A quick question about all pass networks.

[5th element]

Delay for the tweeter a la Linkwitz.

For calculation of these things he says thus on his website.

fo>xover frequency

RC< 1/(2*pi*Xover freq)

Doing this all LW style, my midrange is approx 1" behind the tweeter. The tweeter requiring ~75us of delay.

My crossover is at 5000hz meaning my 'time' on the RC network cannot exceed 32us. This implies I require three delay networks.

My caps are going to be of value 5nF, which means a convenient value for R would be 6k = 30us.

Putting three all pass networks into LspCAD with the first two set to 5nF 6k, requires the third be set to provide 29us of delay to give the required massive reverse null.

Everything follows as predicted so far, much in accordance with what Linkwitz had to do.

Now, this is where my question comes in. I can create the deep reverse null with ONE delay stage, without RC > 30us.

This begs the question, are the other two stages needed?

Intuition tells me yes, because with one stage I have accounted for the 'phase' of the signal, but the drivers are still not 'time aligned'. The tweeters signal will reach my ears before the bass units does, but its phase has been altered such that both mid and tweeter are in phase with one another.

Adding in the other two stages, delays the signal of the tweeter even more such that the wavefronts now supposedly reach my ears at the same time, as well as satisfying the phase.

Is this correct or am I missing something?

[consort_ee_um]

If you are going to this trouble I would have thought it is worth the extra 2 stages to get time aligned Good explanation/discussion here:
http://www.geocities.com/kreskovs/TimeAligned1.html

[Pano]

All Pass? With your user name I would have thought the:

"Leeloo Dallas, Multipass" would have been better.

Analog delay / phase shift is an interesting subject. I'll watch this thread to see what you come up with.

[Ron E]

I say try it both ways. See if you can get a good reverse phase null without the allpass - then tell us which sounds better. I would not be surproised to hear it is the simpler crossover...

[5th element]

Originally posted by Ron E
I say try it both ways. See if you can get a good reverse phase null without the allpass - then tell us which sounds better. I would not be surproised to hear it is the simpler crossover...

Well in the case of simpler, the signal is going to go thru a load of op amps anyway so an extra two isn't going to change much, after all the sound has gone thru many of these before it reaches my ears in the recording stage.

I guess a picture may better illustrate what I am saying.

It seems logical that the other networks will do what 'needs' doing. If this is audible is open to debate and I will certainly try both ways.

[banana]

As long as u use traditional xover network (LR2, LR4, B3, elliptical, etc), phase/time compensation is not possible, or at least not exsist in analog domain.

There is too much undershoot cause by the highpass filter. So that no matter how u align the drivers back and forth, the step response still won't addup.

http://www.goldmund.com/technology/leonardo/

[phase_accurate]

I would not be surproised to hear it is the simpler crossover...

And I bet it is the contrary !!!!

So that no matter how u align the drivers back and forth, the step response still won't addup.

He is not trying to achieve a good step response otherwise he wouldn't use LR from ther beginning.
His intention is to avoid amplitude-irregularities due to the driver offset.

Regards

Charles

[5th element]

Exactly, what Charles said. Looks like I will be adding in the other networks then Now I need to find some more op amps

[banana]

Oooops... I read "time aligned" from your first post, but didn't read detail enough.

I use slant baffle instead. Less hardware involved, but the tweeter have to work off-axis.