Limiter Forms and Definitions

[Kim Lajoie]

A clipper, on the other hand, does not operate on a principle of gain reduction cell and envelope signal - it does not have any time-constant

Mauseoleum is right that clippers usually aren't implemented in the same way limiters are. Clippers usually don't have a gain reduction circuit - their internal design is closer to distortion than compression. For the purposes of this discussion, however, it makes sense to describe their behaviour in terms that can be compared to limiters.

Well, then how can it be described if we go by the definitions I started in the first post?

As I wrote in my first reply: A clipper can be seen as a brickwall limiter with instant release.

No, by 'instant', I mean it really is instant. Any sample that exceeds the threshold is reduced.

Funny though that some so called "brickwall limiters" work with microseconds, some even millisecond values to be called "brickwall". So what's the actual deal here?

If it's only a "brickwall limiter" if we go by two needed facts: being digital and sample based limitng, then why are so many tools called "brickwall limiter"?

Examples: Kjaerhus MPL-1 Pro, T-Racks 3 Brickwall Limiter (works with 0,05ms attack), Waves L1, etc.

Even within the constrains of instant attack and infinite threshold, there are still many variables. A lot of the character of a limiter comes from the release curve and how the curve and time respond to the audio.

Normally a brickwall limiter doesn't have an attack control, but the T-RackS brickwall limiter has one because it actually feeds the limiter into a hard clipper. A longer attack time allows you to clip the transients instead of gain-reduce them. It's a different sound.

Lookahead isn't necessary, but is often implemented to reduce distortion. Without lookahead, a brickwall limiter will operate a bit like a clipper during the attack phase.

So a "clipper" without attack/release, or nearly instant attack/release (microseconds again) is in basic theory a limiter with different knee values, no? (see T-Racks Clipper)

A clipper doesn't have 'nearly' instant attack and release. It *is* instant. A clipper's waveshaping curve operating similarly to a hard/soft knee.

I use my blog so that my posts are easier to find and available to a wider audience.

Understandable, but isn't the discussion in here insteresting as well?

It certainly is!

-Kim.

[Compyfox]

As I wrote in my first reply: A clipper can be seen as a brickwall limiter with instant release.

If we look at it as a compressor "form", then it's understandable.

A clipper doesn't have 'nearly' instant attack and release. It *is* instant. A clipper's waveshaping curve operating similarly to a hard/soft knee.

So if we don't have any time constants to alter, then a clipper must be in it's simple form a distortion device and not another form of a limiter, as it is sometimes known.

The knee then declared how strong the signal is "distorted", and the input gain is like a drive knob on a distortion pedal. The output gain then only declares the maximum peak level.

Then why all the confusion that a "clipper" is a compression form? Well if you look at it from a different angle, it somewhat is (cutting the peaks/limit the peak maximum, etc), but still. This is confusing the hell out of me, even after all those years.

Even within the constrains of instant attack and infinite threshold, there are still many variables. A lot of the character of a limiter comes from the release curve and how the curve and time respond to the audio.

Wait, I thought the "character" (sound) of a limiter is actually no sound at all, since it's mainly cutting peaks. Though it does change the "sound" in terms of gain riding.

Normally a brickwall limiter doesn't have an attack control, but the T-RackS brickwall limiter has one because it actually feeds the limiter into a hard clipper. A longer attack time allows you to clip the transients instead of gain-reduce them. It's a different sound.

This goes a bit off topic, but to my understanding (consulted the manual), the T-Racks3 Brickwall Limiter only "reacts" like a clipper if both the attack and release is put to it's minimal values. But isn't a clipper a distortion module (see above)? I didn't find anything that this plugin uses a clipper and brickwall limiting (unless you have the D/A Distortion Protection activated) in chain.


Leaves still the questions:
Why are certain limiters, even if they don't have instant attack, still declared as "brickwall"?

Who spread the rumor that a "look ahead" function is the only factor that declares what a "brickwall limiter is"?

Why is the LA2A, which uses the same features as known "maximisers" not declared as such?

And why is it, that some limiters (analog models) are used while (pre)mastering if they are clearly not a brickwall limiter, yet they catch all peaks the same way as digital limiters if setup to (very)fast attack?


Personally I think there is a lot of confusion going on in the audio realm.

[IIRs]

So if we don't have any time constants to alter, then a clipper must be in it's simple form a distortion device and not another form of a limiter, as it is sometimes known.

Yes, clipping = distortion.

A clipper is not another form of limiter, it is a possible alternative to limiting.

[Compyfox]

Well, but only if we talk about "cutting peaks", no? Else we distort the broadband signal as well.

I guess this is why a clipper-limiter combo is sometimes wanted to have more control over raising the loudness while using less limiting in the process to have a more transparent sound.

[susiwong]

Not wanting to contradict anybody, just wanted to mention that before the big war of volume at the expense of great sound people were looking at limiters a bit differently.
You were using a bus compressor to increase loudness to a sensible level and add some glue and additional punch to a mix.
If you look at the resulting waveform (normalized) you'll probably notice the average peak level sitting way below 0 dB, sometimes 5dB or 10dB lower.
More often than not you can't even pinpoint a certain musical event to cause this, there are some very short transients causing this phenomenon.
People noticed you could get away just fine without much quality loss if you shaved off these "useless peaks" while keeping 99% of the signal (up to the average peak level) intact and unchanged.
Done correctly you didn't hear any change in the music but you could safely push up the volume of the mix by a few dBs - voila, classic limiting.
Musical dynamics and sound untouched.
Additional bonus: increased s/n ratio, back then the media caused lots of noise.
Depending on how you look at it, it all fell apart when some smartass discovered he could sacrifice some more sonic quality (who owns a good hifi sytem anyway ) and "push the mix in the red", shaving off some of the average peaks too, and thus drown the competition.
The day the music died.
It went downhill from there ...
Of course part of that is an acquired taste now in radio-oriented genres, though, just like the creation of the universe, it's widely regarded as a bad move.
Ymmv,
susiwong

[IIRs]

Well, but only if we talk about "cutting peaks", no? Else we distort the broadband signal as well.

Yes. You can't abuse a clipper in the same way as a limiter as you will get obvious distortion. But for short peaks a clipper can be more transparent then a limiter, the extra harmonics added for the very short duration of the peak being less noticeable than gain pumping.

[Compyfox]

Yes. You can't abuse a clipper in the same way as a limiter as you will get obvious distortion. But for short peaks a clipper can be more transparent then a limiter, the extra harmonics added for the very short duration of the peak being less noticeable than gain pumping.

Like I expected and what susiwong also confirmed.

Though I still wonder why it counts to the "limiter" branch, if it's a distortion device. Because it cuts off the peaks?

If you look at the resulting waveform (normalized) you'll probably notice the average peak level sitting way below 0 dB, sometimes 5dB or 10dB lower.
More often than not you can't even pinpoint a certain musical event to cause this, there are some very short transients causing this phenomenon.

Ah, the good old days indeed. Stuff from the 80ies, early to mid Jackson stuff, early Genesis, Dire Straits, etc. I definitely remember those days.

Though I wouldn't have imagined it be so simple.

People noticed you could get away just fine without much quality loss if you shaved off these "useless peaks" while keeping 99% of the signal (up to the average peak level) intact and unchanged.

I guess with the birth of the CD, this was also the birth of the digital limiters (DAWs came up the same time anyway) and then the death of listenable-to-music. In short, the loudness war began.



But to get back to the initial Q:
How do you define the following three buzzwords according to your knowledge?

Limiter, Brickwall Limiter, Maximiser, Clipper



Oh and before I forget it, thanks to all the users in here for participating and discussing so far.

[mauseoleum]

BLAH BLAH BLAH BLAH BLAH BLAH

One word for you: BIG MUFF

[IIRs]

Limiter, Brickwall Limiter, Maximiser, Clipper

Limiter: compressor with ratio higher then 10:1

Brickwall Limiter: infinity:1 ratio, plus a guarantee not to allow any overs at all.

Maximiser: Brickwall Limiter plus possibly other things like multi-band limiting.

Clipper: distortion.

While there is some overlap between the first three definitions, the clipper is a distictly different beast. The only link between it and the others is that they can all provide ways to tame short peaks and thereby raise the average level.

[Compyfox]

One word for you: BIG MUFF

Two words actually.
And you mean this one (and it's emulation) or rather the... well... other meaning...

Limiter: compressor with ratio higher then 10:1

Brickwall Limiter: infinity:1 ratio, plus a guarantee not to allow any overs at all.

Maximiser: Brickwall Limiter plus possibly other things like multi-band limiting.

Interesting viewpoint on the Maximiser... I've seen multiband mastering limiter that are built like maximisers (e.g. Waves and Izotope), but I try to avoid them as much as possible.

Clipper: distortion.

While there is some overlap between the first three definitions, the clipper is a distictly different beast. The only link between it and the others is that they can all provide ways to tame short peaks and thereby raise the average level.

Then I definitely vote for a category-tag change for the clipper.

[Kim Lajoie]

So if we don't have any time constants to alter, then a clipper must be in it's simple form a distortion device and not another form of a limiter, as it is sometimes known.

That's correct. A clipper *is* a distortion device. Clippers are sometimes used as limiter though, because in some situations they can perform the same function - reducing peak level in order to reduce headroom requirements of the audio.

The knee then declared how strong the signal is "distorted", and the input gain is like a drive knob on a distortion pedal. The output gain then only declares the maximum peak level.

Almost. The 'knee' on a clipper controls the waveshaping curve. A soft knee will clip less harshly, but begin clipping at a lower level (so you'll heat it more). A hard knee will clip more harshly, but will do so less often (only on the peaks, when absolutely necessary).

Then why all the confusion that a "clipper" is a compression form? Well if you look at it from a different angle, it somewhat is (cutting the peaks/limit the peak maximum, etc), but still. This is confusing the hell out of me, even after all those years.

Sorry about that.

Wait, I thought the "character" (sound) of a limiter is actually no sound at all, since it's mainly cutting peaks. Though it does change the "sound" in terms of gain riding.

An ideal limiter has no 'sound' at all. In reality, any process that changes audio has the potential to be audible, and there are many design decisions that go into making a limiter. Each of those decisions has the potential to change the audible behaviour of the limiter.

This goes a bit off topic, but to my understanding (consulted the manual), the T-Racks3 Brickwall Limiter only "reacts" like a clipper if both the attack and release is put to it's minimal values. But isn't a clipper a distortion module (see above)? I didn't find anything that this plugin uses a clipper and brickwall limiting (unless you have the D/A Distortion Protection activated) in chain.

The T-RackS 3 manual doesn't completely describe the internal design of the processors. I can see how that can be confusing.

Why are certain limiters, even if they don't have instant attack, still declared as "brickwall"?

Because they don't, under any circumstances, allow *any* audio to exceed the threshold. That's all a 'brickwall' limiter is.

Who spread the rumor that a "look ahead" function is the only factor that declares what a "brickwall limiter is"?

I don't know. There's a lot of misinformed rumour that gets passed on. The marketing material of most gear/software companies doesn't help.

Why is the LA2A, which uses the same features as known "maximisers" not declared as such?

See my original post in this thread. 'Maximiser' is a marketing term that doesn't have any real meaning. If you want to be generous, you can read 'maximiser' as 'a louder brickwall limiter'. The LA2A is not a brickwall limiter.

And why is it, that some limiters (analog models) are used while (pre)mastering if they are clearly not a brickwall limiter, yet they catch all peaks the same way as digital limiters if setup to (very)fast attack?

Analogue limiters are used for the same reasons any kind of limiter is used in mastering - to reduce the headroom requirements of the audio by reducing the peaks independently of the average level.

Personally I think there is a lot of confusion going on in the audio realm.

Certainly. Music is art, but audio engineering is science. I think a lot of confusion is spread by people who don't accept or understand the science.

-Kim.

[rifftrax]

I'm not sure the term "brickwall limiter" is all that scientific to begin with.

Sounds like a nice little term someone in R&D thought of.

[mauseoleum]

I'm not sure the term "brickwall limiter" is all that scientific to begin with.

Sounds like a nice little term someone in R&D thought of.

Who gives a Flying Funk about "scientific"?
You should know that "brickwall" comes from "thick like a brick".

Mann... this IS science:
http://www.freepatentsonline.com/3223936.html

BIG MUFF in all it's scientific glory!
Slurpy big muff, mann ... in a nurse outfit, with ponytail, fishnets AND with glasses.
The whole package.

[Compyfox]

That's correct. A clipper *is* a distortion device. Clippers are sometimes used as limiter though, because in some situations they can perform the same function - reducing peak level in order to reduce headroom requirements of the audio.

Then it is clearly wrongly tagged as "limiter".

Almost. The 'knee' on a clipper controls the waveshaping curve. A soft knee will clip less harshly, but begin clipping at a lower level (so you'll heat it more). A hard knee will clip more harshly, but will do so less often (only on the peaks, when absolutely necessary).

Much like a soft knee on a compressor compared to a hard knee. Totally understandable. But if it's actually controlling the shape, why call it "knee"?

Another wrong tagging I'd say.

The T-RackS 3 manual doesn't completely describe the internal design of the processors. I can see how that can be confusing.

You don't have to go in-depth, but how did you find that out? C.Budde's Analyser doesn't show everything in that section. Unless the "protection mode" of the T-Racks3 Limiter is actually the "limiter" and the other settings the "clipper".

Then again, not that important - we're drifting plugin-detail-OT here.

Why are certain limiters, even if they don't have instant attack, still declared as "brickwall"?

Because they don't, under any circumstances, allow *any* audio to exceed the threshold. That's all a 'brickwall' limiter is.

But, according to your last posts, only "digital" and "instant attack" declares what a brickwall limiter is. A 0,05ms limiter would still let transients through if we are nitpicky, no? So what about those limites that work on an 1ms to 5ms basis?

The LA2A is not a brickwall limiter.

Well it is an analogue limiter, true. But in limiter mode it's setup to ratio 10:1 and the attack is reduced to 10microseconds as well (according to specs). Unless the factor of the opto compression and the detection characteristic (feed back compared to feed forward) also take a huge role into this.

But it's usage is like that of a nowadays known maximiser at least.

Analogue limiters are used for the same reasons any kind of limiter is used in mastering - to reduce the headroom requirements of the audio by reducing the peaks independently of the average level.

In it's basic form, yes. If we look at old recordings (beginning of the loudness war), yes to that as well. But sometimes I still see people using "vintage limiters" to raise their loudness to nowaday commercial level and not using any additional "brickwall" or "clipper" at al. Though these are rare exceptions.

Certainly. Music is art, but audio engineering is science. I think a lot of confusion is spread by people who don't accept or understand the science.

Well, like rifftrax said, I'm also sure that R&D played a huge role into that as well. Then some crazy people catch that word and spread it like a wildfire. It is later adapted to be the "ultimate answer". To me it's like with the buzzword "warmth" - it can't be scientifically described, yet everyone uses it as declaration of a device's character.


So to sum things up:

Limiter (generic):
- basically a compressor with 10:1 up to infinite:1 ratio
- freely adjustable attack and release times
- freely adjustable makeup gain


Brickwall Limiter:
- "Brickwall" is a buzzword for this limiter setup
- purely digital device, rarely analogue
- basically a compressor with 10:1 up to infinite:1 ratio (later more likely)
- mostly very fast attack (microseconds range), sometimes attack is instantly
- freely adjustable release times
- freely adjustable makeup gain
- look ahead function optional to prevent distortion with the downside to introduce pumping if signal is turned up too strong


Maximizer:
- "Maximizer/Maximiser" is a buzzword for this limiter setup
- basically a compressor with 10:1 up to infinite:1 ratio
- mostly fixed attack and release times (attack in microseconds, release moderately) and therefore reflects the setup of a "Brickwall" Limiter
- controls are simplyfied
- Makeup gain usually corresponds to the input gain (e.g. -3dB threshold/input gain, automatic makeup gain of about +3dB), signal is driven "into the threshold"
- threshold is 0dBFS, unless the output stage can be adjusted additionally
- look ahead function optional to prevent distortion with the downside to introduce pumping if signal is turned up too strong


Clipper:
- "Clipper" is a buzzword
- a Clipper is not a compressor/limiter but a distortion device
- since distortion devices are no time-based effects, the signal is affected instantly (instant attack/instant release)
- "Knee" is another description of the waveshaping form for distortion
- Makeup gain corresponds to the input gain (e.g. -3dB threshold/input gain, automatic makeup gain of about +3dB), signal is driven "into the threshold"
- threshold is 0dBFS, unless the output stage can be adjusted additionally
- a Clipper can be reproduced with a (digital) compressor if it has instant attack (or ultra-fast), instant release (or ultra-fast), and a very high ratio



SUMMARY:
Buzzwords sure add to the general confusion. And clearly not "every new invention" with crazy buzzwords is actually what it's promising.


Again, thanks to everyone for contributing.
Definitely fills some knowledge holes.

[Kim Lajoie]

That's correct. A clipper *is* a distortion device. Clippers are sometimes used as limiter though, because in some situations they can perform the same function - reducing peak level in order to reduce headroom requirements of the audio.

Then it is clearly wrongly tagged as "limiter".

Who's doing the tagging? And on what basis?

I think there's a fundamental misunderstanding that's under the surface here.

A pure clipper works by waveshaping.
A pure compressor works by gain reduction.

A clipper can be used as a distortion device.
A clipper can be used as a limiter.
A compressor can be used as a limiter.

A 'limiter' may use a mix of gain reduction AND waveshaping.

The T-RackS 3 manual doesn't completely describe the internal design of the processors. I can see how that can be confusing.

You don't have to go in-depth, but how did you find that out? C.Budde's Analyser doesn't show everything in that section.

I found it out by analysing the waveform, and knowing what to look for. A slow attack results in clipped peaks, a fast attack doesn't. It's subtle and easy to miss (I almost missed it myself).

Why are certain limiters, even if they don't have instant attack, still declared as "brickwall"?

Because they don't, under any circumstances, allow *any* audio to exceed the threshold. That's all a 'brickwall' limiter is.

But, according to your last posts, only "digital" and "instant attack" declares what a brickwall limiter is. A 0,05ms limiter would still let transients through if we are nitpicky, no? So what about those limites that work on an 1ms to 5ms basis?

A limiter with a non-zero attack can't be a brickwall limiter unless it either has a lookahead delay (so it starts reducing gain *before* the peak) or it's followed by a clipper or other overload protection. Just because a device has an 'attack time' control and is billed as a gain-reduction limiter doesn't mean that's all it does. There's usually a lot more under the hood.

The LA2A is not a brickwall limiter.

Well it is an analogue limiter, true. But in limiter mode it's setup to ratio 10:1 and the attack is reduced to 10microseconds as well (according to specs). Unless the factor of the opto compression and the detection characteristic (feed back compared to feed forward) also take a huge role into this.

But it's usage is like that of a nowadays known maximiser at least.

Who says an LA2A is a maximiser (or a brickwall limiter)? I've never heard of it, and Google seems to think it's *only you* who's saying it.

So to sum things up:

'Brickwall' and 'Clipper' are certainly not buzzwords. They have very specific meanings, which I've described several times in this thread. Your lack of understanding doesn't make those terms meaningless.

Perhaps I haven't been as clear or unambiguous as I intended.

-Kim.