Agree, but a good statistical approximation could be better than nothing ?
No, but there is software out there that analyzes clipping, reduces volume as needed, and “smooths” the waveform for better sound. It works.
Problem is there’s nothing to work with, it’d be pure guesswork.
Identifying clipping and attenuating signal to avoid it I get, easily done, “smoothing the waveform” … based on what, what the coder thought the artist intended before their work was murdered during mastering?
Extrapolation/interpolation. While it cannot necessarily recreate what was lost in clipping, it can reform peaks with moderate success.
Yes, and in real time.
What I mean is something like this, but made as a Roon DSP plugin (I would pay a lot for such a thing )
Avoiding clipping I get, if smoothing was to imply removing/undoing compression forget it.
No, that is not true. Extrapolating/interpolating new peaks – even as just approximations of peaks shaved off in clipping – does restore some crest factor dynamic range.
I agree, but you cannot make anything of something that has been brickwalled, which is what my original reply referred to.
No, “brickwalled” just means raised in level and hard limited to an extreme degree, such that many/most waveform peaks are clipped. Some of that limiting can be reversed by mathematical approximation via the “de-clipping” processes being discussed here.
We’re talking past one another, forget it…you’re right.
Lipstick on a pig…
You can accomplish declipping today by setting a modest (2-3dB at most) headroom adjustment, then upsampling the audio. The upsampling will interpolate between inter-sample clips and recover the peaks. The headroom adjustment ensures that that recovered waveform isn’t clipped again before output.
This is analogous to the method that R128 uses to determine the “true peak” level in the tracks (you can see this value in Roon). It is also the approach taken by Benchmark in their DACs. You can read more about that here.
I’m not sure how much “extra” that product you linked is doing. Would be interesting to know more details…
I have one album that was treated with a similar software. I was very happy with the results. Turned something unlistenable to something listenable, IMO.
I think there might be ways to inflate this punctured tire. Just look at the principles of Dolby noisereduction and DBX, as a couple of examples in the analog domain.DBX is doing EXACTLY what the OP is asking for.
I’m not saying you cannot put some lipstick on a pig, what I’m saying is that once compressed/brickwalled such that there’s no dynamic range in a track, there’s no way to undo that. You can make the whole quieter and get rid of clipping, but you cannot restore dynamic range that’s not there: Metallica - Death Magnetic comes to mind.
My point is: a compressed, brickwalled, clipped today rock/pop is far away from the original artist music, and there is the probability the original content (useful for a future restored, remastered or direct cut version) , is compromised too, like it seems from the latest Roger Waters album. So, every DSP intervention of declipping or something more that can do something better than none is welcomed.
For Metallica Death Magnetic we are lucky because we can easily find on the net the Guitar Hero III game version, that is surprisingly uncompressed and unclipped (!), that is a reward for the many customers who bought a disgraced cd.
“instead of nothing, it’t better instead” is a dictum from our zones
Tried to listen to DM last week as the track suicide and redemption I really love. As in the past, couple of tracks later with ears and brain in distress I had to turn it off. I’ll look into the game version on net but I assume that is ‘legally’ only for streaming via YouTube etc.
What pees me off just as much as those responsible for the ‘mastering’ is the artist is many of these cases is highly influential and could be doing something about it.
They must be blind and deaf to not realise as clearly they know best what their music sounds like.
Of course in these days of numerous reissues it is possible to redo this stuff one way or another. Assuming the uncompressed recording exists.
For now I’ll take the approach a small DSP improvement via guesswork/ estimatation is worth it as the available version couldn’t be much worse
100% agree, Scott.
Cheers duly acquired Rob