When I run the command line version of Replaygain on an album, it gives me an album gain of +4.26 db.

Doing the same thing (same files) from within Foobar yields an album gain of +3.2 db.

Why the difference?

I do know that Frank Klemm made some modifications to the original ReplayGain algorithm when implementing it for Musepack's commandline RG, while Foobar2000 uses the original method (now including on-the-fly resampling before processing where appropriate).

I think Frank intended to make Musepack's estimate more accurate, but I'm not sure if he modified the length of each "instant" in time, the Fletcher-Munson curve approximation or the choice of the 95th percentile loudness as the estimate of overall perceived loudness, something else or a combination of these steps in the calculation.

1 dB isn't very much - it's only just an audible difference at some frequencies. Don't be fooled by the number of decimal places shown in FB2K. Showing RG values to 0.1 dB resolution is already overkill.

I personally wouldn't worry too much about using values from commandline RG for MPC, but I usually use Foobar to scan my files whatever format (and my audio CDs at times).

FB2K's method of calculation is essentially the same as that used for mp3gain, vorbisgain and wavgain.

Edit: P.S. positive values like yours are quite unusual, implying that the original CD was <89 dB SPL. I wouldn't be surprised if you hear no difference at all between the values, because you may well have clipping prevention turned on, which might allow less than +3.2 dB of gain (e.g. if peak value > 0.6918 (or >22670 16-bit sample value)).

That was probably not a good example to use. It was from an audience DAT recording of a live concert.

Most of the time it's a negative value. More negative with Foobar. I was just concerned that the command line version might put me closer to clipping than when using Foobar. I probably couldn't hear the difference anyway.

I like using the command line version because I can scan a bunch of files while still using Foobar to listen to other stuff.

Either way, the great thing with MPC (and Vorbis) is that I can delete the values and do them over again if necessary.

It was only in old versions that Foobar's playlist interface became disabled when scanning RG (meaning you could only play stuff you'd already set going).

Since about FB2K version 0.6, I think, the RG scanning window has been a separate thread and an independent window, and I can put new files on the playlist and skip through my tracks, so I sometimes use it for playing and scanning at the same time, just by moving the window away from the FB2K playlist window. I guess there could be an issue with back-and-forth disk access slowing things down, but it seems to progress OK (and I usually have about 1000 ms of playback buffer, plus 320 KB in the Full File Buffering for slow devices)