Very low THD DAC/AMP makes no sense, as transducers are the primary source of distortion.

[wolfield]

Consider the following: high-quality modern electroacoustic transducers (speakers or headphones) can achieve distortion figures in the ballpark of 0.1% (-60 dB), and this applies to really good units.
Now, imagine two DAC/AMP combos that deliver 0.01% (-80 dB) and 0.0001% (-120 dB) distortion to these high-class speakers with 0.1% distortion.

Right away, there is a catch regarding the difference we can expect from both systems. Harmonic distortions from different components in an audio chain do not simply add linearly in a straightforward way (e.g., -80 dB + -60 dB does not directly result in a single combined dB value). In most real-world scenarios, the phase of the distortion components is not perfectly aligned, so we treat them as uncorrelated signals and combine their power (not amplitude). However, let's imagine an absolute worst-case scenario and assume that harmonics are perfectly in phase, so we add distortion amplitudes directly.

For -80 dB playback system:
Convert dB to Amplitude:

• DAC/AMP: -80 dB = 10^(-80/20) = 0.0001
• Speakers: -60 dB = 10^(-60/20) = 0.001

Combine amplitudes directtly for the worst-case scenario:

• Total amplitude = 0.0001 + 0.001 = 0.0011
• dB level = 20 * log10(0.0011) = -59.17 dB

For -120 dB playback system:

• DAC/AMP: -120 dB = 10^(-120/20) = 0.000001
• Total amplitude = 0.000001 + 0.001 = 0.001001.
• dB level = 20 * log10(0.001001) = -59.99 dB.

And here we have it, folks - hardly any improvement of only 0.82 dB.
What's worse - the real-world number will be even lower.

 

[wolfield]

And if we combine distortion powers using the root-sum-square method, which should be much closer to the real world than our absolute worst case scenario then the difference between -80 dB and -120 dB playback systems on -60dB THD speakers will be around 0.03 dB only.
This is absolutely negligible, comparable to measurement inconsistency.

Until technology allows the production of speakers/headphones with distortions lower by an order of magnitude than the current generation of transducers, the pursuit of very-low distortions in the audio chain is akin to literally applying snake oil to your forehead.

 

[NTK]

As many here are engineers, we celebrate execution excellence. The recent SMSL SU-1 DAC achieves a SINAD of 116 dB for a current retail price of USD $85 at Amazon. What is the excuse of doing worse than this if you're asking for more money? If your device comes in with auto DSP, I'll give you a 10 dB discount for processing headroom.

Given the level of performance already baked into DAC chips, what are the reasons for a final product to perform worse than the reference design? Ditto for amplifiers.

 

[Waxx]

Both of you are right, and the only excuse to do less good than those cheap is a deliberate colouration, sometimes with old tech (tubes, class A). But then don't claim a superior sound on tech level when you build that kind of amplifier. Technically they are inferior, but that does not mean people will not like it. Oldtimer cars (or cars build the old way like Morgan cars) are also technical inferior to modern cars, but they are still fun to drive and have a certain feel that modern cars can't give.

First Watt amps (mostly class A) are not getting there, because they were designed with a certain colouration in mind that the designer (Nelson Pass) intended. He doesn't sell them as technical superior, he sells them as a niche amp with a colouration that he likes and he knows most of his customers like. He said himself (on diyaudio) that if you need super clean amps you better look elsewhere, beause he doesn't make them. He finds them boring (he finds... so subjective).

But if you ask 10K or so because "your amp is superior", and then it measures at 80dB sinad or has other deviations from the technical standards of hifi, then you're selling snake oil because better amps can be had for a few hundreds, or maybe even less.

 

[Mnyb]

We also have noise and FR response . The "usual" design methods that yield low thd tend to also yield low noise and very flat FR and in power amps also low output impedance ?
So they sort of go together ?

 

[Mnyb]

And it does not seems to hard to accomplish . "Bad" data in moderns amps tend to be because of deliberate incompetence using "alternative facts" when designing