5 replies to this topic

[shaun]

Hi , lm trying to nut out how to fix a problem with passive networks. As you increase the order of a crossover you get a 90deg shift in phase eg:6db=90deg , 12db=180deg , 18db=270deg , 24db=360deg. Somewhere a long time ago l remember reading that it wasnt as easy as just reversing the pol of the driver for 12db/second order crossovers and msot of the crossovers that l have made are either 1st or 2nd order. Now l want to make some 3rd order crossovers and am not sure how this phase shift will affect things. Also is it corect in saying that zobel networks are for stablizing the impeadence change so that the amp can better control the driver? Any advice on this and the phase shift question would be really great as lm not sure what to do? Thanks

[Pulse-R]

Zobel network is to normalise the impedance that the x-over sees looking at the driver - you put it at the driver ususally, not the x-over.

the 270deg phase shift is usually done by reversing the polarity of the driver (180deg) then moving the tweeter back(or forward, can't remember exactly), physically the amount of distance of the 90deg at the x-over frequency.
90deg = 1/4 of the wavelength at the x-over freq.

but the 90deg is only 45deg per driver... you won't really notice it much

recommended is 4th order butterworth.

is there a particular reason you want to use 3rd order?

[jds303]

yes, a zobel network neutralises the effects of impedance changes due to changes in frequency (as a voice coil is an inductor), so the amplifier sees a more constant load.
a zobel network is used in parallel with a speaker, therefore can be placed either at the speaker, crossover, or incorporated into the crossover circuitry.

[jas]

if you are using a digital x-over then this is not the case

the linear phase shift in the digital x-over for a FIR filter is half the FIR waveform length. Less FIR points used the shallower the filter slope and a smaller linear phase shift.

If you are concerned just use the same filter slope for both tweeter and mids. OVer-all the most used analogue filter seems to be the 12db/oct butterworth style which is very good in the frequency domain and acceptable in phase.

you also have a very valid point regarding zoble networks or notch filters. Yes from what ive read they are used for FREQUENCY response manipulation or in the case of zoble networks impedence correction. However since they are reactive components they have to affect the ouput in the phase domain aswell. The audio x-over books i own they only discuss the frequency properties of a notch filter or the impedence properties of zoble network. HOwever with a zoble network its designed to counteract the rising impedence of the driver in its top end frequency response. The inductive properties (the voice coil is a single series inductor) can be seen in the top end (high frequency area) of a drivers impedence plot. To counter-act this the designer can use a parallel capacitor (at the drivers side of the x-over) to flatten the impedence of the driver for the passive x-over to be more linear in its roll-off. The careful matching of the zoble network doesnt allow the effects to be seen by the x-over. The slope intended remains the same, not an order out.

just remember if you add passive components to change phase of a x-over you are just changing its order/slope anyway. if you want two slopes to be 180degrees out of phase just choose two slopes that are 12db out. Even better just use the same slope for each.

ok here might be something else to look into. Staggering x-over slopes. SO you might initially have a 12db/oct but after 1 octave you might add an extra 6db slope to make it 18db/oct. YOu can do this and ive seen it used but it is a very rare design.

lots of great x-over design questions im sorry i can not answer them.

maybe try some diy home audio sites

good luck

Edited by jas, 15 January 2006 - 12:41 AM.

[Cyberpunky]

the only way to correct phase shift is either physical time alignment or electronic time alignement
peace
Cyberpunky

[jas]

if you have access to time alignment therefore you are using digial x-overs too therefore you dont have the phase problems like the 90degrees out of phase setup when using a ANALOGUE 12db/oct with a 18db/oct two different drivers playing the same frequencies.

the phase is ONLY yes ONLY dependent on half the co-efficents used. This is not for cyberpunks car but for modern dsp using FIR (old pioneer odr stuff use IIR).
now for the hard part to work out the linear phase of a driver in an active dsp setup you would have to know the number of co-efficients used. NOw we have no idea since we use EQ and different slopes each side for each channel. This results in a very strange looking impulse response. ALso we dont know if the unit compensates for this linear phase problem by delaying each speaker relative to each other after we setup our system. BASically too many unknowns.

IN the end use time delay to setup the sound how your ears like it. Try with and without, play around with it and also try the recommended settings in the book with the distances and delay charts.