Reducing Waveguide Size

[Patrick Bateman]

I posted some thoughts on reducing waveguide size on the "Geddes on Waveguides" thread. In particular, I am curious about the use of domes. While some have used domes in the past to reduce the cost of a waveguide, I am only interested in reducing the size of the waveguide.

So cost will not be a consideration for this project.

Anyways, if you're curious about waveguides, but are daunted by the sheer size of the devices, read on...

I personally own Gedlee Summas, and enjoy the performance of them, and heartily recommend them. At the same time, in the spirit of DIY, I'm curious to see how I can reduce the size of a waveguide.

[Patrick Bateman]

Here is my original post from the other thread.

I gave up on using domes in waveguides a few years ago, but I think I'm ready to take another crack at it. I personally believe that a compression driver is a better solution, but I have a unique problem, and I can't find a way to solve it that doesn't involve a dome tweeter. Here's the situation:

I've built a ton of horns and waveguides, primarily for automotive use. When I first started, I built them full size. We're talking huuuge ol' horns that make the car basically undrivable. This is back in the 90s. As I began building waveguides, I found that you could use boundaries to extend the waveguide. And I don't mean an inch or two. I'm talking about removing over 75% of the waveguide.

Here's the problem I'm running in to: The size of the compression driver itself is really becoming an issue. I can't find compression drivers small enough to fit into corners. The best solution that I've found so far is to bend the throat of the waveguide, so that the compression driver enters at a 45 degree angle. But it's tricky, because the bend creates a dip in the frequency response, and the dip gets deeper as the bend gets bigger. (IE, if the compression driver enters the waveguide at a ninety degree angle you get a big fat dip due to a reflection off the wall of the waveguide.)

Here's the best waveguide I've built. It's an elliptical oblate spheroidal loaded by a BMS 4540ND. It has a horizontal coverage of 108 degrees, vertical of 72, and an average of 90. You can see that the compression driver enters at an angle. This allows me to push it as far back as possible on the dash of my car, which better mates with the angle of the dash and the windshield.

So here's what I'm thinking. If you a neodymium tweeter at the apex of a waveguide instead of a compression driver, you can push that waveguide almost all the way back. By pushing it back further, you can reduce the mouth of the waveguide even further. Basically, you let the windshield and the dash do more of the work.

Again, not saying that this is a good solution for the home, where depth isn't a problem. But my application is rather unique, and I simply can't think of another way to solve this.

By using a dome tweeter I will lose a great deal of efficiency, reduce power handling, and increase distortion. On the upside, it's smaller, looks better, and costs less. I have a hunch that a phase plug could be built to improve the high frequency polar response. Basically get the dome to behave more like a compression driver, and less like a dome.

As I see it, a compression driver is a big dome tweeter with an extraordinarily powerful motor, and a carefully designed phase plug. Here are two compression drivers that I've used. JBL 2470 on the left, back when I was into huge horns in the car. BMS 4540ND in my hand, once I switched to Unity horns.

Here's a TAD compression driver cut in half. They had this laying around their demo room when I was there, a rather amazing paperweight if you ask me. In the cross section you can see that this isn't much different than a dome tweeter, except they've added a very carefully designed phase plug in front of the diaphragm, and the dome is inverted.

To give you a sense of scale, here's the compression driver that I'm using in my car, mounted on a conventional horn. (woodhorn.com btw.) You can see that the TAD absolutely dwarfs it.


Here's a pic of JBL's top of the line components, which use a waveguide that looks suspiciously like an elliptical oblate spheroidal. My design would be similar to this, but using the windshield to bring the soundstage to eye level, and to extend the curve of the waveguide. (In other words, my waveguides will have a narrower coverage angle than JBL's.)

So I would end up with a waveguide that uses the oblate spheroidal curve, but with a footprint that's about 75% smaller than the waveguides I am already using. I haven't decided yet if it will be a Unity. There *is* enough room to mount a couple of midranges to extend the response. The primary goal will be to reduce the waveguides size to the smallest footprint possible, and use the existing boundaries to extend the curve.

Anyone care to comment on these plans?

[Patrick Bateman]

Also, I should mention that this project is a subproject of another one I'm working on.

My Mind's Playing Tricks on Me - More Fun with Waveguides and Psychoacoustics

While I normally wouldn't want two threads on one subject, I believe this particular subject has broad appeal and may interest people. In a nutshell, if I'm going to hack up a waveguide and measure it, I want to be sure people see the results

[dantheman]

This will be a cool experiment for us to enjoy! I can't wait to see what you measure.

Dan

[Patrick Bateman]

To give you an idea of how big waveguides are, here's a Celestion compression driver on a QSC waveguide in my car. As you can see, this clearly doesn't fit! The thing is, we can use existing boundaries to continue the curve of the waveguide. In order to do this, we need a very VERY small source. The smaller it is, the further we can push it back into the corner. And the further we can push it back, the more we can rely on the boundary to augment our waveguide.

This isn't a small detail - doing this can extend the cutoff of your waveguide by two or three hundred percent! So a HUGE win.

One option that I've considered is using a dome instead of a compression driver. A lot of you are interested in domes because they're cheaper. The main problem with domes is that they don't have the efficiency of a compression driver, and because they don't have a phase plug, the high frequency response suffers. The Vifa XT19 is an interesting option, because it *does* have a phase plug, albeit a very small one. In this pic, you can see that it's a bit shallower than a BMS compression driver.

Here is a BMS 4540ND on a QSC waveguide. If I'm not mistaken, the BMS is the smallest compression driver in the world. IIRC it's about $240 a pair from Assistance Audio. This pic demonstrates that we've reduced the depth of our design quite a bit by using BMS.

Brandon (aka Augerpro) posted some measurements of a Celestion compression driver than can give the BMS a run for it's money. It's also less than half the price. There's one problem though - it's about half an inch longer than the BMS. This is a real problem, since I'm trying to reduce the depth as much as possible.

Here's where things get interesting! If you take the Celestion apart, you basically have a high efficiency dome tweeter with a very nice phase plug. It's difficult to tell in these pictures, but that's a 1.75" dome with a sophisticated phase plug mounted in front.

Here's a side view of the Celestion, along with the waveguide I intend to use with it. The waveguides depth will be reduced so that the phase plug of the compression driver is deep inside the waveguide, reducing the footprint of the entire device by at least 50%.

There are a number of other benefits as well. Brandon's measurements showed that the Celestion has lower distortion than the BMS. This is likely due to the larger diaphragm of the Celestion. Also, this waveguide design "pushes" the soundstage back by three or four inches.

This seems like an ideal solution. We reduce the depth of the waveguide, the overall size, and we still have very high efficiency. As a bonus, it *might* reduce high-order-modes. (Based on my understanding of HOMs, they're created by diffraction. By using a waveguide angle that is unchanging, we reduce diffraction. WIN-WIN.)

[flacoman]

They make a line of tweeters with inverted domes and phase plugs, still a bit larger than what you're after but a good place to start.
HTH
Jorge

[dantheman]

Cool illusion: The Levitating Waveguide.

[Patrick Bateman]

Over the weekend I hacked up the Celestion compression driver, and did various measurements. I also measured a direct radiator on a waveguide. In this post I'll cover the following:

• How small can we make the waveguide?
• Which performed better - the direct radiator or the compression driver?
• What is the effect of termination (ie, roundover.)

Before diving into the measurements, I should cover my goals for the project. The whole reason that I want a small waveguide is to control early reflections. In the car I've found that reflections off the windshield and the side windows create audible colorations, reducing intelligibility and skewing the soundstage. But waveguides take up a lot of room. I'd like to use the existing curvature of the dash to extend the mouth of the waveguide. It doesn't take a very large waveguide to control directivity at high frequencies, but getting down to one or even two khz requires a VERY large device. By hacking up the compression driver, or using a conventional woofer, we can let the windshield and the dash do most of the work for us.

The first device that I tested uses a Peerless 830970 in a *very* shallow waveguide. I'm basically using it as a tweeter. The pics above are from my car. The device that I measured used a single Peerless woofer on a flat baffle, with a 1" roundover to reduce diffraction. The waveguide size is identical - about 3" wide and 3/4" deep.

The reason that I opted for this model instead of a conventional tweeter is that the larger diaphragm creates a narrower beam of sound than a conventional tweeter. The pic above is the published polar response, from Peerless. As a rule of thumb, loudspeakers beging to beam at wavelength that's equivalent to the piston's diameter. For the 830970, that means it will start to beam at 6750hz. But can we use a waveguide to control directivity down to two or three khz?

Here's the polar response of my reference, a set of Gedlee Summas. I measure on the same scale as Geddes, but my measurements are in 15 degree increments, his are in 7.5. We're both measuring from zero to ninety degrees. Note that I'm not trying to achieve flat response, I'm trying to achieve directivity control. So I don't care that the response is falling; that can be fixed electronically. Directivity can only be manipulated physically.

Theoretically, a waveguide that's 3" wide and 3/4" deep should control directivity down to 4500hz, with a beamwidth of ninety degrees. It appears that the Peerless waveguide performs a bit *better* than expected. The graph above shows a beamwidth of ninety degrees all the way down to 3khz! By 10khz the beam has narrowed to sixty degrees, which should reduce reflections even further. There's a dip at 16khz, which is also present in the literature from Peerless. This may be reduced via the use of a phase plug perhaps. All in all, I believe this is exceptionally good performance, particularly since the device costs less than ten dollars, including the woofer!

• Pros - very VERY inexpensive, small, easy to build, exceptionally good directivity considering the size.
• Cons - very low power handling, not as efficient as a compression driver, there's a dip at 16khz

[gedlee]

Hi John

It's a performance versus size tradeoff across a continuum. Smaller and wider works better than smaller and narrower - the reasons for this are complicated so I'll just let it lie like that. Basically the narrower and more consistant you want the polar pattern to be the bigger the device has to get. You can trade off those two characteristics at will however. Many people find a very small waveguide arround a dome works great, and it will, as long as the angle is large, but narrowing that angle will cause a continued degradation in performance, unless the size gets larger. How much can you live with? Thats a design decision.

[John Sheerin]

That nasty dip that moves down in frequency from 16k to 10k as you move off axis is similar to what I saw in simulating conical waveguides with no mouth termination. You might try smoothing out the exit of your waveguides and see if that helps. I've found building insulation styrofoam to be good to use for hacking around like this. You can shape it with wood working tools very quickly or use a hot wire.