Scott Kramer
100+ Head-Fier
- Joined
- May 19, 2017
- Posts
- 204
- Likes
- 573
Yes. The sampling theorem (Nyquist-Shannon) states that any band-limited signal can be completely reproduced if sampled at least at double the frequency of its highest frequency content, which implies that using 44.1k sample rate you could perfectly reproduce content up to 22.05k.
As I understand it this has been mathematically proven already, but still seems to be very much under debate as far as HiRez audio goes.
I stumbled upon this video from xiph.org on the tube a while ago and with the usual YMMV and all that, I can’t find anything wrong with it, at least not with the (albeit limited) knowledge that I have. It’s actually very interesting, if you’re so inclined:
Yes. The theorem might be solid enough, but the real challenge is to band-limit the signal, both before sampling it (ADC/in the studio) and after reproducing it (DAC/at home). To accomplish this perfectly you need to construct a brick-wall filter that lets all content below the limit through unharmed and at the same time completely blocks all content above the limit. All without introducing any aliasing or artefacts of any kind in the audible frequency range.
This is really, really hard to do.
One way to make it a bit easier is to raise the sample frequency as to be able to use a less steep output filter, which is less prone to produce aliasing/artefacts.
Yes. When feeding a DAC at it’s maximum frequency there is nothing to oversample as the content it already is at the frequency that the DAC operates at. This would effectively disable the input filter and, as I understand it for Schiit multibit DACs, the comboburrito filter. This would leave the oversampling to the computer’s audio system or none at all if the file already is at the maximum frequency. These can of course possibly yield a different sound depending on the filter being used, owing to differences (or flaws) in their implementation. The Windows system for example, is not very higly regarded in this department as far as I understand.
I discovered that my MacBook apparently automatically set the output to my BF2 to 24/192, so I’ll have to manually set it to 44.1 and listen if I can hear any difference when the comboburrito can work it’s magic fully (now the computer does 44.1 to 192 and the comboburrito would only double up to 384k)
Your point...? This is well known. What if it's recorded (ADC) at 24/96 properly? It's also bandwidth limited so you can reconstruct that perfectly... can 16/44 (DAC) side reconstruct 24/96 perfectly? (Upsampling for filter simplification, HiREZ shenanigans, audiophile dumbassery, and do humans perceive this are other topics outside this snarky (not really) post).
Last edited: