I am no DSD expert (ask @j_j ) but yes it is a one-bit stream. IIRC the idea was to output the delta-sigma's single-bit stream directly then use analog anti-image (low-pass) filters, compared to conventional delta-sigma DACs that use a (mostly) digital filter. It is basically a sigma-delta modulator without the output decimation (low pass) filter. It was originally a 64x bitstream but there are faster versions now. One-bit delta-sigma modulators do have some issues, including higher distortion and susceptibility to "tones" (signal artifacts created by repeating digital patterns in the modulator), that are not a (normally) problem for multibit designs. They also require much higher sampling rates to achieve the same resolution as a multibit design. Virtually all modern delta-sigma converters are multibit designs with digital processing to provide nearly ideal performance.
But fundamentally the time resolution is a function of how closely you can sample the signal, that is how close together in amplitude two points can be. That is a function of converter resolution, i.e. the least-significant bit (lsb) step size, and the signal frequency -- not the sampling rate (though Nyquist must be met). Time resolution is thus how big a step (the lsb, set by the number of bits the converter delivers) and how fast the signal goes through that step (set by the signal frequency).
The link in my signature goes to a post linking a number of introductory articles on sampling and conventional and delta-sigma DACs (among other things) that may provide more background. DSD, like any noise-shaping method, loses resolution as frequency increases (hopefully well above the audio band, thus the need for high sample rates and noise filters).
HTH - Don
Last edited:
