Great Article on Digital Audio, Hearing and Perception

[stubbsonic]

The section that conflicted with my "stable data" and experience is where it says that "16 bits is enough to span the real hearing range with room to spare."

As someone who was doing some pretty intense critical listening to the same pre-masters in 24 bits and 16 bits, there was a noticeable difference in the sound. Where I heard it most was in the stereo image presented by the reverbs and stereo recorded tracks, and in the quality of sound in the tails of reverb and fade-outs.

But I'm no snob. I enjoy listening to CDs. And much of my daily work is in 16-bit, 44.1K. However, when I work on critical projects and am listening at my most intense levels, I can hear a difference. Don't ask me to do a double-blind until I've become more familiar with the recording. Something about hearing something 1000 times in 24-bit and it keeps getting better and better, then you take it down to 16 and for the first time... the overall quality takes a small step backward.

Audio placebo effect? Perhaps. But in most areas of my life, especially in audio, placebos usually have little or no effect on me. Honestly, I have no investment in using 24 bits for audio. On the contrary, I'd rather just do everything in 16 bits. But, somehow I've become convinced that 24 bit sounds better.

It would be tricky to null test a 24 bit track (with a nice dynamic range) against a 16 bit version of the same track (both run through an identical analog stage). Clocking issues or other factors would be tricky to eliminate. But if this was possible, then the results might be telling. I'd be stunned if this hasn't been done many times.

Theoretically, if the difference is inaudible, then 16 bit wins the test. However, it might be interesting to boost the gain and hear the nature of the difference.

[MIDI Life Crisis]

There's also a big difference if you are listening to a delicate and exaggerated track, as in a quiet movie theater or other theatrical and more acoustically designed and controlled space. On an average consumer system or in headphones with 1/4" sound generation membranes I'm sure 44.1/16 bit MP3s will sound as good as they can. Put the same audio thru a $250k theater system at performance levels and compare your delicate mix at 16 and 24 bit and then we can talk.

[mikehalloran]

As someone who was doing some pretty intense critical listening to the same pre-masters in 24 bits and 16 bits, there was a noticeable difference in the sound. Where I heard it most was in the stereo image presented by the reverbs and stereo recorded tracks, and in the quality of sound in the tails of reverb and fade-outs.

There's also a big difference if you are listening to a delicate and exaggerated track, as in a quiet movie theater or other theatrical and more acoustically designed and controlled space. On an average consumer system or in headphones with 1/4" sound generation membranes I'm sure 44.1/16 bit MP3s will sound as good as they can. Put the same audio thru a $250k theater system at performance levels and compare your delicate mix at 16 and 24 bit and then we can talk.

George Martin put it nicely: "All you need is ears."

[Kubi]

Just got back my album from mastering - asked for two copies. One is the original resolution of 24bit 88.2kHz in which it was mastered (my mixes were upsampled fro 24-44.1 to that before the mastering), the other are the 16-44.1 versions for CD. Both are identical in their treatment, except of course the last step of taking it down to the CD standard rate.

The difference is not huge, but provided the playback system and room is such that smaller differences can be heard at all, easily audible to anybody in the room, non-musicians like my wife included. Mostly apparent in the smoothness of the low end and the high end, and the sense of space between the elements. Absolute no-brainer.

So, nice engineer talk - but the conclusion that 16-44 is as good or better than higher resolution and SR simply can't be true. (Don't have any opinion on 192, but have worked with 24bit 88.2/96 enough to know it simply is better - even when using a 16bit 44.1 Kontakt patch playing polyphonically and in chords. Immediately audible to anyone.)

[Kubi]

PS: My mastering engineer played me a few tracks from a DSD project, and I wish he hadn't. That is so superior to PCM that it's scary, and a little bit sad. Heard the two side-by-side using the same material, and it's just sad how inferior PCM audio is. Wonder why we aren't using that, at least as a final medium...

http://en.wikipedia.org/wiki/Direct_Stream_Digital

[Prime Mover]

But it probably doesn't matter. Even if analog signals had the same limited resolution as the digital source, at 16 bits, there are ~65,000 levels. If you assume the effective resolution is 12 bits, there will still be 4,000 levels. How many levels can we reliably distinguish?

To me, that's not really the point. Amplitude doesn't just define volume, but the accuracy of the reproduced waveform. I would guess that higher amplitude resolution could translate into a more accurate representation of the original source. Of course no one can hear the difference if we're talking about bumping the signal up 0.01dB, but it MAY make a difference in whether some of the points are more acoustically natural, or too steep, etc. But I may not be correct on this, this is at the limit of my knowledge.

The one thing to consider is that he's talking about final listening resolution, not pre-processing resolution. In any kind of media, if you plan on doing processing, it's good to have higher resolution than the final product. For instance, in video, if you plan on wanting to do a lot of zooms and digital pans, you'd be better off shooting at 4k (about 3x higher resolution than HD), even if your final broadcast resolution is HD. Similarly, every digital effect can theoretically degrade the amplitude and frequency resolution.

[bayswater]

Eric, you're right, it's not the point. The point, according to the article is that resolution is not reduced in any way by using 16 versus 24 bit, just dynamic range. I was just saying that if he's completely wrong, that the step function in digital audio typically displayed also applies to the converted analog product, it's hard to argue that 16 bit would restrict resolution in any practical sense. We would still get resolution to a few thousands of a db.

[Prime Mover]

Considering that higher sampling means a whole lot more processing, hard drive access time, and space, I'll stick with 44.1/24bits for recording, and 16bits for final mastering. I'm always trying to squeeze every little bit of juice out of my CPU and hard drive access time, so I feel just fine sticking to that. I produced my last album in 48kHz, and realized later that that was probably not worth it, I'll be stepping down to 44.1 from now on.

[HCMarkus]

I find this section of the discussion interesting. How can we explain the conflict between those who hear the difference here at MOTUNation and the Boston Audio Society ABX test discussed in the referenced article (emphasis added):

This paper presented listeners with a choice between high-rate DVD-A/SACD content, chosen by high-definition audio advocates to show off high-def's superiority, and that same content resampled on the spot down to 16-bit / 44.1kHz Compact Disc rate. The listeners were challenged to identify any difference whatsoever between the two using an ABX methodology. BAS conducted the test using high-end professional equipment in noise-isolated studio listening environments with both amateur and trained professional listeners.

In 554 trials, listeners chose correctly 49.8% of the time. In other words, they were guessing. Not one listener throughout the entire test was able to identify which was 16/44.1 and which was high rate, and the 16-bit signal wasn't even dithered!

I sometimes wonder if we may unwittingly fall prey to selection bias... just sayin'.

[bayswater]

Double blind testing?

[stubbsonic]

I'm not aware of the details of the test but here are some possible reasons...

1. Inexperience of the listeners to know what to listen for. I think more experienced engineers can listen quite deeply (e.g., hearing a reverb spread across the panorama, etc.) and might more quickly hone in on differences.

2. Listening examples were dynamically compressed (average level was in the upper range- no quiet parts, no "high impact" parts). The difference with 16 bit is more subtle when the material is at louder volumes. And when the average is loud, the overall listening volume is lower, so the lower-resolution quiet sections are ... quieter (less easy to hear the lower bit stuff).

3. Listeners were unfamiliar with the test examples. I think I can hear the difference because I've been hearing a track repeatedly during a project. I suppose I may be deluding myself. But SO often, I try different audio processes and I don't feel even the slightest hesitation to admit when I can't hear a difference.

I think the null test (if done correctly) should be a pretty reliable indicator. If there is a difference in the signal between 16 & 24, the null should reveal those differences. I know it is not that simple as the differences would now be quite low, whereas with the original signal, they (the differences) would be part of a hotter signal.

I played with bit-reduction effects and have noticed that as I start with lower values and raise the bit depth, the sound goes from terrible to bad to just ok to good (maybe very good). But it's not like I stop hearing a difference from 12-16 bits. The final step from 15 to 16 is an audible improvement. Of course, the effect may not handle some of the low bit filtering that may be needed.

[mikehalloran]

Without knowing the source material, we have no way of being able to qualify the results. This is why I proposed the test that I did.

I remember a live stereo album from the late 1950s with a paragraph on the back that told us that, when examined through a microscope, there were grooves from 16Hz to 25kHz (cycles per second back then). They didn't understand the effect but they weren't going to remove them just in case. I have it in my collection still. It sounds great all these years later played back through 'phones.