Oversampling is the use of a sampling frequency higher than 44.1 kHz. There are many myths surrounding this concept! Is 96 better than 44.1? Or maybe 192 at once? Some claim to hear the difference, others work in 44.1 and don’t care. So what is oversampling really for? When is it justified? Let’s figure it out!
Oversampling is the use of a sampling frequency higher than 44.1 kHz.
A sound wave or its analog representation looks like this:
Audio wave in analog form
As you can see, it is continuous and has a smooth shape. This means that for any arbitrarily taken point, there is an amplitude value, and the number of such points (and, accordingly, the values of the wave amplitude) is infinite.
A digital audio wave looks different.
Sampling
When digitizing, the wave is divided into equal segments, for each of which the amplitude value is determined. The frequency of such amplitude measurements determines their number per unit of time and is called the sampling frequency. For example, with a selected frequency of 48 kHz, there will be 48,000 amplitude values for one second.
Check Out: What is Non Oversampling
Analog and digital audio wave
From the figure, it is clear that the digital wave will always strive to its analog prototype, but will never repeat it.
But there is good news! It is possible to restore an audio signal from digital to analog without loss or distortion, you just need to meet certain conditions:
the sampling frequency should be at least twice the maximum oscillation frequency in the signal (this is what Kotelnikov’s theorem tells us)
the signal itself should be limited in the spectrum and not go beyond half the sampling frequency (the so-called Nyquist frequency)
It turns out that the audible range of up to 20 kHz can be converted to digital and back without losses with a sampling frequency of 40 kHz. But this is theory. In practice, it is impossible to ideally filter the signal by limiting it to 20 kilohertz – any filter has a cutoff steepness.
If part of the signal goes beyond the Nyquist frequency, it will be digitized with an error (the so-called aliasing effect), and this error will lead to audible distortions. Therefore, the sampling frequency is chosen with a reserve – this allows you to avoid distortions in the audible part of the spectrum. This is how the 44.1 kHz standard appeared.
Now comes the fun part!
In theory, 44.1 kHz is enough for the entire audible spectrum to be recorded and reproduced losslessly in a digital environment.
So why do we need oversampling then?
In essence, it simply increases the frequency range far beyond the audible. Sounds pretty useless) However, in reality, there are situations when oversampling is justified, and sometimes even necessary!
not all AD-DA converters are equally good) The above is just a theory. In practice, anti-alias filters and the clock generator (the one that generates the sampling frequency) can be far from perfect, then there is a high probability of not quite ideal operation of the converter in 44.1. In such a situation, oversampling can improve the result.
- Not all AD-DA converters are equally good) The above is just a theory. In practice, anti-alias filters and the clock generator (the one that generates the sampling frequency) can be far from perfect, then there is a high probability of not quite ideal operation of the converter in 44.1. In such a situation, oversampling can improve the result.
- In the digital environment, we not only record and play audio – we also process it! So, any digital processing, or rather its quality, directly depends on the sampling frequency. Any plugin that does not work in oversampling mode produces digital artifacts.
Try to run a simple sine wave into the saturator in oversampling mode and without it. In the second case, you will see how new harmonics will line up not only above, but also below the original sine. This will be very aliasing!
Despite the confident dominance of digital technologies, many still choose the good old analog. Any hybrid setup, including adders, external processing, etc., involves several stages of digitalization-dedigitization of the audio signal. Such multiple conversions are quite capable of leading to signal degradation. And here oversampling comes to the rescue: since it kind of “increases the digital resolution” of the audio signal, the risk of error accumulation in the process of its multiple conversions from digital to analog and back is reduced!
Well, so far it’s all pros, isn’t it? Time to add a fly in the ointment) About the cons of Oversampling:
- file size. It grows proportionally to the sampling frequency. Double it – and the file will weigh twice as much!
- CPU load. In oversampling mode, all plugins work less quickly. And for those who use DSP processing, there is even worse news: its performance steadily decreases with increasing sampling frequency (in other words, the higher the frequency, the fewer UADs you can hang)
- For owners of complex setups of several digital devices connected, there are also nuances. The optical bandwidth decreases with increasing sampling frequency. For example, for a frequency of 192 kHz using the ADAT protocol, you will only have access to 2 channels, not 8, as usual. In addition, many old devices do not support frequencies above 96 (or even 48) kHz.
Verdict:
As you can see, the disadvantages of oversampling are related exclusively to hardware limitations. So if your setup allows you to comfortably work at a high sampling frequency, you can be sure that it will not be worse!
But I have to disappoint those who like to compare sources recorded with different sampling rates! Until you start processing them, you will not hear the difference – it lies beyond the perception of hearing and simply cannot be noticeable! Everything else is self-hypnosis.
Written By Nikita Doroshenko: Sound Engineer.
