FR: MAutoVolume - independent attack and release?

[Tagirijus]

Hey there,

I tried very hard, the last days, to dig into MAutoVolume. I think I understand the principle in general and what the knobs are for etc. Unfortunately I cannot make it to ride the gain without making it really sound like a compressor.

Here comes the problem. I would even accept a latency over 500 ms, if it would realy ride the gain without pumping, but still stabilize the loudness.

Wouldn't it be cool if MAutoVolume had independent attack and release knobs? If I set the speed so that it gets 500 - 1500 ms latency it does not pump anymore, yeah, but it also does not stabilize quite huge loudness differences very well. So it soudns like it cannot change the gain quickly enough or it releases too fast. That's why I thought if independent attack / release would be a good thing.

Or do I want too much from the plugin?

Edit:
Additionally a question: wouldn't be loudness stabilization according to the LUFS be a good idea?

[steve2KVR]

...MAV is for "gain," which could be thought of as "instantaneous"...LUFS is measured over time, momentary>short-term>integrated. Related, but very different concepts.

I use MAV pretty extensively, in that most of my current work is on live multi-track recordings that are essentially completed projects. The clients want/need them "fixed; there's no going back into the studio. In some cases, the age is measured in years.

There are a number of threads here already with really good advise and suggestions, maybe look them up and see if that helps.
"Attack" and "release" are in fact "knobbed" on the control panel, just not named/displayed as such.

Oh yaaa, lookup the ChandlerHimself vid. Pretty good stuff...hth.../s~

[Tagirijus]

...MAV is for "gain," which could be thought of as "instantaneous"...LUFS is measured over time, momentary>short-term>integrated. Related, but very different concepts.

Isnt LUFS in fact just prefiltering with differnet time ms windows? So that could be used for loudness stabilization in theory, right?

There are a number of threads here already with really good advise and suggestions, maybe look them up and see if that helps.

If I search MAutoVolume I just find my thread and the nw tutorial video from Chandler.

"Attack" and "release" are in fact "knobbed" on the control panel, just not named/displayed as such.

Well I just understand "Speed" as attack+release combined, am I wrong? You cannot access both independently, or did I miss something?

Oh yaaa, lookup the ChandlerHimself vid. Pretty good stuff...hth.../s~

Thanks, I know both videos already. Great videos. Didn't help in this case, though.

[steve2KVR]

...perhaps if I could hear/see your use case I could be more specific. I don't view MAV as a compressor/compander/etc., and as such don't use it in that capacity.

Try putting an instance of MLA before and after MAV, that way you can observe for yourself the effect when changing the settings in MAV on both "gain/DBFS" and "LUFS."

No clue on not finding the other threads. I participated in several of them...(sigh)...hth.../s~

[Tagirijus]

...perhaps if I could hear/see your use case I could be more specific. I don't view MAV as a compressor/compander/etc., and as such don't use it in that capacity.

Yeah, sorry. Forgot to mention, stupid me. I want to use MAV for mixing voice overs.

Thanks for your effort though.

[steve2KVR]

...you should/could get absolutely marvelous results in that case...voice over what kind of material? I've done some work like that in the past, no worries.../s~

[Tagirijus]

I cannot upload the material with which I tested it, since it's for a client. I'll make some new test material the next days.

[steve2KVR]

...no worries.../s~

[jmg8]

You probably already know this, however:
The first step should always be to set the input and output gains.
Let's assume that your vocal (or whatever else you are using) is already at the loudness that you want.
Use the input gain to make the "G" meter move around equally above and below the 0.
Now do the opposite to the output gain (if you added 6db to input use -6db on output)

Now MAutoVolume will be working at its best.
After you have made adjustments to the other settings, you may need to go back and adjust the output gain at the end, but leave the input gain as it is.

[steve2KVR]

...^^^^^^^^^what he said^^^^^^^^^...jmg8, well said sir, well said...buried deep in the docs/vids somewhere is the observation that MAV has a "fixed" center threshold of approx -16DBFS, use the pre/post gain knobs to adjust you channel gain to that, which is best done with the meters set to display in "graphs"...hth.../s~

[Tagirijus]

Hey people, thanks for your replies. Yes, I already know all that and I follow these isntructions already, of course - I try not to be that guy asking before reading and trying a lot, hehe.

My problem (why I started this thread) is that MVA either sounds like a pumping compressor (when the speed may be to quick) or it does not gain quickly enough (when the speed is to slow ... even with 100% lookahead). There doesn't seem someting inbetween due to the attack+release-link according to my assumption. I'll try to upload an example soon.

[Tagirijus]

Now I did some further testings. Hopefully this will show either the lack of feature in MVA or the lack of my understanding of the tool.

I did this recording and I just did some EQ on it:

http://www.tagirijus.de/clients/tmp/01_ ... iginal.mp3

Then I did a manual mix, like I always do for my clients - using MTurboComp for some compression and then automating the output level to reach around -23 LUFS before mastering it to the overall loudness the client wants:

http://www.tagirijus.de/clients/tmp/02_ ... al-mix.mp3

Now I tested some medium settings in MVA:

http://www.tagirijus.de/clients/tmp/03_ ... um-mix.mp3

Code: Select all

$eNp1kVFvgjAUhd-9FU19nVgQCySAcXNbTCQayXxncKeNpSW0GN2v30WzLVlm3-qdc3tP741n51qSE7RGaJVQ12GUgCp1JdQ+oZ39GIV0lg7ibN5ZvdOyq8GAtagaMkQ-JcMFWCitbp+0lEVjoLrhldbH+QEKvDLHjzgi3dmms-tCYKfRFEEOyggrTsJeeldwZQ1AlekKEppBJbqakp+Xt2J-sJtCgUR7z5VtUcP0z6p4l31rxK9vy1x8goc5plP2wBmy1TZvpLAW2l5yExq6kcMmEQsmIeN-dSzlzHMiziKPu5FPSVacVy-rR6nLY29I6MSjZPc9tiDkke-TNM5wFq0ht-g4hRxKrSqzgXYjztfUDmP89wRYs1QZWQjTyOKCXUTd1f0vxmmciwruSOvO3lHGtwi4svE-O0sHX+jpnAI=

But I do not like it. It somehow pumps. So I thought I increase the lookahaed and lower the speed to slow:

http://www.tagirijus.de/clients/tmp/04_ ... ow-mix.mp3

Code: Select all

$eNp1UU1vgkAQvfsrNuu14vLhIglgbG0bE4lGUu8Uprhx2SWwWO2v71DaHpp6nPfezHszEy4ulSRnaFqhVURti1ECKteFUGVEO-M2mdNFPAqTZWf0QcuughaMQbYlY9RTMl6Bgdzo5kFLmdUtFAO80fq0PEL2XerO1J0pM4EuEx+BfaZKiKiLhuMUVCuMOAtzHdRpDVAkukBBKvU7Jb-D96I8ml2mQEaU9bgyDXK4wKPKXmXvjvDzyzoVH+DgtNmM3XGG2Gaf1lIYA01P2RGd24HF3ID57pzxvzy2cuZYAWeBw+3AoyTJLpun7b3U+akXYHSHksPP5fw5DzyPxmGC52haMoTHXVLItSraHTQ7cflKbTEvCDyXM27PfI4ta5WQlWhrmV3RRFRd1S8xjcNUFHCD2nbmBjMdEuDTpv98LR59AjTvnLY=

The graph output also shows what I just dislike:



The sentence is not finished, but MVA already rides the gain downwards. What is going on, why is it doing this? Isn't this a thing what could be tweaked by adjusting attack and release independently? Or maybe even something like a dedicated hold value?

[jmg8]

OK. Now I can see it.

Basically, it is the lookahead causing this.

Look-ahead delays the actual signal being processed, but keeps the detector signal intact. This makes the processor use a signal that has not actually arrived for dynamic calculation. This allows the processor to respond even faster, in fact, ahead of time. This feature is useful for mastering, however it naturally induces latency.
Look-ahead can be available in milliseconds (with obvious meaning) or in percentages. In percentages the look-ahead delay is computed automatically based on the attack and hold times. For example, if look-ahead is 100%, attack time 2ms and peak hold 10ms, then the look-ahead is 10ms; 60% look-ahead would be 7.2ms. If the look-ahead is simply an on/off switch, then it is toggling between 0% and 100% values.

Before using look-ahead, you should understand what such a feature does exactly as the results can potentially be damaging to your audio. Look-ahead basically moves the signal back in time, in other words its signal detector measures the input levels ahead of time. This means that when the detector is in the attack stage, the level is rising, the actual signal is not rising yet, but it will do so soon. However, the same applies to the release stage! When the detector moves to the release stage, the actual signal is not falling yet. This can lead to very strange artifacts (which can be used creatively of course).

The common way to fix this is to set the release time considerably higher than the attack time. In this way, the level will rise ahead of time in the attack stage, and same will happen for the release stage and the level will go down, however, since the level is falling slowly, the look-ahead will not be that relevant.

Another option is to use the peak hold feature. It is highly recommended to enable true hold in the advanced detector settings if available. Essentially this feature maximizes the input level over a certain period of time. So for example, if you set look-ahead to 5ms and peak hold to 5ms as well, the actual signal will arrive 5ms later than the detector signal, however the peak hold feature will ensure that the detector holds the highest peaks for 5ms, so the attack stage will be ahead of time, but the release will not! You can consider it a form of latency compensation for the release stage.

So if a hold feature was added this would solve your problem.

I must ask a few things:

1. Do you really need lookahead on? It might be possible to just use the speed to get it right.

2. Why is the Input gain 0db? I explained the importance of setting that first.

3. Could a workaround be to modulate the speed parameter with a follower mod?

[steve2KVR]

...using your mp3 source, but without the preceding compression, try these...
1/
...using another chanel put an instance of MNoiseGen( pink noise@-23DBFS output)
...set the output of that chanel to mute (don't want to hear it, just using it for side chain)
...paste this code into MAV and set the side chain input to the MNG chanel...(don't know you DAW)

Code: Select all

$eNp1kV1rwjAUhu-9FSHebDBrtR9aaCtOtyFYFMu8z2zUYJqU5kR0v36n0zkY8zLv++Y5X-HoVEpy5LURWiW057iUcLXRhVC7hFrYdoZ0lLbibGxBr7W0JTccAF1D2pinpD3lwDeg64mWklWGFxdZqMrCjgmkRvica30Y7zlD13UCFLSFW6ATopBzZQSIo4DzhZCLgk-2GHhR7EPyq1jxa4VbvZXY7WHJFJfIRt3WNVewrJtOZ9OEZuxEzC+cPGyZgUdK3t5nufjkfaT5rv-kRwidr-JKCgBeN1YvoYEbONHAD4IoHA7++vg19DzH7Qe+73o9jxKsNX9dPEu9OTSBhHp9StY-6x0Mw8j3aRpnuLPakEwX32Ol8UxlZCpMJdkZEaK0ZTNLN42bJdyxFhbuON0LH+-W-edwaesL1F+feg==

...should get the result I think you are looking for...
2/
...also don't hesistate to use another MAV in series with the first instance, that's worked for me as well on occasion. In my experience MAV is one of the most misunderstood plugs out there ...hth.../s~

[Tagirijus]

Basically, it is the lookahead causing this. (look-ahead-quote)

Thanks for the quote. I re-read the MVA manual part. I might have missunderstood the technical background a bit, but not totally, hm.

1. Do you really need lookahead on? It might be possible to just use the speed to get it right.

I imagined the lookahead as a look into the future for MVA to better adjust the gain, since it knows how to gain before the parts comes. Practical it's to me like avoiding weird pumping effects, if the sound has a huge dynamic change and MVA had to react to this in realtime - such pumping can only avoided by lookahead, or am I wrong?

2. Why is the Input gain 0db? I explained the importance of setting that first.

Because I set the input gain in the MTurboComp befor MVA - sorry, I should have mentioned this ...

3. Could a workaround be to modulate the speed parameter with a follower mod?

If speed gets modulated, weird pitch effects come up.

...using your mp3 source, but without the preceding compression, try these...

I also tried something like this. I followd your example as well, but unfortunately it pumps even more than in my first MVA try mentioned here. Hm.

...also don't hesistate to use another MAV in series with the first instance, that's worked for me as well on occasion. In my experience MAV is one of the most misunderstood plugs out there ...hth.../s~

This is an interesting idea as well.


A question besides: do you people use MVA for mixing voiceovers and do you get good and natural results out of it? If so: even with VOs, who have a huge dynamic range?