Stereo Tool

Pfftt.
I can load your preset without any issues but I cannot load Wes Keenes properly ???

Since you brought up the CPU, I do want to say that my CPU does support SSE2 and AVX2. I should not be prone to SIMD instruction issues.

Also, I like your Stereotool theme Bojcha.

Tnx, but don't make me talk about theme and all that
CPU: I did't brought anything, Hans said that.
Wes's preset, just load and set MB output level to atleast -7 or lower.

_________________
control point
control point2

Dude. Chill out...

Have you tried tweaking or even disabling the dynamic release speedup at all?

_________________
Imaging producer & Stereo Tool audio processing consultant

Asking me? Is your cpu bad too ?

_________________
control point
control point2

I understand you are frustrated.

I am frustrated too because the changing of dials doesn't change anything for the weird sound that we hear.

Can you direct your energy toward providing constructive criticism that contributes to fixing the problem?

I am trying to provide feedback myself - I am not experienced with this. All I can say is "It sounds quirky" and it is overwhelming for anyone to try to fix issues.

I've worked with SIMD instructions before and it is very likely that something weird to happen for a specific CPU feature. It makes sense with what Hans asked in the moment.

Hopefully to help a little bit: Summarize

1. I gave a preset for feedback because I was still having the "wimpy" (the hard to describe) which is making me frustrated... I sadly agree here.
2. Switching to a new preset changes the action of the multiband compressor video here im bad at explaining
3. The downside to the new compressor (noted from Bojcha) is that it's hard to manage the quick-release and its hard to tell what it is doing.
   • I think personality, there should be a way to tell what things are causing things to release? Am not sure. This is sleepy thoughts.

Time for me to go to sleep since it is currently 1:21 am.

Your bass level going into the clipper is too loud.

My main reason for suggesting that ABDP can be turned off is this: Yesterday a nearby FM station switched over to this new compressor. I've listened to it for about an hour on my car radio, which is extremely sensitive to bass levels - if the bass is just slightly too loud it will start to distort horribly. From more or less clean sound to unlistenable is typically 2-3 steps on the volume setting. What I noticed yesterday when listening to this station is that for the entire hour, the threshold where bass starts to distort was always in the same range of 3 volume steps, regardless of whether they were playing 70's or current music (they play a mix of everything). Low bass caused more distortion than midbass, so they might be using ITU1770, or they might be using 2 MB bass bands, or my car radio might be more sensitive to that, but that appeared to be most of the reason for the 3 dB range. So my thought was: If the bass is this consistent, then there's probaby a bass level that's always safe to send into the clipper without being too low, and if you find that level it might be possible to turn ABDP off completely. With the old compressors the bass was never consistent enough to do that - unless you lowered the bass level a lot.
I guess I need to remove this setting from the list of deprecated settings. We're hiding it if you're in Analog mode, because that was the new default.
The main thing that I see is that you're going into the clipper really hard. Even if I lower the clipper drive by 16 dB the output still peaks to 0 dB, although I don't see the clipper meter move anymore.

The other thing that I just discovered (someone mentioned that a few days ago here, but so far I hadn't seen it, but your preset triggers it) is that Intellilimiter doesn't appear to work in peak mode, and peak mode generally seems to be completely broken. I have not seen this before with other presets, so I'm going to check what's wrong. But it clearly does not work as intended. If I switch Intellilimiter to RMS mode it suddenly does a lot to your bass at -6 dB, but peak mode normally does more. And it seems to do nothing at all.


Just to clarify: The first recording of sine waves that you sent made no sense at all, it looked really weird, and when I tried your preset with your test signal on my end it looked normal. I still have no idea what happened there, and it could very well point at some compiler optimization issue. It has happened before, with massive effects: A while ago someone reported that they constantly had up to 5 dB overshoots in the clipper. They sent me the preset and everything was fine on my end. Eventually it turned out that there was a place where we take the minimum of 2 values in the code. And for some inexplicable reason, the code path for that specific CPU completely skipped that, causing massive overshoots.

So it's not unprecedented. And I still have no idea how your original waveform could look as it did - there were places in the middle of a tone where the level suddenly went up by several dBs, then back down, and then up again. I cannot explain that at all, there's nothing in the algorithm that should be capable of doing that. And the same input signal and preset looked perfectly flat when I ran it on my pc. Since this is still unexplained, I'm still not convinced that your system does what I expect it to do.


What am I seeing here exactly? Of course switching between presets with different MB settings will give different MB behavior - I'm not sure what I'm looking at.

Im not being very specific. On preset change, the multiband glitches and displays that it is compressing really hard (36 of actuation) which is changed after restarting Stereotool. It looks like the input to the compressor like +400db for something.

Thank you so much for your feedback Hans. I really appreciate it.
I'm trying my best here. Hopefully that was okay with you.

After reading back I'm a little confused. Are you experiencing the bug I am talking about? When I reduce the clipper to -6, the output is -6dbs. I try to keep my presets to about 0db of output from the compressors before the clipper unless you are not using my preset???

I just tried the posted preset and it's not hitting the clipper super hard on my end either (beta 101 win10x64). Could this be an example of a CPU computational difference?

I wonder if the plugin or Raspberry Pi versions act any different.

I did find a bug when I was testing this preset though. I, of course, unconventionally use the Simple Clipper as a loudness multiplier I control externally (in addition to the Advanced Clipper).
-When disabling the advanced clipper to enable the simple clipper so I could control the loudness multiplier, the sound became more muffled upon enabling the simple clipper.
-And when I reenabled the advanced, it became clear again.
-Disabling Advanced made it muffled until I disabled the Simple clipper again.

So the simple clipper seems to be making things sound off while it's the only clipper enabled.

Ok, found it. If the RMS size is set to 0, the peak mode data isn't filled in at all. It needs to be set at least to 1 (higher values for better behavior!). I've now fixed that in the code (so if you set it to 0 it will behave as if it's set to 1); also I've made the RMS size sliders enabled in peak mode, since they do have an effect (see the original blog post for why).

OH! And I also have no idea what everything did with that size set to 0. That was supported by the old compressors, but not by the new ones. So the behavior was almost certainly wrong. Aside from this, Sudden Drop is almost contantly active in your preset - that should be extremely rare, basically only if a song ends or the level drops a lot.

With this fixed, output levels of MB1 are about 10 dB quieter than before. So this has a massive effect.


@Bojcha: Your RMS size is set to 40 ms, that's probably way too much. It will make the compressor start to go down before the actual spike, which sounds bad. I recommend to keep it below 20 ms, even lower for higher frequency bands.


This is now more or less confirmed. The station that I mentioned did a quick test and they said that there's some crackle sometimes on some bass kicks, but sustained bass is fine, and this can probably be fixed by tweaking Intellilimiter (and beside that, they tested without changing the bass levels, and without ABDP there's more bass in the output, so the bass threshold in the MB can probably be reduced). Before this update, not using ABDP was completely impossible.

Wes' preset uses MB1 also as AGC and drives it really hard. So seeing something like this is kinda normal, and there's still audio coming out so everything looks fine to me. However, it should behave exactly the same after a restart if you stay on that preset - if it doesn't then something would indeed be wrong.

I did notice that sometimes when switching between presets I get an assert in the code, so something doesn't like preset switching. Haven't found what it is yet (unfortunately the assert occurs far later, in the clipper, but the source is probably the multiband).