I am extracting my CDs with Exact Audio Copy to WAV files but I want to make sure there is no clipping in them.

I have tried to use the program Wavegain but it would be really slow to do that because the windows frontend does not work and its a command line dos program.

Can I use the Normalize function in EAC to do this? If so, what settings should I use for clipping protection?

wavegain works fine here. That is probably the best way to go.

BTW, from the sounds of it, it's the original clipping on the cd that you are worried about and there is nothing you can do about that, it was recorded/mastered that way.

When I press either "calculate" buttons I get the following error:



The reason I think there is alot of clipping in the output is when I made some MP3s previously with Easy CD-DA Extractor and put them through the program called MP3gain. It showed clipping on almost all my MP3s.

Since my Ipod has that problem with high-bitrate LAME-encoded MP3s, I am re-encoding most of my CDs to AAC using EAC as a ripper and iTunes as a encoder but I can't use MP3Gain anymore because it only supports MP3s.

Anyways, can anybody help me with the problem with Wavegain?

1) The clipping in the mp3 is from the mp3 encoder, most likely, and not the input files.

2) From speeks website for the wavegain frontend:

If you get an error when you try to run the frontend, you need one of the following files (the error message indicates which one):

* Visual Basic 6 Runtime files
* mscomctl.ocx

try the files here: http://members.home.nl/w.speek/wavegain.htm

and maybe try redownloading wavegain/frontend.

Also, is wavegain.exe in the same directory as the frontend or the win32 directory?

EAC copies losslessly the CD to the wav file. It doesn't introduce any clipping.
If the original CD clips, EAC respects it, and your wav file clips the same way. In this case, there is nothing to do. PCM Clipping can't be recovered.
If the CD doesn't clip, EAC won't make it clip.

Don't use normalisation. All it can do is introducing more clipping if you set it to a level superior to 1.

Thanks for the replys.

So the clipping must be from a bug in either Easy CD-DA Extractor 7 or LAME 3.95.1 because thats what encoder I used in that program. right?

I installed both VB Runtime and mscomctl.ocx shown and it still gave me the error. All the Wavegain files including the frontend, are in a folder on the desktop. Should I move them somewhere else?

1) The clipping could originate from the cd.
2) The encoder could introduce clipping> replaygain.hydrogenaudio.org and mp3gain.sourceforge.net
3) Never heard of that problem with wavegain, maybe try an older version. And maybe John33 will helpout.

Now that I am using iTune's AAC encoder instead of LAME MP3, is that encoder better in not clipping? Im guessing there is no AACgain

No, but you can use fb2k's foo_rgscan to add Replaygain tags, which will be used to adjust the volume on playback.

http://doc.hydrogenaudio.org/wikis/hydrogenaudio/ReplayGain

How glad I am that I can say: This is wrong.
During the last years, I've now clip restored many songs. To do that, an appropriate program is needed. Some big wav editors offer a clip restoration function. If this is used properly, audible improvements can be achieved. But don't expect any wonders. Much time is needed and some songs are too heavily clipped and can't be repaired.

the clipping problem with mp3 and so:

I once discovered that lossy formats do this in principal in the case that the original file is clipped and/or dynamically compressed like on almost all new CDs. For example with MD-recorders: After recording the signal has to be restored when playing and the volume is few decibels louder than the original. My theory is that the MD-recorder restores the signal in a natural way (and not in an evilly compressed).
You can only avoid this by normalizing down some decibels before recording (or encoding to a lossy format). But the song can then no more be clip restored, because the clipped aereas are now modified.
Take an uncompressed/unclipped recording for encoding to a lossy format (the song can be normalized to 0dB) and I bet the clipping problem won't occur. It's all because of the devilish loudness war.

I have to say this is wrong. Wave editors and plugins can only "guess" what would've been above the flattop. Usually this is just an average of what was before and what is after the clip. If there are multiple clipped samples (many modern recordings), the process becomes less and less precise.

The "improvement" you are hearing is from you're speakers NOT trying to reproduce a straight line.

No, the minidisc knows nothing about what we, humans, consider as "natural", or "artificial". It just tries to reproduce the original signal so that it sounds as close as possible to the original.

Yes, over compressed recordings clip more often when they are encoded. The first and obvious reason for this is that they are recorded louder, which increases the risk of clipping.
But when they are already clipped, they introduce more additional clipping than recordings not clipped.

That is because lossy encoders (psychoacoustic ones) remove inaudible frequencies, which cause the straight line of the clipping to oscillate, thus to clip again, since it is already at the maximum level.

All of this is correct. But there are still audible improvements. I'm always self-critical enough that I can say that this is no imagination. If I could, I would offer an example.



Surely wrong. Of course I perform not only a listening test (always with my headphone). Watching the waveform to see if it looks more 'natural' is equal important. And I like to look at the single samples how they're moved beyond the clipping level / beyond the 0dB boundary (in float format).

I apologize if it sounds too harsh, but to me it seems that you, dreamliner, didn't apply declipping tools properly (like I've said).

Wrong again

What dreamliner said is basically right. The declipping tool can only guess how the signal should look like. Normally the flat "clipped" part is replaced by a sine wave. One way of doing this is to filter apply a low pass filter over the clipping. (The filtered signal will have samples above 0dB)

The result will be a "softened" signal that is easier for your speakers/headphones to reproduce, but it is not the original signal - just a guess of how it could have been. The filtered signal will very likely have samples higher than 0dB.

I apologize if I sound harsh, but it sounds like you don't understand the enough of tha basics to understand what dreamliner said...

Using declipping tools properly? To use them properly would mean NOT to use them on post mix samples. Essentially clip restoration tools are made for very limited use to try save a single track (instrument) that may have just touched slightly over 0dB. They were never intended for reducing intentionally mastered clipping. (This is not to say that I haven't used them for this, I have. Unfortunately some recordings need something done to them. I do agree that it can HELP to reduce clipping, but they can not restore it to the premastered, non-clipped state.

ancl, I guess applying a low pass filter would be a very unprecise process.
But the thought is right. In principal, music is something like a very very complex fourier series. That means, music is the addition of many single sine waves. When creating a rectangular signal with fourier series, more high frequencies are needed. Basically, clipping a waveform means to make it more rectangular, or, in other words, adding high frequencies to the corresponding 'fourier series'. So high frequencies should be canceled out. (This is why clipping distortion often sounds so shrill.)

Now my theory how declipping works: The default process is to turn the flat (or almost flat) line into a full-blown sine peak (if the program can't imagine anything better). Then a fourier analysis is performed using the areas around the clipped area and the peak is appropiately modulated.



In fact, I often discover samples that sound clipped at a single area (mostly the vocals), but the music is at low volume and the sample structures in this area don't look clipped. After using declipping functions, the distortions are still there. So it seems that the original single tracks in the multitrack editor were already clipped. When they're mixed down, they are clipped again. There is no chance to repair the original clipping, the second clipping may be repaired.



Yes, someone would have to be very very lucky to achieve this.

I hope you are now a little bit more convinced that I understand what I'm doing.

Hey, I'm in a good mood today. I guess we're only arguing semantics here.