Personal computing discussed

Moderator: Captain Ned

 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

An experiment in repairing clipped audio

Sat Sep 30, 2017 4:10 pm

A while back, I purchased an album based on a recommendation in the "What are you listening to..." thread: Just A Little Lovin by Shelby Lynne, a Dusty Springfield tribute album.

On the whole it is is fantastically performed and recorded. BUT... in just a few spots scattered around the album, when Shelby belts out a line, there's some clipping distortion. It sounds like static or a faint buzz when it happens. This has bugged me ever since getting this album, since the arrangements are fairly sparse and the mix is very warm, so the moments of distortion stand out like warts on what is otherwise an umblemished, fan-damn-tastic recording.

I was originally going to re-purchase it on CD (I currently have the MP3 download from Amazon), in the hopes that this issue was purely due to a screwup when they created the downloadable rips (I've had that happen before). But according to this thread, the CD has the same problem unless you buy the remastered SACD version (which is hard to find now and stupid expensive).

So I got to thinking... given the sparse arrangements, and the fact that the moments of distortion are very brief (and easy to isolate), maybe I can fix this myself? So I loaded the most problematic track up in Audacity and had a look. At first glance (viewing the entire track), it didn't look too bad -- it hasn't been boosted all to hell and brick-wall limited (look at the top two traces... we'll talk about the bottom two traces in a moment):
Image

If you zoom in on the area where the distortion is audible, it's clear that something's not right; there are some squared-off waveforms (again, focus on the top two traces):
Image

In both of the above screenshots, the bottom two traces are the same audio, after my fix.

Here's what I did: Lower the overall level by 3dB to get some headroom. Then zoom in on the areas where distortion is audible, and scroll through looking for any waveforms that appear clipped (flattened). Then zoom in even further, and use the Audacity "Repair" function to re-interpolate the bad samples.

Here's what a single Repair operation looks like -- bad samples selected:
Image

After invoking "Repair":
Image

It apperas that Audacity is using some fairly sophisticated algorithms, taking into account the overall shape of the waveform in the general vicinity. Bravo, Audacity!

I ended up repeating the above procedure a couple dozen times or so (all within a 15 second section of the track starting at about 3:00).

Sonically, the repair is almost transparent. You can still hear it if you're actively listening for it, but it doesn't jolt me out of being immersed in the music and make me whisper "Damnit!" under my breath any more.

I did similar micro-surgery on a few other tracks which also exhibited some minor clipping, but the above example was by far the most egregious.

Clearly this procedure is impractical if an entire recording is compressed and clipped (no way you could fix something like Death Magnetic this way). But in cases like this where the damage is isolated and brief, it appears to be a viable DIY approach if a better version is not readily available.
Nostalgia isn't what it used to be.
 
morphine
Gold subscriber
Gerbilus Supremus
Posts: 11406
Joined: Fri Dec 27, 2002 8:51 pm
Location: Portugal (that's next to Spain)

Re: An experiment in repairing clipped audio

Sat Sep 30, 2017 4:30 pm

That's some fine surgery. My hat's off to you.
There is a fixed amount of intelligence on the planet, and the population keeps growing :(
 
ludi
Darth Gerbil
Posts: 7446
Joined: Fri Jun 21, 2002 10:47 pm
Location: Sunny Colorado front range

Re: An experiment in repairing clipped audio

Sat Sep 30, 2017 5:07 pm

I had no idea that Audacity included a repairing function. That can be useful.
Abacus Model 2.5 | Quad-Row FX with 256 Cherry Red Slider Beads | Applewood Frame | Water Cooling by Brita Filtration
 
bthylafh
Grand Gerbil Poohbah
Posts: 3906
Joined: Mon Dec 29, 2003 11:55 pm
Location: Southwest Missouri, USA

Re: An experiment in repairing clipped audio

Sat Sep 30, 2017 5:10 pm

Wow, that's really cool.
Hakkaa päälle!
i7-8700K|Asus Z-370 Pro|32GB DDR4|Asus Radeon RX-580|Samsung 960 EVO 1TB|1988 Model M||Logitech MX 518 & F310|Samsung C24FG70|Dell 2209WA|ATH-M50x|Asus Xonar DX
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Sat Sep 30, 2017 5:17 pm

ludi wrote:
I had no idea that Audacity included a repairing function. That can be useful.

It only works on short segments (128 samples max). I'd used it in the past to repair clicks and pops on vinyl rips, and it occurred to me recently that it might be potentially useful for fixing minor clipping too if you lower the overall level first, to give it some headroom to work in. This experiment seems to have confirmed that theory.
Nostalgia isn't what it used to be.
 
Captain Ned
Gold subscriber
Global Moderator
Posts: 26469
Joined: Wed Jan 16, 2002 7:00 pm
Location: Vermont, USA

Re: An experiment in repairing clipped audio

Sat Sep 30, 2017 5:29 pm

Looks like it was over 0dBFS on the digital side, which implies sloppy engineering either in recording or mastering.
If the Earth were flat, cats would have pushed everything off of it by now.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Sat Sep 30, 2017 6:00 pm

Captain Ned wrote:
Looks like it was over 0dBFS on the digital side, which implies sloppy engineering either in recording or mastering.

Yeah. Since it seems to be affecting the vocals, and vocals are high in the mix, it may be that the vocal track was clipped during recording or that it got clipped in mastering. Hard to say, since normalizing tracks with the already clipped vocals to 100% would probably look pretty much the same.

The tilt of the clipped areas could be from signal processing at the studio that occurred post-clip, or the fact that my source is an Amazon MP3 download (i.e. compression artifacts). The album has grown on me enough that I may spring for the CD eventually (in which case I get to do this again); if I do I'll get to find out whether the slope of the clipped areas is a compression artifact! :lol:

Sad thing is, aside from these sporadic clipping episodes, it seems to be a very well engineered recording. It even got nominated for a Grammy in the engineering category. Either the people nominating it missed the audible clipping (which probably accounts for less than 5 seconds scattered over the course of the album), or were listening to a version different from the one used to master the CD. Or maybe the clipping is why it didn't win after getting nominated.
Nostalgia isn't what it used to be.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Sun Oct 01, 2017 1:53 pm

Here are before and after clips of the 2 second segment containing the repair (and several more) shown above:
before
after

If your music player does not obey replaygain tags you'll want to turn the volume up by 3dB on the second one to get a true "apples to apples" comparison.
Nostalgia isn't what it used to be.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Sun Oct 01, 2017 10:52 pm

Was just going back and listening to the entire album to see if I missed anything. It's pretty amazing how even just a few clipped samples can result in an audible artifact. Take a look:
Image

Just one isolated, slightly clipped waveform peak -- a mere 5 samples shifted a small amount from their correct level. The sonic effect of this was a single "click", not unlike what you would get from a minor scratch on a vinyl LP.

The human ear is a pretty incredible device, to be able to detect something like that.
Nostalgia isn't what it used to be.
 
DragonDaddyBear
Gerbil Elite
Posts: 652
Joined: Fri Jan 30, 2009 8:01 am

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 8:10 am

Is all of this being done in Audacity, including the screenshots?
 
Chrispy_
Gold subscriber
Maximum Gerbil
Posts: 4099
Joined: Fri Apr 09, 2004 3:49 pm
Location: Europe, most frequently London.

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 8:20 am

Good work.

I love Adacity (and should use it more than I do), but you'd think in this era we'd have moved on from silly mistakes like audio clipping, right?

Mastering should be done by algorithm rather than by people, meaning zero human error and perfect peak amplitude from the original source rather than wasted fidelity or clipping.
Congratulations, you've noticed that this year's signature is based on outdated internet memes; CLICK HERE NOW to experience this unforgettable phenomenon. This sentence is just filler and as irrelevant as my signature.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 8:22 am

DragonDaddyBear wrote:
Is all of this being done in Audacity, including the screenshots?

Yes.

I imported the original files as 32-bit float Audacity projects; those are my working copies. I export the repaired versions to FLAC for playback (and for transcoding to OGG for my phone).

Audacity screenshots taken with KDE's built-in screenshot tool (Print Screen key), saved as PNG, and cropped in GIMP.
Nostalgia isn't what it used to be.
 
the
Gold subscriber
Gerbil Elite
Posts: 851
Joined: Tue Jun 29, 2010 2:26 am

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 11:22 am

Audacity is a wildly powerful program. Its noise removing algorithm is equally advanced and has saved me several times in post production with base source recordings.
Dual Opteron 6376, 96 GB DDR3, Asus KGPE-D16, GTX 970
Mac Pro Dual Xeon E5645, 48 GB DDR3, GTX 770
Core i7 3930K@4.2 Ghz, 32 GB DDR3, GA-X79-UP5-Wifi
Core i7 2600K@4.4 Ghz, 16 GB DDR3, GTX 970, GA-X68XP-UD4
 
morphine
Gold subscriber
Gerbilus Supremus
Posts: 11406
Joined: Fri Dec 27, 2002 8:51 pm
Location: Portugal (that's next to Spain)

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 1:05 pm

Does it have an actual UI these days, though? :lol:
There is a fixed amount of intelligence on the planet, and the population keeps growing :(
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 1:16 pm

morphine wrote:
Does it have an actual UI these days, though? :lol:

Well it's kind of like GIMP that way. Powerful, if you can figure out how to tell it what you're trying to do. :wink:

It does lock up very occasionally, but its crash recovery seems pretty good. I don't think I've ever lost more than one or two edits, even if I haven't saved the project in a while.
Nostalgia isn't what it used to be.
 
meerkt
Gerbil Elite
Posts: 808
Joined: Sun Aug 25, 2013 2:55 am

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 2:32 pm

Once upon a time I ripped a scratched CD that I didn't have another copy of.

There were tons of short clicks, usually just a sample or two.

Most of the clicks were visible in spectral view, so I didn't have to listen much to find them.
I wanted minimal collateral modification, so I fixed it with Cool Edit by manually dragging the samples to smooth the curve, and verified/tweaked until the click disappeared in spectral view.

(no way you could fix something like Death Magnetic this way).

For that you have the Guitar Hero version. :)
 
Ifalna
Gerbil Team Leader
Posts: 209
Joined: Sat Jan 28, 2012 11:14 am
Location: Celestis

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 3:01 pm

Cool, definitely good to know. Thank you for sharing. :D
The backbone of modern industrial society is, and for the foreseeable future will be, the use of electrical Power.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 3:25 pm

meerkt wrote:
Once upon a time I ripped a scratched CD that I didn't have another copy of.

There were tons of short clicks, usually just a sample or two.

Most of the clicks were visible in spectral view, so I didn't have to listen much to find them.
I wanted minimal collateral modification, so I fixed it with Cool Edit by manually dragging the samples to smooth the curve, and verified/tweaked until the click disappeared in spectral view.

If the skipping is due to a scratch or nick you can often repair CDs by polishing the scratch out with plastic polish (or, in a pinch, toothpaste) and a soft cloth. For deeper scratches I've found that a first pass with fine grit (finest you can get) wet sandpaper cuts down on the amount of time required. IIRC I've repaired 3 or 4 CDs this way.
Nostalgia isn't what it used to be.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 6:01 pm

Chrispy_ wrote:
I love Adacity (and should use it more than I do), but you'd think in this era we'd have moved on from silly mistakes like audio clipping, right?

Mastering should be done by algorithm rather than by people, meaning zero human error and perfect peak amplitude from the original source rather than wasted fidelity or clipping.

Yeah, no kidding. This isn't rocket science. Apply DR range compression if you feel you must (but I'd really rather you didn't), but there's absolutely NO excuse for digital clipping at the mastering stage.

In this particular case, the mix is sparse enough, and the vocals are mixed high enough, that it is conceivable that the clipping is in the original vocal track. If the resulting mix was then normalized to 100% it would look pretty much the same as clipping introduced at the mastering step. However, I'm pretty sure I saw at least a couple of instances of lower-level high frequency content (e.g. cymbals) riding on top of the vocal, and the HF content getting clipped as the signal approached 100%. So that indeed implies a mastering screwup.
Nostalgia isn't what it used to be.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Mon Oct 02, 2017 9:22 pm

Also, a helpful tip if anyone else decides to attempt something like this...

Decouple the stereo tracks (Split Stereo Track in the track dropdown) before attempting any repairs. By default Audacity will operate on both the left and right signal of a stereo pair together. But if the clipping only appears in one channel, you don't want to apply a "fix" to the undamaged audio in the other channel. Even if both channels are clipped, the number of damaged samples may be different.

After making all of your repairs, select Make Stereo Track to join them back together before exporting to WAV/FLAC/whatever.
Nostalgia isn't what it used to be.
 
Noinoi
Gerbil Team Leader
Posts: 228
Joined: Fri Jun 26, 2015 11:31 pm
Location: Somewhere, anywhere!

Re: An experiment in repairing clipped audio

Tue Oct 03, 2017 3:17 am

This apparently even works sort of with a few samples of complete silence in the middle of a track. I had to deal with a file that had a few samples of silence in the middle of the song for some reason; the repair function changed the audible click and pop from the sudden silence into something that isn't exactly correct, but at least not annoyingly so. (audibly, it sounds more like some instruments play earlier than expected, though I'm not sure why it sounds like that even though the damage isn't that long - probably just our hearing?)

I suppose that the repair function is equally handy in fixing clipped audio as is missing audio.
i5-4590 | Kingston 2x8GB | Asus Strix GTX 970 | Asus Z97-Pro Gamer | Kingston Fury 240GB + WD Black 2TB + Blue 2TB | Win 10 FCU
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Tue Oct 03, 2017 5:24 am

Noinoi wrote:
This apparently even works sort of with a few samples of complete silence in the middle of a track. I had to deal with a file that had a few samples of silence in the middle of the song for some reason; the repair function changed the audible click and pop from the sudden silence into something that isn't exactly correct, but at least not annoyingly so. (audibly, it sounds more like some instruments play earlier than expected, though I'm not sure why it sounds like that even though the damage isn't that long - probably just our hearing?)

The longest section the Repair function will attempt to fix is 128 samples, which is about 3 ms at 44.1 kHz sampling rate. 3 ms is probably right on the edge of perceptibility. I could imagine it being noticeable on instruments with a sharp, percussive attack, assuming Audacity has interpolated the missing samples from the surrounding audio.
Nostalgia isn't what it used to be.
 
Aether
Silver subscriber
Gerbil
Posts: 96
Joined: Sat Dec 20, 2014 8:50 pm

Re: An experiment in repairing clipped audio

Tue Oct 03, 2017 7:39 am

Thanks for posting, JBI! It's been a long time since I have played around with Audacity, and even then I did not dive too deeply into it. I had no idea it had this level of sophistication.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Tue Oct 03, 2017 7:59 am

FWIW it also has a clipping identification tool which will tag areas with suspected clipping; if I do something like this again I will use that to help identify the potential problem areas instead of just listening for them.

There's also an automatic clipping fix plugin, which did not seem to work particularly well when I tried it a while back.

While I'm ranting about clipping, another recent acquisition, We Want Groove by Rock Candy Funk Party, is clipped all to hell and back (hard digital clipping at 100%). WTF? Very cool music, badly mastered. Not gonna try to fix this one unless I can find an automated tool that actually works; the damage is way too extensive to attempt manually patching it up. And even though the clipping is all over the place, the nature of the music makes it less obnoxious than it could've been.

C'mon people... if you must play in the loudness wars, at least use some sort of intelligent DR compression. Don't just crank the gain up until you're at +6dB. You can get away with that on analog tape, but digital clipping sounds like crap!
Nostalgia isn't what it used to be.
 
Chrispy_
Gold subscriber
Maximum Gerbil
Posts: 4099
Joined: Fri Apr 09, 2004 3:49 pm
Location: Europe, most frequently London.

Re: An experiment in repairing clipped audio

Wed Oct 04, 2017 5:35 am

just brew it! wrote:
FWIW it also has a clipping identification tool which will tag areas with suspected clipping


LOL, there you go; Proof that an algorithm can recognise the problem. Why do we let humans do rubbish work when we know a machine will do it perfectly?

I get really annoyed that people encoding things can't use the whole volume range. Nothing's worse than having to crank my sound system up twice as far as it should be, just because the audio of what I'm watching only utilises 0-30% of the available channel range. Not only does that amplify the interference from other sources, it also means that when I get a notification or other noise at "normal" volume it startles/deafens me and probably doesn't do my surround system any good either :(
Congratulations, you've noticed that this year's signature is based on outdated internet memes; CLICK HERE NOW to experience this unforgettable phenomenon. This sentence is just filler and as irrelevant as my signature.
 
meerkt
Gerbil Elite
Posts: 808
Joined: Sun Aug 25, 2013 2:55 am

Re: An experiment in repairing clipped audio

Wed Oct 04, 2017 7:51 am

Chrispy_ wrote:
I get really annoyed that people encoding things can't use the whole volume range. Nothing's worse than having to crank my sound system up twice
Normalization won't help much, and dynamic range compression does require decision-making and work.
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Wed Oct 04, 2017 8:16 am

Chrispy_ wrote:
LOL, there you go; Proof that an algorithm can recognise the problem. Why do we let humans do rubbish work when we know a machine will do it perfectly?

It's even dumber than that. There isn't even a need for fancy algorithms to guess where things may be clipped if you do it ahead of time. It's a simple "Will this audio file, when mastered at this level of gain, cause the level to exceed the range of a 16-bit (or 24-bit in the case of HD audio) number? Literally "is A greater than B or less than -B". A modern desktop computer could check an entire CD's worth of audio for this in a couple of seconds.

Sure, you don't have that luxury in a live recording situation, but that's not what we're dealing with here. These are studio recordings, which have been carefully recorded and mixed. Then some dumbass decides to crank it to 11 during mastering, and squares all the peaks off.

Chrispy_ wrote:
I get really annoyed that people encoding things can't use the whole volume range. Nothing's worse than having to crank my sound system up twice as far as it should be, just because the audio of what I'm watching only utilises 0-30% of the available channel range. Not only does that amplify the interference from other sources, it also means that when I get a notification or other noise at "normal" volume it startles/deafens me and probably doesn't do my surround system any good either :(

Time for a digression/rant...

Real-life audio is very "peaky". The loudest transients, like drum hits, can have an instantaneous peak that is many times higher than the surrounding audio; it doesn't sound that way though, because it is so brief. In order to keep the peaks from overloading things (or, in the case of vinyl LPs back in the day, from cutting over into the next groove), the peaks need to be kept at a safe level. But this means the average level (which is what people mostly perceive as the volume level) is much lower.

A few decades ago, some genius figured out that if your song sounded louder on the radio than everyone else's, people were more likely to notice and remember it. So how do we make a song sound louder overall, without making the peaks too big? Well, we raise the overall level, but reduce the gain during the peaks to keep things in a safe range. And this is how the "loudness wars" were born. Originally this gain manipulation was done manually ("gain riding") by the recording engineer. Nowdays it is done via various automated "dynamic range compression" algorithms.

In a nutshell, DR compression makes everything sound louder. But if overused (as it too often is), it squeezes the life out of the music. The transient peaks are part of what makes music sung and played on real instruments by real people sound, well, real. DR compression is effectively clipping's less evil twin. It tries to preserve the overall shape of the waveforms, but varies the gain to maintain near-constant volume. With clipping, OTOH, you just crank the gain, and throw away the tops and bottoms of any waveforms that no longer fit.

Ever notice how commercials often sound louder than the program they appear during? That's intentional too, and it's DR compression -- advertisers are trying to get your attention by being louder.

So where am I leading with all this? Well, the things that you are complaining are "too soft" probably haven't been DR compressed. It is quite possible that they are actually of higher fidelity, in the sense that they preserve more of the original dynamic range of the material.

Which brings us to something called "replaygain". This is an algorithm that analyzes an audio track (or album), using a model of human hearing to estimate how loud it will sound. Then a "gain factor" is placed in the meta-data of the audio tracks. Replaygain-aware playback software and devices can then automatically turn the gain up or down at the start of a track. This eliminates the annoyance factor of having to turn the gain up or down manually, while preserving the dynamic range of the content (since the gain is set just once per track, not instantaneously as the intensity varies within a track). There's also an "album mode", which applies a constant gain setting across multiple related tracks; this preserves the relationship between loud and soft movements in classical pieces, and prevents jumps in gain for albums where multiple tracks run one into another (think the medley at the end of Abbey Road).

Replaygain doesn't alter the original content at all; it's just a hint to the playback equipment telling it where the gain should be set to preserve a constant perceived volume level for a given track/album relative to other content. In a perfect world, all audio content would have replaygain meta-data.

Don't get me wrong; DR compression has its place. For spoken word, it can improve intelligibility by compensating for moments where the person leans away from the mic, for example. And used sparingly in music, it is unobtrusive and can improve the overall listening experience in noisy listening environments, and by keeping the peaks from exceeding the limitations of typical consumer audio gear. But it is just way over-used. The loudness wars need to DIAF.

</rant>
Nostalgia isn't what it used to be.
 
Chrispy_
Gold subscriber
Maximum Gerbil
Posts: 4099
Joined: Fri Apr 09, 2004 3:49 pm
Location: Europe, most frequently London.

Re: An experiment in repairing clipped audio

Wed Oct 04, 2017 9:33 am

I know what you mean about the loudness wars, but if what you were saying was correct, normalisation would work on these uber-quiet tracks because the very loud peaks would give the normalisation algorithm the extents of the dynamic range, and then the majority of the soundtrack that was quiet would be increased in volume to bring it closer to the midpoint of the total 0-100% dynamic range.

Occasionally, I come across a blu-ray rip where this is the case - and normalisation works well, for the higher-fidelity reasons you mention.

When it's just a screw up and someone has forgotten to set their encode volume to 0dB, what happens is that the whole track is too quiet and normalising the track does reduce the difference between the loud and the quiet parts, but the compressed total range means that the midpoint of the 0-30% range is 15% so, I then have to boost the volume by an additional 100-200% to bring the new, normalised 15% average volume level back up to something approaching 50%.

I know percentages aren't really perfect translations to dB values, but that's how the software I tend to use for playback and encoding tend to deal with volumes, so you'll have to forgive the atrocities I'm committing by using linear volume percentages on a PC interchangeably with logarithmic decibel scales ;)
Congratulations, you've noticed that this year's signature is based on outdated internet memes; CLICK HERE NOW to experience this unforgettable phenomenon. This sentence is just filler and as irrelevant as my signature.
 
morphine
Gold subscriber
Gerbilus Supremus
Posts: 11406
Joined: Fri Dec 27, 2002 8:51 pm
Location: Portugal (that's next to Spain)

Re: An experiment in repairing clipped audio

Wed Oct 04, 2017 9:42 am

All you described can be fixed with ReplayGain tagging :)

Add tags, set your player's preamp gain volume to +6/7dB for tagged tracks, go on your merry way.
There is a fixed amount of intelligence on the planet, and the population keeps growing :(
 
just brew it!
Gold subscriber
Administrator
Topic Author
Posts: 49702
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: An experiment in repairing clipped audio

Wed Oct 04, 2017 10:35 am

Chrispy_ wrote:
I know what you mean about the loudness wars, but if what you were saying was correct, normalisation would work on these uber-quiet tracks because the very loud peaks would give the normalisation algorithm the extents of the dynamic range, and then the majority of the soundtrack that was quiet would be increased in volume to bring it closer to the midpoint of the total 0-100% dynamic range.

I think you're misunderstanding the difference between normalization and dynamic range compression.

Normalization just raises the gain until the peaks are at 100%, it does not raise the level of the quiet parts releative to the peaks. If you're changing the relative levels of loud and soft parts on the fly, that's dynamic range compression (or clipping, if you're being stupid about it).
Nostalgia isn't what it used to be.

Who is online

Users browsing this forum: No registered users and 1 guest