speculation: FG-X protection limiter issue

[Jesse Graffam]

Hey guys, I was just going over a sample clip that one of the FG-X beta team guys sent to me, of good ol "MIX1".

Here:
Mix1 FGX v1_1.wav

The settings were basically close to this; the two Transients knobs at 7, Dynamic Perception at 5, and ITP at 0.3. Gain 8.5.

The distortion you'll be listening for is an upper midrange crunch. Apparently known more to users in v1.0, but still present in v1.1.

Going out on a limb that I'm 95% sure of... the following is NOT fact, just observation & speculation.

----

The distortion is being caused by a protection limiter after the clipper (or as Slate likes to word it "saturation"). I drew my conclusion just from hearing it, but let's go deeper for the sake of educational purposes.

Let's check that out... I did some very crude but functional analysis of the leading edge of the transients with the specific distortion, by capturing all of them as graphical waveforms with identical spacing. Then I added them all together with more than enough dynamic range to handle summing. Then I clipped the dynamic range of the image WAY down, to get rid of any disparities, and accentuate only the most extremely similar ones.

The result of the top 1dB zoomed in looks like this:


So we get a block of 4 samples that is very similar during the onset of transients with the specific distortion in question. It's almost identical on the 1st and 4th sample. There are some other samples that are relatively close, but that's a side effect of only doing this on one 1.5 minute clip with essentially very similar peaks, combined with the "saturation" effect's hysteresis across the would-be flat-topped waveform.

In case you haven't put one and one together yet... I'm speculating; that is a protection limiter at work.

----

My recommendation would be to have a look-ahead limiter there, performing the limiting while the audio is still oversampled, and making the attack & release of the limiting action long enough so that any distortion ends up getting mostly filtered out by the downsampling SRC.

Another problem could be that the limiter and/or something before the limiter isn't properly considering asymmetrical waveforms, which is causing the protection limiter to need to react more than it does under "normal" circumstances. You'll notice the peaks this distortion happens on are asymmetrical, as is the distortion. Try out a few trumpet solos if you really want to test that.

Getting rid of the need for a protection limiter in the first place would be a good long-term goal.

----



Anyways... it's great to see this product out there, getting improved from community feedback (glad to help, even tho I'm not on the beta team), and still quite usable and great results if people use their ears. (imagine that)

Cheers & respect to you Fabrice.

[Fabrice Gabriel]

Jesse,


thanks for your feedback !

First of all for this mix I would recommend slightly different settings. Indeed the upper range of the Transient parameters (typically above 5) and an ITP setting above the middle are reserved for the cases where you want a pretty aggressive mastering or when the mix lacks of dynamics.

Here you can let the ITP in the middle or a little below if you want to really push the gain, or even automate it across the song.
This is exaclty the purpose of the ITP parameter, which allows adjusting the behavior of the whole FG-X process with one single parameter. This way you don't have to automate a whole set of parameters to modify the response.

Also it does not really need the LoPunch to be set on 7, since the bass drum is already pretty punchy in this mix.

I or Steven will post the same mix with another parameter set.


Now when you say "The distortion is being caused by a protection limiter after the clipper", I don't really see how a limiter could cause distortion. A limiter is indeed more known to be transparent with a pure sine wave at the expense of killing dynamics.



We are aware that with the FG-X, some people may want to push their mix further, with more punch, without any artifacts of any kind.
First this is beyond the limits of physics since at some point, when you reach a 8, 7 or even 6 dB of Peak to RMS dynamics, you can not do it without scarifying something.
And second our recommendation is to use the FG-X to get more dynamic masterings, rather than louder ones.

At last, I think that there is no "push it and forget it" mastering process out there and people has fortunately still to use their ears!



Fabrice

[Jesse Graffam]

I have no doubt that the settings were "wrong". That's the most obvious part.

This is an exercise in pushing the limits (just a little bit), and finding out where stuff breaks down first. This is how things get improved, and from what I understood this particular distortion was the most common one people experienced.

As far as the LoPunch, and physics, etc... I get all of that. The distortion however is happening because of the sometimes asymmetric bassline in that recording. Would the bassline's sustain also be effected by the LoPunch param?

Originally Posted by fabricegabriel

Now when you say "The distortion is being caused by a protection limiter after the clipper", I don't really see how a limiter could cause distortion. A limiter is indeed more known to be transparent with a pure sine wave at the expense of killing dynamics.

Really though? You're no DSP newb. Come on broh. Even the cleanest brickwall limiters like Kjaerhus MPL-1 have some distortion, even with a sine wave. A limiter that for instance has no look-ahead and a static 4 sample attack & release time could distort like crazy, even if there was no sample clipping involved.

It's ok if you can't confirm or deny the existence of FG-X internal architecture. I totally understand where you're coming from there.

Originally Posted by fabricegabriel

our recommendation is to use the FG-X to get more dynamic masterings, rather than louder ones.

At last, I think that there is no "push it and forget it" mastering process out there and people has fortunately still to use their ears!

I'm with you there.

[Bang]

Originally Posted by Jesse Graffam

The settings were basically close to this; the two Transients knobs at 7, Dynamic Perception at 5, and ITP at 0.3. Gain 8.5.
.

These are really aggressive settings for this mix as Fabrice has said. Try the same mix with Transient knobs off and an ITP of 0 and you'll be really happy with the result

Thanks for posting.

[Fabrice Gabriel]

Originally Posted by Jesse Graffam

Really though? You're no DSP newb. Come on broh. Even the cleanest brickwall limiters like Kjaerhus MPL-1 have some distortion, even with a sine wave. A limiter that for instance has no look-ahead and a static 4 sample attack & release time could distort like crazy, even if there was no sample clipping involved.

Haa ok, you talked about a limiter without lookahead. Well, for me a digital limiter without lookahead is just something gross, so I even did not consider this option.

Well actually it's not totally true that a lookahead limiter is free of distortion, the designer has to usually choose between cleanness and loudness, but anyway limiting *is* unwanted artifacts.

Originally Posted by Jesse Graffam

It's ok if you can't confirm or deny the existence of FG-X internal architecture. I totally understand where you're coming from there.

Indeed yes, even with the revealed architecture, there would be all the settings to adjust in order to get the algorithm working, but that's not a good reason to do so.

Originally Posted by Jesse Graffam

This is an exercise in pushing the limits (just a little bit), and finding out where stuff breaks down first. This is how things get improved, and from what I understood this particular distortion was the most common one people experienced.

That's why I refined the ITP parameter to prevent exactly this kind of case. The first version was less forgiving, now you can set it in the middle while retaining all the punch and at the same time preventing artifacts in most cases. Getting the ITP up just allow you to push the things harder by gradually disengage a kind of "anti distortion system"

So if you mean "how things get improved at ITP=0.3", I would say, "lower the ITP".

Same thing with a limiter, one can say "wow I got my master so much louder with no lookahead, no attack and an ultrashort release. But it distorts, what's wrong with this limiter?".




Fabrice