On the directivity of dipole tweeters

[Rudolf]

Thanks Michael,

I wanted to write it down for people who feel more at ease with pretty pictures and diagrams than with math equations (where I'm rather limited too).

These days I happened to get a glimpse of some Orion++ off-axis measurements. Above 1 kHz they don't look terribly different from the simulations I show in fig. 4.13. At 1-6 kHz both are very much in the same 6 dB corridor. That makes me believe that I'm not completely off the mark.

Rudolf

[twest820]

Originally Posted by Rudolf

Your toroid needs to separate the front and rear radiation of the dipole before the blossoming starts. Too early, and you get a premature beaming - too late, and you get too many side lobes.

Yup; the question I'm interested in answering is whether a good balance exists. Results from horns suggest there should be such a profile, though finding it seems likely to be nontrivial.

The Neo3 on the Note is crossed at 6kHz whereas I'm curious about the possibility of lowering the cross from the 1.8kHz my nude Neo3s are currently set to. A two or three octave shift in target frequency should allow some freedom for a different waveguide design, particularly as I'm not trying to match the high frequency directivity of a ScanSpeak 10F and the crossovers are quite different.

Originally Posted by Rudolf

That makes me believe that I'm not completely off the mark.

I've been meaning to ask; what tool did you use for the simulations in chapter 4?

[Rudolf]

Originally Posted by twest820

... what tool did you use for the simulations in chapter 4?

4.1 is done with Edge. Chapters 4.2 und 4.3 are simulated with Boxsim, a german program that has been developed in close partnership with the Visaton driver factory. The factory measured phase, response and impedance data of most Visaton drivers can easily be imported. Boxsim has not been developed with dipoles in mind, but there is a hack that enables "good enough" dipole simulations.

Rudolf

[twest820]

Thanks; have to look at Boxsim too sometime. Too many tools, not enough time.

As an aside, I've been trying to find the origin of the toroidal dipole baffle idea. Earl's remarks here are probably close. A hyperbolic-exponential throat is probably viable as well, but I'll start the build with just a torus.

[otto88]

For very deep pockets - $820 each:

http://www.solen.ca/pdf/raal/DIPOLE140-15D.pdf

[Telstar]

Very good! A lil too inefficient for my tastes, though.

[twest820]

Lesse here. 2 x Neo3 + 2 x Neo10 + 2 x 18SWS1100 = CAD 892. Right in between the 15D and 15DAM. Hmm, which would I rather have? Drivers for an entire three way or a tweeter?

[Rudolf]

Originally Posted by otto88

For very deep pockets - $820 each:

http://www.solen.ca/pdf/raal/DIPOLE140-15D.pdf

The faceplate of those drivers is 194 mm wide - 3 times the Neo3. You don't want to see the polars of those Raal dipole tweeters.

Rudolf

[mige0]

Originally Posted by Rudolf

The faceplate of those drivers is 194 mm wide - 3 times the Neo3. You don't want to see the polars of those Raal dipole tweeters.

Rudolf

... room for improvement ...


Even Aleksander has to learn dealing with dipole issues...


Michael

[Soundminded]

Originally Posted by Rudolf

There have been some attempts lately to make dipole tweeters work up to high frequencies with constant directivity. How high can we go, if the tweeter housing is already almost as small as its radiating area?

I would like to demonstrate the issues involved with an idealized 1 inch circular tweeter. I am using some Boxsim simulations to show things detached from specific driver peculiarities like FR and dome breakup modes.

Let’s start with the mathematical dipole model – first for a dipole distance of 2,5 cm to keep things within 20 kHz:



This is equivalent to this Linkwitz figure.

A 1’’ tweeter would have a physical diameter of at least 2,5 cm. This would lead to a dipole distance of 1,25 cm. As a 2-point source this would sim like this:



Of course a tweeter is no 2-point source. Let’s first move to a dipole point source on a 25x25 mm square baffle (Boxsim can’t do circular baffles).



We can see that the dipole peaks are slightly lower in frequency because a 25x25 mm baffle is more like a 28 mm disc than 25 mm. More important by far is the change in directivity that is introduced by the baffle. The dipole 8 gets sort of slimmer.

Since a real tweeter is not a point source on a baffle, we make the cone/dome/foil as wide as the baffle. Note that our model is working as a dipole membrane in a housing with no depth now. Any real world tweeter would have to be larger:



Note how the dipole 8 gets even slimmer than before and how the dipole peaks move to even lower frequencies. Strangely the 60 deg peak is at a lower frequency than the on-axis peak. Same for 30 deg. How come? The next diagram shows why by overlaying the last two diagrams:



The wider source starts to beam – lower in frequency at wider angles. From a certain frequency up the directivity of the tweeter is no longer controlled by the dipole function but by the tweeter diameter. The next diagram shows this relationship at the first dipole peak:



A monopole tweeter would have a main beam to the front of almost the same contour as the figure 8 of the dipole. This would be true for any driver whose radiating area is as large as its baffle – regardless of size. Any attempt to manipulate the radiation at the sides of the baffle is quite fruitless for frequencies above the first dipole peak – they don’t show up any longer at the sides.

The only cure is to move the first dipole peak up by making the tweeter smaller. Above the first dipole peak all compact drivers will move from the (constant) directivity of the dipole 8 to the frequency dependent directivity of a beaming driver.

That’s how I see it.

It seems to me that a dipole tweeter defeats the very purpose of constant directivity. Unless the rear radiation is restricted to a reflector, presumably a parabolic one at that, won't that radiation result in reflections whose directions are difficult or impossible to control and will also be a function of frequency? Am I the only one who sees an inconsistency here?