Stereo Tool

Thanks Bojcha! This fix worked perfect. I just went into Winamp EQ, turned it on, and set preamp to -1. No more red/clipping in the input! Thanks man!

Thank you! Disabling the removal of remaining track ends fixed it! I already tried changing latency and the CPU usage slider, but according to Task Manager the whole foobar2000+Stereo Tool 5.00 Beta combo uses max. 20% CPU on my Intel E5200 with the 'Web Radio (Stereo)' preset while displaying the main window. Also thought about it's the Continuator plugin which i'm using for radio like track end detection and crossfading causing it, but even if i remove all other DSP's in foobar's settings, stutter was still there. So i'm pretty happy now. I don't even need this function because continuator already makes my tracks gapless before your DSP even knows about it Still, a native foobar plugin will be nice later when you have some time for it! (and i still have ideas like Linux and inter-application audio input-output support with eg. JACK into the standalone version )

Thank you again for the help! Keep up the good work

The Windows command line version can already stream using stdin and stdout. Unfortunately there's no Linux build yet because there are too many Windows API calls thoughout in the code. (The good news here is that I'm currently working on a fix for VST plugins, that - among others - requires me to split those API calls from the rest of the code. Which means that I'm at least moving towards a Linux-compilable version now)

I've tried the command line version already on Linux with Wine (or wineconsole? i don't remember), it was okay. I could pipe in a wav file from the shell then write the result into an another. Although it works, i don't know how stable it'll be with continuous streaming. Also as far as i remember there are no switches to tell Stereo Tool the sampling rate, bit depth and channel number for raw PCM input samples.

By the way, is the command line version supports registering or it's just always stays a demo? The last makes sense because i don't think to many of us pipes through a file which is longer than 12 hours But with continuous streaming this can be easily achieved.

I don't know if it's worth it, but maybe you can try to compile Stereo Tool with winelib. http://www.winehq.org/docs/winelib-guid ... LIB-WHATIS I guess with this you can even leave the graphical frontend intact. I didn't dig into it too much, i've searched for some kind of shared memory support so i can write a little Win32 gateway for the Winamp DSP version of Stereo Tool into the Linux domain, but didn't found anything promising. Maybe i could use TCP to connect to a JACK audio server but i found it overkill for this, and i'm sort on time because of my diploma work.

The best Linux version would be a standalone version with PulseAudio and JACK support, and (i like this more) a shared library which i can call from my code, optionally without requirement for an X11 display. Well, dreams

I've added an item to my todo list: When sound is removed, I'll make the audio configuration button blink to hint at the cause.

The command line version needs an STS file, if that's a registered STS file it will run in registered mode. (Actually, currently it ALWAYS runs in registered mode....... Don't tell anyone!)

I measure the units using very specific software. Latest beta has 1030 units, compared to 4.22 that has 1001 units in use.

Just for the records - an Core i7 980X cpu, overclocked to 4,4 GHz can do around 8000 Units per core and around 5000 Units per HT core. An old P3 can do around 900-1300 units. So in short Stereotool uses 1/8 of a CPU core or around 20%. Same goes for the BETA, using a preset "FM Stereo" from 4.22 loaded into it.

Dispite of that i would really recommend you to add an Spectrum analyzer graph, as big as the "bars" and refreshing that fast.

By spectrum analyzer I mean something like this with many graphs:

Hi Eurodancetest,

Turn on FM processing and "Direct Sound Card Output" and you'll get a spectrum analyzer

About the processing: For the BETA's on an i7 I can imagine that these numbers are correct, but version 4.22 usually uses a whole lot more (see the responses of other people).

My own performance measurement is: Enable "wave output" in Winamp, then process data and see how fast it runs.(*) v4.22 uses 2 cores, the BETA's use only 1, still BETA's run at 173% of the speed of v4.22.


(*) Make sure that the disk is not fragmented, has enough free space, and no other programs are accessing the disk. That's for precise measurements of course, not needed for comparing v4.22 against the BETA's.

For spectrum analyzer I meant for you to add a seperate one that shows and analyzes in detail the sound from 1 hz up to the samplerate ot the song (usually 44.1 or 48 khz, that's 22050hz and 24000hz per channel)

The one with the FM output is too small and analyzes the band up to 192khz.

I did this "test" over a 15k SAS drive (not fragmented!) and I processed an 3 minutes file for 1:32 min with 4.22 and for 1:28 with 5.00Beta

CPU : AMD Phenom 9950, no SSE4a (i'm testing over debian Linux with wine) !

I'll update you shortly for the results over windows xp.

Hi Hans,
i don't know what happen with last 5.0 betas .. i think something is not good with HighPass Filter...
25Hz tone is not killed even with never-phase-linear Highpass at 35Hz.

Hi Bojcha, how are you testing this?

I've checked this, and what I see is: (for a 25 Hz tone, "never phase linear" filter)

Filtering at 25 Hz: About 90-95% of the tone is still there
Filtering at 30 Hz: About 6-7% of the tone is still there
Filtering at 35 Hz: About 1% of the tone is still there
Filtering at 40 Hz: Less than 0.2% of the tone is still there

Note: If you're processing just a single tone, Pre Amp, Bass Boost and Loudness can increase the volume level of this tone! Even if only 1% remains, this be amplified a lot (Pre Amp x40, Bass Boost x1.57, Loudness x4.00 would still cause a level of 251% of the input level.

In older BETA's I only supported highpass filtering at 20, 30 and 40 Hz. So in-between levels were rounded to these values. In the latest few BETA's, I've supported every multiply of 5 Hz (5, 10, 15, 20, 25, 30, 35, 40, ...). So it's possible that previously when selecting 35 Hz you were actually filtering at 40 Hz, leaving less than 1/5th of the 25 Hz tone that remains now the filtering is done at 35 Hz.

Could this be what you're seeing, or is there really something wrong (and if so, how do you see it - because I don't....)?