Aliasing: why a distortion plugin can sound cheap for no obvious reason

Two saturation plugins can be set identically and one of them sounds brittle in a way that is hard to name. Not brighter, not louder. Cheap. There is usually one cause, and it has nothing to do with the quality of the distortion itself.

Distortion works by adding harmonics above the note you played. That is the point of it. The problem is that digital audio has a ceiling, half the sample rate, and harmonics do not stop politely when they reach it.

What happens at the ceiling

At 44.1 kHz the ceiling, the Nyquist frequency, is 22.05 kHz. Any frequency generated above it cannot be represented. It does not vanish and it does not get filtered away. It folds back down into the audible range and lands somewhere else.

The folded copies are the whole problem. Real harmonics land at whole-number multiples of the note, which is why they reinforce it. Folded ones land wherever the arithmetic puts them, which is almost never a multiple of anything. They sit between the harmonics, unrelated to the key, and they move in the opposite direction to the note: play higher and the aliased content slides down.

That last detail is the giveaway. If a distorted synth line sounds fine in the lower octaves and turns metallic as it climbs, you are listening to aliasing, not to the distortion.

When it gets worse

Aliasing scales with how much you are asking for. More drive means more harmonics, higher up, so more of them land above the ceiling. Brighter sources have the same effect, because their harmonics start higher. Hard clipping is worse than soft, since a sharp corner in the waveform generates a much longer series of harmonics than a gentle bend.

So the worst case is a bright, high-pitched part through a hard clipper at heavy drive. That is also an extremely common thing to do to a lead synth.

Oversampling

The fix is to give the harmonics somewhere to go. The plugin temporarily raises the sample rate, does the distortion at the higher rate where the ceiling is much further away, filters off everything that would not fit at the original rate, and comes back down. That is what an oversampling or HQ switch does, and the numbers next to it are the multiplier: 2x, 4x, 8x.

It works well. Four times oversampling is enough to push aliasing close to the noise floor on most material. It is not free, though: the plugin does several times the arithmetic, so the CPU cost rises roughly in step with the multiplier, and the filters involved can add latency. That is why it is usually a switch rather than something left on permanently.

A working habit

Leave oversampling off while you write and turn it up when you commit. The CPU cost is real, and aliasing does not change your decisions about how much drive you want.

The exception is anything bright and high. On a lead, a hi-hat bus or a cymbal, turn it on before you judge the sound at all, because on that material the aliasing is loud enough to be the thing you are reacting to.

Producer’s note

If a plugin has no oversampling option and you suspect it is aliasing, you can test it without a spectrum analyser. Play a rising line through it and listen to the top of the range. Real harmonics rise with the note. Aliased ones fall. If something is descending while you play upward, that is aliasing, and no amount of EQ afterwards will separate it from the signal, because it is now sitting inside the audible range as ordinary audio.