Wavetables, and why old ones sound old
A wavetable is a row of single cycles and a knob that slides along it. From the PPG Wave to Serum, the same table has sounded different on every machine that read it, for reasons that are now a choice.
What a wavetable is
An analog oscillator gives you a handful of shapes: saw, square, triangle, perhaps a pulse you can narrow. A wavetable oscillator gives you a row of shapes, each one a single cycle long, and a Position control that slides along the row. At one end, say, a soft sine. At the other, something with teeth. In between, every shape the designer put there, in order. Sweep the position with an envelope and the timbre of one note travels the whole row.
That's the entire idea. It isn't sampling: there's no recording of a note, only a single cycle, repeated at whatever pitch you play. And it isn't a filter, although it's often used for the same job: the harmonics change because the shape changes, not because something is being taken away. Tonnau's tables hold up to 256 of those cycles, each 2,048 samples long, which is what the modern end of this history settled on.
Where it came from
The name on the idea is Wolfgang Palm. His company, PPG, in Hamburg, put out the Wavecomputer 360 in 1978: digital oscillators reading tables of waves, with no analog filter after them. It was the first commercial wavetable synth, and almost nobody bought it. The reviews of the time found it harsh. In 1981 Palm put an analog low-pass filter and amplifier after the digital oscillators and called the result the PPG Wave 2. That was the breakthrough. The Wave 2.2 followed in 1982, with the Waveterm computer alongside it that let you make your own tables, and the Wave 2.3 a couple of years after that. If you've heard a Thomas Dolby or Art of Noise record from those years, you've heard a Wave: a glassy, moving, slightly fizzy digital tone with a fat analog filter underneath it.
PPG closed in 1987, undercut by cheaper synths that played back samples. Its distributor, Wolfgang Düren, founded Waldorf, and the first Waldorf product was the Microwave of 1989: the PPG wavetables in a rack box, with analog filters still. The Wave keyboard came in the early nineties as the expensive flagship. Then in 1997 the Microwave II went all-digital, filters included, and the tone changed again. The Blofeld, from the late 2000s, put that engine in a desktop box for a few hundred pounds, which is where a lot of people met wavetables for the first time.
Other instruments were circling the same idea. Korg's DW-8000 (1985) had digital single-cycle waves through an analog filter, but no row to sweep. Sequential's Prophet VS (1986) crossfaded between four waves with a joystick, which it called vector synthesis. Korg's Wavestation (1990) stepped through sequences of waves in time. All cousins, none quite wavetable synthesis in Palm's sense, though Tonnau's joystick owes something to the VS.
Then software. Native Instruments' Massive (2006) made wavetables a bass-music staple. Xfer's Serum (2014) added the thing that changed how people think about them: you can see the table, draw into it, and import your own, and the oscillators are clean right to the top of the keyboard. Matt Tytel's Vital (2020) did much the same for free. Most wavetable synths made since, Tonnau included, use Serum's conventions: 2,048 samples a cycle, up to 256 cycles a table, position as the main modulation target.
Why old wavetables sound old
The interesting part of the history isn't the dates. It's that the same table sounds different on every one of those machines, and the reasons are all specific, all technical, and all now a matter of choice. These are the big ones.
Bit depth
The PPG tables were 8-bit: 256 possible values per sample. Quantising a smooth wave that coarsely adds a layer of noise that follows the signal, a fine grit on everything. The Microwave kept the 8-bit tables; later machines went to 16 bits and more, and the grit went away.
No band-limiting, so aliasing
A bright wave has harmonics far above the fundamental. Play it high enough and some of those harmonics would sit above half the sample rate, where a digital system can't represent them. They don't vanish. They fold back down into the audible range at pitches that have nothing to do with the note you played. That's aliasing: an inharmonic shimmer that gets worse the higher and brighter you go. The PPG did nothing about it, because nothing could be done at the time, and it's a large part of what people mean by "the PPG sound". Modern oscillators prepare a separate, pre-filtered copy of each wave for each octave, so the harmonics that would alias are never there. Clean to the top. Also, at the top, slightly bland.
Stepping between waves
A PPG table had 64 waves, and Position chose one of them. Sweep it and you hear 64 steps, with a small click at each boundary where the shape jumps. Later machines crossfade between neighbouring waves, so a sweep is continuous. Some sounds want the steps: a slow stepped sweep is a rhythm as well as a timbre, and it's a hard thing to fake on a synth that only glides.
Sample rate, and what comes after the oscillator
Early machines ran their oscillators at lower rates than we'd use now, then handed the result to analog filters and amplifiers that were themselves not perfectly clean. The Microwave II moved the filters into software, and the thing many Microwave owners missed wasn't the tables, it was the analog stage. A digital filter can be written to behave like an old one, or like nothing that ever existed. What it can't do by accident is drift.
Interpolation inside the cycle
Between the samples of one cycle, the oscillator has to guess. The cheapest guess is to hold the last value. Better ones draw a line, or a curve. Each guess has its own noise, and it changes with pitch. Nobody chose these things for their sound in 1981; they chose them because the chip could do nothing else. Forty years later, they're flavours.
Era
Every digital synth I've admired dealt with this history in one of two ways: ignore it and be clean, or pick one old machine and emulate it. I wanted a third way, which is why every oscillator in Tonnau has an Era knob. At its default it's today's technology, clean, and the warps run at twice the sample rate with their corners band-limited. Turn it back and the engine gives up its modern habits a stage at a time, so the same table sounds progressively more like it would have on the machines above. From about a third of the way back, a warped oscillator stops oversampling altogether, because by then the aliasing is the point.
There's an Era knob on each oscillator, so one can be a modern sine while another is a fizzing early-eighties saw, a global Era Offset that shifts all three together, and a separate Filter Era for the filters. Era is a matrix destination too, so an envelope can let a note start clean and age as it rings, or a macro can take a whole patch back a decade under your hand.
I'll write up exactly what each stop on the knob changes when the manual's done. The short version is that a wavetable has never been just the table. It's the table plus the machine reading it, and Era is the machine.
Scroll one back.
Tonnau WT-3 is free for Apple Silicon Macs, as AU, VST3 and Standalone. Thirty-eight built-in tables, a slot to load your own to each oscillator and an Era knob for each of those.