First of all, apologies to the OP for having completely hi-jacked this thread! ...At least hope this discussion is useful/interesting for some 
It is strange to me that people believe to hear differences in the sound, but never mention noise or quantization noise.
I hope you don't believe that Italian electro-technical engineers (specially those who have been revolutionizing Italian audio and hi-fi culture in the last 30+ years) don't know the theory and technique of A/D conversion... LOL
Quantization noise is of course a well-known issue even here at South. But it's not what's discussed in the article I mentioned, which refers to recording (and reproducing) a low-level signal (for example late reverberation reflections in a hall).
Quantization values (i.e. the number of bits used when recording) are not fixed, but related to signal levels; in fact, the signal level is expressed by the number of bits used, which corresponds to its amplitude.
The so called quantization noise, which is a reflection of the quantization distortion, occurs anyway, because it responds to the difference between the analogic (continuous) signal and the sampled&held stepped one. It depends both on the sampling frequency (for example, 44,1 kHz) and the quantization (for example, 16-bit). Which is intuitive, because the closer the steps, the less the difference with the original soundshape.
But the point here is that the A/D converter uses the full sampling range (let's say 16-bit, theoretically) for a 0 dB signal, but progressively less bits for expressing the level of a weaker signal. And this is what Eng. Nuti writes.
Now, when this low-level signal varies its amplitude (for example an echo fading away), the resolution available to faithfully reproduce this fading is exactly the number of bits the converter was able to use for representing it. If the signal amplitude corresponded to 4 bits, this fading may just assume the following values: 16, 8, 4, 2, 0. In other words, you just have 5 fixed stepped levels to reproduce the fading.
In a more robust A/D system, with 24-bit or higher quantization resolution, you find that the number of bits available at that same level is much higher (for example, 7), giving a series of 128, 64, 32, 16, 8, 4, 2, 0: almost double the "resolution", and much closer that the previous digitalization to the real signal. The perceived difference is a more solid perception of the room and its volumes, more stable stage reconstruction, firmer instruments positioning in the 3-D space. If the playback system allows it.
As you can see, the issue discussed has nothing to do with quantization noise, which is present nevertheless.
Generally speaking, a 16-bit sampling allows for 65,535 different levels, while a 24-bit sampling allows 16,777,215 different levels: even keeping the same sampling rate, it's 256 times denser. Doe someone really think that this huge amount of space is just ... empty?
Of course, if all you record is a 0 dB signal (like some punk song LOL) the improvement with a higher resolution source is less meaningful than recording a classical concert live, where the ambience is a living part of the program and the dynamics may be outstanding.
QuoteWhere can I find an A/B comparison on the web?
The number of sites that make HD audio samples available for technical comparison is steadily increasing. You might try googling for "HD audio samples". Often you find lower and higher samples in the same site for comparison purposes.
OTHO, many stores are now selling HD files along with the 16/44,1 CD, and you can buy both and compare them on your reference system. One of them is http://www.hdmusicstore.it (Italian only, I'm afraid).
Neil Young has recently acquired 5 (!) companies with the goal of spreading the HD music all over the world. His word is "down with the mp3!", where the density of information in an mp3 is as low as 5% (!!) of the original recording (studio tape).
The point here is discussing the practical limits of a technology, not a single person's ability to tell differences. We know some people have a better discriminating hearing than others, and this is a fact; it's a matter of genetic, health, education, listening pleasure and other reasons.
Eric Johnson can tell the brand of the batteries in his stomps from the way power fades off; I know people able to tell the brand of power cables (the ones between power amp and cabs that is) in a hi-fi system (between a limited number) when you switch them. They are exceptions for sure, but those exceptions are a huge help when it comes to fine-tune real devices and understanding their limits.
As for me, I've always thought that you have to expose yourself to the best possible experiences, in order to get better every day: most possibly, a person accustomed to listening to mp3's won't wholly catch all the improvements in better-sounding material, while the opposite is quite likely to happen. I believe we also become our experiences.
OTOH, many people are perfectly happy with mp3: I absolutely see no bad in this. It's exactly like people happily playing any guitar they came across, and other people always looking for a better pu impedance, a better volume capacitor, a better tube set, or a better come sensitivity... or a better modeller!
What matters, IMO, is happily express ourselves and enjoy our senses 
I have the deepest respect for those experts as they did at least show up. And I don't mean to be sarcastic about it. Most Hifi-people have strong opinions but would never show up to proof them. The guys at the Hifi-Magazines being by far the worst - at least over here in Germany. Whatever I read about CD's, how they work and why it's required to use 1.000 € cinch cables, when I read about Jitter, pits and lands and how they alledgedly represent digital 0 and 1 - it all shows only one thing: Those people don't have a single clue about how that CD-thing actually works technically.
