short or long compression attack times?

[alligatorlizard]

Bob Katz's books sugests using long compression attack times for mastering - 100 to 300 ms. However a colleague recently expressed surprise at this, saying he's always understood you should use very short attack times for mastering. I've just trawled the net for several hours for more wisdom on this subject, and - I suppose unsurprisingly - there seems to be no general consensus on this. Some articles say very clearly to use short attacks, others say use long...

Now I know it's always going to be depend on the type of music etc, but can anyone here offer any general advice? And if both are valid, as is probably the case, would anyone be kind enough to offer examples of where a short compression times might be preferable to longer ones and vice-versa?

[Joe_caithness]

Sometimes long, sometimes short.

Fast attacks often kill sub bass, but often allow you to limit harder as it can tame the transients a bit which it brings up the overall level.

[Cellotron]

Generally with mastering you should use a long attack time when you want to retain more of the initial transients so that you can retain a sense of "punch" and "snap" while still "gluing" elements of the mix that might have been poking out over it back in. Where as you should use shorter attack times when you are specifically trying to tame transients that are overly sharp. So both have their uses. The majority of the time I use a compressor in the chain it is usually with a long attack time though.

Are you sure that BK was actually recommending 100 - 300ms for attack times though? This is extraordinarily long - I think he most likely would have been actually recommending 10 - 30ms.

It should be noted that compressors all react differently and the labeling on the faceplate might not be all that accurate as to what it is actually doing - so that what numerical value works on one comp might be labeled differently on another.

fwiw - the attack on my API2500 gets set the majority of the time at 30ms, my Focusrite Blue 330 at 70ms, and my Pendulum OCL-2 usually ends up somewhere around 20ms. Again - these settings can end up widely varying depending on the source material and desired end results though.

Best regards,
Steve Berson

[aleatoric]

Originally Posted by Cellotron

Are you sure that BK was actually recommending 100 - 300ms for attack times though? This is extraordinarily long - I think he most likely would have been actually recommending 10 - 30ms.

I remember reading him saying 50 - 300ms in his book. I think his whole thing is that most hardware analog compressors list a shorter attack time when by his measurement that attack time is actually much longer (e.g. attack says 30ms, by his measurement it is 90ms).

Ultimately though there is no standard for measuring and labeling compressor attack times. A 40ms attack time on one compressor might be listed as 20ms on another, or even by a more vague term like "medium fast". Let your ears guide you. Be careful with too quick attack times though, most times in mastering you are going to be using longer attack times in comparison to say a mix engineer compressing a drum bus.

[alligatorlizard]

Thanks, very useful replies - sounds like I don't need to use fast attack times for the tracks I'm currently mastering (mainly orchestral soundtrack stuff). Will listen out for those transients though, I guess big cymbal/bass drum crashes are the things to watch in this case.

[jayfrigo]

It's of course whatever sounds right, but generally speaking, shorter attack and faster release often works well for full-band tracks with drum transients. In mastering, compression is also usually used in conjunction with at least some limiting which has very fast attack times, so the two function together to achieve the desired result.

For orchestral stuff (as you mentioned), compression is most often not needed, but if you need a bit due to the real-world concerns of soundtrack music, sometimes upwards will work better for that kind of stuff than the typical rock and roll box. Either way, if it's a live orchestra, I'd be conservative with my application of dynamics processing.

[dcollins]

Originally Posted by Cellotron

Are you sure that BK was actually recommending 100 - 300ms for attack times though? This is extraordinarily long - I think he most likely would have been actually recommending 10 - 30ms.

That's what I was thinking too, it must be a misprint. You aren't going to find many analog compressors with 300ms attack time, anyway. I can only think of one I've ever seen that goes that long.


DC

[Suade]

In general, if you're not sure about setting your compressor attack time I'd say that you might not be hearing the difference all that well when changing the control. It might help if you first turn the compression way up (increasing the ratio and/or lowering the threshold) until you can hear a lot of compression. That way you can hear the difference when changing the attack time more clearly. Once you're happy with the attack (and release) you can then reduce the compression back to sensible amounts....

Hope that helps....

[SPORT]

Originally Posted by Suade

In general, if you're not sure about setting your compressor attack time I'd say that you might not be hearing the difference all that well when changing the control. It might help if you first turn the compression way up (increasing the ratio and/or lowering the threshold) until you can hear a lot of compression. That way you can hear the difference when changing the attack time more clearly. Once you're happy with the attack (and release) you can then reduce the compression back to sensible amounts....

Hope that helps....

thumbsup good advice.

[SSMastering]

Originally Posted by jayfrigo

For orchestral stuff (as you mentioned), compression is most often not needed, but if you need a bit due to the real-world concerns of soundtrack music, sometimes upwards will work better for that kind of stuff than the typical rock and roll box. Either way, if it's a live orchestra, I'd be conservative with my application of dynamics processing.

I just recently picked up an original soundtrack score for reference because I'm going to be mastering an album by a 30-pc orchestra.

Generally, when it comes to orchestral music, less is more.

As for what Bob Katz recommended, the direct quote from the book is "Typical attack times used in music mastering range from 50 ms to 300 ms, with the average used probably 100 ms." He then goes on to say how digital compressors react with textbook speed, etc. But as for what he recommends, I doubt many ME's have set the attack time to 300.

[dcollins]

Originally Posted by SSMastering

As for what Bob Katz recommended, the direct quote from the book is "Typical attack times used in music mastering range from 50 ms to 300 ms, with the average used probably 100 ms." He then goes on to say how digital compressors react with textbook speed, etc. But as for what he recommends, I doubt many ME's have set the attack time to 300.

No surprise I guess, but the slowest attack on my most used compressor is 40ms (indicated) and I never use it there.

Typical?


DC

[lu432]

Compressors are awesome tools, and distructive tools. Everyone here has posted some very helpful information. Something that will also come with experience and time is to learn the different sounds of compressors.

Learn to Identify how a LA2A sounds and how it differs from a Distressor ect ect. Then ask yourself why would you use certain compressors in certain situations? What are you hoping the compressor will do to the sound of the audio file besides the obvious? Once you start answering those very subjective questions on your own, you will find that you will have an arsenal of compressors/limiters that you will routinely use for specific purposes. You'll also become much more critical of your audio tools.

Hope that helps you on your path

[wado1942]

Bear in mind that Bob K does everything from metal to light jazz to modern-classical composition. I certainly wouldn't want to compress an orchestra for a film using a 10ms attack time.

For the record, I'm often 20ms-100ms with most of the music I master.

[Cellotron]

Originally Posted by dcollins

No surprise I guess, but the slowest attack on my most used compressor is 40ms (indicated) and I never use it there.

Typical?


DC

Again - this sounds like a typo. An attack range of 5ms - 30ms (with these numbers representing the nominal figures scribed on the faceplate and not necessarily what they actually measure in a lab test) actually sounds much more what occurs in the vast majority of mastering applications as used by the vast majority of ME's than 50ms - 300ms ever does.

The API 2500's (the most used comp in my rack) longest attack is in fact 30ms. Sometimes I wish I could set it to 40ms - but I don't personally see a need to have longer than this for this particular comp.

And I actually don't have any comps whose attack can be set longer than 100ms other than some digital tools which I never use for mastering applications.

Best regards,
Steve Berson

[dcollins]

Originally Posted by wado1942

Bear in mind that Bob K does everything from metal to light jazz to modern-classical composition. I certainly wouldn't want to compress an orchestra for a film using a 10ms attack time.

For the record, I'm often 20ms-100ms with most of the music I master.

There is probably no attack time that works on orchestra. It never even crosses my mind to use a compressor on such material.

First, I think the use of compressors in mastering is somewhat over-rated, and maybe 10% or less of the time does it really sound subjectively better to me.

But the key to learning to use one effectively on a stereo mix is to juggle the attack and recovery times to get it to 'pump' in a way that re-enforces the beat. This is also why you rarely see anything close to 100-300ms attack times in use.


DC

[jackthebear]

One of the reasons I like compressors that don't show numerical values.......it forces the user to turn the dials until it sounds good......

[alligatorlizard]

Thanks again, lots more useful advice since my last reply.

What I've settled on in the end is indeed pretty subtle - there's nothing really wrong with the tracks mix-wise, so I'm just looking to decrease the dynamic range a little so details don't get lost behind dialogue etc. As I am using a limiter at the end of the chain, it seems to make sense to keep the compression attack time longer, around 100ms, though when I get the studio powered up again tomorrow I will certainly try the exercise Suade suggests, as I'm not sure I if am hearing the difference clearly on the more subtle settings.

However, in this particular instance, due to the way I've decided to use the compressor, I'm not sure if the attack time really is going to make much difference anyway - I'm setting the threshold to -50db and using a very low ratio, 1.2:1

Hence it's always engaged, even in the quieter passages - which means that the attack time is not even really a consideration - right?

As I understand there are - broadly speaking - two main approaches to compression in mastering: either using it to reduce the peaks (and here obviously attack time would be important) or with a threshold about 30db below the rms level and a very low ratio - BK mentions this distinction, and I've seen these approaches described in a few other places too. As the peaks in these tracks didn't seem extreme, and as I didn't want to ruin the impact of any of the crescendos, I opted for the 2nd method (after trying both). Used in conjunction with the limiter, it seems to do the trick, ie reducing the dynamic range subtly, raising overall loudness, and actually bringing out some of the inner orchestral detail which otherwise might get lost in dubbing sessions.

I'm still very much trying to figure out all the possibilities here, so do let me know if this approach sounds drastically wrong, or mis-guided in any way!

[alligatorlizard]

Incidentally, here's the table from the Bob Katz book to which I was referring

Mastering audio: the art and the science - Google Books

SSMastering also quoted the following statement from the book: "Typical attack times used in music mastering range from 50 ms to 300 ms, with the average used probably 100 ms." I can't remember which page that quote is from, but I do remember reading it also. I doubt if these are both misprints?

[Waltz Mastering]

Looked up the attack times on the comps I use for mastering:
(Manley) Vari Mu: 25 msec to 70 msec.
Smc 2b: 1 - 100 msec
CL1b: .5 - 300 msec

[edva]

Originally Posted by alligatorlizard

Thanks again, lots more useful advice since my last reply.

What I've settled on in the end is indeed pretty subtle - there's nothing really wrong with the tracks mix-wise, so I'm just looking to decrease the dynamic range a little so details don't get lost behind dialogue etc. As I am using a limiter at the end of the chain, it seems to make sense to keep the compression attack time longer, around 100ms, though when I get the studio powered up again tomorrow I will certainly try the exercise Suade suggests, as I'm not sure I if am hearing the difference clearly on the more subtle settings.

However, in this particular instance, due to the way I've decided to use the compressor, I'm not sure if the attack time really is going to make much difference anyway - I'm setting the threshold to -50db and using a very low ratio, 1.2:1

Hence it's always engaged, even in the quieter passages - which means that the attack time is not even really a consideration - right?

As I understand there are - broadly speaking - two main approaches to compression in mastering: either using it to reduce the peaks (and here obviously attack time would be important) or with a threshold about 30db below the rms level and a very low ratio - BK mentions this distinction, and I've seen these approaches described in a few other places too. As the peaks in these tracks didn't seem extreme, and as I didn't want to ruin the impact of any of the crescendos, I opted for the 2nd method (after trying both). Used in conjunction with the limiter, it seems to do the trick, ie reducing the dynamic range subtly, raising overall loudness, and actually bringing out some of the inner orchestral detail which otherwise might get lost in dubbing sessions.

I'm still very much trying to figure out all the possibilities here, so do let me know if this approach sounds drastically wrong, or mis-guided in any way!

This is very common, even with ratios lower than 1.2:1, however, attack times definitely still matter in this situation. Play with it until you hear the best transient detail.

[Waltz Mastering]

Originally Posted by edva

This is very common, even with ratios lower than 1.2:1, however, attack times definitely still matter in this situation. Play with it until you hear the best transient detail.

There's only one more useful ratio between 1.2:1 and nothing right?
I guess you could go 1..5:1

[SSMastering]

Originally Posted by jackthebear

One of the reasons I like compressors that don't show numerical values.......it forces the user to turn the dials until it sounds good......

"A digital compressor set to 100 ms may sound similar to an analog compressor set to, say, 40 ms: so it's probably better to remove all the labels on the knob (except slow and fast) and just listen!" - BK

Steve has a point, the API 2500's attack only ranges from .03 ms to 30 ms.

Regarding BK's book, it IS a misprint! It has to be...

Originally Posted by dcollins

No surprise I guess, but the slowest attack on my most used compressor is 40ms (indicated) and I never use it there.

Typical?


DC

[edva]

Originally Posted by Waltz Mastering

There's only one more useful ratio between 1.2:1 and nothing right?
I guess you could go 1..5:1

Good catch, yeah sorry, I was a bit off there.

[Ari-M.]

there is no template rule that applies to all music....attack and release (for my projects) are always dependent on the program material

if I want more attack in a kick drum (hip hop) I will usually open up the attack time

if I want a "radio squish" for a pop tune or guitar driven music, I will tighten that attack time up....

but the one rule i employ, is that the settings on any device are always dependent on the material that is being processed....I am not sure any professional would argue with that approach

[Jean Doe]

I use ultra short attack times when mastering house/techno/minimal etc ALL THE TIME. If the bass drum has been mixed with too much oomph i.e. 100 to 200 Hz (and i get that all the time) it can help using short attacks (shorter than 5ms) and short releases to rebalance the tone of the bass drum. Essentially what you are doing is taking away "body" and yes, that's not a good idea in most types of music. There are always exceptions though

[SPORT]

Originally Posted by dcollins

But the key to learning to use one effectively on a stereo mix is to juggle the attack and recovery times to get it to 'pump' in a way that re-enforces the beat.
DC

thanks for the advice. thumbsup
let me ask a question: this mean that the attack and release setting is same for all songs whit same bpm?

[Sunim Koria]

No rules for getting a great sound, but I generally keep around the 100ms mark (paired with a fast release) to ensure that the compressor is leveling specifically on the sustain portion of the signal and to retain/create "punch". I use fast attack times (below ~5ms) when I want to pull down attacks/drum transients closer to the sustain, but over-using faster attacks in mastering can easily destroy any open-ness and impact that was created within the mix. Faster attacks paired with slow yet appropriately timed releases keeps the microdynamic waveshape relatively correlated and can be good for general calming down/beefing up.

Best,

SK

[MattGray]

Originally Posted by jackthebear

One of the reasons I like compressors that don't show numerical values.......it forces the user to turn the dials until it sounds good......

Or you could just close your eyes or look straight ahead.. works equally effectively. thumbsup

The values on the faceplate are there to help you get where you need to go quicker & also come in handy if you log your settings too.

[heybub419]

Yeah, I was kinda suprised when I read that in Mr. Katz's book as well (50-300!). I typically use between 5 - 25 ms attacks on my projects, about 15 ms average (often with Oxford Dynamics). I find that this works very well on many things; it still allows tansients through with less clipping. I master ITB, since I can't afford much hardware now lol, so I'm not quite sure how the attack times relate with an outboard comp.

[Odeon-Mastering]

Originally Posted by MattGray

Or you could just close your eyes or look straight ahead.. works equally effectively. thumbsup

The values on the faceplate are there to help you get where you need to go quicker & also come in handy if you log your settings too.

+1

One of the reasons I like compressors that don't show numerical values.......it forces the user to turn the dials until it sounds good......

+1

Just think of the compressor as an envelope shaper...