Have been researching - and listening to - DACs fairly extensively for last few months. Will try to distill my impressions here since it seems broadly relevant to the discussion and others may find some interest.
Virtually all DACs used in high-end audio fall into one of two categories: either resistor ladder or delta-sigma.
(a) Designs based on resistor ladder (also called R-2R) chips are typically either based on the legacy Burr Brown PCM1704 family or a similar legacy design from Analog Devices eg AD1865, and delta-sigma designs are based on modern chips from AKM or ESS.
Key points about resistor ladder chips: (1) both the BB and the AD chips are long since discontinued as the industry moved towards delta-sigma convertors which are far cheaper to produce as they don’t require expensive precision laser trimming of on-chip resistor values; (2) both are ‘prized’ by some audiophile manufacturers for their ‘qualities’; (3) when used in high-end DACs the on-chip oversampling/filtering functions of these legacy chips are typically bypassed in favour of using a dedicated DSP chip (such as a Sharc chip in the ND555 and the Soulution 760) which will oversample/filter, with some using a ‘fancy’ spline interpolation algorithm rather than ‘plain old’ zero padding or using a pre-calculated compensation algorithm to correct for the phase errors introduced by filtering. Examples of high-end DACs using legacy resistor ladder chips include Naim ND555, CH Precision C1, Soulution 760; (4) in an attempt to overcome the intrinsic linearity issues which ladder chips manifest, multiple ladder chips are typically used for each channel eg 2 or 4 or 6 or more (paralleling the chip outputs averages the linearity errors of each chip). There are other ways of dealing with the intrinsic linearity issues; for example, the CH Precision C1 has a function which can be invoked to pass a known signal through the DACs chips, measure the output, compare that output to what it should be and then calculate an ‘offset’ which is added to the input signal in order to ‘counteract’ the measured linearity error.
Variation on resistor ladder theme: ‘discrete’ resistor ladder DACs such as TotalDAC which work on the same principle but replace the resistor ladder chip with a set of resistors laid out on a circuit board. No oversampling, no filtering. Schematically simple. MSB DACs are very similar, using a set of ‘modules’ which seem essentially to be discrete resistor ladder arrays. Aries Cerat DACs use legacy Analog Devices chips, parallelled, with no oversampling, no filtering.
(b) For audiophile / high end pieces, delta-sigma designs use ‘premier line’ chips from ESS (eg Sabre 9038PRO) or AKM.
Key point about the delta-sigma chips: they are cheap and ubiquitous, used in everything from that tiny Apple dongle (which converts the lightning output of your iPhone into a stereo minijack for connecting your earphones) to the digital audio workstations used in studios. Examples of high end DACs in this category include Ideon Absolute, Merging NADAC and many others.
Variation on delta-sigma theme: dCS’ Ring DAC which is more or less the same thing as a high-end delta-sigma chip but has its current sources (‘latches’) laid out as discrete components on a circuit board (rather than all being integrated on the convertor chip itself), fed by a signal processing ‘mapper’ on a FGPA which oversamples and filters the incoming signal.
I’ve now listened to most of the DACs listed above and my distinct impression is that, assuming the amps and speakers are good enough to reveal differences between them, those differences are really not big. And it may be that any differences which are present are mostly due to (a) differences in the power supply to the digital and analog sections of the DACs and (b) the quality of the post-convertor analog sections of the DACs. This may be why so many of the high end manufacturers make a song-and-dance about esoteric, complex, over-specified power supply arrangements (Ideon Absolute, Naim ND555 for example) or opt for incorporating tube amplification stages (Lampizator, Nagra, Ypsilon, Aries Cerat) as a means of deliberately imparting a sonic signature which wouldn’t otherwise be present (as intrinsic differences in the ‘sound’ of different convertor technologies - resistor ladder versus delta-sigma - are minimal). It has been my experience, over the last few months of auditioning, that the line stage / power amps in use have a bigger influence on the sound than do different DACs (Since Xmas, I’ve listened to Naim ND555, CH Precision C1, Soulution 760, Ideon Absolute, Lampizator Pacific, TotalDAC d1-seven, Nagra Tube DAC, dCS Rossini + Clock, dCS Vivaldi Apex DAC, MSB Premier, T+A DAC8 DSD). Just to throw the cat amongst the pigeons. (post-script - I see that Simon raised some aspects of what I have written).