Using compression in chamber/classical music

[anabolique]

When do You use compression/expansion-(dynamics processors in general) in chamber/classical mixing process? Off course, I am talking about recording in digital domain, so fitting the dynamic range to recording media is not an issue.

[joelpatterson]

I'd say, in general, like a' so:

Real life has a dynamic range of 150 decibels or so, but CD life lives in a much narrower range-- zero at the top, and when you get to -30 dB's down, that's pretty quiet.

So, it would follow that compression is an aid to squeeze the 150 dB swings into 30 dB swings.

My approach, anyways....

[anabolique]

Mmmm.... well, 16 bit audio has dynamic range of 96dB's (maybe we could say 98 but thats another story) but if we stick to 6dBs per bit, than 24 bit audio has 144dB. By lowering resolution, we can fit those 144dBs to 96dBs.. but that doesnt involve use of dynamic processors... And I am afraid that 150 dBs of dynamic range wont happen even with large orchestras... maybe with some special effects, like cannon fire
And I dont get why 150 dB of dynamic RANGE should end with 30dB SWINGS?

[boojum]

I think that 120dB is about the noise level of a jet engine, far above what a chamber group is putting out. Romantic period orchestras will be putting out higher volumes and Mahler, Wagner, Strauss and so on are quite loud. I doubt that any reach 120dB.

Nevertheless I resort to compression because not everybody will be listening to the CD in a pristine and still environment. Many in a noisy home or a car. Now, how much compression? I do it by ear. Soloists almost as loud as a full crescendo are unreal. I have no out-of-the box answer. Try various amounts of compression until you are satisfied with the result.

[anabolique]

Originally Posted by boojum

I think that 120dB is about the noise level of a jet engine, far above what a chamber group is putting out. Romantic period orchestras will be putting out higher volumes and Mahler, Wagner, Strauss and so on are quite loud. I doubt that any reach 120dB.

Nevertheless I resort to compression because not everybody will be listening to the CD in a pristine and still environment. Many in a noisy home or a car. Now, how much compression? I do it by ear. Soloists almost as loud as a full crescendo are unreal. I have no out-of-the box answer. Try various amounts of compression until you are satisfied with the result.

Yeah, that seems reasonable to me...OK, using compression to taste off course.. but I would like to hear some ideas on possible applications which are "beyond simple " narrowing the dynamic range... maybe some situations when using compression on some channels, not the whole mix? e.g. you have "close" stereo pair for piano , one mic for vocal, and ambient mics and you put compression of certain type on vocal, maybe another on piano or not... some situations where you dont like what you hear from the mic and you try to correct the character,articulation of the instrument? stuff like that..

[rumleymusic]

I rarely put compression on classical music except for solo voice, where it may need a little help to stay consistent in the mix. I view dynamic processing as a form of correction rather than enhancement. Quite frankly, the idea of lowering the dynamic resolution of any piece for the sake of boosting overall loudness is completely insane to me. I have heard too many recordings ruined by poor use of dynamic processing. Chamber music and classical music in general requires intense dynamic contrast to be effective.

[Corran]

On the flip side, I have personally been disappointed in some recordings in which the softest parts were virtually inaudible. I have no qualms about moving the volume knob but if it's excessive it can be annoying.

I usually apply some type of compression on most things I record but it varies wildly on the music, ensemble, etc.

[aracu]

Another approach is to adjust the volume envelope according to your
human perception of what needs to be adjusted, rather than having it done automatically in a machine-like process.

[joelpatterson]

Originally Posted by anabolique

... I dont get why 150 dB of dynamic RANGE should end with 30dB SWINGS?

Well, maybe because the effective listening range given off by a CD playback device is from 30 dB's down (quiet) to zero (deafeningly loud.)

Silence in real life is -150, while silence in CD life is -30.

(If this is at all mysterious or suspect to anyone, God help you.)

Here's a perfect example of what we're talking about, a symphony orchestra at a wedding last weekend, playing Copland's variations on "Simple Gifts."

4shared.com - online file sharing and storage - download Variations on a Shaker Melody -- Copland.mp3


The quietest parts, just a clarinet or two, may and must not be that much way softer than the loudest parts, everyone roaring away at the finale.

So this is where compression comes in-- creating a rise in volume that mirrors the real volume rise, but through a smaller absolute range, say the range from -30 to zero. Rocket, meet not science.

[anabolique]

Originally Posted by joelpatterson

Well, maybe because the effective listening range given off by a CD playback device is from 30 dB's down (quiet) to zero (deafeningly loud.)

Silence in real life is -150, while silence in CD life is -30.

(If this is at all mysterious or suspect to anyone, God help you.)

No man. Youre wrong. There isnt such thing as negative dB value in real life. Negative dB values are refering to dBFS - dB Full Scale, that is maximum load on converter.... As I said, -96 is silence in CD. In real life there is 20 microPascals as threshold of hearing. that is 0dB SPL. and than you have positive values. And , if You think that -30 is silence on CD, then put some loud stuff in your DAW, preferably pop song , something with almost constant level near zero, or just loop some loud part of orchestral tune hitting near zero. Then lower that by 30 dBs . If You dont here anything, then... God help You
Btw, 30 dBS should be only 4 times softer in perceived loudness

[NorseHorse]

WTF does "4 times softer in perceived loudness" mean anyway?

I know this: when listening many environments (normal home / work), dropping the volume 30dbFs is going to make things a lot hard to hear.

Anabolique, please read up on the term "noise floor".

Recording the Wren Organ: What is the "noise floor".

[panatrope]

Originally Posted by joelpatterson

Silence in real life is -150, while silence in CD life is -30.

(If this is at all mysterious or suspect to anyone, God help you.)

While some may recognise that New Yorkers (even the state variety) may be special, the rest of the world generally recognises that the human auditory window lies between the threshold of hearing, 0dB SPL (20 microPascal) and the threshold of pain 120dB SPL (though this may be driven higher by habituation - eg., Metallica fans but the threshold of hearing is correspondingly raised). A figure of -150dB down from this as a threshold of hearing is not supported in any literature I am aware of.

Likewise, while some of us may be lucky enough to work in environments where the monitors may deliver a peak loudness of 115 dB SPL in an environment with a noise floor of 15dBA, the great majority listen in circumstances where the noise floor is higher (30 - 40dBA in a domestic environment) and the maximum sound pressure is lower (due to energy conservation or relationship with the next door neighbours). So while, 120dB maybe the theoretical maximum long-term dynamic range most of us are capable of appreciating, the reality is that 100dB is probably closer to the mark in everyday experience.

In the short term it is even less than that because the ear has its own gain-riding mechanism. Exposure to loud sounds results in temporarily raising the threshold in some or all of the hearing band, and reduces the discrimination of softer components around the louder sound (masking). So instantaneous dynamic range appreciation may be less than that quoted above.

Now, in the world I inhabit a symphony orchestra may have a maximum dynamic of about 50dB from a pp solo to a sfz climax, peaking just over 110dB SPL at the front row of the audience. And they are playing in an environment with noise floor of around 15dBA, which is slightly more than the noise floor of the mics I use for recording . Basically I can set the gain to keep the peak below 0dBfs and touch nothing - and the dynamic will fit nicely in the theoretical 98dB peak to noise window of 16bit recording (and either the hall background noise or the mic noise floor will provide the necessary dithering component). For traditional chamber music the dynamic range is rarely greater than about 30dB - usually about 20B - so the issues are not so great.

Now in preparing for broadcast and playback, I need to recognise that most listeners will listen to this performance in an environment where the noise floor is around 40dBA, and the maximum sound pressure level will be constrained to about 90dB. Experience suggests that in these circumstances an allowable dynamic is about 30 dB where listening is done in 'foreground' mode, and that the quietest sound needs to be at least 10dB above the noise floor. In many circumstances the allowable dynamic may be more like 20dB. Some form of manipulation is required to manage the dynamic range while preserving the impression of the original musical dynamic. This usually consists of manually raising low-level passages to fit within the allowable dynamic, and reducing the gain in anticipation of subsequent peaks so that the instantaneous musical dynamic is preserved as much as possible. (This is the mastering art of 'potting' as applied to the cutting of vinyl masters.)

It it usually not possible to leave this process to automatic processing, but on occasions where time pressures demand it a carefully set up compression process of no more than 1.5:1 can work to fit a 30dB performance dynamic into a 20dB transmission dynamic window without too many audible artefacts. But in the end it all depends on your target market. If it is the ability to listen while driving on the highway at 100km/h with the windows down is your market then the above may not apply to you. (I recall the story of classical station simulcasting on AM and FM - the FM broadcast had a target dynamic range of about 30dB, while the AM broadcast was put through a StaLevel holding about 40dB of gain at lineup, and a resulting dynamic range of about 5dB).

[anabolique]

First I want to say... thanx panatrope! Its nice to hear some info that seems valid.

Originally Posted by NorseHorse

WTF does "4 times softer in perceived loudness" mean anyway?

I know this: when listening many environments (normal home / work), dropping the volume 30dbFs is going to make things a lot hard to hear.

Anabolique, please read up on the term "noise floor".

Recording the Wren Organ: What is the "noise floor".

Its maybe stupid, it means that its stated (after some experimnets) that 10dB drop is perceived as "halve the volume" for majority of people,and 10dB increase of signal is perceived as doubling the volume. So... from 0 to -10 is 2x softer, from -10 to -20 is 3x softer, from -20 to -30 is 4x. Thats just an approximation, not even my personal feeling ... BUT, -30 is definetely NOT CD SILENCE! Especially if compered to some fantom -150 dB "real world"silence. Of course that you wouldnt hear soft passages if you lower the level by -30dB, I said listen to signals near zero-thats why i recommended some "loud" music-heavily compressed and without much dynamic fluctuations.. or to loop loud part near zero dB. I've done that and level is low but clearly THERE. Yes, I know what noise floor is. And there is noise floor of the media , which is , I repeat -96 on CD, noise floor of the hall, etc. ....

[joelpatterson]

Okay. You've officially blinded me with science.

[anabolique]

Cool, I am glad that we are not arguing anymore Now, lets get back to compression !

Edit: I made a mistake!!! Sorry....

If -10dB from OdB is halve the volume,thus 2x softer, then another 10dB drop from -10dB to -20dB is 4x softer then 0dB level, and finally -30dB should be 8x softer... that seems logical... But its CLEARLY audible! Assuming that youre listening at moderate level!

[Yannick]

On topic:

NEVER !
If everyone would at least bother to calibrate to K20, it would be blatantly obvious the eg a string quartet WITHOUT compression/limiting going to 0dBFS is much too loud on CD.

The only time one could use some is on full scale orchestral stuff, but just to get those 3-4 hard hits below 0dBFS.
Better : do it manually.

If some passage is really too soft, mix it properly. But take care, why do 80 people sweat for hours to get that really soft passage really soft&beautiful, to get the louder parts completely destroyed by the next guy with a compressor and a limiter ?

[Yannick]

Dynamic range of an orchestra is about 70dB max.
A string quartet will barely get 40dB
Piano solo: about 30-35 dB

Why would you compress to get it on a 96dB medium ?
The only thing we should do is get some uniformity in replay volume, so some stuff will be slightly to soft.

A lot of stuff will be too loud: harpsichord solo, string quartet playing Haydn, pianoforte playing Mozart etc etc.

[Yannick]

But of course, everybody can continue using compression on classical stuff, that will only make our own recordings better

[DistortingJack]

To be honest, on classical music it will take less time to draw a volume curve than to find the best compressor for the job and still not having it transparent enough. There isn't transient information to shape, just macro-dynamics to manage.

[Petrus]

Originally Posted by Yannick

Dynamic range of an orchestra is about 70dB max.
A string quartet will barely get 40dB
Piano solo: about 30-35 dB

Why would you compress to get it on a 96dB medium ?
The only thing we should do is get some uniformity in replay volume, so some stuff will be slightly to soft.

A lot of stuff will be too loud: harpsichord solo, string quartet playing Haydn, pianoforte playing Mozart etc etc.

As far as I know the largest dynamic range on classical CD:s is about this 70 dB, and I have also been able to record a chamber piece with 70 dB dynamic range, but the ensemble included a gran cassa with hard mallet. So in real life there is no reason to compress anything for technical reasons even for CD (which can resolve almost 110 dB with good dither, by the way).

Actually I can think of only one situation where compression or rather limiting would be beneficial: a few really loud hits like rimshots which, untamed, would result in a CD with average level unnaturally low.

What comes to peak levels with different material I think it is all right to aim for almost 0 dBFS to get the most out of the system, the listener must be able and clever enough to play solo and chamber pieces at a bit lower level than symphonic works. Nobody in real life has a calibrated monitoring setup at home with level pot glued in place, and at the mercy of the mastering engineer. I know where the reference point is in my system, but depending on material and how I feel and what I need the music for (serious or background listening, etc.) I can adjust accordingly, softer and sometimes even louder than the norm (117 dB SLP is the max continuous level I can produce).

I master loud music to -0.3 dBFS max peaks, chamber & solo to about -3 dBFS to reflect a bit what's on the disk.

Negative dB SLP values? Yes, the scale continues past zero, -10 dB SPL is ten times too soft for us to hear but that is quite irrelevant to us humans. Technically it is there, though. So you could argue that the dynamic range of real life is something like 200 dB, from insect whispers at -20 dB SPL to sperm whale's squid-killing yell (180 dB SPL) and let's forget the pressures at the center of a nuclear blast for a moment. But that is certainly academic as far as music recording is concerned.

[Audiop]

[nitpick-mode on]
Actually in the range where we are most sensitive we do in fact
hear about -10dB SPL.[/nitpick-mode off]


/Peter

[David Spearritt]

Originally Posted by Audiop

[nitpick-mode on]
Actually in the range where we are most sensitive we do in fact
hear about -10dB SPL.[/nitpick-mode off]
/Peter

[extreme nitpick-mode on]

Only for pure tones in isolation.

[/extreme nitpick-mode off]

[boojum]

There are two different audiences for the mixes, and ourselves. The majority will be listening on mediocre playback gear and in environments with lots of ambient noise.

Then there is the purist group which is very serious about music, has dedicated room for listening and really good gear, maybe sound treatments and a live-end/dead-end setup.

And then there is us, hunched over the keyboard of our DAW's, trying to make the best possible mix and being ever-so-critical. And I read it here: the mix is never finished, we just give up.

So how much of the serious listener's needs do we sacrifice to make the sound work for the ordinary listener? As for ourselves, there is no pleasing us.

[Audiop]

@David...


@boojum... +1



/Peter

[panatrope]

Originally Posted by boojum

So how much of the serious listener's needs do we sacrifice to make the sound work for the ordinary listener?

This is where that crucial faculty called judgement and experience comes into play. I have always believed that it is important to listen to the results of your handiwork in 'typical' environments, to hear the effect of the decisions you made in production. This means listening to your CD or broadcast in the car or on the family room boombox or kitchen portable (and particularly in mono!) at 'normal' listening levels and, for broadcasting, after your station's transmission processor has had a go at it. And if the result impacts adversely on the appreciation of the musical work, then you know better for next time.

Did someone famous once say: "You can please some of the people all the time, and all of the people some of the time, but you can't please all the people all the time!" (Actually, I think it was me. )

[boojum]

^^^^^^ I do. I listen on the monitors at near and mid field. I check them for fine stuff on my Ety ER-4b's and listen to them on the pretty good set with an old pair of KEF 104/2's. I also take it out to the car and drive around with the CD playing on the car player. It is the car which drives me to add a bit of compression even though I would rather not.

I know I cannot get a really perfect mix for all but I do try to get close. It is all about judicious compromise.

Sailboats are generally either fast and wet or slow and comfy, but we try to have them be both. There's where genius shines forth. The Canadian Nonsuch is my vote in sailboats.

[NorseHorse]

BTW, when I am speaking of compression in this context, I'm also referring to manual compression using envelopes or fader-riding to control dynamics.

Ever listen to http://pipedreams.publicradio.org/ and get frustrated cause you can't have hear anything below mezzo-piano? I do.

Do I always use some form of compression? No. Am I afraid to use it? No.

[rumleymusic]

BTW, when I am speaking of compression in this context, I'm also referring to manual compression using envelopes or fader-riding to control dynamics.

On the flip side, I have personally been disappointed in some recordings in which the softest parts were virtually inaudible. I have no qualms about moving the volume knob but if it's excessive it can be annoying.

Technically, compression is only the process of automatically attenuating peak levels past a certain threshold. This is the process I have a problem with.

Manually boosting an inaudible section for me is no more a sin than spot miking, eq,or reverb. You still achieve all the dynamics that section can achieve, just at a level that fits better in the whole scheme of the piece. While it is nice to have a large dynamic contrast, certain hall acoustics, miking techniques and player skill levels can exaggerate dynamics in an unrealistic way. Fixing these problems is just a matter of necessity for a good recording.

[Ozpeter]

I'd regard overall compression of classical music as unnecessary and undesirable. However, hard limiting to constrain the transients of unique and/or accidental peaks, achieved in such a manner that the performers on hearing the recording would be unaware of its use and would feel that the original range from ppp to fff had been maintained, can assist in avoiding a recording being released with levels well below what would be regarded as normal.

Use of such hard limiting would leave 99.9% of samples unchanged (apart from an overall amplitude increase into the headroom made available by constraining 0.1% of transient peaks).

If you inverted the phase of such a recording vs the original (with matching increase in amplitude only) you'd have silence except for very occasional brief "spitches", almost like clicks, where the limiter had cut in. If there was recognisable musical content in the "spitches" it would mean that the limiter wasn't 'hard' enough or had been set to limit excessively, in my opinion.

[panatrope]

In the above discussion, there is a difference to be observed between CD and broadcast. In the case of CD, one can assume that the listener has more control over the time and place of listening than a broadcast (radio, that is; music on TV is a whole different issue, and TV broadcasters are even less enthusiastic about wide dynamic range - as you are squeezing a picture of the wide world through a small rectangular aperture, you obviously have to do the same for sound).

For my particular station, one has to recognise that 99% of listeners (including those who pay money) are using the station's output as Classical Muzak. If the level stays below mp for more than 30 seconds, they think the station has gone off air. Less than 1% are listening in "foreground", and then only at certain times of the day.

An interesting opportunity presents itself when DAB+ eventuates (Real Soon Now). It will be possible to feed the DAB+ chain with unprocessed output, and listeners can be offered the choice of material with full dynamic range as chosen by the engineer/producer, or modified via the usual FM processing chain. (We will ignore for the moment the damage done to the signal by the DAB+ compression and woefully inadequate bit rate allocated). However, one can also observe that as more than 90% of DAB+ devices currently sold have speakers less than 10cm in diameter, the omens are not propicious. Stay tuned!

One other point affecting radio presentation of music - announcer's voice levels. Stations like to have announcer's voices at maximum level (especially when promoting membership of the station or mentioning sponsors), so the announcer is always at ff level, even though it should come across more like mf or mp. It is part of the perceived dynamic of the station, and sets the level expectation for a piece of music, and limits the impact of say, an ff entry. The BBC did a lot of work on this (albeit in the Golden Age of Radio) to define listeners expectations of voice following music and music following voice. It is an important part of the (critical) radio listener's experience.