Stereo Tool

New version is available as BETA3.

Changes:
1. I found a bug that turned out to be the reason for the very high CPU load in the last BETA3A versions. I've implemented a preliminary fix, the CPU load is a lot lower already. But I'm not sure what the change did to the sound quality. (should now be closer to v4.22 than before, but I don't know if that's a good thing).

2. IMPORTANT The function of the "Slope Overlap" filter has changed. It does NOT reduce the CPU load anymore.
Please use the "Slope Steepness" sliders instead (lowering them has EXACTLY the same effect as increasing Slope Overlap in older versions; Slope Overlap 20% equals Slope Steepness 80% for both slopes).

"Slope Overlap" DOES have an effect on the sound, but not on the CPU load. It now changes the shape of the slopes. Setting it a bit higher reduces Slope Steepness artifacts by a few dB. I don't know yet what the ideal version for this slider is - and if it's always the same value or if it depends on the Slope Steepness and/or Latency setting. When I do I'll just use it always and remove the slider.

(What I do know: If you set Slope Overlap to 100-Slope Steepness, the artifacts are ALWAYS reduced compared to no Slope Overlap.)

In lower latency modes, using a bit of Slope Overlap might improve the sound quality, even if Slope Steepness is set to 100% (this is also true for higher latency modes, but I think you won't notice the difference there).

NEW version:
Winamp DSP plugin: http://www.stereotool.com/download/dsp_ ... _BETA3.exe
Stand alone version: http://www.stereotool.com/download/ster ... _BETA3.exe
VST version: http://www.stereotool.com/download/vst_ ... _BETA3.dll

Previous version:
Winamp DSP plugin: http://www.stereotool.com/download/dsp_ ... BETA3A.exe
Stand alone version: http://www.stereotool.com/download/ster ... BETA3A.exe
VST version: http://www.stereotool.com/download/vst_ ... BETA3A.dll

That's odd. I've just tried your settings here, and I don't get hiccups, so I don't think it's the processing (do you see the hiccups in the output waveform?).

You might want to try the new BETA3 I've just uploaded, it has a much lower CPU load. I don't really see how that would help but who knows...

I tried the new version, cpu usage is indeed much lower, and I don't hear much difference in sound quality!

Changing the first 3 sliders vs cpu usage:
18 % cpu usage when sliders are left
20 % sliders about 1/4th
26 % sliders about 1/2
34 % sliders about 3/4th
45 % sliders to the right

No buffer underflows, and I can bring the output buffer size back to 0.03 s without underflows

Also the stuttering is gone in this version... The only stuttering that happens, is when all the sliders (including the 4th) are slided to the right. But I guess that's not a bug?

now i have stutter when Slope Overlap is on full right side .. but no stutter on Slope Overlap 99% just one step to left!
I'll post what i found one by one

- MPX output is again 100.0% and not 100.6% (by some peeks) in beta3A
- LUKE is right - There is small peek on 38kHz. like reflection from pilot or rds (guessing)
- Slope Overlap - not good on 100% (full right) but on 95% good and much les artifacts in lowlatency mode.
And there is no cpu usage changes (left or right) so 95% is good
- Block overlap still good at 25% which is good for minimum best quality. (maybe 30%), this slider should stay in ST because who wants and can to use it at 100% where evry little artifact is not audible.
- and one bad thing: I put it on transmitter (cpu e5200) and no less cpu usage!
Still 50-55% if i rise Block overlap 60% and stuttering starts .. same as BETA3A.

Also, in the 'as much as needed' setting, when i lower the upper frequency of the filter to below 10 khz i hear nothing. The other settings are fine.

So:
- MPX output is better? (I really don't understand why...)
- Will check the 38 kHz peak. Are you using Stereo Boost? It can't be reflection - but it might be caused by an offset in the stereo signal. Does this happen always, or mainly when low bass sounds are present?
- Slope Overlap @ 100% is probably causing a division by 0 or something like that. Which is not the case at lower settings. Will check that.
- Agree with your comment about Block Overlap.
- NO less CPU usage?! I really don't understand that - before I was processing some data twice, and now I'm not anymore. Or are you using the "extra latency" mode - there nothing is changed...
- Note that with Slope Overlap you should be able to lower the Slope Steepness setting with MUCH less effect on the sound than before (and I measured it with a Slope Overlap of 10% only).

Do you hear a difference in sound quality (4096 setting, no Slope Overlap) between the two versions? So I can choose which of the two I'm going to use. (CPU load is not an issue, I can reach approximately the same level for both processing methods).

Kewl. Note about the sliders: The 1st and 2nd slider should stay further to the right, the 3rd slider doesn't need to be set high (20-30% should suffice, higher might be better but the difference is VERY small).

"Don't hear much difference" - do you have a preference for one of the two? I think I can achieve the same lower CPU load for both versions.

By the way - really, 95%?! Is that the best value? Then I've been wrong all along - in that case I don't need a 'smooth' window over my processing block but almost a rectangle! I had expected that such high values would cause much more artifacts, instead of less... (I haven't had time to test it myself yet, will get back on this).

Plese check, specialy lowlatency mode .. from 0 to 100 best sound is on ~95 .. 100 will be best actualy but it stutters.
And bass rings less too.

- 38kHz peek .. i save picture of it ..and there is a peek but now no. i don't know what happen.
- cpu usage.. i'm little confused about all .. what is "extra latency mode" ? i use ..
*4096
*slider 1/2 80%
*slider 3 25-40%
*slider 4 ~95%.
Which i found minimum best quality vs cpu usage.

"Extra latency" = part of the Loudness processing is performed using whatever latency is needed. I made this when I thought it wasn't possible to get the v4.22 quality using only 4096 samples latency - but I think I can remove it now (but on the other hand, the CPU load is lower so I could keep it as a 'best quality, lower CPU usage" mode).

With slider 4 at 95%, you can probably set slider 1/2 lower (80% corresponds to the image you posted before, but the purpose of slider 4 is to reduce the artifacts that occur when slider 1/2 are set lower).