DSP Robotics Support

Hi all,

I've built a kind of resonator filter and now I want to create a mirror image of it.
so to put it simply: if the res filter emphasizes the frequency 300hz by 24db for example I want an anti resonator to take the input and do exactly the opposite.

My goal is(and maybe someone can point out a better solution) to be able to blend to res and the anti res with a knob because my res is doing it's job too good and a lot of the character of the input signal is lost so I want the user to be able to have sort of a "spectrum compressor" that will leave the peaks created by the resonators but with an option to tone them down with the anti-res' knob.
I don't want to use a simple res+in mix cause it doesn't accurate/"tight"/"togeather" IMHO, it sounds like two different signals put togeather...

hope I'm clear, thx for all the previous & future help

This is quite a complicated problem to cope with. When handling with bandpass filters the problem is simple - divide the output by Q and the peak will be snapped to 0dB. With highpass and lowspass is not that simple. In lowpass the resonance peak is present below the cutoff frequency. When you center a peaking filter on it and tune it down, you will notice that the peak is actually not symmetric (the left half falls slower than right half) and when you tune it down you create a notch roughly at the Cutoff. Mixing Wet+dry solves nothing - that is simply how shelving filters work and they have even two resonance peaks - one going up and one down.

To reduce the peak simply cascade multiple filters with lower Q (basically creating higher order lowpass/highpass with steeper response at low resonance).

Have ypu tried "signal" minus "filtered signal" procedure? It should give you the exact opposite of filtered signal.

@tester: it turns out the same as the dry/wet method...

@KG: how would I calculate the Q? could only find bw(in ocataves) to q online. my bw is in hz and I'm not sure about the conversion...

zoobooboozoo wrote:@tester: it turns out the same as the dry/wet method...

@KG: how would I calculate the Q? could only find bw(in ocataves) to q online. my bw is in hz and I'm not sure about the conversion...

I've found this:
http://www.rane.com/note170.html

basically the very first chapter "Given the -3 dB Points, To Find BW and Q" shows what to do.

Dry/wet is a combination of a sum between unfiltered signal and filtered one. I said - substract one from another. You may test it on lowpass/highpass filters, one is in opposition to the other. "signal" - "lowpass filtered" = "identical sounding like highpass filter" (if you substract true highpass from s-l then you will get only some background artifacts).

Can you post your filter?

Unless your filter isn't normalised or you absolutely require correct phase shifts, that what tester described is in most cases the best way to go. It doesn't require fancy recalculation of filter coeffs and is also easy to wrap your brain around it after months you haven't worked on the part.
Depending on your filter design you might have to delay the dry signal some samples before subtracting the filtered path from it.

I tried that method, somehow it doesn't sound natural just yet. like the resonator is too seprated/emphasized compared to the dry signal.

Why should I sample delay the subtraction? and by how many samples?