Stereo Tool

The only thing I know (see in its GUI) is that the Breakaway Rustonium preset (default settings) uses ultra fast AGC down speed (at position 3:28).

Whether it's fast or slow, the net result there is that there is less of a volume change than what is present in the original, unprocessed track. There's about 20 decibels of change in the unprocessed track, and a maximum of maybe 8 decibels in what Bojcha posted. To me, this indicates that most likely that preset will have less Dynamic Range, thus trading a perceived "better" handling of this specific situation for a generally accepted "worse" handling of the dynamics of the audio. The only way that might not be true is if that AGC does more "look-ahead" than StereoTool's AGC.

Apologise for my intrusion. Found the on-going discussion intriguing.
Now, I have not tested the soundtrack in subject. But in my vain attempt to strike consistency that other hardware processors exhibit, I am leaning to infer that an AGC with large look-ahead might not present the solution by itself.
That also begs the question - do other processors use RMS based multiband processing? For if that's the case, things might be different I guess.

_________________
visit website

I can give you one hint for MB...
Try more "LIMIT"mode with very slow speeds.(.0001). Also make sure that you adjust "clippers" good, especially for first 2 bands (mostly first).

The limiting can end up decreasing the dynamic range. I keep bringing this up, because I think your perception of what is "good" is more in line with the concepts of the "loudness war", which is a higher total RMS, but a lower dynamic range.

Also, what comes in strong there is not necessarily the first 2 bands. This is what I keep trying to mention to michi. Percussion extends across the sound spectrum. What needs to be clipped is the crash / attack range of percussion, which is anywhere from 2.5K clear on up to 10K... It is not a coincidence that I have a stronger clip at 5.5K

well.. what "orbans" do.
But actually not quite true.
Since ST multiband on compress mode sounds "better" in same tine there is big rms and also peek loss in some tracks since ratio is bit strange, also "Up speed" curve is bit strange also
I also mentioned "very low speeds"!
I am actually testing it. Almost full 'limit mode' in MB. - You can check it - http://goo.gl/Lh4xN
What TT says?

I'd have to get the VST bridge working. Writing the stream out via disk writer, I can try that, I suppose...

Nope. It specifically says [Sorry cannot write streams to disk]

I think what I'm getting at is I have a fundamental disagreement with the desire to limit the dynamic range. Yes, it might limit bursting, but I'd rather listen to a more dynamic sound than a less dynamic sound.

I think the best approach to this individual track is work on clipping the crash / attack range from 2.5K - 10K (or 11K), rather than subject all tracks to a reduction in dynamic range.

I didn't miss this comment before, it was just that I was tinkering.

I think this is a valid point. We know that multiband in ST is currently peak-based, and that Hans has mentioned an interest in changing it to RMS-based, or at least testing RMS-based. It would be interesting to know what type of multiband is in use for the comparisons, because if they're not the same, then direct comparisons are invalid.

Sorry, again you use a wrong definition of limiting (based on sciolism).
Limiting = very high ratio + very high threshold.
And this does not decrease a dynamic range more than traditional compression (low ratio + low threshold) ! Only very high ratio + low threshold decreases a dynamic range dramatically.
But this is not called limiting - this is ultra compression.

The question is what happens (what is possible) in Stereo Tool`s multiband ?
There are no special sliders for ratio and threshold.
But the Compress-Limit slider combines somehow both values in one (ratio and threshold).

The actually characteristic is also dependent on the input level (the AGC output).
The AGC output acts more or less like an offset for the threshold.

->
1. Press Reset
2. Turn Multiband on
3. Play some tracks of music
4. While you listen drag the slider Compression-Limit back and forth
5. Can you hear the difference ?

But why can Brian measure (better ?) higher DR values with the TT DR Offline Meter ?
Because his presets (the additional processing besides multiband) have the goal to create an output with relative low maximal RMS.

And the internal weighting of the TT DR Offline Meter prefers these low RMS levels.
Technically it calculates correctly (based on 0 dB as maximal output level), but practically it fails to identify the perceived loudness (the loudness a human ear hears - virtually extending far beyond the 0 dB digital barrier) and the relative differences between quieter parts (intros, solos, etc.) and the maximal loudness of high energy parts.

Besides that important auxiliary condition, the measured DR value contains no information about the dynamic structure on the timeline.
It only calculates sums.
Theoretically a processing could invert the original dynamic balance between quieter parts (intros, solos, etc.) and high energy parts - but the TT DR Offline Meter would measure the same DR value.

Starting from scratch (-> Reset) I agree that compression sounds better than limiting characteristic, because the range of reaction with compression for individual frequency bands is much broader than with limiting.
With compression you have better (much simpler) control of the spectral structure.

But for my taste the usage of compression characteristic in Stereo Tool`s Multiband leads to a dynamic mismatch with many presets.
Originally quieter parts (intros, solos, etc.) get XL-size and high energy parts get S-size.
That characteristic might be useful for music in the elevator/departement stores or low/medium volume via headphones.
But with speakers and loud volume settings (the way music is meant to be played - as close as possible to a live audition) IMO that characteristic is not suitable.

I've been making some changes here and there, and am noticing that the AGC probably needs to be looked at more than Multiband. Some time ago I mentioned a problem with female vocals. It also impacts male vocals, particularly if there are heavy guitars going on at the same time as the vocals. I had isolated it to somewhere between 800 and 4K. The thing is, if I turn AGC off and just rely on multiband and final limiting's volume control, things are much better, at least in my admittedly hobbyist opinion.