Hi there,

first of all: English isn't my native language.
I already tried to use the search-function, but my formulations didn't find the correct answer/result :-(

My situation:
1) I'm musican and I'm limiting my music to -0.1 dB to prevent digital clipping.
2) If I encode a song (latest recom. version of LAME) with "fast standard", "fast extreme" or "insane", i's blasted/overamplified/overdriven in WinAMP, footbar2000 and other MP3-decoders. Only some peaks are affected, but it's annoying with digital measure-equipment (like DAT-recorders or PPMs).

I converted some MP3s back to WAV using the fixed LAME-builds, but they've got the same problem.

My question:
Why does LAME (over-)amplifiy the songs? I could upload some demos, if you want.

Best,
Tobias

The problem is not in LAME, not even in MP3, but in lossy compression.

Lossy compression does vary the audio waveform. Since you are limiting that near to the limit, the decoded signal can (and as you say does) go out of the digital limit.

There is nothing you can do to fix this, other than reducing the amplitude of the signal before, or after encoding (first method, Wavegain, second method, MP3Gain or replaygain (Foobar2000 is a good program for that)

Oh okay. I thought, there would be a hidden switch to set a maximum peak.
I mean something like a dynamic processor, which kills/reduce these peaks.

But okay, I can live with it :-)

Ahem, the purpose of lossy encoders is not to act as dynamic range compressor-DSPs - we already have more than enough of that in the recording-phase!

It may be interesting to know for you, that the sound inside of the mp3 is NOT clipped.... this is because mp3 can store higher volume(because of floating point) than your source audio. So, the audio inside of the mp3 is not damaged and does not clip - it will only clip when it is decoded/played, because then it wont fit anymore into the space.

There are two solutions to this - both involve using replaygain:
1. wavegain the audio before encoding - this will NOT guarantee that the resulting mp3 will not clip upon decoding - but it works most of the time.

2. (the proper solution) use something like mp3gain AFTER encoding. In a nutshell, MP3s store something similiar to a master-gain value. Thus, it is possible to change the volume of mp3s without actually changing the stored audiodata itself.

- Lyx

Hey Lyx: Frohes Neues!! Happy new year to anyone!
Professionals use compressors not for compressions, but for great sound :-)

If the "sound" inside the MP3 is not clipped, it's okay. Seems like a good headroom. Thanks for information!

If the sound in the WAV was not clipped, then the sound in the MP3 is not clipped either. It may surpass 0dB (which will cause playback to clip), but the MP3 does not clip.

uh, NeoRenegade ... it cannot surpass 0dB digitally* :-)
but ofcourse, I understand what you mean

*from the digital point of view, a fixed amount of 0dB peaks in a shot time is interpreted as digital clipping:
-0.2dB 0dB 0dB 0dB -0.1dB 0dB -0.3dB => it will show clipping on the 4th value (0dB)

No, you don't understand.

MP3 doesn't have a bit depth per se, so signals don't get clipped when amplified too much in MP3 as they do in WAV.

If i remember right, there is indeed a "maximum" - and reaching it would mean clipping - but it's so high that in practice it cannot be reached. To make it simple: it's virtually unlimited.

If audio was normalised to 0dB then hard limited at -6dB, would flat peaks at -6dB count as clipping? Would they produce the same distortions, just at a lower volume? Sorry, just something I don't understand.

Yes, because clipping is the introduction of unwanted distortion (harmonics that weren't present in the original signal) due to the "flattened" peaks. The amplitude of those peaks does not matter.

Thanks.

I just found out that a part of my question wasn't answered.



Okay, that means MP3 cannot clip at 0dB, but when decoding it back to PCM, it gets a bit-depth (normally 16bit) and then these MP3s are clipped!

Does ReplayGain help on this situation?

Decode them with foobar, using replaygain. You should then be fine.

slightly off topic - does vorbis have a similar playback clipping snafu with high saturation content?