Why is ”bad phase tracking” really that bad?

[lasse]

Hi all,



A thought came to me when responding in a thread concerning work on a ongoing system.



Ideally, it is stated in litterature, on websites and on this board,that the phase of the two drivers in the overlap region should be as alike as possible, resulting in the predicted net combined output over the region.



Well, what happens if you design the XO-parts to make the outputs as close to the targets, corresponding to the chosen alignment, as you can and you still get a discrepancy in phase from the two drivers?



And better still if, in spite of this, the measurements show a good response over the overlap region.



My question is: Is this really a problem and if so: What bad results will come up when your drivers are not in phase in the overlap region?



I illustrate this with a few pictures from my ongoing project:

-The total design

-The measurement

-The targets for the two drivers



Best regards//lasse

[Chris Roemer]

From your 1st graph, I can see your Fc is near 2.5kHz (don't know if this is a sim, or actual measurement... whatever).
Right at your Fc, it looks like the 2 drivers are "summing" +2-1/2dB (LR is typically +6, and BW is +3).
I can also see that at Fc your drivers are OOP (out of phase) by about 100*. Theoretically, I'd reverse the tweeter to get them only 80* OOP. (They'd probably sum +3.5dB @ Fc then.)

(With good "tracking") your (black) summation line should stay above the individual driver plots through the XO region. You've got interference around 1kHz and again around 5-6k.
(The drivers are summing in a negative way.)

IF these are measured plots, then this is what your FR will be EXACTLY WHERE YOUR MIC was when you took them (if you put yours ears right there).
You must remember that the phase relationship (IN / OUT - and everywhere in-between) continually changes w/changes in listening postion. For vertically aligned drivers (TM or MTM) the phase changes more between a standing and sitting position than from side-to-side. On a horizontal MTM (CC for example), the phase changes mostly horizontally (think of sitting in different positions across a long couch).

IF your phase plots sat on top of each other (target for LR slopes) throughout the XO region, then there's less (noticeable) change in phase relationships as you move off-axis (compared to YOUR current situation. SOMEwhere (moving L to R, or between standing and sitting) these drivers most likely WILL sum +6dB, while other positions might actually give you a "null"). Listening to a horizontal MTM (CC) already has you listening to the drivers from 2 differnt angles (the distance between your ears!).

You can get better "tracking" (resulting in less variation in summing w/respect to listening postion) by flipping the tweeter polarity, AND/OR bumping EITHER filter order UP/OR DOWN.
If your tweeter is "marginal" (distortion rising approaching its Fc) then I'd bump its filter up by one order and see how it tracks. If you tweeter is "safe" (or very robust) but this is an Al coned mid w/a bad breakup, then I'd probably up the LP filter by one order and see. This is the greatest power of running sims IMO. WAAY easier to hash all this out in a sim onscreen than swapping parts on a breadboard.

Does your sim software make use of (X, Y, Z) driver offsets?
SOME of the softwares can actually show you the FR change moving different amounts off axis (I think some can generate polar plots as well).

[lasse]

Thanks Chris,



As always I value your comments and I am confident that by posting and discussing on a general subject, such as this, we all learn.



Sometimes it also helps to see your thoughts in writing and looking on the one moer time…



I should have been clearer on that

-The total design (is a simulation, in Speaker Workshop))
-The measurement (is a measurement, in SIRP)
-The targets for the two drivers (Is a simulation in SW and baseline for optmization against targets, LR2 2500Hz)



What I cannot get into my head is, given a measurement looking that “decent” as the one in this post, what is wrong with two drivers contributing to a common output but in different phase angles depending on the frequency. Note that the difference in phase between the drivers are more or less constant over the overlap region.



Please correct me where I am wrong in my reasoning, whereas this is neither the first nor the last time this happens…



Regards//lasse

[dkalsi]

Chris R,

Thank You! Very well written and educational post.

D

[jhollander]

With the appropriate phase alignment, you can understand if or where off axis peaks occur. In other words there's nothing wrong with flat on your design axis, but unless the phase alignment is appropriate you might have peaks off axis.

[fpitas]

As a sort of historic note, many vintage speakers used Butterworth alignments, and the driver acoustic outputs were designed to be in phase quadrature (90 degrees apart) on the normal listening axis. The big flaw in that design was the big off-axis peak (3dB) where the drivers were in phase. Hopefully it didn't point at a listener, or reflective surface.

[Dave Bullet]

In my opinion :-) (here we go)....

Phase is about trying to get the drivers to play in unison over as wide a frequency range as possible and as wide a listening axis (vertical and horizontal) as possible. Why in phase? For an "unsmeared and unaltering" sound. If two drivers are playing out of phase, the audibiiity of this (I don't have the education or the data) will become apparent - depending on the degree of "out of phaseness" and frequency response (as our ears are more susceptible at different frequencies). Thats the effect of lobing as you move around and cross in phase and out of phase lobes.

Thats why designers emphasize various off axis measurements. In a real room environment - you want a smooth power response - dips (esp in the 2 - 4Khz region I've found) are far more acceptable than peaks and make for a non-fatiguing speaker.

You basically want to walk around your room - left to right (45 degrees either side) and not hear the music "peak and dip". It's understandable higher frequency drivers will roll off - so a relaxing of upper treble is expected.

[Chris Roemer]

lasse,

Looking (roughly) at your phase plot (+/- 2octaves from Fc):
near 600Hz: your drivers look to be about 130* OOP (OUT of phase) - reversing the tweeeer would actually get them closer (only about 50* out)
1.2kHz: still about 110* OOP
2.5k (your Fc): this is actually your "best" phase situation - I THINK they're about 90* "out" right here. As Dave said, this "used" to be the "standard" (Butterworth) alignment. Called "quadrature" since 90* is one/forth of a full 360. Two drivers that are "spread" 90* in phase add (or "sum") +3dB (as, indeed, yours are very close to)
5k: Looks like 120* OOP to me. Again, I'd probably reverse the tweeter if I HAD to use this XO.
10k: again, about 120* out.

(also building off what Dave said) Spatial "imaging" (the ability for a stereo pair to "sound" like different instruments are occupying different locations (across the room) out in front of you) depends on accurate phase "duplication" (trying to replicate what you'd hear if you were "live" at the recording session).
What makes something sound like it's "stage center"? I propose that it's because both sounds reach your (two - spread) ears from 2 different speakers at the same time. As a sound moves left (on "stage"), it gets closer to your left ear (and further from your right). By introducing phase errors (RIGHT at the source(s), your speakers), the phasing (or "timing" -essentially) of these signals dances all over the place, throwing good imaging out the window.

[Dave Bullet]

If you persist at crossover modeling, you will get good phase tracking. This usually involves being flexible about moving the Fc to get the drivers to naturally "gel" without overdoing crossover components. Let the drivers tell you what works as you move and change things. It usually involved in a crossover component reduction as you are not "fighting the drivers". In your case though - you're already at minimum and its a good place to be with tweaking to adjust

There are times though that a specific Fc is demanded for. In my case I have my drivers and "forcing" them to work together. This may not be optimal but it is where I am. I won't go into the details as this is your thread.

Suffice to say, with persistence, nice acoustic targets can be reached. With drivers that have bad breakup nodes, this demands either the ability to measure or guaranteed QA / consistency in manufacturing (so you can use others measurements).

It may be at the expense of extra crossover components to "beat" the drivers into your intended acoustic target..

I ask about metal drivers as they add complications that need to be considered. HAving non-linear distortion measurements are very useful to guide a maximum upper limit depending on breakup severity and especially H3 distortion product peak further down (primary breakup / 3).

[Dave Bullet]

Below is my 3 way with phase data included. I had to use HBT extracted min phase as I could not use measured phase below my gated frequency to adequately design the W/M crossover. The curves are smoothed to 1/48th octave as that is what the HBT extraction process did.

As you can see - woofer / mid line up nicely. This hit the LR2 target I was going for. I needed LR2 for healthy impedance in the midbass

M/T isn't as good. I wanted steeper slopes for 2 reasons. 1. Midwoofer breakup node. 2. The seas tweeter did not like 2KHz (I don't care what people say - it wasn't forgiving on certain material and dominated the upper mid / lower treble with a harsh sound).

[lasse]

Hi,

I just realized one (embarrassing, perhaps one of many…) mistake in my modelling.

I have not extrapolated the tweeter response 2nd order at low fq´s before extracting phase……

Will be back with better phase information.

Regards//lasse

[jhollander]

I would add the baffle to the tweeter and woofer to improve your prediction.

[scottvalentin]

Below is my 3 way with phase data included. I had to use HBT extracted min phase as I could not use measured phase below my gated frequency to adequately design the W/M crossover. The curves are smoothed to 1/48th octave as that is what the HBT extraction process did.

As you can see - woofer / mid line up nicely. This hit the LR2 target I was going for. I needed LR2 for healthy impedance in the midbass

M/T isn't as good. I wanted steeper slopes for 2 reasons. 1. Midwoofer breakup node. 2. The seas tweeter did not like 2KHz (I don't care what people say - it wasn't forgiving on certain material and dominated the upper mid / lower treble with a harsh sound).

That looks really good - I recall seeing your thread on HTGuide - you've put a lot of work into that crossover and it shows. Nice job!

[Mark65]

Below is my 3 way with phase data included. I had to use HBT extracted min phase as I could not use measured phase below my gated frequency to adequately design the W/M crossover. The curves are smoothed to 1/48th octave as that is what the HBT extraction process did. As you can see - woofer / mid line up nicely. This hit the LR2 target I was going for. I needed LR2 for healthy impedance in the midbass M/T isn't as good. I wanted steeper slopes for 2 reasons. 1. Midwoofer breakup node. 2. The seas tweeter did not like 2KHz (I don't care what people say - it wasn't forgiving on certain material and dominated the upper mid / lower treble with a harsh sound).

I think the mid/tweet relationship is pretty darned good. Sure, the vertical "wrap" lines are not directly on top of each other, but the tracking above and below Fc is spot on. Lasse, my opinion on phase relationship is much the same as others here, in that having it track as closely as possible at least an octave above and below Fc is the goal. Will you reach that goal? And if not is it a problem? Only you can answer that, based on your preferences and your listening area. I also agree with Dave about letting the drivers tell you where they want to cross. I always start with a goal, and I sim xo's for the individual driver targets without looking at the overall summation, then adjust from there for phase and power response and largely ignore the individual targets and original Fc goal (while paying attention to potential driver issues, of course). So the end result is sometimes significantly different from the initial target. Now, if you want to see what difference your phase relationship makes, do some off axis measurements, and observe any differences at Fc. Then decide for yourself if they're objectionable or not, and whether is worth fussing over.

[dlr]

I wrote a small program that allows you to examine and experiment with different acoustic crossover combinations (ideal drivers, of course). It's interesting to see the summed response of differing crossover types, slopes and offsets with sometimes surprising results.

WinFilters

dlr