Rescuing Recordings, Take 2: Unscrambling an Audio Egg

[fas42]

As mentioned in a previous thread, this is an exercise in recovering decent sound from recent recordings that have been overcompressed. Tim has complained many times about the quality of the recent Bruce Springsteen album, Magic, so I've had a first go at sorting out the initial track, "Radio Nowhere". What I ended up doing is getting a 30 second section of the track at a key area, adding to that the decompressed version of same, and finally the first clip again, reduced in volume to match that of the altered clip.

So at:

0 secs: Original clip
31 secs: Decompressed version
62 secs: Original at -6dB


The link to download from is https://www.filemail.com/dl.aspx?id=AWUDBJFIZPZHYTT.

Please do not play or download this file unless you're really interested in assessing progress made: I'm trying out another file transfer website, and the free use limitations are that the file is only up 3 days, for a maximum of 20 downloads. We'll see how it goes ...

This is mainly for Tim and Bruce, but anyone else with comments to offer, please do so! If people want the clip to be in a different format or with different content, give me a yell.

Please be gentle ... :b ,
Frank

 

[JackD201]

Are you talking about file or signal compression Frankie? If it's the latter what expander did you use?

 

[fas42]

Are you talking about file or signal compression Frankie? If it's the latter what expander did you use?

The latter, Jack, dynamic or signal compression. So far it appears, despite what Tim said, that the current problem of the loudness wars causing overcooked albums to be released is not an issue for the members here, no-one has downloaded my sample file yet. You're very welcome to see what you think of the result, Jack, and anyone else who may have been dissuaded by the cautionary note in the first post is now encouraged to download it.

How I've done it is a combination of techniques, it's a work in progress. Hopefully I'll get some feedback, to see if it's worthwhile persisting with the exercise ...

Frank

 

[JackD201]

This is what I heard Frank. In the second part, the snares came out of hiding, it was slightly easier to tell apart the guitar which turns out to be lead on the left and the other playing rhythm on the right. When the sax and lead guitar launch into a sync'd up riff, they were also easier to follow individually. Springsteen's voice seemed to have sunk back into the mix. The third part, well, sounded just like the first part just softer.

Thanks for the clips. For one it illustrates how the softer parts of instruments panned in the less dominant channel can come up and hog the instruments placed opposite. It sounds similar to mic bleeding. So you get spatial congestion like you would in a very reverberant venue except the assault is of direct sound.

As for unscrambling the egg, well, in my opinion the egg was still left scrambled. To me, despite the noise of the original, it had more energy and in a bizarre way the futz'ed up guitars actually worked with the radio theme of the song. Expanding it made the song, shall I say, emasculated. So while I felt there were sonic benefits, all in all I felt that the loss of the message was not worth the sonic benefit.

Just one guy's honest opinion Frankie.

 

[fas42]

That was excellent, Jack, that's exactly what I'm after: as many points of view as possible. Ultimately, the final mix and mastering of an album is the decision of the artists, producers and record company, and that is what was decided for Springsteen's Magic. What I'm focusing on at the moment is getting back to the mix prior to the last round of heavy duty compression; actual extraction of the individual tracks prior to the final mastering, to be reassembled in some way, is a very, very heavy duty exercise, to be left for further down the track. The latter is the sort of thing the doomed Lexicon processor was on about ...

As you point out, the compression allows the softer elements to compete, unfairly, with the more dynamic parts of the mix. Now I just need a Springsteen fanatic to tell me whether that's how his band should sound ...

Thanks very much for going through the exercise, Jack!

BTW Jack, would you have a "worst" album that you feel has been "damaged" by over-happy fingers in the control room?

Cheers,
Frank

 

[Phelonious Ponk]

I listened to all 3 of them in a row, 3 times. First time through, I heard pretty much what Jack did. The lead guitar in the right channel and the sax solo in the left channel seemed to come forward a bit, and the vocals seemed to recede in the "de-compressed" version. By the time I got through the third listen I wasn't sure I was hearing any difference but the obvious volume drop.

In all fairness, though, "Radio Nowhere" is a very busy, "wall of sound" song. It probably wouldn't sound that different with the compressor turned off. If you're going to experiment with this for awhile, Frank, let me give you some good material to work with. My favorite new act is The Civil Wars, an Americana duo. Mostly spare, acoustic arrangements, and they both sing and play with a lot of subtlety and nuance. I was listening to their one and only CD the other day and noticed that they haven't escaped the loudness wars. There are some places where you can tell, just by the string attack of the guitar, or the whisper in the voices that a passage was sung/played very softly. But it's as loud as the more forceful passages. That would probably be a much better recording to work with, as any progress you might make would be much easier to hear.

Tim

 

[fas42]

Thanks for that, Tim, that's excellent feedback. I deliberately chose a very busy section, because for me it was easier to see whether I was making progress, particularly in the area of the crash cymbals and the acoustic around the sax. I just realised that the mp3 of this song is available as a free download, so there shouldn't any problem with making a decompressed variation also available in this format: having the whole song to play will probably make it easier to pick up the differences. So I'll proceed with creating an mp3 at the highest quality level, and see what you think ...

I will certainly have a look at The Civil Wars: could you name a track that you see as the worst offender?

I certainly know what it's like listening to the same material over and over! I was totally unfamiliar with the album, so the first track came across as a total mess on first hearing. By now I know it almost backwards, so I also have the problem that my head is filling in the gaps in the sound, it "knows" what the song should sound like; people doing mastering must have this problem also ...

Frank

 

[bblue]

I haven't been able to the files in question as they have expired on the download site. But as a general comment, Springsteen stuff is almost always audio garbage. Mis-engineered, overly compressed individual tracks and then hard peak limiting on top of the mix. Just awful, usually. But I've not heard this album so maybe he's finally learning? (doubtful)

I personally can't imagine how basic de-compressing or trying to undo heavy peak limiting would be very effective.

I'd love to hear the samples, though.

--Bill

 

[fas42]

Bill, I uploaded the sample at the end of a first "heavy" session of fiddling when I had run out of puff, it sounded OK at the time and I just wanted to get something happening, so put it out there. Now, in the cool light of day, I'm not happy with that round of processing, specifically because at a key point some elements of processing distortion are intruding, so now doing another session to refine the methodology.

So, I could put up that sample again, or have you wait until I'm happy with the next round: which would you prefer?

Frank

 

[Phelonious Ponk]

I will certainly have a look at The Civil Wars: could you name a track that you see as the worst offender?

I'll listen carefully over the next couple of days and see what I can come up with. I wouldn't look for the worst offender, though. I'd look for the most revealing one. One in which obviously soft passages just don't play softly.

Tim

 

[bblue]

Bill, I uploaded the sample at the end of a first "heavy" session of fiddling when I had run out of puff, it sounded OK at the time and I just wanted to get something happening, so put it out there. Now, in the cool light of day, I'm not happy with that round of processing, specifically because at a key point some elements of processing distortion are intruding, so now doing another session to refine the methodology.

So, I could put up that sample again, or have you wait until I'm happy with the next round: which would you prefer?

Would it be possible to do both? Along with what you started with so I can be up to speed with what you originally posted?

You might find it far more rewarding to try your unscrambling ideas on tracks that at least have a possibility of being decent audio to start with.

What software are you using for all this? And which de-compressors or de-clippers? I have quite a few of those types of plug-ins for audio repair and may be able to help in some way. Are you upshifting to 24 bit and higher sample rate before trying to perform any of this? 44k 16bit would be far too coarse for good results.

--Bill

 

[Bruce B]

Unfortunately I haven't been able to listen to the file. We lost power due to the ice storm.

 

[fas42]

Unfortunately I haven't been able to listen to the file. We lost power due to the ice storm.

So have you actually got a copy of the file downloaded, to listen to later?

Frank

 

[fas42]

Would it be possible to do both? Along with what you started with so I can be up to speed with what you originally posted?

You might find it far more rewarding to try your unscrambling ideas on tracks that at least have a possibility of being decent audio to start with.

What software are you using for all this? And which de-compressors or de-clippers? I have quite a few of those types of plug-ins for audio repair and may be able to help in some way. Are you upshifting to 24 bit and higher sample rate before trying to perform any of this? 44k 16bit would be far too coarse for good results.

--Bill

Okay, I'll put it up again. As mentioned in my first post, the sample is comprised of three clips: a bit of the the original track, the same decompressed, and then the original again, attenuated to match the RMS level of the altered section.

I'm using a combination of software tools, and Audacity and Reaper figure in the mix: there appear to be about 3 or 4 decent declipping mechanisms out there, but I haven't focused too strongly on that yet. But what amazes me is the frequency of clipping in material for which it's completely unnecessary: Tim's "The Civil Wars" first track clips at one point for a period of 7 samples in the first 30 secs.

As said before, this is a work in progress; my techniques will evolve -- I'm not happy with the uploaded example. So far I haven't felt the need to do upsampling, and I've confirmed that the software works at a 24 bit precision level.

Thanks for your offers of assistance!

Frank

 

[fas42]

File's back up, at same website, exactly the same file as before, same conditions: 20 downloads, available 3 days: https://www.filemail.com/dl.aspx?id=GTATCBHZGEVHAOL

Again, remember this is a first version of the exercise on this sample of music; but all feedback about it is very welcome!

Edit: Sorry, muffed it slightly, should have specified 20 downloads (the maximum), but I'll leave it like this for the moment ...

Frank

 

[bblue]

File's back up, at same website, exactly the same file as before, same conditions: 20 downloads, available 3 days: https://www.filemail.com/dl.aspx?id=GTATCBHZGEVHAOL

Again, remember this is a first version of the exercise on this sample of music; but all feedback about it is very welcome!

Edit: Sorry, muffed it slightly, should have specified 20 downloads (the maximum), but I'll leave it like this for the moment ...

Frank

Got it, Frank, thanks.

It's as I suspected given the artist, complete crap audio and a totally lost cause. It's not really clipped per se, but at the final stage of mastering, peak limited with the threshold at 0db. There are many types of these obnoxious super loud peak limiters, some are single band, some are multii-band, some include (optionally) mild compression to keep the inbound levels consistent, some introduce a lot of distortion and some don't, depending on whose hands are operating them.

I have an anticipative de-clipper that tries to determine by sampling the waveform in real time, when to extend the peaks. (That's Izotope's RX 2). It produced a version with more peaks, but it looks like the audio was peak trashed before it ever got to mastering, by another limiter or series of limiters during mixing. Probably on a digital workstation, but that's just a guess by the sound. That plus individual track compressing/limiting and distortion would make this virtually impossible to descramble. It's no good to start with... My opinion of course.

I saved RX 2 output to my FTP server if you are interested in checking it out. Will pm you with details. I'll be interested in following your progress if you try to pursue this.

--Bill

 

[fas42]

Thanks, Bill, I've got the files, sorry I didn't respond earlier. Will have a chance a bit later to listen properly to what the RX2 did ...

Bob Ludwig did the mastering, and he hasn't exactly got the worst name in the business; interesting how things have got in the recording industry. At the moment I'm looking at Barton Hollow, by The Civil Wars, and that's copped pretty severe treatment for something that's labeled folk, with a bluesy feel ...

Frank

 

[bblue]

Thanks, Bill, I've got the files, sorry I didn't respond earlier. Will have a chance a bit later to listen properly to what the RX2 did ...

Bob Ludwig did the mastering, and he hasn't exactly got the worst name in the business; interesting how things have got in the recording industry. At the moment I'm looking at Barton Hollow, by The Civil Wars, and that's copped pretty severe treatment for something that's labeled folk, with a bluesy feel ...

Frank

Bob is well respected, no doubt there. But a mastering engineer can only do so much, and I'm sure his instructions were to make it as loud as possible. Imagine being in a mastering engineer's shoes and having to endure trying to master an album like this. Your hands are tied! Probably 1 in 5 (at the most) projects worked on are with good quality original masters. What a frustrating job, but quite rewarding when you work on the gems.

--Bill

 

[fas42]

Of interest, I have no Springsteen myself, but yesterday got a copy of "Human Touch" from the library. Not quite sure where you got the idea that his recorded quality was always "crap", Bill, it's a very impressive recording. Some tremendous acoustics, sound staging, clarity of instruments, etc, are spot on, no complaints from me ...

Frank

 

[bblue]

Of interest, I have no Springsteen myself, but yesterday got a copy of "Human Touch" from the library. Not quite sure where you got the idea that his recorded quality was always "crap", Bill, it's a very impressive recording. Some tremendous acoustics, sound staging, clarity of instruments, etc, are spot on, no complaints from me ...

Hi Frank,

I just purchased a copy of this to eval. IMHO, as far as Springsteen goes, it's a lot better than most. It is only moderate vs heavily peak limited like most of his, and has a lot more bass guitar in the mix. Some of the songs are actually somewhat enjoyable to listen to (surprise). Exceptions are Gloria's Eyes, Roll of the Dice, Long Goodbye and Real Man. The first three are more pushed into peak limiting and have a lot of HF eq and distortion. The last is incredibly over eq'd in the HF range, shrill and distorted.

Those four songs would probably tame down considerably when played on a mid-fi type system with a rolled off high end, and conventional stranded wire cables, but still would be pretty distorted. Just enough to avoid peeling paint off the walls. Here, they're virtually un-listenable.

-Bill