Stereo Tool

I can just reapet myself from erlier, to answer as short as possible:

AGC:
This new Bass AGC is full opposite of what i suggested erlier. What i sugegsted is Broadband AGC (aka parametric EQ for AGC)
Simply explenation what should this do is, adjust on what freqs will AGC react. Why is current bassAGC opposite? Because AGC should have much less respond on lowest bass, more respond bass, mid-bass, mid and midhighs. Why? Because with simple AGC some track can easily trick AGC to work at wrong reduction/amplifying, just because track have just a bit stronger bass but good dynamic.
Then after AGC like this some MB AGC (MB pure compressor) is welcome.


1, 2, 3 and 4. - Soon or later everybody gets in that problem. All plugins that we ever used (limiters compressors .. ) uses, more or less, same scheme. Words like, ratio, knee, milliseconds, Attack release, threshold, dB, VU meters all have rulers with clear measurements. When it comes to ST you must forget all this and start from scratch.

5. huh.. not so important actually, question is how much bands are enough, It's not if it's too much, and not good if ther are less. Find the universal number and stick to it. There are many processors - we don't need to think too much
Even Omnia3 can sound good with 3-band compressor/limiter

6. Yes!

hans, stand alone download link does not work...so I cant test. fixy .


SORRY...JUST SEEN NOT AVAILABLE. MY BAD.

Hm, this 2nd AGC was only intended to protect Loudness (and at low latencies also Multiband - at higher latencies I think I should place it after Multiband) against bass artifacts. And it does seem to help - a lot actually, especially on lower latencies. It is NOT intended to change the 'coloring' of the audio (actually you should try to configure it such that it has minimal impact on that).

You should see it as a 'de-basser' like there's also a 'de-esser'. And it's placed before Multiband to also reduce Multiband artifacts.

I think Multiband really needs a big redesign - if I compare a single band AGC to the singleband compressor (Loudness&Limiting window) AGC does a much better job. Multiband uses the 'singleband compression' algorithm for each band - so there's a lot to gain there, either by doing what Bojcha (and now also Phoenix) suggests, or even by replacing the processing by the type of processing I'm doing in the AGC.

But that was NOT the intention of this change.

Why? The statement is so counter-intuitive by itself. Be it some $30 plugin, $300 mastering suite or DAW or a quarter-million dollar sound mastering hardware, MB dynamics is based on the same principles with proprietary technologies thrown in. So why discard the basics? How could you possibly start from 'scratch' if there is no common groundwork on which all MB processing based. Stereo Tool does not employ some alien technology that you forget the basics that are out right from the bible of sound processing, unless you are trying to present this opinion that entire calibration of MB is merely playing with fancy knobs or sliders in trial and error fashion endlessly. Do understand this - not every user has trained musical ears or has experience of over 20-30 years in music industry that they can immediately identify the signature to look for, as when their aural sense supersedes their visual sense. Reason why time and again some newbie throws this question at Hans as to how to work with MB is not that the documentation is unclear but because the functionality vs usability of MB is so mediocre as compared to clear intelligibility of other software/hardware based MB processors.
Well if Hans decides to rigidly quantify the number of bands so be it...I've no qualms about it. But do remember there are other users(like Michi) who use it for processing their soundfiles. Moreover, forum members draw comparisons with BBP. Recall that some presets use 5 bands and some use 7. Unless Hans thinks it's high time to roll out different versions targeting different audiences it's wise to stick to 10 band design except leveraging user flexibility to activate/de-activate certain bands. Moreover I reinstated the need for Band specific Q because of the fact that bands are working on Center-frequencies. Ideally it's be good to implement Cross-over frequencies then we probably won't need band specific Q.
Right now say if I want to work with 7 bands, so I set the last 3 to insanely high values to disable them. Doing so the effect of EQ or compressor of even the last(10th) band(say set to 250000) affects the 7th band if using default global Q=0.30. Now only if I change steepness to very high value I can minimize the effect, but then I don't want such high steepness for other bands which are functional. Hence band specific Q should be employed when working with Center-frequencies as opposed to X-over frequencies.

_________________
visit website

Indeed this was the first thing i ran into problems with when i started using stereo tool. Please change it.

Hans,
I don't test each new beta like others here but I do have one, long overdue comment. The "Warm & Smooth" preset is badly made. It either needs to be updated or removed. When I load it (since version 5.xx) and even now in 6.10, it ALWAYS distorts and gets red bars no matter what.
For the sake of new users of ST, I think it needs to be removed or fixed.

Multiband is the only filter (except for Stereo Image, which is hardly used anymore) that is basically unchanged since Stereo Tool 1.0.

Ok, I started with 8 bands and I added some filtering and clipping, but the basic behavior never changed.

At the time when I created this filter, I didn't know a lot about audio processing. As far as I can see Multiband suffers from the following issues:
- Behavior is different at different input levels. So after a big drop (due to a loud sound) the volume rises at a much higher rate than after a smaller drop. This might actually be good...
- Multiband is PEAK based instead of RMS based. This is less relevant because it works on small frequency bands, but changing it might improve the behavior anyway (peak level is always bad, since a small peak can have big effects). This issue can be alleviated a bit by setting the 'down speed' lower, but:
- Multiband does not look ahead. So when a loud peak starts, the volume gets dropped DURING that loud peak. Setting the 'down speed' lower makes the peak stand out more (Clipping removes it above a certain level), setting it higher causes a 'flatter' sound.

All these things are fixed in the (single band) AGC filter (just compare it to Singleband compression, at the lowest latency setting, and you'll notice the difference - and I don't have the impression that the AGC reduces 'punch', even though it does look ahead and reduces the level before a peak starts). That's why I mentioned replacing the current processing by AGC processing (but: That would again lead to a very non-standard behavior. Which might be good. Or not.)

Anyway, much more will need to be done than just improving the GUI.


Note: Using AGC (ST implementation)-like compression instead might have another big benefit: No clipping is required, so there's no need anymore to use 'single octave' frequency bands. And AGC seems to 'pump' less than other filters, so I could use less bands, which would give a more constant 'total output' RMS level before Loudness, which in turn should help Loudness (which is RMS level based, higher = less quality, so more constant = higher total output level without more quality effects). Less bands (esp. at bass frequencies) also reduces artifacts - hence reduces the need for the 'steepness' filter.

Having said all this, having less bands also means less control over bass frequencies...

I never mention RMS based MB bands. I think nobody mention that.
But if that will fix output to be more constant, and if that can replace REAL compesion/lookahead limiting per band, then ok.

Constant otput is much depend on how will AGC work too.
Current behavior of main AGC is very good. Just is needed that EQ in sidechain to be tuned for High compresset vs high dynamic track.

Yes, but current bass bands are OK. So i would not touch it. Just less bands in midhighs/highs and slighlty re-aranged freqs.

No, but I think (expect) that it will sound better (more constant).

See this image:

From: "Mina04" at 0:16.6 (loud voice sound).

If I adjust the volume based on PEAK level (the peak at the center in this case, the volume will be dropped a bit at that point), I'll get different results than when I limit based on the average (RMS) level over a (slightly? - maybe configurable) larger time frame.

On the other hand:
- doing so will make it impossible to respond extremely fast.
- Very short spikes will not be reduced at all and stay very loud. Actually that's probably also the case if I lower the volume slowly (low down speeds) in the current implementation... Maybe I need a hybrid solution (maximum of RMS and 50% of current sample value or so).

I remember that during this long period of AGC enhancement discussion and Beta development Bojcha mentioned the (good or better) behaviour of Audioproc's AGC.
Now (and with Bojcha's ongoing criticism about Stereotools' multiband) I wonder what experiences Bojcha has with Audioproc's multiband ?
Audioproc has also 10 bands and besides EQ similiar pre gain I think all the additional options (threshold, ratio and even lookahead) Bojcha wants for Stereotool's multiband.

So, Bojcha what are your experiences with Audioproc's multiband ?
Is it this (something similiar), that you want for Stereotool ?

I just wonder.


My opinion:
I can imagine that for the standard purpose of (loud) broadcasting Stereotool's actual multiband design is not the best solution.

But for people like me (with special interests in audio restoration, re-mastering and not broadcasting) the way Stereotools' multiband works until now seems to be a good alternative to traditional multiband compressor designs (because it is so different and can work more as a dynamic EQ that can be tweaked in every detail).
I have just bought some of the 2009 Beatles (so called) remaster releases.
It's quite astonishing what details you can restore with Stereotool's multiband (in combination with noise gate) without intensification of the already present overemphasized snare, hihat and cymbal sounds.
Tweaking high center frequencies in conjunction with intelligent compression/clipping can even reduce these harsh tones while voices and instruments are enhanced.
It is more or less impossible with other tools like Magix Cleaning Lab, Diamond Cut 8 and even Izotope RX or any other advanced audio editor/plugin combination (I have tried) to do the same (especially not that easy as with a perfect tuned Stereo Tool 6.10 specialized restoration preset).
It is even possible to do it without any noticeable change of the original dynamic range structure (as it was mastered on the remaster CD's).
With heavy compression it is always simple to somehow 'enhance' old music.
But to keep the original (wide) dynamic range structure is the holy grail of serious (professional) audio restoration.