Stereo Tool

(Transport: TODO list from page 80):

What remains for version 5.10:
- Disable separate pre-emphasis and relaxed clipping when FM processing is enabled. It's currently too confusing. ■ DONE
- Remove remaining spikes -> behavior is incorrect, it should partially maintain spikes that are really big (= difficult to get rid of). ---> Looking into this I found that I need to do something very different, and I also need to do that for normal (non-web based) processing: ■ DONE
- Add 'maximum mirror reflection' and make it configurable. ■ DONE
- Add an extra slower Hard Limit phase, or let the volume come up slower when the peaks get really loud, to avoid bass artifacts in xylophone sounds. Easy to do where hard_limit_volume[] is used. ■ DONE
- Also fix the singleband behavior (AGC style on/off). ■ DONE, default is OFF again, built-in presets with Loudness get ON
- Fix Loudness distortion caused by louder bass and lower highs. ■ DONE
- Use 'large file mode' for command line version. ■ BUSY, waiting for reply ■ DONE
- Fix hangups for switching between Singleband AGC/non-AGC switching. ■ DONE
- Save all new settings to disk (INI and VST): 'maximum mirror reflection', 'multiband constant frequency', ... ■ DONE (?)
- Reduce Loudness window height a bit. ■ DONE (?)
- Fix Punch artifacts ■ DONE
- Fix 'dirty punch' due to different handling of the very deep lows and deep lows. -> Must follow Lowpass filter settings. ■ DONE
- Attempt to make bass filtering below 90% more strict in Loudness ■ DONE
- Enforce limiting strictness even if Punch is enabled. ■ DONE
- (Optional) add FM hiss filter - removes FM hiss from input signal, useful for re-transmitting ■ DONE
- Fix Very deep bass distortion protection artifacts (or remove it) ■ DONE - not perfect but won't get better than this
- Set Allow louder lows to +10000 (or remove it) ■ DONE
- Fix settings in presets to closer match modern CD sound (less highs etc.), using the Median display. ■ BUSY
- Fix Allow louder highs behavior. Attempt to replace it by a highs AGC. And rename to de-esser. ■ DONE
- Test what happens if I (constantly) allow more distortion! ■ DONE
- Move strictness to lowest freqs only? Currently audio is also kept below configured midlow level (100%) instead of below configured verylow level (90%). ■ DONE, using separate settings instead now -> Low freq area is still needed for voice vibrations!
- Check if voice vibration filter can be set much softer due to strict bass clipping. ■ DONE, no... But since it's not needed to also reduce the bass during Punch it's not too bad.
- Replace current filtering by bass clipping in each clipper step? Or at least lower the number of filtering steps. (Lower CPU usage and better "S" protection expected). ■ DONE
- Fix De-esser in 512 samples mode ■ DONE
- Disable Very Strict button when Strict is disabled ■ DONE
- Change 'Allow louder highs' name to avoid problems with older versions, add dB display, and save as real value. ■ DONE
- Loading and saving new setting in STS file ■ DONE
- VST: Loading and saving new setting in VST plugin ■ DONE
- VST: Add setInitialDelay() call to compensate for Stereo Tool processing delay ■ DONE
- VST: Make VST version working again on all hosts! ■ DONE
- VST: Check loading and saving of int/float parameters in VST plugin hosts ■ DONE
- VST: Bypass mode: Copy input to output when out-of-place processing is selected ■ DONE
- VST: Keep window at position (0, 0) with its specified size. Change borders. Separate VST plugin main window without close icon ■ DONE
- Display de-esser and voice vibration filter effects ■ DONE
- PRESET: Check the new presets (including Bojcha's). ■ DONE
- Check FM Calibration 20 Hz band ■ DONE, same as previous versions
- Check which 'New Loudness bass processing' sliders must be kept, and place them at correct location. ■ DONE
- Max highs on loud the parameter is not saved ■ DONE
- Move identical presets to a separate function. (All Gtxx's, Bojcha's, ...) ■ FORGET IT
- Fix Final Limiter diff ■ DONE
- Add Loudness diff ■ NO
- Check audio (bass processing) in lower latency modes ■ DONE
- Loudness CPU load is 29% higher than in version 5.02, needs to be lowered. 1/3rd of that is Strict bass filtering, 2/3rds is Voice vibration filtering. ■ BUSY, now 27% - Back to 29% - Now at 26%; TODO: 1. Add more efficient 2/3 band filtering (2*FFT + CLIP + ADD + 2*FFT + CLIP + CLIP + ADD + CLIP --> 2*FFT + 1*FFT + CLIPCLIPCLIPADDCLIP_LOOP). 2. Check if 1st Loudness call is really needed for non-strict bass processing. - Now at 24% (21% without 1st Loudness call). - TOTAL CPU load is only 6% higher without Final Limiter ■ DONE
- Fix voice vibration filter in 512 sample latency mode ■ DONE
- Fix dynamic Loudness input level behavior of Punch ■ DONE
- PRESET: Update FM presets for louder highs ■ DONE
- PRESET: Check and add good new presets (also in the 'share your settings' forum). ■ BUSY (5 added, 1 to go (maybe), name Gtxx preset must be changed, non-webradio (generic) must be added); dubstep party presets for FM don't turn on FM! Probably remove Vamprecords (too much changed, no longer needed), Check O-84 AGC up speed; fix Phoenix' BBC Radio 1 preset name
- VST: Check audio issues (muffled, different AGC up speed). ■ DONE (for now)
- Release 5.10
- UPDATE DOCUMENTATION

Later:
- Protect Multiband against steep valleys
- 192 kHz FM output
- 400 Hz tone, Bessel null tones? [eldoradofm]
- Stable scope (see viewtopic.php?f=14&t=2726&start=10 . Probably solution is to find the loudest tone, with boost for the lows, and then lock to that).
- Cleaner FM hiss filter output (currently some artifacts remain, I think they can easily be removed)
- AGC rolloff slider for Michi - see viewtopic.php?f=10&t=2712&p=7667#p7667

Even with 44.1 kHz I hear (and see) differences.
Maybe it depends on the processing intensity and the caracter of the source ?

Please, Hans check it with my TranS-EnhanCe 1.6 HiFi Full DR100 preset I have just uploaded to 'Share your Stereo Tool settings'.
Try older (with a wide dynamic range) music with this preset and compare VST versus DSP !
viewtopic.php?f=10&t=2792

Michi:
I have processed the same track in both the VST and Winamp plugin, in the same host (one with VST-to-DSP bridge). I've stored both output waveforms and subtracted one from the other, and they are identical except for some rounding differences (the difference is at -94 dB).

I did get more differences at first, that turned out to be because the Quality slider was set differently (in the VST it was still at its default level of 80%, in DSP I had moved it to 100% a long time ago. And this slider is NOT part of the preset!)

At different sample rates I do see issues, which are apparently caused by the fact that the bridge that I use doesn't support passing the frequency to my plugin! (Apparently it always gets '44100' as answer to that question). It does work properly in other VST hosts, so it must be a problem of the bridge in this case (in this case the one to run VSTs in Winamp, 'VST Host DSP v1.0 for Winamp [dsp_vst.dll]').

I've measured the difference using 'FM Loud Europe (Bass)', because with that setting nearly every filter is turned on and is used aggressively (ANY difference causes big differences in the output).


Edit: Hint: There's currently no way of seeing the current sampling rate in Stereo Tool, but the following trick works:
- Enable FM processing, but NOT 'Direct sound card access'.
- Now if the input sampling rate is > 120 kHz, Stereo Tool will display a frequency analyzer next to the waveform in the main window. If it thinks the sampling rate is below that, it won't.

With due respect Hans, but I'd beg to differ on this account. I really have no clue how to actually compare 2 waveforms(or subtract one from another), but the difference from the 2 plugins with same settings at 44.1 KHz is so perceptible that I can take a blind test and tell one sound from the other(as much as I can tell the sound played through Winamp vs sound played from Adobe Audition). And this is not a miniscule one like comparing the quality of one beta version of ST with another!
Here again lot of things probably come in to picture. Most importantly the kind of soundtrack being tested with. Please check with one which is not very loud and has clear vocals. The reason being when I play old tracks(pop/rock/instrumental) the difference is 'so' there as against the modern tracks (clipped CD/mastering). And when you check it with a preset like FM Europe(BassBoost)/Extreme, the sound has gone through such an extensive chain of processing/filtering, it's like you're MAKING the sound from the 2 plugins 'sound' the same. I checked with preamp+MB+little loudness. The difference is obvious.

Lemme see if I can upload sample tracks encoded from both plugins at the same settings. My internet connection has been very intermittent since last night.

_________________
visit website

Phoenix: There must be something else that I'm missing then. Maybe even different behavior in different hosts? Like I said, in the comparison I did the difference level was at -94 dB which is extremely low (just rounding on the last bit or so); there's no way you can hear that.

But maybe other hosts do something different (the ones that I tested didn't, and I used an old track with huge dynamics for testing - not that that would make any difference if the diff level is around -94 dB).


Some recordings might indeed help to pinpoint what's going on. Please make sure that you use the LATEST beta! Older ones could have issues with sample rates etc.

Phoenix/Michi: Is it possible that the differences are caused by the programs that you use to send the audio?

For example:
- Winamp is usually 16 bits, VST by default 32.
- Different players have different MP3 decoders, which sound very different
Such differences are small, but Stereo Tool, especially at aggressive settings, will exaggerate them.

For the most reliable test:
- Use a .WAV file, without compression, as source. (Winamp can handle 16, 24 or 32(float) bit .WAV files, so you can use any of these).
- Write the output directly to a .WAV file, so don't go through the sound card.
- Then play both output .WAV files using the same player.

If I do this using the setup I described above, both .WAV files are identical but for the last 1-2 bits. (Average difference level -94 dB).

This sounds reasonable, but ....
But in practice it is getting difficult to tell any difference (to hear or to watch the GUI meters) with such a heavy processing at all.

Please try it with my TranS-EnhanCe HiFi Full DR100 preset !
And something like Duran Duran - Wild Boys or whatever you have (as long as you use my preset !).
I don't think so.
As I already have written it happens in Wavelab and with BA Live.
Concrete:
I can use three instances of Stereo Tool at the same time:
One DSP Beta70A hosted with BA Live (FX slot 1).
One DSP Beta70A hosted with the Acon chainer using xlutop's DSPtoVST wrapper (Acon chainer FX slot 1).
The Acon chainer VST (wrapped by Cakewalk VST Adapter to DX) is hosted with VB FX-4 (winamp DSP version) and this is hosted with BA Live in FX slot 2.
(You don't need the Cakewalk VST Adapter $ and can use the native Acon DX version instead !).
And finally native VST Beta70A loaded in Acon chainer FX slot 2.
So I can use any player I want (winamp, foobar, J.River or whatever).
And using my TranS-EnhanCe HiFi Full DR100 preset (of course all quality sliders up and same sample latency) the two DSP versions sound almost the same (I cannot hear or see a difference).
But the VST version output is obvious different.
It is so obvious, that I agree that is not necessary to measure it by comparing written WAVs bit by bit (though I understand your problem Hans, that you must measure this difference somehow before any further steps - the question is: difference ? Or no difference ?).
But even my sister notice it (and she is really a stupid girl in other situations of life).
Maybe not with FM Loud Europe (Bass) preset, but with my preset (DR100 !) it is there !

So, if it would be only me who complains.
But phoenix agrees that there is this difference.
Where is Bojcha when you need him ?
Bojcha ?
Difference VST versus DSP or no difference ?
@ all
Try it especially with my DR100 preset !

Thanks Hans- this is getting very exciting. I've been following this thread for several weeks and hoping that the ST experts would provide some recommended "starter" presets for the new version. Very glad to hear that the presets are coming soon, and can't wait to try them.

Just did it, same result as before: Difference between Winamp and VST version is now even less (at -96 dB).

I've also tried it with ACON, but my DSP/VST bridge doesn't work when I select ACON (I get silence). So I coudn't test that.

I think it must have something to do with some (?) VST hosts (definitely not all, because I'm hearing the sound without any difference with the DSP/VST bridge) - but I have no idea what it can be.

Hi

About my preset, yes it's outdated. ST loudness changed in last 20 betas. Beside that i have new settings.
Tonight I'll finish new one, viewtopic.php?f=10&t=2712