I really enjoy listening to all my mp3's at the same level after using mp3gain. However, I was wondering anyone could tell me why there is clipping on my original mp3's? (link below has before and after I used the program) Under options I have "Don't clip while doing track gain" Is that the correct setting? I just have read some post about mp3gain and I am confused at what clipping is and is mp3gain making my mp3's worse sounding. I am a basic user, so please try not to get super technical on me. Thanks for your help. Marc

http://home.comcast.net/~m.gibbs/mp3gain.jpg

All songs have a volume peak somewhere.

On CDs, this peak has a ceiling of "1.0", but when you convert tracks to MP3s, then the ceiling goes away. Also, the conversion process can increase the peak.

If the peak turns out higher than 1.0, this might (or might not) result in distortion or noise. This is what clipping is.

You can get rid of the clipping by lowering the volume of the MP3.

I've never really had cases where this clipping was a problem, though.

The only times clipping was a problem for me was when the original CDs had clipping! Eeeek!

you mean, the WAV technically clipping... or not techically clipping but peaking at 0dB on almost all samples? (eg. a very clip pressed song)

For clipping at source (mastering) versus clipping at "user-end" i.e. lossy decoding have a look here.
A useful explanation of the clipping via decoding ("user-end") is here.

C.

In my opinion, this is the correct setting. What does the setting do? It will decrease the adjustment to below the reference value (89 db) only in these cases where a correct adjustment would cause clipping. Simpler said: it will make the track a bit less loud than it should be, because if it were adjusted to the correct loudness, some clipping would occur.

This setting will only take effect with highly dynamic tracks that already have a perceived loudness at or below the reference level, i.e. tracks that would need to be made louder. Indeed, you can only introduce clipping when making a track louder, never when making it more silent. Consequently, it is practically never going to happen with modern, compressed pop music, where usually a negative adjustment, frequently in the order of 7 dB and more, is needed.

Neither of these are true statements. Please read the links that have been given if you need an explanation as to why.

I favor making tracks equal volume over introducing clipping that you cannot hear and FWIW, my reference is set to 92 because the level of my iPod's line out is so damn low. If you can hear the clipping, that's another story but I really get the feeling many of you still listen with your eyes and not your ears.

Theory and practice. Is there any chance that a highly compressed track that has been replaygained at -7db will still clip because of that possibility of "overshoot" during decoding? The links given explain the issues very well conceptually, but not quantitatively. The amount of overshoot that can be introduced during decoding will probably be very small (and indeed the resulting clipping mostly inaudible except perhaps for the finest ears on the finest equipment). This is my feeling only. If someone can show that a peak may be increased as much as 0.3-0.5 dB after decoding compared to that of the original uncompressed audio, I will stand corrected.

You just need to open foobar2000 and check the replaygain values:

Track Gain : -3.65 dB
Track Peak : 1.222659

1.0 is full scale and 0.0 is silence. Do your math.

Anyway, the second link pointed by carpman should have been enough.

I forgot about the replaygain_peak value, that indeed is the easy way to quantify the effect. Yes, it is much more important than I expected. Still, it is limited: if my math is correct, the signal in your example should be attenuated by 1.72 dB to avoid clipping. The track replaygain is -3.65, and thus is more than enough to prevent clipping.

Scanning my collection for some loud recordings (-8, -9 dB replaygain), I see that the album peak value (which would be the loudest peak in the collection) rarely exceeds 1.3 (+2.3 dB) (wait, found a nasty one with album peak of 1.45, replay gain -10 dB). For older not remastered albums with replay gain values in the order of -2, -3 dB, I find album peak values to be at most just above one (e.g. 1.01, 1.07). This would still lead me to conclude that the issue is of practical importance for dynamic tracks only, these requiring a replay gain value of only -2 dB and up (compared to the 89 dB reference level, and as far as they are really dynamic, i.e. have also peaks near the maximum).

Because of the magnitude of potential clipping on decoding, I would be even more inclined to opt for the "Don't clip" option except where loudness rather than quality is important (such as when on the road or in the car).


Is it? It was a conceptual example, but in no way a real world, quantitative example. Thank you for pointing me to the real extent of the effect anyway.

Even if you can't hear the clipping?

Unless you've analyzed every album available, how can you make such conclusions?

There would be no need if you cannot hear the clipping, but as for now, I cannot be sure that I would never hear the clipping.


Do you routinely see track peak values of 2 in your collection?

[edit]Quote tags were wrong

No way in a real world != routinely. You have no business making statements which you cannot back up.

Well, he was partly right. The example shows two equal-frequency sine waves, and says that the encoder removes one because of it being masked. That's not a real life scenario

Something that's more like real life is clipping (which is pretty similar to a square wave), and removing harmonics of this clipping. The result is something that's not a sine, nor a square.

The "Don't clip while doing track gain" is in my understanding only relevant when applying the gain to the mp3 audio frames, not when replaygaining using tags. In the latter case, you can instruct replay gain aware playback software to handle it (do nothing, reduce gain or apply hard limiting).

[edit]Removed reference to previous discussion