Re: [linux-audio-user] Fwd: [Jamin] Re: soft clip: Achieving Gain, inconsequential overloads

On Sun, 2004-04-11 at 19:03, R Parker wrote: No, normalization only scales the data points so that relative amplitude is maintained. The WHOLE file is scaled so that -3db becomes -1db, -6db becomes -5db, -9db becomes -7db, etc. This does raise the overall volume by 2db, but it won't normally get you all that much volume if there are a lot of dynamics in the file or else you'll clip those peaks. Normalization is more like an after the fact tweak of the gain control on the preamp before you recorded. Insted of the people being at -3db, you can 'turn up the volume' so that same peak becomes -1db. What you are asking for is exactly what a compressor does, but the way you ask the question above is in reverse from the way it works. Think of compression (for now just a single band that covers the complete audio range) as first raising the valley's from -12db -6db, and then looking for the peaks and shrinking them down so they don't clip. When the energy gets too high the compressor 'turns down the volume quickly' (the attack setting) and when the energy goes down it turns the volume back up. (the release setting) The 'range' setting is basically the amount you want to turn the volume down (compress) on the peaks. In the old analog synth days (and still in the world of analog compressors) the signal does through a voltage controlled amplified. The voltage control comes from a circuit that looks at the amplitude of the input, among other things. Yes, normalization creates a new file with new data values. It needs to be non-real-time since it has to look at the whole track to understand how to scale all the data values. I really don't think normalization is going to be all that important to you in this case. It's more useful (IMO) when a specific track is recorded low, or came from a source that maybe you didn't have control of, but it's really nothing more than a volume control. Turn it up too much and you still get clipping.