MAutoVolume - some experiments

[DarkStar]

I have been looking at MAutoVolume to see what it is doing. You might be interested in my findings and they may help you use the plug-in. I used an MOscilator sine wave and an MAutoVolume range of 24dB

The first thing to be aware of is that input levels below -39.8 dB are reduced in volume. I am not 100% sure that that should happen. Edit: That is because of the Noise Floor parameter.

Have a look at these curves, showing the effect of Sensitivity on the volume change:


The green line shows the volume for 0% Sensitivity. With 100% Sensitivity (the red curve) the Output volume is pretty constant but only for Input levels in the range -20 dB to +4 dB (shouldn't the effect cover Range * 2?). 50% Sensitivity (the purple curve) flattens the Output curve across a wider range and 75% (the blue curve) flattens it more.

It may be worth noting that it seems that the "target output level" (my term) is about -14.1 dB. Input levels above that are reduced towards it and input levels below it are increased towards it.

These curves show the effect of the Output parameter value.


Changing it just changes the transformed Output level by the specified amount. For example, setting the Output to -3 dB, sets the "target output level" to-17.1 dB (as shown by the lime green curve)


And now we come to the Range parameter.

For an Input of -24 dB, 100% Sensitivity and 24 dB Range, the Output Volume is about -15 dB (the dark red curve)
For an Input of -24 dB, 100% Sensitivity and 6 dB Range, the Output Volume is about -18.5 dB (the dashed orange curve), a gain of about 5.5 dB
With 50% Sensitivity, the Output Volume is about -20 dB (gain of about 4 dB)

-------------------------------------------
To me it seems more intuitive if the Output parameter were replaced by a "Target" parameter and the output level calculated appropriately. For example, a Target value of -12 dB would be the same as if the current Output value were set to 2.1 dB (= -12 - -14.1). Then we could set the desired level.

[Tyrs]

Excellent analysis, many thanks for sharing. It would definietley make sense to add proposed "target" parameter. Also a graphic interface would be helpful, like we see it in some compressors.

Could you please tell also something about your setup for the test?

[steve2KVR]

...there is a "mention" in the docs that -16db was the pivot point, really easy to miss, but it's there.../s~

[DarkStar]

I used MOscillator generating a sine wave and set the Wet level there (Dry level was "silence"), feeding that into MAutoVolume. That's it.

For a graphic interface, check out the Time-graph and display the Input and Output levels and the Gain.

I did hear Chandlerhimself mention -16 dB in his video, but my tests show it is about -14.1 dB.

t input levels below -39.8 dB are reduced in volume. I am not 100% sure that that should happen

Oops, my mistake, that's because of the Noise Floor parameter.

[DarkStar]

And here are examples of the monitoring available in the MAutoVolume time-graph (the left-hand scales refers to the Gain, I think, not the levels):



And here is a comparison in MMultiAnalyser:



You can see that the level and waveform has been smoothed out (the orange curves)

[DarkStar]

Last one for today, with the level adjustment operating both ways:



(I set the MAV Input to +12 and the Output to -12.)

[DarkStar]

TIP: in the above you will see that the loud peaks are actually boosted as the detected gain at those points is positive. This can be dealt with by increasing the Look-ahead parameter:

[vectorwarrior]

Thanks for sharing this. Yesterday I was trying out a 32 band Auto-Volume to see what effect it would have on a sound and this info is helpful to understand it on a data level. For what it's worth, it's been a fun experiment and could be a useful alternative to a compressor.

[steve2KVR]

Nicely done @DarkStar, very nicely done. I've been using PluginDr for testing.../s~

[Babymetalrocks]

I used MOscillator generating a sine wave and set the Wet level there (Dry level was "silence"), feeding that into MAutoVolume. That's it.

For a graphic interface, check out the Time-graph and display the Input and Output levels and the Gain.

I did hear Chandlerhimself mention -16 dB in his video, but my tests show it is about -14.1 dB.

t input levels below -39.8 dB are reduced in volume. I am not 100% sure that that should happen

Oops, my mistake, that's because of the Noise Floor parameter.

Does this mean you should set your input to -14.1 for MAutovolume to most accurately bring the top and bottom to the middle point?

[MeldaProduction]

Ok, so some technical info - MAutoVolume's "target" is -20dB prefiltered, which in some values of Speed follows the EBU values, so it could be translated to LUFS (but then there's another operations involved...). So honestly I'd suggest not trying to dig too much in that