Audio Clipping

The WordSmith has an audio production question. Here is a situation. I just mixed a new track. Set the audio volume to -0.01bd with a limiter. Bounced the track from Cakewalk down to a wave file. The wave file is not clipping because of the limiter. Now it is time create the MP3 file. LAME is the most popular MP3 software used by Napster, Amazon….. and is free. I use it to convert a WAVE file into an MP3 file. So I run the command line prompt like: lame.exe "File.wav" "FilE.mp3" –option --option --clipdetect. Notice the ClipDetect option. In several case I have noticed that despite having created a wave file that is not clipping. Lame has detected a volume above -0.01db which is clipping. So I fix that problem with --scale 0.99 or what ever Lame suggests. Now I have a MP3 song that is not clipping according to Lame. But I am just thinking that each audio playback software is different and may detect a clip instance even though a new LIMITER has been put on the MP3 file. So my questions are: 1) Should I bounce the wave file down to some thing less than -0.01db then see what LAME say about clipping? Or is it safe to get right on the clipping edge? 2) When clipping is detected what is lost? 3) Do you lose on the vocal or instrumental side or both?

thewordsmith


 The WordSmith has an audio production question. Here is a situation. I just mixed a new track. Set the audio volume to -0.01bd with a limiter. Bounced the track from Cakewalk down to a wave file. The wave file is not clipping because of the limiter. Now it is time create the MP3 file. LAME is the most popular MP3 software used by Napster, Amazon….. and is free. I use it to convert a WAVE file into an MP3 file. So I run the command line prompt like: lame.exe "File.wav" "FilE.mp3" –option --option --clipdetect. Notice the ClipDetect option. In several case I have noticed that despite having created a wave file that is not clipping. Lame has detected a volume above -0.01db which is clipping. So I fix that problem with --scale 0.99 or what ever Lame suggests. Now I have a MP3 song that is not clipping according to Lame. But I am just thinking that each audio playback software is different and may detect a clip instance even though a new LIMITER has been put on the MP3 file. So my questions are: 1) Should I bounce the wave file down to some thing less than -0.01db then see what LAME say about clipping? Or is it safe to get right on the clipping edge?  

The sample points in digital audio do not always fall on the peaks of a waveform. You can have intersample peaks that might be +2dB (or a little more) over the peak values displayed by the meters over the course of a typical piece of audio.

If your limiter doesn't see these peaks, it can't reduce them. So you need to reduce the limiter to give yourself a bit of headroom. Or, if you use a limiter that upsamples (or upsample your audio, then limit it), the maximum intersample peak levels will be less.

2) When clipping is detected what is lost? 3) Do you lose on the vocal or instrumental side or both? 

Clipping means the top of your wave from is clipped off. So whatever part of the waveform was above 0dB is replaced by a generally nasty sounding distortion.

droddey


But those shouldn't matter really at the D/A stage. That's a digital thing. It matters in the DAW, hence why you don't push things too hard on the master bus or any given track, because any downstream processing of the signal might end up with a peak above 0dBFS. At the D/A converter stage all that has to happen is that the actual points be below 0dBFS. The regenerated wave happens in the analog domain and isn't limited by the FS scale.

Conversion of the file to another format is downstream digital processing as well and has the same problems. Anything that takes the original data, processes it and writes it back out, might cause a digital clip.
 
Some hardware folks argue that getting too close to 0dBFS can cause bad results on lower quality converters, and that keeping the digital samples at -0.3dBFS or lower helps deal with that. Maybe if the D/A super-samples and doesn't account for the possibility of inter-sample peaks it could be affected I guess.

Correct, but the point is if you upsample/resample, then your new samples will fall in different places and you will get some peak sample values that are higher than the original.

All I can do is go by what a manufacturer of music type program stated. mp3's are only 87.13% the output rate for a wave file or else they can not handle it.