Compressor release setting - please explain to a thickie...

[mikeyboy75]

Okay before anyone jumps on my neck - I have searched for info already on this forum and the interweb in general, as well as having a number of books/mags etc that apparently explain how a compressor work, but one thing I just can't comprehend is how on earth the Release amount is supposed to work.

Let me give you an example of my confusion, using the example of compressing a long slowly evolving then slowly dropping off pad/string sound.

So the sound starts, slowly increases in amplitude until it reaches the threshold, and then after the Attack time the compressor starts to attenuate the signal. Let's say the ratio is 2:1, so for every db over the threshold the amplitude is only increased by half a decibel.

Okay, all fine so far. Now then.......

Let's say the volume has risen 4 db over the threshold, so the compressor is outputting a signal 2db over the threshold. Now if the input signal drops down 2 db........ what happens??


I used to assume that the signal would drop by 1 db (to follow the ratio), in which case when you go drop back past the threshold amount the input would be back to the same volume as the post-compressor level and the Release setting is irrelevant.

So something else must be happening - I'm guessing that if signals OVER the threshold drop back down (but not under the threshold) then the same compression still remains. Is this right??

I know at the end of the day, all that matters is 'does it sound right', but I'd really like to understand this as it's been puzzling me for years. Honestly!!

I hope my explanation made some sense, and that someone can be arsed to reply!! Ta..

[Echoi]

This might help you somewhat

http://www.dubstepforum.com/viewtopic.p ... ompression

[mikeyboy75]

Cheers dude, but to be honest it still doesn't explain to me what happens in the exact situation I described (as no description of compression that I've found does, I feel like I'm missing something...)

That is, when the level starts to drop but is still above the threshold....

I'm thinking particularly for sounds without sharp transients and drop-offs, like pads, which may lower in volume slowly back down to the threshold...

[Sharmaji]

really it depends on the compressor-- feedback/feedfwd/ opto/VCA/ all behave differently. plenty of compressor's don't even have attack/release settings.

but in general:

compressor listens for a signal to pass across the defined threshold/ acts on the signal after the attack setting/ stops acting based on the release setting. the ratio is how hard it acts.

[Echoi]

Cheers dude, but to be honest it still doesn't explain to me what happens in the exact situation I described (as no description of compression that I've found does, I feel like I'm missing something...)

That is, when the level starts to drop but is still above the threshold....

I'm thinking particularly for sounds without sharp transients and drop-offs, like pads, which may lower in volume slowly back down to the threshold...

Ah sorry, i only really read the thread title

[mikeyboy75]

Thanks chap

...

compressor listens for a signal to pass across the defined threshold/ acts on the signal after the attack setting/ stops acting based on the release setting. the ratio is how hard it acts.

I get all that, but that still doesn't explain what happens after the attack time has passed but before the volume has dropped below the threshold - imagine the volume dropping very slowly (several seconds) down to the threshold, and maybe even rising up again even higher before that has happened....?

[Sharmaji]

Thanks chap

...

compressor listens for a signal to pass across the defined threshold/ acts on the signal after the attack setting/ stops acting based on the release setting. the ratio is how hard it acts.

I get all that, but that still doesn't explain what happens after the attack time has passed but before the volume has dropped below the threshold - imagine the volume dropping very slowly (several seconds) down to the threshold, and maybe even rising up again even higher before that has happened....?

in practice, if the release is so long that the signal never gets back up to the threshold level, you'll hear that the compressor progressively pushes the entire signal down, rather than just the peaks (or whatever) you're listening to-- it'll be extremely obvious. try putting a high-ratio comp on seomthing like a 4x4 kick pattern, and set it to have a fast attack so that it grabs the front of the sound, and a REALLY long release-- you'll hear that each hit progressively gets quieter.

so in that case... turn down the release.

[krispy]

This might also be helpful, don't know how many times i've read this

http://www.tarekith.com/assets/dynamics.html

[narcissus]

the release setting... on compressors i've used.. adjusts how quickly the signal drops back to the original level, after being compressed.. like how you described, OP, the signal being 4 db over the threshold then falling to 2, yes, you're output would at first be +2b, then would sink to +1 db, and the release setting is how quickly it falls... set to 500 ms, it would take half a second to fall from 2db to 1 db, even if the change in the original signal was abrupt. depending on the source material, you will want the release of the compressor to be more abrupt or gradual. on a pad you'd most likely use a longer release setting.. but there are no rules really.

[macc]

It makes more sense if you think of the attack and release as 'time it takes for the compressor to respond to changes in signal level' than 'time it takes for the compressor to realise the threshold has been crossed in whatever direction'. Cos largely, that's how it is. Think of a soft knee compressor, acting way below the threshold - it still responds according to those times even though the threshold has nowhere near been reached.

In your example, taking a hard-knee comp for simpicity, as long as the release time is sufficiently longer than the period of the lowest frequency present (so signal modulation rather than gain modulation isn't an issue), and not way too long, then when you have your 'slowly-dropping-towards-the-threshold' signal the compressor's gain reduction will simply track the signal. It will be doing progressively less compression as it drops, before the signal drops below the threshold and it properly releases. If the signal slowly/gently rises again, the compressor will respond according to a its particular design and also its topology - feedback or feedforward (whether it is listening to the input signal or the output signal to derive its gain readings). Generally though, assuming the time constants are set rationally, it will simply track the signal, doing progressively more/less compression depending on how the signal varies. If the release time is way too long it will 'lag', if it is way too short it will incorrectly read the oscillation of the signal itself as drops/rises in the level and give the signal modulation distortion I mentioned above. So even when ostensibly above the threshold, the release has a say in matters. I'm assuming we're talking about a sort of pad-type sound here btw, nice and smoothly up and down.

If you're talking about a signal that chops up and down violently while above the threshold, towards the limits of a compressor's tracking ability (ie, as fast as the shortest release time available), then it starts to depends even more on what time constants are set and the particular design of the compressor. Compressors with programme-dependency have the means to know what's going on, either by using a second-order filter in the release circuit (so it depends on rate of change rather than absolute signal level) or other means. So some compressors might track the signal more accurately than others, which stick to their assigned attack/release times. Generally programme dependency circuits are a bit slower acting, but some plugins (I think!) can be adaptive at faster times. I know all the bands on the Maselec MLA-3 are programme dependent, and some of those are pretty nippy to say the least. 'Programme dependent circuits reduce the distortion * and speed up the release for short duration
transients', it says here

SO.... I have gone way off the point and am about to write another 500 words of bollix The first bit said it best

* (Presumably lengthening the actual release time when required)

[mikeyboy75]

Thanks chap
in practice, if the release is so long that the signal never gets back up to the threshold level, you'll hear that the compressor progressively pushes the entire signal down, rather than just the peaks (or whatever) you're listening to-- it'll be extremely obvious. try putting a high-ratio comp on seomthing like a 4x4 kick pattern, and set it to have a fast attack so that it grabs the front of the sound, and a REALLY long release-- you'll hear that each hit progressively gets quieter.

so in that case... turn down the release.

Yep I get you, but with a kick pattern, after each kick the signal will go below the threshold so the release time will kick in - that makes sense to me, I'm really interested in sounds which slowly return to the threshold.

The reason I ask is that I do a lot of ambient bed/pad backings which I'm trying to tame, it's not really for percussive sounds.

[macc]

The reason I ask is that I do a lot of ambient bed/pad backings which I'm trying to tame, it's not really for percussive sounds.

Use an adaptive (programme dependent) compressor, ideally soft knee, and set the release time for the middle of next week. Sorted.

[mikeyboy75]

Ok cheers all for the responses guys - really appreciate it. I always thought the release setting was just tied to the threshold but apparently it's not that simple - life would be boring if things _were_ that easy

The reason I ask is that I do a lot of ambient bed/pad backings which I'm trying to tame, it's not really for percussive sounds.

Use an adaptive (programme dependent) compressor, ideally soft knee, and set the release time for the middle of next week. Sorted.

So have I finally got to accept that Reason's compressors just aren't up to the task?? Damn..

[Isturite]

One thing that can help you hear what the release is doing... excuse me if somebody already mentioned this, I read through most of the posts but shamarji said something about setting a really high ratio to hear the release... this is somewhat how i learned compressors...

I learned mostly with an 808 kick and a snare I made in Absynth... I set the threshold way down, like -40/50dB.. then the ratio as far up as it will go, on the ableton compressor it's -inf... so it compresses it all the way down to the threshold.... once you have this the volume will be quieter so you might have to turn up the output (turn it back down before you turn the threshold up or you are in for some loudness)... anyway, when the setting are at max like this play around with the attack and release settings... you will REALLY be able to hear what it's doing, if I put a short release on an 808 it turns it into a flat volume subby boom sound, if I turn the release up it starts to sound like a normal 808 that decreases in volume... this is because with a fast release the compressor is recompressing the signal every amount of milliseconds that is set on the release, when the release is longer the compressor doesn't compress the tail of the sound because the release time was long enough for the compressor to release once the sound is low enough not to go over the threshold...

the opto/vca/etc... setting i'm still figuring out... but the opto setting is useful when using a really short release to prevent the clicking sounds produced when the compressor is resetting too fast... I use opto almost whenever i drop the release under 120ms

[Sharmaji]

The reason I ask is that I do a lot of ambient bed/pad backings which I'm trying to tame, it's not really for percussive sounds.

Use an adaptive (programme dependent) compressor, ideally soft knee, and set the release time for the middle of next week. Sorted.

and them limit

[Sharmaji]

So have I finally got to accept that Reason's compressors just aren't up to the task?? Damn..

reason's M-Class compressor is actually quite nice.

that little red box one... isn't.

[mikeyboy75]

So have I finally got to accept that Reason's compressors just aren't up to the task?? Damn..

reason's M-Class compressor is actually quite nice.

that little red box one... isn't.

I always use the M-Class one, yep, and have also always told myself that if I can't blame Reason's tools for the quality of my productions...

It's not an adaptive compressor though is it? Boo

[mikeyboy75]

The reason I ask is that I do a lot of ambient bed/pad backings which I'm trying to tame, it's not really for percussive sounds.

Use an adaptive (programme dependent) compressor, ideally soft knee, and set the release time for the middle of next week. Sorted.

and them limit

Was that serious??! I can't tell

[mikeyboy75]

One thing that can help you hear what the release is doing... excuse me if somebody already mentioned this, I read through most of the posts but shamarji said something about setting a really high ratio to hear the release... this is somewhat how i learned compressors...

I learned mostly with an 808 kick and a snare I made in Absynth... I set the threshold way down, like -40/50dB.. then the ratio as far up as it will go, on the ableton compressor it's -inf... so it compresses it all the way down to the threshold.... once you have this the volume will be quieter so you might have to turn up the output (turn it back down before you turn the threshold up or you are in for some loudness)... anyway, when the setting are at max like this play around with the attack and release settings... you will REALLY be able to hear what it's doing, if I put a short release on an 808 it turns it into a flat volume subby boom sound, if I turn the release up it starts to sound like a normal 808 that decreases in volume... this is because with a fast release the compressor is recompressing the signal every amount of milliseconds that is set on the release, when the release is longer the compressor doesn't compress the tail of the sound because the release time was long enough for the compressor to release once the sound is low enough not to go over the threshold...

the opto/vca/etc... setting i'm still figuring out... but the opto setting is useful when using a really short release to prevent the clicking sounds produced when the compressor is resetting too fast... I use opto almost whenever i drop the release under 120ms

Cheers boyo. But like I said, it wasn't really for percussive sounds where I was confused - as in these cases the amplitude drops back below the threshold, the release time kicks in, and all is simple and it is indeed fun (and essential) to do what you suggested to really see what the controls are doing. With a much longer slowly building and fading sample (over several seconds) then I never knew what was going on in that no-man's-land from when the gain started dropping but was still above the threshold. Seems that the release isn't entirely based on the going below the threshold, and it's also down to the individual compressor itself..

[lowpass]

It makes more sense if you think of the attack and release as 'time it takes for the compressor to respond to changes in signal level' than 'time it takes for the compressor to realise the threshold has been crossed in whatever direction'. Cos largely, that's how it is. Think of a soft knee compressor, acting way below the threshold - it still responds according to those times even though the threshold has nowhere near been reached.

In your example, taking a hard-knee comp for simpicity, as long as the release time is sufficiently longer than the period of the lowest frequency present (so signal modulation rather than gain modulation isn't an issue), and not way too long, then when you have your 'slowly-dropping-towards-the-threshold' signal the compressor's gain reduction will simply track the signal. It will be doing progressively less compression as it drops, before the signal drops below the threshold and it properly releases. If the signal slowly/gently rises again, the compressor will respond according to a its particular design and also its topology - feedback or feedforward (whether it is listening to the input signal or the output signal to derive its gain readings). Generally though, assuming the time constants are set rationally, it will simply track the signal, doing progressively more/less compression depending on how the signal varies. If the release time is way too long it will 'lag', if it is way too short it will incorrectly read the oscillation of the signal itself as drops/rises in the level and give the signal modulation distortion I mentioned above. So even when ostensibly above the threshold, the release has a say in matters. I'm assuming we're talking about a sort of pad-type sound here btw, nice and smoothly up and down.

If you're talking about a signal that chops up and down violently while above the threshold, towards the limits of a compressor's tracking ability (ie, as fast as the shortest release time available), then it starts to depends even more on what time constants are set and the particular design of the compressor. Compressors with programme-dependency have the means to know what's going on, either by using a second-order filter in the release circuit (so it depends on rate of change rather than absolute signal level) or other means. So some compressors might track the signal more accurately than others, which stick to their assigned attack/release times. Generally programme dependency circuits are a bit slower acting, but some plugins (I think!) can be adaptive at faster times. I know all the bands on the Maselec MLA-3 are programme dependent, and some of those are pretty nippy to say the least. 'Programme dependent circuits reduce the distortion * and speed up the release for short duration
transients', it says here

SO.... I have gone way off the point and am about to write another 500 words of bollix The first bit said it best

* (Presumably lengthening the actual release time when required)

When's the rest of your book coming out?

Interesting to hear about programme-dependancy, not come across this before. You mention it would use a second order filter what would that be like a highpass? so that you don't get a huge buildup of LF content come in and mess up the carefully tweaked release time? Not too sure what it would be filtering here.