Stereo Tool

@eldoradofm: Could you send me the settings that you are using for testing? The HPF doesn't cause DC on its own, and so far I haven't seen this issue.

@phoenix: No, not yet solved. I thought I had placed of jn my treo list but I don't see it there..

Bojcha: What do you mean by not clipping the bass? That's basically all that bass boost does....

Don't know for others, me always keep "max bass peek level' to 250%.
meaning... maybe it's better to be something like bass parametric EQ, to be possible some custom bass shape after MB. In MB EQ we are limited to frequency of bands. Anyway.. maybe not time for this... keep going...

latest 110 beta standalone gives me a free robot if latency <2048


Damn, checked with Bojcha, seems like presets constantly mess with the program.
Starting from scratch.

Hans is it possible to put the HPF before the multiband?
Now it seems like it is behind it so the first band might actually be processing something that will be filtered out later, doesn't make sense to me.
Also what happens when you place the phase rotation before the AGC of just behind it, but certainly before the MB?
I was just listening to 'old' 80's records that looked absolutely not symmetrical on the scoop and the higher MB bands went too much down to my liking.

I have compared the Mika track with your settings between BETA028 and BETA101.

The difference in DC offset between the two is very small. What I did was: I processed the file with both versions, converted the result to mono, then subtracted the two from each other, and finally I lowpass filtered the remaining (difference) signal.

The *absolute* difference in DC level between the two versions is:
- Below 15 Hz: 2.5% (+1.9 kHz, peak level 76.9 kHz)
- Below 10 Hz: 2% (+1.5 kHz, peak level 76.5 kHz)
- Below 5 Hz: 1% (+0.8 kHz, peak level 75.8 kHz)
These peak levels are assuming that the sound card COMPLETELY ignores the frequencies below the mentioned frequency - in reality it should be less bad (unless bad stuff starts to happen at higher frequencies already).

There is another difference though that may have much more effect, especially if the sound card has issues at low frequencies: The signal hits the minimum and maximum level FAR more often in BETA110 than in BETA028, and for longer periods of time (multiple adjacent samples are at peak levels, this does not occur in BETA028). You can clearly hear it if you compare the two, the highs are much louder in BETA110. So in BETA028 the chance that the peak level is reached at a point where it will cause issues because there is a (minor!) DC offset is much smaller than for BETA110, and it could be that the Pira only reports overshoots that last multiple samples.

Anyway, there is only slightly more DC offset in BETA110 than in BETA028.

The absolute DC offsets (not diferences!) in BETA110 are:
- 6% (4.5 kHz) for 15 Hz
- 4% (3 kHz) for 10 Hz
- 2.5% (2 kHz) for 5 Hz

So which filters cause this offset? Biggest cause seems to be Bass Boost. Multiband and Stereo Boost also contribute. Why? Because the HPF is the very first filter in the chain (well, directly after the declipper that is). So any asymmetrical bass signals might lead to DC issues.

Maybe it would be a good idea to perform a phase linear highpass filter step even if the non-phase linear HPF is selected. It won't cause a lot of artifacts because there are nearly no low frequencies to filter out, but any new DC offset will be removed.


Edit: Hm. Surprisingly, using the phase linear HPF filter, the results are actually better. Absolute DC content is at 2.7% using 15 Hz, and rapidly drops further if the frequency is lowered. So yes, combining the two should make things a lot better. I've put it on the TODO list.

Hi,
110 betas? woooow. Good work!

I have a question:
What exactly doing the Dynamic Increase slider? Difference between the maximum and minimum set is immeasurable and inaudible. (at least in my LUCIDITY FX4 - Ready to release beta ):

Left = 0%, Right = 100%: http://www.abload.de/img/incdynamicwvoc.png

On average, loud audio comes out 0.3 dB louder at 100%. But there's clearly more distortion, so you probably don't want to set it much higher than 10%, and 0% is usually the best unless you are in an extremely competitive market (with regard to audio level) or you _want_ (?) the extra distortion.

I've tried both in the past and didn't like either.

Moving the HPF before Multiband makes the filter much weaker and increases artifacts. Say there's a tone at 30 Hz, and the HPF is configured to filter at 35 Hz. Then the 30 Hz tone will be reduced by - say - 4 dB. And there will be artifacts. Next, Multiband will increase the 30 Hz tone again so in effect nothing happened - except that the artifacts are still there. On top of that, Multiband and Bass Boost can cause new low frequency signals that won't be removed if the HPF filtering is done before it. (I just discovered this, see my response to eldoradofm).

Phase rotation usually doesn't affect Multiband behavior much. At least as far as I have seen. And again, phase rotation causes weird effects that are your ears (normally) won't notice because they are masked, but if compression is performed afterwards, the masking may fail and you could get really weird noises.

If you could send me one of those recordings where the highs go down too much, I'll take a look to see if moving phase rotation could help. But I would really prefer not moving it.

BTW: There's still a Multiband redesign planned, the new Multiband should act on RMS level instead of peak level, and that would fix this issue completely.

1 - so far unable to reproduce
2 - what does that mean? If something is wrong I need to fix it!

@eldoradfm: Could you check if the DC offset issue is indeed less bad when you use the phase linear HPF? That would at least confirm that it really _does_ have something to do with DC offset.

Edit: If I use *both* filters, I get 1.1% at 10 Hz, 0.7% at 5 Hz.