3-Way Linkwitz-Riley Electronic Crossover (please comment)

[Hicoco]

First Plot with second order all-Pass Filter and second Plot without.
Input 1V - 1kHz. Traces Output Crossover

 

[Hicoco]

With All-Pass Filter

No All-Pass Filter

 

[MarcelvdG]

I calculated 50 degrees of phase difference at 2 kHz and you simulate about 310 degrees, but my calculation wraps, that is, full 360 degree cycles are ignored. 310 - 360 = -50.

 

[MarcelvdG]

Theoretically, with ideal components, the crossover frequencies should be 600.0860595 Hz and 2000.169546 Hz. Do the simulated values get closer to the theoretical ones when you simulate with more steps per decade (or per octave)? It isn't of any practical importance, I'm just curious.

 

[MarcelvdG]

If you should want to try it with an all-pass filter, there is a useful document on the Analog Devices site called MT-202, https://www.analog.com/media/en/training-seminars/tutorials/MT-202.pdf Figure 5 and the equations below it show how to design one, except that the equation for R₄ is wrong, it should have been R₄ = Q/(2k).

In your case, that leads to this circuit:

View attachment 1130916

with values

C = 33 nF
R₁ = 5683 ohm
R₂ = 11366 ohm
R₃ = 5683 ohm
R₄ = 2841.5 ohm

to be inserted between U_7.2 and C₃₈. The circuit attenuates by three so you will have to make up for that somewhere else in the chain, for example by reducing R₄₀ of your original circuit to 2 kohm.

For your information, R₃ = 2 kohm and R₄ = 1 kohm should work equally well, as only their ratio is in the transfer function. 1 kohm and 2 kohm are both standard E96 (and E24) values.

 

[Hicoco]

Theoretically, with ideal components, the crossover frequencies should be 600.0860595 Hz and 2000.169546 Hz. Do the simulated values get closer to the theoretical ones when you simulate with more steps per decade (or per octave)? It isn't of any practical importance, I'm just curious.

i use .ac dec 100 1 100k

 

[MarcelvdG]

That should be enough. No idea where the inaccuracy comes from then.

 

[Hicoco]

Thanks for your advices, PCB on the way to the factory, will see with ideal components.

 

[Hicoco]

Input 600Hz 0.5V Medium & Low

Input 2000Hz 0.5V Medium and High with all-pass filter to correct the phase error

With Medium and high the all-pass filter been by-pass

Exact values
R1= 5.683K, R2= 11.366K R3= 2K, R4= 1K, C= 33.12nF, C'= 33.2nF

 

[Hicoco]

i still having a little error, i will try to change a little the value of R1 and R2 to correct this phase difference..any ideas!!

 

[MarcelvdG]

What does it look like at the inputs of the 1 uF DC blocking capacitors? Are the level potmeters set equally?

 

[Hicoco]

2kHz 500mV at each input, the high level are set to be adjusted at the same level as the med, for FC 2khz around 386mV

 

[MarcelvdG]

I meant the remaining phase error, is it already there before the 1 uF capacitors? The RC times and hence the phase shifts of the 1 uF capacitors will depend to some extent on the level settings.

 

[Hicoco]

i am not sure i have to check that

 

[Hicoco]

Yes look like it is there before the 1 uF

High & Med before the 1 uF without all pass filter

High & Med before the 1uF with all pass filter

 

[MarcelvdG]

I'm sorry, but I haven't a clue what the cause may be.

Are your oscilloscope probes properly calibrated, if they are 10:1 probes? What type of op-amp do you use?

 

[Hicoco]

Yes everything look good, i use LME49720

 

[MarcelvdG]

So the 1 kHz calibration square wave looks exactly as a square wave through your probes, no noticable over- or undershoot. Pity, that would be easy to fix.

With LME49720's, you have plenty loop gain at audio frequencies, so that can't be the cause either.

 

[Hicoco]

yes exactly the same, i did a test at different frequencies, and i am in perfect phase at 2.9KhZ and up see pics test from 2.1khZ to 3.9khz with 1V input. That is at the output High and Med after the 100 Ohm.