Stereo Tool

Bojcha: I just discovered that your suggestion of using a Slope Overlap of 100% (even without the bug that's currently there) causes horrible artifacts for difficult sounds like xylophones. So it should be set lower (at 80% most of the artifacts are gone, and that's using a latency of 256 which is used internally for some of the filters - but it still sounds considerably worse than my original filter). This probably means that the maximum values are 90% for 512, 95% for 1024, 97.5% for 2048 and 98.75 for 4096. BUT - at these values MOST of the artifacts are gone, not all of them, so I would suggest using considerably lower values to stay on the safe side.

This also means that some of the CPU usage improvements that I made aren't working properly and cause the sound (well, some sounds anyway) to deteriorate quite horribly. So I have no new version for tonight (I think I know how to fix it though, so I hope have a new version tomorrow. I had the CPU load down a few % before I discovered this problem, I think I can still reach that same CPU load with no effect on the quality.

I skipped some versions, but the beta3 i downloaded on jan 7 had much more bass in it than the beta3 i downloaded today. The high frequencies are alot more present which seems to push the rest away. Could be my unprofessional ears but still :p

Probably caused by the 'block shape' problem I mentioned in my previous post. I've just compared the last uploaded version with the version of january 5th, and at very low frequencies (below 30 Hz) it did indeed have about 2 dB lower bass levels - and some 'noise' around 0 Hz. I'm now building a new version where the frequency response is nearly identical (+/- 0.1 dB) to that of the january 5th version - will be online in an hour.

Edit: Where did you get that really cool avatar? It could almost have been a member of this family:



(See the 2nd one - but yours is cooler)

ok i am confused...
Where is Slope Overlap good ? 80 or 98.75 or ....

New version is available. (BETA3)

I *know* that low values (10%, 20%) are good. And I know that 50% is better than 0% (because these are the values that I've measured). However, if you're approaching 100%, instead of a nice gradual slope, the slopes become very steep. This basically corresponds to just 'cutting' the audio at some point which - as you probably know - causes a click.

The processing then thinks that that click (all frequencies!) is part of the sound, and frequency analyses fail. The higher you set this value, the stronger this effect will occur (because the sound caused by it gets closer to a click).


If the slopes are big, you can set Slope Overlap (IN PERCENT!) steeper because the cutoff won't be that sharp. The smaller the slopes get (which happens when the latency is lower), the more problematic high Slope Overlap values become. Basically, for every /2 latency reduction, 100-Slope Overlap must double to get the same steepness.


Note: I'm currently thinking about another solution that won't need Slope Overlap at all anymore, so maybe it will disappear in the next version.

ok clearer now .. then, nice place for slope overlap and slider can be removed. But maybe not same place for all latency values ?

i just compared Bass! (standalone normal output) beta3 (08.jan) and beta3 (10.jan)..
And bass is the same .. (?)

But is it also the same as in v4.22? (According to my measurements, it is almost the same)

Also, it may depend on the settings...

Deeper bass is a bit better in v4.22 .. very small difference. about 35-40Hz and under. At more then 35-40hz it's same as latest beta3.

That could be caused by the highpass filter - it's less precise than before due to the lower latency. (Assuming that you have the highpass filter turned on). Can probably be fixed by setting the first equalizer slider of the Multiband window a bit higher.