lattice filter question

[eccdbb]

Hi, I have a few questions on fixing the phase changes that a crossover creates using lattice filters.

Can you connect multiple lattice filters together to get greater maximum phase difference? For example if X is a lattice filter, X X X R_load.

Do you calculate the lattice filters for R_load regardless of how many there are?

[speaker dave]

You can definitely cascade a number of latice filters together. I assume you are talking about the style where the 2 capacitors are a straight link and inductors are a crossover link (or vice versa). One latice group gives a 180 degree phase flip and the group delay that goes along with it.

As to impedance effects of a multiple string, I would give a qualified yes to using the end resistor as a calculated load all the way through. I haven't actually built or simulated a multiple, but I think that a single lattice looks resistive at the input. If that is the case then you are good to go because a latice plus resisitor is a resisitor, so 2 latices plus a resistor is still a resistor, etc. etc.

It would be good to try and simulate it just to be sure. That may be difficult since a lot of simulators like 3 terminal layouts (ladder networks). Perhaps someone else has played with that?

David S.

[bear]

What are we trying to acheive?

Time delay?

Bandpass??

High slope?

_-_-bear

[speaker dave]

Originally Posted by bear

What are we trying to acheive?

_-_-bear

The latice is a pure all-pass, so phase shift or time delay.

http://en.wikipedia.org/wiki/Lattice_phase_equaliser

[bear]

Sure, but I was asking the OP... to see what he was trying to do.

_-_-bear

[eccdbb]

What I'm trying to do- fix the phase part of the transfer function caused by the crossover. My ultimate goal is to have a N-way speaker that has perfect impulse response. For example in a 2-way, if you make sum(magnitude(frequency response)) flat via a simple 2nd order crossover, then next up is making the phase for both 0. Then the transfer function for the system will be ideally 1 for the audible frequency range.

[bear]

That's what everyone wants.

Best to not try to do it in passive, though... therein lies the rub... "n" active elements added to the signal path... and you can't really correct everything.

In DSP it starts to get really possible, but then you have inserted "n" DSP active elements in the signal path... which is starting to get closer to transparent. But beyond my ability to engineer, even if it can be made transparent.

_-_-bear

[eccdbb]

I suppose in DSP you could create a looong filter and convolve, but that's just intense.

[phase_accurate]

The best that you could achieve is time-alignment to some degree, but I doubt that you could achieve perfect impulse response with these when used within a passive crossover.

You could actually do group-delay equalisation (that's what they were originally intvented for actually !) with these passive allpasses but this would give high losses. So it is best done on the low-power side of things. There is one Swiss manufacturer of active studio monitors that uses group-delay equalisers. But they are made with active 2nd order allpass sections in front of an active crossover:

PSI AUDIO: Technology CPR:Compensated Phase Response system, providing a constant group delay response

Regards

Charles

[speaker dave]

Originally Posted by phase_accurate

You could actually do group-delay equalisation (that's what they were originally intvented for actually !) with these passive allpasses but this would give high losses.

Regards

Charles

Why is this? We are talking about unity gain all-passes.

The real issue is circuit complexity required. It may not be practical but it certainly is possible.

David S.