Stereo Tool

BETA031 is available.

Really pissed now so I keep it short. Over 200 of our presets failed verification with 9.30. Just measured SB2 with peak detection for now but attack curves are off somehow, release is fine though.
I heard a major sound change first with beta 27 which does make sense: It's probably the change of RMS size behaviour where you screwed up the peak detection as a side effect. (I tested the RRD switch already but it doesn't change anything.) So back to 9.23.

I think I found out what has happened.

First, as far as I can see, Peak is not broken. If I load settings in 9.23 and then go to 9.30, I get the same behavior. No changes at all. (The other way around CAN be different though!).

But the RMS size behavior for RMS based compressors is indeed different. I thought at first that it was aways better, and it probably still is, but the levels vary too much to just update without changes...

Any chance to copy "Test Signals" to "calibration" section?

_________________
control point
control point2

Main issue was that the RMS size of the Auto EQ was set to a default value (20 ms), while it was 0 before. For some tracks this could lead to upto 4 dB of difference in levels coming out of Auto EQ. Set it to 0 to get the old behavior back; I'm releasing a new version tomorrow that does this automatically when loading an old preset.

Beside that there are 2 other things that are different between 9.23 and 9.30:
- RMS behavior fix. In all tests that we did this was an improvement of the audio.
- RMS overreaction caused by bass fix. This was a real bug that could cause compressors to drop down without a good reason (almost randomly) when using large RMS sizes - introduced in 9.23.

This does mean that the output will be different for 9.23 and 9.30. But, except for the Auto EQ issue, the difference should only be improvments.

Hi Hans:
This problem in "Location" persist on v9.30 - 64bit - Standalone Best regards

Sorry for being so slow. Still analyzing the peak one, havn't had a look at RMS yet.
Anyhow the attack curve is somehow different and it doesn't seem there's any way to compensate for that with the given settings. So unless you want us being stuck at 9.23 we definitely need the pre-9.30 peak mode back. There are many legacy options and RMS-modes, so why not add another peak mode?

I've done multiple tests (Peak, Symmetric Peak, FB^2 on and off) and see no difference at all, so please send me your settings if you still do see a difference. I'm either missing something or it's the change in how settings are saved. (Which will cause issues if you reset the settings in 9.30 and go back to 9.23, but not if you reset the settings in 9.23 and go to 9.30).

For clarity: There's not supposed to be any change in how Peak is handled. There is no need for a legacy mode - if there's a combination of settings that's different from before that's a bug. But I've subtracted the output waveforms of 9.23 and 9.30 with SB set to Peak and got only silence.

Please just send me a preset where the two are different so I can analyse it. I don't see anything in the code that would explain any difference.

I tried to find a preset with as few changes from default as possible and found out that no changes are necessary:

Preparation with ST 9.23 Winamp/DSP:
Factory reset, turn on SB, export all settings. *No other changes necessary.*

Preparing a test signal:
1kHz sine wave, dual mono, wav, 32bit floats, samplerate = 44.1kHz or 48kHz (tested both, doesn't matter); volume factors: 0.5 for 1000 samples, then jumping to 1.0 for 1000 samples, then switching back to 0.5 for 3000 samples; (I verified that both channels are identical and the numbers in the file comply to the mathematical model by at least 7 digits;)

Then I processed the file with the command line version, both 9.23 and 9.30, using the above preset and subtracted the output signals. For convenience I divided the difference signal into 3 parts according to the volume factor samples and played them separately in loops: 1st and 3rd part are silent, but the 2nd one gives an audible clicking.

Am I doing something wrong or stupid here? If so, please tell me!

How do you subtract them? I know that there are versions of Audition where subtracting 2 (nearly) identical waveforms that hit the minimum or maximum value causes loud clicks. That's a bug... I can't think of any other reason why you would get audible clicks.