I know it's not possible to 'remaster' a CD when all you have is the original CD as source, but I'd be interested to know of any techniques that might improve the sound of a CD that suffers from over-compression and/or clipping, a.k.a. 'clipression' (a term I first read this very forum). For example, is there an effective expander that can undo some of the damage caused the by heavy-handed use of a compressor..?


EDIT: I'm only asking for suggestions (of software/techniques to try). I appreciate that an improvement in sound is a matter of opinion.

Adobe Audition has a clip restoration feature that works well on lightly clipped waveforms, but if the clipping is too severe the restored wave just sounds weird.

A tried a clipped peak restoration plugin in SoundForge 7. Like you say, if clipping is too severe the restoration isn't really an improvement. I was wondering if there was an effective expander of some sort, i.e. something that can undo some of the damage caused by over compression. Probably not

Replay Gain a track then export it to wav (using Replay Gain) in Foobar. After that use the bass boost effect in Audacity. It hardly sounds great (actually it's terrible compared to a properly mastered track), but at least there is a bit more boom with drums.

Edit: Added extra point about using Replay Gain when exporting the track to wav.

Seems to me that if a CD has been overprocessed, clipped, and otherwise squished, much of the pleasant part of the recording has actually been removed entirely.

I'm not sure that one could bring it back.

I use an old dbx 117 expander in my living room system (not for headphone listening though), it can give a nice small improvement on badly mastered CDs.

A small improvement may be just enough. I usually try to listen to music as it was intended to sound, but unfortunately some CDs do require a little 'enhancement'. I quite like the Ozone plugin for Winamp.

I have to say, the only reason for 'de-clipping' might be to save your speakers, but the result likely won't sound any 'better'.

A thread here once mooted the idea of a class-action for damaged audio equipment - it seems more than a few drivers, particularly in-car, have been fried by what amounts to large amounts of DC from 'modern' CD releases with excessive clipping.

Myth alert!

Clipping isn't DC. Secondly, a clipped signal won't damage a speaker, unless the continuous 'RMS' power of the signal is greater than what the speaker can handle. But that will occur regardless of the 'shape' of the signal - a 50w 'RMS' square wave is no more likely to damage a speaker than a 50w 'RMS' sine wave. The square wave simply contains harmonics of the fundamental tone.
(Realistically, the only time when the content of the signal can make it more likely to damage a speaker is when a low frequency tone of sufficient power is sent to a speaker that has been poorly designed - ie, an enclosure that allows the woofer to bottom out at low frequencies, or a tweeter/midrange that has been crossed over too low)

The reason why many people are confused over this issue is because when you clip an amplifier, you can get significantly more power out of it than you can without clipping. It is this extra power that is more likely to damage the speaker, rather than the shape of the signal.

Come again?

R.

Edit "come again" meaning;- what?

True.
See this thread for an example of successful restoration with cool edit/audition:
http://www.hydrogenaudio.org/forums/index....91&hl=declipped


Tried soundforge too and some others. Forget it. The only one that sometimes works is cool edit.
Slight clipping, or at least a slight sort of clipression where clipping is the main part of the damage, may be repairable, especially the vocals if their damage occured during mastering instead of mixing. If not, you can't do anything about that cause the clipped peaks are hidden inside the mix.

As for compression/limiting:
I tried a lot of expansion settings in cool edit's dynamics processor, but nothing really works. Most likely it becomes worse. The biggest problem here is that mostly hard limiters have been used, that means infinite compression ratios. There's nothing left of the "different peak" information that might be expanded/restored.

Somewhat, at least, but.

True, but. Some amplifiers, with some kinds of driver stages, can clip in a way that the output "sticks" to one of the power supply rails. This usually happens in a fashion that means it "sticks" more on one side than the other, and this can cause "DC" of sorts, i.e. a temporary DC bias to a clipped signal.

Note, this is not a linear phenominon or anything like it, it has to do with having NPN output stages for both sides of an amp with bipolar supply, how it's driven, and how you should not design an amplifier.

Dead wrong. A clipped signal can develop much more high frequency energy than an unclipped signal. This can cause tweeters, which are often sized to handle much less power than woofers, to fry.

What's worse, we get into nonlinearities of amplifiers again. Some amplifiers, when going into a "protection" mode, oscillate at high (usually low ultrasonic, sometimes higher) frequencies. This energy can be enormous, and at barely or not-audible high frequencies.

Again, this can fry tweeters.

I've seen systems in which each of these has happened.

Now, a speaker that has a good "gain structure" is hard to break, driverwise, this way, but then you often find that you can fry a resistor or 2 in the crossover the same way.

So, this part is not a myth. Perhaps the reasoning given before was wrong, but this is not a myth, and I've fixed systems that have met this problem and fried.

Indeed, and these harmonics, especially when clipping is severe, are what can cause the problem. Sometimes. Protection oscillation at least used to be pretty common, and was way bad for both speakers and the amplifiers. (Hint: Look up "storage time", and consider what happens when both halves of a totem pole try to be on at the same time.)



Now, I don't know if that part is right or not. The difference in power between a sine wave and a square wave of the same peak amplitude is only 3dB, after all, while the increase in high frequencies can be many dB (well, from zero to lots of energy, hard to set a dB value) when you clip a midrange signal.

I will say that what the other person said was perhaps not myth, even if the way it was expressed was either ignorant or negligent in the details.

In ideal equipment, perhaps we might get a different outcome. Today, we hardly have ideal equipment.