Digital Music Jargon Explained – Oversampling

This second part of a series on music technology terminology deals with the burgeoning issue of oversampling. Many of you may have seen this attribute on Compact Disc players in days of old, but what does it actually mean?

Oversampling The Goods

There are actually two forms of oversampling – at the input stage and at the output stage (digital to analogue conversion). It is the latter form that is of interest here.

It is fairly intuitive that higher sampling rates lead to better digital representation of analogue waveforms. However, oversampling actually refers to the generation of additional samples from a waveform that has already been digitally encoded.

Nyquist And Shout

Disc drive
Creative Commons License photo credit: aussiegall

On a CD, we can only play back frequencies up to 22kHz (as per the Nyquist Theorem), as everything above that is subject to aliasing (and beyond the range of human hearing anyway).

Therefore, a low-pass filter is applied to remove the aliasing. However, as music producers know, every EQ (filter) applied to a signal colours the sound in some way, particularly if you have a steep gain/attenuation curve.

If you are rolling off from 22kHZ to -90dB at 20kHz, you will probably introduce some unwanted phase shift artefacts in the high frequencies of your recording, due to the steepness of the filter curve. However, if you had sampled the original recording at a higher rate, you could have much more of a range in which to apply this same attenuation, using a gentler filter slope which would result in less distortion.

Interpolating Steps In

This is where the concept of interpolation is applied. Once the signal is in the digital domain, it is possible to apply a phase linear digital FIR (finite-impulse response) filter and reconstruct the waveform at a higher sampling rate.

In essence, this process uses the original samples to draw a continuous waveform (as is the case in the analogue domain) and then fills in the gaps with extra sample data. This ‘guessing’ of the waveform’s inter-sample positions is interpolation, and can be used to synthesise any desired sample rate.

So, if your CD player uses 8x oversampling, when it applies the low-pass filter it can use a range of 158kHz to achieve the desired attenuation rather than only 2kHz. This is a cheaper and less computationally intensive method of achieving a good output signal, and dramatically reduces phasing problems in your music…


Home | Canabrism | Guides | All Music Technology Posts | XML Sitemap