Stereo Tool

that's it ..
quick test, and things are much better not only with this kind of songs but all!!
Only for better detecting High dynamic songs (lower rms power) and for High clipped songs (higher rms power) is needed better peek detecting .. so higher rms power songs will stay down .. and songs with more dynamics will go up

For program material that has low dynamic range (or high RMS and low peak levels), there will be more activity in
the AGC sections and less activity in the Limiters. This behavior is due to the AGC sections response to high RMS
energy, while the Limiters are reacting to lower peak energy.

Cool!

About RMS: I had something like that in the dualband version, but I was treating highs and lows differently... Right now everything is just thrown together so it's very likely that the algorithm doesn't do its job anymore.

Also, the 'remove remaining peaks' filter might cause artifacts because it expects that the lows and highs stay below certain levels. For lows it's probably fine, but highs might very well be (continuously) louder than expected in some cases - which could lead to some distortion.

True, but this is quick test for 1-band prelimiter ... i'll continue testing next few days.
Now Multiband, Final Limiter and Loudness act much different .. so i gues that tewaks is needed there too.

I think that setting the 'soft limit' levels of the pre limiter a bit higher should help a lot (the output level of the pre limiter is probably a bit lower than before, and because it doesn't do any high/low equalization, multiband will have to act a bit stronger, which can also be achieved by increasing the pre limiter's 'soft limit' levels.

Plan for the RMS filter:
1. I first need to fix the singleband compressor/limiter. This is basically the same filter as the multiband filter, but a bit simpler (no splitting in bands). ■ DONE, uploaded as BETA3
2. Then I'll convert it to RMS. That's a bit problematic, because for that I need to look at pieces of sound of about 50 ms. So either the response (drop in volume) will not occur exactly when a peak occurs in the sound (and in that case I should drop the volume afterwards, not before!), OR I need to find a way to 'estimate' the RMS volume of a single sample. I think I have an idea for that, but I'll have to try how well it works.
3. Then, if I have an 'RMS estimate' for each sample, I can use that to do exactly what I'm doing now. But with a better RMS estimate.

Trick to calculate the estimated RMS volume of a sample:
- For a block of data, calculate the RMS volume
- Also calculate the average peak level
- Calculate RMS / AvgPeak. The result of this is higher for more compressed input sound
- For each sample, to determine whether compression is needed, multiply it by the RMS/AvgPeak value.

Some peaks may be way too high, but since the response is very slow if it's just a few peaks it won't matter much for the volume that comes out.


After that I'll add the AGC and Multiband gating.

--> BETA3: Singleband is working now (and uploaded)

Instead of using an RMS estimate I'm using a simpler algorithm now: I just calculate the RMS of a piece of sound, relative to the highest peak level. Then I compress as if the sound is louder (or softer) based on the RMS value. Doesn't work too bad - almost everything ends up within 1 dB of each other - but I do get a lot more pumping so I do need to improve it a bit. (Or just set the compression a bit less agressive, I'm now just using some random values and it's probably a lot faster than the original dualband limiter.)

Hi Hans,
Here is another example song where level Drop is not caused by Pre-Limiter much but on multiband more!
Atention to part where is voice is unbeliveble high level (1:07 - 1:11)
http://www30.zippyshare.com/v/14304946/file.html
I know that production of this song is ultra bad .. but i tryed other processors and this Level-drop is not happening.
Best to notice this phenomena is to play without Loudness.

Wooh - ultra bad is an understatement!

But, I don't hear what you're describing at all. *See EDIT below!* And when I look at the Multiband sliders, I see that during this really loud voice the 700 Hz band volume drops a lot, but the others stay more or less constant (which is what I would expect - note however that in a case like this the 350 and 1400 Hz bands will drop a lot as well. It's not visible in the display but it does happen.).

Which setting are you using?


Edit: At which setting did you set "limit output"? That's basically a 'smart' singleband compressor (it limits the maximum volume of all the bands combined, by looking at which bands contribute most to high peak levels). If I set this at a low value, or its down speed slider at a high value, I do clearly hear what you're describing... If you set "limit output" to 100% it's disabled, and the effect should be gone.

I play/test it with "Web radio Soft Compression" preset without Loudness... Singleband Prelimiter - first two sliders to 25.00.

Yes, 700Hz band drops as should.. but whole level is droped. Looking at Breakaway RTA Scope, i clearly see whole level is down in that moment. It's better when i up Final Limiter to Make-it-up .. but i must put it to 1.50 and then it goes to hardRed (especially when FM output is used) which again is not good.
When i turn ON Loudnes seems it recover it a lot but .. what about when is Loudness is not needed. (as for my stream).

edit1:
I just put dsp ST default v4.22 to try fix it from SingleBand Compressor.. and seems it works very well, but little pumped, maybe i need to fine tune it .. ..

edit2:
Tuned! Almost ideal! i think this singleband comrpessor/limiter somehow need to be little smarter. tested over 100 songs of different genres .. and anyway much better.
btw: beta3 and beta3a are so better! i'l' wait singleband compressor/limiter fix and "gate"

edit3:
oh, i see Singleband is already fixed.