There are a lot of elements that make one DAC sound different from the next, however there are two elements in particular that cause first-order effects IME: (Most others are second order, and I will list these later)
1) Master Clock jitter and jitter rejection of the DAC
2) Digital filtering or lack thereof
You will notice that I did not even mention the D/A chip used. These actually dont vary hugely, except for the AKM series that sound a bit electronic no matter what you do in the DAC IME.
You will find that once you have a really low jitter source to feed the DAC, that most of them will sound virtually identical, particularly if you are using a preamp in your system, which masks a lot of detail and causes compression, even the best of them. I can point you to three TAS reviews where Steven Stone drove various different DACs (ranging from $1k-$7K) with the same low-jitter source and they were not only improved, they were virtually indistinguishable from each other, and Steven uses a near-field system too.
Digital Filtering is what generally sets-apart older ladder D/A chips and the like from newer Delta-Sigmas. Generally, in newer chips, the digital filtering is automatically selected and nearly impossible to override. There are a few exceptions. On the other hand, it is just missing in older chips like the PCM1702 and TDA1543. As a result, if you dont add a digital filter, these older chips sound more like analog. The downside is that they dont achieve the detail rendering that the newer chips do.
It would seem like the goal is to eliminate or reduce the effects of digital filtering using a modern D/A chip, and this is exactly what my goal was.
Okay, now lets talk about second-order differences in DACs in general. Here is a list of things that set them apart:
1) power subsystem - this has a major effect on the dynamics and bass tightness. Too few power decoupling caps, the wrong caps or wrong sizes of caps in the wrong places all has an effect. Voltage regulator speed and regulation accuracy has a huge effect also. Both of these are critical to achieving slam as well as control.
2) analog output stage - this can compress if there are too many stages or too many op-amps, as well as adding distortion and noise.
3) I/V converter - most D/A chips need this and the more perfect this is, the better the SQ
4) PC board effects - this is something that I discovered by modding for about 10 years. It was obvious when I modded a Mark Levinson #38/380 preamp. This preamp had long analog signal traces running from the front panel all the way to the back, more than a foot long. This preamp had a very dark sound due to this. Rewiring all of these traces by cutting them and replacing with pure silver wire in teflon tubing made a huge difference.
5) volume control - next to jitter, this is the next MOST IMORTANT THING to address in your audio system. Using digital volume control is not a good idea because more than about 9 dB reduction in volume starts reducing resolution. Using an active preamp (which is usually what is inside a DAC) is not good either because of the added noise, distortion and compression. This can and usually does suck the life out of the music. Passive linestages based on resistors are a bad idea also. The only ones that sor to work locate the resistors on the back of the amps, eliminating the capacitance of the cables in the response equation. Resistor passives generally reduce dynamics by limiting the current that reaches the amps. So what is the solution:
There are two actually:
a) Transformer-based linestage - if the transformers are well-designed, this is the most transparent of the choices. The drive from the DAC must be sufficient however. A benchmark DAC1 with op-amp output will not cut-it.
D/A control by changing the reference voltage - this does not add any parts to the signal path. No distortion, compression or noise is added, and no resolution is lost. In fact, as you reduce volume, the distortion decreases. This is the opposite response of any other volume technology. There is only one DAC on the market to my knowledge that has this.