education lab · 02
Why digital distortion can create frequencies that were never in the original signal — and how sample rate, Nyquist frequency and oversampling keep them under control.
Aliasing is a digital audio artifact that occurs when a signal contains frequency components above the Nyquist frequency — half of the sampling rate. Instead of disappearing, those frequencies are represented as lower frequencies inside the available spectrum. The result is called frequency folding or aliasing.
At a 48 kHz sample rate, the Nyquist frequency is 24 kHz. A clean 1 kHz sine wave is no problem: it is far below the 24 kHz limit. But if a process creates a new component at 30 kHz, the digital system cannot represent 30 kHz directly. That component folds back and appears at 18 kHz. Your DAW now contains an 18 kHz tone that was not present in the original signal.
This distinction is important: aliasing is not simply “too much high end.” It is a mapping error caused by sampling. The unwanted frequency is effectively given the identity of another, lower frequency. Once aliasing has been created, you cannot remove it perfectly with a normal EQ because the aliased component now lives inside the valid audio band.
The most important source in modern music production is nonlinear processing. Saturation, clipping, distortion and waveshaping change the waveform shape. A changed waveform contains additional harmonics. With enough drive, those harmonics can extend far beyond the original audio bandwidth.
For example, a 10 kHz sine wave processed by a nonlinear waveshaper can produce harmonics at 20 kHz, 30 kHz, 40 kHz and beyond. At 44.1 kHz, Nyquist is only 22.05 kHz. Everything above that boundary is at risk of folding back into the audible spectrum.
Linear processors such as a conventional static EQ or gain stage do not create new harmonics by themselves. They can move or attenuate existing frequency components, but they do not normally generate an entirely new harmonic series. This is why aliasing is especially relevant to distortion, saturation, clipping, waveshaping and other nonlinear audio processes.
EQ, gain, panning. Moves or attenuates what is already there — no new harmonic series, nothing new that could exceed Nyquist.
Saturation, clipping, distortion, waveshaping. Reshapes the waveform and therefore creates harmonics — the ones that can fold back.
Aliasing can sound like a subtle increase in high-frequency roughness, metallic texture, digital grit or loss of clarity. In some processors it is barely noticeable; in others it can become a strong part of the sound. The problem is particularly easy to hear on exposed cymbals, bright synths, vocals, distorted guitars and mastering chains.
Not all aliasing is automatically bad. Some classic digital distortion sounds deliberately exploit it, and some producers like the character it adds.
Push the tone past the line and watch it come back
education lab · 02
Play a tone above half the sample rate and watch it come back as a different tone — one you never played.
amplitude 1.03 cycles shownideal reconstruction
The dashed coral line is Nyquist, half the sample rate. Amber stems are the harmonics the waveform actually contains, drawn wherever they fall — including beyond the line.
Everything beyond Nyquist cannot be represented. It folds back at the sample rate and lands somewhere inside the band: those are the coral stems. The two time plots show the same thing without any maths — above the line, the curve that comes back is not the curve you played, even though it passes through every single sample point.
The sample rate determines how many samples per second represent the audio signal. Common rates are 44.1 kHz, 48 kHz, 88.2 kHz and 96 kHz. A higher sample rate moves the Nyquist frequency upward, giving nonlinear processors more room to generate harmonics before they fold back.
At 44.1 kHz, Nyquist is 22.05 kHz. At 48 kHz it is 24 kHz. At 96 kHz it is 48 kHz. Higher sample rates therefore reduce the amount of newly generated harmonic content that immediately lands above the Nyquist boundary — but they do not automatically make a nonlinear processor alias-free.
The Nyquist frequency is half the sample rate. It is the highest frequency that can theoretically be represented without aliasing in a sampled system. A 48 kHz system therefore has a Nyquist frequency of 24 kHz.
Think of Nyquist as a ceiling rather than a sonic brick wall. Frequencies above it cannot be represented as their original frequencies. If they enter a sampled system without being filtered first, they can fold into the range below Nyquist.
When a frequency crosses the Nyquist boundary, it is mirrored back into the digital audio spectrum. The simplest example is a 48 kHz system: a 25 kHz component appears at 23 kHz; 30 kHz appears at 18 kHz; 40 kHz appears at 8 kHz.
Mathematically, the aliased frequency can be described as the distance of the original frequency from the nearest multiple of the sample rate, expressed inside the 0-to-Nyquist range. You do not need the equation to mix music, but understanding the mirror behavior makes the phenomenon much easier to predict.
A nonlinear processor creates harmonics at integer multiples of the input frequency. A 5 kHz sine wave can therefore generate components at 10 kHz, 15 kHz, 20 kHz, 25 kHz and so on. The higher the harmonic order, the more likely it is to cross Nyquist.
This is why aliasing becomes increasingly relevant with aggressive saturation and clipping. A gentle nonlinear curve may generate mostly low-order harmonics that remain safely below Nyquist. A hard clipper can create a much denser spectrum and push substantial energy above it.
Oversampling temporarily increases the internal sample rate of a digital processor. A plugin might run its nonlinear algorithm at 2×, 4×, 8× or higher than the session sample rate. The extra bandwidth gives the processor space to generate harmonics above the original Nyquist frequency without immediately folding them into the audible band.
A proper oversampling stage also uses filtering when returning to the original sample rate. This removes out-of-band content before downsampling. In practical terms: process higher, filter, then return to the session rate. That is the core idea behind anti-aliasing in nonlinear audio plugins.
If a distortion, saturation or waveshaper plugin offers oversampling, higher settings generally reduce aliasing at the cost of additional CPU usage and potentially more latency. 2× can already help; 4× is a useful practical setting for many situations; 8× or higher can be valuable for aggressive nonlinear processing and critical mastering work.
Do not treat the highest setting as automatically “better.” Compare at matched level and listen to the actual material. Some processors change their character slightly when oversampled, and CPU efficiency matters in large sessions.
A pre-filter can remove very high-frequency content before a nonlinear stage. This can reduce the amount of energy available to generate high-order harmonics. It is especially useful when a bright signal is about to hit aggressive saturation or clipping.
Filtering before distortion is also a creative decision: whatever you remove before the nonlinear stage cannot be distorted and regenerated in the same way. A low-pass filter before a waveshaper can therefore change both the amount and the character of the resulting harmonics.
A post-filter can tame excessive high-frequency content after nonlinear processing. It can make a distorted signal smoother, but it cannot reliably undo aliasing that has already folded into the audible range. An EQ cannot tell whether an 18 kHz component came from the original source or was created by a 30 kHz harmonic and folded down.
For that reason, post-filtering is best viewed as tone shaping, while oversampling and anti-aliasing design address the cause of the problem inside the nonlinear processor.
More drive usually means more nonlinear behavior, more harmonics and a greater chance of generating content above Nyquist. This does not mean high drive is wrong. It means that aggressive settings make oversampling more important if you want the result to remain clean and controlled.
Working at 96 kHz can reduce aliasing because Nyquist moves to 48 kHz. However, the sample rate of the DAW is only one part of the equation. A well-designed plugin with internal oversampling can perform far better than a poorly designed nonlinear algorithm simply running at a higher session rate.
Think of the nonlinear processor as a harmonic generator. Every time you bend, clip, saturate or otherwise reshape a waveform, you are asking the processor to create additional frequency content. At 1× sample rate, the available frequency space ends at Nyquist. Oversampling moves that temporary ceiling upward, gives the harmonics room to exist, and filters them before the signal is returned to the original sample rate.
Aliasing matters whenever a process is nonlinear: saturation, distortion, clipping, waveshaping, bit crushing, some compressors with nonlinear stages, analog-modeling plugins and other processors that intentionally reshape the waveform. It is particularly worth checking when processing bright material, high-frequency instruments, aggressive distortion or a final mix/master.
For transparent processing, choose a plugin with a well-designed anti-aliasing strategy and use oversampling where appropriate. For creative distortion, decide whether aliasing is part of the sound rather than assuming it is automatically a defect.
The most important practical rule is simple: do not judge aliasing by the amount of processing alone. A subtle saturation stage can alias significantly if it generates high-order harmonics without adequate bandwidth, while a carefully designed nonlinear processor can remain remarkably clean even when driven hard.
Aliasing is frequency folding. Nonlinear processing creates new harmonics. The Nyquist frequency defines the upper limit of what a digital system can represent. Harmonics that exceed that limit can fold back into the audible spectrum. Oversampling raises the internal Nyquist limit and allows those harmonics to be filtered before they become audible aliases.
Once you understand that chain, aliasing stops being a mysterious “digital problem” and becomes a predictable part of digital audio engineering.