Digital audio is always sampled, which means that any digital audio file is created at a fixed sampling rate (and resolution). The sampling frequency of the Red Book Audio CD is 44.1 kHz (16-bit resolution). Audio on DVD is sampled at 96 kHz (24-bit resolution). Sample rate conversion is necessary when you need to convert a digital audio file (i.e. an analog signal that has already been digitized) from a given sample rate to another frequency (resolution may remain unchanged or change). Upsampling (or interpolation) is the process of converting from low to higher (for example, from 44.1 kHz to 48 kHz); downsampling (also known as decimation) is the process of converting from a high frequency to a lower one (for example, from 96 kHz to 48 kHz). It is important to note: our studio performs mixing and mastering at a frequency of 96 kHz, therefore thinning is used to provide additional formats intended for distribution on social networks and on streaming services.
Changes in bit depth, for example from 16-bit to 24-bit or vice versa, can be performed simultaneously with resampling, but care should be taken to avoid mixing the two principles. Poor resampling algorithms, whether upsampling or downsampling, can introduce artifacts that are clearly audible during playback. A typical low-quality, but extremely fast algorithm will be based on linear interpolation. High-quality oversampling algorithms use more processor time because they require conversion to the frequency domain. Modern PC processors (clock frequency ~ 2 GHz) can easily cope with very high-quality resampling in real time. Sound cards that perform real-time resampling require a good DSP.
Oversampling is often required, which is actually part of the mastering process for audio CDs, as professional audio equipment uses 96 kHz or 192 kHz for masters, while the Red Book Audio CD specification uses a sampling frequency of 44.1 kHz. Different media are recorded at different sampling rates (44.1 kHz CD, 48 kHz DAT, 96 kHz DVD audio, etc.). Digital mixing of different sources, selected at different speeds, will require re-sampling to the overall speed and resolution.
Many PC audio cards (in particular, Creative Labs codecs based on 10k1 and 10k2) and AC97 codecs can only enter, output or process audio data at a frequency of 48 kHz and force the recalculation of information at one stage or another. Sometimes sound software, plug-ins and drivers add the ability to re-sample, which allows for a cleaner sound (at different levels, product algorithms can show themselves in different ways). For example: the same synthesizer, at different levels of resampling can give different shades of the same signal.