Ive heard some people say that preset standard has some bloat and the v1 or v0 settings have less bloat. Is there any truth to this? How could this be? Isnt the lower quality setting supposed to give less data per sample than the higher setting meaning less bloat? Pls explain

Also, some posts on hydrogen say that on lame 3.90.3, the fatboy sample sounds better with the standard setting than extreme, Is this still the case with 3.97b1?

While I can be wrong, I belive that these people meant is that preset standard ( v2 ), can have a higher increase of bitrate in specific type of samples compared to v1 and v0. This doesn't mean that v2 would use a higher bitrate than v1 or v0, but rather that it can reach values similar to those of v1 and v0 in the special cases where this occurs.


In numbers:

Normal V2 ~ 190kbps
Normal V1 ~ 220kbps
Normal V0 ~ 240kbps

"bloated" V2 ~ 220kbps
"bloated" V1 ~ 240kbps
"bloated" V0 ~ 260kbps

So doesn't this actually make the encoder more efficient since it might be giving complex passages more data and less to the simpler ones? More range in bitrate

I think you're referring to sfb21(?) bitrate bloat. I saw it referrenced in a test done by Guru and it occured with -v 2 and higher settings. For a test, I loaded up my flac/cue archives, replaygained them, then converted, via Foobar and the command line encoder, to -v 2 --vbr-new and -v 3 --vbr-new. I also did the same thing, but with no replaygain on the flac/cue combo. Results...-v 2 came out with a smaller filesize when the gain was applied. IIRC, there was no change with -v 3.

To read more about it, search for a thread that's something like "Wavegain vs. mp3gain".

what does this mean?

Well, for a much better explanation I suggest you check out the thread I referrenced. Guru did a test and plotted all the codecs on a graph. -V 2 and above experienced this "bitrate bloat".

The bloat you're thinking about is most likely the sfb21 bloat that is possible with -V2 through -V0. Due to faults in the design of the Layer III format itself encoding audio of 16kHz and higher will cause more bits to be needed than normal (it's complicated and I don't know the easiest way to explain). Metal, a genre I have a lot of, has this problem sometimes. -V3 through -V9 do not have this problem as they have the -Y switch activated by default. This switch basically makes it where if the data above 16kHz will bloat the bitrate it is not encoded.

Basically instead of the 190-200kbps I get files in the 220-230kbps range with -V2 sometimes.

The WaveGain vs MP3Gain thread clintb brought up shows how lowering the gain pre-encode with Wavegain brings the data above 16kHz to a less audible level and it is most likely not encoded. The reason -V3 wasn't affected much if at all was because the -Y switch was already doing the same thing. Since MP3Gain adjusts the audio volume post-encode it has no effect like WaveGain in lowering the bitrate.

Adding to what DreamTactix291 has done so much better than myself...

The bitrate bloat can be avoided by converting a ReplayGain'ed lossless file, via Foobar, to a -V 2 and above LAME mp3.

I have the following:
1. Single FLAC/CUE files.
2. The .cue files are loaded into Foobar and an album based ReplayGain is done to each.
3. In Foobar preferences, go Components>Diskwriter>Processing>"Use ReplayGain"...check this box.
4. Setup your commandline encoder, LAME in this case and convert away.

By doing the ReplayGain analysis on a lossless file it seems to be doing the same thing as WaveGain.

Ah I forgot the Replaygain way with foobar2000. I'd wager to say that's the advantageous way of doing it since your lossless files aren't actually altered. Unlike Wavegain which is not a reversible procedure.

The resultant temporary wave file is still affected much like the wavegained file.

Yes, but since you want your resultant mp3 to be gained down it's alright. And like I said your original losslessly compressed file is still unaffected in this case. That's where the advantage over Wavegain itself comes in.

So basically you guys are saying that v3 is more efficient than v2 and I will be saving more disk space for a minimal audio difference using it

You can also add -Y to V2, V1 or V0 to make it more "efficient" WRT sfb21 if you want better quality than V3.

Yes, But What Will It Degrade??? High Frequencies?

DreamTactix291 has already explained what the sfb21 is, and pointed that it affects the frequency band from 16Khz to 22khz (in a 44.1Khz source, the bandwidth is from 0Hz to 22Khz, so this means the Higher end)
He also added a brief description of what the -Y switch does.

-Y tells the encoder to behave differently on this band, reducing its quality level. It is not a filter, it just selects differently what to encode.

Generally, this translates in the said reduction of bitrate (read: using the expected bitrate, not the extra "bloated" one), and no noticeable change in sound quality, because the whole sound of the song masks the differences.


Edit: Btw... my initial comment was refering to this phenomenon too. I just preffered not to enter in the technical aspect of sfb21.

actually, -Y tends to prevent sfb21 from being encoded at all.

the only stuff you'll get above 16k with -Y is in short-blocks (which don't have an sfb21 to bloat. i think short blocks have 15 scalefactor bands rather than 21).

No, no and no.

oops, sorry. how does it work then?

I have a couple questions. I'm able to see the 'bitrate bloat' between v2 and v3 after running some conversion tests. I've tried v2 with the -Y switch and it does seem to bring the bitrate down quite a bit.

So if I'm understanding this correctly, I can replaygain my FLAC files prior to encoding them to MP3 in foobar using v2 without the -Y switch and this will give me essentially the same results as just using the -Y switch?

by encoding mp3s from replaygained FLAC files, what effect does this have on the resultant mp3 file? Will that file not require replaygain info?

OK, I ran a conversion test on an album I have in FLAC format and converted to MP3, I've listed the average bitrate for each song on the album (this happens to be a country album):

using lame 3.97b2

using V2 only on non-replaygained FLAC files:
193
199
187
207
207
185
172
199
211
219
200
212
190
188


using the -Y flag on non-replaygained FLAC files:
159
160
161
171
173
159
152
166
166
164
158
171
160
161

using v2 on replaygained FLAC files without -Y switch:
170
175
166
181
188
172
154
179
193
190
171
188
173
173

Not quite. There are two different issues. One is the -Y switch, the other is transcoding from replaygained Flacs.



Correct. The Flac file output will be scaled to replaygain standard before being input to mp3 encoder. The mp3 file will turn out with no replaygain tags, but already at the replaygain level anyway, so it doesn't matter.

The bitrate decrease you saw by encoding from a replaygained FLAC is because some signal fell below the ATH due to replaygaining, therefore the mp3 encoder did not need to process it. Not to worry, that means it would have been inaudible anyway.

This is a different and unrelated bitrate savings from using the -Y switch during mp3 encoding.

Thank you for the info, I'm starting to understand. I like the idea of pre-replaygaining my FLAC files prior to mp3 transcode, will offer some benefits in my application.

If I go that route and want all my mp3s encoded the same way sounds like I'll have a couple extra steps because some of my sources are FLAC, but others are direct from CD. I had planned to use EAC to rip and convert directly to mp3 but sounds like I'd need to rip to FLAC first.