Search form

[The Elf]

Have I successfully scared everyone into stupefaction?!

[blinddrew]

We're cogitating...

Or waiting until we're at a screen where we can zoom in enough on the screen shots to see what's actually going on!

[The Elf]

[Wonks]

As clear as Brexit so far.

[Wonks]

Possibly because you've posted two pictures of a low pass filter, when you wanted one low pass, one band pass and one high-pass filter.

[Wonks]

And without being able to see where the outputs of the channels are being sent to (I can see where aux sends are going) it's impossible to work out the full routing path.

[The Elf]

Possibly because you've posted two pictures of a low pass filter, when you wanted one low pass, one band pass and one high-pass filter.

Nope - it's correct as shown. It's all about polarity and routing. Take another look!

[The Elf]

And without being able to see where the outputs of the channels are being sent to (I can see where aux sends are going) it's impossible to work out the full routing path.

It's all there - top of the channels.

[Wonks]

No, not getting it. And the second photo shows the audio has already come from a high pass filter, so it is effectively a band-pass filter for the mids.

I know you know what you're doing, but it's not as easy to let us know what you are doing.

[Wonks]

And without being able to see where the outputs of the channels are being sent to (I can see where aux sends are going) it's impossible to work out the full routing path.

It's all there - top of the channels.

missed that - all rather small.

[The Elf]

OK...

What is happening is this:

The original audio is split, using sends, to three 'Mult' channels, each containing a filter. The original channel has no output destination (though it could be routed to its inversion channel, I suppose - more on that later).

'Mult Lo' gets a LPF that limits it to frequencies below 200Hz.
'Mult Md' gets a LPF that limits it to frequencies below 2kHz.
'Mult Hi' gets a instance of the same filter, but I don't switch on any processing in there - it's simply to ensure that any oddities of the plug-in (delays, distortions, etc) are present in each band. Strictly speaking it needn't be there.

'Mult Lo' is sent to a channel where it is inverted in polarity and this inverted version sent to 'Mult Md' and 'Mult Hi'. This cleanly removes anything below that set 200Hz frequency from both the 'Mult Hi' and Mult Md' channels.

'Mult Md' is sent to a channel where it is inverted in polarity and this inverted version sent to 'Mult Hi'. This cleanly removes anything below that set 2kHz frequency from the 'Mult Hi' channel.

(The inverted channels are not routed to an output, as I only need them for their sends).

Essentially the splitting job is now done. I now route the three 'Mult' channels to corresponding 'Process' channels, where I can do whatever I want to each band - trim, compress, expand... anything.

The last channel on the right is simply an inverted version of the original source. Un-muting this should give an absolute null at the stereo output - proving that no artefacts are being created by the splitting process - and it does do just that.

The whole point of this is to avoid using LPF, Band and HPF filters, since these create anomalies at their cross-overs. My method nulls all such anomalies.

Hope that begins to make sense!

[blinddrew]

Aha, it all makes sense now.

[Hugh Robjohns]

I've done a block diagram which might be helpful to reveal what's going on:

[Watchmaker]

I obviously need to quit my day job.

[The Elf]

I've done a block diagram which might be helpful to reveal what's going on...

...which proves beyond any reasonable doubt that Hugh is far cleverer than me!

Nicely summed up!

[Hugh Robjohns]

It's only a diagram! I think seeing the signal flow is often easier to understand than the words...

The hard part was trying to work out what was going on from your Cubase screen-grab. I should have waited until you'd written your description!

This idea of subtractive filtering is actually used more commonly than it might seem, and does have useful benefits when it comes to seamless integration through the cross-over regions, as you have explained.

H

[Wonks]

OK, Ive worked out what's going on but:

The whole point of this is to avoid using LPF, Band and HPF filters, since these create anomalies at their cross-overs. My method nulls all such anomalies.

This system uses two LPFs. These will still impart their 'artefacts' to the resulting audio signals around the crossover point.

With no extra processing, the artefacts from the LPF cancel out because of the various inversions and recombinations once the three signals are re-combined. But they are still there within in the three process mults. It's only when recombined without further processing that everything cancels out.

But any signal boosting or cutting that affects the 200Hz and 2000Hz areas on one of the mult tracks will not allow full cancellation, so those artefacts then come through; albeit at the level of the boost/cut. So any level alteration would let some artefacts through that could cause slight interaction with the mult above or below once the three mults are recombined.

So with processing on the mults, then filter interaction caused by the band splitting is then significantly minimised (rather than completely eradicated), and it only occurs at two points in the audio spectrum (rather than at the four points that would exist if a LP filter, BP filter and HP filter were used instead to create the mults).

And relax.

[The Elf]

Any artefacts will have *some* effect as soon as you move away from the unity/zero processing position, but better that than have the audio compromised before you've even moved a control. I'd argue that this is better!

[Wonks]

I didn't say it wasn't! I was just trying to get a fuller understanding of how it worked.

[The Elf]

There's no reason why this technique can't be expanded into more frequency bands. You can use whatever filters that take your fancy, change slopes, crossover frequencies, etc. You effectively have your own completely customisable multi-band processor.

Another advantage I hadn't considered until now is that by un-muting the 'Inverted Original' channel you get a hint at what your processing is doing, since anything you hear is the difference between the original audio and your processing.

When I use this for 'mastering' I hide all of the mult and inversion channels to allow me to concentrate on the important stuff.

I've found polarity tricks quite useful in many situations. Recently I used some keyboard parts in opposite polarity to remove the same parts from a mix file where the original tracks without keyboard were not available.

Just realising that this hijack should probably have its own thread! Admins, feel free...