Stereo Tool

Oh!, on a different context...umh, it's difficult to opine on that as my system is pretty powerful. Consequently a surplus overhead of 4% would still not be enough to complain for now. So, perhaps, feedback from those who are running the new beta on older(less powerful rigs) counts.
I presumed it to be concerning sound quality metric which has significantly improved even for high latencies.

_________________
visit website

Ah, no, I just meant CPU usage.

But I found 3 things I can improve which will probably restore the original performance - and the last 2 will probably also (slightly) improve the sound quality!
- Multiband steepness: Moving UP can be done inside current loop, no separate loop needed
- Upsampling: Use existing UpsampleMultipleParts() call (also for FM processing, should improve Hard Limit behavior)
- Downsampling: Create Downsample() code.

I can't test Last betas as much as I wanted, because i have some problems with my tooth last days.

but, i can see/hear 3 things,

-Upsampling - 44.1*4 or 48*4 or 88.2*2, -Downsampling 192/2 or 176.4/2 or 88.2/2, now ..
Where feets 128 and 32 kHz, and what happens when this rates are used? I don't think that there is some special multiplier then round number. For me two totally unnecessarily rates.

- Multiband: For low latency this is place where most impact to sound happens! What i think is that current ST is not a solution for 1024 and 512 samples mode. Bass and SubBass are totally out of control. and constantly pushed down by "band 3" or so.
Right now I don't have ASIO soundcard to check how BA Asio works at LowLatency mode, but if i remember correctly it works just fine. "all" sound was there, even in non-phase Linear mode.

- I check bit, thing is little strange. Standalone at FM mode is about 16-20% cpu usage on my PC. All is ON, clipper, transmitter calibration ..etc.. When I press reset (where all is OFF and 192kHz rate), there is 3-6% CPU usage (even minimized). Why is that? Should be 0%?
I mean, Very strict bass clipping uses 1% of my CPU, maybe less, idle Stereo Tool uses almost 6%.

Hi Bojcha,

1. Might indeed be useless. But 32 kbit is one of the 'standard' sample rates that's used a lot (not that much anymore though), anyway IF a sound card supports it, and also 128 kbit (4*32), it can be used and the CPU load is lower.

2. Non-phase linear is actually easier. But I agree that this needs more work - anyway, it's a lot better than in 6.00 I think?

3. True, that's what I need to fix. This is downsampling + upsampling. Which is currently WAY too expensive.

Hans i see red bar on last beta on input... and i dont hear any diff on steepness like last beta...

_________________
Sexy Girls in your city

By the way: Dropping phase linearity in case of adding artifacts in case of large differences between bands might be the way to go here.

I've been thinking: None of the steps (except the stereo widener, which I could move to another place in the processing chain) that are performed before Multiband compression should cause big differences in the bass content relative to the rest of the content. (Except AGC in rare cases). Which means that it might be possible to remove the 'processing window' at this stage without much effect on the peak levels. And if that's possible I can use non-phase linear filters for the lowest Multiband bands, then do what I do now for the higher bands.

This would however disable clipping for those lower bands. (That's why I don't want to use it for ALL bands). And the separation between the bands will be extremely poor.

(Note to Bojcha: Remember the 0-latency version I created before I went on to create version 5.00? This would be a hybrid between those versions: 0-latency for bands 1-3, normal for bands 4+).

Anyway, not for version 6.01.


Question: What happens if you increase 'steepness' in low latency mode? I know there are lots more artifacts, but does it solve anything?

On input just means that your input is (potentially) too loud - it reaches the maximum values (-32768 or +32767 for 16 bits audio). So there's potentially clipping distortion on your input signal.

Actually there was a bug in the previous BETA's that caused Steepness to behave badly. But badly might actually be good (it means more difference between bands is allowed).

Which latency are you using? And can you check the multiband output display to see what Steepness does? (Output bars will display a less bright area, which is cut off by steepness).

but on last beta's i did nt had red bars with same input

_________________
Sexy Girls in your city

i think last steepnes sounds better... my opinion!
i use 2048 latency.
On steepnes below 20% i see LOW freqs affect on bars 40hz and 90hz only.... if i add steepnes no freq affect

_________________
Sexy Girls in your city