Stereo Tool

Uploaded another update - BETA3A. HARD LIMIT is now almost finished (far less artifacts, and much better peak control). There are still some artifacts left, I'll have to see if I can get rid of those as well...

OOPS! BIG MISTAKE! I had some test code still turned on in that last version - so it sounds really crappy. I'm rebuilding it now, will upload it again tomorrow morning (if the build doesn't fail), otherwise tomorrow evening.

Fixed - Found some bugs that caused artifacts. The Hard Limit filter sounds better now if FM processing is enabled.

The links again:
Winamp DSP plugin: http://www.stereotool.com/download/dsp_ ... BETA3A.exe
Stand alone version: http://www.stereotool.com/download/ster ... BETA3A.exe
VST version: http://www.stereotool.com/download/vst_ ... BETA3A.dll

Note: I've added an item to my TODO list: The bass protection filters work less good in Overlap=0 mode. Dunno why yet.

Minor update: Now in Overlap=0 mode there are NO peaks left after Hard Limit. Has no effect on the sound (basically, a few times per minute a volume drop of about 1% = 0.1 dB which lasts a few ms may occur - this removes the peaks that remained in Overlap=0 mode after Hard Limit.)

I think I do now. The highpass filter gives very different results, with phase linear filtering it filters less strong than for Overlap>0 which causes the deep lows (around 30 Hz) to be a few dB louder, and with non-phase-linear filtering, it filters at exactly the requested frequency, which causes frequencies at 32 Hz (with filter frequency set to 30 Hz) to be over 6 dB louder - and that's in the output signal, after Loudness! (It probably doesn't help that when non-phase-linear filtering is used, the filtering is done BEFORE anything else, so it's ignored also by the pre limiter - phase linear filtering is done AFTER the pre limiter).

Edit: I think I can confirm this now. Without highpass filter the difference is very small (I still think that Overlap>0 sounds slightly better, but it's very difficult to even be really sure.

I also found that - due to the much less steep filter - setting the highpass frequency to 20 Hz for Overlap=0 is closer to v4.22 30 Hz than 30 Hz is. I'll have to check if I can improve the phase linear Overlap=0 version of the highpass filter further.


Edit: Did some more tests, and v4.22 does still outperform the BETA3A w.r.t. bass distortion, even if I turn the highpass filter off. Also, BETA3 with Overlap=0 has something a bit annoying in the high frequencies (disappears when Overlap>0). I don't know what's causing it. The bass distortion problem is also present in the Final Limiter if I overdrive it, which might indicate that it's a fundamental problem of the filtering method - something that cannot be solved.....

But, if I turn the highpass filter on, the difference immediately becomes much smaller.

Note 2 to self: Test the output of all the filters UPTO Loudness combined, and check if the spectrum is identical to 4.22 (or even Overlap>0) - if possible find where it goes wrong this way...

Hm - ok here it is. In v4.22, in case of non-symmetrical bass sounds there was a DC offset which caused the peak level to be lower - which made the sound better. In v4.23, there's no such DC offset anymore - which causes spikes in one direction (hence distortion) if the bass (and probably other sounds as well) is not symmetrical.

In the sample I'm looking at right now these peaks correspond to a bass level that's 1 dB louder in v4.22 - which explains the distortion. Will have to think about a solution...

At least now I know what to look for when comparing the ~ 20-30 different filters that make up the Loudness filter in v4.22 and BETA3A.

Since solving this means introducing a 0 Hz component to the sound (which is there in v4.22!), I guess I should at least make it possible to disable it...

GOT IT! Fix is coming in about an hour - quality @ Latency 4096 is noticeably improved (less bass distortion, probably also a better sound overall). Low latency sound seems to be improved a lot more - but I have to do an A/B comparison to be sure (and it's getting too late now).

As it turns out, there was actually a but in v4.22 and all earlier versions that caused part of the Loudness filter to be skipped at 0 Hz. But as it turns out, skipping that part leads to a better sound quality. When re-implementing parts of the code for low latency, I didn't make the same mistake again...

New BETA3A with improved bass distortion handling (now similar to v4.22):

Winamp DSP plugin: http://www.stereotool.com/download/dsp_ ... BETA3A.exe
Stand alone version: http://www.stereotool.com/download/ster ... BETA3A.exe
VST version: http://www.stereotool.com/download/vst_ ... BETA3A.dll