Digital Audio: The Line Between Audiophiles and Audiofools

This vid starts a little slow and gets a little slower as it progresses but I watch this stuff hoping to land on the correct side of the title -> "Audiophiles and Audiofools"

The presenter provides down to earth explanations of the complex paradigms that define 'Hi Res' audio. He also explains how these paradigms work together to exceed the capacity of the human ear. IE: Sample rates; Bit depths; Lossless CODECs; Lossy CODECs are all defined in simple terms and related to the human aural capacity.

https://www.youtube.com/watch?v=IiZqYnd5g8M

I agree if a lossless FLAC CODEC is used to create a FLAC file from a CD that was mastered with the standard Phillips CD format and you're not planning on remastering the FLAC file then it seems reasonable to say that paying more for a 24 bit FLAC over a 16 bit FLAC is a waste money and storage space.

However the presenter barely touches on PCM but he does not address SACD, DSD and the latest DSD DAC offerings at all.

if everything that can be heard by the human ear is included in the standard Phillips CD format and we use a DAC that accurately decompress's the FLAC without alteration how do SACD, DSD and the latest DSD DAC's add to human listening experience?

I'm no expert, but I'm always a little uncomfortable with the way Nyquist is thrown about. It stated that the minimum sampling period required to get a unique reproduction (my terms here) is at least twice the frequency being sampled. I'm not sure that translates into perfect reproduction of arbitrary wave forms up to that frequency. Even in a simple sine wave case, I see how it might, or might not, do a 'good' job. Imagine a 20 KHz sine wave. Now, sample it at 40 KHz. Should be possible to get it back, perfectly, using a filter that cuts off at 20 KHz, right? So let's consider that those sampled points were right at the zero crossings of the sine wave. The digital sample sequence is 0, 0, 0, 0, ... Now, shift the timing (phase) of sampling by 12.5 usec (a quarter wave), and the samples become 1, -1, 1, -1, ... HOW are you going to get the same analog reconstruction from these two fundamentally different sample patterns? Of course, there IS NO magic to get that sine wave back from the sequence of 0's - ALL frequencies, at zero amplitude (silence) have that same pattern of all 0's.

:confused:

Dramatically increasing the frequency, IS likely pointless, but moderate increases above 2x the max does not seem crazy to me. Repeat the example sampling at something higher, like 80 KHz (4x). Now you cannot get 0, 0, 0, 0 from 20 KHz. You get something between 0, 1, 0, -1, etc. and 0.71, 0.71, -0.71, -0.71, etc. I can readily see how fitting a sine wave of up to 20 KHz to the various sample sequences possible at 80 KHz always comes up with pretty good approximations for phase and amplitude of the original signal.

I have an analogy in some of the work I do, where, theoretically, I can achieve a prefect analysis of the components of combined signal (spectrum) regardless of relative levels, but it only works with PERFECT data, which I cannot get in the real world.

Now, those who ARE experts can school me on where I'm wrong.

:eek:

This is way too technical for me to understand, at least at this point. But it's fascinating nonetheless. Can you recommend more videos like this btw? Cheers!