Bending wave, low order crossovers and tradeoffs


Stereophile just reviewed an interesting bit of kit, the Manger P1. The mid-tweeter is a bending wave transducer, while the woofer is conventional. Crossover point is around 400 Hz.

The DIY community has a similar type of design called Woofer Assisted Wide Band.  It is 90% wide-band, plus a woofer.

https://www.stereophile.com/content/manger-p1-loudspeaker-measurements

Of course, you have to listen to make any sort of real assessment as to the value of the speakers and their suitability to your own home, I just wanted to share with the measurement readers how I might look at a crossover and the paths not taken so others can gain some insight into just how much is happening in the crossover design of a speaker.  Let's take a looksie ...


There are many who feel the best crossovers are 1st order (6 dB / octave) as high order rob dynamics or something. This is not an effect I have heard. Lots of 1st order, time aligned speakers I have not like at all, and one with active woofer system, was spectacular. So for me this is not a compelling sales pitch.


But still, lets say low-order or no-order filters is a very desirable characteristic, so lets talk about the negative consequences of having a very simple crossover, as apparent here.


Comb Filtering
Those armchair speaker designers who get frothy mad at driver arrays, claiming "comb filtering" when there's no evidence in the measurements are oddly silent when it's right in front of them. This is a good example. Look at figure 5. Plenty of comb filtering visible here.  What's going on? The low-order filters used is letting the woofer interfere with the mid-tweeter. Lots of great speakers do much better off-axis than here.


Next, lets look at the impedance chart, Figure 1. See that 3 Ohm dip around 200 Hz? With a low rise above 8 Ohms around 1,500 Hz? This is evidence of a minimalist crossover. It's quite possible that the woofer is run full-range with no low pass filter at all, and the mid-tweeter may have only a cap.


I will say that I do not like speakers with a dip in this region, as I find them quite demanding of amplifiers. My usual reaction is "WHY WOULD YOU DO THIS?" and then I am reminded that audiophiles LIKE demanding speakers. We are rather masochistic when it comes to the care and feeding of speakers. Some manufacturers deliberately drop the impedance in this range for exactly this reason.


Lastly, lets look at the overall shape of the output, Figure 4, showing a subdued mid-range. Not exactly recording studio attributes here, but possibly a good speaker at lower volumes. The shape here is a function of the cabinet, drivers and crossover design. No one thing contributed to the speaker's tonal balance, which we can lay it the foot of the designer, but if you choose to use a minimalist crossover as evidenced here there's only so much you can do to change things.


So, to recap, the designer picked a minimalist crossover and accepted poor lateral response and a low minimum impedance, and a lumpy frequency response curve as trade-offs. Or we could say he/she wanted all three.




Best,

E
erik_squires
PS - If you like the idea of a minimal crossover, may I suggest building the classic Seas A26 kit?

https://www.madisoundspeakerstore.com/2-way-speaker-kits/seas-a26-10-2-way-kit-pair-based-on-the-cla...

It has 1 capacitor and 1 resistor, and you can experiment by adding a 2nd order to the tweeter. Building a kit like this is a great thing to do with a friend or loved one you want to introduce to audio.


Best,

E
It's quite possible that the woofer is run full-range with no low pass filter at all, and the mid-tweeter may have only a cap.
I have never designed or built speakers, but am trying to understand this statement. In figure 3, it appears that the woofer is rolling out at or near 12 db per octave above 300 hertz. It is also showing no output greater than -40 db above about 3.5khz. How could it be running full range if this is the case? There would almost certainly be a lot of output above that with an unfiltered 8 inch driver. 

This is a good example. Look at figure 5. Plenty of comb filtering visible here. What's going on? The low-order filters used is letting the woofer interfere with the mid-tweeter. Lots of great speakers do much better off-axis than here.
How is the woofer contributing to this, if it's contribution is more than 40 db down at 3.5 khz? All of the comb filtering I see in figure 5 is well above that frequency. Are you sure that is not something associated with the way the bending wave driver operates? I have never seen measurements on one of these types of drivers before and am wondering if it is something due to it's design. Could the poor off axis performance be laid at the feet of the driver instead of the cross over design? I agree that the off axis performance looks bad. So does the spectral decay plot. 

Could this excerpt from the body of the review explain the comb filtering seen in the measurements?

Unlike most loudspeaker drivers, the diaphragm of the Manger Sound Transducer (MST) does not operate pistonically. Instead, voice-coil excitations generate transverse waves along its flat surface, like ripples in a pond.

Read more at https://www.stereophile.com/content/manger-p1-loudspeaker#Wb5r3k9kVdizr0LV.99


Lastly, lets look at the overall shape of the output, Figure 4, showing a subdued mid-range. Not exactly recording studio attributes here, but possibly a good speaker at lower volumes. The shape here is a function of the cabinet, drivers and crossover design
Again looking at figure 3, this would appear to be all due to the bending wave driver with likely some cabinet contribution. I am not trying to be argumentative here, rather trying to learn how to interpret provided measurements and understand how you reached the conclusions that you did.
It is also showing no output greater than -40 db above about 3.5khz. How could it be running full range if this is the case? There would almost certainly be a lot of output above that with an unfiltered 8 inch driver.

Depends on the woofer, really. I imagine a custom unit could be built that way, or which needed only 1 additional component. Also, the output may seem low, but it is not too low to avoid interference. Could that comb filtering be caused by the bending wave transducer? Possibly, but then why is it different horizontally and vertically?

Honestly the shape of the roll off measured, the impedance and lateral measurements aren’t aligning very well. If it’s using a 2nd order low pass, then why is it interfering as much, and why is the impedance curve the way it is?

Perhaps 1st order filters? It’s hard to know without knowing the impedance of the original drivers.

Again looking at figure 3, this would appear to be all due to the bending wave driver with likely some cabinet contribution.

Right, and if you have a very simple crossover, you don’t have the ability to flatten any such shape, so you are left with whatever the driver starts with.

My point here was to reinforce something I said elsewhere, most crossovers incorporate some amount of equalization, but very simple circuits have to avoid them.

Though it can be said that even a 1st order filter, judiciously picked, can do both EQ and filter.
Could that comb filtering be caused by the bending wave transducer? Possibly, but then why is it different horizontally and vertically?
Look at the scales on the graphs. The horizontal goes fully to 90 degrees off axis. The vertical only goes to 15 degrees off axis. If you look at the plot line on the horizontal that would correlate with 15 degrees, it looks suspiciously like the vertical at 15 degrees. 

Right, and if you have a very simple crossover, you don’t have the ability to flatten any such shape, so you are left with whatever the driver starts with.

My point here was to reinforce something I said elsewhere, most crossovers incorporate some amount of equalization, but very simple circuits have to avoid them.
After my first post in the thread, I went back and read the reader comments at the end of the review. There is a lot of good discussion about this very thing and some of the problems with equalizing this particular type of transducer. 
See that 3 Ohm dip around 200 Hz? With a low rise above 8 Ohms around 1,500 Hz? This is evidence of a minimalist crossover.
no its not. I see no evidence it has to do with the crossover. 

The review by John Atkinson ends with the following

The p1's measured performance nicely correlates with its sonic character.—

So Erik, this looks to me like sour grapes. Lets have a look at your designs to see if theres any comb filtering. You yourself used first order slopes on your own design if i recall correctly. 

Kenjit:


I am not sure who you are talking about:

1 - Not sour grapes. As I said before, the value and sound quality of the speaker cannot be determined by specs or measurements.


2 - You’ve accused me of using 1st order crossovers before, and I’ve corrected you before. I have never designed a speaker with 1st order crossovers. Maybe you don’t know what that means?


3 - So JA writes a non-committal "they sound like they measure" sentence and you think that puts us at odds? No. JA is an awful and biased speaker reviewer, but this statement is so bland and nondescript you can’t possibly take its meaning.


Erik

Erik,

For some really interesting technology, look beyond the Manger to the MC AudioTech Forty-10 speaker.  I heard it at CAF 2018 and it was pretty much my best of show.

"Hidden in the high-frequency spaced array are 10 identical bending- wave drivers 'of the latest generation,' covering everything above 100 Hz. In its Folded Cube low-frequency enclosure reside two 'industrial type' woofers in a dipole arrangement. The speaker uses a dedicated hybrid external crossover with low-frequency level/contour controls."

So if the comb filtering is solely the result of the mid-tweeter driver, that implies at least 2 discrete surfaces with a significant distance between them, no?

How very peculiar.
Hey @twoleftears - Where were you when we were talking bending waves a few months ago?? :)



I dunno. Frankly, I don’t really understand how they work. At CAF the designer Paul Paddock was there with a guy doing the dog and pony, and PP was pretty cagey about the details.

This is him:

https://patents.justia.com/inventor/paul-w-paddock

They were a real sleeper at the 2018 CAF, but now this year after the 2019 CAF everyone is oohing and aahing over them.  Go figure.

I'm probably going to offend people who actually know bending wave transducers, but I think the Ohm Walsh drivers are great examples.


In a normal cone, you want the whole thing to move, like a solid piston, but without the mass. :)  That is, ideally, the voice coil, dust cap and cone all move at exactly the same time.

The Ohm Walsh driver looks like a piston, but here's what's different: The cone bends. The signal starts at the voice coil and then, over time, travels down the length of the cone to the ends. I'm not even sure if the end of the cone is allowed to move or not, it may be fixed.

And to me this is the overall principle of a bending wave transducer. They are excited at one point, and then at some point in time the signal finally reaches the edges.

With a cone, ribbon, ESL, AMT the ideal situation, the signal reaches the edges of the driver at the same time it reaches everywhere else.

Of course, there are no absolutes, but that's the definition I work on.