Re: Oh dear.
@Vic - again sorry for inability to quote correctly. A few points:
"Class A stages are very, very close to linear - close enough to say that they are indeed so. Class B is nothing like as close."
On the other hand, a class B power amp can be produced with linearity good enough to be audibly irrelevant. E.g., http://www.douglas-self.com/ampins/dipa/dipa.htm
It seems to me that once you've got good enough performance on any parameter, it doesn't make sense to try to improve that parameter: a much better idea is to try to improve other parameters - in any engineering context.
""Surely the thing to consider is how it sounds?"
No. If your recording equipment colours your sound, you've already got fidelity problems. The recording/mixing/processing stages should be as transparent as possible. This means generating as little noise as possible, and in the digital domain, that means high sample rates and high resolution."
I seem not to have got my idea across. My thinking is that if an improvement in measurable engineering terms of a particular engineering parameter does not produce any detectable change in how it sounds, then maybe it's a good idea to put your efforts into improving other parameters.
Of course the recording and processing stages should be as transparent as possible. The question is how best to achieve that. Certainly high resolution in the amplitude and time domain when using digital recording are helpful, but how high is high enough? But high resolution is worthless unless your measurements (which is what digital sampling is) are also precise and accurate.
I'm no expert in the field of digital sampling. You tell me: what are the tradeoffs between precision, accuracy, and resolution when doing audio digital sampling?
""If a given combination of resolution and sampling frequency produce results audibly indistinguishable from the best available analogue technology, maybe the thing to do is improve linearity and noise (etc) before bothering with top-line buzz-numbers?"
Why? Analogue recording technology pushed the capabilities to the limit to achieve acceptable performance. Digital techniques allow *much* higher fidelity at negligible cost; it would be foolish to restrict those capabilities just because the previous technology couldn't match them."
You talk about negligible cost - well, maybe. I'm just wondering how much improvement you actually get from increasing resolution in the amplitude and time domains beyond a certain limit - what might that limit be? What compromises are being made to increase those resolutions? At what point should engineering efforts be spent on improving other aspects of the sampling process?
I don't know, but I do have doubts that there's much point in going beyond 32 bit / 192 kHz sampling from the point of human audio recordings.
Biting the hand that feeds IT
