Why aliasing is a ghost frequency problem

You load up a saturation plugin to warm up a vocal or add grit to a bass line. You push the input gain. Instead of the smooth, pleasing density of analog tape, the high frequencies start to sound metallic. A cold, static glare envelopes the track, making it sound thin. This is not the sound of digital audio being inherently bad. It is the sound of aliasing.

Your plugins are generating harmonics that the digital system is mathematically incapable of representing. Instead of disappearing, those frequencies fold backward into the audible spectrum, creating digital ghost signals.

When you push a digital clipper or saturator, it generates new harmonic content. If you start with a 10 kHz tone in a 44.1 kHz session, the second harmonic is at 20 kHz. That is fine. But the third harmonic is at 30 kHz. Since the maximum frequency a 44.1 kHz session can represent is 22.05 kHz, that 30 kHz harmonic cannot exist in your DAW.

Instead of being ignored, it bounces off the Nyquist ceiling and folds back into the audible range at 14.1 kHz. This folded frequency is not harmonically related to the original chord or key of your song. It is discordant noise. If you stack multiple saturators across your mix bus and drum buses, these ghost frequencies pile up, ruining the clarity of cymbals and vocals.

The boundary of a digital system is the Nyquist limit, which is exactly half of your sample rate.

$$f_N = \frac{f_s}{2}$$

When a nonlinear process generates a frequency $f$ that is higher than $f_N$, the system folds it back. You can calculate the resulting alias frequency using this formula:

$$f_{alias} = |f_s - f|$$

If you are working at a sample rate of 48 kHz, your Nyquist limit is 24 kHz. If a saturator generates a harmonic at 35 kHz, the folded frequency lands at 13 kHz. The further you push the saturation, the more high-frequency harmonics you generate, and the more alias noise leaks back to corrupt your midrange.

You can isolate and view aliasing using a simple sine wave generator and a spectrum analyzer.

A common mistake is turning on 8x or 16x oversampling on every plugin in the session. Oversampling works by temporarily multiplying the sample rate inside the plugin. This pushes the Nyquist limit far above the audible range, allowing the plugin to filter out the high-frequency harmonics before they can alias.

However, oversampling is CPU-heavy. If you run it on forty channels during tracking, you will introduce massive monitor latency and cause audio dropouts.

Another mistake is believing that analog modeling plugins do not alias. Many vintage emulations do not have built-in oversampling, meaning they leave digital fingerprints all over your transients.

Use oversampling selectively. Keep it turned off while tracking to ensure zero-latency monitoring. Enable oversampling (typically 4x is sufficient) only during the mixing and mastering stages, focusing on nonlinear plugins like clippers, saturators, and limiters. If a plugin does not offer internal oversampling, you can render that specific track at 96 kHz to print a clean file, then bring it back into your main session.