Stereo Tool

I mentioned AudioProc's AGC only because it have "fixed" eq detection in sidechan, same what i proposed for ST AGC, just adjustable. Same thing is in BBP! Sidechain EQ.
I never tryed AudioProc's multiband.

Michi, just try this. Play track from Abba - Gimme, Gimme. and Alex Gaudino - I'm in love (vocal_edit)
And tell me .. does this two track sounds same loud, same good, for your "home" listening?

1. I don't have these and I don't want these,
because I am not a radio guy

2. Your question is unprecise (with or without Stereo Tool ? And if with what preset ?) and seems to me like a rhetorical question.

3. It would be much easier if you simply say what you dislike while comparing these two tracks under a (your !) concrete circumstance of listening (and its conjunction to Stereo Tool).

4. I do not deny the problems you have described.
And my previous post was not meant as a negative reaction to your ideas.
But it is not my special interest,
because I am not a radio guy
I do confess that if I listen to radio it is most often news, political or cultural background information, concerts or very special music - but very seldom standard pop/rock/...../...../..... commercial radio broadcasts.
I really can't stand music with very heavy compression broadcasted or mastered on CDs.
Maybe it is a problem, because I am musician and I know the true definition of a natural dynamic impact of real music.
So, if you want I am here the black sheep under the white (radio) ones.

nc

But you have it (Audioproc) ?
And so you can look and say if I am right that everything (extra options) you have asked for (ratio, threshold, etc.), that Hans should add to Stereo Tool's multiband is inside Audioproc's multiband compression module ?

I ask again, because I think that this could help Hans somehow to improve Stereo Tool's multiband the way you want it.

Bojcha, this forum and the progress of Stereo Tool development is based on our discussions here (and sometimes especially our controversies).
Though I do not think that we really disagree (if you had the impression then it is a misunderstanding).

I think that everybody that found his way to Stereo Tool and this forum is very happy, because the things (Stereo Tool) is getting better and better.
And almost everybody can contribute to this progress.
It (Stereo Tool) might be not perfect.
But almost every day you can feel the change (speculate if Hans has build a new Beta ) and everybody knows that you are a very important reason for the improvements during the last year.

I like this place and I like the people here (especially Hans and Bojcha).
We are on the way .....

Ok people, don't start fighting ok


Now after reading all this I think I'll add what I *think* (and please correct me if I'm wrong, as I have no experience with these other devices).

The current Stereo Tool Multiband filter, with 10 bands
1 Positive: Brings out details in the sound
2 Positive: In cases of extremely compressed input material, can somewhat restore dynamics in the audio
3 Positive: Does not easily cause pumping due to the large number of bands.
4 Negative: Output signal still contains big spikes (problematic for Loudness)
5 Negative: Probably does not do the best job at keeping the total audio at a constant level (since its working on 10 separate bands).
6 Negative: Certain sounds get reduced too much (loud women voices)

The same Multiband filter, reduced to - say - 3 bands
1 would be more or less lost
2 too
3 will cause far more pumping than currently
4 would be better
5 would be better
6 would be better, but may also cause other sounds to be reduced more which would sound annoying.
Clipping is not possible with this setup!

The current AGC filter, expected behavior with 3 bands
(Note: This is not the current AGC set to '3 bands', because the bands are locked such that it behaves as a single band AGC unless the input gets very extreme. I'm talking about 3 completely separate bands here - which by the way CAN be configured using the current AGC... I will do some tests to see if I can confirm what I'm writing here...)
1 would be more or less lost
2 too
3 could contain a bit more pumping, but probably not that much (see current AGC behavior); might be worse though at more aggressive settings.
4 would be better (especially with low 'remove remaining peaks' levels, also note that for Loudness RMS levels are FAR more important than peak levels, and this would directly respond to those RMS levels)
5 would be better, especially for the 'apparent' (RMS) level - so also better than the Multiband-3 solution
6 would be better, and probably also better than for the Multiband-3 solution.
Also, due to the look-ahead, no clipping would be needed (although I'm not really sure if this is good or bad, for getting a sound with strong 'punch')


An alternative would be to have BOTH filters implemented, so Multiband + 3-band AGC or vice versa. That might be the solution to combine the positive things of both filters.


Note: I have not mentioned things like ratio, knee, hold times etc, yet, but adding those (in case of hold times, making them configurable) to Multiband might also make things a lot more powerful.


What will probably also help: The AGC responds identically at all input levels - so increase the volume by 10 dB and you'll still get the same output. For Multiband this is NOT the case - but it probably should be.

I have tried to improve that in the past but I failed at the time - actually that's BECAUSE Multiband doesn't use RMS values but only single sample values - I will have to look at it again. Changing this will also make the audio more constant in volume, and probably increase the audio quality.
--> Made this bold because after thinking about it again I think this might be the single biggest issue in the current Multiband implementation, even more important than knee, hold times and all kinds of stuff like that.

Short explanation of the issue to make things clearer:
- Say I have a loud spike, such that the volume is dropped to -6 dB.
- Then I get softer sounds. The volume rises back to 0 dB, say with 1 dB / 10 ms.
- Next, I have a much louder spike, and the volume is dropped to -60 dB.
- Then I get softer sounds. The volume rises back to 0 dB, with 10 dB / 10 ms. --> COMPLETELY DIFFERENT.
Now, this is good because quickly raising the level after a loud peak reduces pumping.
But it's also bad because for continuous loud sounds the output will be FAR more compressed and 'dense', which also makes it difficult to configure Multiband correctly for all types of sound.

The ideal solution would be to check the 'wanted' volume directly after the loud sound: If the volume immediately drops to 0, it's perfect that the volume is increased very rapidly to avoid pumping. But if it stays loud, the audio should come out as in the -6 dB example.

Now the issue in the current Multiband implementation is that I'm filtering based on single sample values, and each time when the waveform passes '0' it seems that the volume drops to 0... Some RMS like level would help a lot.


This is why I'm currently trying to keep the Multiband input level as low as possible (using AGC target output level) - if this issue would be solved, the AGC would be far less important for the output sound.

Ok, I've been running some tests with the 2-band AGC at pretty extreme settings and I hear the following differences:
- Multiband (10 bands) causes very loud audio to sound a bit 'distorted' (that is because very loud sounds get reduced and the rest remains, there's no real distortion taking place but it sounds somewhat similar). AGC (2 bands) causes the audio to sound very dense and compressed, but without seeming distorted. The output level is very well controlled.

Actually, 2 bands AGC sounds like some FM stations (which' sound I don't really like, but they do have a specific recognizable 'sound image' that really stands out).

To the point:
- More bands = more open sound
- Less bands = denser, 'typical, recognizable' sound
- More bands = no need for AGC, peak level works fine
- Less bands = AGC needed for better results

Using a single band (in the Loudness window, turning AGC style on or off, listening to unprocessed input), AGC sounds cleaner (which makes sense since it looks ahead).

...and I thought my higher-order highpass was controversial...

Based on the responses in this thread and some measurements that I did I think I can now conclude that there are 2 important issues (there are more, but these seem to have the biggest impact on the sound):
- AGC has an issue with lower input levels. This causes tracks with big dynamic content to come out upto 2.5 dB softer than they should! (Note: Pre Amp of 10 is still considered 'lower' here...)
- Multiband issue described earlier.

After reading all this about the AGC and multiband issue I think I agree on just about all your points Hans.

The RMS level of the sound is definitely more important than the peak. I think any one can "compress" a sound with any compressor to the level they choose, but the tricky part which I think everyone looks for is a consistent volume output (whether they want 0db or -2db).

I also agree, the threshold, release, etc. settings I think are not that important. After trying many other compressors/limiters/etc. that only offered those settings, it's not enough to make a great sound and I don't think having those settings implemented into ST would make that much of a difference. I say this because--ST NOT having those options available to tweak has been able to create a sound no other audio processor (software) has been able to achieve. The big name brands that create those expensive VST plugins that cost $400 and up can't come close to what ST can produce.

As I said, I don't know the answer to the AGC vs Multiband debate BUT I do think that one cannot co-exist without the other when making that smooth, volume consistent, clean, non-pumping sound.

Also on a little side note, there is a little GUI problem in ST. Not a big deal but something I noticed.



If I come up with any ideas as to the AGC/Multiband issue, I'll be sure to add my 2 cents

All I know is nothing I've used or seen being used has gotten me the quality of sound that ST has to this date. I may not have the ear to tell the differences you guys are talking about, but maybe thats a good thing. The people who are listening to these streams whether through FM or WEB can't tell the differences either when they are so small. I mean they can tell if something sounds good or bad, and 99.9% of my listeners tell me that it sounds great on a wide variety of sound systems. Are most of these adjustments being made affecting FM broadcasting or everything? Reason I ask is I only do web broadcasting and I only use 128kb and 64kb AAC+. I try to hear these imperfections that people claim are there, for the life of me I can't. I play every kind of genre of music out there all mixed together randomly, and ST make each song sound great with constant levels across the board.

So Hans. You have an amazing product, and the fact that you are constantly making things better is what sets you way ahead of the others out there. Keep up all the good work, but don't work too hard. Remember there is no such thing as perfect sound. As long as the listeners think it sounds terrific, then I think thats good enough.

-n4-

Also I'd pay for a registered copy, but I like to advertise for you. So I don't mind when you come on once and a while and say "this sound is brought to you buy ST...." You should add an option to the paid version to allow your advertisement in there at a custom point maybe. I like people to know that I am using Stereo Tool to get the great sound they are hearing.