Most Transparent Multiband Splitter

VST, AU, etc. plug-in Virtual Effects discussion
Aesaire
KVRist
58 posts since 13 Sep, 2020

Post Fri Sep 24, 2021 8:15 am

What's the most transparent multiband splitter keeping crossover filter slopes constant (eg. you can only compare splitters with -12db/octave vs other splitters set on -12db/octave)?

Is there a test you can do to to check how unaltered the audio is once processed by the splitter?
Last edited by Aesaire on Thu Sep 30, 2021 3:31 am, edited 5 times in total.

User avatar
] Peter:H [
KVRian
1443 posts since 22 Sep, 2016

Post Fri Sep 24, 2021 8:54 am

Hm, if we assume the Audio is completely unaltered by the splitters then inverting the output after the splitter and adding it to the original signal before the splitter should be all silent. I have never tried it out myself. Afaik all filters apart from linear phase ones add phase shifts of certain frequencies. Therefore I think the experiment is only theoretical and most probabaly you get the sound altered. Using google I found this page here where splitters are discussed: http://www.dsprobotics.com/support/view ... f=4&t=8081.

I own Melda toys and in their MultiBand Compressor you can chose the splitter slope and even set it to linear phase ... might be interessting to experiment with it. https://www.meldaproduction.com/MDynamicsMB

Aesaire
KVRist

Topic Starter

58 posts since 13 Sep, 2020

Post Fri Sep 24, 2021 9:14 am

] Peter:H [ wrote:
Fri Sep 24, 2021 8:54 am
Hm, if we assume the Audio is completely unaltered by the splitters then inverting the output after the splitter and adding it to the original signal before the splitter should be all silent.
Tried it with non-linear phase crossovers, it never nulls. There are way too many phase shifts in a three band crossover.

rafa1981
KVRian
591 posts since 4 Jan, 2007

Post Fri Sep 24, 2021 10:04 am

It is only possible with FIR/FFT crossovers, but you should expect them to have latency.

So it is either phase shift (not a problem if you aren't mixing with the dry signal) or latency.

Aesaire
KVRist

Topic Starter

58 posts since 13 Sep, 2020

Post Fri Sep 24, 2021 9:36 pm

rafa1981 wrote:
Fri Sep 24, 2021 10:04 am
It is only possible with FIR/FFT crossovers, but you should expect them to have latency.

So it is either phase shift (not a problem if you aren't mixing with the dry signal) or latency.
Could you recommend any FIR / FFT multiband splitters? Also, don't FFT crossovers have trouble accurately rendering transients on top of the latency?

MogwaiBoy
KVRAF
4041 posts since 26 Nov, 2015 from Way Downunder

Post Fri Sep 24, 2021 10:20 pm

Before we discuss the most transparent one, how about a list of what's available? They seem pretty rare. As far as I know there's not that many dedicated plugins for this task - although I guess it's not that hard to DIY one yourself.

Actually this is the only one I can think of off the top of my head:
https://klevgrand.se/products/gaffel

What else is there.. Melda?

My favourite overall multiband plugin is Fabfilter's Pro-MB in 'Dynamic Phase' mode. They should use that tech to make a multiband splitter.

MogwaiBoy
KVRAF
4041 posts since 26 Nov, 2015 from Way Downunder

Post Fri Sep 24, 2021 10:29 pm

Actually I've done it with ISOL8, which recombines perfectly I believe.

IIRs
KVRAF
6741 posts since 25 Mar, 2002 from sheffield, england

Post Sat Sep 25, 2021 12:41 am

MogwaiBoy wrote:
Fri Sep 24, 2021 10:20 pm
My favourite overall multiband plugin is Fabfilter's Pro-MB in 'Dynamic Phase' mode. They should use that tech to make a multiband splitter.
The problem with that is, dynamic phase mode is really dynamic EQ, and the audio signal isn't actually split at all.

rafa1981
KVRian
591 posts since 4 Jan, 2007

Post Sat Sep 25, 2021 2:14 am

Aesaire wrote:
Fri Sep 24, 2021 9:36 pm
rafa1981 wrote:
Fri Sep 24, 2021 10:04 am
It is only possible with FIR/FFT crossovers, but you should expect them to have latency.

So it is either phase shift (not a problem if you aren't mixing with the dry signal) or latency.
Could you recommend any FIR / FFT multiband splitters? Also, don't FFT crossovers have trouble accurately rendering transients on top of the latency?
I don't know of any, just multiband plugins that use it internally. It doesn't mean that they don't exist.

By FIR/FFT I wanted to mean a FIR filter doing the convolution by using a FFT. As the number of FIR filter coefficients grow after some point it's more efficient to use a FFT instead of naive convolution. But that doesn't mean that the filtering itself is done on the frequency domain, the results are the same, it's just an implementation detail/optimization. This is how e.g. r8brain does it.

I don't know if the phase shifts of a Linkwitz Riley crossover can be heard on an A/B test on most cases/material. If mixing the original signal is desired, you can apply the same crossover in parallel to the Dry signal and sum it back, this way both the dry and wet signal will have the phase shift (there are more efficient ways to do it inside a plugin).

I'm almost ready to release a version of MixMaxtrix that can be used as a Linkitz Riley crossover with selectable band slope, but there is a bug on the Windows version (Win/Linux only) that I have to catch and very little free time to do it.

I have been toying with the idea of adding a crossover based on forwards-backwards filtering (as ljkb QRange does), but I don't think I'd use it a lot.

FapFilter
KVRian
891 posts since 29 Oct, 2015 from Jupiter 8

Post Sat Sep 25, 2021 2:43 am

not trying to start this old fad again, but Reaper has both IIRC, a minimum phase and linear phase splitter (in the form of JS FX)
You can do it manually with any linear phase EQ aswell (there's a tutorial on Youtube by Dan Worral to do this inside Reaper for instance. Your DAW might do this slightly different)

if your minimum phase splitter features different slope steepnesses for the crossovers, you could try to flatten them to 6db per octave and see if this helps

so TL;DR: have you tried Reaper? :party:
There's no Logic to Windows

Ploki
KVRAF
5024 posts since 17 Dec, 2009

Post Sat Sep 25, 2021 3:41 am

I’ve recently discovered that a lot of multiband operations (dynamic and saturation for example) are better in single band with compensated emphasis or creative sidechain
Image

Aesaire
KVRist

Topic Starter

58 posts since 13 Sep, 2020

Post Sat Sep 25, 2021 7:21 am

MogwaiBoy wrote:
Fri Sep 24, 2021 10:20 pm
Before we discuss the most transparent one, how about a list of what's available? They seem pretty rare. As far as I know there's not that many dedicated plugins for this task - although I guess it's not that hard to DIY one yourself.

Actually this is the only one I can think of off the top of my head:
https://klevgrand.se/products/gaffel

What else is there.. Melda?

My favourite overall multiband plugin is Fabfilter's Pro-MB in 'Dynamic Phase' mode. They should use that tech to make a multiband splitter.
There's a JS 3 band splitter in Reaper.
There's the RS-MET Crossover. http://www.rs-met.com/freebies.html

Vast majority of these should be Linkwitz/Riley filters, so they should be linear in the frequency domain but have phase shifts.
Actually I've done it with ISOL8, which recombines perfectly I believe.
ISOL8 has phase shifts when you put it into multichannel mode. The manual says it uses Linkwitz/Riley filters too.

Aesaire
KVRist

Topic Starter

58 posts since 13 Sep, 2020

Post Sat Sep 25, 2021 9:00 am

rafa1981 wrote:
Fri Sep 24, 2021 10:04 am

I don't know of any, just multiband plugins that use it internally. It doesn't mean that they don't exist.

By FIR/FFT I wanted to mean a FIR filter doing the convolution by using a FFT. As the number of FIR filter coefficients grow after some point it's more efficient to use a FFT instead of naive convolution. But that doesn't mean that the filtering itself is done on the frequency domain, the results are the same, it's just an implementation detail/optimization. This is how e.g. r8brain does it.

I don't know if the phase shifts of a Linkwitz Riley crossover can be heard on an A/B test on most cases/material. If mixing the original signal is desired, you can apply the same crossover in parallel to the Dry signal and sum it back, this way both the dry and wet signal will have the phase shift (there are more efficient ways to do it inside a plugin).

I'm almost ready to release a version of MixMaxtrix that can be used as a Linkitz Riley crossover with selectable band slope, but there is a bug on the Windows version (Win/Linux only) that I have to catch and very little free time to do it.

I have been toying with the idea of adding a crossover based on forwards-backwards filtering (as ljkb QRange does), but I don't think I'd use it a lot.
Haven't tried FIR/FFT filters before so I can't comment.

I don't think I'd be able to tell the difference in a blind A/B most of the time. Linkwitz Riley crossovers should be good enough for the majority of use cases, I was just wondering if there was an even more transparent splitter available.

(Checked out MixMaxtrix, cool stuff)
FapFilter wrote:
Sat Sep 25, 2021 2:43 am
not trying to start this old fad again, but Reaper has both IIRC, a minimum phase and linear phase splitter (in the form of JS FX)
You can do it manually with any linear phase EQ aswell (there's a tutorial on Youtube by Dan Worral to do this inside Reaper for instance. Your DAW might do this slightly different)

if your minimum phase splitter features different slope steepnesses for the crossovers, you could try to flatten them to 6db per octave and see if this helps

so TL;DR: have you tried Reaper? :party:
Flattening them to -6db will solve transparency issues, but the crossovers are too shallow to separate the bands significantly. That's why I mentioned keeping slope constant, I know gentle slopes will induce less artifacts, but a many applications call for steeper slopes.

I've found the JS:3 band splitter, can't find the minimum phase and linear phase splitter. Found a FFT splitter too, downloaded it from Reapack I think.

Did some testing, ISOL8, RS-MET and the Reaper FFT Splitter don't null against an inverted signal, the lowest slope they go to is -12db/octave. The Reaper JS:3-band splitter actually nulls with a inverted signal. IIRC this splitter is -6db/octave. The Melda analog and linear phase crossovers also null when set to -6db/octave. Anything above -12db/octave doesn't null, this goes for linear phase crossovers and hybrid crossovers too.
Ploki wrote:
Sat Sep 25, 2021 3:41 am
I’ve recently discovered that a lot of multiband operations (dynamic and saturation for example) are better in single band with compensated emphasis or creative sidechain
Yup, the multiband alternatives are usually cleaner. Multiband does have a lot of interesting creative applications though, so it's worth discussing.
Last edited by Aesaire on Sat Sep 25, 2021 10:04 am, edited 2 times in total.

FapFilter
KVRian
891 posts since 29 Oct, 2015 from Jupiter 8

Post Sat Sep 25, 2021 9:19 am

i've only tested it once, because of a payware splitter i was using had extremely audible crossovers.
Checked the JS splitter and to my surprise even the non FIR one completely nulled to infinity with the unprocessed signal.
Might be though that the slopes of the JS one are indeed very flat.
Maybe i was lucky as those crossovers might have been set to “perfect“ frequencies, but i set both the payware splitter and the JS one to the exact same frequencies and the payware one had very audible phasing issues while the JS one nulled.
As long as i'm not hearing any issues it's good enough for me anyway.
And of course if you are doing some heavier mangling, even audible phasing is not a big problem anymore (for me anyways)
There's no Logic to Windows

rafa1981
KVRian
591 posts since 4 Jan, 2007

Post Sat Sep 25, 2021 9:29 am

Aesaire wrote:
Sat Sep 25, 2021 9:00 am
Haven't tried FIR/FFT filters before so I can't comment.
I see that you are a Reaper user. Sake's NOTT JSFX effect has a linear phase mode and LR. It is not a crossover though, but a multiband upwards compressor.

Return to “Effects”