[iamivancb]

A portion of the signal that got clipped is lost data. Nothing can get back lost data.
It’s a similar issue with compression. There is no clear way to find out what compression settings were applied based only on the resulting signal. Compression tools have a variety of settings after all. Software trying to reverse compression or correct clipping are really guessing and making up new signal with the hope that it’s going to be more enjoyable. Those are tools for subjective preference, not actual restoration of data because said data is usually lost for good.
For dynamic compression I would personally leave it as is. For clipping, if it’s audible then it probably sounds quite bad and a declipper seems like the way to try and make it listenable again. I would still first try to find a non clipped version if it exists(older master?).


I’m not sure about the meaning behind vinyl losing dynamics when transfered to CD. Digital recording from a turntable sounds the same to me as the turntable itself. Vinyl allows less dynamic than CD anyway. The only reason why vinyl would have a lot of dynamic is if the needle turns into Tony Hawk and uses the bumps on the vinyl to take off.

Those programs claim that they can "restore" the original master, which means going back to it just before it started to be over-compressed. So it seems to be snake oil, then? That's the problem with newer music, not the older masters available.

Maybe I got a bit confused with the vinyl part you told me about, a sin of youth/haven't had one in my hands yet, only CDs. I guess a wave of rabid audiophiles will jump after this last thing I said.

 

[iamivancb]

I have an old dynamic expander from the 80s. It works pretty well if you use it sparingly.

With every kind of recording?

 

[iamivancb]

The loudness war results in the over application of compression. A de-clipper is designed to remove digital overload clipping/distortion, not compression. So unless your over-compressed recordings also have audible digital overload distortion a de-clipper won’t help much, if at all.

That’s not entirely true. Because:

Yes, compression tools do have a variety of settings but digital overload clipping doesn’t. The waveform peak is simply chopped off at the point of 0dBFS. It can therefore be a relatively straight forward process of interpolating what those missing samples would have been if there were not a 0dBFS limit (which is possible in practice using 32 or 64 bit float processing). Of course, it depends on the amount of clipping, with continuous clipping over a long period there’s not enough data to reconstruct the waveform accurately.

The issue with compression is different because it’s not just a case of some missing samples, the waveform shape has been changed depending on the compressor settings (attack and release envelopes for example but other parameters as well) and there’s no way for the declip software to know what those settings were or therefore exactly what’s missing.

G

Alright. So using a declipper, if the recording has no distortion will just make the mix sound rather bland and thinner?

 

[bigshot]

With every kind of recording?

I've used it primarily on pre-hifi recordings, but it was designed to use with LPs. It has several degrees of decompression. Even the most aggressive one works well. It doesn't work miracles, but it helps. There is probably a digital plugin that does the same thing or even better.

 

[castleofargh]

The loudness war results in the over application of compression. A de-clipper is designed to remove digital overload clipping/distortion, not compression. So unless your over-compressed recordings also have audible digital overload distortion a de-clipper won’t help much, if at all.

That’s not entirely true. Because:

Yes, compression tools do have a variety of settings but digital overload clipping doesn’t. The waveform peak is simply chopped off at the point of 0dBFS. It can therefore be a relatively straight forward process of interpolating what those missing samples would have been if there were not a 0dBFS limit (which is possible in practice using 32 or 64 bit float processing). Of course, it depends on the amount of clipping, with continuous clipping over a long period there’s not enough data to reconstruct the waveform accurately.

The issue with compression is different because it’s not just a case of some missing samples, the waveform shape has been changed depending on the compressor settings (attack and release envelopes for example but other parameters as well) and there’s no way for the declip software to know what those settings were or therefore exactly what’s missing.

G

I made a scenario in my head and only talked about it, which I admit would have really deserved more.

A portion of the signal that got clipped is lost data. Nothing can get back lost data.

I said a portion so I wasn't completely lost in space, but I was also thinking we'd have enough clipping that it's clearly annoying and can't be handled by lowering the gain or really just the low pass filter. So I assumed a situation grave enough that not all the sines could be fully deducted from the samples, which of course is not what we usually encounter outside of butchered amateur recordings. Clearly I wasn't optimistic when I made that post. Or maybe I initially planned to explain various cases and forgot along the way? IDK. Thanks for pointing it out and sorry @iamivancb for giving you the wrong impression.

 

[gregorio]

Those programs claim that they can "restore" the original master, which means going back to it just before it started to be over-compressed.

No, compression is added as part of the mastering process. So if a recording is over compressed, “restoring the original master” will restore the over-compressed master, not the final mix or some other point before the mastering compression/limiting was applied.

With every kind of recording?

An expander will work on every recording. However, it depends on the settings used and it’s impossible to perfectly undo the compression that’s already been applied. Generally, the results of using an expander are mixed, the best results are usually obtained from a light application of expansion but then it’s only going to have a minimal effect on highly compressed/over-compressed material. It’s not a commonly used tool.

Alright. So using a declipper, if the recording has no distortion will just make the mix sound rather bland and thinner?

Declippers don’t remove distortion, only digital overload distortion. If a recording has no digital overloads (clipping) then a declipper won’t do anything. However, this can depend on the exact algorithm/s applied by the declipper and the settings chosen.

A declipper will try to restore the waveform peaks chopped-off due to hitting the max level (0dBFS), so these restored peaks will be above 0dBFS and will need to be reduced in level. Many declippers will add compression/limiting and make-up gain, so the restored sections don’t suddenly sound quieter than the unaffected sections. Obviously, adding more compression isn’t the solution to an already over-compressed recording but it’s less objectionable that overload distortion or certain peaks suddenly sounding quieter than the surrounding material.

G

 

[gregorio]

So I assumed a situation grave enough that not all the sines could be fully deducted from the samples, which of course is not what we usually encounter outside of butchered amateur recordings.

Yes, particularly with studio recordings that situation should never occur, it shouldn’t occur even in the case of live recordings. The most common use of declippers is when processing Production Sound (the live sound recorded during the shooting of TV/Film) because the levels are not known beforehand and the recording environments are relatively poorly controlled.

G

 

[bigshot]

An expander will work on every recording. However, it depends on the settings used and it’s impossible to perfectly undo the compression that’s already been applied. Generally, the results of using an expander are mixed, the best results are usually obtained from a light application of expansion but then it’s only going to have a minimal effect on highly compressed/over-compressed material. It’s not a commonly used tool.

It isn't really needed much for production use. If you're mixing, you're applying compression, and if you need more dynamics, you just reduce the compression. Expanders are more useful to consumers who are looking to make already mixed music a little more dynamic. I found it to be very useful with classical 78s. A little peak expansion combined with a little pattern based dynamic noise reduction sounded natural and improved the sound.

 

[iamivancb]

I've used it primarily on pre-hifi recordings, but it was designed to use with LPs. It has several degrees of decompression. Even the most aggressive one works well. It doesn't work miracles, but it helps. There is probably a digital plugin that does the same thing or even better.

Sounds interesting. My idea was to fix some modern recordings, though. For pre-hifi Adobe Audition seems to be a good choice for removing some artefacts. Didn't know about that tool btw.

 

[iamivancb]

I made a scenario in my head and only talked about it, which I admit would have really deserved more.

I said a portion so I wasn't completely lost in space, but I was also thinking we'd have enough clipping that it's clearly annoying and can't be handled by lowering the gain or really just the low pass filter. So I assumed a situation grave enough that not all the sines could be fully deducted from the samples, which of course is not what we usually encounter outside of butchered amateur recordings. Clearly I wasn't optimistic when I made that post. Or maybe I initially planned to explain various cases and forgot along the way? IDK. Thanks for pointing it out and sorry @iamivancb for giving you the wrong impression.

Don't worry, I got the idea. I just have to sharpen my ears to see which recordings have audible distortion (apart from obvious ones such as Death Magnetic).

 

[iamivancb]

No, compression is added as part of the mastering process. So if a recording is over compressed, “restoring the original master” will restore the over-compressed master, not the final mix or some other point before the mastering compression/limiting was applied.

An expander will work on every recording. However, it depends on the settings used and it’s impossible to perfectly undo the compression that’s already been applied. Generally, the results of using an expander are mixed, the best results are usually obtained from a light application of expansion but then it’s only going to have a minimal effect on highly compressed/over-compressed material. It’s not a commonly used tool.

Declippers don’t remove distortion, only digital overload distortion. If a recording has no digital overloads (clipping) then a declipper won’t do anything. However, this can depend on the exact algorithm/s applied by the declipper and the settings chosen.

A declipper will try to restore the waveform peaks chopped-off due to hitting the max level (0dBFS), so these restored peaks will be above 0dBFS and will need to be reduced in level. Many declippers will add compression/limiting and make-up gain, so the restored sections don’t suddenly sound quieter than the unaffected sections. Obviously, adding more compression isn’t the solution to an already over-compressed recording but it’s less objectionable that overload distortion or certain peaks suddenly sounding quieter than the surrounding material.

G

So I guess we have nothing to do with these cases then, just to screw us with that side of the issue.

Just as declipping programs do not work with the same effectiveness each time, thus I wouldn't use it for modern stuff for sure. It seems more appropriate, as his owner pointed out, for older mixes.

Yes, so it would just work in those cases then. Certainly, it is not the best thing to do; however, after messing with the master it doesn't seem to be anything better to do to fix it somehow.

 

[71 dB]

I’m not sure about the meaning behind vinyl losing dynamics when transfered to CD.
Digital recording from a turntable sounds the same to me as the turntable itself.
Vinyl allows less dynamic than CD anyway.

Yes, vinyls transferred to CD without any processing sound the same. I used to transfer underground electronic dance music from vinyl to CD 25-20 years ago. I got so fed up with vinyl back then I try to be as digital as possible and have as little to do with vinyl as possible. I didn't want vinyls to sound the same on CD, so I had to process them. I cleaned pops MANUALLY, because automatic removers made the kickdrums weaker. It was very tedious restoration work. I also cut 2 dB below 200 Hz with a low shelf filter, because my father's turntable I used to transfer the vinyls has 2 dB bass boost for some reason. I bought those vinyls used and they were in bad shape after being played for years in clubs by DJ's. Even after heavy restoration some of them were barely listenable. Anyway, vinyl sounds exactly the same transferred on CD if no processing is done.

However, a dynamically brickwalled master will sound more dynamic on vinyl than on CD! This is because CD will reproduce the brickwall dynamics as they are, but vinyl will distort the brickwall shape into something more dynamic. Phase distortion alone can do this and "accidentally" expand the signal. So, a DR6 master is DR6 on CD, but perhaps DR8 on vinyl. Maybe this is why people assume vinyl transferred on CD has reduced dynamics? No, if you transfer the DR8 vinyl made of DR6 brickwall master on CD, the CD will also be DR8, identical to the vinyl with the good and the bad.

 

[bigshot]

A lot of old LPs were compressed to limit the peaks. This was to maintain a consistent groove width and to fit more grooves on a side. Those albums are what dynamic expanders were designed for. They weren't for transferring to digital. They were for real time playback before the digital era.

 

[Ryokan]

Threads on new DAP's coming out often mention the way it can upsample, a lot of talk follows about converting PCM to DSD, what is the advantage to simply playing flac or mp3 files?

 

[castleofargh]

Threads on new DAP's coming out often mention the way it can upsample, a lot of talk follows about converting PCM to DSD, what is the advantage to simply playing flac or mp3 files?

All but outdated R2R NOS DACs will turn a PCM signal into something like 3 to 5bit + massive oversampling. While DSD DACs if they're not a decade old will also use at least 2 bits (because a one bit system is just dumb, always was, and only made sense before we found another way). The actual processes end up being more similar than most people suspect.
You'll still find objective differences (even more so if the designer wants some audible ones to be found), but no more than what has always existed with all DACs and DAC chips. Difference in the final sample rate, difference in the filter (frequency and type), differences in how much of some upsampling event is used for anti jitter/reclocking action. I suspect that a bunch of those differences are there to avoid patent infringement, and not to try to reinvent the wheel.
I'm sure you can have fun oversampling anything up to DSD1024 equivalent or whatever, but why?

MP3 and flac are converted to PCM before being sent to the DAC/DAC chip. It doesn't change anything to what can or will be done with it.

I'm guessing people who are dearly interested in such discussions are already fans of the very special DSD "sound", or have firm beliefs about oversampling settings and how they "sound". But for us cardboard ears who never listen to anything and only look at graphs, it's rarely worth anything because if a device can benefit from oversampling/upsampling and the designer knows his stuff, the DAC will do it when and where it's most relevant. Like just about all DACs already do.

I can feel new 'friends' cursing me even before I post this.