@Krunok and
@Headphonaholic were talking about transparency earlier. When working with less than fully capable equipment, it's fun to think about ways to squeeze more out of it... within reason.
Looking at Amir's graph of THD+N vs power output, we see what seems to be a typical profile for solid state amps, with distortion falling with increased output until we hit the point of clipping. Sure there's a little rise before clipping, but distortion still stays relatively lower than at low output.
In the case of the Nobsound into a 300 ohm load, we see distortion of 0.0043% at about 116 mW. Driving an HD650, that works out to 119 dB SPL with distortion happening at -87 dB, or 32 dB, which one would suspect could well be audible.
However, people don't typically listen at such high levels (even accounting for 20 dB peaks). I personally listen at levels of around 70 dB, so we could assume peaks around 90 dB. On the HD650, that would require only .14 mW, which puts us in the vicinity of 0.05% distortion, -66 dB, so about 24 dB. Again, this might very well be audible.
What only occurred to me today is that, using negative digital pre-gain, we can effectively move into the amplifier's sweet spot. In this particular case, we'd apply -29 dB of pre-gain to push the 90 dB peaks to require the same output as 119 dB peaks without pre-gain. At that point, we've got THD+N of -87 dB against an output of 90 dB, so the noise is only 3 dB above silent. Even better, at the average level of 70 dB, we're drawing 1.12 mW of power, putting distortion at 0.006% or -84 dB, or 14 dB below silent. That seems pretty darned transparent!
Am I missing or miscalculating something here? If I'm right, it seems like a good general strategy with solid state gear to apply negative pre-gain to push the amplifier into its optimal performance range given the listener's preferred volume. If I were to build a DAC/amp, it would be interesting to include some knowledge of its distortion profile and then automatically apply negative pre-gain in the digital domain in order to operate in that optimal range from a distortion perspective.
EDIT -
I need to get better sleep! What I'm missing is this - the distortion is a function of the output power. For a given load (headphone), the output power for a given SPL is always going to be the same. If I digitally attenuate the input signal, I end up having to turn up the volume dial, which just reduces the analog attenuation of the input signal prior to amplification. So the amplifier is doing the same amount of work and will produce the same amount of distortion.
To move the amplifier into the sweet spot of distortion without increasing SPL, I would need to increase the load, either by using less efficient headphones or adding some sort of analog attenuation after the amp (which I assume creates new issues).