IMHO it's very clear and natural choice -- used throughout the industry -- to measure a DAC vs. its input levels, not its output level like
@amirm does. So, you dial in 0dBFS or the more typical -3dBFS to see its behavior at high levels where it usually is worst wrt distortion.
If the DAC has a variable output based on analog stage after the DAC chip (which we have in this case), you set that to a reasonable level you have to inspect that seperately if and how much this affects the measurements. It clear a close to zero volume will degrade signal to noise ratio, whereas too high a level can add additional distortion.
Since normally ASR measurement procedures aim to extract the best performance the device is capable of in full isolation on the bench, a DAC output should measured in this way:
- Set up a typical high send level you want to test at, for example -3dBFS, say, for a 1kHz spectrum,
- Scan through DAC output level settings, if any, to find the region where the performance is stable and constant and then leave it there (during that scan one would have quickly seen that the DAC is not the problem at higher output levels, the level/buffer stage is. The same would have come to light if the transfer line-in --> line-out were inspected).
Amir's referencing to output levels (as of late) has produced misleading data at times. What makes it even more problematic is that Amir seldom states digital input levels when he adjusted the input level to get his "nominal" output levels of 2Vrms for unbalanced and 4Vrms for unbalanced... these choices don't make much sense either because it
automatically assumes that a) a balanced output is signal-balanced (which is NOT the definition of a balanced output, impedance balance is) and b) the balanced output is derived from the unbalanced output with an added inverter. You shall not make such assumptions, output level is output level and when the device doubles the output level for balanced then measure at this level, if not then don't (repeat: the reference is input level).
This are just a few points why I'm saying Amir's measurement lack some thoroughness. Not specifying most details of his measurement setups doesn't help either to have confidence. Why not create and attach a professional test report?
A real point that speaks for Amir, though, is that he strives for some sort of consistency so comparisons and ranking (for those who feel the need for such) are somewhat easier to make, but we must never forget that his data doesn't always represent "the truth". I personally wished Amir produced less tests per month but more complete and thorough ones.
I still think his data is very useful, despite all the criticism, just don't jump to conclusions, and fellow posters please don't be silly claqueurs.