Crossover and phase importance

[BossBunos]

Hi, I'm new in the speaker DIY world and i'm trying to create my first speaker.
I made a 2-way prototype to learn about measuring drivers and creating crossovers.
In this example i'm using the SB26ADC tweeter in an augerpro waveguide and a NBAC15 5" woofer in a sealed enclosure. Again this prototype is just for testing and learning. I'm planning to add a third way with a 6" NBAC woofer from SB, but that's out of scope right now. I measured the horizontal and vertical axis of both drivers. If I just add an active LR4 high pass/low pass on the tweeter and woofer @2Khz I get the following phase result in vituixcad:

That seems pretty much aligned. However when i create a passive crossover that results in good directivity and on axis response there is a phase mismatch. What are the implications of this mismatch, is it a bad thing? It's a concept I don't understand yet. I gues it's because the slopes are not symmetrical?

 

[Keith_W]

Remember that the slope is a convolution of the mechanical bandpass of the driver with the electrical bandpass of the crossover. Since both are minimum phase, the result is additional phase rotation.

Phase mismatch at the XO point results in cancellation. I assume that the top left graph is woofer (green), tweeter (red), and summation (black). There is no cancellation at the summation.

 

[DVDdoug]

I've forgotten most of my filter theory but the low-pass creates a phase-lag and the high-pass a phase lead. If there are 90 degree phase shifts at the crossover frequency (double-pole filters?) the two waves will be 180 degrees out-of-phase at the crossover frequency. The solution is to flip the polarity of the tweeter (or the midrange in a 3-way).

 

[dalexander]

Seems like neither response is getting at what op is wondering which I often wonder myself.

Is there an ideal phase graph you want to see? Or is it simply whatever it takes to get the simulated response you want?

I’ve made multiple crossovers and one had a phase graph that was very tightly aligned with each other while the other not at all, while both crossovers certainly sounded different (one was 3rd order filters while the other 4th) neither sounded clearly “better” so what’s the goal to aim for? Or is it purely contextual in the real world.

 

[Lars Risbo]

the phasing of the xover mainly affects the vertical polar pattern. phase match at the xover frequency gives a main lobe pointing straight forward and is what you get eg with ideal LR filters. You can probably adjust your passive filter to get better phase match. one simple test is to flip the polarity of one of the drivers and measure on axis. if phase is matching you get a deep null at the xover frequency (aka reverse null)

 

[Salt]

Correct. For in-phase-crossover: - 6 dB at xover frequency for both transducers and a deep notch there when one of the two is inversed in polarity

 

[Keith_W]

Is there an ideal phase graph you want to see? Or is it simply whatever it takes to get the simulated response you want?

I will start by defining some terms and then i'll present both sides of the argument. First, there is a difference between intrachannel phase distortion (phase distortion in one speaker, playing in mono) and interchannel phase distortion (phase of one speaker is different to the other). Nobody disputes that the latter is audible - wire one speaker out of phase, and you will easily hear it. The debate is about intrachannel phase distortion, which is what we are dealing with here.

Traditional thinking: it's not audible. There is something called "Ohm's Acoustic Law" which says it isn't. Linkwitz and Toole believe that it isn't. Multiple studies, which all used minimum-phase speakers and electronics (which by themselves distort phase), have failed to demonstrate a difference in normal listening rooms. It is only audible with special test signals, or with headphones, or with speakers in anechoic chambers. Phase only matters if a phase mismatch results in cancellation.

More modern thinking: it is audible. For this, you need to compare it with a linear-phase filter, which you can only find with digital FIR filters. Pay attention to what JJ says in the last 10-15 pages of this thread.

 

[bmc0]

when i create a passive crossover that results in good directivity and on axis response there is a phase mismatch. What are the implications of this mismatch, is it a bad thing?

The mismatch at the crossover frequency is only about 45°, which doesn't affect the summation very much (~0.7dB). If you look at the predicted vertical directivity around the crossover frequency, you should find that the main lobe is not symmetrical.

You can see that the ~110° phase mismatch around 6-7kHz causes slight cancellation (i.e. the summed response is below that of the tweeter alone).

 

[dalexander]

I will start by defining some terms and then i'll present both sides of the argument. First, there is a difference between intrachannel phase distortion (phase distortion in one speaker, playing in mono) and interchannel phase distortion (phase of one speaker is different to the other). Nobody disputes that the latter is audible - wire one speaker out of phase, and you will easily hear it. The debate is about intrachannel phase distortion, which is what we are dealing with here.

Traditional thinking: it's not audible. There is something called "Ohm's Acoustic Law" which says it isn't. Linkwitz and Toole believe that it isn't. Multiple studies, which all used minimum-phase speakers and electronics (which by themselves distort phase), have failed to demonstrate a difference in normal listening rooms. It is only audible with special test signals, or with headphones, or with speakers in anechoic chambers. Phase only matters if a phase mismatch results in cancellation.

More modern thinking: it is audible. For this, you need to compare it with a linear-phase filter, which you can only find with digital FIR filters. Pay attention to what JJ says in the last 10-15 pages of this thread.

Thank you for the info I’ll check this thread for sure, I’ve yet to mess with FIR filters but would really like too, my currently journey is working on active filters as opposed to passive ones.

 

[Lars Risbo]

lets not confuse polarity and phase

 

[terryforsythe]

The solution is to flip the polarity of the tweeter (or the midrange in a 3-way).

Not with 4th order LR filters, which the OP is using.

 

[bmc0]

I’ve made multiple crossovers and one had a phase graph that was very tightly aligned with each other while the other not at all, while both crossovers certainly sounded different (one was 3rd order filters while the other 4th) neither sounded clearly “better” so what’s the goal to aim for? Or is it purely contextual in the real world.

Assuming nonlinear distortion was not a significant factor, the differences you heard were probably due to differences in on-axis frequency response and/or directivity around the crossover frequency.

Traditional thinking: it's not audible.

To be clear, I don't think that anyone who is well-informed would argue that intra-aural phase distortion is never audible (this idea can be trivially proven false, as j_j has pointed out many times), just that the kind of phase distortion you get with "normal" crossover slopes isn't audible, at least in real-world conditions.

 

[terryforsythe]

That seems pretty much aligned. However when i create a passive crossover that results in good directivity and on axis response there is a phase mismatch. What are the implications of this mismatch, is it a bad thing? It's a concept I don't understand yet. I gues it's because the slopes are not symmetrical?

Drivers have complex impedances, including phase shift. Passive filters themselves introduce phase shift. Moreover, the phase shift introduced by a passive filter is dependent, in part, on the load impedance, which in this case the impedance of the driver. With passive filters and real-world speaker drivers, it is challenging to get the drivers perfectly phase aligned while also obtaining the desired crossover slopes. Back in the day I spent many hours in LEAP modelling various speaker systems. Some driver combinations worked out better than others.

If you have resistors on your tweeter filter to get the SPL aligned with the woofer, you can try moving those, and perhaps use an L-pad (series resistor and resistor L-G).

You also can add impedance compensation as the last stage of the crossovers to try to bring the load impedances closer to one another at the crossover frequency. Essentially, you are changing the load impedance as seen by the filters. That may help, but it will affect the tuning of the other crossover components, so those would have to be tuned to account for the new impedances.

Another thing to try is playing with the values of the last elements in the 4th order filters. Sometimes that can help change phase shift without really messing up the frequency response. If it is effective, you can go back and fine tune the values of the other filter elements.

Finally, if you just want to simplify things, go all active. Tuning is much simpler, and there are other benefits as well. The downside is that you need to add active crossovers (preferably with DSP) and additional amplification.

 

[terryforsythe]

Also, try removing the last inductor in the tweeter filter to see how that affects the tweeter's phase. It will be the L-G inductor closest to the tweeter. In VituixCAD you can just use the icon to open that component, which effectively removes it.

 

[Jukka]

VituixCAD simulates phase into the frequency response graph, so it's easy to not f... go wrong with it. You can check how inverting driver polarity affects the FR by selecting a driver in the crossover view and clicking Invert. You should immediately see the null everyone talks about.

Also check that you got driver positioning X Y Z set, because that will affect responses a lot. Always check by measuring.

If you are interested in phase in an academic way, go ahead, but with LR4 you are set for a good crossover.

 

[Salt]

Vituix is a simulation to be crosschecked in reality...

 

[BossBunos]

Thanks for all and the many replies. It looks like the phase is more of a result of a crossover rather than a goal if I interpret the comments correctly. I can indeed see that changing the components values changes the horizontal and vertical directivity and phase. But as long as the directivity is as I desire the phase doesn't matter much. Like @bmc0 pointed out the 45 degree mismatch at the crossover frequency can also be seen in the horizontal directivity chart @4khz and the listening window response.
The first image I simulated with the active LR4 crossover @2khz resulted in an aweful horizontal and vertical directivity plot in general and shows that phase optimalization alone is not a good option. It's more a nice to have.
Eventually it's a good (flat) on axis response with a (as good as possible) smooth dispersion that we're aiming for. And this will automatically result in a aligned phase response in the ideal world

 

[BossBunos]

This is the result with the tweeter inverted. I'm not sure how deep the null is supposed to be. To me it looks pretty deep with -7.3dB

 

[ThatSoundsGood]

This is the result with the tweeter inverted. I'm not sure how deep the null is supposed to be. To me it looks pretty deep with -7.3dB
View attachment 440103

Is this a simulation or an actual measurement?

 

[BossBunos]

Is this a simulation or an actual measurement?

Crossover simulation based on actual polar measurements of the drivers in a prototype cabinet