Teaching demo
Filter Lab
You have received a distress signal, but it is very noisy. Consider the frequency spectrum and filter out the noise to recover the signal.
Press Play. For the first two seconds you hear only the interference — six pure tones and broadband static. Then the signal fades in underneath. Drag the red markers (once you add a notch) to hunt down each whistle: the spikes on the spectrum give them away.
The loud spikes that never move are pure tones — single frequencies. A notch filter is the surgical tool for exactly that: it removes a razor-thin slice and leaves everything else alone.
Stare at the spectrum for a while: which parts change over time, and which parts stay frozen? The signal is the part that moves. Maybe the band-pass can keep just that region and throw the rest away.
Your toolkit
A notch cuts one narrow slice of frequencies — sharpness (Q) sets how narrow. The band-pass keeps only what lies between its two edges and discards everything else. The output view shows what your filters leave behind. Filters can only remove; the trick is working out what to keep.
Re-implementation of the public-domain teaching-dsp demo (a TeachEngineering project by Connor McKay, Christopher Light & Ayoade Adewole), rebuilt to run on this site with the original audio. Notch filters are biquads with adjustable Q (wide to aim, sharp to be surgical); the band-pass is two cascaded biquads whose Q follows the selected bandwidth. The white curve is the true frequency response of your filter chain, straight from the audio engine.