Oversampling is an increasingly common function in most plugins, which increases the sampling rate of the signal it’s processing by a fixed multiple like 2 or 4. So if the sampling rate of your session is 48kHz, a 2x oversampling setting will make the sampling rate 96kHz.
The process of oversampling can be CPU intensive and can cause performance issues if too high of a rate is used.
Simply put, oversampling increases the maximum frequency your processors can handle and increases the accuracy with which the signal is encoded and processed.
This makes it a great option for reducing distortion and subtly increasing the clarity of your signal.
Oversampling reduces or completely gets rid of 3 forms of potential distortion a signal can have: aliasing, clipping, and quantization distortion. Although these forms of distortion are often mild and difficult to consciously hear, they’re often noticed when using a lot of processing or pushing a processor harder.
First, oversampling reduces aliasing distortion by allowing a processor a higher frequency at which to process the signal.
Second, It reduces clipping distortion by making the amplitude of the signal more accurately encoded, which if the signal is close to 0dB, can be the difference between clipping and not clipping.
And third, it reduces quantization distortion, or the difference between the actual signal and the encoded signal, by causing more accurate quantization.
If you’re using a lower sampling rate for your session, but you still want to use a fair deal of processing, it helps to use oversampling to reduce distortion. Oversampling should be used both in mixing and mastering sessions when either aggressive or a lot of processing is being used.
For example, oversampling causes more accurate encoding and quantization of the signal’s amplitude - so if you’re using a limiter, and you’re pushing the signal up to 0dB, use oversampling to make the signal encoded more accurately.
This may cause a peak that clipped to actually not clip, in turn reducing clipping distortion. In this instance, the more you push the signal up against 0dB, the higher the oversampling you’ll want to use.
Aliasing distortion occurs when the signal fed into a processor has exceeded the maximum frequency that that processor can handle or accurately represent as determined by the sampling rate. When it does occur, harmonics are reflected down the frequency spectrum, which causes both unpleasant harmonic and phase distortion.
For example, say my sampling rate is 48kHz. Using the Nyquist theorem we know that the maximum frequency which can be represented is 24kHz. So if the signal goes above this, say to 32kHz, we’ve gone above this by 8000Hz.
If we reflect this value back down the spectrum, we know our distortion will be 8kHz less than 24kHz. In other words, the distortion will show up at 16kHz.
You’re probably wondering if you even need to be concerned about aliasing distortion, considering most signals only go up to 20kHz, and a 44.1kHz sampling rate covers that frequency; however, it may still be an issue. If you’re processing high frequencies with distortion plugins, you’ll likely cause aliasing distortion.
Say you’re distorting your high-frequency range, and your distortion starts at 15kHz. Now, let’s say this distortion causes a 2nd order harmonic.
This harmonic will be at 30kHz, far above the range that can be handled by our sampling rate - resulting in aliasing.
Similarly, let’s say you’re using aggressive limiting, and clipping occurs to your high frequencies. This too can create harmonics, resulting in aliasing.
If you want to more accurately emulate analog equipment, you’ll most likely need to use oversampling to reduce aliasing. Although aliasing is subtle, it can add up amongst multiple signals and processors, resulting in a distinctly digital sound, often characterized as sounding metallic or brittle.
If you’re using a form of analog emulation, check to see if oversampling is an option - it’ll greatly reduce the impact of aliasing, and result in a less digital sound.
This is why a lot of developers, especially analog emulation developers, are beginning to emphasize oversampling in their plugins.
For example, you’ll notice that FabFilter uses oversampling for their analog emulation plugins.
The same could be said for the indie-developer analog obsession.
Admittedly, oversampling seems like a pretty insignificant function - it doesn’t do something super noticeable, especially when observed on one plugin; despite this, it definitely plays a big part in creating clean mixes and masters. Although you may not notice it at first, oversampling remedies distortion in a big way.
If you value reducing clipping distortion, aliasing distortion, and to a lesser extent, lowering quantization distortion, you should definitely use oversampling.
Additionally, if you want to have accurate analog emulation without the negative impact of digital sounding aliasing distortion, use oversampling.
The higher your native sampling rate, the less you’ll need to use it, but if you’re working in 44.1 or 48kHz sampling rates, you should definitely utilize it when available.