My note about frequency splitting.

[mromgwtf]

I've seen many people splitting their basses into a few frequency bands with FL Studio's EQ2 or other equalizer and then processing them differently.
Your typical EQ uses IIR (infinite impulse response) filters - which are used commonly because of their speed. But they have a non linear phase response, meaning that they alter the phases of different frequencies in the spectrum. And so if you split your signal into n bands using an IIR filter, and then reconstruct it - some of the frequencies will cancel each other out, because of the phase differences, leaving you with a fucked up signal. The proper way to split frequencies is to use a linear phase EQ, which uses FIR (finite impulse response) filters - much slower but they preserve the phases.

First one is a casual square wave, second one is a square wave after being split into two bands at 600 hz and then combined together:


Edit: I should have also said that FIR filters add delay to the signal.

[Fowles]

Good post.

[NinjaEdit]

A second-order Linkwitz-Riley filter, which is a kind of IIR, is used for cross-overs (Pirkle 2013, p186).

[mromgwtf]

A second-order Linkwitz-Riley filter, which is a kind of IIR, is used for cross-overs (Pirkle 2013, p186).

From what I'm reading that's just two butterworth filters cascaded. But what about the phase response? You can execute the second butterworth on reversed signal, and with that canceling the phase of the first filtering operation, which will result in linear phase response. But that requires you to have access to whole signal saved for example on disk, and that's not possible in real time. Am I missing something here?

[NinjaEdit]

It just says the response is flat. That may not mean there is no phase cancellation. Using and bypassing Ableton's Multiband Dynamics sounds at least so similar that I can't tell the difference.

[bouncingfish]

Is this stuff audible though, or can you just see it with analyzers?

[mromgwtf]

Is this stuff audible though, or can you just see it with analyzers?

Definitely audible. You can't hear a phase change from single harmonic, but when more than one harmonic are in the signal, the phase change becomes audible.
But when other person listens to the track, he can't tell the difference because he didn't hear the first version (not shifted). So that's not the case. The thing is that such phase shifts change the shape of the waveform and more important, the RMS of the waveform. This kind of modification of a waveform can absolutely change the way a limiter interprets the signal, and that means: muddy mixes/masters.

[Banesy]

Do you have a sample to compare two different filtering styles so we can hear it happening? I believe you that it happens but how prominent is the result?

[mromgwtf]

Do you have a sample to compare two different filtering styles so we can hear it happening? I believe you that it happens but how prominent is the result?

I will try to provide a sample today.

[bouncingfish]

THank you!

[GregoryTJ]

I wonder if a FFT could be made to run fast enough while preserving the 'phasiness' of a signal to be used in frequency splitting.

The Frequency Splitter Template that comes with FL Studio's 'Patcher' doesn't distort the wave very badly, no idea what filters are behind it though.