Now that poorly mastered CDs with excessive clipping have been a fact of life for well over 15 years now, a good amount of attention has been paid to making this music sound the best on the radio, with the least amount of objectionable artifacts added by the station's audio processing. For example, major changes to AGC structure has been implemented in leading audio processors, in order to better deal with this highly pre-processed program material.
But perhaps due to grumblings about "it's like putting Humpty Dumpty back together" and "the record companies should just give us unclipped material to being with," I have seen little attention paid to the concept of using DSP to reconstruct the clipped-off waveforms and peaks of the CD audio. Of course this cannot remove the spectral "trash" that was added by the clipping, but it can make some surprisingly good improvements to the audio, by reducing distortion and adding back some of the lost peak intensity and dynamic impact.
One song that is now the cornerstone of a marketing push of the benefits of the new Omnia 11 processor is Kelly Clarkson's "Because Of You." Certainly, improvements in processing can go a long way towards making this song sound more acceptable on the radio, but you can't make it sound any better than what is on the CD. Or can you? A while back I came across a "declipping" application from a company called "CuteStudio." (link: http://www.cutestudio.net/data/products/audio/seedeclip/index.php ) At the time, it was just a simple command-line utility, but now it has involved into a full graphical application for Windows. Basically it attempts to use DSP declipping to restore up to 6 dB of lost peaks. But how well does it work? Well, I'll let your eyes and ears be the judge.
Here is "Because Of You", directly ripped from the "Breakaway" CD:
http://i42.tinypic.com/28157xw.gif
Here is the "declip" command line utility in action -- it found 6,234 instances of clipping in the CD audio:
http://i42.tinypic.com/wc0hma.gif
And here is the waveform, after "declipping":
http://i41.tinypic.com/i6w6bp.gif
For comparison, here is the "declipped" version overlaid with the original CD audio dropped down by 6 dB, to exactly show the restored peaks:
http://i44.tinypic.com/1zd82vc.gif
And now finally, some audio clips demonstrating the difference...
First, a sample of the direct CD audio, reduced by -6 dB:
http://mysite.verizon.net/tekel/kelly-cdrip-6db.flac
And now the "declipped" audio of the sample:
http://mysite.verizon.net/tekel/kelly-declip.flac
A pretty obvious difference, no? Instead of just going "splat... splat... splat", the kick drum actually sounds like a kick drum, and it no longer causes ugly hole-punching into Kelly's vocals. The difference is even more obvious when processing is applied.
Any what about the infamous "bacon frying" pre-emphasis clipping distortion on the intro? Well, applying the CD de-emphasis curve to the intro cured that problem for me. I only applied it to the intro, up to the first drum beat, since the rest of the song does not exhibit this problem. The change in HF EQ at this point is not noticeable in the processed on-air audio. The only thing that you do notice is the near-elimination of the "frying", when suitably processed in accordance with the FM 75 uS pre-emphasis curve:
http://mysite.verizon.net/tekel/kelly-intro-cdrip-processed.flac
vs.
http://mysite.verizon.net/tekel/kelly-intro-declip-deemph-processed.flac
For reference, here is the CD de-emphasis curve:
http://home.comcast.net/~jahearn79/images/cd-deemphasis.gif
So, what does the future hold? Might we see audio processors with built-in on-the-fly "declipping"? It might seem counterintuitive to "declip" the incoming audio only to then follow it with several layers of clipping, but for decades we've had AM processors which remove assymmetry from the incoming audio waveforms, and then add it back in the form of assymmetrical +125% / -100% modulation.
But perhaps due to grumblings about "it's like putting Humpty Dumpty back together" and "the record companies should just give us unclipped material to being with," I have seen little attention paid to the concept of using DSP to reconstruct the clipped-off waveforms and peaks of the CD audio. Of course this cannot remove the spectral "trash" that was added by the clipping, but it can make some surprisingly good improvements to the audio, by reducing distortion and adding back some of the lost peak intensity and dynamic impact.
One song that is now the cornerstone of a marketing push of the benefits of the new Omnia 11 processor is Kelly Clarkson's "Because Of You." Certainly, improvements in processing can go a long way towards making this song sound more acceptable on the radio, but you can't make it sound any better than what is on the CD. Or can you? A while back I came across a "declipping" application from a company called "CuteStudio." (link: http://www.cutestudio.net/data/products/audio/seedeclip/index.php ) At the time, it was just a simple command-line utility, but now it has involved into a full graphical application for Windows. Basically it attempts to use DSP declipping to restore up to 6 dB of lost peaks. But how well does it work? Well, I'll let your eyes and ears be the judge.
Here is "Because Of You", directly ripped from the "Breakaway" CD:
http://i42.tinypic.com/28157xw.gif
Here is the "declip" command line utility in action -- it found 6,234 instances of clipping in the CD audio:
http://i42.tinypic.com/wc0hma.gif
And here is the waveform, after "declipping":
http://i41.tinypic.com/i6w6bp.gif
For comparison, here is the "declipped" version overlaid with the original CD audio dropped down by 6 dB, to exactly show the restored peaks:
http://i44.tinypic.com/1zd82vc.gif
And now finally, some audio clips demonstrating the difference...
First, a sample of the direct CD audio, reduced by -6 dB:
http://mysite.verizon.net/tekel/kelly-cdrip-6db.flac
And now the "declipped" audio of the sample:
http://mysite.verizon.net/tekel/kelly-declip.flac
A pretty obvious difference, no? Instead of just going "splat... splat... splat", the kick drum actually sounds like a kick drum, and it no longer causes ugly hole-punching into Kelly's vocals. The difference is even more obvious when processing is applied.
Any what about the infamous "bacon frying" pre-emphasis clipping distortion on the intro? Well, applying the CD de-emphasis curve to the intro cured that problem for me. I only applied it to the intro, up to the first drum beat, since the rest of the song does not exhibit this problem. The change in HF EQ at this point is not noticeable in the processed on-air audio. The only thing that you do notice is the near-elimination of the "frying", when suitably processed in accordance with the FM 75 uS pre-emphasis curve:
http://mysite.verizon.net/tekel/kelly-intro-cdrip-processed.flac
vs.
http://mysite.verizon.net/tekel/kelly-intro-declip-deemph-processed.flac
For reference, here is the CD de-emphasis curve:
http://home.comcast.net/~jahearn79/images/cd-deemphasis.gif
So, what does the future hold? Might we see audio processors with built-in on-the-fly "declipping"? It might seem counterintuitive to "declip" the incoming audio only to then follow it with several layers of clipping, but for decades we've had AM processors which remove assymmetry from the incoming audio waveforms, and then add it back in the form of assymmetrical +125% / -100% modulation.
