Propeller Head Plaza

Excellent discussion there. A must watch for anyone wondering about hi-rez digital audio

182.53.241.86

This Post Has Been Edited by the Author

In Reply to: RE: A brief subjective study on digital signal quality by AM fidelity posted by rick_m on February 28, 2011 at 09:00:43

I am very pleased to see the great measurements brought up by Keith Johnson. He wanted to cover all things that go wrong with CD quality audio with little time explaining his slides.

My one little thing I was trying to find out about was chopped up upper sidebands at higher frequencies. I had been thinking about the problem for years, but I finally sat down and with little math convinced myself there's a problem with dynamic music signals and insufficient bandwidth, to let things move about in the small brick-walled space.

Now I have some verification from Keith that indeed, the actual bandwidth as stated by Nyquist is ONLY for steady state sinewaves. I kept seeing so many people publish that Nyquist's math was for every dynamic signal, but I could not see how that can be, and doubted it.

There's no place for a moving amplitude or phase for a sinewave signal up high. The rise time for a signal approaching Fs approaches infinity, and bandwidth given to it zero! That is how I see things with the equations I provided that Johnson talked about. To capture full dynamic capability out to 20 KHz, you need a minimum 4 times that plus margin to help. Even more when you count the possible FM that can spread that need even wider. Just in AM, that is where the 88 KHz rate comes about from Johnson's own words.

But interestingly (and perhaps a bit obvious), more is always better to do editing, digital equalization, etc. out to 20 KHz as pro audio has had to deal with first.

All the slides were very good measurements of that which original CD player manufacturers had failed to implement well. And they were thought out measurements designed to find the culprits. Not an overall scan with too simple measurements that cannot detect these things. Brilliant work Keith showed there, and lot of it. He connected real measurements with audible results every time and beautifully isolated. Bravo!

Maybe the minimal sampling for dynamic signals reproduced in a space Fbw (20KHz here) should be called the Johnson theorem, an adjustment to the Nyquist theorem.

Minimum Fsample > 4 x Fbw. Add margin for better results. That understanding is really a new modified theorem for Nyquist's. I do believe now that the Nyquist theorem was derived without concern for anything but steady state. The theorem needs a new revision to take in more understanding of newer assumptions into account, I am convinced. All theorem's are only as good as the assumptions applied to it.

That video made my day. Thanks so much for providing that, Rick.

-Kurt

Follow Ups: