A Lesson in Dynamic Range (or Why 32 Bits per Sample Should Never Catch On)

re: A Lesson in Dynamic Range (or Why 32 Bits per Sample Should Never Catch On)

Tricky — Wed, 23 Aug 2006 10:55:25 GMT

Ok, I'm not an audio guy ... but I can see how extreme audiophiles can benefit from 32 bits per signal sample.

Why have the 6.02 dB increment per bit? Why not reduce this amount, or make it variable?

I'm asking instead of going the standard way it's done 6.02dB per bit .. why not implement so that 32 bits per signal sample doesnt have to mean that you need to go up to 192dB.

That is, sample only the range of human hearing. low end is 1 dB ? and then the upper end .. 120 dB?

so then you split that 119 db into 4 billion slices (aka a 32 bit number). If you know the regions that human hearing is best suited for you can increase the resolution in those predetermined regions 4 dB per bit? (losing resolution at the less significant zones 7 dB per bit?).

re: A Lesson in Dynamic Range (or Why 32 Bits per Sample Should Never Catch On)

Vorn — Wed, 23 Aug 2006 21:28:09 GMT

the 6.02dB increment per bit is not a thing that /can/ be changed. a 6.02 dB increase is /exactly/ a doubling in sound value, and so doubling the number of values you have also doubles the amount of dynamic range you have, which is what's being looked for.

On the other hand, there are sort-of workarounds, if you're willing to do strange things - imaging technology uses a value called gamma, which is essentially a power to which all the input values are raised to when being output. However, sound is different: the input is a signed value, so raising it to a power is unacceptable (a sine wave, squared, gets its frequency doubled; raising a negative number to a noninteger power gives a complex number). Even if we change the baseline, applying a gamma-like function will create distortion.

On the other hand, one of the most common formats for digital audio is a 32-bit float, which is essentially a 24-bit number with 8 bits of gain. Though the gain is rarely used, it allows you to get something along the lines of 256-bit range - or about 1,500 decibels - without actually dropping 256 bits of data on each piece of audio. Not that you'll ever find a microphone that can cover a range that large.

Vorn

re: A Lesson in Dynamic Range (or Why 32 Bits per Sample Should Never Catch On)

Vorn — Wed, 23 Aug 2006 21:45:23 GMT

er. more than that. 256 + 24 = 279; 280 * 6 = 1680 dB.

Vorn

32 bit audio redux

The Audio Fool — Thu, 24 Aug 2006 02:44:18 GMT

In my previous post, I don't think I explained very well why a 32-bit signal wouldn't work on the low-end.&nbsp;...

Audio Fidelity: Output Level

The Audio Fool — Tue, 05 Dec 2006 06:10:43 GMT

Output level is one of the simplest fidelity metrics to understand, but don't take that to mean it's

re: A Lesson in Dynamic Range (or Why 32 Bits per Sample Should Never Catch On)

Greg — Tue, 10 Jul 2007 18:55:45 GMT

The technique Tricky mentions is of course possible. Something like it has been used in telecomm for years where more bits really hurts because of limited transmission BW. e.g. mu law and A-law. Vorn's says 6dB per bit can't be changed, which is true if you are using standard ADC chips when the original music is captured. If instead a whole new digital audio infrastructure was in place utilizing a new analog to digital map... not a small undertaking, and not likely to ever happen. The masses are going the other direction - MP3/AAC as opposed to say SACD. Of course AAC uses techniques much more sophisticated to increase effective resolution per bit per second.

re: A Lesson in Dynamic Range (or Why 32 Bits per Sample Should Never Catch On)

Arcticboy — Wed, 22 Aug 2007 14:03:07 GMT

This is interesting!

I follow your reasoning. And there is a but, of course...

If you want to mix a lot of signals in the digital domain, you need to have a great dynamic headroom simply to make space for all the signals (BITs) present at any one time. This is why some audio manufacturers even work with 48 bit internal resolution.

By the way; the physiological/electrical limit for dynamic range in the anaolgue domain is around 134 dB, so Playing out anything with a greater dynamic range is futile. Electronic components will distort.

2007 year-end link clearance

The Old New Thing — Mon, 31 Dec 2007 19:53:26 GMT

Random stuff.

MSDN Blog Postings » 2007 year-end link clearance

MSDN Blog Postings » 2007 year-end link clearance — Mon, 31 Dec 2007 20:27:20 GMT

PingBack from http://msdnrss.thecoderblogs.com/2007/12/31/2007-year-end-link-clearance/

2007 year-end link clearance

The Old New Thing — Wed, 02 Jan 2008 11:06:16 GMT

Random stuff.

Digital Audio Series « xosfaere

Digital Audio Series « xosfaere — Sun, 06 Apr 2008 21:14:15 GMT

PingBack from http://xosfaere.wordpress.com/2007/07/06/digital-audio-series/