This link was brought to my attention recently:
http://people.xiph.org/~xiphmont/demo/neil-young.html
Although the article seems to make sense, I have tried ABX testing using the Foobar utility and can accurately pick the higher resolution file (192/24 vs 48/16) 90% of the time (although I notice that accuracy decreases rapidly with ear fatigue).
This leaves me wondering what is always true in the article versus what might be more complicated/variable. I'm not looking for black and white debate over dogma and opinions here; rather I am hoping to hear the perspectives of level headed and open-minded people who know more about this topic than me. Is hi res audio an unnecessary and detrimental waste of space or is it worth using?
I don't consider Monty's article to be open-minded. His mind is clearly made up, and this is an advocacy article. OK fine, if it were presented that way, but it isn't. Anyway, I'm not going to argue all the points I disagree with in his article, but I will contribute a few perspectives that you don't find there. And you can decide for yourself whether I'm being either level-headed or open-minded.
First, the Nyquist Sampling Theorem is correct, and it will always give you the right answer. However, it has no power to tell you whether you have asked the right question. If your goal is to record and play back music that sounds natural, then "How high a sampling rate do I need to record the highest frequency of interest?" is definitely the wrong question. If instead you ask, "How high a sample rate do I need to record and play back with minimal distortion of the waveform?" you find that you need a sample rate that is quite a bit higher. To simplify a bit, any practical implementation of a sampling system is based on filters, both on the record end and the playback end, and these filters, whether they are digital or analog, affect the signal in the time domain. This is completely ignored in the Theorem, because it is only concerned with what frequencies can be recorded. But our ears certainly don't ignore it in listening.
Discussion of high sample rates always seems to center on the ability to record ultrasonic frequencies. Look, we know that doubling the sample rate raises the high frequency limit, we're not beginners here and we all understand that much about digital audio. However, the guys who are pushing digital audio to better levels of performance don't say they need to record ultrasonics. They need the high rates to get digital audio to pass the waveform with less temporal distortion. One of the best A/D/A's ever developed, the Pacific Microsonics Model Two, doesn't even record that much into the ultrasonic range at high sample rates. The designers instead used that extra bandwidth to implement gentle rolloff filters that have less temporal distortion. So please ignore all that hooey about ultrasonics; it's a smoke screen. And once you quit worrying about ultrasonics, there really isn't much left in the Xiph article.
So go ahead and do your own listening comparisons if you wish. There are perfectly valid technical reasons why you MIGHT hear differences, and if you do then you are free to have preferences as well.
Paul Stubblebine
Paul Stubblebine Mastering, San Francisco
The Tape Project, LLC
serious student of the audio arts
