Volume Gain and Normalizing - Artifacts Introduced? -Sound Forge - PC

Subject:Volume Gain and Normalizing - Artifacts Introduced? Posted by: rockrev Date:3/18/2003 8:40:57 AM One quick question - when using volume gain or normalizing to a level BELOW zero db, are inaudible/audible artifacts introduced? And what if you do these processes more than once (and still remaining below zero)? Also - if I use Volume Gain to raise my peaks to -4db, is it the exact same as using Normalize to -4db? By exact I mean will the waveform have the same exact properties? Thanks, rockrev
Subject:RE: Volume Gain and Normalizing - Artifacts Introduced? Reply by: Chienworks Date:3/18/2003 9:28:51 AM Manipulating a digital file is always going to make some changes from the original that don't preserve perfect fidelity. This is due to the "granular" structure of digital information (i.e. you can't have partial bits). These differences can also be cummulative if several processes are performed, therefore it is always better to process less. Also, the fewer bits you have to start with, the less resolution is available and the greater these errors can be. Having more bits makes these errors smaller. How much processing can you do before these artifacts become a problem or noticeable? That depends on lots of things including the material you're working with, what sorts of processing you're doing, how well it was recorded (does it make good use of the most bits available?), and what your tolerance for artifacts is. We can't really tell you "don't do any more than 4 processes". Normalization is nothing more than a volume control that automatically calculates the change for you. Once the change has been calculated, there is no difference between the normalization process and the volume process.
Subject:RE: Volume Gain and Normalizing - Artifacts Introduced? Reply by: Sonic Date:3/18/2003 12:21:34 PM Volume and peak normalization are the same thing, just that normalize scans for the proper adjustment for you. Moderate to high rms normalization often will introduce gains that drive single peaks over 0 dB. The Normalize tool lets you choose what you want to do in this case, i.e. bail, clip, normalize to 0, or, most typically, compress, which definately changes the characteristics of the audio. But Chienworks is right in that, any attenuation can effectively reduce the dynamic range, while a gain leaves it pretty much unchanged (assuming it doesn't clip anything), but both can still introduce small changes due to the limitations of finite arithmetic and re-quantization. So if I take a piece of quiet audio and go -6, -6, -6, +6, +6, +6, I am not guaranteed to get the same thing I started with unless the intermediate bit-depth has a significantly higher precision than the source and/or output bit-depth. Sound Forge actually has a feature in preferences called "Use floating point temporary files" which retains much higher precision values between operations (at the cost of disk space and some performance). You may want to check it out. J.

Subject:Volume Gain and Normalizing - Artifacts Introduced?

Posted by: rockrev

Date:3/18/2003 8:40:57 AM

One quick question - when using volume gain or normalizing to a level BELOW zero db, are inaudible/audible artifacts introduced? And what if you do these processes more than once (and still remaining below zero)?

Also - if I use Volume Gain to raise my peaks to -4db, is it the exact same as using Normalize to -4db? By exact I mean will the waveform have the same exact properties?

Thanks,

rockrev

Subject:RE: Volume Gain and Normalizing - Artifacts Introduced?

Reply by: Chienworks

Date:3/18/2003 9:28:51 AM

Manipulating a digital file is always going to make some changes from the original that don't preserve perfect fidelity. This is due to the "granular" structure of digital information (i.e. you can't have partial bits). These differences can also be cummulative if several processes are performed, therefore it is always better to process less. Also, the fewer bits you have to start with, the less resolution is available and the greater these errors can be. Having more bits makes these errors smaller. How much processing can you do before these artifacts become a problem or noticeable? That depends on lots of things including the material you're working with, what sorts of processing you're doing, how well it was recorded (does it make good use of the most bits available?), and what your tolerance for artifacts is. We can't really tell you "don't do any more than 4 processes".

Normalization is nothing more than a volume control that automatically calculates the change for you. Once the change has been calculated, there is no difference between the normalization process and the volume process.

Subject:RE: Volume Gain and Normalizing - Artifacts Introduced?

Reply by: Sonic

Date:3/18/2003 12:21:34 PM

Volume and *peak* normalization are the same thing, just that normalize scans for the proper adjustment for you. Moderate to high *rms* normalization often will introduce gains that drive single peaks over 0 dB. The Normalize tool lets you choose what you want to do in this case, i.e. bail, clip, normalize to 0, or, most typically, compress, which definately changes the characteristics of the audio.

But Chienworks is right in that, any attenuation can effectively reduce the dynamic range, while a gain leaves it pretty much unchanged (assuming it doesn't clip anything), but both can still introduce small changes due to the limitations of finite arithmetic and re-quantization.

So if I take a piece of quiet audio and go -6, -6, -6, +6, +6, +6, I am not guaranteed to get the same thing I started with unless the intermediate bit-depth has a significantly higher precision than the source and/or output bit-depth.

Sound Forge actually has a feature in preferences called "Use floating point temporary files" which retains much higher precision values between operations (at the cost of disk space and some performance). You may want to check it out.

J.