FreeDSP OCTAVIA

I'm stuck on designing a Linkwitz-Reiley 8th order crossover. Simulating in SigmaStudio, 4th order crossover is perfect as below...
スクリーンショット 2025-04-29 13.59.49.png

But just cascading the same one, the phase is perfectly matched, but the crossover point would be a -6dB dip in total system response.
スクリーンショット 2025-04-29 13.58.24.png

So I try to move the cut-off frequency toward the crossover frequency.
スクリーンショット 2025-04-29 14.12.01.png
The total frequency response has a -7.5dB dip. This is worse than before!

Does anyone know how to solve this problem?

CyberPit
 
Last edited:

uriy-ch-San,​

Fixed crossover frequency is inconvenient in the case of cut-and-try speaker tuning.
So, I just want to build a voltage-selectable variable crossover-frequency dividing network.
Current implementation is like this...
スクリーンショット 2025-04-29 16.09.31.png

I believe L-R8 filtering is possible because you suggested an example.

CyberPit
 
Last edited:
  • Like
Reactions: uriy-ch
LR crossovers are built from Butterworth crossovers of half the LR order, not Linkwitz-Rileys. For example, LR4 is composed of two B2s in cascade. So for creating an LR8 you would cascade two 4th order Butterworth filters, not LR4s.

Remember LR transfer functions are 6 dB down at the crossover frequency, so two of them in series means the xovr point is 12 dB down, as you saw. Compare this with Butterworth filters being 3 dB down so cascading two of those yields the desired 6dB down at crossover.
 
Thanks, DSP-Geek-San and torgeirs-San:worship:

The inventor says each stage Q equals 0.54, 1.34, 0.54, 1.34
So I put into those values...
スクリーンショット 2025-04-30 11.03.00.png
Not bad. But there is +0.5dB peak. I thought this might be some fraction errors.

So I spread the crossover frequency a little...
スクリーンショット 2025-04-30 11.21.27.png
It's almost a flat response. I guess this will be enough for actual use.
Thank you guys, now I can update my design.

CyberPit
:cheers:
 
  • Like
Reactions: torgeirs
A special thanks to Mir. Linkwitz. And forum members who supported me.
Now I can build an LR8 Channel Divider, which has a tunable Crossover frequency. At the center position of the pots, the frequency response was measured.
OCTAVIA_VariCross_CentePos_Response.gif

(Please ignore the noises below 500Hz. It depends on my measurement system.)
Git-hub Wiki

CyberPit
 
Last edited:
Uriy-ch-San,
Maybe it depends on AUX-ADC noise, I checked the Full-Range output with the sine wave. The ourput spectrum was like this...
スクリーンショット 2025-05-01 12.39.02.png
An extremely high-level spread noise floor existed. And it comes from the Volume blocks. This AM nose disappeared when I changed the CV to a DC source.
So I try to put various LPF blocks after the AUX-ADC input. But most of the LPF blocks do not pass DC.
Finally, I found the First Order Filter Block passes DC. It works fine as I expected. The value is stabilized very well.
スクリーンショット 2025-05-01 15.47.19.png
After this modification, the output spectrum is changed drastically.
スクリーンショット 2025-05-01 15.32.03.png

The output became very clean as I expected. This modification was applied to Balance and delay control.
Now you can clone the latest SigmaStudio project from the GitHub page.

https://github.com/freeDSP/freeDSP-...Balance_Sweep-Cross_TW-Delay_Improvrd.dspproj

CyberPit
 
Last edited: