The extended gain analysis tables in Foobar2000 (and the derivatives that copied it) are wrong.
Here is how to show that it's wrong. Use <http://www.daniweb.com/software-development/python/code/263775> to create a 1 kHz signal using a 48 kSamples/sec rate (48.wav) and and a 192 kSamples/sec rate (192.wav). I modified the script to generate 2s of samples. Because the underlying signal is the same (1kHz with a fixed amplitude) the perceived loudness should be identical, independent of the sampling rate. However, using the Foobar2000 tables, the computed values for the 48 kSamples/sec case is: REPLAYGAIN_REFERENCE_LOUDNESS=89.0 dB REPLAYGAIN_TRACK_GAIN=+4.11 dB REPLAYGAIN_TRACK_PEAK=0.12207031 REPLAYGAIN_ALBUM_GAIN=+4.11 dB REPLAYGAIN_ALBUM_PEAK=0.12207031 while the values for the 192 kSamples/sec case is: REPLAYGAIN_REFERENCE_LOUDNESS=89.0 dB REPLAYGAIN_TRACK_GAIN=+1.40 dB REPLAYGAIN_TRACK_PEAK=0.12207031 REPLAYGAIN_ALBUM_GAIN=+1.40 dB REPLAYGAIN_ALBUM_PEAK=0.12207031 The difference between the +4.11dB and +1.40dB gains is surprising for essentially the same underlying signal. The problem is that the number of taps in the filters is insufficient at the higher sampling rates (192 kSamples/sec) to produce a good result for the comparatively low frequencies at which the filter operates. Rather than generating a more complex filter with more taps, use the fact that most of the higher sampling rates are either multiples of 2 or multiples of 4 of the sampling rates already determined: 192000 = 4 x 48000 176400 = 4 x 44100 144000 = 4 x 36000 128000 = 4 x 32000 96000 = 2 x 48000 88200 = 2 x 44100 64000 = 2 x 32000 This is likely to be perfectly acceptable since the ReplayGain adjustment is purely an estimate of loudness. The only values in the Foobar2000 table that are not accounted for are: 112000 = 4 x 28000 56000 = 2 x 28000 To accommodate this, a 28 kSamples/sec entry is required in the base table. I'll save this for a separate patch. ________________________________ From: Earl Chew <[email protected]> To: "[email protected]" <[email protected]> Sent: Saturday, January 7, 2012 9:23:35 PM Subject: [Flac-dev] Support 56kHz to 19.2kHz gain analysis (Patch v2) Revise patch to correctly allocate enough buffer for gain analysis. ________________________________ From: Earl Chew <[email protected]> To: "[email protected]" <[email protected]> Sent: Saturday, January 7, 2012 11:12:17 AM Subject: [Flac-dev] Support 56kHz to 19.2kHz gain analysis Copy additional filter values from Foobar2000 as found in <http://code.google.com/p/sirens2/source/browse/trunk/libwavpack-4.32.psp/wvgain.c?r=32> to allow metaflac to perform gain analysis on high sample rate audio. _______________________________________________ Flac-dev mailing list [email protected] http://lists.xiph.org/mailman/listinfo/flac-dev _______________________________________________ Flac-dev mailing list [email protected] http://lists.xiph.org/mailman/listinfo/flac-dev
flac.diff
Description: Binary data
_______________________________________________ flac-dev mailing list [email protected] http://lists.xiph.org/mailman/listinfo/flac-dev
