butterworth filters design

[p2mine]

What do you think about these filters for a surround sound system( Subwoofer + 4 satelites)
the cutoff frequency is 120 for the subwoofer. the schematics looks like:

the filter for the satelites looks like:


the sum of the output signals is( as is computed by matlab)

[officeboy]

As I see it, your crossover points are too close, Your 180 out of phase, and those are seriously steep crossover slopes.
I'd move the sub down to 100ish crossover to help kill that peak.
And I would think about why you have such steep slopes, since I don’t think it’s generally the suggested method and may result in being able to pinpoint the crossover frequency very easily. Maybe tame things down to a 12, 18- 24 db/oct slope..

Just my crappy opinion of course.

[Bill Fitzpatrick]

Originally posted by officeboy
and those are seriously steep crossover slopes.

Those aren't seriously steep crossover slopes, p2mine just speaks the foreign language of "decade" rather than the familiar "octave."

[neutron7]

from what i can see, the easiest thing would be to gradually decrease sub xover till it is flatter, i would think a small dip would be much preferable to a 4db boost especially around that frequency that makes things sound "boomy"

[p2mine]

ok now. thank you guys for the comments. I've revised my schematics now. The filter frequency is 150 for satelites and 100 for subwoofer. that gives a sum almoust flat, a peak is sensed at 78 Hz of +1 dB.
I know that the filters are not in phase. here is the new diagram of phase and gain.
I think that these slopes are ok. because i've tried with 8dB/oct filters and i was able to hear some high frequencies at high volume.

the signals are 270 degrees out of phase.
now, I know that there are circuits that can bring your signal in phase. will it sound different if i would have such a circuit? what will be the difference?

My opinion is that because the frequencies are from different sources, you cannot sense the differency. I mean that it would sound the same if you have to signals in phase at completly differerrent frequencies, as it would be to have to signals 90 degree out of phase at completly different frequencies.

Is it my point correctly or not?

thank you.

[Bill Fitzpatrick]

You have yet to take into consideration the acoustical roll-off and phase of the satellites.

[RHosch]

Why not aim for symmetrical and in-phase 12 or 24dB/oct Linkwitz-Riley alignment? The 24dB/oct variant is nearly the same as the cascaded Butterworth filters you've modeled, with only minor corrections in component values required, and the 12dB/oct variant is simpler if you can tolerate the more gradual slope.

[sreten]

The two set ups should be connected out of phase to reduce
phase error at crossover to nominally 90 degrees, (disregarding
the phase response of the satellites).

L/R second order is also connected out of phase as difference
is 180 degrees at crossover, but L/R 4th order is connected
in phase as the difference is 360 degrees at crossover.

sreten.

[p2mine]

what particularities has linkwitz-riley filters?
butterworth is characterised through it's flat amplitude response at the crossover point.

sreten, i didn't get it quite right. can you explain me in plain english?

what are you saying is that if I have a 90 degree delaying circuit, then the signal (considering that the signal is a sine wave) will be in phase?

thank you!

[sreten]

L/R filters are 6dB down (rather than 3dB) at the crossover
point so the resulting acoustic output sums to unity.

I was just saying for 3rd order crossover circuits where
phase difference at crossover is usually 270 degrees the
drivers are usually connected out of phase.

sreten.