Skip to content

Why aliasing is a ghost frequency problem

That harsh digital top end is often aliasing: saturators and clippers create harmonics above the Nyquist limit, and they fold back into your mix as ghost frequencies.

Updated · 5 min read Lesson 4 of 14 in Audio Science

In short

  1. A digital system can only store frequencies up to half its sample rate, the Nyquist limit.
  2. Saturators and clippers create harmonics above that limit. They fold back down as tones unrelated to your key.
  3. Oversample nonlinear plugins while mixing and mastering, and leave it off while tracking.
In this lesson (8)

Saturation that turns the top end metallic

You load a plugin to warm up a vocal or add grit to a bass line. You push the input gain. Instead of the smooth density you hear on tape, the top end turns metallic. A cold glare settles over the track and it starts to sound thin. What you are hearing is .

The plugin is generating the session cannot represent. Instead of disappearing, those frequencies fold back into the audible range as ghost tones.

Why it matters

Harmonics with nowhere to go

A digital clipper or saturator creates new harmonics. Start with a 10 kHz tone in a 48 kHz session and push the saturator. A harmonic at 20 kHz is fine. The harmonic at 30 kHz is not: the highest frequency a 48 kHz session can hold is 24 kHz, so 30 kHz cannot exist in your DAW.

It does not vanish. It folds back below the limit and lands at 18 kHz. That folded tone has no musical relationship to the note or the key of the song. Stack several saturators across the drum bus and the mix bus and these ghost frequencies pile up, smearing cymbals and vocals.

Figure 1. A 10 kHz tone through a saturator in a 48 kHz session. The 3rd harmonic at 30 kHz and the 5th at 50 kHz cannot be stored, so they land at 18 kHz and 2 kHz, frequencies that have nothing to do with the note.

The science

The mathematics of foldback

The ceiling of a digital system is the , exactly half the :

fN=fs2f_N = \frac{f_s}{2}

A frequency above fNf_N is stored as a lower one. For a frequency between fNf_N and fsf_s, the alias lands at:

falias=∣fs−f∣f_{\text{alias}} = \left| f_s - f \right|

At 48 kHz the Nyquist limit is 24 kHz, so a harmonic at 35 kHz lands at 13 kHz. Higher harmonics keep folding back and forth across the range. In general the alias sits at ∣f−kfs∣\left| f - k f_s \right|, where kk is the whole number that brings the result below fNf_N. The harder you drive a saturator, the more high harmonics it makes, and the more of them fold back into the range you can hear.

Figure 2. A wave with nine cycles, measured only ten times. The dots fit the fast wave, and they fit a slow one-cycle wave just as well. The converter keeps only the dots, so the slow wave is what plays back.

Listen

Hear a frequency fold back down

A tone sweeps up from 500 Hz to 15 kHz. At a 16 kHz sample rate everything above 8 kHz is stored as a falling ghost tone, the same fold a saturator causes when its harmonics pass the Nyquist limit.

Loading demo…
100%

Start with your speakers or headphones low.

Try it

Isolate alias noise

You can see aliasing with a test tone and a .

0 of 6 steps done

Common mistake

Oversampling everything, or nothing

You turn on 8x or 16x on every plugin in the session. Oversampling runs the process at a multiple of the sample rate inside the plugin, which pushes the Nyquist limit far above the audible range, then filters off the extra harmonics before converting back. It costs CPU, and many plugins add while it is on. Across forty channels during tracking, that can mean dropouts and a monitor delay the performer feels.

The opposite mistake is assuming analog-modelled plugins never alias. Some do not oversample at all, or only do it when you export. Test the ones you rely on with the experiment above.

Takeaway

Oversample where it counts

Use oversampling selectively. Keep it off while tracking so monitoring stays immediate. Turn it on while mixing and mastering, and focus on nonlinear plugins: clippers, saturators and . 4x is often enough. If a plugin offers no oversampling, render that track from a 96 kHz session to print a cleaner file, then bring it back into your main session.

3 questions

Check yourself

  1. In a 48 kHz session, a saturator creates a harmonic at 30 kHz. Where does it end up?
  2. Why is aliasing worse than ordinary harmonic distortion?
  3. What does oversampling do inside a plugin?

Sources

Show all 2

Next lesson · 5 of 14

Fourier shows the ingredientsAny sound can be described as a sum of sine waves. See how harmonics build a waveform, how to hear them, and what an analyzer leaves out.

Previous: Where 32-bit float headroom ends

Portrait of Virzy Guns

Written by Virzy Guns

Songwriter and producer, founder of Virzy Guns Production. Now building HealingWave.