Originally Posted by flatmap
I'll try to make it on July 26. It would be great to hear Dan Lavry say a few words.
One idea would be to ask him to speak on the merits of different resolutions for digital. Doesn't he state that simply higher isn't necessarily better? I'd love to hear more about that.
No new toys, but I'd bring along my K-702's and Proton. I would like to hear the matchup with the AKG's and Big Poppa's Darkvoice amp again.
Higher resolution is a good thing. Higher resolution means more bits, and with more bits, each sample is closer to the correct value.
Of course there are practical limitations. For example, the noise floor can be a limiting factor. There is not much point in trying to be accurate to say 0.1uV, if the noise riding on the signal is 10uV.
But as a rule, high resolution is a good thing.
Unfortunately, the rather loose use of phrases such as "high resolution" or "high definition audio" causes much confusion, and some folks confuse "high sample rate" with "high resolution".
What I have been saying is that sample rate too slow or too fast reduces the resolution. Too slow sampling means you lose audible bandwidth (can not hear the highs). But too fast sampling costs you accuracy over all the audible frequency range.
More is better when it comes to bits, pixels, money in my pocket... :-) But more is not always better. One needs only 2 dots to define a straight line. One needs only 3 points to define a circle. For digital audio, digital video, digital medical gear, digital instrumentation and so on, once you define the bandwidth you need, your sample rate requirement "falls into place". If we can agree that the ear can hear say 20KHz (or 30KHz or whatever), in theory, you need to sample twice that rate (thus 40Kz or 60KHz or whatever). In practice we need to have some reasonable margin. That is why 44.1KHz CD tends to cover around 20KHz audio, and 96KHz is already an overkill.
I know that it is not easy to grasp, but if you have enough samples, (just a little over twice the bandwidth), a proper DA will "connect the dots" correctly, and will re make the original signal, not just at sample time but at any time between the samples.
There is no need to have more samples then at a little over twice the bandwidth, it adds nothing to the signal. In theory, it does not add or take away anything. The analogy is again -You can use 2 dots for a straight line, or 10 dots for the same line. A ruler and a pen will yield the same line.
But in practice, going faster and faster, means getting less accurate result. That is why a 1GHz converter is at best capable of 4 bits; a 100Mhz is not true 14 bits, and so on. The slower the rate, the more accuracy you get. Of course we can not go too slow either because we need to cover the audio range...
So slower is no good, and faster is no good. What is left is something between slower and faster - the OPTIMAL rate, which for audio would be in the 60-70KHz. unfortunately, there is no 60-70KHz standard. 44.1KHz is a bit slow, and 96KHz is a bit fast, but they are not too far from the optimal. 192KHz is way off the mark.
To summarize: High resolution audio means the ability to capture and reproduce audio precisely. Conceptually, more bits are always better, they add to the accuracy. But as Nyquist proved, making the sample rate faster beyond a certain point (twice the required bandwidth) adds nothing to the outcome. In practice, one can not keep adding bits, at some point, the various physical causes or errors limit the number of “real bits”. In practice, sampling too fast brings about errors, while adding nothing – there is a down side and no up sides.
Thus - sampling too fast is LOWERING the accuracy. Calling 192KHz high resolution is misleading.