AGC - automatic gain control RMS window

[JJ303]

I’m a big fan of automatic gain compensation in plugins and we also seen this becoming more popular lately.
The AGC in Melda plugins is great but have its own know limitations as described in the documentation, and this is that it may cause pumping.

From experience I find that enabling AGC on EQ works great, but even then it can cause strange volume fluctuation and even sudden loud volume increases in transient material.

However on dynamic processing as with the MDynamics plugin range, or dynamic bands in MAutoDynamicEQ, it will cause “pumping” so you cannot judge the actual compression applied with AGC enabled.


QUESTION:
What I’m asking is if there would be a possibility to increase the RMS window for the AGC, like a manual setting, or in other way make the gain compensation of the AGC to react slower?


As for some example of great gain compensation implements we have (however do NOT misunderstand me, I still prefer Melda plugins over these):
- DMG Audio TrackComp 2 that only correct the gain directly after you change a parameter but not like pre-defined auto gain but actually measuring it

- TbProAudio AB-LM Lite implementation in their plugins where you have the choice or Fast and Slow compensation

- izotope Ozone were the actual gain compensation is not that great, and actually quite buggy, but the great feature is that the plugin automatically freezes/sets the applied gain compensation when you close the plugin UI so you do not need to worry that you forgot to turn it off

Hope above make sense

[Held]

There's an option for this in MCompare, MAGC's big brother.

ALC window controls the length of the loudness calculation window. The longer it is, the slower the response to loudness changes, but it also avoids quick fluctuations, smoothing them out. Set it to a longer value if the playback seems fluttering, or contains unnatural level changes.
Range: 1000 ms to 30000 ms, default 3000 ms

[JJ303]

Thanks Held,

I was actually just playing around with MCompare after realising that it also allowed gain matching loudness between two instances and not just reference tracks.
I have to say it is much better than any other similar tool including LetiMix GainMatch that I currently use.

The default 3000ms RMS window setting in MCompare is much better suited for gain matching than the 400ms RMS window in the AGC function included directly inside the Melda plugins.

So if there would be an option to increase the AGC functions RMS window to 3000ms, then that would be great and would love to see that as a feature. And I also believe that may increase the attraction of the Media plugins.

As mentioned above, I currently use LetiMix GainMatch but may now start using MCompare instead.
However the issue with this approach is that you always have to insert two instances, one before and one after the plugin you want to gain match the in and out of. You are also unable to set the applied gain when done and have to manually apply it when the instances for gain matching are removed.
All this will eat up time and stop your flow when mixing, so do not use it that much. And this is why I wish you could do this directly inside the Melda plugins.

It is much easier and quicker to have gain match function included directly in the plugin, and Melda plugins are almost there except that the RMS window is in my view to small.

[Held]

Right, you were talking about the AGC setting inside each plugin. For some reason I thought you were talking about MAGC (the plugin).

In that case, I agree that it would be great to have more settings available for AGC.

At the moment, I mostly use the 'set' button for dynamic plugins, but having to do that after each adjustment is a bit annoying. On the other hand, if the RMS window is 3000ms, you'd have to always wait 3s until you get the adjusted volume which isn't very transparent.

An option could be to change the color of the AGC button every time a value in the plugin is changed, until the RMS has adjusted. That way, it's easy to tell when you can start comparing again.

[JJ303]

@Held,

I would agree that 3000ms is not always the best option, so adjustable RMS window for the AGC like in MCompare would be great so you can use smaller window when suitable.

I do however not fully follow what you mean by "change the color of the AGC button every time a value in the plugin is changed, until the RMS has adjusted".

[Held]

I do however not fully follow what you mean by "change the color of the AGC button every time a value in the plugin is changed, until the RMS has adjusted".

When you have an RMS of 3 seconds, it takes 3 second until the volume is matched after you change a setting in the plugin. So if you change something and start bypassing immediately to compare it to the input signal, the volume might be different.

So I thought it would be cool to have a visual signal, e.g. the color of the AGC button, to tell you when it's safe to start comparing again.

[Hexspa]

Wouldn't you want a shorter AGC time? Ideally, you have instant volume matching for every altered parameter. You could have that with a buffer (latency), right? Like a lookahead function.

A neat feature to include might be to have the sound not change until you release the mouse button. Kind of like how graphics programs used to just show the outline when you moved an object instead of continually redrawing the whole window. That, coupled with instant AGC would give more contrasting impressions of parameter changes; like a pulse response instead of a sine.

[Svama]

+1 for a faster and „better“ agc mode, without pushing the set button. But 3000ms doesn’t make sense too me and the pumping behaviour as it is for now, although not. CraveEq hadn’t nice agc and the other already mentioned and many more.
But this a debate we‘ve already made, but maybe this time?!
Would be greater than any new plugin and GUI to me.

[JJ303]

I’m happy to see that it’s not just me that would like an improved AGC

@Held:
That make sense, I thought I missed something with the current implementation on AGC.


@Hexspa:
What I mention about RMS window is not about how quick the AGC corrects the gain but rather the time it measures.
Shorter time may cause pumping/fluctuation with the AGC when doing dynamic processing as smaller the RMS window the closer to peak measurement we get.
However not knowing all ins and outs of the AGC coding I may be incorrect on some parts here, so this is good discussion to have and maybe the Melda guys/girls may have some input.

Your description “sound not change until you release the mouse button” sounds like how the agc function in DMG Audio TrackComp 2 and it is to me the best one and would of course love that, but believe that would be a lot of reprogramming for Melda to implement and and the possibility to change the RMS window may be an easy and more plausible implementation.


@Svama:
Let’s hope that we get an improved AGC this time
And maybe 3000ms is a bit extreme and 1000ms may work better in most situations, but nice to have the option like in MCompare.

As for Crave EQ. I have and sometime use Crave EQ due to the nice pan per band feature (hint hint Melda), however the agc in is not based on measurement like AGC in Melda but rather predefined curve based on white noise (or similar). So a cut/boost in a region that do not contain any “signal” will still affect the applied agc gain. However much better than FF Pro Q agc as Crave exclude any high/low pass in the agc while Pro Q do not.

[Hexspa]

Shorter time may cause pumping/fluctuation with the AGC when doing dynamic processing as smaller the RMS window the closer to peak measurement we get.

I understand that. It's no different than setting the compressor RMS to 'Peak'; more liklihood for distortion.

As it is, a 3s integration time will level out the plugin 3s after you make the change.

With a 'lookahead buffer' (adding latency), you'll hear what you did 3s ago but it'll be volume compensated.

I don't know. Seems weird now that I think about it. But - if the plugin didn't 'execute' the command until you let go of the mouse button, with a variable RMS (however long you hold down the mouse button), you'll hear a volume compensated result as soon as you let go. In this case, AGC would only work from the time you click until the time you release. It wouldn't be continually adjusting outside of that window of time between mouse events.

Or it could but the variable AGC would let you A) define the AGC window in real time and B) give you an instant result once you let go which includes your parameter change (might require a buffer).

Like I said, I'm spitballing.

[JJ303]

Yes, there are so many different approaches you could take for agc. This as we can see in many different plugins, some good some bad, and one solution may not fit all.

I however find the functionality in both MCompare and MAGC working great for at leats my needs, especially as you can change the speed/window to suite the situation. And believe that the detection behind MCompare and MAGC may be similar to the per Malda plugin AGC, so may be the easiest thing for Melda to implement and maybe it just require us to get access to change the RMS window of the AGC.

Again, I may be off here.


My dream AGC implementation would be something similar to either:

- DMG Audio TrackComp 2 that only calculate and correct the gain directly after you change a parameter, and then I looks to be disengaged so you do not need to worry about strange behaviour if you forget to turn it off

- The detection of MCompare, MAGC or TbProAudio AB-LM Lite, and then the izotope Ozone functionality that it sets/freezes the applied gain and disengage the agc when you close the plugin GUI


If anyone else have a dream agc function, please post as it may be hundred times better than mine above. Easy to overlook things.

[Hexspa]

That sounds nice. For example, click AGC and it flashes at an accelerating rate over the time period you defined. Once it stops flashing, AGC has been completed and now the output level will remain at what it set.

Honestly, I don't use the AGC feature in the plugins. I'm sure you've making good suggestions.