What causes "steep Q" sibilance?

[rasputin1963]

Here's a noob question for you:

 

Let's say I have a digitally-recorded track of some polyphonic music.... like a guitar or piano.

 

Using sophisticated filter plugins, one can actually notch out a very specific band of frequencies, as you know.

 

Yet, if the "Q" of this notch is steep, a nasty telltale sibilance creeps in. This is why, as you know, it's usually better to have gentler Q-slopes when both boosting and cutting bands of frequencies.

 

So my question is... Why? Why do steep-Q's start to usher in unpleasant sibilance? [This may be like a small child's Why-Is-The-Sky-Blue question...]

 

I tried "getting creative" one day by breaking up a single track into three separate bands-- low, mids and treble-- so I could distribute them to three separate tracks for further experimentation. I separated each band very sharply and steeply, thinking I was doing some very precise "surgery"...

 

Imagine my surprise when the result was not "musical" at all, but filled with nasty sibilant artifacts.

 

Why IS this? And why do gentler slopes sound OK?

 

Thanks, ras

[Danny (NS::U)]

Caution: I may be barking up the wrong tree.

 

However:

 

It's been my experience that many filters (depending on design) will tend to show ringing/ overshoot at the point(s) where the frequency response starts to "fall off the cliff." (That is, when the slope of the filter is extremely high.)

 

If you take into account that you may have some pretty gnarly phase artifacts, and the ringing causing some odd emphasis of an area that might already be a problem (that's why you're notching it out, right?) then you can see how you might actually cause more problems.

 

This isn't to say that this is a universal problem, though. We live sound reinforcement types have been using 24 dB/ octave filters in our crossovers for ages - but those filters are specifically designed to sum at the crossover points beautifully, and to be phase coherent. It is very possible to make extreme filters that aren't completely ugly - but it takes work (and usually a much larger amount of DSP processing power than "simpler" filters.)

[philbo]

 

Caution: I may be barking up the wrong tree. However: It's been my experience that many filters (depending on design) will tend to show ringing/ overshoot at the point(s) where the frequency response starts to "fall off the cliff." (That is, when the slope of the filter is extremely high.) If you take into account that you may have some pretty gnarly phase artifacts, and the ringing causing some odd emphasis of an area that might already be a problem (that's why you're notching it out, right?) then you can see how you might actually cause more problems. This isn't to say that this is a universal problem, though. We live sound reinforcement types have been using 24 dB/ octave filters in our crossovers for ages - but those filters are specifically designed to sum at the crossover points beautifully, and to be phase coherent. It is very possible to make extreme filters that aren't completely ugly - but it takes work (and usually a much larger amount of DSP processing power than "simpler" filters.)

 

 

The phase coherency is the key - - without it, the steeper the filter, the greater the phase shift. When the phase shift starts approaching 360 degrees, the filter becomes resonant and tends to 'ring' at it's tuned frequency when excited by audio.