≋ overtoneTHE PHYSICS PLAYGROUND02 / WAVES & SOUND
MAKE A WAVE. FIND YOUR FREQUENCY.

Sound, unpacked.

A tiny sound studio. A living physics lesson.
Build a signal, sculpt its spectrum, and hear the mathematics.

Sound off · plots are active

01 / The waveform

original after EQ
Time → · relative amplitude · automatic scale
How to read the waveform

What happens moment by moment?

The horizontal axis is time in milliseconds. The vertical axis is the instantaneous, signed signal amplitude. Above and below zero are opposite directions of the oscillation—not positive and negative loudness.

Every frequency component contributes a sine wave. At each instant, those values add. Components can reinforce or cancel each other, making peaks, troughs, and the overall shape.

After EQ: boosting an existing component gives its oscillation more weight in the sum; cutting it gives it less. Emphasizing higher harmonics can add fine ripples or sharpen transitions, while removing them can smooth the signal. The exact shape also depends on relative phase.

Watch the scale: both traces use the same automatic vertical scale. A louder signal can make the axis expand instead of looking taller. Compare the original and processed traces and read the axis labels; do not compare pixel height alone across different settings.

02 / The spectrum

original after EQ
20 Hz → 20 kHz · log frequencyPeak amplitude · dB re A = 1
How to read the spectrum

Which frequencies are present?

The horizontal axis is frequency in hertz, on a logarithmic scale: equal spacing means equal frequency ratios. Moving from 100 Hz to 1 kHz spans the same distance as moving from 1 kHz to 10 kHz.

Each vertical line marks one frequency present in the synthesized signal. Its top shows that component’s peak amplitude in decibels, referenced to amplitude 1. For example, 0 dB means amplitude 1; −6.02 dB means about 0.5; +6.02 dB means about 2. Negative dB means smaller than the reference, not a negative oscillation.

Muted violet = original. Amber = after EQ. Paired lines are shifted a few pixels apart so you can compare them; that tiny visual offset is not a pitch change. Boosting EQ raises an affected amber line, cutting lowers it, and its frequency stays fixed.

This is a calculated line spectrum of the demo tones, not a music recording, microphone measurement, or live FFT. The display spans −60 to +24 dB. Silence has no line; components below the floor are not resolved. The spectrum shows magnitude, not phase, so different waveforms can share the same spectrum.

A

Build the source

Harmonics are the ingredients of timbre.

Eight partials only: these are finite Fourier approximations, not ideal sharp-edged waves. Each slider sets a harmonic’s peak amplitude before EQ.

Enable the second wave to explore beats and cancellation.

B

Shape the equalizer

Boost or cut a neighborhood of frequencies.

6 overlapping bell filters · Q = 1.2
Combined response · ±24 dB plotEQ active
How EQ changes the signal

Why a slider is not a spectrum bar

An EQ slider sets a gain in dB around its labeled center frequency. It does not set that frequency’s absolute amplitude. The bell-shaped band also affects nearby frequencies, and bands overlap. The magenta response curve shows their combined effect.

Lout(f)=Lin(f)+Gtotal(f)L_{\mathrm{out}}(f)=L_{\mathrm{in}}(f)+G_{\mathrm{total}}(f)

For a nonzero component, its output level is its input level plus the total EQ gain at that frequency. A line starting at −12 dB with +6 dB of EQ ends at −6 dB. If the source has no component there, a boost cannot create one.

The causal chain: change a band → change the response at nearby frequencies → change their amplitudes and phases → add the components again → get a new waveform. The EQ response is a property of the filter; the spectrum is a property of the signal.

Loads a 250 Hz square-like source and cuts the 750 Hz band by 12 dB. Compare the H3 spectrum line and waveform, then toggle Bypass EQ. Nearby harmonics also change slightly because the filter has width. Sound starts only with Listen.

LISTENING MODES Illustrative curves · not genre rules

Flat

Listen for:

The trade-off:

Why do equalizers use genre names?

Names such as Rock, Jazz, and Soul are listening shortcuts: they suggest a tonal balance that a preset designer associates with that style. They do not identify a special kind of sound wave, and they do not turn one genre into another.

There is no universal Rock or Jazz curve. Different players and manufacturers can use different gains, band frequencies, and filter widths under the same name. The curves in Overtone are our own teaching examples, not copies of a particular device.

Choose a preset by what you hear, not just the track’s genre. The recording, headphones, speakers, room, and your preference all matter. Start from Flat, compare with bypass, and use small adjustments.

Try this here: select Sawtooth-ish, choose a listening mode, and toggle bypass. A pure sine has only one frequency, so most EQ differences will mainly change its level; a harmonic-rich signal makes changes in timbre easier to hear. This lab synthesizes tones, not rock or jazz recordings.

The same equation applies to every mode: A′(f)=A(f) 10G(f)/20A'(f)=A(f)\,10^{G(f)/20}. Genre names only choose the band settings. The band values below are each filter’s gain setting; overlapping filters make the combined gain curve different from those individual numbers.

EQ changes existing components; it cannot create a missing harmonic. The curve is the combined response, so neighboring bands interact.

PERIOD OF THE FUNDAMENTAL
T=1f0T=\frac{1}{f_0}
WAVELENGTH IN AIR
λ=cf0\lambda=\frac{c}{f_0}

c ≈ 343 m/s at 20°C

BEAT FREQUENCY
fbeat=∣f2−f1∣f_{\mathrm{beat}}=|f_2-f_1|

With second tone enabled

PLAYBACK HEADROOMAutomatic trim · plots stay untrimmed

Turn a knob. Test an idea.

FOUR GUIDED EXPERIMENTS
Start with a sine. Add a little third harmonic and compare the sound and shape.

The equations behind the sound.

FROM ONE SINE TO A SPECTRUM
01 / A TRAVELING WAVE

One repeating motion

y(x,t)=Asin⁡(kx−ωt+φ)ω=2πf,k=2πλc=fλ\begin{gathered}y(x,t)=A\sin(kx-\omega t+\varphi)\\[6pt]\omega=2\pi f,\quad k=\frac{2\pi}{\lambda}\\[6pt]c=f\lambda\end{gathered}

AA is amplitude, ff frequency, φ\varphi phase, λ\lambda wavelength, and cc wave speed. At a fixed point, this becomes a sinusoid in time (up to phase and direction convention). The scope shows a time signal, not a spatial snapshot.

02 / SUPERPOSITION

A recipe for timbre

s(t)=∑n=18Ansin⁡(2πnf0t+φn)T0=1f0\begin{gathered}s(t)=\sum_{n=1}^{8} A_n\sin(2\pi n f_0t+\varphi_n)\\[8pt]T_0=\frac{1}{f_0}\end{gathered}

n=1,…,8n=1,\ldots,8 here. Harmonics are integer multiples of f₀; changing their amplitudes changes the waveform and timbre. Square-ish uses odd An=1nA_n=\frac{1}{n}. Triangle-ish uses odd An=1n2A_n=\frac{1}{n^2} with alternating signs (180° phase). Sawtooth-ish uses An=1nA_n=\frac{1}{n} for all eight. Overall normalization is arbitrary.

03 / BEATS & CANCELLATION

Adding can subtract

sin⁡(2πf1t)+sin⁡(2πf2t)=2cos⁡ ⁣(π(f2−f1)t)⋅sin⁡ ⁣(π(f1+f2)t)fbeat=∣f2−f1∣\begin{gathered}\sin(2\pi f_1t)+\sin(2\pi f_2t)\\=2\cos\!\bigl(\pi(f_2-f_1)t\bigr)\\\qquad\cdot\sin\!\bigl(\pi(f_1+f_2)t\bigr)\\[8pt]f_{\mathrm{beat}}=|f_2-f_1|\end{gathered}

This identity assumes equal amplitudes and zero initial phases. The magnitude of the envelope swells at the beat frequency. For equal frequencies and amplitude A, the resultant amplitude is 2A∣cos⁡ ⁣(Δφ2)∣2A\left|\cos\!\left(\frac{\Delta\varphi}{2}\right)\right|: twice A at 0°, zero at 180°. Cancellation here is at one mixed signal, not everywhere in a room.

04 / DECIBELS

A ratio, not a volume unit

GdB=20log⁡10 ⁣(AoutAin)AoutAin=10GdB/20\begin{gathered}G_{\mathrm{dB}}=20\log_{10}\!\left(\frac{A_{\mathrm{out}}}{A_{\mathrm{in}}}\right)\\[10pt]\frac{A_{\mathrm{out}}}{A_{\mathrm{in}}}=10^{G_{\mathrm{dB}}/20}\end{gathered}

+6.02 dB doubles amplitude; −6.02 dB halves it. Power ratios use 10log⁡10 ⁣(P2P1)10\log_{10}\!\left(\frac{P_2}{P_1}\right). Doubling amplitude is not the same as doubling perceived loudness. Spectrum levels here reference signal amplitude 1, not calibrated sound-pressure level or your speaker volume.

05 / FREQUENCY RESPONSE

What an equalizer does

Y(f)=H(f)X(f)H(f)=∏j=16Hj(f)An′=An∣H(nf0)∣\begin{gathered}Y(f)=H(f)X(f)\\[6pt]H(f)=\prod_{j=1}^{6} H_j(f)\\[6pt]A'_n=A_n\left|H(nf_0)\right|\end{gathered}

Each filter changes amplitude and phase: φn′=φn+arg⁡H(nf0)\varphi'_n=\varphi_n+\arg H(nf_0). Cascaded filters multiply their complex responses, so their gains in dB add. The magenta line shows that combined response. The spectrum combines same-frequency components as phasors, so opposite phases cancel.

06 / THE DIGITAL FILTER

Inside each bell

H(z)=b0+b1z−1+b2z−2a0+a1z−1+a2z−2z=ei2πf/fs\begin{gathered}H(z)=\frac{b_0+b_1z^{-1}+b_2z^{-2}}{a_0+a_1z^{-1}+a_2z^{-2}}\\[10pt]z=e^{\mathrm{i}2\pi f/f_s}\end{gathered}

Six peaking biquads use Q = 1.2 and gains from −12 to +12 dB. This lab evaluates their steady-state complex response and applies it to every synthesized tone, including phase. It does not process microphone recordings. Before listening fₛ = 48 kHz; afterward it uses the device’s AudioContext sample rate.

Model notes, plot conventions & further reading

The waveform and line spectrum are calculated from the same sinusoidal model used for playback; they are not microphone measurements or a live FFT. The animation advances slowly for readability, while the time labels show the actual signal window. “Freeze plots” pauses only the visual time offset. Both waveform traces share one automatic scale. At wide time windows, the plot preserves each screen pixel’s minimum and maximum samples to show the envelope without hiding narrow peaks.

Playback starts only when you press Listen. Output volume is separate from the scientific plots. A conservative automatic headroom trim bounds the sum of component amplitudes before playback; it is shared between original and EQ paths for a useful bypass comparison. It can reduce overall playback level when you boost bands. Plots always show the untrimmed relative signal. Start with a low device volume.

Audio changes use a brief crossfade. Harmonic phases and the second-tone phase are relative to a common source start. No input recording, upload, account, or network connection is needed.

Filter equations: W3C Audio EQ Cookbook. Audio synthesis: Web Audio specification.