Your Saturator Is Playing Notes That Aren't in the Song

You've got a bright synth lead that sounds a bit polite, so you throw a saturator on it and push the drive. For a moment it's great, thicker and with more attitude. Then you push a little further and the top end turns to sand. It's grainy in a way that doesn't sound like tape or a cranked amp. It sounds faintly out of tune, like there's a second, smaller synth hiding behind the first one, playing something slightly wrong.

There is. Your saturator is generating frequencies that the session can't hold, and they're folding back down into the audible range as tones that have nothing to do with the note you played. That's aliasing, and once you know how it works you can calculate where it lands and get most of it out of your mix.

The Ceiling You Can't See

Every digital session has a hard frequency ceiling called the Nyquist frequency, which is half the sample rate. At 44.1 kHz that ceiling sits at 22.05 kHz. At 48 kHz it's 24 kHz. Nothing above it can exist in the file.

Normally that's fine, because your recordings were filtered on the way in and nothing above 20 kHz matters to your ears anyway. The problem is that a saturator doesn't just change the level of what you give it. It creates new frequencies. Drive a sine wave into a clipper and out come harmonics at 3x, 5x, 7x the input frequency and on up the series. Hard clipping keeps generating them a long way up.

When a plugin generates a harmonic above Nyquist, the harmonic doesn't politely disappear. It reflects off the ceiling and lands somewhere below it. For a frequency between Nyquist and the sample rate, the alias appears at the sample rate minus that frequency. Push even higher and it keeps folding back and forth like a ball bouncing between two walls.

Where the Junk Actually Lands

Let's run real numbers. Session at 44.1 kHz, a 3 kHz tone (think of the bright upper edge of a lead or a cymbal-heavy hat), driven hard into a symmetrical clipper so it throws off odd harmonics.

The 3rd harmonic is at 9 kHz. The 5th is at 15 kHz. The 7th is at 21 kHz, just under the ceiling. All legit, all musically related to the source.

The 9th harmonic should be at 27 kHz. That's above Nyquist, so it folds to 44.1 minus 27, which is 17.1 kHz.

The 11th should be at 33 kHz. It lands at 11.1 kHz.

The 13th should be at 39 kHz. It lands at 5.1 kHz.

The 15th should be at 45 kHz, which is past the sample rate itself. It folds again and lands at 900 Hz.

Read that last one again. A harmonic that should sit way above anything a human can hear ends up at 900 Hz, right in the middle of your mix, and it isn't a multiple of 3 kHz. It doesn't belong to the note. That's why aliasing doesn't sound like extra brightness. It sounds like dirt, ghost tones, a slight detuned ring. The higher the source and the harder you drive, the more of these strays you get.

Real harmonics get quieter the higher they go, so that 15th-harmonic alias is a lot lower in level than the fundamental. But stack up a dense, bright mix with a few saturators, a clipper on the drum bus and a limiter on the master, and these little strays add up into the grainy, slightly smeared top end that makes a mix feel cheaper than it should.

See It for Yourself

You don't have to take the math on faith. This takes five minutes.

Put a sine sweep on a track. Either load a sweep audio file or automate a test oscillator to glide from about 1 kHz up to 15 kHz over ten seconds. Put your saturator after it with oversampling off and the drive up high. After that, put a spectrum analyzer with a fast, detailed display (Voxengo SPAN is free and works fine).

Play it and watch. The real harmonics all move up together as the sweep rises, nice parallel lines climbing to the right. The aliases give themselves away: some of them slide downward while your sweep goes up, and others cross over the real harmonics at angles that make no sense. Those are the reflections. Now close your eyes and just listen. That descending whistle you can hear under the sweep is the sound you've been calling "harsh."

Then switch oversampling on and play it again. Most of those wrong-way lines vanish or drop way down into the noise.

What Oversampling Actually Does

Oversampling is a simple idea. The plugin temporarily raises the sample rate inside itself, say to 4x or 8x your session rate. Now the ceiling is much higher, so the harmonics your saturator creates have room to exist without folding. Before handing the audio back, the plugin filters out everything above the original Nyquist and drops back to your session rate. The harmonics that would have folded into your midrange get cut cleanly instead.

Lots of plugins have this, though it's often hidden. FabFilter Saturn 2 has a High Quality button, and its manual is refreshingly blunt about it: "Good" mode is 8x oversampling and enough in most cases, "Superb" mode is 32x for the absolute minimum of aliasing, at a real CPU cost. Ableton's Saturator and Dynamic Tube have a Hi-Quality mode tucked away in the right-click context menu, which a lot of Live users have never seen. Limiters like FabFilter Pro-L 2 have oversampling too, because fast gain changes and clipping on peaks create the same kind of out-of-band content.

It's not free. Oversampling eats CPU, often adds latency, and the filters that make it work can slightly change the very top of the spectrum. Some engineers leave it off on certain plugins because they like the extra grit. That's a valid choice, as long as it's a choice and not an accident.

When You Can Stop Worrying

Aliasing isn't a reason to panic every time you load a distortion plugin. It matters most when three things line up: a source with a lot of high-frequency energy, a lot of drive, and a nonlinear process with hard edges.

Low sources are mostly safe. Saturate a bass at 100 Hz and it would take the 221st harmonic to cross the 22.05 kHz ceiling of a 44.1 kHz session. By then the harmonic is so quiet it doesn't matter.

Gentle colour is mostly safe. A tape or console plugin adding a dB or two of warmth generates weak upper harmonics. Not much to fold.

Higher session rates help. At 96 kHz the ceiling moves to 48 kHz, giving every plugin in the session twice the headroom before anything folds. That's one of the few real arguments for running a mix at a high sample rate, and the benefit comes from cleaner processing, since none of us hear above 20 kHz anyway.

Where to switch oversampling on: saturators and clippers on bright sources (lead synths, vocals with a lot of air, hats, overheads, the drum bus), and limiters on the master when you're pushing hard. Where to leave it off to save CPU: gentle saturation on bass, kick and low pads.

A Listening Habit Worth Building

Next time a saturator makes something sound harsh, don't reach for an EQ cut at 5 kHz first. Flip oversampling on and listen again. If the harshness smooths out and the source sounds more like itself, that harshness was aliasing, and an EQ cut would only have dulled the good harmonics along with the bad ones.

The skill underneath all this is noticing small changes in distortion, the difference between grit that belongs and grit that doesn't. SoundGym's Distorted Reality game trains that exact comparison: two sounds, one with more distortion, pick it. It's a good way to get your ears to the point where the A/B on the oversampling button becomes obvious instead of a coin toss.

Your level meters won't flag any of this, and no plugin pops up a warning when its harmonics start folding. The analyzer test above and a quick oversampling A/B are the only alarms you get, so make them part of how you set up any bright, driven sound.

---