An example for a dongle:
Output voltage: at least 2 volts on unbalanced, 4 volts on balanced.
SINAD: 100 dB or better, 1 kHz, 22.4 kHz bandwidth
SNR at 50mv: 85 dB (?)
SNR at full 2/4 volt output: 110 dB (?)
Output impedance < 1 ohm
Example for Amplifier:
SINAD >= 80 dB
SNR >= 110 dB (?)
Channel balance < 0.5 dB
Crosstalk > 70 dB @20 kHz
About SINAD I have a personal train of thought:
Distortion audibility is best when
- testing with pure sine waves
- the order of the harmonic is very high, which minimizes the chance of the distortion being masked by the fundamental
- the distortion product(s) land(s) in the region where our hearing is most sensitive
- the level is sufficiently high that the distortion product also reaches a meaningful level
- the level is NOT so high that equal loudness contours take "flatten" our perception, which means the relative difference in level between a low frequency fundamental and a high order distortion becomes small again
Looking at the distortion audibility level graphs
here (you need to pay to view all), the absolute worst case of the shown examples and orders comes down to 12th order harmonic at a 300Hz fundamental (3400Hz harmonic) at 100-110dB and falls to 0.03%, which equates to 70dBr.
Based on that information, that means a device with more than
70dB SINAD will be audibly transparent at all times.
For reference, I encourage anyone who reads this to take the
Klippel listening test for distortion audibility.
Now, several devices with similar distortion in series will quickly worsen that figure. But realistically how many devices will someone have in series? DAC, EQ, Preamp, Amp? +3dB on an uncorrelated addition on each, which would net us 79dB of SINAD in this case. I don't see any benefit in a device having a SINAD better than 80dB based on these conclusions, as long as DNR is on point.
About SNR@50mv I think a simple noise level in volts would be a bit more convenient to work with. Just take the sensitivity of the IEM, multiply it by the figure and see what level the idle noise will reach with the IEM in question. But SNR@50mv is fine too, I guess. Just a bit more abstract for absolute comparison, but it works perfectly well for relative comparison.
Channel balance sounds fine, I wouldn't want a device with 1dB channel imbalance.
Output impedance I don't know. I'd say take the lowest impedance headphones/IEMs with the highest reactance (impedance swing) and calculate the impact of amplifier output impedance and make sure that the impact stays below 0.5dB? Somebody else can do the maths here, maybe
@staticV3 or
@oratory1990 feels like doing that
SNR I think is the main thing for me when it comes to DACs. If it's sufficiently high, you don't have to worry about gain staging, running lots of negative preamp gain or having a long chain of devices.. 110dB sounds like good to me.
Crosstalk highly depends if you're listening on sterophonically (speakers) or binaurally (headphones/IEMs). The more crosstalk, the more 'Mono' the signal becomes. 20dB crosstalk is audible but didn't bother me in my tests, 30dB I found inaudible (on speakers). Idk about headphones, I did not test that so I can't really comment any further on that, but 70dB sounds excessively high to me, if we're talking about audibility only. Maybe 50dB? It's not a tough bar to clear, but why cast off a device that is otherwise flawless if it has a crosstalk of 60dB?