[Lex2]

   
I'm not an expert on music theory... but I was under the impression that classical composers used (what we call) dynamic range for this very purpose... to create a sense of emotion or drama. I don't know what said composers actually call this, but the idea is the same. 
 
I have no horse in this race, so feel free to ignore this particular opinion. 

 
You have no argument from me there, with regards to classical recordings. I was mainly referring to "mainstream" (for lack of a better word) music. Even then, even for classical, the mix/mastering guy has to take into account the typical noise floor of a listening environment for classical, which may be as high as 30-40 dbA (crickets, outside traffic, dishwasher downstairs, etc). This would suggest a minimum of about 40-50 dBA for the quietest passages (about 10 dbA above the noise floor). One can raise the lowest passages manually to preserve the musical dynamic to appear like the original. Or another way to do it is to set up a low ratio compression (perhaps 1.1 to 1.5 max) to fit a wide performance dynamic into a smaller transmission window. Also, you'd need some hard limiting to catch any stray or accidental peaks in recordings that might otherwise be missed by the audio guy.

 

[RRod]

   
You have no argument from me there, with regards to classical recordings. I was mainly referring to "mainstream" (for lack of a better word) music. Even then, even for classical, the mix/mastering guy has to take into account the typical noise floor of a listening environment for classical, which may be as high as 30-40 dbA (crickets, outside traffic, dishwasher downstairs, etc). This would suggest a minimum of about 40-50 dBA for the quietest passages (about 10 dbA above the noise floor). One can raise the lowest passages manually to preserve the musical dynamic to appear like the original. Or another way to do it is to set up a low ratio compression (perhaps 1.1 to 1.5 max) to fit a wide performance dynamic into a smaller transmission window. Also, you'd need some hard limiting to carch any stray or accidental peaks in recordings that might otherwise be missed by the audio guy.

 
That lines up with the stats on my classical collection. Quietest passages (with actual music) are at about -65 dBFS RMS; setting that at 40dBA gets me 105dBA peaks which seems reasonable. My listening room is at about 35dBA.

 

[spruce music]

  Regarding paragraph 1: My fault for not discerning that the thread was full of people in the recording business.  My comments were targeted at the listener.

Yeah, no kidding.  With all respect to gregorio, who makes many excellent points, and who provides us with a real insight into those in the business, it seems his approach is like many.  They are going to give us loud properly compressed music for our own good no matter how much we don't think we want it.
 
I would say more, but don't want to go around in circles already discussed.  It does make me think the effects of the loudness war are permanent and the war was lost forever.  Loud it is, loud it shall be since with digital we can be louder than ever before.  Heck Sony/Philips should have made the CD an 8 bit medium, more than enough dynamic range once you use dither.  I guess the market agrees too.  The more compressed the sound gets, the more money and sales there have been in music.  OH WAIT a Minute...........!!!!!!

 

[pinnahertz]

  Yeah, no kidding.  With all respect to gregorio, who makes many excellent points, and who provides us with a real insight into those in the business, it seems his approach is like many.  They are going to give us loud properly compressed music for our own good no matter how much we don't think we want it.
 

The "funny" thing about the loudness war, regardless of which battle field you're on, broadcast or recorded music, is that every single listener has this thing called a volume control, and they all use it...a lot.  They adjust "loudness" to whatever they want it to be without regard to how the recording was made.  The loudness war is only a relative war, one tune relative to another.  And thats strange because...again...every listener has that volume control, usually easily within reach, and they will dive for it if they need to.  
 
Just before someone says something about background music in a bar or restaurant...it's all been leveled, completely and purposefully, so that a tune mastered "loud" doesn't ever play any louder than the others.  New idea?  Nope...juke boxes have done that automatically for 60+ years.  Same for DMPs with Replay Gain and SoundCheck...it's a done deal. 
 
User volume control is a fact of life in all audio except public venues, and theoretically, they've done it some other way.  Thus negating the point of the war just about completely, outside of the collateral damage that everyone is stuck with.  

 

[gregorio]

  This will be overly simplistic and probably shot at yet again, but...peak energy provides impact and excitement.  Lack of it greats a flat wall of dynamics with no variation.  No variation is constant and unchanging stimulus, and that translates to lack of excitement and emotion.

 
I think generally we're on the same page. I will just mention one further point: Yes, with heavy compression one looses pretty much all the transients and I agree with you that a flat wall, with of dynamics with no variation would be boring. However, we need to remember that compression/limiting just reduces the range of variation of the waveform's energy and that the amount of energy is only one of the properties the brain uses to perceive loudness. In other words, even with little/no variation in the waveform's energy (caused by very heavy compression/limiting) we can still create considerable variation in perceived dynamic range. The brain doesn't just use energy levels to discern loudness, it also uses changes in the number of tones and the number, relative amounts and distribution of harmonics and of course, the human ear is more sensitive in the mid band of frequencies than the high and low bands. Classical music composers knew (or at least intuitively felt) and used these facts in their compositions but the composers/creators of modern popular music use/push these facts way further, aided by modern technology. In other words, EDM (and some other genres) is specifically composed for heavy compression. This is why a heavy but acceptable amount of compression on say EDM is almost certainly unacceptable on classical.
 
Having said all this, there are some tracks of some genres where there appears to be little/no energy variation and no dynamic range either, some metal sub-genre tracks for example! Personally I find that extremely boring but it's not my job to dictate what genres should exist or how composers/creators should create and indeed I have mastered some thrash tracks.
 

  [1] They are going to give us loud properly compressed music for our own good no matter how much we don't think we want it.
 
[2] It does make me think the effects of the loudness war are permanent and the war was lost forever.  Loud it is, loud it shall be since with digital we can be louder than ever before.

 
1. As I've mentioned in previous posts, compression doesn't just affect amplitude, it affects tonal balance and positioning and above I've broadly explained that compression is a very significant part of the compositional process itself (of many modern genres). Saying that the amount of compression should be a decision for the end user (listener) is like telling painters they can only use very intense colours in their painting and that the viewer of the art should be free to desaturate, reduce contrast colour curves/balance, etc. It's never going happen, artists will never willingly accept such creative limitations/restrictions.
 
2. No, I don't believe the loudness war is permanent. In the TV world, loudness has to a very large extent been controlled in recent years. Prior to a few years ago, levels were controlled in TV but as mentioned above, levels and loudness are not the same thing. There used to be a very large variation in loudness, particularly between TV programs and TV commercials, even though the levels were usually much the same. Today, in many countries, this is no longer the case as TV audio is now spec'ed to loudness (rather than to levels). This same change to loudness is gradually being applied by many music streaming and online services like youtube, and the AES has recently published specs to standardise this. It's going to take a while though, the AES specs are recommendations, not legal requirements and it's going to take a good long while for composers/creators of modern popular genres to fully appreciate all the implications, how best (and worst) to work with this new paradigm. Because very heavy compression is effectively punished with the new loudness specs, ultimately, some genres will evolve, maybe even to the point of new sub-genres evolving, just as they did to take advantage of very heavy compression in the first place. It's not going to happen overnight though.
 
G

 

[Lex2]

The loudness range (LRA) of songs has not changed very much in the last 20-30 years so much of this may be a red herring. Crest factors have fallen, and RMS values have risen, no doubt, but that is akin to stylistic changes prevalent during the day, with more movement of the RMS to the right of the distribution. Kind of like changing Levels in Photoshop - aren't photographs so much brighter now compared to say a decade ago because everyone can manipulate the bits much, much easier now? You get all these highlights bunching up near the right, although the range of values stays more or less the same. The movement now is to get the RMS to move back to the center, and increase the crest factor again.That may well be happening now, even though not signed into law, apart from broadcasting. Online distributors such as Apple are standardizing loudness for iTunes tracks, and have been for quite a while.

 

[Don Hills]

...Having said all this, there are some tracks of some genres where there appears to be little/no energy variation and no dynamic range either, some metal sub-genre tracks for example! Personally I find that extremely boring but it's not my job to dictate what genres should exist or how composers/creators should create and indeed I have mastered some thrash tracks....

 
Even metalheads appreciate good sound:
http://www.metal-fi.com/
 
On the general topic, I recall some words from an expert:
 
"Do you think the loudness wars have finally reached critical mass? I'm having that feeling, yeah. Hopefully this whole loudness war thing that we've been through with the CD, there's no more room to go any louder. These things are just stupidly loud and annoying to listen to. There's quite a big backlash. I'm thinking people are realizing that there is a musical price to pay to have your iPod on "shuffle" and have your song be the loudest thing. I admit, to put out something that won't sound as loud as what comes before and after it on a iPod shuffle does take a certain amount of guts as a producer, but more and more producers and artists are getting back into dynamic range again.
As a mastering engineer, people sometimes blame us for things we have no control over. Mixers themselves have found that A&R departments, and certain record company executives, wouldn't approve their mixes unless they were already that loud. So the mixers, in self defense, started premastering stuff before they send it to me, and pre-squishing it and compressing it. And then we get it, and everybody in the band is used to this by now, and if you don't give them something back that's at least that loud, they think you're not very good.It's a very deadly situation.
So what do you do? Well, you try to educate. There's a couple of bands I work with, like Tool- I remember Danny Carey, the drummer from the band, walked into the studio session and said, "We don't care if our record's the loudest record on the radio, we just want to have the quality of what we've achieved in the mix," and I just have to admire that." - Bob Ludwig

 

[pinnahertz]

  The loudness range (LRA) of songs has not changed very much in the last 20-30 years so much of this may be a red herring. Crest factors have fallen, and RMS values have risen, no doubt, but that is akin to stylistic changes prevalent during the day, with more movement of the RMS to the right of the distribution. Kind of like changing Levels in Photoshop - aren't photographs so much brighter now compared to say a decade ago because everyone can manipulate the bits much, much easier now? You get all these highlights bunching up near the right, although the range of values stays more or less the same. The movement now is to get the RMS to move back to the center, and increase the crest factor again.That may well be happening now, even though not signed into law, apart from broadcasting. Online distributors such as Apple are standardizing loudness for iTunes tracks, and have been for quite a while.

'The loudness range (LRA) of songs has not changed very much in the last 20-30 years so much of this may be a red herring.'
Please cite a reference that backs this statement up.
 
'Crest factors have fallen, and RMS values have risen, no doubt...'
Please explain how RMS values can rise, but loudness hasn't changed much in the last 20-30 years.
 
'aren't photographs so much brighter now compared to say a decade ago because everyone can manipulate the bits much, much easier now? You get all these highlights bunching up near the right, although the range of values stays more or less the same.'
Please provide a reference that backs up this statement.
 
'The movement now is to get the RMS to move back to the center, and increase the crest factor again.'
If you're referring to the "movement" against the loudness war...hardly.  It's been pretty much ineffective as it targets the listener who has no ability to influence how music is produced, other than not to buy it.  Nobody does that.  
 
"That may well be happening now, even though not signed into law, apart from broadcasting."
If only.  In radio broadcasting loudness remains completely unregulated.  There's no evidence in recorded music that the war has even slowed slightly. 
 
"Online distributors such as Apple are standardizing loudness for iTunes tracks, and have been for quite a while."
Again, if only.  Apple downloads are simply AAC versions of whatever they get as the standard release.  They apply no changes, no loudness adjustments at all, and nothing is standardized.   However, iTunes (and the version of iTunes found on IOS devices) has a feature called Sound Check, which works like Replay Gain, and scans a track for loudness, then applies a volume offset during playback only.  The file is unchanged.  Sound Check is off in iTunes by default, but on in iTunes Radio by default.  What they've done is provide a means for listeners to negate the loudness offset incurred by aggressive processing by making a volume control correction.  It does not undo the side effects of that processing at all, and makes no actual modifications to the files themselves. 

 

[spruce music]

   
1. As I've mentioned in previous posts, compression doesn't just affect amplitude, it affects tonal balance and positioning and above I've broadly explained that compression is a very significant part of the compositional process itself (of many modern genres). Saying that the amount of compression should be a decision for the end user (listener) is like telling painters they can only use very intense colours in their painting and that the viewer of the art should be free to desaturate, reduce contrast colour curves/balance, etc. It's never going happen, artists will never willingly accept such creative limitations/restrictions.
 
2. No, I don't believe the loudness war is permanent. In the TV world, loudness has to a very large extent been controlled in recent years. Prior to a few years ago, levels were controlled in TV but as mentioned above, levels and loudness are not the same thing. There used to be a very large variation in loudness, particularly between TV programs and TV commercials, even though the levels were usually much the same. Today, in many countries, this is no longer the case as TV audio is now spec'ed to loudness (rather than to levels). This same change to loudness is gradually being applied by many music streaming and online services like youtube, and the AES has recently published specs to standardise this. It's going to take a while though, the AES specs are recommendations, not legal requirements and it's going to take a good long while for composers/creators of modern popular genres to fully appreciate all the implications, how best (and worst) to work with this new paradigm. Because very heavy compression is effectively punished with the new loudness specs, ultimately, some genres will evolve, maybe even to the point of new sub-genres evolving, just as they did to take advantage of very heavy compression in the first place. It's not going to happen overnight though.
 
G

Well this is the sort of thing I had in mind.  You tell me compression effects all these other aspects and I agree.  What is up to the end user is whether or not that user likes the result.  Quite a few don't.  The level of very, very compressed music which is currently the norm doesn't make me listen and think, "how creative that is", it simply sounds bad to me.  You insist I don't get it.  I don't.  I don't like it either.  I hear some artists with a 20 year history, they do a new album, it is squashed.  It doesn't seem to improve their artistry or sound.  I hear it and think, "this could have been good, but it is so damn compressed it takes life out of it. It compresses some of the artist's musical skills into oblivion." I fail at finding how the current trend in mastering or recording or whomever made it this way is any benefit whatsoever.  None of which has any effect on changing the current fashion.  Unlike you I don't see the fashion abating or ending.  I hope I am wrong about that.

 

[Lex2]

  'The loudness range (LRA) of songs has not changed very much in the last 20-30 years so much of this may be a red herring.'
Please cite a reference that backs this statement up.
 
'Crest factors have fallen, and RMS values have risen, no doubt...'
Please explain how RMS values can rise, but loudness hasn't changed much in the last 20-30 years.
 
'aren't photographs so much brighter now compared to say a decade ago because everyone can manipulate the bits much, much easier now? You get all these highlights bunching up near the right, although the range of values stays more or less the same.'
Please provide a reference that backs up this statement.
 
'The movement now is to get the RMS to move back to the center, and increase the crest factor again.'
If you're referring to the "movement" against the loudness war...hardly.  It's been pretty much ineffective as it targets the listener who has no ability to influence how music is produced, other than not to buy it.  Nobody does that.  
 
"That may well be happening now, even though not signed into law, apart from broadcasting."
If only.  In radio broadcasting loudness remains completely unregulated.  There's no evidence in recorded music that the war has even slowed slightly. 
 
"Online distributors such as Apple are standardizing loudness for iTunes tracks, and have been for quite a while."
Again, if only.  Apple downloads are simply AAC versions of whatever they get as the standard release.  They apply no changes, no loudness adjustments at all, and nothing is standardized.   However, iTunes (and the version of iTunes found on IOS devices) has a feature called Sound Check, which works like Replay Gain, and scans a track for loudness, then applies a volume offset during playback only.  The file is unchanged.  Sound Check is off in iTunes by default, but on in iTunes Radio by default.  What they've done is provide a means for listeners to negate the loudness offset incurred by aggressive processing by making a volume control correction.  It does not undo the side effects of that processing at all, and makes no actual modifications to the files themselves. 

 
This was a paper I read many years ago but I found the reference, finally.
 
http://www.nature.com/articles/srep00521/figures/4
 
The one you want is the interquartile plot on bottom right, which stays constant. The plot above makes sense, you see the median shifting to the right over the years as tracks get louder (the sample space was some 400,000 songs). Please don't use loudness and loudness range interchangeably. I was refering to loudness range calculated using moving, average windows.
 
I was only referring to online loudness, e.g. Spotify, Youtube, or Apple iTunes streaming. Any user can knock themselves out by downloading tracks and listening to them at a level that they find suitable, with or without Soundcheck. No law can ever be passed for controlling something like that. Apple also came out with Mastered for Itunes which is a set of guidelines and tools for processing and to ensure proper encoding. It's not law, but it's a start. It says to the mastering guy - this is all the loudness and crest factor you're going to get, so make full use of it. Any louder or if it clips, we'll turn it down anyway so get with the program.
 
Legal loudness specifications for broadcast (mostly TV) have received more headway in the EU (based on EBU guidelines). There is a CALM Act in the USA, but that covers commercials, I think. Radio in the USA is not regulated, you are right. The self-perpetuating belief is that people want to listen to the loudest channels. I have the suspicion that kids today are getting deafer by the day, but I have no data to back this up.

 

[pinnahertz]

   
This was a paper I read many years ago but I found the reference, finally.
 
http://www.nature.com/articles/srep00521/figures/4
 
The one you want is the interquartile plot on bottom right, which stays constant. The plot above makes sense, you see the median shifting to the right over the years as tracks get louder (the sample space was some 400,000 songs). Please don't use loudness and loudness range interchangeably. I was refering to loudness range calculated using moving, average windows.
 
I was only referring to online loudness, e.g. Spotify, Youtube, or Apple iTunes streaming. Any user can knock themselves out by downloading tracks and listening to them at a level that they find suitable, with or without Soundcheck. No law can ever be passed for controlling something like that. Apple also came out with Mastered for Itunes which is a set of guidelines and tools for processing and to ensure proper encoding. It's not law, but it's a start. It says to the mastering guy - this is all the loudness and crest factor you're going to get, so make full use of it. Any louder or if it clips, we'll turn it down anyway so get with the program.
 
Legal loudness specifications for broadcast (mostly TV) have received more headway in the EU (based on EBU guidelines). There is a CALM Act in the USA, but that covers commercials, I think. Radio in the USA is not regulated, you are right. The self-perpetuating belief is that people want to listen to the loudest channels. I have the suspicion that kids today are getting deafer by the day, but I have no data to back this up.

The paper will take a little study.
 
"Apple also came out with Mastered for Itunes which is a set of guidelines and tools for processing and to ensure proper encoding. It's not law, but it's a start. It says to the mastering guy - this is all the loudness and crest factor you're going to get, so make full use of it. Any louder or if it clips, we'll turn it down anyway so get with the program."
 
No, that's not what "Mastering for iTunes" says.  
 
From the document "Mastering for iTunes", subheading, " Be Aware of Dynamic Range and Clipping"
 
"Although iTunes doesn’t reject files for a specific number of clips, tracks which have audible clipping will not be badged or marketed as Mastered for iTunes."
 
(and from the last paragraph in that subheading):
 
(regarding loudness processing) Whatever you decide—exquisitely overdriven and loud, or exquisitely nuanced and tasteful—we will be sure to encode it and reproduce it accurately. We only ask that you avoid clipping the signal.
 
Earlier in the document, "clipping" is defined as resulting from levels above 0dBFS, and is NOT defined as sub 0dBFS deliberate clipping or brick-wall limiting.  Apple never says they will turn anything down.  

 

[Lex2]

iTunes Radio turns down to about -16 LUFS. Which is significant in itself because they reach about 20 million listeners. Individual tracks, however, are all over the place in terms of loudness, at least the last time I checked. You'll notice that even audio previews on iTunes are much louder compared to the same on iTunes radio. An automatic Sound Check, while good for turning down, may not be as good for turning up "quiet" tracks due to clipping.
 
You're probably refering to the deliberate clipping usually done to avoid artifacts due to the AAC codec. 1 dB level drop is normally a good rule of thumb, and that is what the Apple RoundTripAAC plugin (mentioned in the document) is for, i.e. present to the engineer sample and inter-sample peaks before and after encoding.

 

[gregorio]

Just to be clear, music on the iTunes Store has not been turned down or changed. It's only changed if the end user selects "replay gain", "sound check" or whatever they're calling their version of loudness normalisation. Streaming services such as iTunes Radio, Spotify and others apply loudness normalisation automatically and the end user has no option to disable it (as far as I'm aware). BTW, from about a year or so ago Youtube has also started loudness normalising all uploads.
 
The loudness war has long since gone beyond just extreme compression added in mastering, loudness is also maximised with EQ and as I've already mentioned, with the structure, composition and production itself. The loudness war is not going to end, it's going to continue BUT what's clever about loudness normalisation is how one achieves loudness. Loudness Normalisation effectively turns the loudness war on it's head! All those time honoured tricks and techniques used to make audio (sound or music) sound louder/stronger/fatter, with loudness normalisation actually makes audio sound quieter/weaker/thinner. For the loudness war is to continue (under the loudness normalisation paradigm), the techniques of achieving loudness are effectively, pretty much the opposite of what they are now. A wide variation in dynamic range (at the structural level as well as just less compression) is a good example because it can/will create the perception of being louder than a song with little/no dynamic range. Compression, even heavy compression, can still be added for artistic reasons on individual tracks but that has to be compensated for (under loudness normalisation) rather than leading to an upward spiral of more loudness/compression on all the other tracks in the song.
 
Of course, artists will still be able to abuse compression/loudness on their own CD's (or files they offer for download) but streaming services are a growth area, while download sales are starting to stagnate and CD sales actually reducing. Therefore, getting one's music to sound good (including if that means as loud or louder than other releases) on a streaming service will gradually gain a higher and higher priority for artists/labels and that means dealing with the consequences of loudness normalisation, which in effect, means the end of the loudness war. That future is not absolutely certain but does appear very likely.
 
G

 

[castleofargh]

yup, to me the most significant blow against loudness war are stuff like EBU (european broadcasting union) R128. I believe it mostly started to stop having the commercials on TV sounding twice as loud as the movie, which was pissing everybody off, but it's what the audio industry needs.
and indeed as it's the perceived loudness that is regulated, low dynamic audio ends up being lowered the most. so that's what needs to be enforced everywhere as it effectively kills the original purpose of the loudness war. which a digital peak limit like "mastered for itune" does not, that just took care of most potential intersample clipping in the playback. and I don't know if it does anything else?

 

[gregorio]

  ... it's the perceived loudness that is regulated, low dynamic audio ends up being lowered the most. so that's what needs to be enforced everywhere as it effectively kills the original purpose of the loudness war. which a digital peak limit like "mastered for itune" does not ...

 
As this is the science forum and many come here for a better understanding, it's maybe worth expanding a bit what loudness normalisation actually is:
 
Until recently, all audio was limited by peak level. Limiting by peak alone allows for a wide range of perceived loudness, for example the more time we spend at or very near peak level, the louder it will sound and of course the ear is far more sensitive in certain frequency bands. Obviously then, we can make something sound much louder than something else (with an identical peak level) by creating music (or soundscapes) with a consistently loud structure, by adding lots of compression and by maximising the EQ in the most critical hearing bands. Loudness normalisation addresses all these areas. About 15 years or so ago, the ITU (International Telecommunications Union) started work creating a loudness measurement, they took existing data (such as the Fletcher/Munson equal loudness contours) and combined that with their own studies to come up with a loudness curve, from which they created an algorithm for a filter (called the k-weighted filter).

To measure the loudness of a piece of audio, first the k-weighted filter is applied, then the RMS of that result is measured. The end result is a measurement called the Integrated Loudness and is at the heart of the ITU's published standard (ITU-R BS, 1770). Note that both the EBU's R128 and the ATSC (Advanced Television Systems Committee) A/85 specs are entirely based on ITU-R BS. 1770, although confusingly these two organisations have decided on different names for the ITU's Integrated Loudness scale. LUFS (Loudness Units relative to Full Scale) in Europe and LKFS (Loudness K-weighted relative to Full Scale) in North America. Along with various other measurements, the ITU-R BS. 1770 also includes a new peak measurement, True Peak (dBTP) rather than sample peak (dBFS). Although both the LUFS and LKFS scales are identical the actual specs required by the EBU and ATSC are slightly different: -23LUFS and -1dBTP in Europe and -24lKFS and -2dBTP in North America. These specs are for TV and would not be suitable for all devices (such as mobile devices) so within the last year, the AES has published recommendations for music/audio streaming (-16LKFS to -20LKFS) and the EBU is currently working on specs for Europe. BTW, iTunes radio fits nicely in that range but YouTube doesn't, being approximately -13LUFS.
 
So what in practice does all this mean? Let's take hypothetical song "A", a massively compressed victim of the loudness war, which peaks at 0dBFS, has virtually no dynamic range (and consistently averages let's say -11dB RMS). Hypothetical song "B" has a much large dynamic range, it has very quiet sections as well as very loud sections (which peak at say -2dBFS). Currently (without loudness normalisation), song A sounds louder and thicker/fatter than song B, even the loudest sections of song B would sound marginally quieter than pretty much any part of song A. If we apply loudness normalisation, say at -20LUFS/LKFS this is what happens: Song A will be significantly reduced in level. Not only will it be reduced by the 9dB difference between it's -11dBRMS measurement and the -20LUFS target but notice in the k-filter graph above that from 1kHz the filter actually increases the loudness measurement by up to 4dB or so. If, as would be likely, song A has been EQ'ed for loudness, then it's integrated loudness could be as high as -7LUFS. At the end of loudness normalisation to -20LUFS, song A would have been reduced by 13dB (and therefore peak at -13dBFS). Song B on the other hand might not have been reduced at all (and it would still peak at -2dBFS, 11dB higher than song A) and it's loud sections would now sound very obviously louder than any part of song A!
 
BTW, when looking for the k-filter graph image above, I came across this article in Sound on Sound, which is the best I've seen on the subject.
 

  You tell me compression effects all these other aspects and I agree.  What is up to the end user is whether or not that user likes the result.  Quite a few don't.  The level of very, very compressed music which is currently the norm doesn't make me listen and think, "how creative that is", it simply sounds bad to me.  You insist I don't get it.  I don't.  I don't like it either.

 
I understand but what you don't seem to understand is that much of the time I don't like it either! However it's not up to you or the "quite a few" because the "quite a few" you quote are not in reality actually very many. It's also not really up to me either, sure I have some influence over the tracks I master but that influence is constrained by the wishes of my clients (artists and/or distributors) and the mix I'm presented with to master.
 
The take away from the above and from what I've mentioned previously is two fold:
1. The loudness war, as commonly characterized to/by consumers, is not just a simple matter of compression ruining music and therefore the solution is not simply to abolish the use of compression. The DR database is good in that it highlights to Joe Public that the loudness war exists, in a format which is simple enough to understand at a cursory glance. On the other hand, it's so over simplified, it can easily lead to erroneous conclusions if used as the basis for any determinations of audio quality, and indeed there's ample evidence of this occurring here on Head-fi!
2. I don't doubt the loudness war will continue but to maximise loudness under the loudness normalisation paradigm will mean creating songs like song B rather than song A. Which as mentioned, is at least as much, if not far more about compositional structure and mix/production than it is about mastering compression/limiting! It's going to take a while to filter through the entire chain of content creators but looking to the future: The loudness war is dead, long live the loudness war!
 
G