the machinery that makes spectrum analysis practical for real, non-repeating signals. Instead of a fixed ladder of harmonics, it works with F(ω) = ∫f(t)e^(−iωt)dt, a continuous integral that in the digital world becomes the Discrete Fourier Transform — an O(N²) calculation the Fast Fourier Transform (Cooley–Tukey, 1965) collapses to O(N log N), the difference between 3 hours and 0.02 seconds for a million-point signal. This is also where the practical headaches of spectrum analysis live: spectral leakage, caused by a signal that doesn't repeat cleanly inside the analysis window, tamed by Hann, Hamming or Blackman windowing functions at the cost of frequency resolution; and the time–frequency uncertainty principle, which says a short window pins down when something happened but blurs its frequency, while a long window does the reverse.
Want to explore more? Browse the full library of 1000+ interactive, browser-based simulations, or see other side-by-side comparisons.