QUOTE (Yamabushi @ Apr 3 2005, 09:46 AM)
Otto, thank you for your informed response. I guess what I'm getting at is that I want to better understand the difference between creating an AAC file with Diskwriter with the ReplayGain button engaged vs not using the replaygain and then as a second step using AACGain. I believe there is a difference but am not 100% sure what it is. So... what if any is the difference in the resultant lossy file.
Okay, conceptually, the diskwriter is basically transcoding the file. It converts the file to WAV, then compresses it using whatever format you want. The important bit here is that it always first goes to WAV (even if this WAV file is only stored in memory and never gets written to disk).
When you tell the diskwriter to use the ReplayGain information, it will scale the intermediate WAV file according to the ReplayGain adjustment needed. It's like WAVGain in this respect. The scaled WAV is then what gets passed to your compressor.
MP3/AACGain, on the other hand, work differently. They adjust a parameter in the lossy file itself which changes the volume, in 1.5 dB increments. They also can write metadata to the file that allow fobar to deal with it properly and so forth.
Both of them actually change the song data itself. Diskwriter does it before compression, MP3/AACGain does it after. Of these two, Diskwriter's method (WAVGain) is a lossy process. There is always some round off error involved in scaling a WAV file. This error is minor, but additive. MP3/AACGain is a lossless process. It's only adjusting the global gain values, and it's doing so in a fully reversible way. However it is limited to 1.5 dB adjustments, where the other one is not.
The ideal approach is to rip the CD bit for bit, and do absolutely nothing extra to the data. The data from the CD is as good as it will ever get, so you want to do as few lossy things to it as possible. Compressing to lossless is what you have done, so this is not harmful. Then you want to compress to lossy. This is where you say you get clipping, but again, I ask, are you really hearing clipping? Just because MP3/AACGain says that it clips doesn't mean that it's really audible.
When you rip a modern CD, especially a rock/alternative/pop one, the tracks are quite likely to be somewhat clipped already. Not a lot you can do about it. However, this is not clipping that a program will be able to see easily. All it can see is that several samples in a row are at digital full scale. The actual clipped peaks are gone. When you then do lossy compression on this track, the resulting MP3 or AAC might indeed be clipping. What clipping means, in this context, is that if you decode the MP3/AAC, that some samples will be above digital full scale and will thus be clipped. But this is an artifact of the encoding process itself. The waveform you get out of the lossy compressed file is not the same as the one you put in. Those clipped sections will be approximated, and yeah, you can end up with a peak higher than full scale. Does this mean it's clipped? Sure. Does this means it sounds different than the original? Not at all.
And then clipping can be minor too. If on decoding an MP3, one sample comes out at digital full scale + 1, well, that's technically clipping, but there's no way in hell anybody could ever hear it. But MP3Gain/AACGain, these will report the thing as clipping nonetheless.
If you listen to a song, and hear no problems, and then MP3Gain/AACGain tell you it's clipping, don't worry about it. What you hear is more important than what the program tells you. The real reason MP3/AACGain warn about clipping is because you can adjust the gain of a song upwards and cause clipping when there was no clipping before, and this is usually very, very bad sounding.
Okay, so this is kinda long, but the short of it is that the foo_pod clienc really doesn't need RG in it, because the RG->SoundCheck functionality takes care of it, and applying RG before the compression process is not the best way to do it anyway.