Audio dithering typically involves adding a small amount of noise before downconverting to lower resolution samples.
But there's another form of audio dithering that uses error diffusion (like TFA describes) rather than adding noise. If you use a single-bit ADC but sample much faster than Nyquist and keep track of your errors with error diffusion, you preserve all the audio information in the original with a similar number of bits as a (e.g.) 16-bit ADC sampled at Nyquist, but with the additional benefit that your sampling noised has moved above the audible range where it can be filtered out with an analog lowpass filter.
This is one-dimensional dithering but in the audio world it's called Sigma-Delta modulation or 1-bit ADC.
But there's another form of audio dithering that uses error diffusion (like TFA describes) rather than adding noise. If you use a single-bit ADC but sample much faster than Nyquist and keep track of your errors with error diffusion, you preserve all the audio information in the original with a similar number of bits as a (e.g.) 16-bit ADC sampled at Nyquist, but with the additional benefit that your sampling noised has moved above the audible range where it can be filtered out with an analog lowpass filter.
This is one-dimensional dithering but in the audio world it's called Sigma-Delta modulation or 1-bit ADC.