24/192 Music Downloads are Very Silly Indeed

https://people.xiph.org/~xiphmont/demo/neil-young.html

by mbrubeck • 13 years ago

4 5 13 years ago

people.xiph.org

natch• 13 years ago

Wow this is such bullshit. The sample graph he shows where he (correctly) debunks the stairstep theory gives no acknowledgement of the real-life situation, which is that many waves are interacting with each other in a signal, and it's not the stairstepping of the pure simple waves he showed that is the problem, it's the stairstepping of the interacting waves, which are complex enough that a low sampling rate can throw things way off (meaning off far enough to cause artifacts at audible frequencies) when rounding errors come down on the wrong side.

mbrubeck• 13 years ago

No, even if you sum any number of signals, as long as each component signal is bandlimited then the resulting signal is also bandlimited and can be reconstructed perfectly according to the Sampling Theorem. This may be counterintuitive, but it's provable. (I could prove it using just what I remember from a couple of semesters of undergrad signal processing courses.)

Of course a low sampling rate can result in errors, but what the sampling theorem says is that we can calculate exactly how high a rate we need to always perfectly reconstruct signals in a given band. This doesn't change just because your signal is constructed from many component signals.

Of course, if you think the math is wrong, then you have an easy way to disprove it: Construct a counterexample out of "many waves interacting" and show that the resulting signal isn't perfectly reconstructed as the Sampling Theorem claims. (Or perform a double-blind test that shows any human can detect the difference between a 192 kHz recording and the same recording downsampled to 48 kHz.)

davidgerard• 13 years ago

You're arguing with mathematics here.

jemeshsu• 13 years ago

Many years ago my audiophile friend said amplifier should not have remote for volume control, it reduced the sound quality. So I asked him 50-100 years down the road, does human being still need to manually walk to their amp to adjust the volume.

It's true that most people can not tell the difference between a 320bps mp3 vs 24bit lossless audio. But 50-100 years later, will we still be listening to 320bps lossy quality audio? That will sound really silly.

pasbesoin• 13 years ago

I appreciate the lengthy exposition; however, I've found that 320 kbs with cross channel bit sharing (or whatever it's called) is where I'm finally reasonably happy, at least with rock/pop.

Perhaps as with eyesight and the "fourth type of cone", we'll eventually measure/document that not all listeners are equal.

P.S. Also, pay attention to the subsequent analogue amplification stages, not to mention speakers. Your digital metrics will prove less important when your analogue pathway suffers.

24/192 Music Downloads are Very Silly Indeed

Related Stories