Geddes on Waveguides

diyAudio Editor
Joined 2001
Paid Member
Earl Geddes will be leaving the country in a few days, but he wishes to post some information on waveguides to contribute to our common knowledge. Let's let him lay out his points as he wishes and not introduce subjects that will make this thread veer off topic..

Here you are Earl:
 
I'm most interested in how to create the expansion profile of an Oblate Spheroid WG. I've been quite happy with my 90° conicals lined with foam to attenuate HOM's, but closer to optimum is worth a shot. I'd like to reach as low as possible in similar size constraints as your Summa.
 
johninCR said:
I'm most interested in how to create the expansion profile of an Oblate Spheroid WG. I've been quite happy with my 90° conicals lined with foam to attenuate HOM's, but closer to optimum is worth a shot. I'd like to reach as low as possible in similar size constraints as your Summa.

/second Particularily what dimensions govern *usable* frequency response.


Lynn Olson said:
To the moderator: if it's OK with Earl, why not reprint (in this thread) the technical postings he's already made on the BTA thread?

/second this one too
 
You can post anything that you want from this site, thats fine.

It turns out that OS waveguides are quite constrained by the math. Once you have a driver and then the coverage pattern, the rest is fixed. OS waveguides do not have a predicted "cutoff" so the low end tends to be dictated by the compression driver. The coverage angle can only be held down to where the mouth dimension is too small to control it. In the ESP15 (a Summa) this coverage narrows a bit at about 1000 - 2000 Hz. The waveguide is about 16 inches across. In the smaller ESP12 and ESP10 the waveguide is smaller with a notable raising of the lowest frequency of coverage. It is really important in these devices to have a sizable waveguide - too small is a lot of compromises.

So the design procedure is simple. Pick your Comp driver, and then your coverage pattern then the contour becomes:

y(x) = sqrt(throat radius^2 + x^2 Tan(coverage_angle)^2)

x is the distance along the axis. Note that at x = 0 the angle is zero and the radius is "throat_radius".

For extreme accuracy, which appears to be important, one wants the initial angle and radius of the wavegiude to match the exit angle and radius of the driver. This is tricky, but one who is competent at design can work out the correct numbers from the above equation. I did it in MathCAD. I can generate the contour for any radius and exit angle, (I would have to charge a fee for that). Its not an intractable problem however and trial and error on a piece of paper or in a spreadsheet can get you what you need.

Note that larger throats on the compression driver invariably lead to a lower frequency of falloff when the waveguide is true CD. Many drivers show good HF response out to 20 kHz on a plane wave or on a non-CD horn, but when put on a true CD device like an OS waveguide the response dies above 10-12 kHz for a 1.5 " driver and 9-10 kHz for a 2" driver. This is why I have only used 1" drivers. I have not found one that goes out far enough in any larger throat sizes.
 
Another point that I think needs to be understood is the impotance of CD, Constant Directivity. We must consider that without CD we cannot have a flat power response and a flat axial response.

Most reserchers agree that the power response is very important for tone coloration while the direct response tends to be the major factor in imaging. The industry is all too focused on getting a flat "axial" response, but to me this is probably the least important measurement.

In a polar diagram, the axial response respresents a very small portion of the radiated sound field, its a small disk at the center, but the off axis points represent every greater areas - anuluses (sp?) of increasing area. The axial point is therefor the least significant point for the power response - it has almost no effect on the power response. Further, there are very good reasons for one to not be directly on-axis of the loudspeaker (another topic), and the classic "sweat-spot" approach to sound design is kind of hedonistic. In a home theater there can be six or eight people listening - a sweat spot is simply not viable in that venue, and lets face thats the venue of the future.

For these reasons the power response and the polar responses must be smooth and flat even if the axial response is not. In my deisgns I pretty mush ignore the axial response seeking to get the best 22.5 degree response with smooth and flat polar/power response. Typically the axial response is not ideal in this scenario.

Now in a small room the situation is even more constrained. Thats because the sound system in a small room needs to avoid the very close by room boundaries to as great an extent as possible. The very early reflections and the lack of a gap between the direct sound and the reveberant field will create confusion in the image and a poor timbre of the sound. This means that in addition to needing CD, we need CD with a very narrow coverage angle - not a trivial task.

In my years of research I have only ever found one way to get CD and narrow directivity at the same time and that is with a horn. But classic horns sounded, if not terrible, certainly collored, distorted, not good! I spent nearly 20 years on this problem since I could see that horns were the solution to the CD and narrow coverage problem, but only if we could solve the sound quality issues.

The solution, that I have found, is a waveguide, with a foam plug. This device has the sound quality of the very best direct radiators, but a much better pattern control - true CD. Then lest we not forget about signal power and power compression. A compression driver will have a fraction of the power compression of a small tweeter and loads more headroom. Whats not to like!!

Our imaging and timbre perceptions are nearly dominated by the sound above 1 kHz. The musical content, the rythm, etc. are carried by those frequencies below 1 kHz, but our "perception" is inordinately weighted by the response above 1 kHz. There is a very good reason for this psychoacoustically, and it has to do with the way the nureons fire in the ear (an interesting topic in and of itself, but the important point to note is that we process sound differently above and below about 1 kHz).

So getting the 1 kHz. and up right is paramount to a good perception of coloration and imaging. I feel that far too little attention is paid to this critical region in the market place because it is here that we see all kinds of problems and yet it is here that we should be the most concerned.

I view sound system design in three major frequency ranges - low frequencies, where modal effects and the room dominates, there is no imaging or psychoacoustics to worry about, its simply a matter of adequite output and smooth spatial and frequency response (more on this in another thread); 200 Hz - 1000 Hz, probably the most forgiving of the three regions, our auditory system is only just begining to be capable of resolving spatial aspects (localization) and it is not yet very good at resolution of time delays, reflections and frequency response. If you are going to compromise something do it here as it will have the least noticable effect. Above 1 kHz is where we live as far as music is concerned. This region is ultra sensitive to time delays, reflections, frequency response, diffraction, all the things that tend to mess up coloration and imaging. Mess up this frequency region and you won't be able to recover the sound quality. Here is not where you want to make compromises for sound quality.

Thats all for now.
 
Paul W said:
Earl,
Very glad to see you participating in this forum.

Would you expound on the root cause(s) of HOMs in waveguides?

Are HOMs generated only in the WG/horn, or do the mechanisms extend to the compression drivers?
Paul


Thanks

HOMs will be generated to some extent in any device with non-straight walls. In essence they are the diffraction of the throat wave around the curves in the walls. But these are not the only sources. The wavefront that feeds the waveguide also contains HOMs to the extent that this wavefront does not match the throats aperature. Thus even though a straight walled horn will not generate HOMs it is not possible to feed a conical horn with a wavefront free of HOMs. there appears to be no way of getting a curved wavefront from an axial source without incurring the generation of some HOMs. But there are better and worse ways of doing this. The OS waveguide can be proven to be that shape which generates the least HOMs in curving the wavefront from flat to spherical. All other shapes will genetate more HOMs.

Diffraction horns generate huge amounts of HOMs and internal reflections. They may be good at directivity control, but they are terrible at the sound quality aspects of the problem. There are many compromises seen in the marketplace, but none of these will be the minimum generator of HOMs. That honor goes to the OS waveguide. The OS waveguide is whats called a "catenoid of revolution".

I have long been aware of the fact that the drivers themselves generate HOMs. Unfortunately, the prevailing model for phase plugs insures that HOMs are generated. That's because the so-called "Bob Smith" design was never intended to generate a flat wavefront, it was intended to minimize standing waves in the cavity in front of the diaphragm. In so doing its job it actually generates a wavefront that is not of constant amplitude across it, even if the phase is correct. In a patent application of mine I show how to design a phase plug that does generate a wavefront of constant amplitude and phase - minimum HOM. But alas, its going to be awhile before the driver manufacturers catch on to the importance of this factor. We at Ai may actually be the first to make drivers with this new phase plug concept. It's in the plans, but plans can change. It all depends on how much budget there is for non-product R&D.
 
AJinFLA said:
With a 16" WG & 15" woofer AC's over 1 wavelength apart @ 900hz, won't you have poor correlation, with vertical response anomalies and lobing off axis? What about time domain performance?

cheers,

AJ


There are two unavoidable polar response lobe holes in the vertical direction - this is the worst area for power response because of this, but it's not as bad as many, if not most speakers. A lot of time was spent with the crossover to optimize this response aberation. One of the lobe "holes" is aimed at the floor bounce.

There is no problem at all in the horizontal plane. From the impulse response, the time alignment of the two drivers is quite close - the difference is in the usecs. The combination of physical offset and time delay from the LP filter makes the matchup of the delays almost exact. The woofer could go forward an inch or so to be ideal, or the waveguide back, but thats very difficult to do at this point and for a few usecs its probably not worth it.
 
OS Worksheet

I can provide an excel worksheet (size~35k) which allows the display of any variety of profiles based on the oblate spheroid formula. JoshK did most of the work to get it to the current state. That includes a roundover transition to the baffle. I'd be glad to e-mail it by request and/or send it to someone who could provide a "host" site for it.

A question for Earl: In this worksheet the roundover is a circular arc that is tangent with the WG and the baffle. Is there a better shape (maybe a segment of a spiral) for the transition from the waveguide to the baffle?
 
gedlee said:

For extreme accuracy, which appears to be important, one wants the initial angle and radius of the wavegiude to match the exit angle and radius of the driver.

Dr Geddes,

I'm still quite new with regard to waveguide and compression driver concepts, but what is the exit angle of the compression driver? Is it a physical property determined by the phase plug geometry or is it wavefront curvature at the compression driver exit?

And just out of curiosity, will the new speaker series use a new compression driver with a plane wave or is it not worth the additional effort?
 
Hennie said:


Dr Geddes,

I'm still quite new with regard to waveguide and compression driver concepts, but what is the exit angle of the compression driver? Is it a physical property determined by the phase plug geometry or is it wavefront curvature at the compression driver exit?

And just out of curiosity, will the new speaker series use a new compression driver with a plane wave or is it not worth the additional effort?


Compression drivers have exit "tubes" which flare outward. The most typical is 6.5 degree, which is what almost all B&C drivers are. For a particular driover you may need to get this from the manufacturer as there is no standard.

For our initial release we will use the B&C drivers that I have ghad so much success with. I believe that the improvement in a compression driver from using a modified phase plug may well be worth the effsrt, but initially I simply don't have the time to put in that effort. We are not big enough to get a major manufacturers attention to make these design changes for us. It will happen, just when I don't know.
 
Re: OS Worksheet

Ed LaFontaine said:
A question for Earl: In this worksheet the roundover is a circular arc that is tangent with the WG and the baffle. Is there a better shape (maybe a segment of a spiral) for the transition from the waveguide to the baffle?

If you send me a copy I'll check on the math.

I just use a "fillet" tangent to the waveguide and the baffle as you suggest.
 
Earl, Have you looked at BMS

I wonder how the use of a ring radiator based compression driver affects the formation of HOMs? Here is a link to a gentleman who tested dome tweeters loaded by conical waveguides and found advantages to using an XT-25 ring radiator.
http://www.aeronet.com.au/waveguide.htm
The BMS 4552 has a following for sure.
http://www.bmspro.info/index.php?show=item&usbid=10278&id=54364

The real question is if a ringradiator produces a wavefront that is less disturbed by an OS waveguide.
 
Ed

The math that you sent is great for the baffle radius, but neglects the finese at the throat. Your waveguide has a slope of zero at the throat, but the exit of the compression driver has a slope of 6.5 degrees. This mismatch in slopes will cause more HOM. You want to match the input of the waveguide radius and slope to the same on the exit of the driver, otherwise there is a mismatch. This math is tricky, but it can be done with a cut and try approach.

Preliminary data of ours suggests that the throat is very sensitive to small perturbations of the wavefront and that the transition from the driver to the waveguide must be as smooth as possible. We actually fill in the small cracks here with clay so that the transition is seamless. No real hard data on this but the "soft" data suggests that smoother is measurably better.
 
Re: Earl, Have you looked at BMS

Jazr said:
I wonder how the use of a ring radiator based compression driver affects the formation of HOMs? Here is a link to a gentleman who tested dome tweeters loaded by conical waveguides and found advantages to using an XT-25 ring radiator.
http://www.aeronet.com.au/waveguide.htm

Well first off I disgree with his terminology. I first used the term "waveguide" to differentiate it from a "horn" - the two things use different equations and approaches and should go by different names. Then he calls an exponetial horn a waveguide, which, by my deffinition it is not. This made me sceptical.

His measurements look way better than measurements that I get, and well, I trust mine, how he got his is not clear. If his measurements are correct then it appears that using just about any form of "waveguide" on almost anything works pretty good. This has not been my experience.

I would not have expected a ring radiator to have worked very well, but I suspect that all of his waveguides are very short wide devices, yet he still seems to show high directivity. Very curiuos.
 
Posted by gedlee:
Your waveguide has a slope of zero at the throat, but the exit of the compression driver has a slope of 6.5 degrees. This mismatch in slopes will cause more HOM. You want to match the input of the waveguide radius and slope to the same on the exit of the driver, otherwise there is a mismatch. This math is tricky, but it can be done with a cut and try approach.

Cut and try it will be. I think you reference the B&C driver. So, for another driver I assume the angle could be different from 6.5 degrees.

Thanks for all.