Audio Fidelity: Dynamic Range
I want to talk about noise for a minute. We all know what noise is, and that we don't want it in our audio. Unfortunately, noise is always present, whether we want it or not. In pseudo-technical terms, noise is random unstructured variations in a signal. In a digital signal, that variation comes from quantization, the breaking of the nice smooth wave into distinct integer-sized levels. This introduces a kind of stair-step to the signal, causing variations, and hence, noise.
But what about analog waves? No quantization means no noise, right? Remember how sound is transferred. Air pressure waves reach your ear and make it vibrate. On the smallest level, that pressure wave consists of millions of individual air molecules bouncing off your eardrum, resulting in a type of natural quantization. Even an electrical signal is made up of distinct electrons.
So there's always noise. Always. The amount of background noise can vary, but it's there no matter what. The only real question is how loud the noise, whether you're in a quiet library or in a 747 right over the engine. (well, there's also how your noise is shaped, but let's not get ahead of ourselves). The technical term for the level at which the noise drowns out all other signals is called the noise floor, and it determines exactly how quiet a signal can get before it is lost.
On the other end of the spectrum is the output level, or reference level. The exact definition of the reference level varies depending on whom you ask, but for simplicity sake, it's the loudest useful amplitude that a signal can have. (This is generally a function of where the distortion passes a certain threshhold, or a certain maximum-safe-wattage at a speaker, or, in some cases for digital audio, it's just defined as all ones)
So what is Dynamic Range? DR is simply the ratio between the reference level and the noise floor, expressed in dB. DR is often called Signal-to-Noise Ratio (SNR) by novices and marketers (but not engineers - more on that in a later post). The dynamic range of a signal describes the ratio between the loudest point of the signal and the noise floor recorded in the signal, usually introduced by the recording equipment. After you've captured a recording, the DR of your signal effectively limits how much you can amplify it, because you amplify the noise floor with it. Obviously, it pays to keep the noise low and the DR high when you record a signal.
DR is equally important in your playback equipment. The more DR you have, the more you can amplify (or attenuate) a signal and still reproduce it correctly. When you play a signal, your playback equipment absolutely must have at least as much DR as the signal. If you don't, then either the quieter parts will simply drop out of the signal (they get swamped by the noise floor), or worse, the louder parts will distort and become unintelligible and ugly (or worse still, both at once). Either way, not a fun experience.
Dynamic Range is sort of like oxygen for audio. You don't usually have to think about it, but if you can't get enough, it can really ruin your experience.
Comments
- Anonymous
June 27, 2006
The comment has been removed - Anonymous
June 28, 2006
BobTurbo: Right you are. The state of PC audio has been absolutely atrocious for years, and on any given WinXP system, you're likely to hear several of the artifacts I'm describing in these articles. With the rewritten Vista audio stack, the bottleneck is not in the software anymore, but the hardware.
The reason I write these articles is to bring attention to audio fidelity and raise awareness of where PC audio should be. This is also why, starting with Vista, Microsoft is enforcing a minimum fidelity requirement for any systems or devices that want the "Made for Windows Vista" logo. If your audio sounds as bad as most WinXP devices, then you can't get a Vista logo. - Anonymous
August 07, 2006
Distortion in audio is very closely related to noise. Both  "distortion" and "noise" are used... - Anonymous
October 18, 2006
Quantization adds noise. Taking a nice continuous signal and expressing it as distinct integers will