Ok, I performed 3 decode tests last night on an album of MP3s and here's what was found:

• Test 1 From A Suggestion With Strong ATH Noise Shaping
  
  -Used FB2K Convert
  -Disabled DSPs
  -WAV (PCM 16bit dithered)
  -Used ReplayGain
  -Album Gain (with clipping prevention, and pre-amp set to 89.0 dB)
  -Dither (Strong ATH noise shaping)
  
  Results:
  
  .WAV files sounded clear but were low in sound (lower sounding than their MP3 counterparts)
  
  
• Test 2 Based On Above Suggestion With No Noise Shaping
  
  -Used FB2K Convert
  -Disabled DSPs
  -WAV (PCM 16bit dithered)
  -Used ReplayGain
  -Album Gain (with clipping prevention, and pre-amp set to 89.0 dB)
  -Dither (NO noise shaping)
  
  Results:
  
  .WAV files sounded clear but were again low in sound (lower sounding than their MP3 counterparts)
  
  
• Test 3 Something I Thought To Try
  
  -MP3 Gained All MP3s first (using default settings) & "Album Gain"
  -Opened files that I used MP3 Gain on in FB2K
  -Replaygained files in FB2K using 1st test suggestions (Strong ATH noise shaping)
  
  Results:
  
  .WAV files sounded clear and were noticeably louder; they were actually a bit louder than their MP3 counterparts

Does any of this make sense and should my results have been as such?

My question at this point is, should I needed to have MP3 gained AND Replaygained in order to get a good sounding .WAV that was equally or more loud than its MP3 file? Just seems like extra work in getting this outcome.

Perhaps I should change some setting in using ReplayGain in FB2K so that I won't have to MP3 Gain first (and have my files be permanently changed) and could just do this whole process in FB2K?

Thanks...

yup...


BWT:
well. am more interested in why there`s
so huge diference in mp3 gained,
and wav gained.. at the same setting.

it is huge diference in the files.. (format. compress/none)
but i theory.. the gaining prosess..
should give the same overall volume..

right..?

Hi n68:

I didn't do any WAV gains, just MP3 gains.

I'm just wondering though if my outcomes should have been what they were and if after gaining (regardless of the method used), if the WAVs should be loud or louder than their MP3 counterparts?

I'm thinking that the 3rd test I tried was overkill. I did that 3rd test assuming that any WAVs I create from MP3s should be just as loud as their MP3s. I don't know though if that should be the goal, along with removing any clipping.

I'm trying to get this process down smoothly for decoding MP3s to high quality WAVs and then burn the WAVs to CD-Rs for DJ use. Nothing more, nothing less, just want this to be a simple, smooth process.

In essence the FB2K ReplayGain is like a WAV gain - it's done after decoding, but at high resolution before dithering back to 16-bit to reduce rounding error.

It's quite normal for mp3gain to make mp3s quieter than the original album. That's the whole point of it (because most modern albums are mastered too loud).

[EDIT: I stand corrected. The 26 ms thing isn't true - see Snelg's post below] Also mp3gain bases its assessment of loudness on 26 ms (1 frame of MP3) chunks of sound then can only apply gain in 1.505 dB steps. FB2K bases it on 50 ms, I believe, so there are slight differences. Also, mp3gain can only apply either track or album gain. With FB2K you choose at the time of playback.

Unless the album was clipping after mp3gain, there shouldn't be very much difference (very subtle due to 1.5 dB steps of mp3gain). Of course, you had to have scanned the whole album as an album in FB2K for it to playback in album mode, otherwise it will just use the track gain it calculates automatically when you play the tracks. You can see whether it has album gain tags using the Info tool. Also, note that the latest mp3gain beta will write the tags, too, but it will have used its own calculations, not FB2k's. FB2K might or might not recalculate for files that are already RG-tagged.

The choice of dither is probably inaudible in 16-bit audio. "No noise shaping" will be theoretically hissier than "strong ATH noise shaping", but as 2Bdecided pointed out elsewhere, if DJing you'd better not use strong ATH if it involves pitch-shifting, unless the CD is low-pass filtered at about 16kHz in the pitch-shifter before shifting. That's because the HF dither noise of Strong ATH Noise Shaping would be brought into the audible band when shifting down in pitch.

Actually, mp3gain uses 50 ms windows. I wrote the initial ReplayGain functions as generically as possible, not for a particular audio format. Frank Klemm did some clever optimization to my original code, but after that everyone else has used pretty much the same template (at least as far as I know). Frank played around with the window sizes, but I haven't seen any conclusive "this is better" changes, so I haven't fiddled with that part of the code for quite some time.

And yes, the current BETA version of MP3Gain writes the ReplayGain info in tags. I intentionally mimicked the format (and a good chunk of the source code) that foobar2000 uses for these tags. When writing the tags, MP3Gain does not "round" the ReplayGain values to 1.505 dB steps. So you should get identical (or at least practically identical) results in the tags using either program.

-Glen