Stereo Tool

I'm currently doing some tests, and I'm getting the impression that - for low latencies - I get the best results when I use a relatively steep slope and a very high block overlap (20% vs 80%). (Of course, both at 100% would give the best result; what I mean is that it's apparently better to reduce the slope width than the overlap, for low latency modes. In high latency modes it's the opposite!).

ohhh... this is awesome fix .. great job Hans .. there is new explenations for CPU usage/latency ... nice


edit:
i have feeling that overall sound from multiband with this new clipping now is much better.

If I had known that those sounds were caused by the clippers I would have (tried to) solve(d) it long ago already... But I never realized that it was caused by that!

I still need to finish the CPU usage control (the 2nd slider isn't working yet, and I want to add another slider to reduce the latency at the cost of increasing the CPU load). I think I should also add a few buttons to go to some "presets" to make things easier. And I need to do something with the Quality/Latency box in the main window.


Question about the quality: After all the recent fixes, is the quality now still lower than the quality of v4.22/Hi quality mode (normal latency)?

About quality, can i answer next year? in few minutes i'll put this beta on transmitter .... and listen awhile...

Happy new Year to all and all Best from me!

edit:
oops, my E5200 can't work at max quality and cpu load is 65% with block overlap at ~40% works. and cpu is at 55% same cpu usage as v4.22

That makes sense - 100% is really too much (in fact without any optimizations it would require twice the CPU load of v4.22!).

The best way to determine the required Block overlap value is to play a really-really-really low tone (say 20-40 Hz) and then listen if you can hear artifacts.

I noticed that with the highpass filter turned OFF, an overlap of 20% is already enough (in 4096-latency mode).

With the highpass filter, I need an overlap of 60% to get rid of all the artifacts, which is way too much. I'm looking into it.

(At 40% overlap the amount of artifacts is still very low, and probably never noticeable in normal music).

Update: With the Highpass filter enabled, I'm now getting good quality with Overlap set to 20% (better than overlap at 50% in the previous version!)

Update coming in about an hour!

New update is online. For Latency setting 4096, around 20% overlap should suffice now (higher is theoretically still better, but I don't hear any artifacts anymore).

Also, the left and right slopes can now be configured differently if needed. (Making the left slope steeper _slightly_ increases the latency, that's not the case for the right slope).

I think I found another CPU load reduction that will have no or very little effect on the sound. Will implement it next year...

Happy new year!

New idea isn't going to work....

no, actually something is better and something is not...
what i can hear Loudness is better in 4.22, and after 2.20x cpu load growing.
I am listening transmitter.. and again peeks not goin' to 100% as in 4.22.
Overall ... in my car i can hear that hole band from 1kHz and down is lower, or 1khz and up is higher, but when i measure all is OK! And somehow Highs are stronger and higher but not precise and clear...
Whole Dynamic is different .. so for multiband i must to set "attack" from 00005 to 00003 ..

i feel loudness is doing all this... i keep it on 2.00x now i set it ti 1.50x to keep sound clearer and prelimiter is up now to maintain level.

BUT, all this with so small latency .. respect.

Still testing and comparing..

edit:
Just checked stream with latest beta3 too...
Sounds just good! and maybe better then 4.22 .. but i don't use Loudness! only other difference is that on stream "Block Overlap" is at 50% and on transmitter is 20%, i can't push it more.

one minor bug from 4.22 and beta3:
Meters in "Stereo Boost" know to freeze after some time ... 1 day, half day or few hours.