Compressor's attack 0ms vs 0.5ms (different behavior)

I've experienced this in the past, where the 0ms attack added distortion to the sound, while 0.5ms or 1ms did not. And when I say "distortion" I don't mean "saturation" or any pleasant distortion. It's just like annoying, crackling distortion.

Today I noticed something different, so I would like to know if any of you guys know why this happens, if it's normal or a "bug".

 

I created 2 audio files with the same length and same amplitude, one at around 80Hz and one at around 4kHz (I guess).

Added the compressor and noticed that the Vintage VCA and the Vintage OPTO would compress the 4kHz A LOT when set to 0ms, but at 0.5ms it would behave the same way as the 80Hz.

This only happens with those 2 models.

 

Is this normal? If so, why?

 

Here's the project:

Test Compressor's Attack at 0ms.logicx.zip

It could be just distortion that was present on the original analog model, that the Logic developers tried to emulate. Or not. In any case even emulations don't always sound like the analog model they were modeled after so I think of these models almost as if they're new models in their own right, with their own identity, their own idiosyncrasies, and try not to make too much "sense" of how they sound. Rather I dial them to try and find different sound, switch models, adjust, until I get the sound I'm looking for, or something interesting at least.

It could be just distortion that was present on the original analog model, that the Logic developers tried to emulate. Or not. In any case even emulations don't always sound like the analog model they were modeled after so I think of these models almost as if they're new models in their own right, with their own identity, their own idiosyncrasies, and try not to make too much "sense" of how they sound. Rather I dial them to try and find different sound, switch models, adjust, until I get the sound I'm looking for, or something interesting at least.

 

Yes, I do the same, especially because I’ve never used the hardware versions anyway, so I wouldn’t be able to compare. I’ve even seen people online saying that even the best emulations, with the exact same settings don’t sound the same as the hardware. Not worse, just different.

 

To me these 2 behaviors I mentioned are just weird and I don’t see how they would be implemented on purpose because one is unpleasant (and unusable in my opinion) which is that initial distortion almost like when you have an audio file and cut it in the middle of a cycle, you know?

the other is also weird because at 0ms of attack the gain reduction is 30dB, for example, while 0.5ms is 5dB. Huge difference

And it’s just on those 2 models and only on the high frequency version, not the 80hz.

 

Anyway, I was just wondering if anyone could confirm this is an actual thing or a bug. I always try to stick to 1ms of attack whenever I remember. Just in case

Yes I've also noticed some unusable distortion with some settings on certain compressor models. However I've also noticed the same kind of distortion on analog compressors.

I seldom use fast attack times (note, VCA compressors are generally "fast" and Opto one are generally "slow", which is intrinsic in their design, the technologies used, and other historical factors), as fast attack times (intentionally) kill your transients. Useful for when you *want* to do that, perhaps, but it kills a lot of the character of the sound in general. Probably the only time I'll really hit faster attacks is on some aggressive parallel compression, but that's about it.

Yes I've also noticed some unusable distortion with some settings on certain compressor models. However I've also noticed the same kind of distortion on analog compressors.

 

I see. The only times I used analog gear was a loooong time ago, when I still didn’t even know anything about mixing or anything. It was just moving knobs and see what sounded good so I don’t have that analog reference.

I seldom use fast attack times (note, VCA compressors are generally "fast" and Opto one are generally "slow", which is intrinsic in their design, the technologies used, and other historical factors), as fast attack times (intentionally) kill your transients. Useful for when you *want* to do that, perhaps, but it kills a lot of the character of the sound in general. Probably the only time I'll really hit faster attacks is on some aggressive parallel compression, but that's about it.

 

I use fast attack times a lot to control peaks or to reduce the transients of certain samples (pretty much what the Enveloper does).

But also, it only kills a lot of the transients depending on the ratio and threshold so it’s not just the attack’s “fault”.

But in my example, the difference between 0ms and 0.5ms, which is ridiculous small, makes a huge difference of maybe 20dB. It’s not gradual.

Also, it’s only doing that to a high frequency audio snippet, not the 80Hz, even though they are at the exact same level.

Maybe I’m wrong here, but it seems like a weird and unusual behavior, I guess.

I use fast attack times a lot to control peaks or to reduce the transients of certain samples (pretty much what the Enveloper does).

But also, it only kills a lot of the transients depending on the ratio and threshold so it’s not just the attack’s “fault”.

 

Well sure, that's how a compressor works. A zero attack time won't do anything if you haven't the set threshold and/or ratio to do any reduction!

 

But in my example, the difference between 0ms and 0.5ms, which is ridiculous small, makes a huge difference of maybe 20dB. It’s not gradual.

Also, it’s only doing that to a high frequency audio snippet, not the 80Hz, even though they are at the exact same level.

Maybe I’m wrong here, but it seems like a weird and unusual behavior, I guess.

 

I haven't looked at your files, so I can't really comment on that. Ultimately, it depends on the incoming audio, and the behaviour characteristics of the compressor. If it's too fast, back it off a bit. If you genuinely think there is some bug or issue at a certain setting, send in a feedback note.

I use fast attack times a lot to control peaks or to reduce the transients of certain samples (pretty much what the Enveloper does).

But also, it only kills a lot of the transients depending on the ratio and threshold so it’s not just the attack’s “fault”.

 

Well sure, that's how a compressor works. A zero attack time won't do anything if you haven't the set threshold and/or ratio to do any reduction!

 

But in my example, the difference between 0ms and 0.5ms, which is ridiculous small, makes a huge difference of maybe 20dB. It’s not gradual.

Also, it’s only doing that to a high frequency audio snippet, not the 80Hz, even though they are at the exact same level.

Maybe I’m wrong here, but it seems like a weird and unusual behavior, I guess.

 

I haven't looked at your files, so I can't really comment on that. Ultimately, it depends on the incoming audio, and the behaviour characteristics of the compressor. If it's too fast, back it off a bit. If you genuinely think there is some bug or issue at a certain setting, send in a feedback note.

 

 

I mentioned that, just because you said it kills the transients. And you said it kills a lot of the character. Well, that isn’t actually true, because of what I said. If you don’t go hard on the threshold and/or ratio, you can even get to a point or you don’t actually hear the difference. I think the word “kill” when something is actually just slightly attenuated, is probably a bit too much in a general context If I attenuate a certain frequency on an EQ, I’m not killing it or not killing the overall character of the sound.

 

When / if you have the time, open my files and see what I mean.

If two audio files have the exact same length and the exact same amplitude, shouldn’t the compressor behave the same way? Especially because it does behave the same way above 0.5ms and only on 2 compressors does it behave in that weird way.

 

But it’s easier if you open the files and see it for yourself

I mentioned that, just because you said it kills the transients. And you said it kills a lot of the character. Well, that isn’t actually true, because of what I said.

 

it is true, it's a generally known characteristic of super fast attacks. Of course you can be more suble about it, and sometimes toning down that attack transient is intended, as I mentioned, say to soften up a ticky hihat or something. So all tools have their uses - I wasn't saying otherwise. Just that I don't tend to use them that much, because I don't generally like how it sounds. ymmv of course.

 

If you don’t go hard on the threshold and/or ratio, you can even get to a point or you don’t actually hear the difference. I think the word “kill” when something is actually just slightly attenuated, is probably a bit too much in a general context If I attenuate a certain frequency on an EQ, I’m not killing it or not killing the overall character of the sound.

 

Of course, I understand that.

 

When / if you have the time, open my files and see what I mean.

If two audio files have the exact same length and the exact same amplitude, shouldn’t the compressor behave the same way? Especially because it does behave the same way above 0.5ms and only on 2 compressors does it behave in that weird way.

 

But it’s easier if you open the files and see it for yourself

 

Yeah, I don't have the time to do an investigation right now...

But in my example, the difference between 0ms and 0.5ms, which is ridiculous small, makes a huge difference of maybe 20dB. It’s not gradual.

Also, it’s only doing that to a high frequency audio snippet, not the 80Hz, even though they are at the exact same level.

Maybe I’m wrong here, but it seems like a weird and unusual behavior, I guess.

 

Just some naive math: let's assume that the threshold of the compressors lies somewhere inbetween zero and the maximum amplitude of the sine waves.

 

The 80Hz sine has a period of 12.5ms, the time from rising from zero to the first positive maximum amplitude is 3.1ms. No matter if your attack time is 0ms or 0.5ms, the 80Hz sine is quickly dampened before reaching the maximum.

 

For the 4kHz sine the period is only 0.25ms. Thus an attack of 0ms might quickly dampen this sine before reaching the maximum the first time; an attack of 0.5ms means, that two full peroids of this sine have already passed i.e. that both the positive and negative maximum amplitude has each been reached twice before the level is reduced.

 

Possibly this model is oversimplified but could describe the main effect though.

But in my example, the difference between 0ms and 0.5ms, which is ridiculous small, makes a huge difference of maybe 20dB. It’s not gradual.

Also, it’s only doing that to a high frequency audio snippet, not the 80Hz, even though they are at the exact same level.

Maybe I’m wrong here, but it seems like a weird and unusual behavior, I guess.

 

Just some naive math: let's assume that the threshold of the compressors lies somewhere inbetween zero and the maximum amplitude of the sine waves.

 

The 80Hz sine has a period of 12.5ms, the time from rising from zero to the first positive maximum amplitude is 3.1ms. No matter if your attack time is 0ms or 0.5ms, the 80Hz sine is quickly dampened before reaching the maximum.

 

For the 4kHz sine the period is only 0.25ms. Thus an attack of 0ms might quickly dampen this sine before reaching the maximum the first time; an attack of 0.5ms means, that two full peroids of this sine have already passed i.e. that both the positive and negative maximum amplitude has each been reached twice before the level is reduced.

 

Possibly this model is oversimplified but could describe the main effect though.

 

The math and the logic (no pun intended) behind your explanation makes sense... BUT!

If you download the project you will see that both waves are peaking at -3dB. The ratio is set to 5:1 and the threshold to -10dB. How does the high pitched wave creates a 20dB gain reduction, if it only goes 7dB above the threshold? It doesn't make sense... does it? Am I missing something?

Danny, I've extended your project with exported audio files in order to see the effect of the compressors in the wave forms.

Overall observations: the target level (after the settlement phase) is about the expected level (-8dBFS).

The VCA behaves for 80Hz as expected. With 4kHz and 0ms attack, the level starts a bit too low and ramps up over 20ms; with 0.5ms it operates as expected but the attack time results in a exponential drop over around 2ms (mathematically the 0.5ms attack time are probably the time contant for an exponential descrease to the target level).

The OPTO seems not to support 0ms as one would expected, it transmits the first 2ms without dampening. In addition for the 4kHz it dampens then 3dB too much and ramps up to the target level over 35ms.

It seems that 0ms attack time got some special behaviour.

Test Compressor's Attack at 0ms with exported regions.logicx.zip