Stereo Tool

Wow. You might also want to check Ocen Audio.

Leif's Gap Protection wouldn't do much for non-FM, because the level of the highs is a lot lower without pre-emphasis. The issue with loud frequencies is that they kill all other sounds. For loud bass to kill the rest that sounds somewhat natural, if a loud S kills all other sounds that sounds really really annoying.

I don't really know how much latency got added, and I'm actually not really sure if it's necessary. There have been some people who had some issues with audio glitches with Wasapi or Kernel Streaming (I don't remember which of the two, but on some systems only one of the two functioned properly). The thing is, we also just updated to a new PortAudio version, and in the change log it lists a Wasapi glitches fix in some modes, so it might be ok to use what we used before. Anyway, PortAudio gives us 2 values, which can be system dependent, one for low latency and one for high latency use. We used the low latency version before, and now the high latency. There are much bigger things that contribute to latency than this, so I doubt it will be noticeable.

O, and BETA004 is posted. The curves for Leif's Highs Gap are slightly smoother now, and I have added a settings that gives a small extra priority to highs when other sounds are very loud, which might reduce highs IMD distortion. The difference in audio level is small (0.1 dB or so), so it's probably safe to turn it on.

I maded one test tone contain 3 sine tones 100Hz, 1kHz and 10kHz. Now pink noise looks decent, but when i push clipper with this 3 tones it always gives priority to 10kHz and very quickly pushes bass down even without ADBP OFF and Bass protection off.. That's why i asked what is pushing bass down. Only thing that brings it back is 'Highs/Rest Threshold" under 150%. That slider is overall not good under 200 and now i'm scratching my head.

_________________
control point
control point2

Tones are not good to test this. For tones, after clipping the remaining RMS value can be a lot higher than it can be for noise, when clipping at the same threshold. The impact that high (or any, actually) frequencies have on the total audio after wideband clipping is based largely on their RMS levels.

Having said that, it shouldn't specifically push the bass down, so I'll check why that happens. Highs/Rest threshold should affect *all* non-highs audio, not specifically bass. (But, since in music bass tends to be much louder than mids, bass also tends to be affected more).

With 3 identical tones at 100 Hz, 1000 Hz and 10000 Hz, that can't happen so it has to be something else - the 100 and 1000 Hz tones should be affected equally.


Edit: Doesn't reproduce here. 100 and 1000 Hz come out at roughly the same levels. There is something weird: If I jump in time in the file the levels move by upto a few dB. But on average they are the same.

Ok, nothing wrong, it's just interference between the 100, 1000 and 10000 Hz tones and the 19000 Hz pilot. So, this won't happen with music, just with tones.

I have also noticed that "High vs Rest" didn't do anything in non-composite mode. Turns out it was hard coded to a value that corresponds to about 180% in non-composite mode. I've fixed it for non-legacy mode.

Demodulated, all default in clipper. Abaut 5dB in clipper bar.

_________________
control point
control point2

This is similar to what I see, but when I restart the stereo coder it can be vice versa, or identical. For an honest comparison, set the pilot and RDS levels to 0, that removes most of the difference.

Enabling ABDP does have this effect, even if I set the bass frequencies to 0 and levels to 200%, and increasing "High priority" there makes it worse. So there's something left in ABDP - but without that I don't see this at all anymore with pilot and RDS at 0, or without composite clipping.

the previous highest setting for de-esser was 7.23dB. now the highest setting is 7dB.
manually entering 7.23dB value into the STS file shows the de-esser to be completely off but with the new highest default value adjustable by the slider of 7dB, de-esser is not completely turned off.

Yes i saw pilot interfere with this 3 tones, and yes i disabled both pilot and rds for test. More interesting is that there is much difference with different clipper level. so at 1dB clipper bar is kinda still ok.. at 5 bars is like on that picture.. but at 6 or more bars mids and some bass is comming back. And yes something in ADBP is doing this and its not any settings we have. Highs priority yes a bit, but it's not only that. Btw, "Leif's gap" off, is totaly vice-versa. Without composite things looks ok.
And to be clear, im testing all this because i am trying to find easiest way to make clipper as much as possible balanced over spectrum and response for given loudness level(s).

_________________
control point
control point2

I found what's causing it, it's something that should not have such a big effect on the bass. I think (haven't tested yet) that these tones are kinda a worst-case situation for it, but aside from that, this behavior is affecting way higher frequencies than it needs to. It's a filter that protects the highs against IMD, and because I never expected it to do a lot it actually affects the bass upto around 220 Hz. Oddly, in earlier tests, this filter actually *increased* the amount of bass. But that was with music, not with test tones.

Edit: It's actually audible upto at least 200 Hz... but far less than at lower frequencies so I can probably at least use a different response curve.

Edit: I've done more tests. For low frequencies (< 100 Hz) the effect isn't strong enough, they typically need to be reduced by another 1-2 dB to make room for highs. For higher frequencies (120-250 Hz) the effect is too strong. Some more tweaking is needed, mainly to avoid reducing the bass more than needed if the level of the highs and bass is not very high - currently the algorithm starts to pull back when the input gets louder and that's actually the opposite of what it should do.

Sorry Hans, stil no go. But atlest now it shows an error. Last version to work was 9.62.


Regards

MrDonT