"Although it is not immediately obvious, this circuit’s resolution is not restricted by the DAC’s quantizing limitations. The loop forces the DAC’s LSB to oscillate around the ideal value. These oscillations are integrated to DC in the loop compensation capacitor. Hence. the circuit will track input shifts much smaller than a DAC LSB."
witchcraft.
@niconiconi So, if I'm reading this correctly, it claims that by *introducing instability*, *resolution* (and by implication precision) is somehow increased?
I can think of words other than "witchcraft" to describe such a claim...
If you have a slowly-changing signal, you can use a 1 bit ADC to get arbitrary resolution, by adding white noise of a known distribution to the signal and ADCing the sum (the averaged value you get from your ADC will correspond to the noise's PDF at negative input signal value).
Amusingly, many GPS receivers (used to?) play a similar trick, because they deal at some point with something like a signal with very low SNR that they have to average over a longish period of time (although they don't care about increasing resolution, but only increasing precision), so they already had to deal with a noisy signal. (See https://s53mv.s56g.net/navsats/theory.html for the kind of simplifications that yielded in old receivers.)
