5

I'm using digital filters to apply spectral mangling-type special effects to audio.

When using a digital filter (vsts/standalone DSP programs/outboard digital filter, etc.), especially when using narrow transition bands/brickwall filters, are there any effective ways to remove ringing artifacts introduced by the filter? Please bear in mind I am new to DSP and my level of understanding is very basic.

jonsca
  • 1,994
  • 3
  • 21
  • 39
Dale Newton
  • 287
  • 3
  • 11
  • 4
    The short answer is "don't make them brickwall filters". See here for the longer answer- http://dsp.stackexchange.com/questions/2170/why-do-i-see-ringing-in-the-output-of-a-digital-filter-with-a-narrow-transition – Jim Clay Apr 26 '12 at 12:55
  • Can you hear the ringing? – endolith Apr 26 '12 at 16:52
  • @endolith Sounds like a song from Les Miserables "Do you hear the filter ring?" To the OP, ringing is the price to be paid for sharp-cutoff filters. Once upon a time, filters were broadly classified as High-Q or Low-Q (Q essentially denoted quality in some broad sense), and that High-Q filters are prone to ringing was a mantra or rule of thumb taught to all engineers, technicians, etc. If you can taper off the input signal (tell the diva to close her mouth gradually instead of snapping it shut at the end of her aria), you will greatly reduce the ringing. – Dilip Sarwate Apr 26 '12 at 17:38
  • Now that you mention it, it does conjure up images of Valjean locked up in a dark room. – Dale Newton Apr 26 '12 at 20:29
  • @dilip I would like to ask: is there nothing that can be done about the ringing in its various forms that's inevitably added to the sound? For example (humour with me here!), if the filter knows its own properties and cut offs etc, can it not be designed to cancell out any artefacts like ringing it introduces by introducing the same signals but 180 out of phase so they are cancelled out? Im guessing that 'brilliant idea' has been thought of by at least 1 filter designer (and is probably a ridiculous idea for some reason 8) ) – Dale Newton Apr 26 '12 at 20:35
  • @endolith I can hear a lot of ringing, also filtered channels seem to come in early and rise like a tide of sound whenever theres a big step change in the input level . To be honest though, with the ringing I dont know how much of that is the rigning from the filter, and how much is just the effect of hearing the extremely narrow bands of sound which are coming through, or other such pscyho-acoustic effects. Im experimenting with that to find out! Generally though I want to know if the problem of ringing, which is surely going to be a major factor in what Im doing, can be tamed somehow. – Dale Newton Apr 26 '12 at 20:47
  • 1
    @DaleNewton Yes, you could do the 180 degrees out of phase thing, but you would be changing the frequency response of the filter in such a way that it would no longer have a sharp transition. – Jim Clay Apr 27 '12 at 02:44
  • 1
    @DaleNewton Some sound examples would really help here. It's hard to tell at the moment whether what you're hearing is an artifact or the natural response of a narrow-band filter. – datageist Apr 27 '12 at 02:45
  • OK. I'll put some samples up as a new question. Thanks. – Dale Newton Apr 28 '12 at 14:48
  • 1
    @DaleNewton Probably not a solution to your problem, but I have some thoughts about ringing in this question: http://dsp.stackexchange.com/questions/6492/how-much-of-a-problem-is-the-gibbs-phenomenon-and-would-this-solve-it – Mats Granvik Jan 12 '13 at 19:57
  • Thanks. Above my head though to be honest. how would your functions be used to remove the artefacts caused by the gibbs phenomenon? – Dale Newton Jan 20 '13 at 01:16

1 Answers1

7

Sticking with linear systems, removing the ringing is nearly the same as adding back some of the spectral content that your really steep transition filters removed. Why use some crazy scheme to add back the stuff in the "softer" transitions that your hard-edged filters cut out? Just use a more reasonable total filter response in the first place.

Going to non-linear systems, you could use some sort of AI pattern matching to determine what kind of sound waveforms would be perceived by a human a ringing, and just gate those waveforms out. But that might just add weird sounding artifacts of its own (as well as also spreading spectral content outside of your really steep filters).

hotpaw2
  • 35,346
  • 9
  • 47
  • 90
  • The narrow transition bands seem to be the only way to get some of the effects I'm after. The sound is often close to how Id like it, but the ringing is spoiling it. – Dale Newton Apr 27 '12 at 00:40
  • You might want to specify more precisely the audible result of the "effect you are after" in another new question, and see if it can be implemented without a bad filter. – hotpaw2 Apr 27 '12 at 00:49
  • Regarding your gating specific waveforms idea, I had no idea that was possible. Could you explain a little how that would be done? Would it be very different from how a regular filter works? Aside from my uses, If its possible to gate specific waveforms by some method other than that used by 'everyday' commercial Equalizers like I use (parametric filters / notch filters etc) I imagine such a filter could be very useful for many applications (removing mains hum, notch filtering, effects) etc. – Dale Newton Apr 27 '12 at 00:50
  • For example completely removing / augmenting specific frequencies or narrow bands of frequencies in order to change the spectral content / timbre of instruments. – Dale Newton Apr 27 '12 at 00:57
  • @Dale : if that was a question, you might want to ask it as a question, not make another comment under the current question. – hotpaw2 Apr 27 '12 at 13:54