DSP Robotics Support

I guess if I should do it that way I had to have allpass on the "dry" signal with the same phase respone. There must be a smarter way to do it?

In your example if you have a dry signal and subtract a band pass filter you get a positive peak at f0 just as you do with addition of the same.

You are correct, Allpass + bandpass ,) Diffuse focus worn to the brink, but got some nice results as a result.
But hey at least you get a nice shelve Also you can cause feedback into the allpass filter for resonance. But I gues the results will be unpredictable. And ringy..

tor wrote:Here you have an oversampled version of it:
Can sombody (Martin?) please point me in the right direction on how to make peak and shelving filters with this structure, if possible?

Cool! No resonance peak bandwidth squeezing due to frequency warping, all the way up to Nyquist! But why did you remove the denormal blocker?

As MH points out, the other filter types can be obtained by mixing (and for the peak and shelving some parameter rescaling). Here is a collection of some filter types:
- Lowpass
- Highpass
- Bandpass 1 (constant skirt)
- Bandpass 2 (constant peak gain)
- Notch
- Allpass
- Peaking EQ
- Low Shelf
- High Shelf
Have fun!

Okay, so... what are the cons of ZDF filters again? In what areas they are worse than the old ones?

Thank you again Martin

martinvicanek wrote:

tor wrote:Here you have an oversampled version of it:
Can sombody (Martin?) please point me in the right direction on how to make peak and shelving filters with this structure, if possible?

But why did you remove the denormal blocker?

Simply because I use another method

tester wrote:Okay, so... what are the cons of ZDF filters again? In what areas they are worse than the old ones?

They're actually kinda cool, if you dont find ways around the problems you would have otherwise.
Personally I like my TPT blocks to be somewhat at a radian frequency, and require one integrator and not so much math to do so. But this can be solved by stacking inside a new code module and recycling the math.
Heard some pretty nifty implementations, so.. I guess this would be a way to do it.
Not so shure I did this for the same reasons in other projects, under another mode of operation.
But that reflects on abstract concepts of adding memory to integrator, the world must find it usefull to name it.

*Removed double post.

Here you have a double precision variant of the ZDF filter.

I have been trying to make a version that is both double precision and oversampled, but I have failed so far

If anyone want to try help me work this out, I would be happy.

There seem to me like there is an issue with the feedback/onesampledelay.

As far as i understand it is properly wired but still there is issues near nyquist.

Troggie? I know you have helped me with a similar issue once before If you have the time, please, please help me.