Just to test out a hunch, I generated a 1kHz tone at full amplitude (48kHz sampling rate), then WaveGain, VorbisGain, and MPC ReplayGain'd it in the respective formats...

WaveGain and VorbisGain both come out with a -14.16dB adjustment, while the MPC ReplayGain comes out with -14.10dB adjustment. I realize this is completely negligible in this case, but if the problem is with the implementation of the algorithm, could there be problems with the adjustment on real music?

I also realize that ReplayGain is still in major development, and many members of this board don't fully support it... But I find it particularly useful for transcoding MPCs to portable players, and am just curious about the differences in the results... is there any "perfect" implementation?

Frank Klemm has made some enhancements to replaygain for mpc - that's why they don't give exactly the same value. I think there was talk about merging these changes into wavegain and vorbisgain, but apparantly it hasn't been done yet. Anyway, when the difference is that small it's impossible to hear it.

Yeah, I read that the human ear can't detect a 1dB difference in volume...

But, looking at some of the results I've gotten in transcoding from MPC, VorbisGain usually ends up with a gain of at least -2 to -3 dB beyond what the original MPC value was... Is this because of the transcoding? What with Vorbis's current low-q high-frequency boost... =P

Anyway, just asking. =)

See my little ReplayGain analysis in this thread:

http://www.hydrogenaudio.org/forums/index....t=ST&f=1&t=7981

It's more like 0.1 dB, as a "round" number. I believe 0.3 dB have been detected in blind tests.

Another reason for different result is that current MPC replaygain scanner thinks that all files use 44.1kHz samplerate.