Mastering recordings on CDs

[AudioSceptic]

No, it's horrible mastering. Thank god the vinyl mastering exists, and that's only because they can't push the levels so hard in this format due to the technical limitations of the Vinyl medium.
View attachment 555954

Indeed, most of the Masterings on digital media have an extremely high level of loudness because they all are afraid of "losing to the competence" so they all believe that in order to be "competitive" you have to play their game too: how loud and compressed can I make this sh*t?

Vinyl, on the other hand, has technical limitations as I said, this forces engineers to decrease the limiter levels of their masterings. This is also another reason of why people wrongly feels that "Vinyl sounds better", thinking that the medium is "more musical", but most of them ignore the reality: it's simply because Vinyl has the best Masterings. CD's and digital streaming platforms could have these masterings too if they just decided to use the same vinyl mastering.....

But what puzzles me is that vinyl has a *much higher* noise floor and therefore a *much lower* potential DR, so it shouldn't it be the other way round?

 

[AudioSceptic]

Yes... Ian Shepherd wrote a little on this subject years ago. Vinyl always measures higher on the dynamic range scale (crest factor) because the type of noise that plagues phonograph record playback is impulsive, which causes the crest factor measurements to read high.

Nevertheless, it has been shown that record mastering leaves more dynamic range due to the fact that the needle cannot track a heavily clipped recording--it will simply jump out of the groove. This is why vinyl mastering always requires checks after cutting the master to make sure that a real cartridge needle can track the entire album.


In what way? Better is better. If you haven't heard the difference personally after a good demastering, you can't really profess its usefulness in increasing listening pleasure.


Sometimes this is true for albums and tracks done since the advent of more invasive mastering operations (since 1991)--and/or have a great deal of compression and clipping (typically DR values less than ~6 or 7 are usually lost causes).

But declipping, removing droning HVAC and electric power noise, re-EQing to minimize the negative effects of mastering EQ for low cost/low fidelity loudspeakers (also euphemistically called "translation"), and the use of a multiband expander all combine to give you a tool chest of techniques available to address common recording and mastering damage to recordings due to poor mastering practices.

I've used these practices on over 30k music tracks since 2015. They work very well. But effectively using these practices does require a little expertise. Perhaps that's what is bothering you...? Simply practicing using the techniques is the solution to that issue, I've found,

Chris

That definitely requires reading!

 

[Ste_S]

How much does DR actually matter? It always seems to be a stick vinyl enthusiasts wield to talk about the superiority of that format over digital. One of my favourite semi-recent albums is Low - HEY WHAT, and on CD it gets a DR score of 2. https://dr.loudness-war.info/album/view/195915.

Listening to it again now, and yep, the closing The Price You Pay sounds fantastic and is giving me goose-bumps. The distorted sound is very much part of the album. So what gives?

 

[Ste_S]

Another case in point - some music naturally has a lower DR (such as RHCP, I'd imagine). Here are the last four Floating Points albums

https://dr.loudness-war.info/?artist=floating+points&album=

Mere Mortals and Promises are done with orchestras, whilst Cascade and Crush are electronic dance floor-oriented, and unsurprisingly, they have a much lower DR.

 

[Chris A]

How much does DR actually matter? It always seems to be a stick vinyl enthusiasts wield to talk about the superiority of that format over digital. One of my favourite semi-recent albums is Low - HEY WHAT, and on CD it gets a DR score of 2. https://dr.loudness-war.info/album/view/195915.

Trying to understand what is occurring here, I undertook a quick analysis of the lead track (White Horses) from this album. Here is a screenshot of the waveform view within Audacity:

Its DR rating is 3 (dB) as-is.

Here is the waveform view after de-clipping (2.25 dB of total clipping in this track):

Its DR rating improved to 5 (dB) just declipping and normalizing the track. Here is the "Plot Spectrum" view from within Audacity, after declipping:

Here is the spectrogram view after declipping:

There is a fair amount of "splatter" on the vocals before declipping, and much less after declipping, although I didn't feel the need to re-EQ the track after declipping (as I've run into with other highly clipped tracks).

Even after declipping the track is not listenable to my ears between minutes 2:20-3:30--being mostly noise at that point.

As far as further actions to this track, I'm not sure what I'd add.

I'll stop here to wait for comments.

Chris

 

[krabapple]

When did this even start?

glimmerings in the early 1990s, charging ahead circa 1995.

Look up 'loudness wars' It's an extremely well-trodden subject in audio circles. There's already over a dozen threads about it here on ASR

 

[Tell]

Another case in point - some music naturally has a lower DR (such as RHCP, I'd imagine). Here are the last four Floating Points albums

https://dr.loudness-war.info/?artist=floating+points&album=

View attachment 556024

Mere Mortals and Promises are done with orchestras, whilst Cascade and Crush are electronic dance floor-oriented, and unsurprisingly, they have a much lower DR.

RHCP doesn't have naturally lower DR, it's just very compressed. Though there are music that are naturally lower in dynamic range, bass heavy music for example, having constant low bassnotes playing in some electronic stuff stops the DR meters going especially high. I do listen to psychedelic bass/glitch/dub that do sound very dynamic, open and awesome, yet scores quite low.
And on the other side of the spectrum a simple piano piece will naturally score very high on dynamic range, yet neither of these sounds "better" than the other, just different.

 

[krabapple]

Even after declipping the track is not listenable to my ears between minutes 2:20-3:30--being mostly noise at that point.

As far as further actions to this track, I'm not sure what I'd add.

I'll stop here to wait for comments.

Chris

The online "DR rating" is a crude measure. And 'dynamic range' -- the difference between the peak level and lowest level in the content -- often isn't literally what is being measured. More like 'crest factor' (peak vs rms). There are issues with that, too.

For example, you can alter those numbers simply by changing EQ, no compression needed.
And comparing LP to CD 'DR' is a mess, as Ian Shepherd has documented.

What DR tool are you using?

 

[AudioSceptic]

How much does DR actually matter? It always seems to be a stick vinyl enthusiasts wield to talk about the superiority of that format over digital. One of my favourite semi-recent albums is Low - HEY WHAT, and on CD it gets a DR score of 2. https://dr.loudness-war.info/album/view/195915.

Listening to it again now, and yep, the closing The Price You Pay sounds fantastic and is giving me goose-bumps. The distorted sound is very much part of the album. So what gives?

It's tracks like this that show that DR is only one part of SQ in music reproduction, and that it is very much dependant on genre. It sounds pretty damn amazing!

 

[kemmler3D]

In what way? Better is better. If you haven't heard the difference personally after a good demastering, you can't really profess its usefulness in increasing listening pleasure

I think it's questionable in terms of fidelity. Better is better but it's subjective. Any method of increasing dynamic range of a mixed song, including stem separation before expanding, involves new distortion. This may be less bad than the excess limiting you're trying to get rid of, but that's a matter of opinion, and the fact of reducing fidelity to the recording is not.

This may be totally fine in your view but it's questionable from a POV that values fidelity first.

 

[Chris A]

What DR tool are you using?

I'm using the older TT Offline Meter plugin for foobar2000 (usable in foobar vers. 1.5.1, not usable in vers. 2.0 and later, IIRC). It measures the second highest crest factor for the track (also IIRC).

The online "DR rating" is a crude measure. And 'dynamic range' -- the difference between the peak level and lowest level in the content -- often isn't literally what is being measured. More like 'crest factor' (peak vs rms). There are issues with that, too.

However, the clipping real.

This may be totally fine in your view but it's questionable from a POV that values fidelity first.

Usually, by the time I get to declipping and significant re-EQing, notions of track fidelity are usually a figment of the imagination. I just try to get the tracks to the point where I can listen to them, if they are that damaged.

YMMV.

 

[AudioSceptic]

Trying to understand what is occurring here, I undertook a quick analysis of the lead track (White Horses) from this album. Here is a screenshot of the waveform view within Audacity:

View attachment 556025

Its DR rating is 3 (dB) as-is.

Here is the waveform view after de-clipping (2.25 dB of total clipping in this track):

View attachment 556026

Its DR rating improved to 5 (dB) just declipping and normalizing the track. Here is the "Plot Spectrum" view from within Audacity, after declipping:

View attachment 556027

Here is the spectrogram view after declipping:

View attachment 556030

There is a fair amount of "splatter" on the vocals before declipping, and much less after declipping, although I didn't feel the need to re-EQ the track after declipping (as I've run into with other highly clipped tracks).

Even after declipping the track is not listenable to my ears between minutes 2:20-3:30--being mostly noise at that point.

As far as further actions to this track, I'm not sure what I'd add.

I'll stop here to wait for comments.

Chris

Thanks for that. Can that amount of clipping ever be justified?

 

[krabapple]

I think it's questionable in terms of fidelity. Better is better but it's subjective. Any method of increasing dynamic range of a mixed song, including stem separation before expanding, involves new distortion. This may be less bad than the excess limiting you're trying to get rid of, but that's a matter of opinion, and the fact of reducing fidelity to the recording is not.

This may be totally fine in your view but it's questionable from a POV that values fidelity first.

Far be it from me, a habitual upmixer, to raise an eyebrow at 'demastering' everything -- whatever floats your boat -- but there's no need to make compression a demon.
As noted, it's been used in recording, mixing and mastering, since before the CD era -- to make music sound *good* as well as a less noble but legit use, to make it 'competitive' with background noise (less noble still: to make it louder than a competitor recording). Like any tool, it's possible to abuse, and there are some very bad examples out there. But heavy compression can also be an intended *sound*, esp as applied to a stem or instrument in a mix

I know of one remix where such compression was removed from the lead instrument (Hammond organ) -- making it sound like weak sauce indeed. Yechh.
(FWIW, it's ELP's Tarkus as remixed by blessed holy St. Steven Wilson. He decided not to remix any more ELP after that.)

 

[Chris A]

Thanks for that. Can that amount of clipping ever be justified?

That's the question that I punted back to the community. I've never really tried to repair any tracks with that amount of clipping, so it really becomes an issue of "what do you like to listen to" or "how much noise or induced odd-order distortion are you willing to listen to".

Again, the notion of "fidelity" isn't something that enters the mind's eye at that juncture--at least in my experience. We're talking about levels of manipulation of the tracks that are very far from the notion of acoustic fidelity and freedom from distortion and noise. I know that there are a lot of people that "go there", it just doesn't include me, however...

YMMV.

Chris

 

[krabapple]

I'm using the older TT Offline Meter plugin for foobar2000 (usable in foobar vers. 1.5.1, not usable in vers. 2.0 and later, IIRC). It measures the second highest crest factor for the track (also IIRC).

I think there are better ways developed since then

e.g.

However, the clipping real.

Yes, clipping real. But clipping real doesn't always mean clipping audible. The duration/clustering of clipping matters too.

And one could of course make a file that clips but still has a high DR.

 

[kemmler3D]

Far be it from me, a habitual upmixer, to raise an eyebrow at 'demastering' everything -- whatever floats your boat -- but there's no need to make compression a demon.
As noted, it's been used in recording, mixing and mastering, since before the CD era -- to make music sound *good* as well as a less noble use, to make it 'competitive' with background noise. Like any tool, it's possible to abuse, and there are some very bad examples out there. But heavy compression can also be an intended *sound*, esp as applied to a stem or instrument in a mix

I know of one remix where such compression was removed from the lead instrument (Hammond organ) -- making it sound like weak sauce indeed. Yechh.
(FWIW, it's ELP's Tarkus as remixed by blessed holy St. Steven Wilson. He decided not to remix any more ELP after that.)

Compression is a very useful tool and outside of classical you will be hard pressed to find any recordings that don't use *any*. But like any studio tool it's bad when it calls attention to itself in a non artistic way. I think that is unfortunately the case for a lot of loudness war recordings.

 

[Chris A]

I think the album with the greatest amount of sustained clipping that I've dealt with is an older album from Infected Mushroom called Vicious Delicious. I remember a fairly deep amount of clipping before declipping, and a very high DR rating after declipping. I can listen to those tracks for a few minutes now, but the genre isn't my favorite, so I can't say that I really listen to them voluntarily.

Yes, clipping real. But clipping real doesn't always mean clipping audible. The duration/clustering of clipping matters too.

You talk as if you make a living doing this to music tracks. If this is a true statement, it is my suggestion to simply agree-to-disagree on what constitutes "acceptable mastering practices".

Personally, I find that the illusion of realism in acoustic tracks basically disappears once the DR rating falls below ~17 (dB). I've written about this on diyAudio in years past. For instrumentation that doesn't support a 17 dB crest factor, I find the illusion of realism to be solely based in psychoacoustic cues that give the listener a sense of ASW and surround channel delays--meaning, it takes more than two independent channels of music to create a sense of realism, in my experience.

YMMV.

Chris

 

[Erogen]

Low - HEY WHAT
White Horses

1. Reduced RMS volume level.
2. A 20 Hz/-24 dB high-pass filter was applied

Nice vocals, but everything else is terrible. Drama.

 

[AudioSceptic]

I'm using the older TT Offline Meter plugin for foobar2000 (usable in foobar vers. 1.5.1, not usable in vers. 2.0 and later, IIRC). It measures the second highest crest factor for the track (also IIRC).


However, the clipping real.


Usually, by the time I get to declipping and significant re-EQing, notions of track fidelity are usually a figment of the imagination. I just try to get the tracks to the point where I can listen to them, if they are that damaged.

YMMV.

We have some real experts here so this seems a good time to ask about a python script I have that uses ffmpeg to assess DR. Is this a valid process?

"Implements the classic TT Dynamic Range Meter (DR) algorithm:

1. Split each channel into non-overlapping 3-second blocks.
2. For each block, compute RMS (scaled by sqrt(2)) and the peak sample.
3. Take the loudest 20% of blocks (by RMS), average their RMS in the
power domain -> RMS_top20.
4. Find the 2nd-highest peak sample in the whole channel -> Peak2nd.
5. Per-channel DR = 20 * log10(Peak2nd / RMS_top20).
6. Track DR = round(mean of per-channel DR)."

 

[Chris A]

Note that my experience with dynamic range measurements of the Offline Meter type is that they are really measuring bass dynamic range, not really psychoacoustically relevant dynamic range for higher frequencies.

Chris