Zero attack compressor - What did I do wrong?

[Simone Lombardo]

First of all, there's an older post that's very similar to this, but it's from more than 3 years ago, so I preferred creating a new post.

Yesterday I was watching the same video mentioned in that old post. The title is a bit misleading, since the video actually explains how to create look ahead and not exactly zero attack, but the goal is the same: force the compressor to act instantly when the signal goes above the threshold.

To recreate that technique in Reason I had to split the parallel input of the (muted) audio track with a Spider Audio and send the first output into a mix channel, then second one into the sidechain input of the compressor, which is used as an insert in that channel and is set to the minimum attack (1ms). Before the compressor there are two DDL-1s (one for each channel, otherwise for some reason the sound changes) both set to 1ms in order to match the compressor's attack.

Here's the front view, rear view and sequencer:



Despite doing basically the same thing as in the video, the volume still goes over the threshold at the very beginning. And raising the delay time to a higher value (for example 2ms) seems to be useless. I don't really know what I'm doing wrong. What do you think?

Edit: forgot to say that the threshold was 0db and the ratio was infinite:1 when I tested it (basically a brick wall limiter), but the signal was still exceeding 0db at the beginning.

[orthodox]

Despite doing basically the same thing as in the video, the volume still goes over the threshold at the very beginning. And raising the delay time to a higher value (for example 2ms) seems to be useless. I don't really know what I'm doing wrong. What do you think?

First make sure that the (MClass) compressor is set to Ratio=Infinite, SoftKnee=Off. Otherwise, it's not limiting the output.
Then, the attack time does not imply that the signal envelope reaches the peak level in the specified amount of time. It only goes part of the distance to that level.

May everybody forgive me, but let me suggest this compressor, which internally implements both the delay-compensated look-ahead and the true zero attack modes:

https://www.reasonstudios.com/shop/rack ... -dynamics/

[Loque]

You just need to delay the main signal and not the side chain. Set the compressor to a fast attack too.

[orthodox]

You just need to delay the main signal and not the side chain. Set the compressor to a fast attack too.

Judging by the screenshot, he does delay the main signal. And he wrote that Attack=1ms, the same as the delay.

[Loque]

You just need to delay the main signal and not the side chain. Set the compressor to a fast attack too.

Judging by the screenshot, he does delay the main signal. And he wrote that Attack=1ms, the same as the delay.

You are right. The Threshold is pretty high (or low, depending on the viewpoint ), probably the Compressor does not react at all.

[Simone Lombardo]

The Threshold is pretty high (or low, depending on the viewpoint ), probably the Compressor does not react at all.

I raised the input gain of the compressor in order to bring the incoming signal over 0db, which is where the threshold is set. I thought the signal would have stopped there immediately, but at the very beginning the two red lights at the end of the Big Meter were still being turned on.

[orthodox]

The Threshold is pretty high (or low, depending on the viewpoint ), probably the Compressor does not react at all.

I raised the input gain of the compressor in order to bring the incoming signal over 0db, which is where the threshold is set. I thought the signal would have stopped there immediately, but at the very beginning the two red lights at the end of the Big Meter were still being turned on.

As I said, regardless of all the other settings like Ratio, the compressor in the Attack phase does not really reach the peak level of the signal, only partially. You can't expect it to truly limit the signal at the threshold.
The look-ahead method is used to suppress the pumping effect resulting from the attack envelope delay. Usually, you set the look-ahead delay and then start adjusting the attack time to get the sound you like the most. Then you adjust the output level, so that it won't clip.

[Loque]

The Threshold is pretty high (or low, depending on the viewpoint ), probably the Compressor does not react at all.

I raised the input gain of the compressor in order to bring the incoming signal over 0db, which is where the threshold is set. I thought the signal would have stopped there immediately, but at the very beginning the two red lights at the end of the Big Meter were still being turned on.

You raised the gain of the input signal, not of the side-chain signal, which is now triggering the compressor. in fact, i expect the sound just to be much louder without any compression.

[orthodox]

You raised the gain of the input signal, not of the side-chain signal, which is now triggering the compressor. in fact, i expect the sound just to be much louder without any compression.

As I can recall, the Input Gain knob of the MClass Compressor gains both the Main and Sidechain inputs equally.

[Loque]

You raised the gain of the input signal, not of the side-chain signal, which is now triggering the compressor. in fact, i expect the sound just to be much louder without any compression.

As I can recall, the Input Gain knob of the MClass Compressor gains both the Main and Sidechain inputs equally.

Ok, i am not sure atm and cannot test. You need to test it for yourself. and it wont hurt you to lower the threshold until you see any gain reduction in the meter

[Simone Lombardo]

You raised the gain of the input signal, not of the side-chain signal, which is now triggering the compressor. in fact, i expect the sound just to be much louder without any compression.

Well... this was me when I realized that:

As I can recall, the Input Gain knob of the MClass Compressor gains both the Main and Sidechain inputs equally.

Loque might actually be right. I remember that even if the input gain was higher the sound was almost unchanged, and it was only compressed when the threshold was lower than 0db. In short words, I might have had a massive brain fart...

[orthodox]

As I can recall, the Input Gain knob of the MClass Compressor gains both the Main and Sidechain inputs equally.

Loque might actually be right. I remember that even if the input gain was higher the sound was almost unchanged, and it was only compressed when the threshold was lower than 0db. In short words, I might have had a massive brain fart...

I have just checked it, and it really does affect both inputs. So there is nothing to worry about, the input gain and the threshold knobs are interchangeable in your case where the signal is the same.

[Simone Lombardo]

I have just checked it, and it really does affect both inputs. So you can calm down, the input gain and the threshold knobs are interchangeable in your case where the signal is the same.

If you say so... But what version of Reason are you using? I use R10 (see signature), and maybe it can differ depending on the version.

BTW, I checked out the compressor you suggested. It seems really nice, but that's still 59 bucks, so I'd wait some time before deciding to buy it. Maybe I'll use the 30-day trial.

[orthodox]

If you say so... But what version of Reason are you using? I use R10 (see signature), and maybe it can differ depending on the version.

It is taboo to change the behavior of any existing device in Reason. The songs must sound the same. And I remember I did measurements of this compressor 5 years ago.

BTW, I checked out the compressor you suggested. It seems really nice, but that's still 59 bucks, so I'd wait some time before deciding to buy it. Maybe I'll use the 30-day trial.

I tried to implement in it most everything I know about compressors. By the way, It can also be set up to exactly emulate any one of the built-in Reason compressors.

[Simone Lombardo]

I tried to implement in it most everything I know about compressors.

Oh, didn't know you were the developer. This might be a bit off-topic, but what are the requirements to develop a Rack Extension? Besides an IDE that supports C++, do you need SDKs and stuff like that?

[orthodox]

This might be a bit off-topic, but what are the requirements to develop a Rack Extension? Besides an IDE that supports C++, do you need SDKs and stuff like that?

Yes, the process is described here: https://www.reasonstudios.com/developers (at the bottom of the page)
You create a business entity, register yourself as a RE developer at Reason Studios, and download the SDK from their site.
An IDE is not really needed, I don't use it.

[Simone Lombardo]

This might be a bit off-topic, but what are the requirements to develop a Rack Extension? Besides an IDE that supports C++, do you need SDKs and stuff like that?

Yes, the process is described here: https://www.reasonstudios.com/developers (at the bottom of the page)
You create a business entity, register yourself as a RE developer at Reason Studios, and download the SDK from their site.
An IDE is not really needed, I don't use it.

Awesome! Thanks for the help!

[Benedict]

From what I can ascertain (your images are awol), Selig already answered your problem in the older thread.

Potted & simplified version:

A Compressor has to "hear" & "judge" the event against its rules before it can start to swing into action. This takes time. The faster the judging circuit the faster the process can begin but time is a constant so it still ticks long as the process does its thing. It may take the compressor 0.002 ms to realize it needs to do something, 0.001 ms to start doing it, 0.002 ms to get to full processing power. That is 0.005 ms to get in-control of things. In that time, nature is holding sway.

A compressor can NEVER truly be 0ms Attack without doing some mighty nasty things as it warps the waveform at what is essentially Audio Rate (that 0.005 ms), therefore is waveshaping similar to AM or FM which is never as pretty as you hope. Nature is not interested in kind, it just does.

If you play with the perception of time using Lookahead logic then you can give the impression of a thing happening without lag by introducing lag and then delay compensation to make the lag appear to disappear. Good luck understanding that sentence fully.

The other method is to use the waveshaping method mentioned above and allow distortions in a way that is more pleasing than displeasing. This is what the Soft Clip section does in the M-Class Maximizer. That can "soak" up quite a lot of level with no particular output (but distortion - which you'll probably like if you don't push it too hard).

In practical terms, I have the Soft Clip on during Composition & Mixing to ensure nothing horrid hits my speakers (and I refuse to mix loud enough to be clipping so I am not hearing it) but turn it off and swap to Limiter with Lookahead as I get the last point of my Masters.

[orthodox]

A compressor can NEVER truly be 0ms Attack without doing some mighty nasty things as it warps the waveform at what is essentially Audio Rate (that 0.005 ms), therefore is waveshaping similar to AM or FM which is never as pretty as you hope. Nature is not interested in kind, it just does.

Can't agree with you. I do the literal zero attack compression without displeasing side effects. It does eat the front transients on kick drums, so I sometimes mix it with dry signal, but otherwise the result sounds pretty clean.

[Benedict]

A compressor can NEVER truly be 0ms Attack without doing some mighty nasty things as it warps the waveform at what is essentially Audio Rate (that 0.005 ms), therefore is waveshaping similar to AM or FM which is never as pretty as you hope. Nature is not interested in kind, it just does.

Can't agree with you. I do the literal zero attack compression without displeasing side effects. It does eat the front transients on kick drums, so I sometimes mix it with dry signal, but otherwise the result sounds pretty clean.

The physics doesn't change here. Suggesting otherwise is not exactly being helpful to those who hope to learn.

Your personal feelings about the results are just that: feelings. I won't attack a logical argument with my feelings without it being clear what is fact and what is perception. I will, of course, agree that to you it is immutable fact that you like the things that you like and dislike the things that you don't (probably me right now as I'm equally not delighted at being dismissed based on your feelings, even as you show that the facts are in fact true in your case too yet somehow I am still presented as wrong).

Did you read the later part about Soft Clipping and how we tend to like it? This covers what you are talking about but seems to have been ignored. You like or are prepared to live with the artifacts of the process but it doesn't mean that they are not there and technically quite "terrible" (assuming we are looking for a clean result - which is always the baseline assumption in processing).



p.s. if you had quoted only "is waveshaping similar to AM or FM which is never as pretty as you hope. Nature is not interested in kind, it just does." then I would have been less likely to be offended. Still, not taking into account all I wrote, therefore misquoting me, is not pleased-making but presenting me as foolish - not knowing what I showed I do know. Ask Selig to moderate directly (or pop over to that link in the OP I mentioned).

[Benedict]

I tried to implement in it most everything I know about compressors. By the way, It can also be set up to exactly emulate any one of the built-in Reason compressors.

I also didn't know you were a Dev and that device was yours. I haven't tried it but it does look interesting as a sort of do anything device. I like that one to rule them all kind of device rather than needing 78 one-trick ponies.

[orthodox]

I do the literal zero attack compression without displeasing side effects. It does eat the front transients on kick drums, so I sometimes mix it with dry signal, but otherwise the result sounds pretty clean.

I tend to agree with this guy (is it because our names look similar?). If he feels the result sounds clean enough, there might be others who would feel the same. Some might say that zero attack compression warps the waveform so badly that it can never be as pretty as we hope. But here we are, deeming it pretty.

Some may suggest that the artifacts introduced by this method are comparable to those of soft/hard clipping. Well, we can actually compare these two methods to find that they sound different. Unlike Maximizer, which has non-zero attack time, a zero attack compressor does not clip the waveform in the long run. It affects only peak fronts, and the distortion is unobtrusive.

Can't agree with you.

Who knows what this guy meant to say. Did he just want to tell us about some different experience he'd had? Or did he take one thesis out of context to portray himself right and others wrong? Is he an ignorant person? Or did he mean to show disrespect?
We will never know. So let's assume the worst, for that sort of attitude can never be tolerated.