Can anybody tell me the default settings in Audacity for ‘Normalize’?
I am using Audacity to edit a podcast (speech).
The settings shown in the screen grab on the right in the photograph are the default settings in my Audacity, whilst the settings shown on the left are the default settings in Audacity on another computer.
Usually, “Normalization” adjusts the volume for 0dB peaks (the “digital maximum” of 100% or 1.0 on the scale to the left of the waveform). But for some reason Audacity defaults to a slightly lower level of -1dB.
I have another application that calls it “maximize” which is a better English word, but “normalize” is the correct audio terminology.
The peaks don’t correlate well with perceived loudness and there is also Loudness Normalization. But with loudness normalization you can end-up pushing the peaks over 0dB (into clipping/distortion) so loudness normalization is often used with limiting to hold the peaks down.
There were several factors involved in that design choice.
I wish I could find it again, but a few years ago I read a very interesting article that looked at the performance of CD players, including some very expensive CD players. The surprising result of their tests was that around 9 out of 10 CD players suffered from non-linearity distortion slightly below 0 dB.
Another interesting phenomena is that some compressed audio formats (such as MP3) may slightly increase (or decrease) peak levels in the round trip of encoding ↔ decoding. This may cause clipping distortion if the audio is normalized to 0 dB prior to encoding.
Audacity’s Normalization calculates the audio level based on sample values. Assuming an “ideal” (“perfect”) D/A converter, the analog level will pass in a smooth continuous line through every sample level, which means that analog peaks may be slightly higher than the sample values at the crest of peaks. Some audiophiles argue that the “True Peak” (analog peak) should not exceed 0 dB, so the “digital peak” (sample value) should always be below 0 dB. (Personally I think this argument is spurious because D/A converters can be designed to handle “True Peak” levels greater than 0 dB, even with integer sample formats).
In terms of “perceived loudness”, -1 dB is almost indistinguishable from 0 dB, so -1 dB was adopted by the Audacity team as a reasonable compromise between “maximum level” and “safe level” when normalizing.