Talk:Dynamic range compression

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Professional sound production (Rated B-class, Top-importance)
WikiProject icon This article is within the scope of WikiProject Professional sound production, a collaborative effort to improve the coverage of sound recording and reproduction on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
B-Class article B  This article has been rated as B-Class on the project's quality scale.
 Top  This article has been rated as Top-importance on the project's importance scale.
 
edit·history·watch·refresh Stock post message.svg To-do list for Dynamic range compression:
  • Clarify difference between non-linear transfer methods and "locally linear" variable gain methods, since both seem to be valid for various definitions of compression, expansion, limiting, and companding. Clipping is definitely non-linear transfer function, but others can be either.
  • Add images and information referred to on talk page discussions.

Side-Chain?[edit]

I believe the words "Side-chain" aren't used correctly here. The way it's defined right now is what I would call "Ducking" (there's a quick note about this alternate name). The page goes on to define "Parallel compression" as what I would call a "Side-chain effect." This page backs up my definition of Side-Chain: [1] I would define the three terms like this:

Side-Chain - using two copies of the same signal (only one with an effect applied), and mixing the two together.

Parallel Compression - This is exactly the same as Side-Chain Compression in my book.

Ducking - Using the amplitude of one signal to control the gain of another. The example of having background music fade out while someone is talking and fade back in when they finish talking is perfect.

I hope I'm not being too pro-active, but I'm going to change these right now.

Do you agree / disagree? Blue Dinosaur Jr 21:01, 12 November 2007 (UTC)

  • I've just found contradicting links that support the original defined term. As a result, I am not going to edit the original page. What kind of authority can we find on this subject to provide ourselves with a good source? All I've done is confused myself. Blue Dinosaur Jr 21:07, 12 November 2007 (UTC)
I don't see a huge problem with Side Chaining as it stands now. Perhaps it isn't as clear as it can be for the layman, but it isn't wrong. I'll give some thought to how it could be made more clear. Binksternet 21:14, 12 November 2007 (UTC)

'Dogshit' reference[edit]

Is it really necessary? I don't think it adds anything to the article and just has the ability to offend, but as a newbie to editing Wikipedia, I'll leave it to this article's regulars. 65.107.212.130 (talk) 22:23, 14 January 2008 (UTC)

I like it as a nice bit of straight-talking, though I won't insist on keeping it.--Father Goose (talk) 03:52, 15 January 2008 (UTC)
The quote should be cited. -- Another Stickler (talk) —Preceding undated comment was added at 18:22, 28 October 2008 (UTC).
I agree very strongly that Rip Rowan's quote should remain too. Indeed, it's a perfect word to describe the sound of many modern masters and the dreaded 're-masters' as well. Over compression is wrecking recorded music and that quote perfectly reflects the strength of feeling amongst (good) sound professionals who genuinely care about their work, and are repeatedly forced by know-nothings into ruining it with this idiotic practice. Paulzon (talk) 20:51, 10 April 2009 (UTC)

I laughed when I read that, and it's a very apt description. Keep it. —Preceding unsigned comment added by 75.164.238.106 (talk) 01:02, 31 May 2009 (UTC)

poor definition[edit]

This article is a good introduction to the topic, but it could use some editing. For example, its definition of a compressor would more-accurately describe a limiter.

Also, the lack of hardware brand names and a description of their products is a major omission. A detailed description need not be advertising.

WilliamSommerwerck (talk) 12:30, 23 March 2008 (UTC)

If you are capable of making those changes, by all means, do.--Father Goose (talk) 02:27, 24 March 2008 (UTC)

"Music production" and "Marketing" sections are reduplicative[edit]

The "Music production" and "Marketing" sections seem somewhat reduplicative. They should probably be combined. —Christian Campbell 02:22, 2 April 2008 (UTC)

Compression in Digital Arena[edit]

I removed:

"Compression in the digital arena is essential, as digital recording levels are very intolerant to clipping (over-modulation); they much more sensitive than recording tape used previously in analog recordings."

Because:

  1. It is confusing. Is it referring to compressing an analogue signal prior to conversion to digital, or digital compression when it is digital audio?
  2. It is not correct. That is, compression is not essential for digital recording. 24 bit digital has more than 140dB of dynamic range, so plenty can be spared for headroom to preventing clipping at A/D conversion. For 16bit or lower, it may be useful, but not essential. Proper gain structure may be all that is necessary. Regarding DSP compression for a digital audio signal, proper gain structure should prevent the need for compression to prevent clipping. Clipping is unlikely as bit-depth in modern DAWs is high enough to provide adequate headroom. In other situations it maybe useful, but not essential.

Iain (talk) 08:05, 17 May 2008 (UTC)

Good move. Thanks. Binksternet (talk) 08:09, 17 May 2008 (UTC)

Main Definition change[edit]

I changed:

"Dynamic range compression also called DRC (often seen in DVD player settings), audio level compression, volume compression, compression, or limiting, is a process that manipulates the dynamic range of an audio signal. Compression is used during sound recording, live sound reinforcement, and broadcasting to alter the perceived volume of audio. A compressor is the device used to create compression."

to:

"Dynamic range compression also called DRC (often seen in DVD player settings), or simply compression, is a process that reduces the dynamic range of an audio signal. Compression is used during sound recording, live sound reinforcement, and broadcasting to alter the perceived volume of audio. A compressor is the device used to create compression."

Because:

  1. I don't know of anyone who speaks about "audio level compression"
  2. It has been a common recommendation not to use the term "volume" with relation to professional audio. while "amplitude" is the bi-polar magnitude of signal, "level" is the uni-polar, and "loudness" is the perceived level, "volume" was never properly defined and is only common in domestic applications.
  3. Dynamic range compression is not limiting. Limiting is a subset of dynamic range compression, but the two terms are not interchangeable.
  4. "manipulates" changes to "reduces" as compression always reduces dynamic range.
  5. "perceived volume" change to "level" as the operation of a compressor is actual and not only perceptual.

Izhaki (talk) 13:41, 28 July 2008 (UTC)

Those aren't bad changes. By the way, I moved your Talk entry down to the bottom of the page as is the norm. Your rearranging of the Talk entries to put the most recent on top was not optimal. Binksternet (talk) 15:11, 28 July 2008 (UTC)
Looks good to me.--Father Goose (talk) 18:42, 28 July 2008 (UTC)

Attack / Release Image is misleading[edit]

Audio Compression Attack and Release.svg

If we want to be 100% correct, then the image above is misleading or possibly even incorrect. Reason is that the attack and release never correspond to the exact time it will take gain reduction to reach full effect or dive back to 0 dB. Instead they are always linked to a set amount of dB. Please see the main article for detailed explanation of this, and feel free to try it out with your compressors to see that 10 dB of gain reduction will take longer to diminish compared to 20 dB.

This image can be made correct if "phase" is used instead of "time"; so "attack phase" and "attack time". Iain - it's your illustration. Izhaki (talk) 13:40, 29 July 2008 (UTC)

Is that image trying to convey such specifics? I think it's more general than that. It doesn't look wrong to me. Binksternet (talk) 15:04, 29 July 2008 (UTC)
One can assume by this illustration that the actual figure you set on the control (say, the attack time 10 ms) is really the time it takes gain reduction to reach full effect. Completely up to you guys. Izhaki (talk) 15:41, 29 July 2008 (UTC)
Rane's Chapter 3 on Dynamics Processors says about the attack control that it can be thought of as specifying "...the time required for gain to settle to a defined percent of final value. Typical are 86% or 95% of the final value." The image above has no specified time of attack, so we aren't looking at whether it has hit within ~90% of its mark by that time. Binksternet (talk) 16:37, 29 July 2008 (UTC)
And then it goes on to say: "It is important to understand the difference between release rate -- as determined by this control -- and release time. There is no industry standard and different manufacturers define this control differently. Rane defines this control, in a compressor for example, as how long it takes for the gain to change by 10 dB, not how long it takes to return to unity gain (no gain reduction). To calculate the actual release time requires a little math: Release Time = (Gain Reduction x Release Setting) / 10 dB"
The main problem is that most compressors use the term "release time" to what is in practice "release rate" (and this is how Rane call it) - This is the core of the issue, wrong terminology. If you ask most people what is "release time" they will say that it is the time they set on the control on the compressor, not many people are aware that what they are setting is the release rate. At the moment the article defines "attack time" as what being set by the control, it does not fit with the current illustration. Either the illustration needs to change to "attack phase" or the article has to change to "attack rate". The latter change, so I believe, would cause confusion. I hope this is clearer now.Izhaki (talk) 18:40, 29 July 2008 (UTC)
I think 'attack phase' is the best term. To avoid any confusion, can we use 'attack phase' in the article text as well. For example: "A compressor does not react instantaneously to changes in the signal. The 'attack phase' is the period when the compressor is increasing gain reduction to reach the amount that is determined by the ratio. The 'release phase' is the period when the compressor is decreasing gain reduction to the level determined by the ratio, or, to zero, once the level has fallen below the threshold." What do you think? Iain (talk) 07:31, 30 July 2008 (UTC)
That's very good (although it does not resolve the original issue with the illustration), but it teaches the reader nothing about the actual attack and release controls. I think that the first thing we need to decide is the terminology to be used. How do we call what is set by the attack and release controls on the compressor? Attack time? attack rate? attack settings? Izhaki (talk) 08:14, 1 August 2008 (UTC)
  • The illustration is easily fixed, but I did not want to make changes if there is not general agreement on terms.
  • I intended the explanation of the controls to follow my proposal above. For example:
"A compressor does not react instantaneously to changes in the signal. The 'attack phase' is the period when the compressor is increasing gain reduction to reach the output level that is determined by the ratio. This happens when the input level increases (and is above the threshold). The 'release phase' is the period when the compressor is decreasing gain reduction to the level determined by the ratio, or, to zero, once the level has fallen below the threshold. This happens when the input level decreases. The length of each period is determined by the rate of change and the required change gain reduction. For more intuitive operation, a compressor's attack and release controls are labelled as a unit of time (often milliseconds). This is the amount of time it will take for the gain to change a set amount of dB, decided by the manufacturer, very often 10 dB. For example, if the compressor's time constants are referenced to 10 dB, and the attack time is set to 1 ms, it will take 1 ms for the gain reduction to rise from 0 dB to 10 dB, and 2 ms to rise from 0 dB to 20 dB.
  • What do you think? Can we agree on the terms 'attack phase' and 'release phase' at least? Iain (talk) 09:42, 1 August 2008 (UTC)
The explanation above and the use of 'phase' in it makes perfect sense to me. I have to non-related reservations though:
  • "A compressor does not react instantaneously to changes in the signal." - A digital compressor is perfectly capable of acting instantaneously on input signals. Not only that, but with the standard DAW dual-buffer operation, compressors can use look-ahead that lets them act on signals 'before they happened'. What's more, the attack and release are not acting on the input signal but on the gain reduction amount (and the sentence, as it stands, is more linked to peak/RMS sensing). I would change this opening sentence to: "A compressor might provide a degree of control over how quickly it acts."
  • "This happens when the input level decreases." This sentence has to be omitted as it is incorrect: the gain reduction might still increase when the input signal drops in level. Think of compressor with a ratio of 1000:1 and a burst noise that overshoot the threshold by 10 dB and then drops to 5 dB above the threshold; but due to slow attack, the gain reduction was only at 3 dB when the signal dropped by 5 dB - the gain reduction will still rise to 4.99 dB. Again, the link between the attack and release and the input signal is dangerous - compressors, by design, have time stage following the scale stage (ratio) and the time stage will only be linked to the input signal if auto-attack or auto-release is in force.
  • If all is good and sound, we can apply the corrected explanation; accordingly, I believe that in the illustration 'time' will have to change to 'phase'. Also, I think it could be improved if you draw values on the level axis and use a ratio of, say, 2:1; If you are in a lazy mood, I can illustrate it myself.Izhaki (talk) 11:43, 1 August 2008 (UTC)

Audio Compression Attack and Release-2.svg

  • Here is a new version of the image. I think 'time' is an appropriate term for the x-axis, as the graph shows the change of level over time. I also changed the length of the attack and release so they do not match, hoping to convey that they are indenpendant things. I have not replaced the image on the article page.
  • Agreed "This happens when the input level decreases." should be removed. I was trying to simplify for readers who have no experience with compressors, but failed. Iain (talk) 05:49, 3 August 2008 (UTC)
Great job! Izhaki (talk) 13:08, 3 August 2008 (UTC)

Hard/Soft Knee Image[edit]

The Hard and soft knee picture is misleading too please see [2] for a good image

the image in this article shows that the compressor anticipates the threshold which is not the case. The image on [3] seems more plausible. —Preceding [Wikipedia:Signatures|unsigned] comment added by Timburly (talkcontribs) 14:26, 4 September 2008 (UTC)


The image is correct according to Rane [4]. They say:

Soft knee begins applying a small amount of compression just before the threshold point is reached, continues increasing compression through the threshold point and beyond, finally applying full compression to the highest level signals.

Iain (talk) 19:14, 4 September 2008 (UTC)


I looked at the Rane graph. I also looked at the graph in my (analog) dbx Model 163x OverEasy Compressor/Limiter owners manual , .pdf here [5]. Both graphs show the soft knee surrounding the threshold, not starting at the threshold. The illustration at [6] might be an artistic mistake. My dbx 163x manual says "With dbx's OverEasy characteristic, every possible ratio is available, from 1:1 to infinity:1, dependent on the signal level." That means the sbx 163x's output curve really only meets the idealized straight line portions at zero and infinity, like a logarithmic curve, so the compression ratio changes smoothly throughout the entire input range, not just near the knee. That means the sbx 163x has a different curve than the Rane. The Rane has a "span" (their term) of 10db surrounding the threshold that contains the soft knee portion, beyond which it's straight lines. I think you could say that the dbx 163x has an infinite-width span, with no straight lines. The illustration [7] currently used in this article is therefore closer to the Rane's curve than to the dbx 163x's, but it does show the curve centered around the knee which is the point. Is the Rane digital? I think if one wanted to write one's own digital signal processing programming code, any arbitrary curve could be implemented, even zig zags, using a look-up table or something, in which case, terms like knee and threshold become blurry. -- Another Stickler (talk) 19:15, 28 October 2008 (UTC)


High Threshold versus Low Threshold[edit]

Just wondered about this bit -- "Engineers wishing to achieve dynamic range reduction with few obvious effects might choose a relatively high threshold and low compression ratio so that the source material is being compressed very slightly most of the time."

If the threshold is high then wouldn't it be the case that the compression is not acting most of the time, but only some of the time? (I know it says "relatively high" rather than "high", but I don't see that this changes things much).

(Scott, 27-10-08). —Preceding unsigned comment added by 90.217.112.218 (talk) 09:16, 27 October 2008 (UTC)

Your point makes sense to me. Why not go ahead and correct the sentence to say relatively low threshold? Be sure to click "edit" next to the section heading rather than "edit this page". -- Another Stickler (talk) 19:36, 28 October 2008 (UTC)

I agree that this part was wrong. I have corrected it. Potatoj316 (talk) 19:23, 20 November 2010 (UTC)

That looks better now but what we really need is a citation. The whole paragraph looks like WP:OR to me. --Kvng (talk) 15:45, 21 November 2010 (UTC)

Black background when print[edit]

I tried printing the first five pages using Mozilla/SeaMonkey and the graphics are almost entirely solid black. How do I get around that? —Preceding unsigned comment added by Solarblast1 (talkcontribs) 19:09, 11 December 2008 (UTC)

compressor in foobar?[edit]

how do you know foobar has builtin compressing? it should be made clearer —Preceding unsigned comment added by 86.103.209.229 (talk) 18:54, 12 May 2009 (UTC)

citation found[edit]

Hey Folks, I don't seem to manage it on my own, but I would like to add the citation to the "upward compression" part. The book I am talking about is Bob Katz' "Mastering Audio - the art and the science", Focal Press 2002. There is an interesting part about dynamic range manipulation. So whoever is routined in w. go ahead! —Preceding unsigned comment added by Trikko (talkcontribs) 22:40, 29 September 2009 (UTC)

Broadcasting has no references[edit]

The "Broadcasting" section of "Common Uses" has no references and frankly all sounds like first-person research. Can someone cite references? Otherwise it shouldn't be on here.128.183.243.137 (talk) 13:51, 2 August 2010 (UTC)

I added a banner to the section. --Kvng (talk) 14:40, 2 August 2010 (UTC)

I have added references to the Broadcasting section + extended it a bit (as loudness differences *within* channels are relevant too). I have kept the banner with the main part of the original content. --Wearealmosthere (talk) 10:11, 5 October 2010 (UTC)

Use[edit]

The Use section needs some work. There are other important reasons for using compression. I've removed the incomplete summary of uses from the lead until I have time to address this. --Kvng (talk) 12:36, 27 April 2011 (UTC)

Attack and Release[edit]

"The 'release phase' is the period when the compressor is increasing gain to the level determined by the ratio, or, to zero dB, once the level has fallen below the threshold"

This language confuses me. Why would the compressor be "increasing gain" during the release phase? Isn't that the attack phase? — Preceding unsigned comment added by 67.183.109.15 (talk) 15:01, 20 August 2011 (UTC)

A compressor reduces gain during the attack to make louder sections proportionally not quite so loud. It therefore increases or restores gain during release. --Kvng (talk) 19:53, 22 August 2011 (UTC)

History of the Fairchild 660 and 670 Compressor/Limiters[edit]

One of the earliest and arguably most classic compressor/limiters is the Fairchild 660 and 670. The design was created by Rein Narma as he was building Les Paul's first 8-channel mixing console. The design was then licensed by Sherman Fairchild who hired Rein Narma as the company's chief engineer:

[Rein Narma Fairchild 670 Compressor Limiter]

It includes a link to a podcast with Les Paul about its creation.

Robert.Harker (talk) 00:04, 15 November 2012 (UTC)

These are probably notable enough to deserve their own article. -—Kvng 14:37, 16 November 2012 (UTC)
Agreed. An article about the Fairchild family of compressors would be good. Beatles engineer Geoff Emerick loved the unit for its fast attack, soft knee and sonic signature. He always used it on vocals.
The 660/670 has been cloned digitally in plugins.[8][9][10][11] Bono's vocals were processed with a Fairchild plugin for the U2 album No Line On The Horizon. Binksternet (talk) 17:05, 16 November 2012 (UTC)

I created an article for Fairchild Recording Equipment Corporation just so the story of where the design came from would not get lost. Not up writing a article about the compressors themselves. Robert.Harker (talk) 19:20, 16 November 2012 (UTC)

Types and Design[edit]

First, I think a naming convention for the headers would be nice. Singular or plural, not arbitrary. It should be "Type", "Design", "Use", etc. or "Types", "Designs", "Uses", etc.

Type: There's also upward and downward expansion, if this article wishes to be thorough.

Design: Optical is commonly referred to as "opto". Not mentioned is this article is OCA (Operational Transconductance Amplifier), Valve, Tube, Vari-Mu, and few other designs. It might be helpful to mention famous compressors for the more popular designs. The dbx 160 and Distressor are VCA compressors. The LA-2A is an opto compressor. The UA 1176 is an FET compressor. — Preceding unsigned comment added by 64.223.125.197 (talkcontribs) 20:09, 31 August 2013 (UTC)‎

Expansion should really go in the Noise gate article which in turn probably needs a title change. I don't personally think that unifying the headings is all that important as they currently appear to specify plurality when it makes sense to. For example, Design and Use are both nouns that could refer to actions rather than classifications in context. If it really bothers you, I'd suggest to WP:BEBOLD and change it yourself. Radiodef (talk) 18:16, 1 September 2013 (UTC)