Music Into Noise: The Destructive Use Of Dynamic Range Compression

I think lex's comments are, as usual, not only terrifying but very succinct and dead on. you don't notice what's being done to the dynamic range if you don't have a halfway decent stereo or quality headphones

This isn't really true. Hot Fuss sounds flat to me played through my tinny computer speakers or my shitty iPod earbuds or my car stereo w/busted right speaker. It sounds just as flat through good speakers or headphones. I think it does partially boil down to the music the lex listens to because minimal hip-hop doesn't sound terrible through hypercompression (though I can't say the same abt the pop and R&B he's into -- that would probably benefit from improved dynamics).

i notice it even in rap, the clipse last album was tiring as hell to the ears.

Yeah, well I know that a difference between me and some of my peers is that listening to music is its own activity for me. I try to avoid "background music," but that's what compression does, really - it makes everything into background music.

I've heard lots of people say that they don't mind compressed music for listening in the car, because they need to drown out the road noise or whatever. But all you have to do is turn it up loud enough so you can still hear the quiet bits over the road noise. Now the consequence of that is that the loud parts become really loud, meaning you can't have a conversation with someone in the car. That's fine with me - I'd often rather listen to music than converse on a car trip. But I know not everyone feels that way. (He Poos Clouds is not very good road trip music)

er, multiple xposts there

I think the music is compressed by frequency range to some extent as well -- that's probably why it seems more convenient for car listening, b/c just cranking up an uncompressed record means that you're getting a lot more unwanted piercing treble. (Or something, I'm just talking out of my ass on this one.)

I like lots of elements, lots of texture and timbre, and lots of little sounds. I like to hear horsehair, fingers on strings, rattling wires in a snare drum, the hum of the electricity in a sequencer.

Let's not go overboard here, I think what you're talking about maybe has more to do with recording (and mixing) techniques than mastering compression.

he's not saying that it has anything to do with mastering compression -- he's just saying the kind of music he likes is the kind of music that is harmed most by hypercompression

Are there some good examples of recently remastered/reissued albums which suffer from this over-compression? Other than Raw Power, where it seemed to be the whole point of the exercise.

-- everything, Wednesday, 6 June 2007 19:45 (10 minutes ago) Link

There's a whole audiophile-ish subculture based around comparing different generations of CD reissue, which has a tendency to get very anal, tends to automatically decry any new reissues, or so it seems to me, and it's a bit snake oil-ish/follow the guru-ish as well. I believe the last lot of Roxy Music remasters were pretty bad, but I didn't buy them, so can't report 1st hand.

I'd be interested to hear from anyone who purchased the new versions of Pulp's mid-90s albums or something of that nature. That seems to be the kind of music that suffers from this. Also, "This Is Hardcore" was the first album I remember thinking sounded terrible.

Some of the Cocteau Twins remasters were pretty horrendous.

Re: Pulp-- Rest assured their final album isn't compressed much, since Scott Walker was the first person I ever read decrying compression. It was in some article around '97 or '98 where he was asked what he thought about Radiohead, and he replied something like "I think they're really good, but they use too much compression just like everybody else."

I can't get a handle on Pulp's last album, soundwise. Sometimes I think it's great, othertimes I think it's shit.

fig 4 Rihanna ft. Jay-Z - Umbrella

http://farm2.static.flickr.com/1119/533686637_55de8ef108_o.jpg

fig 5 mims - this is why im hot 2007

http://farm2.static.flickr.com/1173/533592022_a4267f218b_o.jpg

You should probably annotate those. What do they illustrate.

I don't think the Roxy Music remasters were particularly compressed, and I don't think they clip, but there is some dispute as to whether they are better or worse than the original. They are HDCD and probably sound best on a compatible player.

Oh, though actually I don't know if that was teh "last lot" of RM remasters--they were from the mid-90s.

er, late 90s. That's all.

he's just saying the kind of music he likes is the kind of music that is harmed most by hypercompression

Okay, that makes sense.

Compression doesn't really matter for a lot of the music I listen to (rap, electronic music, etc. where it's all programming and synths with not a whole lot of dynamics anyway) but it would be a problem with jazz, brass band, etc. Esp. because it can make cymbals and horns sound like shit.

well the vertical is volume. sort of. when i make recordings that how i see it. i could be wrong though. the thing about the hip hop pop stuff i posted is they seem to have a fair bit of space between the beats whilst the indie stuff up thread is all cramped up into one block. still too loud though. some else may explain better.

the only roxy i've ever known are from the early 00s. i love them. i wonder what they ought to sound like.

Play Autoditacker back-to-back with Radical Connector and say that compression doens't effect electronic music.

Hmmm? I'm intrigued. (Not intrigued enough to buy Radical Connector after what's been said about it, though).

MoM use loads of elements, loads of layers, loads of textures, etcetera; not dissimilar from, say, Guillemots, except that one is electronic and one is acoustic in predominant source. Turn each of those individual elements up and instead of occupying distinct space they start competing and overlapping, blurring with each other. Sure, a MoM song may not have the same, let's call it 'narrative dyanmic' in terms of quiet-loud-quiet-loud 'song structure', but the integrity of the individual sonic layers can suffer from over-compression, it can be clipped just as much as anything else.

Autoditacker is a more organic sounding album, though?

I do remember RC sounding kind of harsh and busy, but I think it would still sound harsh and busy if it wasn't really compressed.

xpost

The harshness and busyness is what compression does! There's, I'd wager, less going on in Radical Connector in terms of number of elements. The Von Sudenfed, while ace, is a touch too loud, loses some definition. Go back to Orbital Brown, Artificial Intelligence, etcetera.

no you can play basically any old 80s-90s idm/detroit stuff vs. post-2002 whatever and the same thing is going on. electrohouse etc. gets so hammered to the wall through multiband comps it's really really punishing on some sound systems IMO

I know their stuff up to and incl. Idiology. Maintaining space among clutter does seem to be the knife edge they're required to walk. By Idiology, I'm starting to feel like it's tipping too far into the antic.

Cos by Idiology they're too compressed. Tombot's totally OTM here.

xposts

I think the music is compressed by frequency range to some extent as well -- that's probably why it seems more convenient for car listening, b/c just cranking up an uncompressed record means that you're getting a lot more unwanted piercing treble.

Mastering compression doesn't (in theory) have anything to do with that. It operates strictly on dynamics. mp3 compression affects frequency range, though.

The images acrobat posted are graphs of volume vs. time in the left and right channel of a stereo audio file. I'm a little confused about this, though - audacity has two waveform views. acrobat's images come from the default view, called "waveform." I feel like the second view, called "waveform (dB)" is what I'm used to seeing in audio applications. The documentation doesn't do much to clarify the difference:

Waveform Traditional display of audio material. It displays the amplitude of the audio over time. This is the default display mode.
Waveform (dB) Like Waveform, but logarithmic instead of linear vertical units . It displays the amplitude in dB of the audio over time.

Reasons this isn't so much of a problem with rap/r&b hits:

- synths, programming, and short samples tend to have a smooth, constant level to begin with
- rap/r&b hits don't have a lot of internal dynamics
- the whole style and architecture of the genre are already built around this kind of treatment (and it could be argued that other genres are getting louder partly in a wrongheaded effort to compete with that)
- something like a Rihanna single is built for this kind of bursting assaultive energy, more so than for album-length close-listening

It's not an issue with SOME electronic music, but I think it's a huge issue with a lot of it, both the stuff that pumps super-loud (Goldfrappy-type -- haha xpost TOMBOT) and the stuff that actually wants you to be listening to its organization in the sound spectrum and whatnot -- electronic music has delivered some amazing organizations of sound and space available in few other formats, and loudness-maximizing kills a lot of the options for that! I mean, erasing the front-to-back part of the sonic field is possibly even worse than if you just erased the side-to-side part (i.e., went mono).

And of course the worst examples of this come along when music that wants to be sedate and comfortable and open to close listening -- whether its folk, classical, deep-and-warm indie, or Norah Jones -- pegs its levels to the norm and winds up flat and undynamic.

I'm increasingly chafing at the feel of modern masters, which are just flat and claustrophobic -- they have no air, no space, no depth -- but it's one of those things you wind up just having to bracket in order to follow music. It's strange to think that I used to be able to go to sleep comfortably when listening to music, and not because of overwhelming ear fatigue (when I wanted to be knocked to sleep by overwhelming ear fatigue I'd just put on too-dense shoegazer bands or Melt-Banana or something!) -- these days I'd be driven nuts by even the more sedate and cozy superloud records I own.

But where this really nags at me is with a lot of modern rock that goes really over the top with it, which might actually be pushing old-man buttons for me: maybe it sounds good to some kids when the snare drum makes everything in the mix squish and duck under it because there's no extra space in the mix -- maybe it makes it sound to some like the band is just TOO AWESOME AND ENERGETIC to fit in conventional sound standards -- but that'll just always be a "mistake" to me, and an ugly one. (Cf that "Tea Leaf Dancers" song by Flying Lotus that makes intentional use of Extreme Ridiculous Ducking Compression, which I find painful to listen to but some people find interesting.) Any music that even half wants to sound "natural" or like something that might really happen simply CANNOT accept a mix where the loudness of each part is constantly squishing around to add up to 100% peak level, because no guitar in the world rises and falls in presence/volume according to whether a tom has been struck in the past half-second.

This is a total side note, but it occurred to me a while back that certain mixing and stomp-pedal antics of 90s rock might have set the stage for people not taking note of certain things -- I kept thinking of the opening of "Smells Like Teen Spirit," which appears to use some fading and the channel change to do the whole compression thing where a single instrument might be louder than when the others come in. (CLEAN CHANNEL CLEAN CHANNEL okay here's the whole band)

But of course the average listener isn't going to notice plain ear-fatigue compression issues -- like Lex says, you're just gonna go "eh, this music's kinda tiring and uncomfortable-sounding" -- especially since the ear-fatiguing qualities of this sound remarkably similar to the qualities of a small, crappy speaker, like the ones on a cheap TV. The worst rock offenders seriously just sound like you've turned your TV up too loud and it's kinda blaring and assaultive.

xpost to Steve: mastering compression is TOTALLY frequency-band based -- that's one of the "advantages" of the digital software for this stuff, that people can work with the compression and EQ at the same time. All this stuff is getting multi-band compression.

Was it Nick who introduced the metaphor of a normal face vs. a face squished up against the glass? Cause that really is the perfect way of describing what this does.

Yeah, I know what multi-band compression is. But that just allows different frequency bands to be dynamically compressed with different settings. It doesn't, as Curt1s seemed to be implying, mean that the frequency range gets lessened.

But mp3 compression, if not taken down too far, doesn't necessarily squash the soundspace in the main way we're addressing here... is that right?

Face into glass wasn't me, I'm afraid!

oh steve g i just changed the view hmmm yeh now it all looks a lot messier

xp

No, any kind of lossy audio data compression shouldn't have a big effect on dynamic range (although depending on the material and the algorithm and the amount of data compression any number of weird artifacts can be created). Although I guess it kind of does in the sense that you lose resolution on quiet sounds, like reverb tails and cymbal decays and stuff.

The multiple common meanings of the word "compression" certainly don't help to make this issue clear.

no it doesn't...

dynamic range compression simply squashes peak volume, leaving the frequency range intact. mp3 compression actually does omit certain frequency ranges on the principle that the louder signals are more important than the quieter ones, and the listener won't miss the quieter ones -- so it subtracts out bass & harmonic frequencies, simplifying the signal -- that's why certain music sounds 'thinner' on mp3 -- there's actually less there. also you get raspy aliasing artifacts at lower kbps rates.

Autoditacker & Radical Connector are two great examples by the way -- Autoditacker is a fantastic headphone album with dozens of quiet details bouncing around and lots of creative use of dynamic range, and Radical Connector is just a huge monster truck

but I think it was meant to be that way, that was their intent. same with that Flaming Lips album you lead your article with, which was produced and mastered by the same person -- you may not like the aesthetic, but it was pretty clearly chosen. a lot of electronic & hip-hop music does not suffer from this approach because they are going for sounds that haven't been heard yet, and you have to admit, the sound is new. the problem is when other genres have to compete and albums are destroyed by mastering them to the same volume levels.

There's also a principle, the name of which escapes me, that puts a ceiling on the maximum frequency you can represent digitally - I believe the it's equal to half of the sample rate. So if the sample rate is 44.1 kHz (standard CD quality), you can represent frequencies up to 22 kHz, which is fine for almost all purposes, as the upper limit of human hearing is 20 kHz.

But a funny illustration of this principle is the dog whistle at the end of Sgt. Pepper's. I grew up listening to the CD version, and the sound was always clearly perceptible. But I'm told that on the vinyl version it's much less audible to human ears, and that due to the above principle the sound on the CD actually consists of subharmonics of the original sound.

I'm not quite sure at what point or in exactly what manner mp3 compression further limits potential frequency range.

You know, I dunno if it would be legal to do such a thing on Stylus or wherever, but Nick or someone should post a DLable mix (high bitrate by necessity of the subject) of tracks exemplifying the different points to be made.

There's also a principle, the name of which escapes me

Nyquist Frequency - http://mathworld.wolfram.com/NyquistFrequency.html

as for something that makes what we're talking about audible, this post from upthread (already rolled into concealed messages) is perfect -- listen to what happens to the drum intro when you squash the track, the reverb on the drum tail suddenly becomes huge, and the 1989 production suddenly sounds like a 2007 production, and it's entirely due to the mastering

great illustrative video here:

http://www.youtube.com/watch?v=3Gmex_4hreQ

-- Johnny Hotcox, Wednesday, June 6, 2007 12:39 AM (21 hours ago) Bookmark Link

well as lots of people who know about audio are on one thread, a question: i made the mistake of recording an interview in a pub. halfway through the interview the bass of the background music starts getting in the way of the voices, what is the best technique, using audacity hopefully, to turn down the bass? isolating speach would be great but i think that's a bit of a pipe dream. this is a WAV file btw.

acrobat, that's what equalizers are for. Try using a highpass filter around 100 or 120 Hz (i.e., cutting everything below those frequencies) and see if that helps. Although I think the audacity EQ is pretty awful.

is there any other freeware that would do the job better?

I was just talking to my dad about the Olympics logo and the epileptic thing, and logos in general and why the Olympic one had to be 'evolving', etcetera, and he said "it's because people are so obsessed with technology that they forget what they're meant to be doing". And I thought, that's what's happening here.

acrobat: well, I'm thinking of the default EQ plugin that comes with Audacity. There are links to other Audacity-compatible plugins here. I can't recommend one specifically, but I'm sure there are some good freeware EQ plugs out there.

this was mentioned on another thread, but since it hasn't been mentioned here yet, the logical solution is: car stereos and iPods with built-in compression algorhythms which the user can switch on & adjust themselves, so you can listen in your car to classical or jazz or chamber pop records that normally you'd only be able to listen to at home

the algorhythms are already out there (Volume Logic was initially called Octiv -- company was way ahead of their time, founded by people who used to work at Orban, who invented the sound of FM Radio compression in the 70's/80's with the Optimod, but Octiv was too ahead of their time so the company keeps getting sold off, but their time could still come).

It doesn't, as Curt1s seemed to be implying, mean that the frequency range gets lessened.

this isn't what I was implying, but I didn't word my post very well at all.

The Best of Fad Gadget, or at least the copy I have, clips horribly. I'm not even sure it's compression that's the problem, they just got careless with the volume!