Music Production
Aliasing
Aliasing occurs when different frequencies become indistinguishable after sampling. In ordinary digital audio, content above the usable sampling band can appear as unwanted lower-frequency components that were not present in the original sound.
Digital audio measures a signal at regularly spaced times. If those measurements cannot distinguish a fast oscillation from a slower one, the faster component appears under a different frequency identity: an alias. This is a sampling problem, distinct from the level-rounding error of audio quantization.
A numerical example
At a sample rate of 48 kHz, the Nyquist frequency is 24 kHz, half the sampling rate. For reliable reconstruction of ordinary baseband audio, the signal bandwidth must stay below that limit. Practical filters also need a transition region.
If an ideal sampler receives an unfiltered 30 kHz tone at 48 kHz, it aliases to 18 kHz: 48 minus 30. This is a paper example, not a recommendation to generate ultrasonic tones. Raising the stored bit depth would not fix it, because the problem is which moments were sampled, not how many amplitude values are available.
Prevent it before it folds
An anti-alias filter attenuates unwanted high frequencies before analog-to-digital sampling. Digital downsampling also needs suitable filtering before reducing the rate. Once unwanted content has folded into the wanted band, an ordinary filter cannot generally separate it from legitimate audio at the same frequencies.
Aliasing can arise inside a DAW too. Nonlinear processing, such as distortion or fast limiting, can generate new high-frequency components. Oversampling runs a processor at a higher internal rate with appropriate filtering, reducing the risk when returning to the project rate. It costs processing power and is not a promise to remove all artifacts or repair aliases already present.
Recognize the tradeoff
Some sound-design effects deliberately use aliasing for a rough digital color. For a clean workflow, compare a processor's oversampling options at matched loudness and inspect its documentation instead of assuming the largest setting is always best.
Common questions
Will more bits prevent aliasing?
No. Bit depth concerns amplitude representation. Aliasing depends on the relationship between signal frequencies, sampling rate and filtering.
Can I remove aliasing by filtering afterward?
Not reliably once it overlaps wanted frequencies. Prevention before sampling or downsampling is different from filtering a signal that already contains aliases.