Can Bad Recordings be Fixed?

[fas42]

Just now, miguelito said:

 

What tools have you used for decompression? I fear that compressed music has also a lot of clipping, which would make decompression impossible.

 

As mentioned above, the main software I used was Reaper - every track, album will be different in how it's been 'mangled', and I used a combination of visual inspection, intelligent guessing, and trials around the likely parameters for best results. From memory, I declipped, then decompressed - on the few tracks I played with seriously I was pleased with with what I got out of them - but I certainly knew that the compression settings could get very fancy with some material, and that it would be a major exercise trying to work out what had been done for those.

 

What I took away from the exercise was that it would be possible to do the analysis, if one had plenty of time, and motivation - do it with software, rather than manually; just build up the sophistication of the unravelling algorithms, over time, until most anything could be 'repaired'.

[fas42]

8 minutes ago, miguelito said:

 

.... If there is clipping, then you’re screwed as that loses info.

 

Info may be lost, but it's extremely transitory. Interpolating, guesstimating, copy and paste from similar parts of the waveform can do a lot - my belief is that reversal of vicious clipping could be finessed to the point where losses were essentially inaudible.

[miguelito]

13 minutes ago, fas42 said:

 

Info may be lost, but it's extremely transitory. Interpolating, guesstimating, copy and paste from similar parts of the waveform can do a lot - my belief is that reversal of vicious clipping could be finessed to the point where losses were essentially inaudible.

 

Very interesting on clipping. I would have thought clipping was basically the end of the road.

 

I can certainly imagine someone coming up with software that would analyze the entire track - probably making some assumptions on the compression type used - and estimate compression parameters. Maybe a few different variants for the user to select and listen for the most “natural” result. I can certainly see a possible use as a DSP plugin in some players (Roon for example).

 

 

[gmgraves]

6 hours ago, Audiophile Neuroscience said:

 

I still have a dBX from the 80's I think.

 

I used dBX as an example because I don't know of any current hardware com-panders. I know that Audacity (at least the Mac version) has a variable compressor and things like attack and release can be changed, but I see no expander function in the "effects" menu. That doesn't mean that someone hasn't written an expander module for Audacity, it just means that one isn't bundled with the program (and I find that a bit odd). 

[gmgraves]

5 hours ago, miguelito said:

 

Interesting, though I wonder if compression is really that fine tuned as opposed to using a universal setting. I also wonder if analyzing an entire song can be helpful to estimate compression parameters. There are many instances in data analysis where transformed can be “deconvoluted”. If there is clipping, then you’re screwed as that loses info.

 

Clipping is what limiters do. If you listen to FM radio, you hear clipping all the time as the Federal Communications Commission (FCC) has decreed: "thou shalt not over modulate - ever!" Used to be that every radio station had a CBS Labs Automax for compressing a radio station's dynamic range, and a Volumax for clipping signals that would otherwise over modulate the carrier. If they didn't have the CBS Labs duo, they had the even older and even worse sounding Fairchild model 660 and 670. Today, most American FM stations use Bob Orban's Optimod equipment which combines limiter, compressor and stereo multiplexer all in one unit. Orban's equipment is much less brutal than the earlier stuff (one radio station chief engineer that I used to know always said that the CBS modulation control duo was akin to driving 10-penny finishing nail with a sledge hammer).

 

When I was a teen, growing up in the suburbs of Washington DC, there were only about 10 FM stations in that market. The stations were so few and far apart that there was no need for modulation control. Since the stations couldn't run into each other like they can now (because there are so many and they are so close together), nobody cared if they over modulated. The result was a glorious age where FM sounded magnificent. Live broadcasts of concerts were the highest-Fi signal that an audiophile could play through his system. Before stereo, American FM could do 20-20KHz, with an 80 dB dynamic range and live broadcasts were relayed from the performance venue to the studio using Class-A phone lines. When stereo came along, the bandwidth went to 50-15KHz and the dynamic range (with full quieting) dropped to the low 70 dB range. 

 

I know I digressed. I wanted to show that not all clipping is readily audible. It can be done in such a way (on peaks only) so that the listener will never hear it. 

[sandyk]

3 hours ago, gmgraves said:

 

I used dBX as an example because I don't know of any current hardware com-panders. I know that Audacity (at least the Mac version) has a variable compressor and things like attack and release can be changed, but I see no expander function in the "effects" menu. That doesn't mean that someone hasn't written an expander module for Audacity, it just means that one isn't bundled with the program (and I find that a bit odd). 

 Hi George

 Even my older version of Sound Forge, (SF9) permits Expansion at various ratios.

 You wouldn't expect the freeware Audacity version to have a module like that.

 

Regards

Alex

 

 P.S.

 I have available a DL of before and after versions of a complete track in .wav format, using SeeDeclip Duo Pro,  if anybody is interested to see if they can hear any difference from before and after. The track had only minor clipping, although the posted waveform suggests otherwise. The differences are quite subtle, and you will need to increase the level of the declipped version by a couple of dB.

PM me if you would like the DL  links.

[sandyk]

3 hours ago, gmgraves said:

The result was a glorious age where FM sounded magnificent. Live broadcasts of concerts were the highest-Fi signal that an audiophile could play through his system. Before stereo, American FM could do 20-20KHz, with an 80 dB dynamic range and live broadcasts were relayed from the performance venue to the studio using Class-A phone lines. When stereo came along, the bandwidth went to 50-15KHz and the dynamic range (with full quieting) dropped to the low 70 dB range. 

 

 Hi George

 I remember those days well, which is why my rant elsewhere about degraded SQ from FM stereo (use of .mp3 carts etc.) ,  AM radio, and poorer Video definition from HD TV with the recent Industry change to .mpeg 4 ,where the bit rate has often been lowered too much, compared with the original HD TV transmissions with .mpeg 2,  in order to squeeze more channels of Lifestyle TV  ( (mainly long advertisements) in the available spectrum.

 

Regards

Alex

[fas42]

8 hours ago, miguelito said:

 

Very interesting on clipping. I would have thought clipping was basically the end of the road.

 

I can certainly imagine someone coming up with software that would analyze the entire track - probably making some assumptions on the compression type used - and estimate compression parameters. Maybe a few different variants for the user to select and listen for the most “natural” result. I can certainly see a possible use as a DSP plugin in some players (Roon for example).

 

 

 

When I was exploring, that's essentially what I was doing - a smart guesstimate, done manually, using largely visual clues, and then iterate to "just right". On either side of the correct settings, the sound was "peculiar"; get it right on the button and everything would sound "normal" - it was also possible to see the "sweet spot" visually; the waveform would not show an awkward shaping to the envelope of the peaks.

[fas42]

57 minutes ago, sandyk said:

 Hi George

 Even my older version of Sound Forge, (SF9) permits Expansion at various ratios.

 You wouldn't expect the freeware Audacity version to have a module like that.

 

Regards

Alex

 

 

It is possible to directly use Audacity - Reaper makes the set of plug-ins available for download, which includes the key compressor module. I've installed these, and they work fine.

[sandyk]

Just now, fas42 said:

It is possible to directly use Audacity - Reaper makes the set of plug-ins available for download, which includes the key compressor module. I've installed these, and they work fine.

 

 

And the price ?

[fas42]

Just now, sandyk said:

 

 

And the price ?

 

A very nice, zero!  ...

[fas42]

And here you go, https://www.reaper.fm/reaplugs/ ...

[Audiophile Neuroscience]

As mentioned previously Barry Diament recommended Reaper over Pro Tools and others. He was of the view that the DAW imparts its own sound to the end result and among other things favoured Reaper. He uses a bunch of more sophisticated tools.

[fas42]

13 hours ago, sandyk said:

 I have available a DL of before and after versions of a complete track in .wav format, using SeeDeclip Duo Pro,  if anybody is interested to see if they can hear any difference from before and after. The track had only minor clipping, although the posted waveform suggests otherwise. The differences are quite subtle, and you will need to increase the level of the declipped version by a couple of dB.

PM me if you would like the DL  links.

 

I have had a look at sandyk's offering, and to my eyes this looks like classic compression - of the sort that I looked at some years ago. Audacity reports zero clipping of the before, and technically this may be the case - but the dynamics have been savagely squashed, by about 10dB from eyeballing the waveform.

 

The SeeDeclip fix only slightly rectifies this, by about 1.5dB - I think I might have a go at refreshing my techniques, and try and do a lot better than this, on this particular posted track, . If I think I've got somewhere, I post it up, and invite criticism ... .

[Audiophile Neuroscience]

8 minutes ago, fas42 said:

 

I have had a look at sandyk's offering, and to my eyes this looks like classic compression - of the sort that I looked at some years ago. Audacity reports zero clipping of the before, and technically this may be the case - but the dynamics have been savagely squashed, by about 10dB from eyeballing the waveform.

 

The SeeDeclip fix only slightly rectifies this, by about 1.5dB - I think I might have a go at refreshing my techniques, and try and do a lot better than this, on this particular posted track, . If I think I've got somewhere, I post it up, and invite criticism ... .

 

Very, very interesting to compare the 'corrected' versions. I already have Alex's (SandyK) SeeDeClip version downloaded.

[miguelito]

14 hours ago, fas42 said:

 

When I was exploring, that's essentially what I was doing - a smart guesstimate, done manually, using largely visual clues, and then iterate to "just right". On either side of the correct settings, the sound was "peculiar"; get it right on the button and everything would sound "normal" - it was also possible to see the "sweet spot" visually; the waveform would not show an awkward shaping to the envelope of the peaks.

 

Geezus Xist! That’s a lot of work!

[fas42]

8 hours ago, miguelito said:

 

Geezus Xist! That’s a lot of work!

 

I wasn't fussed ... . The point, for me, was to see if it could be recovered, using a logical, iterative approach - which did work out. Obviously, this would be madness if every single recovery process was completely manual - but it means that intelligent, analysing algorithms could be devised ... the next step, one day ...

[gmgraves]

On 10/7/2017 at 1:45 PM, sandyk said:

 Hi George

 Even my older version of Sound Forge, (SF9) permits Expansion at various ratios.

 You wouldn't expect the freeware Audacity version to have a module like that.

 

Regards

Alex

 

 P.S.

 I have available a DL of before and after versions of a complete track in .wav format, using SeeDeclip Duo Pro,  if anybody is interested to see if they can hear any difference from before and after. The track had only minor clipping, although the posted waveform suggests otherwise. The differences are quite subtle, and you will need to increase the level of the declipped version by a couple of dB.

PM me if you would like the DL  links.

 

Well, Audacity has an adjustable software compressor, why not a similar expander module? 

[gmgraves]

On 10/7/2017 at 2:52 PM, fas42 said:

 

A very nice, zero!  ...

 

But still, no expander function that I can see. The compression ratio control, starts with 1:X, See no control marked X:1.

[fas42]

2 minutes ago, gmgraves said:

 

But still, no expander function that I can see. The compression ratio control, starts with 1:X, See no control marked X:1.

 

Look at ReaXcomp - the ratio goes either side of 1:1.

[fas42]

Just to mention, I'm making better progress that I thought I would!  It's been ages since I've done this, but it's coming back to me ...

 

Instruments are a big improvement, now getting the sweetness through - the 'recovery' is at a point where the variation between right, and not right, is in a fine balance - in the end, one gets the "right numbers", and none other are better - shouldn't be long!

[Audiophile Neuroscience]

Okay I'll go first.

 

I compared the original and treated (SeeDeClip) files from Alex of Better be home soon.

The treated I thought was a winner, not night and day but, with impression of less congested sounding, the voice a little more separated and front, and the intro tending to bloom as in build a little more noticeably. That was matching volume by ear with the treated needing a boost of 2dB (2 notches on Volume).

 

With Volume levelling engaged on JRiver (R128) the differences almost disappeared, maybe in favor of the treated. I fiddled with "adaptive volume" etc but couldnt get a clear winner.

 

Now the obvious conclusion is that when volume was accurately matched = no difference. I am not so sure as I have never consistently used JR volume levelling. It seems to veil the sound somehow. It needs another 10bB up on the volume control when used and I suspect it is either messing with the sound or I am incorrectly using it. Overall I prefer the sound listening without it. Without it I prefer the treated version.

 

In the past when volume matching I have been satisfied volume matching with a sound level meter and that was concordant with my subjective impression of a match. Crude I know.

 

 

[Audiophile Neuroscience]

6 hours ago, fas42 said:

Just to mention, I'm making better progress that I thought I would!  It's been ages since I've done this, but it's coming back to me ...

 

Instruments are a big improvement, now getting the sweetness through - the 'recovery' is at a point where the variation between right, and not right, is in a fine balance - in the end, one gets the "right numbers", and none other are better - shouldn't be long!

 

There's quite a loud and strident section around 2 min in. Interesting if you can tame this

[sandyk]

22 hours ago, Audiophile Neuroscience said:

Very, very interesting to compare the 'corrected' versions. I already have Alex's (SandyK) SeeDeClip version downloaded.

 

 Unfortunately, I didn't have the program set up properly with that one,  and the average level was too low.

[sandyk]

1 hour ago, Audiophile Neuroscience said:

I compared the original and treated (SeeDeClip) files from Alex of Better be home soon.

The treated I thought was a winner, not night and day but, with impression of less congested sounding, the voice a little more separated and front, and the intro tending to bloom as in build a little more noticeably. That was matching volume by ear with the treated needing a boost of 2dB (2 notches on Volume).

 

 David

 That is how I heard it too.

 However, this was when using cPlay

A while back, I did the complete album this way, and preferred the DeClipped versions.

 However , It's not just about DeClipping !

Regards

Alex