LINKWITZRILEY 4th Order Problem

RWE TexM3 · Messages: 212 Registration: 03.14.2003

Hi guys, I have a little problem with a frequency splitter. Found this code on musicdsp and the class in an other forum (I don't remember where..) Code:

Mistermotorist · Messages: 651 Registration: 03.25.2001

I tried using the same code and had similar results. I tried to lowpass the frequency controls, but wasn't too successful. In the end, I modified this . It is C++ class for the RBJ filters. If you put two filters back-to-back with a Q=0.707, you should get a 4th order filter that sums flat. I know it doesn't directly answer your question, but I found the RBJ code very useable and flexible.

Chongus · Messages: 2,241 Registration: 10.22.2002

If the Q is set to 0.707, the filters will be Butterworth. When 2 2nd order filters are stacked together to make a 4th order filter, they will be -12dB at fc. The end result is that you would have a LR4 filter. The other thing you could do is modify the RBJ filter coefficient code to make a specific LR2 filter for the 2nd order biquad calculation. I have this in my RBJ filter coefficient code, but I am not using it because I switched to using higher order filters (in other words, I think it was working, but I am not really sure). If you use the 2nd order filter, remember to flip polarity of one of the bands. Code:

Zinka325i · Messages: 204 Registration: 02.26.2001

Polarity....

DarkSideofLeigh · Messages: 114 Registration: 12.12.2002

How could I forgot that! Well, I think since I bought some DSP related books, I'm "over-readed". Simply forgotten the polarity of them.... Thanks for your help and tips on this, but I think I will try to "correct" the Linkwitz.. simply because it sounds good. i will post the final file if I have some success.

TC M3 · Messages: 372 Registration: 05.26.2002

OK, I tried that with the rbj biquads. But I'm not shure if the Q set to 0.707 returns a linear output. I think it's about 0.500 to 0.600. (I have no analyzer on my mac)

M3Armand · Messages: 922 Registration: 08.07.2003

I am not sure what is happening. Are you putting two identical filters back-to-back with a Q of 0.707? That should sum flat (when 2 lowpass are matched with 2 highpass with the same frequency). If you are using the the LR2 coefficients, the Q is not used (and you need to flip the polarity of that one). I'd post a pic, but a flat line seems like a pretty lame pic.

UR2SLOW · Messages: 389 Registration: 07.26.2003

After I broke my head a long time about this Linkwitz Riley code and how to make it stable, I decided to put away and tried another thing. Well, I have made a x-over filter class which I want to share with you guys. It's based upon a biquad second order LP/HP filter but has -6dB at the cutoff point. So it comes very close to the LR 4th order. And it is stable :) I hope you find it useful! Code: