PDA

View Full Version : Clipping analysis



liffy99
2013-02-25, 02:02
I'm interested in analysing some audio CDs to see if clipping is present in their recording. Can anyone recommend software that will allow me to analyse this (Mac) ?
I have tried Audacity which has an 'Analyze Clipping' function which identifies possible culprit peaks. On the waveform graph however none of the waveform exceeds the +1 or -1 levels - so I'm not quite sure what I'm looking at. Should, for example, the +1 level equate to 0db (the max level CD can be recorded at without clipping) ?
Thanks

Julf
2013-02-25, 03:34
I'm interested in analysing some audio CDs to see if clipping is present in their recording. Can anyone recommend software that will allow me to analyse this (Mac) ?
I have tried Audacity which has an 'Analyze Clipping' function which identifies possible culprit peaks. On the waveform graph however none of the waveform exceeds the +1 or -1 levels - so I'm not quite sure what I'm looking at. Should, for example, the +1 level equate to 0db (the max level CD can be recorded at without clipping) ?
Thanks

Yes, Audacity is pretty good. There is no way you will ever get anything beyond +-1 shown on the waveform display, because it represents 0 dBFS (or full scale). A digital signal can never go beyond that. The analog output from the DAC van go beyond 0dB FS as a result of the interpolating filter filling in the "missing" clipped peak. What the Audacity clipping detector does is detect places where the full scale was not enough to represent the wave, and the signal got clipped as a result.

liffy99
2013-02-25, 04:44
Yes, Audacity is pretty good. There is no way you will ever get anything beyond +-1 shown on the waveform display, because it represents 0 dBFS (or full scale). A digital signal can never go beyond that. The analog output from the DAC van go beyond 0dB FS as a result of the interpolating filter filling in the "missing" clipped peak. What the Audacity clipping detector does is detect places where the full scale was not enough to represent the wave, and the signal got clipped as a result.

So does the 'interpolating filter' reconstruct an approximation of the waveform ? In which case why do we hear clipping at all ? Is it that a reconstructed output above 0db is sent to an amplification stage that exceeds the amp's stated capacity, overloads it, and causes subsequent clipping at that point ?

Julf
2013-02-25, 05:15
So does the 'interpolating filter' reconstruct an approximation of the waveform?

Yes, but it will be an incorrect approximation, because the data itself is "wrong" (as in clipped - limited to 0 dBFS instead of whatever higher value the original sound wave data contained).


In which case why do we hear clipping at all?

Because the reconstruction will not match the original wave - but the reconstruction will cause the analog signal peak value to go above 0 dBFS.


Is it that a reconstructed output above 0db is sent to an amplification stage that exceeds the amp's stated capacity, overloads it, and causes subsequent clipping at that point ?

That is unlikely to happen with an analog amp, as an analog amp should have at least some headroom. Might be an issue if the signal is re-digitized.

cliveb
2013-03-03, 10:38
I'm interested in analysing some audio CDs to see if clipping is present in their recording. Can anyone recommend software that will allow me to analyse this (Mac) ?
I have tried Audacity which has an 'Analyze Clipping' function which identifies possible culprit peaks. On the waveform graph however none of the waveform exceeds the +1 or -1 levels - so I'm not quite sure what I'm looking at. Should, for example, the +1 level equate to 0db (the max level CD can be recorded at without clipping) ?
Thanks
As others have pointed out, you'll never see any sample greater than 0dB. And since in principle samples at 0dB might be genuine (ie. not clips), then any analysis of a CD for clipping can only use some kind of rule to determine whether a section of waveform is likely to be a clip. A common rule-of-thumb is that 4 consecutive full-scale samples represent a clip. But in fact many CDs are clipped below full scale. Moreover, many clips on CDs are not absolute flat-tops, but are either slightly "ragged" (indicating subsequent processing after clipping) or sloped (indicating EQ after clipping). Audacity's clipping detection is actually pretty primitive. It only notices full scale clips, so will often miss clipping on CDs.

Here is not the place for me to advertise my own software, but I do have a program (Windows only, unfortunately) that does some pretty sophisticated clipping analysis of 16 bit PCM WAV files (ie. the sort you get from CDs). PM me if you are interested.