Stereo Tool

I wasn't asleep yet - new version is available as BETA2.

Tomorrow I'll first try to fix the dualband AGC issue, then I'll continue with the Multiband CPU load.

Dualband AGC pre-echo is fixed
Multiband CPU load did indeed not (or hardly) go up

Something seems to be wrong with the Hard Limit filter though. The CPU load is going up a lot if I turn it on with more aggressive settings (so I've configured it less aggressively for now, but that reduces the output volume). Will check that tomorrow.

I'll upload the new version later tonight.

Uploaded as BETA2 again.

Pooh, this was far more difficult than I had expected.

I've been looking into the "CPU load is much higher when Hard Limit is enabled" problem. As it turned out, if I switched between different latency values, the CPU load varied wildly, sometimes it even became twice as high.

This was caused by memory locations that were used - which were (in part) allocated dynamically when needed. So, depending on what you did, the memory addresses could be different.

Now here's something really annoying: In caches of Intel (probably also AMD) CPU's, if two blocks that are used at the same time are a multiple of 64 kB apart, they cannot both be placed in the cache. And if that happens the processing (at that point) will be may times slower (in the order of 50 times!). Which seems to be what happened (at least to some of the filters in Stereo Tool) - which isn't surprising if you know that a block of memory that holds 4096 samples is exactly 32 kB in size...


Since everything was allocated dynamically, this happened sometimes, and I couldn't control it. So now I've allocated everything statically at start up. On top of that, I had to re-organize the allocations to reach higher speeds. (And that was basically trial and error, with 4 different latencies AND for each of them with and without display to check).

Anyway, it looks like I have now found a setup that works, and gives the best possible performance for each setting. While I was looking into the memory usage anyway, I've reduced the memory usage, which means that Stereo Tool now uses less memory and probably that it starts up faster.


Will post an update in an hour.

Uploaded as BETA2 again:
Winamp DSP plugin: http://www.stereotool.com/download/dsp_ ... _BETA2.exe
Stand alone version: http://www.stereotool.com/download/ster ... _BETA2.exe
VST version: http://www.stereotool.com/download/vst_ ... _BETA2.dll

Sound is identical (except that I restored the more aggressive hard limit settings from 2 versions ago), CPU load might be (depending on the actions that you perform) a bit lower.

Hi Hans...
last few days I have listened last beta(s), AGC processing mostly, and i found that second band in AGC is not good in some/many cases.
For some cases is lifesaver (like song i post erlier) .. but for most is not.
I have many songs with voices stronger than music and when voice with higher level comes "second AGC band" lowers whole band after bass more than should. So i started to use only first band.
I am thinking now, should removing second band in Prelimiter was good idea. Maybe now will work better as new AGC.

I notice another thing with "Highs". Mostly, loud "S" for few milliseconds lowers whole band in multiband not only that particular high band and sound goes down. Sometimes that can be seen in multiband's meters. i tryed to set clippers for highs lower but no help.
Another things with highs is they are distorted in too many songs. i checked whole my chain Studio-to-transmitter. From players, cables, link (switched to 48kHz PCM wav), even today i replace transmitter with backup one. i just hear same thing. Highs knows to be distorted. It's not classic distortion.. its more like late reaction of some processing. Don't know is it multiband or limiter or Loudness .. today i have doubt in Hard Limit filter. Really don't know exactly.
Loudness is at 2.00. If i push it more .. distortion is frequently more occured, but again if i turn off multiband and level output to be as should, distortion is less even when loudness is >x2.30.
Strange is i am hearing this distortion on some tuners/receivers and on some not, and i know i have exactly 75kHz output.

Vibrating voices: curently i fix (not 100%) that by lowering first 3 bands in multiband.

One thing i did not try, is to try ST org.v4.22. - and i'll do that tomorrow.

edit:
new beta2standalone not working, output dead fm/normal, loudness dead, final-limiter dead. @ 4096 !
512, 1024, 2048 works, sometimes

edit2:
After "reset" and starts to work .. but strange things like RDS output is 2.8 not 4.5%, 38kHz is at about 41kHz .. clicks clicks ... weird all !

About the latest version:
I just started the stand alone version and I do indeed see weird things. For example, I see 2 RDS signals, one is inside the stereo signal range (which probably explains some of the weird sounds you're hearing). And I see 2 stereo signals! I don't know if this is a problem in the signal itself or just in the code that measures it - but something is definitely wrong. I think the Winamp version is fine, other versions might not be.

My Juli@ driver in Vista64 is broken, which makes it difficult to test...


Edit: FM Calibration is broken, also in the Winamp version. Causes the same effects that I was describing above.

About the 's' distortion:
In the Winamp plugin version, I've checked the output and it looks perfect (I've recorded the output and upsampled it to 4*176400 = 705600 Hz): No spikes, and sound between the bands (between 15 and 19, 19 and 23 and 53 and 54 Hz) is between -80 and 90 dB. So that doesn't explain distortion.

Question: Are you using "FM Calibration"? If so, if some output levels are set higher than the center value, make sure that the sound card output level is set lower (below 50% is always safe), otherwise you might get distortion at the sound card already. But that wouldn't explain that you don't hear distortion on all receivers.

Note that if there are problems in the stand alone version that are not present in the Winamp version (which might be indicated by the fact that weird things are happening in the latest version, I would expect older versions to work properly but maybe they also don't), that might also cause this.

About the volume drop in Multiband:
Multiband should be identical to older versions (4.22 and probably even back to 3.xx).
I'll check tomorrow to be sure.

About dualband AGC:
I'll see if I can find a solution for that. First idea: Don't allow the 2nd band to drop below the level of the first. (Does it drop below that level when you hear what you're describing, based on the output level display?)

- I am using FM Callibration but only small for phase difference (0.007499993) all bands
- For transmitter i only use Standalone version so i did't even try on my PC yet. Yes, actually most of time is same as first band. If, let's say, louder voice from song comes in, second band goes down a bit. and sound from 200Hz to 16kHz is down. Strange is that multiband is not recovering it.

For Highs distortion ...
One thing i forgot to check, before all i must get this songs from studio to check it without processing .. i just found one that has distortion already (rip from CD) blah.

Even if you only use a phase difference you should still set the output level to the sound card a bit lower.

Tonight I'll try to fix the things that are broken in the last version.

like this .. ?