CLI tool to detect amount of clipping in audio file

replica9000 · 11-07-2012, 10:22 PM

http://ompldr.org/vZzZyZw

The above image is the same song from two different sources. The red shows the audio that is being clipped, which doesn't sound very nice.

What I'm looking for is a command line too that could output some numbers showing how much clipping a song has, or how compressed the dynamic range is.

onebuck · 11-09-2012, 08:22 AM

Hi,

You might look at WavPack 4.60 beta available, Windows and Linux versions older version, not sure if project is still active.

I think you can use 'CLI' or 'GUI'.

HTH!

replica9000 · 11-09-2012, 08:40 AM

Thanks for the suggestion, but I'm not sure how this helps me...

H_TeXMeX_H · 11-09-2012, 10:23 AM

The 'sox' 'stat' and 'stats' commands can be used to determine clipping. See the man page.

replica9000 · 11-09-2012, 04:05 PM

Thanks, this is what I was looking for.

So this is the output for the files from the image shown in the OP.

From a compilation CD: [Pictured first]

Code:

             Overall     Left      Right
DC offset   0.000253  0.000253  0.000107
Min level  -1.000000 -1.000000 -1.000000
Max level   0.999969  0.999969  0.999969
Pk lev dB       0.00      0.00      0.00
RMS lev dB    -10.56    -10.51    -10.61
RMS Pk dB      -4.38     -4.83     -4.38
RMS Tr dB     -78.57    -78.28    -78.57
Crest factor       -      3.35      3.39
Flat factor    17.75     17.25     18.17
Pk count       8.68k     8.18k     9.19k
Bit-depth      16/16     16/16     16/16
Num samples    14.1M
Length s     319.805
Scale max   1.000000
Window s       0.050

From another compilation CD: [Pictured second]

Code:

             Overall     Left      Right
DC offset   0.000287  0.000167  0.000287
Min level  -1.000000 -1.000000 -1.000000
Max level   0.999969  0.999969  0.999969
Pk lev dB       0.00      0.00      0.00
RMS lev dB     -8.44     -8.36     -8.52
RMS Pk dB      -4.09     -4.09     -4.17
RMS Tr dB    -257.19   -256.63   -257.19
Crest factor       -      2.62      2.67
Flat factor     9.55      9.21     10.03
Pk count       5.82k     6.90k     4.75k
Bit-depth      16/16     16/16     16/16
Num samples    14.2M
Length s     322.213
Scale max   1.000000
Window s       0.050

So looking at "Pk lev dB" for both files, it would indicate there would be some clipping going on. Looking at "RMS lev dB" and "RMS Pk dB", the first file isn't as loud and should have less clipping going on. Sound right?

H_TeXMeX_H · 11-10-2012, 04:06 AM

Yes, if the "Pk lev dB" is 0, then there is clipping.

For example try:

Code:

play danosongs.com-sunspark.mp3 stats
play danosongs.com-sunspark.mp3 gain -3 stats

Clipping will say:

Code:

In:2.21% 00:00:05.39 [00:03:58.78] Out:254k  [!=====|=====!] Hd:0.0 Clip:173  
             Overall     Left      Right
DC offset  -0.004690 -0.004614 -0.004690
Min level  -1.000000 -1.000000 -1.000000
Max level   1.000000  1.000000  1.000000
Pk lev dB       0.00      0.00      0.00
RMS lev dB     -9.13     -9.20     -9.06
RMS Pk dB      -6.43     -6.54     -6.43
RMS Tr dB     -13.61    -13.22    -13.61
Crest factor       -      2.88      2.84
Flat factor     1.35      1.68      1.02
Pk count        47.5        47        48
Bit-depth      29/29     29/29     29/29
Num samples     238k
Length s       5.387
Scale max   1.000000
Window s       0.050
play WARN rate: rate clipped 90 samples; decrease volume?
play WARN dither: dither clipped 83 samples; decrease volume?

Note the exclamation marks, the 0.00 "Pk lev dB", and WARN.

With gain adjustment it says:

Code:

In:1.75% 00:00:04.27 [00:03:59.90] Out:200k  [-=====|=====-] Hd:2.5 Clip:0    
             Overall     Left      Right
DC offset  -0.003329 -0.003232 -0.003329
Min level  -0.707946 -0.707946 -0.707946
Max level   0.707946  0.707946  0.707946
Pk lev dB      -3.00     -3.00     -3.00
RMS lev dB    -12.26    -12.31    -12.22
RMS Pk dB      -9.43     -9.54     -9.43
RMS Tr dB     -16.61    -16.22    -16.61
Crest factor       -      2.92      2.89
Flat factor     1.57      1.82      1.27
Pk count        40.5        43        38
Bit-depth      32/32     32/32     32/32
Num samples     188k
Length s       4.272
Scale max   1.000000
Window s       0.050

The sox man page recommends a gain -3 to -6 or in rare cases up to -9 dB.

replica9000 · 11-10-2012, 01:35 PM

I don't plan on modifying any of the audio files I have, only to gather info on them. If I have a the same song across multiple CDs, I can look at the info and decide which one would probably sound better.