question on phase

[lord_darkhelmet]

I've looked around and can't find an answer that I understand. Hopefully some kind soul can break it down for me.

I took some measurements of well known manufactured speakers and compared them to some I designed. The measurements were done at the same time under same conditions. They are useless below 500hz but I do not believe that has any importance in this exercise. The measured phase data was stripped off and re-calculated before appropriate amount of delay was modeled back in. This was worked out by taking measuremeants of each driver without any crossover and comparing to a measurement of both drivers connected. Delay was added to the midrange until the summed response matched the measured response of both drivers.

First up is a direct on-axis test of the tweeter. Shown is the mid/tweeter curves and their summation.



..And here is a shot of the phase data.


Notice how the phase wraps around at almost the exact crossover point between the mid and tweeter.


Now here is a shot of my design and phase data, same test conditions.



Mine clearly do not wrap the phase at the xo point of 2300hz. Question is - what does that translate into for sound or is it not important? The mid and tweeter of my design do have reasonably close phase response at the crossover point which I thought was the important part.


At any rate, hats off to the guy who designed the commercial speaker I tested. Getting the drivers to follow near perfect slopes while maintaining the phase response must not have been easy.

[peter_m]

...hats off to the guy who designed the commercial speaker I tested...

I prefer your design... your phase looks perfect. What is important is the phase of both the tweeter and woofer to be close matched around the x-over area...

[Dave Bullet]

What do the impedance curves look like?

[lord_darkhelmet]

I can post those after work. Is there a connection ?

[Davey]

I'd agree with Peter and say yours looks pretty good.

That looks like a higher-than-4th-order crossover you've created so you shouldn't see a 180 degree phase wrap at the crossover frequency. It should be nearer zero degrees at that frequency with 180 wraps above and below the xover frequency. That's just about what you have.
(A 4th-order crossover would have a 180 phase wrap at the crossover frequency.)

The midrange phase curve matches with the tweeter phase curve over fairly wide range so it looks pretty good.

Cheers,

Dave.

[Deward Hastings]

It’s hard to judge a crossover design based on a single characteristic like phase tracking around the crossover point. A notch filter (or an elliptic filter) will mess with tracking, but the benefits (or avoiding other “problems”) may justify it. Even at the crossover point a designer might opt for a phase mis-match to steer the lobe, or simply because, when final voicing was done, the whole thing just sounded better.

As long as the tracking is reasonably good around the crossover point exactly where total phase (relative to what?) passes through “0” doesn’t matter, since the “0” is arbitrary . . .

[Curt C]

Notice how the phase wraps around at almost the exact crossover point between the mid and tweeter.

Mine clearly do not wrap the phase at the xo point of 2300hz. Question is - what does that translate into for sound or is it not important? The mid and tweeter of my design do have reasonably close phase response at the crossover point which I thought was the important part.

Phase wraps are arbitrary…
Try this: Change the phase of both drivers in your CAD program. The phase plots will still show the same relative phase data, but the wraps will move by 180 degrees.

Where the phase wraps on the plot is just a function of the way the phase is displayed on the graph, and has no bearing on the perceived sound. Since the vertical axis of the plot only shows +/-180 degrees, when the phase goes to -180 degrees it 'wraps' up to the 180 degree. If the plot’s scaling was changed to display +180 / - 720 degrees, you'd see both phase plots continuously going down and no phase wraps. Between these two plots the data doesn’t change, just how it is displayed. Consequently, while the phase wraps are an explicit indication of phase tracking at 180 degrees, they are otherwise meaningless. It is far more important the phase closely track over the entire stop bands of the drivers.


C

[lord_darkhelmet]

Thanks for the responses. To illustrate it for myself I generated several active crossover transfer functions in the CAD software without any actual frequency response data or inter-driver delay. The phase data matched my real design (pics are in my first post above) perfectly when I simulated an under-lapped 6th order butterworth transfer function.


So provided the individual driver's phase response agree near and at the crossover point, is it fair to say:

The vertical radiation pattern or "axis tilt" is not a result of phase directly but is a result of the chosen crossover point and slope, plus all of the fun variables like horizontal/vertical driver spacing on baffle, delay time, and of course the physical properties of the drivers themselves.

[Deward Hastings]

The vertical radiation pattern or "axis tilt" is a result of the relative phase between the two drivers, which is influenced by driver spacing on baffle, delay time, differing acoustic centers, physical properties of the drivers, and of course the crossover itself.

[Dennis H]

The vertical radiation pattern or "axis tilt" is not a result of phase directly but is a result of the chosen crossover point and slope, plus all of the fun variables like horizontal/vertical driver spacing on baffle, delay time, and of course the physical properties of the drivers themselves.

It is directly a result of phase and phase (on whichever axis you choose of measure it) is a result of all those other things.