Oh shit! SINAD can suck my ****s

Yes, take a THD+N (SINAD) at lower voltage levels to check for noise for gear which is much more voltage sensitive (which of course would be much more sensitive to noise). In fact, do this: take THD and N separately!

Any THD/N measurement must be tailored for the intent of the device: expected load, sensitivity, etc.

--

There is no such thing as a 0.003%THD+N or 120db SINAD amp. However, there is such a thing as an amp that has 0.001% THD+N at rated power (10W) into 8-ohms (with 20kHz bandwidth implied). We need to be very careful and specific when we use the term SINAD or THD+N.

I mean, we don't go around saying the HD650 is a 60db SINAD headphone. If anyone did, the reply would be this: "go f**k yourself" or "what da f**k are you saying bitch?"

--

Now I get the need for simplification. This is why I prefer the use of the GONAD matrix:

 

1. FFT process gain will show a lower noise floor on the spectrum, but the harmonics won't be affected. The calculated SINAD / THD+N isn't affected by FFT process gain. So this is a non-issue.
2. If we lower the output voltage level from rated max from a power amp, most likely we will get better THD because rated max power is usually taken when 1% THD is reached.
3. THD will likely do down as voltage output is lowered into a given load. However, at some point THD+N (with the noise component) will go up.
4. To sum it up THD increases with output level. THD+N is usually V shaped, at lower levels starts off kinda of bad because we are so close to the noise floor, starts to get better at a "sweet spot", and then increases as we go up in output level. Opamps exhibit a checkmark V, when THD+N gets lower lower lower, until the shit suddenly hits the fan.

However the above is generalizing. Not everything hits follows the expected behavior. See below. The real answer to your question is rather than speculate what happens to THD when the output level is changed: measure it and see! It's best not to always assume.

 

Answer is more complex. The problem is that we cannot measure distortion in transients (for example, what I've been trying to do with headphone burst responses), or at least provide a handy number.

The math used to calculate SINAD requires taking 16k to 32k samples typically. On top of that, we then do that 3 to 5 times and average the results to get rid of noise. The problem with transients is that maybe we have a handful of samples to play with. The math can't work on this.

I'm convinced the distortion profile, or distribution of harmonics, or even out of band ultrasonic noise may provide hints of transient distortion.

Otherwise, the Magni 3+ and Heresy should sound exactly the same. Both have similarly super high THD+N into typical headphone loads at typically voltage levels.

Yet in a level matched triple blind test, I could easily tell the difference between them.

https://www.superbestaudiofriends.org/index.php?threads/magni-3-heresy-and-3-blind-listening.8663/

 

For Verite, the results, use the Sennheiser row. Similar sensitivities and impedance.