A square wave is the simplest signal imaginable. It's 1,
then it's 0. A child can draw it.
Fourier analysis says: to represent this, sum infinitely many sine waves.
But the sum can't reproduce the jump. At every discontinuity it
overshoots by ~8.95%. Not at 10 terms. Not at 100. Not at 10,000.
At every finite N, forever. This is the
Gibbs phenomenon — and it causes ringing artifacts in
JPEG images, MRI scans, audio processing, and radar.
The signal was always finite. The overshoot is the model's problem,
not the signal's.
Key Interaction
Drag the Fourier Terms slider from 2 to 512.
Watch the peak overshoot. It doesn't move. 10 terms, 100 terms, 500 terms
— the Fourier sum always overshoots by ~8.95%. The symbolic gate says 1.
The Fourier gate says 1.089…The infinity was in the model, not the signal.
1Select a measurement point — peak overshoot, center, or edge
2Watch the SymbolicGate look up the exact value in finite steps
3Drag the slider — the FourierGate adds more sine waves but the overshoot persists
SISO — Fourier's Overshoot
The Gibbs phenomenon — finite signal, infinite model, permanent error
Configuration
16
Gates in Stream
SymbolicGate
Piecewise lookup. Is x in (0,π)? Then f(x)=1. Finite rule. Exact answer. No series.
FourierGate
Partial sum of N sine harmonics. Infinite terms needed — but infinity isn't available.
Event Stream
📥
📐
🔊
📤
📐 Symbolic — Piecewise Lookup
The signal is defined. The value is known. No computation required.