Combating Pattern Flip

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Horns and waveguides increase their output by directing it into a narrow beam. For instance, a compression driver on a horn may produce as much as 110dB with a single watt, by focusing all it's energy in a narrow beam.

JBL_PRX425_2.png


One of the downfalls with this is 'pattern flip.' Pattern flip is when the waveguide or horn loses 'pattern control' on one axis but not the other

For instance, the waveguide on this JBL PRX 425 will control horizontal directivity down to about 1350Hz, but it's vertical directivity will collapse at about 3300Hz.

This is caused by the dimensions of the waveguide. It's waveguide is approximately 25cm wide, which means it will control the pattern down to 1350hz. (34,000cm per second / 25 cm = 1360Hz.)

'Pattern flip' can really make a speaker sound weird. You basically have an octave or two where the directivity of the speaker is completely asymmetrical. IMHO, this asymmetry makes it VERY difficult to come up with a crossover that sounds natural. When there are a couple of octaves where the directivity is all over the map, it's not possible to 'voice' the speaker properly.

In my personal experience, 'pattern flip' is one of those 'unsolvable' problems of loudsepeaker design. If you try and fix it with EQ, the speaker just winds up sounding weird. The only appropriate solution is to avoid pattern flip entirely.
 
  • Like
Reactions: 1 user
An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


The obvious solution: use a bigger waveguide. The JBL DD55000, M2, and 530 all go that route. The larger waveguide prevents pattern flip. For instance, the M2 waveguide controls directivity down to about 900Hz, in both the vertical and horizontal axis. The woofer begins to beam at about the same frequency, so you have a seamless directivity match at the crossover point.
 
There's a couple of challenges with the approach that's laid out in post #2:

1) sometimes you want an asymmetric pattern. For instance, my new listening room has walls that are 9' tall. But it's very deep and wide. This means that if I use a waveguide with a pattern of 90x90, I'm going to get a lot of reflections off the ceiling and the floor. For instance, if the waveguide is one meter off the floor, and the vertical pattern is 90 degrees, you're going to get a 'floor bounce' just one meter away from the loudspeaker. If you narrow the vertical beam to 30 degrees, the 'floor bounce' occurs at 3.73 meters in front of the speaker. Narrowing the beam from 90 to 30 degrees 'pushes' the floor bounce to nearly 4X further back, which is likely less audible.

2) Waveguides and horns with wide patterns can be difficult to cross over. For instance, if you have a compression driver on a 60x40 waveguide, it's only "working" about 25% as hard as a compression driver on a 100x100 waveguide. (Because it's radiating into 1/4 the space.)
 
aGAujNi.jpg


01xZAMm.jpg


H4lJhB2.jpg


It occurred to me that you can combat pattern flip by using a cylindrical enclosure.

For instance, this waveguide has a pattern of 120x80 and it's 30cm wide. That means that the horizontal pattern is controlled down to 1133Hz. In a rectangular enclosure, the same waveguide would 'flip' at a very high frequency - above 2khz - and the pattern flip would make the octave between 1-2khz sound "weird."

But by putting the waveguide in a cylindrical enclosure, you push pattern flip all the way down to 850Hz, where it's inoffensive.

Obviously, all of this is geometric. If you used an enclosure that's 15cm wide instead of 30cm wide, the horizontal pattern would be controlled down to 2266Hz and the vertical pattern down to 1700Hz.

The only real limitation that I can see, is that the horizontal pattern must be greater than about 100 degrees. This is due to the depth of the cylinder; if the horizontal pattern was 90 degrees, the throat of the waveguide would extend beyond the cylinder. So this design is particularly appropriate for waveguides with a wide horizontal coverage angle.
 
  • Like
Reactions: 1 user
So basically you build a huge horn and 'saw off' the sides which you do not need? I like it :) Would the polars measured along the diagonals still be OK?

Another solution would be to put woofers both above and below the waveguide and use Horbach-Keele filters. Vertical center-center spacing is a problem, so use a waveguide without a proper mouth roundover on it's top and bottom lip (like the Econowave waveguide) and use small woofers. In order for the woofers to remain beaming horizontally, multiple woofers might need to be arrayed horizontally.

Like this, but with an asymmetrical waveguide:
Image taken from Pseudo-coaxial with narrow directivity (and Horbach-Keele filters)
 

Attachments

  • _MG_2018_klein.jpg
    _MG_2018_klein.jpg
    145.6 KB · Views: 237
Last edited:
So basically you build a huge horn and 'saw off' the sides which you do not need? I like it :) Would the polars measured along the diagonals still be OK?

Another solution would be to put woofers both above and below the waveguide and use Horbach-Keele filters. Vertical center-center spacing is a problem, so use a waveguide without a proper mouth roundover on it's top and bottom lip (like the Econowave waveguide) and use small woofers. In order for the woofers to remain beaming horizontally, multiple woofers might need to be arrayed horizontally.

Like this, but with an asymmetrical waveguide:
Image taken from Pseudo-coaxial with narrow directivity (and Horbach-Keele filters)

9H1Eb48.jpg

Actually, the other project I have on my plate is almost exactly like that.

The reason that I went with a cylinder is because I have a really difficult time making decent looking speaker boxes.

I tried a dozen different ideas to make the box look good, but none of them were palatable:

1) I tried making a good looking box, by taking the time and effort to cut everything properly. But it still looked ugly.

2) I looked on Craigslist for speakers that I could use just for the box. Basically buy an old JBL, remove the drivers, and put my own in there. The problem with that is that a lot of the old speakers have ratty enclosures. And the ones that are in good shape, people want $500-$1000 for them.

Image_047.jpg

3) My current speakers are Vandersteens, and I came close to doing what they do: basically wrap the entire thing in grill cloth, so that there's no need for expensive / time consuming finishes.

4) Another idea I had was to buy a NEW set of speakers, and remove all the drivers and put my own in there. The problem with going that route is that most new speakers have a very narrow baffle. No good for a waveguide speaker.

After a lot of pondering, I'm leaning towards a cylinder. Because the shape is made in a factory, it's just about perfect. Ideally I'd finish it like my old Summas:

IMG_0661.JPG


Basically wrap the cylinder in a layer of fiberglass, then a layer of bondo. Sand the whole thing and then paint it.

If I'm not able to make the finish look good, at that point I could wrap the entire speaker in grill cloth a la Vandersteen.

That would make it look a lot like this:

hsusubsTN1220HO.jpg
 
Another solution would be to put woofers both above and below the waveguide and use Horbach-Keele filters. Vertical center-center spacing is a problem, so use a waveguide without a proper mouth roundover on it's top and bottom lip (like the Econowave waveguide) and use small woofers. In order for the woofers to remain beaming horizontally, multiple woofers might need to be arrayed horizontally.

To me, that sounds a lot like the Rey Audio Warp monitors :
WARP
 
if I use a waveguide with a pattern of 90x90, I'm going to get a lot of reflections off the ceiling and the floor. For instance, if the waveguide is one meter off the floor, and the vertical pattern is 90 degrees, you're going to get a 'floor bounce' just one meter away from the loudspeaker. If you narrow the vertical beam to 30 degrees, the 'floor bounce' occurs at 3.73 meters in front of the speaker. Narrowing the beam from 90 to 30 degrees 'pushes' the floor bounce to nearly 4X further back, which is likely less audible.

Of course with most living room acceptable speaker sizes, changing the vertical pattern to something narrow only keeps treble frequencies from bouncing off the ceiling and floor, since the directivity goes away below there -- a narrower pattern needs to have a larger dimension to keep the same control frequency, and the asymmetric waveguide instead makes the dimension smaller, pushing that frequency upwards with two effects. Keeping treble reflections off the ceiling is probably easier with an absorber up there.

About 5 years ago I almost got around to making a waveguide similar to your cylindrical cut idea --
Sketch%20of%20Assym%20Box.jpg


...but only 'almost'. Seemed like a lot of work and really big.
 
diyAudio Moderator
Joined 2008
Paid Member
Keeping treble reflections off the ceiling is probably easier with an absorber up there.
Yes it is. I went that way and reduced the reflections by about 3dB. The waveguide was much harder to build, and to get right, and it didn't help with the amount of space it occupied. I could make a better sounding round waveguide and the absorption was much more effective.
 
Hi Patrick,

Do you think your argument holds up for other types of controlled directivity speakers though? D'Appolito, line arrays, large ESL panels, etc.

These do the opposite of what you are concerned with. They control vertical, but not horizontal. Could that be key?

Erik
 
diyAudio Moderator
Joined 2008
Paid Member
Is this not a radial horn with its mouth wide open?
Almost. A radial is more like the Iwata, so this is slightly elliptical, maybe not so much like comparing a round horn with an equivalent square one.
iwata horn looks similar. i don't know the design principle of it though.
Imagine a 360 degree horn radiating about one plane, then quarter it.
 
...In my personal experience, 'pattern flip' is one of those 'unsolvable' problems...The only appropriate solution is to avoid pattern flip entirely.

Have you considered the interaction with the woofer?
Around the crossover frequency the effective radiator includes both the horn and the woofer.
The usual horn-stacked-on-woofer thus increases the effective radiator vertical size and should narrow the vertical pattern.
I suspect this can be done to eliminate or minimize pattern flip, if the woofer size, horn vertical size, woofer to horn distance and crossover frequency and slope are all correctly matched.
I haven't done any calculations yet but I expect it to work with the H x V around a ratio 1.6 : 1, say 72 by 45 or thereabouts.
Sound reasonable?

Best wishes
David
 
Have you considered the interaction with the woofer?
Around the crossover frequency the effective radiator includes both the horn and the woofer.
The usual horn-stacked-on-woofer thus increases the effective radiator vertical size and should narrow the vertical pattern.
I suspect this can be done to eliminate or minimize pattern flip, if the woofer size, horn vertical size, woofer to horn distance and crossover frequency and slope are all correctly matched.
I haven't done any calculations yet but I expect it to work with the H x V around a ratio 1.6 : 1, say 72 by 45 or thereabouts.
Sound reasonable?

Best wishes
David

image


Doing this stuff without a spreadsheet always gives me a headache, but let me take a crack at it:

In a speaker like the LSR708i, you have a center to center spacing of 15cm. If you used "single octave spacing" you'd have a crossover of 2267Hz. The LSR708i uses a 7" woofer, but the piston itself is about 6". So that means it will start beaming at (you guessed it) 2267hz.

So at the crossover point, the woofer has a vertical beamwidth of about sixty degrees. You can see this effect in the polar response of any decent woofer.

SB17NRXC35-4-freq.jpg


Here's the polar response of an SB Acoustics 7" woofer, SB17NRXC35. It's beamwidth doesn't "narrow down" to 60 degree until 3800Hz. At 2267Hz, it's beamwidth is about 120 degrees.

So I'm guessing that's probably part of the "JBL Secret Sauce" here: I'd speculate that you can get a seven inch woofer to start beaming at a higher than expected frequency, likely by manipulating the surround.

sb17mfc35-4.jpg

What I mean by that is that if you have a modern high-excursion woofer, and it has a big fat lossy surround, that driver may have the beamwidth of a fairly small piston. Just a guess.

The JBL waveguide, with a height of about 12.5cm, will start to lose beamwidth control at around 2720Hz.

At this point in my ponderings, I had to bust out the simulators, there's just too many variables here.

5SfYxk7.jpg

Here's how the vertical polars look in a design like the JBL, with a spacing of 15cm and a xover point of 2267Hz. (One wavelength spacing.)

WKDVupg.jpg

Here's how the vertical polars look in a design like the JBL, with a spacing of 15cm and a xover point of 1700Hz. (3/4 wavelength spacing.)

NjTDUe5.jpg

Here's how the vertical polars look in a design like the JBL, with a spacing of 15cm and a xover point of 1138Hz. (1/2 wavelength spacing.)

6PMLudO.jpg

Here's how the vertical polars look in a design like the JBL, with a spacing of 15cm and a xover point of 756Hz. (1/3 wavelength spacing.)


To me, it basically looks like vertical beamwidth gets wider and wider and wider as you push the crossover point lower and lower and lower. Naturally, there's a ton of variables which XDir ignores. It doesn't consider the directivity of either radiator, it assumes they're omni. It has no facility to introduce delay. You can manipulate the phase of the radiators, but as far as I can see, that only rotates the beam, it doesn't WIDEN or NARROW the beam. Just moves it up or down. Manipulating the phase at the xover point could be useful if you haven't compensated for the offset of the tweeter in the xover design.

Have I ever mentioned about how I hate making crossovers? :(
 
I just had a 'eureka' moment:

1) When the waveguide loses pattern control, it's beam becomes wide. AKA, "pattern flip."
2) When you have two devices radiating the same frequency, their beam narrows. See post #15 for sims.
3) The shape of the beam from two devices depends on a ton of variables: Are they both playing at the same SPL? What is their spacing? What is the phase difference between the two devices? What is the beamwidth of the two devices?

If it was 1990 you'd probably ignore all those variables and just pick a specific spacing for a specific crossover point. IE, you might space the woofers at a distance of 15cm and cross them over at 2267Hz. (One wavelength spacing.)

But in 2018, you can create a 3D sim in something like COMSOL, and then just go crazy with all the variables.

For instance, you could have a waveguide that's going omni at 2720Hz, which then crosses over to a woofer at a specific frequency. The tweeter and woofer array will have a narrow beam (because they're interfering with each other.) Probably the most important variable is the slope of the crossover; by manipulating the slope you can 'tweak' how quickly the vertical beamwidth goes omnipolar.

I know Bill Waslo has written about this. Since he writes programs to model crossovers, I'm pretty sure he knows a lot more about this than I do :)

WKDVupg.jpg

For instance, here's what the beam of the two-way JBL looks like, if it uses a spacing of 15cm and a xover point of 1700HZ (3/4 wavelength spacing)

gJFPwOw.jpg

hmexWYX.jpg

At one-half octave below the xover, here's how the polars look. You can vary the vertical beamwidth by varying the xover slope. For the most part, it appears that the faster you roll the tweeter off, the faster the speaker goes omnipolar. So one could use a shallower slope to vary how quickly the beamwidth goes omni. Naturally, all of this is limited by the physical size of the array. The two pics above compare the same spacing, but note that the level of the tweeter is varied, to simulate the effect of a steeper slope. I did NOT include the effect of the phase shift, which introduces another interesting effect, because the phase shift appears to 'steer' the beam UPWARDS, or even create an on-axis null when the phase difference reaches 180 degrees.

Yowza this stuff is complex.
 
Last edited:
YO7bIxX.jpg


Wow the amount of weird things you can do with this seems unlimited. Just realized you can use a little bit of phase shift below the crossover point to point the beam DOWN towards the floor.

Then you can 'correct' that phase shift with the baffle of the woofer, by sloping it.

The net effect is that the beam is pointed towards the listener, but there's an off-axis null pointed at the ceiling.

history.jpg

I heard these once at CES. I generally don't like conventional two-ways, I like waveguides and horns. These didn't sound like conventional two-ways. I wonder if Wilson was doing some trickery like that in the xover? Might explain why the baffle is so aggressively sloped. (Besides the obvious, which is lining up the voice coils.)
 
Jeah, there seems to be lots of options to explore, following your threads with great interest :)

Ps. Wayne Parham jugless the crossover with his pi speakers so that nulls in the crossover region aid the vertical directivity. I'm sure you already know his thinking and writings already but here is one such thread: http://www.diyaudio.com/forums/multi-way/123426-horn-vs-waveguide-138.html#post1878402

The link points to a post by Tom Danley where he mentions the 1.6 ratio that David mentioned earlier this thread so I thought this was worth posting here.

Carry on :)
 
Jeah, there seems to be lots of options to explore, following your threads with great interest :)

Ps. Wayne Parham jugless the crossover with his pi speakers so that nulls in the crossover region aid the vertical directivity. I'm sure you already know his thinking and writings already but here is one such thread: http://www.diyaudio.com/forums/multi-way/123426-horn-vs-waveguide-138.html#post1878402

The link points to a post by Tom Danley where he mentions the 1.6 ratio that David mentioned earlier this thread so I thought this was worth posting here.

Carry on :)

I'm actually a bit of a noob when it comes to crossovers. I'm pretty good at fabrication and I have the 3D programs down cold, but crossovers have always been an uphill battle for me.

Thank you!
 
I just had a 'eureka' moment:

1) When the waveguide loses pattern control, it's beam becomes wide. AKA, "pattern flip."
2) When you have two devices... their beam narrows. See post #15 for sims...

Yes, that was the key idea, to let the two effects cancel each other out to maintain more or less constant beam-width.
I don't think one can just do all the horn directivity work and not consider the interaction with the woofer.
But yes, that results in a lot of variables to play with - the 1.6:1 H:V beam-width was an idea to limit the search space a little.
Not only mathematically simple but 72 by 45 seem reasonable practical values, decent sweet spot horizontally, not too much bounce off the floor, not so different as to lead to an impractical horn shape.

The link points to a post by Tom Danley where he mentions the 1.6 ratio that David mentioned earlier this thread...

The 1.6:1 ratio should be the Golden Ratio but I was hesitant to mention that because it's almost an audiophile fetish.
I was amused to see that it occurs in the maths quite naturally in this situation, curious to read Tom's comment but the link takes me to a post by "Zilch" instead, some problem?

Best wishes
David
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.