Nova phase issues

[TreedB]

First of all I can't express enough, my appreciation for your excellent plugins. Due to my meager means along with enthusiasm for making music, your free versions are amazing. I use several, and plan on going to GE versions soon, particularly Nova. I need input on use, it's such a help with keeping tracks with similar spectrums more discernible from each other, particularly sidechaining them, dynamically compressing frequencies more prominent in each others spectrum. However I'm finding lack of correlation when used, I've not pinned down a consistent setting or behavior, but I think it may be worse in sum mode. Obviously a correlation meter reveals the problem dramatically, but checking in mono by ear does verify the problem, with a discernible volume drop in the indicated track. Any thoughts on this, verification there are phase issues with use, possible advice on more proper use to avoid problems, anything at all, would be helpful.
Thanks again for the wonderful music tools

[Smittefar]

Nova is a minimum phase EQ. Any EQ move will inflict some change in the phase. The solution might be a linear phase EQ, but they come with other drawbacks.

[TreedB]

yes agree with all that, but I've never seen it quite this severe with any EQ. Except perhaps, Marvel GEQ, which defies this theory as it is linear phase. However, again, primarily in M/S use, which may be the real problem with both. M/S by nature messes with phase, but I'd have thought summing would not be a problem.

[Smittefar]

There is also the thing that Nova has built-in saturation. This can only be disabled in the GE version. This can really mess with the phase and L-R relationship, when you do mid-side / sum-diff processing.

If you take a sine wave e.g. 1 kHz and pan it all the way to the left. Then add nova in sum or diff mode. Even with all filters at 0 dB, you will start hearing and seeing some signal in the right channel. This is due to the built-in saturation.

[TreedB]

actually I find no mention of saturation or the ability to turn it on or off anywhere in the GE manual. Can you provide any further info on that? A purchase would be in order knowing if I could address this issue

[ilmai]

AFAIK (and owning the full GE collection) Nova doesn't have saturation, it's only SlickEQ that has it. Both SlickEQ GE and the free version have togglable saturation, both in the EQ and output stages. Nova doesn't have any saturation options.

[imrae]

I still don't quite understand what OP is asking about. Is this correlation between stereo channels or between parallel tracks? Have you tried comparing with equivalent (static) EQ curves for a stock EQ?

I can get a 50 dB null on pink noise between Nova GE (eco mode, no EQ gain) and ReaEQ with a single band. With RenEQ I get a 96 dB null below 1k. In both cases the difference smoothly increases up to Nyquist which is interesting. This suggests that the phase shift is following the same rules as other minimum-phase EQs, but there is some other processing difference affecting the high-end. Could be the oversampling/antialiasing setup (which is present even in Eco mode). It's subtle though and shouldn't be a problem in use.

Behaviour in GE's "Precise-" mode is the same as Eco, but in regular "Precise" mode it won't null below about -12. According to the manual, the difference between those modes is "internal nonlinearity". Looking in PluginDoctor, Precise mode has a gentle highpass effect (about 3 dB/octave) below 50 Hz which comes with a small phase shift. If you have lots of subs this might be noticeable?

[TreedB]

stereo track, with several plugins including Nova, put a correlation meter at end of plugin chain. Easy to see the problem, mono the track as a test, and if the problem is severe, the level drops dramatically.
But I think I've found the problem, and it's not Nova per say. It seems to be a result of any M/S eq, linear or not. I knew this was a problem with other imaging manipulation, but not M/S eq, now I know. Whether Nova or any other eq, dynamic, or otherwise, M/S use must be tempered with awareness of this issue, and judgement of what to compromise. For some it may be no issue at all.
Thanks all for your input, and thanks TDL for an amazing tool

[Smittefar]

Regarding the saturation, the manual mentions (Page 23)

NOVA GE offers five different quality modes:
 Eco: Economic mode, with an internal bandwidth ~100kHz
 Precise: Default quality, with an internal bandwidth ~200kHz
 Insane: Very high quality, with an internal bandwidth ~400kHz
 Precise-: Precise quality, without internal nonlinearity.
 Insane-: Insane quality, without internal nonlinearity.

The internal non-linearity is actually a bit of saturation. Try using one of the two last modes to see, how it changes.

[imrae]

The sides channel IS the anti-correlation between L/R, that's how sum/diff processing works. 100% L/R correlation means mono, which is probably not what you wanted either.

[Hermetech Mastering]

Regarding the saturation, the manual mentions (Page 23)

NOVA GE offers five different quality modes:
 Eco: Economic mode, with an internal bandwidth ~100kHz
 Precise: Default quality, with an internal bandwidth ~200kHz
 Insane: Very high quality, with an internal bandwidth ~400kHz
 Precise-: Precise quality, without internal nonlinearity.
 Insane-: Insane quality, without internal nonlinearity.

The internal non-linearity is actually a bit of saturation. Try using one of the two last modes to see, how it changes.

Yep, this is true, Nova is ALWAYS applying saturation unless you switch it to one of the "-" modes. I always have it turned off.

[imrae]

"Eco" also seems to be saturation-free.

[TreedB]

I purchased GE. No practical difference in this regard in any of the modes. I can't comment on the saturation issue, but I have to conclude M/S EQ work by nature messes with phase, and therefore stereo correlation be an issue worth watching when using. I have to admit it's an impressive enticing mixing tool, now I will use it more carefully. So, all this resulted in me learning a bit more about M/S editing issues, and acquiring the more advanced version of the tool. So, the original topic is not exclusive to this plugin at all as far as I can tell.

[TreedB]

The sides channel IS the anti-correlation between L/R, that's how sum/diff processing works. 100% L/R correlation means mono, which is probably not what you wanted either.

right, took a while to sink in for practical use, and apply that to improving practices. thank you. This is the crux of the issue, and not exclusive to this tool

[Smittefar]

Mid-side processing has been quite the new hip thing on the internet the past few years. It has some great uses, but it is not without drawbacks. There really is no free lunch