Geddes on Waveguides

[gedlee]

Another update - I have added the Orion and a two way Behringer to the list of systems. Check out the Behringer! I am impressed - and that's hard to do!

[Patrick Bateman]

Originally Posted by Douglas WInker

Here is the text from my post. Let's see if it works.

I see from this thread that people were asking questions about the setup in my BMW. Here are some answers to your questions.

First, my dissertation was on a constant beamwidth, wide bandwidth array and had nothing to do with my car. JBL was nice enough to send me a bunch of speakers to use in the array though. I think it is cool that some of your have actually found and downloaded it btw.

No, I don't work for Harman. I work for this company ETS-Lindgren | EMC and RF Microwave Testing, Measurement and Shielding We build test chambers and I do all of the chamber design for the acoustics side of the business - anechoic chambers, reverberation rooms, noise control rooms, studios, audiometric rooms, etc.... So I live primarily in the industrial test and measurement world. If you are ever in the Austin area, send me a PM and I can give you a tour of our factory and our lab. We have a very large hemi-anechoic chamber with a >0dBA noise floor. Nice and quiet in there. Anyway, on to the 2118 setup I ran.

So the 7 series is no more. It was in a fire at Audio FX in Georgetown, Texas. Some of the equipment was recovered and I was able to save the diffraction rings for the mids. I gave them to someone in Florida with a Civic as long as he promised to be nice to them. That was a few years ago.

No this was not part of a thesis and was never published. I lost all of my test data when my old school pc died and don't have a need/desire to test again. I have the math and my notes somewhere but no promises on finding that and posting.

Yes, it what Gary used in his Regal, but due to the crossover points and size of his mids, it was more for points that sound. The technique I used only works when trying to mate a large speaker with a tweeter. If I had the room, I would have run the 2118 with a 2105 and then the Scan tweet, but I did not have the room. So unless you are going to run an 8 in a 2 way setup, don't waste your time trying this.

Now for the educational portion of the program....(no math)

If you look at sound radiation from a baffled piston (loudspeaker), you will see a relationship between loudspeaker diameter and wavelength as they relate to beamwidth. As the wavelength approaches the diameter of the piston, the main lobe narrows and side lobes appear. Now that is a simple algebraic relationship and nothing in acoustics starts out as a simple algebraic equation. It starts with differential equations with certain boundary conditions. One of the boundary conditions is that the edge of the loudspeaker in not clamped, which is what you have with a loudspeaker that moves freely at the edge. The other boundary condition is that the distance from the center of the loudspeaker to the edge of the loudspeaker is constant. It is this condition that determines loudspeaker beam patterns. If you are at a point directly in front of a speaker (say 1 m) and measure to the center of the dust cap, you get a distance of 1m. Now if you measure from the same point to the edge of the cone, you get a longer measurement. The difference between the two is the key because that difference if the same for any point at the edge of the cone. When the speaker moves (assuming pistonic motion), a wave front appears about the surface of the cone and travels toward you. The wave motion from the center of the cone gets there first and the wave from the edge (called the edge wave) gets there a little later. This difference or delay is also a phase shift. When you combine both waves at your position, they can combine constructively and destructively depending on the amount of delay. If the wave length is very large, there is little or no change in the directivity. If the difference is on the order of a wavelength, then you get interference and the beam pattern changes.

Now, if the edge of the loudspeaker is random, then the interference is reduced and the beamwidth stays omni-directional at higher frequencies. Also, the side lobes, when they appear, are 27 dB below the main lobe in amplitude. Since it is hard to build a random edge woofer and make it work well, you build the next best thing – a lens or diffraction ring as I call it. Of course, if you do the math you will see that an 8” speaker is good to about 1 kHz before you need this. My 8s play to almost 3 k so I need a ring. When you get to a 6”, you are at ~3k before this would do you any good so it is not necessary. They work for me because I go from an 8 to a tweeter. If you put them on anything smaller, they are not worth the effort.

I have to give credit where credit is due on these. Gary Biggs helped me make mine. Actually he made them because my aluminum skills at the time were not so hot. I did polish them myself though.

So....who wants to see a picture? I would like to post one.

Feel free to ask questions, but I don't spend much time on forums. So send me a PM if I don't reply.

-Doug

In a previous post Geddes mentioned the possibility of using a waveguide mouth with an irregular shape to break up the dip in the response. Here's a similar idea which was used by Douglas Winker, who used to work for Harman iirc. The original post is here:

DIY - reducing speaker beaming - Page 2 - DIYMA.com

Winker used it to extend the upper frequency response of an 8" woofer, but the idea is similar. It seems like this might be useful with the woofer if it's intended to be listened to off axis. (Like these speakers are.)

[Pallas]

Originally Posted by gedlee

Another update - I have added the Orion and a two way Behringer to the list of systems. Check out the Behringer! I am impressed - and that's hard to do!

For those of us who can't see your undoubtedly interesting program - I think you might be underestimating the number of people into high-performance speakers who prefer OSX - any highlights? What model was the Behringer?

[bwaslo]

These techniques only exist in C++ and FORTRAN

FFT can be done in about anything, though not always very efficiently. I've seen FFT subroutines in Visual Basic even. There are some very fast FFT algorithms for Pascal/Delphi (which I use). It can't do the multiprocessor trick, though, at least not with the code I've used.
If using .NET compilers, all the languages reduce to one underlying language, the .NET common runtime -- VBasic,Net, C#.net, Delphi.net all are just dialects in that situation. I don't know whether there's a Fortran.NET.

[gedlee]

Originally Posted by Pallas

For those of us who can't see your undoubtedly interesting program - I think you might be underestimating the number of people into high-performance speakers who prefer OSX - any highlights? What model was the Behringer?

I don't remember the model number. How many small powered Behringer two ways are there. I don't understand "prefer OSX". The argument over OS's is not going to go away over this issue. No matter what you do someone is unhappy.

Originally Posted by bwaslo

FFT can be done in about anything, though not always very efficiently. I've seen FFT subroutines in Visual Basic even. There are some very fast FFT algorithms for Pascal/Delphi (which I use). It can't do the multiprocessor trick, though, at least not with the code I've used.
If using .NET compilers, all the languages reduce to one underlying language, the .NET common runtime -- VBasic,Net, C#.net, Delphi.net all are just dialects in that situation. I don't know whether there's a Fortran.NET.

I searched for VB6 FFT implimentations and couldn't find any. I saw that there was a Java one. What about Singular Value Decomposition? FFTs are trivial by comparison.

The FORTRAN is not .net, but does run in the same IDE and uses the same debugger so the difference is transparent (albeit no Managed Code). It does not compile to the common language of .Net, but compiles down to assembler, which is, of course what makes it non-transportable. But its also what allows for lightning fast implimentations especially when running on Intel chips. The compiler knows all about how to optimize for Intel's strengths. From what I understand virtually all of the large FEA packages use this compiler for the number crunching. FORTRAN is obviously not GUI capable (although it can do COM.)

[gedlee]

Originally Posted by Patrick Bateman

In a previous post Geddes mentioned the possibility of using a waveguide mouth with an irregular shape to break up the dip in the response. Here's a similar idea which was used by Douglas Winker, who used to work for Harman iirc. The original post is here:

Yes this is very similar to my comments. But the quoted statement above is incorrect. The random edge will remove any "coherent effects" such as an axial hole, or lobes, but in general the beam width stays the same. This can be proven from in a number of different ways, but that will be left to the reader.

[jcx]

the "Truth" series?, size?, ported, kevlar, ribbon? I assume biggest woofer "small monitor" is better for directivity integration?

looks like "Eurolive" series also has a elliptical waveguide w/12" woofer...

BEHRINGER: Products

[gedlee]

It was one of the Truth 203X - other than that I don't know.

Nice attention to low diffraction for a commercial product.

I should mention that in a recent blind listening test it beat out the Orion. The data supports that result.

[catapult]

Originally Posted by bwaslo

FFT can be done in about anything, though not always very efficiently.

Yup. In my day (home office) job, I use an FFT routine to look at stock market data in Tradestation's 'Easy Language', which has its roots in Pascal.

Automated Trading System & Platform - Tradestation

[John Sheerin]

Nice program Earl. But isn't the DI wrong for the piston? DI is shown decreasing to a negative value as frequency increases.