Article: The Denafrips Design Philosophy Part 1

[miguelito]

Interesting. Would love to see a more precise characterization of digital noise, I assume some of it is quantization noise (critical for low level signals), jitter, etc. It is interesting the point about vinyl noise and that the brain is able to separate it out, that is indeed my experience.

 

I would say there are ways to make measurements better: Most music occurs at low volumes. Yet, we measure THD, SINAD, etc with max volume input signals. I would like to see these measurements with a large set of input signal levels, even just the sine waves normally used. In fact, my understanding is that the original AES spec was to measure THD with an array of input levels. I speculate the industry didn't like this because it showed the defects of crossover distortion in the first solid state amps - they would look a lot better if just the max input signal was measured.

 

To illustrate the point, if you plot how much time a 2V sine wave spends at each level of signal, you get the yellow line below. But most music follows a distribution given by the red line.

 

As for the argument that real music is not used because it is hard to characterize it - as opposed to predefined waveform which anyone can reproduce. We could come up with a complex set of waveforms, with level distributions that resemble music, and use that. As a scientist (I have a PhD in Physics) all of these things should be measurable. However, it is clear that the simplistic measurements we do today are not enough.

 

Finally, the key arbiter of it all is what we hear: the ear is a very precise tool and it is after all the one we ultimately care about.

[fas42]

1 hour ago, miguelito said:

Interesting. Would love to see a more precise characterization of digital noise, I assume some of it is quantization noise (critical for low level signals), jitter, etc. It is interesting the point about vinyl noise and that the brain is able to separate it out, that is indeed my experience.

 

 

Digital noise is much harder for the listening brain to ignore - it contaminates the entire sense of the musical thread; and this is why there is a general belief that so many digital recordings are 'bad'. If, and this is a mighty big if, a digital replay chain is finally of a high enough integrity that the remaining noise is attenuated enough, then the brain can also separate out what it wants to concentrate on, the music.

 

The trouble for makers of components is, that so much noise originates from poor implementation of the overall system; things like cable and their connections that are too flawed, and inferior resistance to exterior electrical noise. Only products which use OTT methods are usually good enough to be unaffected by the issues; and of course these are very expensive ...