Great Article on Digital Audio, Hearing and Perception

[HCMarkus]

http://xiph.org/~xiphmont/demo/neil-young.html

Interesting... x-ray audio anyone?

[MIDI Life Crisis]

Thanks for that interesting read. Always the skeptic, I kept looking for flaws in the arguments, but he addressed my concerns logically and I was in total agreement with many of his arguments - especially as concerns when and why 24 bit is useful.

[bayswater]

http://xiph.org/~xiphmont/demo/neil-young.html

Interesting... x-ray audio anyone?

How about those 30kH 500kW radio stations? They keep me awake all night.

[Prime Mover]

Yup. Completely agree.

One of the gotchyas that gave people the wrong impression was the first mass distribution of digital audio in the mid-late 80s with CDs. Immediately there were cries by audiophiles everywhere that CDs were no match for vinyl. It's true that side-by-side comparison of albums often revealed CDs to be pretty poor, but it had almost nothing to do with digital audio and everything to do with rushed remasters, degraded old masters, 16bit mastering, and poor understanding of digital mastering. Also, vinyl does add it's own color and saturation, but is that a good thing? That sound wasn't there when the mixing and mastering engineers decided it was the perfect cut.

I think people equate higher sample and bit rates to being "closer to analog", as smaller stair steps look more like analog waveforms. But the reality is, everything you HEAR at the other end is analog anyway, those jagged steps are only theoretical and in the encoding; the DAC isn't generating that shape, it's back to smooth anti-aliased waveforms, as this paper points out.

The one thing I'm still a little unclear about, after all this time, is the difference in bitrates. At times I hear people talking about higher bitrate being higher resolution of amplitude gradations, other times I hear it being described as a larger amplitude range. I don't see much point in having a larger amplitude range than human hearing can tolerate, but I could definitely understand an argument for smaller gradations being slightly more accurate at reproducing content, and being more resilient under lots of processing. When I read Mastering Audio, I got the distinct impression that bitrate translated to resolution, but most of the posts seem to suggest it just adds range. Which is it?

[cuttime]

So is dithering useless?

[stubbsonic]

With the higher bit-rate (24 bit) you get both higher amplitude resolution and a wider dynamic range (it is essentially the same thing).

If music is recorded... let's say... "nicely"--- there can be this kind of average listening level that is clean and comfortable. Then there could be this impactful loud passage that is not distorted, and a very quiet section that is clean and clear. At all levels, there is ample bit resolution to represent the music clearly.

With 16 bit, if you had that kind of recording, you could have the "average" listening level that might using about 12 bits. then the 16 bit is the loudest part. With 16 bit the very quiet sections would be pretty low bit resolution. It is the quietest parts of 16 bit that can benefit from dithering (for example long fade outs).

Now I'll go read that article and see if I need to come back and edit out the moronic from this post.

[HCMarkus]

My understanding is, when talking about PCM Audio, Bit Rate is synonymous with Sample Rate x Sample Size. However, it appears this term is generally used in connection with lossy compression algorithms like MP3. From Wikipedia:

In digital multimedia, bit rate often refers to the number of bits used per unit of playback time to represent a continuous medium such as audio or video after source coding (data compression). The encoding bit rate of a multimedia file is the size of a multimedia file in bytes divided by the playback time of the recording (in seconds), multiplied by eight.

The term Bit Depth is synonymous with Sample Size, which governs resolution and dynamic range. From Wikipedia again:

In digital audio using pulse-code modulation (PCM), bit depth is the number of bits of information in each sample, and it directly corresponds to the resolution of each sample. Examples of bit depth include Compact Disc Digital Audio, which uses 16 bits per sample, and DVD-Audio and Blu-ray Disc which can support up to 24 bits per sample.

In basic implementations, variations in bit depth primarily affect the noise level from quantization error—thus the signal-to-noise ratio (SNR) and dynamic range.

Bit depth is only meaningful in reference to a PCM digital signal. Non-PCM formats, such as lossy compression formats like MP3, AAC and Vorbis, do not have associated bit depths. For example, in MP3, quantization is performed on PCM samples that have been transformed into the frequency domain.

[HCMarkus]

http://xiph.org/~xiphmont/demo/neil-young.html

Interesting... x-ray audio anyone?

How about those 30kH 500kW radio stations? They keep me awake all night.

You are the Golden Eared person the researchers are looking for Bays!

[bayswater]

In the article, he appears to argue that the arguments about resolution differences is a mistake. The reproduction is perfect, and only dynamic range is limited, and not in any practical way. I have to assume getting analog resolution that isn't in the digital representation comes from interpolation in the D-A process.

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? In the original psychophysical experiments, a db was linked to a just noticeable difference (JND), so that would make it somewhere around 120. Apparently, we can hear differences of less and 1 db at some levels and frequencies. But my bet is no one can distinguish more than 4,000 levels.

[bayswater]

http://xiph.org/~xiphmont/demo/neil-young.html

Interesting... x-ray audio anyone?

How about those 30kH 500kW radio stations? They keep me awake all night.

You are the Golden Eared person the researchers are looking for Bays!

Along with most dogs.

[stubbsonic]

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? In the original psychophysical experiments, a db was linked to a just noticeable difference (JND), so that would make it somewhere around 120. Apparently, we can hear differences of less and 1 db at some levels and frequencies. But my bet is no one can distinguish more than 4,000 levels.

I'll be right back. I'm going to post a sine oscillator climbing in amplitude in increments of .1 dB. Let's see if we can hear that.

[bayswater]

I'll be right back. I'm going to post a sine oscillator climbing in amplitude in increments of .1 dB. Let's see if we can hear that.

If we can, we'll still only need 10 or 11 bits to represent it digitally.

[stubbsonic]

Ok. This is a 600 Hz sine recorded at 24 bit, 48K. It is 20 seconds long. It starts at -30 dB. Every two seconds it increases by .1 dB, the .2 dB, then .3 dB ... up to 1 dB in the last two seconds. I have NO idea what this proves. Just was curious.

http://www.jonstubbsmusic.com/misc/SineTest.wav

[stubbsonic]

http://www.jonstubbsmusic.com/misc/SineTest8bit.aiff

[mikehalloran]

Yet one more article by someone who hears with his eyes.

There are two fundamental flaws: Although the ear supposedly can't hear above 20kHz, the brain uses frequencies in the 22-28kHz range to locate stereo images. The other is that high frequency sounds are not sampled enough at 44.1 or 44.8k - this leads to the 'one hi-hat phenomena' where nearly all recordings make it sound the same. Although I have little patience for working in analog anymore, it's nice to hear real cymbals when I listen to a quality recording that has never been digitalized.

I recently did a hearing test where I discovered that I no longer detect anything above 12.5kHz. So what? I still know real cymbals when I hear them and enjoy quality analog through good 'phones. There is more to human hearing than an audio curve.

Just because he can't measure it, doesn't mean it can't exist. It can mean that his hypothesis and methods are flawed.

My friend, Rick Turner, is fond of saying, "CDs elevated mediocrity to a pretty good level."

I agree with much of what Monty writes in this article even if much of it is wrong.

The real test would be to take an analog machine that is fairly flat from 16-28k (I have one that I could loan) and run a series of tests with high end stereo mics. Compare with high end systems at 44.1/48/96/192. Now analyze that in double blind tests.