[tortugaranger]

While off-topic from LDR preamps, I decided to share the post below because of its outstanding clarity and quality on an aspect of digital audio over which I suspect most (including myself) are a bit confused if not misinformed. Very informative.

My favorite quote from the thread's author follows. Here he's trying to provide some perspective on the practical significance (there is none) of the increased dynamic range afforded with 24 digital (vs. 16).

A very rough way to think of it; if the maximum level of a digital system were set at the sound of a truck going by from 10ft away then 144dB quieter (in 24bit) would be roughly the level of noise produced from two hydrogen atoms colliding!!

http://www.head-fi.org/t/415361/24bit-vs-16bit-the-myth-exploded/15

[robertopisa]

Yap, the declared 24 (or even 32) bit resolution cannot be perceived by humans. The best DACs can reach around 20 bits, mine is a very good DAC and reaches slightly less than 20 bits. This is due to the quantisation noise, which masquerades everything beyond 20 bits. In other words, in the best treated rooms with the best rigs, you can get the leftmost 20 bits of each sample of digital audio.

This is the reason why I do not like the digital volume and opted for a good analog passive volume control. Jriver has an extremely good digital volume on 64 bits. However, only the leftmost 20 bits can be actually heard in the best acoustical environments. Drawback is that each approximately 6db of level reduction means that the bits are shifted to the right by one position. Since only the first 20 bits can be actually heard at most, this means that for low-level volumes the first, say 5 or 6 bits are always zero, and thus the leftmost 20 bits contains only 15-14 bits of audio signal that can be heard. This is even less in normal rooms, not treated professional studios. This means that the low-level details, such as ambience and 3D soundstage, are getting lost, which makes the difference in hiend audio.

[tortugaranger]

thus the leftmost 20 bits contains only 15-14 bits of audio signal that can be heard. This is even less in normal rooms, not treated professional studios. This means that the low-level details, such as ambience and 3D soundstage, are getting lost, which makes the difference in hiend audio.

The revelation for me personally in this article was how I had been confusing bit depth with the general concept of resolution and so of course 24 HAD to be better than 16. Once I understood that bit depth is about dynamic range and that the dynamic range available via 16 bits is sufficiently deep (down into the noise floor) such that adding more bits (going deeper into the noise floor) doesn't buy you anything. Whereas pushing up the sampling rate beyond 44khz does make sense up to point.

The other somewhat related thing I've finally tumbled to doing research on DACs is how the push into hi-res DACs for the general consumer and even high end audio effectively abandoned the R2R Ladder DAC which ironically sounds better than most (all?) of those higher-res delta-sigma DACs. I had an opportunity recently to listen to a comparatively modest 20 bit R2R Ladder DAC running 16/44 files and decided right then and there that I'm going back to the future and getting an old-school R2R. No doubts at all.

Don't mean to start a big long argumentative thread on this topic so I think I'll stop here. Peace out.

[Mike Nomad]

Don't mean to start a big long argumentative thread on this topic so I think I'll stop here.

Probably too late.

[*Scotty*]

It probably wouldn't hurt to look at bit depth and resolution from an analogue distortion standpoint.
In a 16 bit system your lowest distortion in terms of THD occurs when you have all 16 bits present, which is at 0dB VU
Peak signal. This equals a level of THD equivalent to a distortion attenuation level of -96.33 dB or 0.0015258% THD.
 In a 24bit system the theoretical lowest THD possible is equivalent to -144.49 dB or 0.000006%THD, once again when all the bits are present at 0dB VU.
The best D to A converters are running at about 20 to 21bits of resolution or a worst case THD equivalent -120.41 dB or 0.0000954%THD at 0dB VU.
 The crunch comes when we don't have all of the bits present which is what happens as the recorded signal level drops away from the peak volume at 0dB VU. Each bit equaling 6.02dB of signal level. In a 16bit system when you are down at -66.22 dBdb you have lost 11bits of information leaving 5 bits to describe the information which doing the math gives us a THD equivalent to -30.11 or 3.1153009 %. This is a level of distortion that most of us would find objectionable.
 With even 20 bits of resolution from a 24 bit system, the distortion situation at -66.22 below 0dB VU is much better. While we have taken away 11bits we still have 9 bits left to describe the information which gives us a distortion attenuation of -54.18 or 0.1954339 % THD which is almost 16 times lower distortion than a 16 bit system at the same -66.22db signal level.
In system with a true 24bits of resolution taking away 11 bits to reach a -66.22dB signal level still leaves 13 bits to describe the information would yield a distortion of 0.012218% this is level of distortion approximately 255 times lower than what a 16bit system has at same signal level.
 A good case can be made for having at least a 20 bit level of resolution in a D to A converter and 20 bit resolution source material.
I did not discuss the effects of dither as that subject goes beyond simple mathematics. 
Scotty

[Russtafarian]

Here's a link to the source document. 

http://benchmarkmedia.com/blogs/news/15121729-audio-myth-24-bit-audio-has-more-resolution-than-16-bit-audio?utm_source=Application+Notes&utm_campaign=f80542b9fc-AppNotes_15aug-+Audio+Myth+%28LINK%29&utm_medium=email&utm_term=0_e28f43b8aa-f80542b9fc-125367429&mc_cid=f80542b9fc&mc_eid=d0e0f50302

Please note the final statement:

"Long word lengths do not improve the amplitude "resolution" of digital systems, they only improve the noise performance. But, noise can mask low-level musical details, so please do not underestimate the importance of a low-noise audio system."

John Siau is not saying that 16 bit is as good as 24 bit.  24 bit is clearly better, but the difference is due to lower noise rather than increased resolution.  It sounds like two sides of the same coin to me.  What's the difference between "more bits resolves more music" and "more bits lowers noise to reveal more music"?

Russ

[*Scotty*]

He views things in terms of SNR within the context of a digital recording and playback system. I don't think this is as accessible as viewing the loss of bits as the signal level decreases in terms of its effect on the THD at a specific signal level. At the end of the day in a PCM system when you have 0 bits you 100% distortion. In a 16bit system this is what occurs at -96.33dB.
Scotty

[DaveC113]

Thanks for the info Scotty, what dB VU levels do you consider average average? I know it varies a lot, I'd guess a super-compressed pop recording might only drop to -12 dB while a symphony might actually have a lot of information at the -66.2 dB point you referenced?

I don't think I've ever compared the same recording at different bit lengths while keeping the sampling rate the same...