And I agree it's a convert-to-converter thing...though the underlying theory (for whatever it's worth) implies that more samples per second = much finer "slices" of the audio you are sampling…
Actually what the underlying theory - specifically, the Nyquist theorum - states is that in order to
losslessly reproduce an analog frequency, one needs to double the frequency for the sampling rate.
For real-world reasons such as aliasing and physical low-pass filter characteristics, real world sampling rates are slightly higher than the theory stipulates; this is why we sample at 44.1k instead of an even 40k in order to cover the audible audio spectrum.
But the "more slices" interpretation is anything from a bit misleading to downright incorrect the way it's represented in many textbooks. A 44.1k sample wont reproduce a (for example) 10k signal any better than a 25k sample rate will. Assuming all other physical design characteristics of the circuit being equal, the difference in sample rate will have zero effect at that frequency. Any increase in "resolution" translates into an increase in frequency response, not in accuracy at that frequency.
The same holds true for 44.1 vs 88.2 or 96. Assuming you have a 44.1k D/A that is designed well enough not to add any artifacting at 20k because it's filters and such are designed well enough, an equally-designed 88.2 D/A won't do any better of a job with 20k than the 44.1 will. It will simply increase the frequency response to 40k.
Then when you figure in that, unless you are recording a clean synth with HF information direct, your recording signal chain *before* the converter probably won't take you to 20k anyway, the extra sample rate is superfluous.
...and heck, we all know that hard salami tastes MUCH better when it's sliced very thin.
That, OTOH, is very true!
It's just not an accurate analogy for digital theory. I'm not knocking you for selecting 88.2; go ahead and use whatever sounds best and works best for you. I'm just relating that if 88.2 sounds better for you, it's probably because of a poor design at 44.1, not because of the actual sample rate itself, that's all.
I guess the more important point he should be aware of is the bit-depth
Also agreed. And I think there's a pretty good consensus that 24bit is the way to go there, with the option of using plugs that can go to 32bit floating point of you want.
But his issue/concern isn't so much about what rate is best, anyway (from what I'm reading)...it's more to do with him trying to sample MP3's at above-48kHz rates and wondering why it won't work.
True, but he did raise the question/point of he thought it was silly not to use 96k as long as it was there. I (and others here) we're just pointing out that it's not really necessarily as silly as it might seem at first blush, and that it can in just as many cases be silly to automatically use it by assuming it is automatically better.
As far as the converting to MP3 thing, I'd recommend saving a target file as WAV first, and make that your master copy. Then if you want to step down to MP3s for distribution, do it from there.
I'm not intimately familiar with Sonar, but I wouldn't be completely surprised if it's MP3 encoder simply isn't designed to work at bitmapped rates of higher than 48k. It probably would not be the first nor the last encoder out there to have that kind of limitation. This should be documented in the apps online help, I would think.
If that's the case, what I would recommend, if he decides to stay at 96k/24-bit for his work, is to save his masters as 96k/24-bit WAV, just so he has a clean, unadulterated copy of his final work. Then for duplication purposes, make a duplication master WAV at 48k or 44.1k and 24-bit, and use that to make his MP3 encodings.
G.
) as WAV files, and then convert my final CD WAV files to 16/44.1...but for MP3 files, it's a much lower quality than that.
