A bit off-topic however I feel I have to say that:
To design a good sounding electronics is not a trivial task. There is much more to it than appears to an electronics engineer who has little to no experience in that very special area. Just as a simple example - yes, transducers are in theory the most non-linear parts of the audio chain. However for that very reason the electronics facing those devices on both ends is difficult to design right. Even a "perfect" amplifier has to work not with a dummy resistive load but with a very non-perfect speaker. So the real importance is how these two devices work together (and that is not an easy thing to get right, believe me, I've designed a number of amplifiers, including some very good ones, and some not so good). It is like in any specialised area of electronics, as soon as you start to dig deep you have to change your perspective and to learn some nuances.
As an amplifier designer, how much should a good audio amplifier with, say, 70-100W of power cost on the street?
A quick google search for "class ab 100w audio amplifier IC" came up with this in the first few results.
Low noise ... 0.01% THD ... 6 Euros in one-off quantity on Mouser.
I suspect the 'amplifier' problem has been studied, the design compromises have been made and it's now available in a monolithic packages from a dozen different manufacturers. 
In a way, yes. Monolithic amplifiers can be very good and have excellent performance/price ratio. Discrete amplifier designs are still almost always
crap that simply won't last (e.g. class AB output stages with Ube multipliers
)
In commercial applications class D is extremely important, too, of course. Not just for low end applications, but also for high-end applications. Today you will often find that even ICs designed for low end markets incorporate not simply PWM modulators but actually quite good SDMs. Class D can surpass 0.005 % THD+N today (actually, ten years ago, but let's not nitpick on time lines).
There is a fuzzy crossover point where true pursuit of high-fidelity reproduction becomes a completely subjective fetish. Certainly a more socially acceptable fetish than many others.
Well said!
If you reduce the resolution to 16bits and 48kHz, the quality of that recording noticeably suffers to my ear.
You should immediately visit a university clinic. Someone there is probably interested in doing some hearing tests with you.
As a matter of fact, increasing the Fs from 48 kHz to XYZ kHz does not add any information in the audio band (<20 kHz)
at all, and never has.
Calling bullshit. To, like, all points you make. Your dynamic range calculation is just utterly wrong (at 30 dB SPL average you are unable to perceive a 100 dB dynamic range). Studios, just by the way, use 24 bits or 32 bit float not because 16 bits aren't enough, but because dozens of post-processing filters simply add a lot of noise. Using a high resolution format keeps that noise waaaayy below anything someone can hear.
When I was talking volume range, I meant rms volume of sine wave. You need more bits to represent sine waveform, don't you? So the total dynamic range should be volume range AND snr of the waveform itself.
Uh, no, that's not how digital volume control in practical DACs/SDMs work. They won't reduce the effective dynamic range until quite some volume reduction, typically -50... -60 dB, because they directly fiddle around inside the modulator and don't simply scale the input PCM stream (which, yes, would proportionally reduce dynamic range. So in systems where no DAC volume control is used 24 bits makes sense - it still doesn't make sense to distribute 24 bit media.)
Home
Help
Search
Login
Register
