Oversampling is to sample the input signal at a frequency far greater than the Nyquist sampling frequency. Let the original sampling frequency of digital audio system be fs, usually 44. 1kHz or 48kHz. If the sampling frequency is increased to R×fs, R is called oversampling rate, and R > 1. In this sampled digital signal, because the number of quantization bits has not changed, the total quantization noise power has not changed, but at this time, the spectral distribution of quantization noise has changed, that is, the quantization noise that was originally evenly distributed in the 0 ~ fs/2 frequency band has been dispersed to the 0 ~ Rfs/2 frequency band. The figure on the right shows the power spectrum of quantization noise when oversampling. If r > > 1, then Rfs/2 is much larger than the highest frequency fm of the audio signal, so that most of the quantization noise is distributed in the high-frequency area outside the audio frequency band, and the quantization noise distributed in the audio frequency band will be reduced accordingly. Therefore, the signal-to-noise ratio of the system can be improved by filtering out noise components above fm through low-pass filter. At this time, the maximum signal-to-noise ratio of the oversampling system is shown on the right.

Where fm is the highest frequency of the audio signal, R fs is the oversampling frequency, and n is the number of quantization bits.

As can be seen from the above formula, when oversampling, the signal-to-noise ratio of the system increases by 3dB for every doubling of sampling frequency, in other words, the number of quantized bits increases by 0.5 bit. It can be seen that increasing the oversampling rate can improve the accuracy of A/D converter.

However, the effect of improving SNR by this oversampling method alone is not obvious, and noise shaping technology must be combined.