Stereo Tool

Now I do!

- Not heavily compressed songs sound good in both versions
- Already compressed songs go 'up and down' much more in BETA, and at the start they often sound FAR too loud

And here's the cause: ITU 1770!

Songs that are compressed, are maximized in volume based on ACTUAL RMS level, not on ITU-1770 corrected RMS level. If I correct based on ITU-1770 afterwards, that gives a lot of volume increases and decreases.

So, the new mechanism should:
1. Work as 5.00 (that is, without 'rms level smoothing' AND without ITU-1770)
2. Also measure the volume based on smoothing (and ITU-1770?)
3. Then, if output level stays too loud (already compressed) for a longer period of time, lower the volume by at most 2 dB, responding VERY slowly.

More specifically,
- RMS smoothing causes a reduced response time. If a new track starts, it can be upto a few dB too loud for a second or so.
- ITU-1770 causes volume drops when loud voices kick in (!), because songs have already been optimized for non-ITU RMS volumes. (*)

So I need to turn both off, and then check the output level and - if that's a bit too low or too high - very slowly adjust the volume to correct for that.


(*) Edit: Only for the head correction part of ITU-1770. The bass reduction (which has far more impact on the sound!) does NOT seem to have too much ill effects (only 0.2 dB louder peaks at start of new track). So I can keep using the bass correction, just not the head correction.

I hope that everything will be fine

I'm building an update now.

ITU-1770 head correction is switched off for now, except for that (and actually I'm not really sure if I _should_ turn it on again) everything seems fine now.

Changes:
- First, the normal processing of version 5.00 is used (with some improvements from the BETA versions, such as gating and improved 'remove remaining peaks' etc.)
- Then, the output level is measured.
- The average output level over a period of time (configurable to upto 60 seconds, default value 10 seconds) is compared against the requested volume
- If the average output level is HIGHER than a configurable percentage (default: 70%, approx. 3 dB) of that output level, the volume is dropped.

Note: I also noticed that by lowering the 'down speed', the RMS level of compressed and not-yet-compressed songs gets closer. (That didn't work properly in version 5.00 due to a bug).

Uploaded:
Winamp DSP plugin: http://www.stereotool.com/download/dsp_ ... r_BETA.exe
Stand alone version: http://www.stereotool.com/download/ster ... r_BETA.exe
VST version: http://www.stereotool.com/download/vst_ ... l_BETA.dll
VST version (No SSE2): http://www.stereotool.com/download/vst_ ... 2_BETA.dll
Command line version: http://www.stereotool.com/download/ster ... d_BETA.exe

Press RESET to load the new default settings - for some sliders the meaning (and hence also the default value) has changed!

hi Hans,
Did you check Beyonce's song with new AGC ?

I'll reply more later must go to sleep now.

That song should be banned!

Anyway, shouldn't be too difficult to fix - the 2nd layer will have to behave similarly to the first.

Will look at it today. By the way, I'm building an update now with ITU-1770 support.


Edit: Beyonce track fixed now. Re-building.

Uploaded as BETA again.

In the previous version, the last step ('density measurement') was accidentally based on RMS level instead of ITU-1770 (bass only) level. Which has little impact, unless there's a lot of bass present.
That's solved now. The 'target sustained level' should be set about a bit lower now to get similar behavior as before (70% before -> 65% now).

Also, the band 2 density is now calculated separately.

And there's a new ITU-1770 switch; enable this for "full" ITU-1770 support instead of only the bass part of ITU-1770. (This means that highs are counted stronger).


Notes:
- The new settings are not yet saved to disk!
- In this version (and the previous as well), the volume is dropped if the average level gets too high. This results in a LOWER AGC output level. So you probably need to boost the 'target level' a lot. I will do some measurements to calibrate the levels again, so the selected level matches the output.

Much better. Beyonce: Just another torture song it was really bad in previous beta.

I still think that you still need to lower Low frequency detection little in 1st agc band by 3-4dB around 50Hz and down.

Anyway good for now.

mistake; http://www.stereotool.com/download/ - 4nd

That would deviate from the ITU-1770 proposal. And, worse: It might increase the effect that a loud voice pushes all other sounds down. Or do you hear similar (bass) effects in other songs as well? Because for now, this song doesn't sound bad, and I don't want to introduce things in all other songs just to improve this one.

TODO list:
DONE Change ITU button position
DONE Save new settings, possibly remove some old ones
DONE Correct AGC output level
DONE Fix band 3 behavior
DONE Bug fix: Too low setting causes clipping/extreme output levels?!
SKIP Select RMS / Reduced bass / ITU-1770?
- Update version number and release