What a wavetable actually is
A wavetable is one cycle of a waveform stored as a list of numbers. A 2048-sample table holds 2048 amplitude values describing the shape of a single loop, and that is the entire object. It has no pitch, no timbre and no character of its own until something plays it.
Playback rate is what creates pitch. If a table holds one cycle and the instrument emits it once per 128 samples, you get 375 Hz at 48 kHz. Emit the same table once per 256 samples and you get 187.5 Hz. The stored data is untouched; only the number of samples per cycle changed.
This is why a wavetable is not a sample. A sample is seconds of recorded audio and carries whatever pitch and timing it was recorded at. A table is a single cycle with no time information at all, which is what lets it produce any note in range without changing speed.
Table position is the timbre control
Moving through the table changes which partials you hear. Table designers sweep from a pure sine at one end to a bright, harmonically dense waveform at the other, usually keeping the fundamental steady so pitch stays clean while brightness moves.
- Position 0 to about 30 percent is typically near-sine and usable at any pitch.
- The middle of the table usually carries a few strong low partials and a few weak high ones.
- Above roughly 70 percent the harmonic count climbs quickly and the usable pitch range drops fast.
- Crossfading between two positions smooths the jump, at the cost of running two oscillators per voice.
Band-limiting is what forces region selection
Every digital system has a Nyquist limit, half the sample rate. At 48 kHz that is 24 kHz. A table position loaded with partial content up to 20 kHz cannot be played at A6, because the fundamental alone sits near 1760 Hz and the partials above it are already past the limit.
When a partial lands above Nyquist it does not disappear. It folds back into the audible band at a frequency equal to its absolute difference from a multiple of the sample rate, which is aliasing. The instrument has two bad options: play the bright position and get aliased garbage, or refuse to play it.
| Sample rate | Nyquist limit | Practical bright-table ceiling |
|---|---|---|
| 44.1 kHz | 22.05 kHz | Roughly F5 and below |
| 48 kHz | 24 kHz | Roughly A5 and below |
| 96 kHz | 48 kHz | Roughly F7 and below |
| 192 kHz | 96 kHz | Roughly two octaves higher again |
This is why instruments ship multiple tables sliced by region. A single global table forces the whole instrument to the most restrictive region, so the bright half is only reachable at high sample rates. Raising the project rate to 96 kHz genuinely extends the range, and that is a measurable effect rather than a marketing one.
Supersaw and the cost of unison
The other common modern oscillator trick is not a table technique at all. Unison stacks several detuned copies of one voice, typically five to seven, spread by a few cents either side of the centre pitch. The slow beating between them thickens the waveform on playback without adding any low frequency.
The trade-offs are concrete. Seven oscillators per note on a sixteen-voice polyphonic instrument is 112 oscillators to process, so CPU cost climbs steeply. The detuned copies are often spread across the stereo field, which narrows as they converge in the middle and can collapse in mono. And the beating is a chorus effect that loses its thickness when the notes fall far apart in pitch.
What is measurable, and what is not
Everything above is testable with a spectrum analyser. Cycle count, playback rate, Nyquist ceiling, oscillator load and stereo width all produce numbers. If a plugin claims a brighter wavetable, that is a spectrum reading. If it claims a more musical one, that is a preference.
Synth marketing does attach more than that. Scale names, aura descriptions and claims about a table position producing a specific feeling are not physics and do not survive a blind comparison any better than any other unverifiable description. What does survive is the arithmetic: one cycle, a rate, a position, and a hard ceiling above which the content cannot be played cleanly.
The bottom line
A wavetable is one cycle stored as numbers, and the whole instrument is a two-axis control surface: rate picks pitch, position picks timbre. Band-limiting is not a detail on top of that, it is the constraint that makes the top octave playable. Supersaw unison is the other common trick, and it trades CPU and mono width for thickness rather than adding a new note.
Hear timbre without pitch moving
432Hz MASTER shifts a whole production by one ratio, so you can audition oscillator characters against a bed that never slides underneath them.
Frequently asked questions
What is actually inside a wavetable?
A single cycle of a waveform stored as an array of sample values, usually 256 to 4096 points. Nothing else. The apparent complexity comes from stepping through that array at the rate the note requires.
Why does a wavetable change sound when I move through it?
Each stored cycle contains a different set of partial amplitudes, so position is selecting a different spectrum. Two positions can share a fundamental and still differ completely in brightness.
Does a wavetable replace additive synthesis?
No. Additive builds partials numerically one at a time. A wavetable can represent an equivalent spectrum in far less memory, but it is a fixed snapshot, not a live set of partial controls.
Why does my high-register wavetable sound thin or wrong?
Usually because the top of the table contains partials above the Nyquist limit for that pitch, and the instrument is either folding them or dropping them. Either way you are losing content, and the fix is to use a different table position.