Skip to main content
Topic: Alternative ReplayGain (Read 671 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Alternative ReplayGain

Hi, what's the exact purpose of Alternate RG scanner, is it only the improved speed (what;s FFT good for?), and can I choose which one to use or does it completely replace the standard RG scanner?
Alternative ReplayGain engine. Should be faster as it does not use FFT. Please note that reference frequency (set at -20dB) is 2kHz, and not 1kHz, so gains are generally higher than standard algorithm. You may want to use a -6dB or lower preamp level with classical music if you dislike limiters/compressors.


Re: Alternative ReplayGain

Reply #1
Bumped because I'm also interested in this, since it is indeed several times faster.

foo_arg appears to replace the builtin RG scanner, but you can turn it off from Advanced > Tools > Alternate ReplayGain > Do not use. Other than the verbose logging, it does not seem to have any other configuration options. Since it writes onto the same tags as the original scanner, I'd rather not use it on permanently kept files.

FFT (Fast Fourier Transform, just google it) analyses a signal into frequency and phase components. Together with windowing, this is often used for spectrograms and such. My confusion here is that FFT isn't involved in the EBU-recommended ITU -R BS.1770 normalization algorithm, nor in the libebur128 implementation of it that's supposedly used by foobar2000. So I don't really know where the speedup comes from.

As for accuracy: I did some testing on my own (foo_arg 1.0), and it seems like foo_arg produces lower (or is that higher?) gain values than the standard one: on my data, -11.4 dB vs -10.6 median / -10.5 dB vs -10.0 mean. For the absolute differences, the standard deviation is 1.26, but the statistical tails are slightly uneven. I'll leave it to you to test how noticeable the difference is in listening.

SimplePortal 1.0.0 RC1 © 2008-2019