Friday, July 22, 2011

Loudness Wars

As I just pointed out in a comment on the Atlantic article The Loudness Wars, this isn't about loudness at all. It's about clipping. The absolute loudness of a track is under your control, not the engineers. All the engineer can do is compress the dynamic. Here is an excellent short explanation:

The advantage to compression is that it makes a track using it stand out against tracks that don't. It also makes tracks more audible in an environment where there is a lot of background noise. So, big pluses in a lot of situations in the modern world. But, honestly? It makes music sound like shit. Sorry to be blunt. I read about this a while back, but the Atlantic article lays out the context. I've been wondering why so much pop music sounds so crappy and this is part of the reason. Another part is pointless hyperactivity, but I've already posted, briefly, on that. Meaningless arm-waving is a vice all music is prone to, but this punching up of everything, or chopping of the peaks, is a vice unique to pop music.

Here is some truthiness for you:
  1. You can't actually listen to music when there is a lot of background noise--all you can listen to is punched up sludge
  2. Music uses a number of different techniques to express things: a range of pitches, a range of harmonies, a range of rhythmic values and a range of dynamic values--it makes no more sense to flatten out the louds and softs than it would to use just one note in the melody
  3. Musical expression is not a function of digital or other compression, or of costuming, or of dance routines, or of whatever else you can find to stuff into the video
  4. I'm all in favor of sexiness, but sexiness and musicality are not actually the same thing.
Speaking of dynamic range, the right use of it can be one of the most powerful resources in music. The most amazing example I can think of is the first movement of Shostakovich's Symphony No. 7. I cannot embed a clip that will demonstrate this. You will actually have to physically go to an orchestral performance of it. The reason is that, after a six minute introduction, a long passage starts--usually known as the "Invasion Theme"--that lasts for about eleven minutes. The entire passage, a long passacaglia on a single theme, is a crescendo. It starts with a simple rhythm on a single snare drum and by the end there is the whole orchestra including three snare drums, five tympani, bass drum, cymbals, tam-tam, triangle, tambourine, xylophone, eight French horns, six trumpets, six trombones, tuba, full woodwind section including contrabassoon, and at least sixty strings. These instruments are slowly added, bit by bit, until all of them are playing at full volume. This incredibly wide dynamic range, from a single snare drum playing quietly to the huge orchestra playing loudly is UNRECORDABLE! No recording technology we possess can capture that enormous a range. So go find a performance, buy a ticket and get back to me.... OK? I have had the experience of hearing this in a live performance and it is not one you soon forget.

As I said, it cannot be experienced in a recording, but just to give you a slight taste of it, here is a YouTube clip of the first part. The first movement of the symphony is 25 to 30 minutes long, depending on the performance, so it has to be chopped up. The crescendo I am talking about starts around 6:52 in this clip:

and continues...


Doghouse Riley said...

Well, you make several important points, but compression and clipping are two different things.

You can compress the dynamics of an audio file for good or for ill, but you cannot compress anything into clipping.

For that you need to push the signal "into the red". You can do this between the mic and the pre-amp, the pre-amp and the recorder, or the recorder and the mastering phase. Or all of them.

With analogue systems you often get a pleasant increase in third-harmonics and a slight euphonic compression (IF you know what you are doing).

With digital recordings you end up with an unpleasant screechy buzzy distortion.

There are several valid reasons for a mastering engineer to compress the dynamic range of a file. I won't go into all that here. But I will note that the CBC did a study which found that there should be no more than 15dB between the lowest and loudest sound ... if you want to have everything heard properly in a moving car. And, yes, that included classical music. (I.E. the needs of radio are different than the needs of home audio.)

So that, as you mentioned, "compression makes tracks more audible in an environment where there is a lot of background noise."

But, as to your next assertion, "It makes music sound like shit", you are simply wrong. Shitty compression done by shitty engineers and/or shitty musicians and/or shitty producers makes music sound like shit.

When Bob Ludwig (for example) compresses the sound of a trumpet so as not to have it overwhelm the rest of the big band, he makes the recording better ... and you won't ever notice that he did it.

As to the 7th Symphony, I notice that an Amazon reviewer said: "I didn't notice any adverse effects of the extreme dynamic range that apparently bothered others; to me it seems the recording rather faithfully reproduces the extremes heard in the hall." And another said: "Gergiev indeed has put his personal stamp on this work as being "major league" Shostakovich, and the dynamic range of this Philips recording is phenomenal enough."

So apparently, it IS possible to get it all down with modern digital technology.

We won't even get into the way most conductors compressed the orchestra's dynamics when making analogue records for the 78 and LP markets.

Doghouse Riley said...

But the topic of conductors altering dynamics for recording is a fascinating one!

Bryan Townsend said...

Thanks, Doghouse! (Or should I say Mr. Riley?) I have spent quite a lot of time in CBC recording studios, but it was all on the other side of the glass. I appreciate you correcting me on clipping. So, if I understand rightly, clipping occurs when a dynamic spike exceeds the capacity of the medium: "goes into the red". Whereas dynamic compression takes the full range of dynamics--nothing in the red--and flattens it. OK.

I look at all this from a performer's point of view. A sound system has a certain capacity. A well-engineered recording is organized so as to use this capacity without exceeding it, correct? Therefore, according to, say, the CBC, you take whatever you have in the studio and fit it into a 15dB spread? And this means that everything can be "heard properly in a moving car". I honestly don't think you can listen to much classical music in a moving car. In order to hear, say, the pianissimos in a Haydn string quartet you have to crank up the volume on playback so they are audible over the car sounds. Then the forte blasts you out the windshield. Doesn't work for me. To listen to classical music with a range from, say, pp to ff, you need to be in a quiet environment.

That's one thing. The other thing is that I think--correct me if I'm wrong--that most recordings of classical music, especially orchestras, compress the dynamic range to some extent because otherwise, it would not fit into the capacity of a home sound system. If you sit in the third row listening to, say, the Shostakovich 7th, you will hear a much wider range than you will on a recording at home. This is just my feeling from having done both. And I had a pretty good system at home, though not worth thousands of dollars.

One last example: I was at one concert that I would bet would be pretty much unrecordable. It was for six percussionists and piano. Again, I was sitting near the front. About twenty feet from the piano. At one point all six percussionists were vigorously playing several large gongs each--an incredible wash of sound. The pianist was playing the piano as hard as he could. The lid was bouncing up and down a few inches! But it was quite inaudible. Gongs have an amazing effect live. They are like an almost physical blanket of sound--but it doesn't seem that loud!

In any case, just some experiences. I don't really know the engineering end of it, so educate away!

RG said...

Coming from my research in language translation, aspects of this discussion strike me as analogous to the oft heard claim that some things (words, phrases or idioms, works) are untranslatable. There is a logical error in that propostion. One need ask only what aspect of the original is not represented by the (allegedly inadequate) translation. Then use the indicators in the reply to make up the difference. [The fact that such algebraic formulation of a translation may be "ugly", does not affect the logic or ontology of its adequacy.]

In the case off music, though, it will not be possible even to indicate the aspects that are allegedly "unrecordable". That is, it is logically or epistemically impossible to identify something in the performance is not in the recording. Unless recourse is had to that pre-verificationis: Je ne sais quoi!

Bryan Townsend said...

Yes, I know what you mean, but surely there are more unrecordable--in practice, let me hasten to say, in practice, not in theory--sounds than recordable ones. We have, over the last roughly one hundred years developed ways of recording sound waves and playing them back that are suitable to those kinds of sounds that are in the range of human hearing and the kind that we like to listen to. There are a myriad of other sounds we have, mostly, not bothered to even try to record. Whale sounds are one of the few we have. (There is also a lot of technology developed for underwater surveillance for military use...) But we haven't tried to record, for example, the sound of a volcano erupting or the fusion reaction in the core of the sun, or the sound of a hummingbird's wings (or have we?).

Even in the case of concert music, I have heard many things in the concert hall that I have never heard on record. The listening experience in live concerts is quite different than it is with a home sound system. I suspect that spending a great deal of money on a huge home sound system and installing it in a large and well-designed listening space would partly span the gap. But only partly... If you had ever heard a live performance of the Shostakovich 7th Symphony, you would probably grasp my point immediately.

I can summarize my observation in this way: setting aside the capacities of the recording process itself (the limitations of microphones and storage media) the playback system itself has significant limitations. Speakers always have specifications: 60 to 20,000 Hz is a typical one. Also, dynamic range: there is both a ceiling and a floor to what a speaker can reproduce.

Now you can reply that the specifications of speakers are tailored to the human range of hearing, which is true, but I suspect that in the concert hall we have a more complete sensorium.

I do know that I don't sound the same recorded as I do in concert (nor the same from one hall to another).