SOS sept 2011 Dynamic Range & The Loudness War

[svenneke]

Hi all,

I did read the article in SOUNDONSOUND sept 2011 about "Dynamic Range and the Loudness War" and I have some remarks concerning what Emmanuel Deruty wrote. But first of all excuse me because english isn't my native language.

-1-

In magazine in the box "Loudness Range & Limiters" he wrote that by using a compressor/limiter you can raise the RMS level and this way add dynamic range. That way, he continues, one can get perhaps 100 or 105 dB instead of the normal 96 dB of a 16-bit file.

This is technically impossible I think. The loudest level in a digital file is 0 dB full scale. The amount of volume we perceive with such a file is determined by the level you set on your speaker/amplifier combination. The bitdepth determines how big the difference between the loudest level is without distortion and absolute silence. In case of a 16-bit file that difference is 96 dB. The bit-depth determines the amount of steps an audio-stream uses to change the levels. What he is saying is that when you raise the RMS level you increase the dynamic range. This is not the case, you simply raise the level of the average sound levels in the audio file and leave the peaks unaltered. The only way to raise the dynamic level is to use a higher bit depth (24 bit for example), allowing to use more steps and so to set a higher volume on your amplifier and allowing to play louder levels without distortion, again compaired to absolute silence. Well, this is pretty simple explained but I guess you get my point.

-2-

Of course we will experience a higher level if you raise the RMS, but it is not a natural way to hear sound like that.
Emmanuel Deruty compared the situation with photography and fortunately I am a photographer and have 25 years of experience with analog and digital imaging. The picture of the tower (which is missing in the printed article btw) doesn't seem natural because it does not relate to the way we see things in reality. The average level is raised to much and everything seems to bright. In real life our eyes adapt to our environnement in such a way that we see an average level between the brightest parts and the darkest parts of what we are looking at, but the brightest parts will always dominate and dictate that behaviour. Something similar is happening with our hearing and this fact is detrimental for the way we perceive "hot" mixes.

I explain how.

We adapt our sensitivity of our sight with the help of the aperture in the iris, our pupils. If what we see has brightness differences within a normal range, our pupils will open to a certain diameter to allow the brightest part and the darkest to be visible at the same time, without hurting our retina.

However if the difference is very high (a high dynamic range) like in situations with strong lights and lots of dark shadows, we will experience the average level of our view to be much darker because our pupils adapt to the dominant bright lights to protect our eyes. That is the reason why night-scenes with light spots look almost black or very dark in most parts of the image. I like to compare also with a view of someone with the sun behind him/here. We can hardly see the person an will only see a silhouet. We are only able to see a certain level difference at the same time and no more than that. If the real world situation exceed that difference we adapt to the brightest part and the mid, darker and the darkest parts will shift towards the dark, without detail.

The same phenomenom is happening with our hearing. The muscles and filtering in our ears initiates a process that performs according to the same law. If we hear loud transients, they have the same effect as bright lights: our ears adapt themselves to those peaks and we experience an average level that is somewhat lowered. However if we use heavy compression/limiting and we raise those average levels again to unnatural hights, our hearing system tries to protect our ears again even more and stiffens our muscles/sensors and lowers our sensitivity.
At the same time by using compressors/limiters we shift the quiet parts upwards and certainly the mid-level volume parts of the sound to levels we never hear in the real world, compared to the peak levels.
Also, to be heard equally loud, mid to high frequencies do need a lower amplitude than low frequencies (bass, kick-drum). If compression/limiting raises all those levels of mid/high frequencies because they belong to the mid level amplitude part, we will perceive the mid to high frequencies much more pronounced than we would normally do. To me this is the most disturbing effect of the increase in levels of todays mastered music. The balance between frequency and natural level spread of those frequencies is altered in an unpleasant way.

Also, to compare with photography again, there is a certain subtle effect in image level editing that, when you raise the average level of a histogram to much, (compare that to the 2007 mean distribution curve in the audio graph), it causes steps that become to big in the low levels. With audio, a smooth curve is then streched so much that the digital steps that represents the audio signal are getting too big and begin to look like a stair. This adds tiny amounts of harmonics that were not present in the original file. Unfortunately most of the time these harmonics do sound unpleasant to our ears because they are not musically related to the songs frequencies, unlike the harmonics introduced by some hard/software designed especially for that.

-3-

The graph in the article with the mean distrubution curve is interpreted as giving the same dynamic variability between 1967 and 2007. This is a wrong interpretation. Even if the blue line seems pushed towards the right, the same lenghts of those lines cannot be interpreted as the same dynamic variability because those lines doesn't take the whole curve into account, but just a by the author well choosen part of it. It especially doesn't take into account the behaviour of our ears to adapt to the loudest levels.

Compared to pictures it is like you just look at the average color/brightness levels and forget to look at the bright and dark parts of the picture. It is not because the brightest and darkest parts are at both ends of the curve in a picture histogram, that they don't exist and don't count in the way we see the picture.

The same thing applies to sound. We hear the whole range from left to right in the curve from 1967 as can be seen in the article, but only if the average levels are indeed "average" (normal gauss-curve). But if the average level is raised (shifted 2007 gauss-curve to the right), the tendency to adapt to the louder levels, causes loss of information in the low levels because also our hearing has a limited range of level difference we can experience at one moment. The "deep blacks" Emmanuel is talking about, dissapears because of the domination of the loudest levels. Therefore the "contrast" does NOT remains intact but is also lowered. In other words, you are loosing information in this case, just like in the picture of the night-scene I wrote earlier about, simply because our own ears are triggered too intense and protect themselves against damaging by lowering our sensitivity.

Therefore the difference we have to look at is not the blue line Emmanuel drawed, but the distance from the middle of the top of the gauss curve that represents the average sound levels, to the utmost right that represents the loudest peaks in our music. The left side of the top of the curve that represents the lower level sensitivity, will also be smaller because our hearing is becoming less sensitive for low-level sound due to the higher average levels. Less sensitive means a smaller width on the graph. If you look at the curve that way your can see clearly that the range is much less in the year 2007 versus the year 1967. Again, remember that our sight/hearing always adapt itself to the brightest/loudest parts of the information we see/hear, meaning the utmost right position in the graph.

Looking at the loudness war problem this way, we can definitly conclude that the dynamic difference between the peak levels and the top of the gauss curve that represents the average level, is much less in 2007 than in 1967. Because the top of the gauss-curve represents the biggest part of the level of the samples of the song, we can see that the loudness war has indeed lowered the dynamic range (difference) of todays music because we adapt our hearing to the loudest parts and therefore loose information in the lower level parts of the music.

I hope that this explanation can help to understand what is happening because I found a lot of comments on the internet about this article asking at the author why, if the dynamic range hasn't changed, the hot mastered songs do sound so fatiguing and not as nice as "normally" mastered ones.

Although I have to admit that there is some clever mathematics done in the article and of course I could be totally wrong... in the end I am only a photographer.

Greetings

Sven Bontinck
Belgium