Issues with/alternatives to BV512 EQ for multiband processing

[Sus4]

Hi all-
I've thrown together a combi using a few BV512s in 16-band EQ mode to split a signal into four bands for some basic multiband processing. But today I noticed some pretty high frequencies still making it through the "low-pass" EQ, and I'm trying to figure out why, so I can work on some alternatives.

Since I can't find any details on which bands are controlled by the GEQ faders I can't say exactly where attenuation should be kicking in, but that instance of the vocoder is set to attenuate the 12 right-most bands. Some sound that I'd say is centred around 3kHz (or 6 bands up from the last band not set to zero) is still making it through, and pretty audibly too.

I'd like to think I've got a more-than-basic understanding of signal flow, but there's clearly a gap in my knowledge here. I know not to expect a brick wall filter from a GEQ, but... I would expect the signal to be getting pretty close to zero so far from where attenuation starts. Do the vocoder's filters just have a very gentle slope? Is this related to phase problems that the EQ introduces?

Then I guess a follow-up question is whether anyone can suggest some better REs or VSTs (free would be ideal! but not necessary) for a roll-your-own multiband unit with more surgical (but distortion-free) filtering.

Thx

[Oquasec]

Depending on what you're doing that would basically duplicate the entire signal causing a phasing issue since the whole signal is duplicated with no offsets since it's the same signal duplicated again by any splitters or something. I usually stick to the splineEq or mclass.
Although...Keeping each EQ in it's own frequency space solves the phasing issues...

[selig]

Hi all-
I've thrown together a combi using a few BV512s in 16-band EQ mode to split a signal into four bands for some basic multiband processing. But today I noticed some pretty high frequencies still making it through the "low-pass" EQ, and I'm trying to figure out why, so I can work on some alternatives.

Since I can't find any details on which bands are controlled by the GEQ faders I can't say exactly where attenuation should be kicking in, but that instance of the vocoder is set to attenuate the 12 right-most bands. Some sound that I'd say is centred around 3kHz (or 6 bands up from the last band not set to zero) is still making it through, and pretty audibly too.

I'd like to think I've got a more-than-basic understanding of signal flow, but there's clearly a gap in my knowledge here. I know not to expect a brick wall filter from a GEQ, but... I would expect the signal to be getting pretty close to zero so far from where attenuation starts. Do the vocoder's filters just have a very gentle slope? Is this related to phase problems that the EQ introduces?

Then I guess a follow-up question is whether anyone can suggest some better REs or VSTs (free would be ideal! but not necessary) for a roll-your-own multiband unit with more surgical (but distortion-free) filtering.

Thx

The BV512 is an EQ, and as such not good at splitting a single into bands for multi-band processing. You need to use filters for that, specifically filter that will re-combine with a flat response such as the MClass Stereo Imager.

No EQ can attenuate even close to zero, most give 20-24 dB of cut at most. Filters OTOH can give you that much attenuation per octave (or more)!

[Sus4]

Depending on what you're doing that would basically duplicate the entire signal causing a phasing issue since the whole signal is duplicated with no offsets since it's the same signal duplicated again by any splitters or something. I usually stick to the splineEq or mclass.
Although...Keeping each EQ in it's own frequency space solves the phasing issues...

I'm not totally sure what you mean by "keeping each in its own frequency space." Right now I'm splitting one signal through JP's Shelob into four vocoder EQs, each with a different set of bands left at unity and every other band fully attenuated. I'd expected some phasing no matter what but it seems like there might be even more than I'd anticipated.

The BV512 is an EQ, and as such not good at splitting a single into bands for multi-band processing. You need to use filters for that, specifically filter that will re-combine with a flat response such as the MClass Stereo Imager.

I had considered using the MClass, but the nice thing about the BV512's EQ is the ease of setting the bands visually, even with a tradeoff in precision. Should have guessed that wouldn't be the only tradeoff

Let's say I use something like the Nova with 96dB/o -pass filters. What trouble will I be getting myself into then? Is that steep enough to expect major problems with phase or resonance? (As I said, I've found a gap in my knowledge here, so forgive me - some research suggests it's not a totally obvious question, but I could be wrong!)

[dioxide]

Note that the different devices in Reason give different results. A while back I did a comparison of the different band-splitting devices in Reason.
viewtopic.php?t=7492956

The BV512 isn't in there, but from what I recall the cutoff slope is very steep so you will need something similar to replace it. Also there are some issues with the Flower Audio device which are unfixed as far as I am aware.

[mbfrancis]

If you split w/ the BV512 and re-assemble the track totally dry/no processing, it won't sound *anything* like the original signal. For band splitting I use the stereo imager or 4dyne. Yoko (such a good name) is great too.

[selig]

Depending on what you're doing that would basically duplicate the entire signal causing a phasing issue since the whole signal is duplicated with no offsets since it's the same signal duplicated again by any splitters or something. I usually stick to the splineEq or mclass.
Although...Keeping each EQ in it's own frequency space solves the phasing issues...

I'm not totally sure what you mean by "keeping each in its own frequency space." Right now I'm splitting one signal through JP's Shelob into four vocoder EQs, each with a different set of bands left at unity and every other band fully attenuated. I'd expected some phasing no matter what but it seems like there might be even more than I'd anticipated.

The BV512 is an EQ, and as such not good at splitting a single into bands for multi-band processing. You need to use filters for that, specifically filter that will re-combine with a flat response such as the MClass Stereo Imager.

I had considered using the MClass, but the nice thing about the BV512's EQ is the ease of setting the bands visually, even with a tradeoff in precision. Should have guessed that wouldn't be the only tradeoff

Let's say I use something like the Nova with 96dB/o -pass filters. What trouble will I be getting myself into then? Is that steep enough to expect major problems with phase or resonance? (As I said, I've found a gap in my knowledge here, so forgive me - some research suggests it's not a totally obvious question, but I could be wrong!)

The goal with multiband is to have a flat response when you re-combine the filters. Not all filter types do this. Stereo imager does this perfectly as do some of the filter types on the others mentioned (Yoko and 4dyne) - but not all the filter types in those devices are "complementary" filters (combine 'flat'). You really don't want super steep filters with multiband any more than you want to butt-splice sustaining audio - you need a smooth transition so that you don't get abrupt changes across the spectrum. And yes, the steeper the filter the more phase shift and possible irregularities when using them in a multi band setting.

[Sus4]

Cool. Thanks everyone for the insights.