Geddes on Waveguides

This is why we use software simulation to first understand what's going on before building. AxiDriver is sufficient for this kind of task even though it is not exact. But it allows you to sort of figure out what is effecting which part of the spectrum.
Thanks Earl. I hadn't thought of it as a gradual change...duh!

This is my starting point. You can see an on-axis dip of about 4db centered around 5k...the dip decreases to about 2db at 10 degrees off-axis and is completely gone before 20 degrees.

ND1460A30inch80degreeWG4meters01020.gif


The WG is 30" OD x 12" deep, 1.4" entry, OS with a 10 degree overall entry angle, mouth "roundover" is a prolate section (not a constant radius). Measurement is 0, 10, 20, 30, 40 degrees from 4 meters.

Does this provide any clues on the amount of asymmetry needed? I don't picture myself building "scores" of molds since it took me weeks to get something usable for the first set of WGs :dead:.
 
Yup, I've gained a better overall understanding by using both Axi-Driver and Hornresp. However, since they both sim only axi-symmetric devices, they weren't particularly helpful with the asymmetrical question (at least not to me).
It is possible to eliminate on axis dip with axisymmetric device. My past posts show this. The issue is related with how the waves expand. However, if you think that asymmetric is the only way to go, then it's a different issue. maybe ABEC can do it, but the modeling takes some effort.
 
Last edited:
The only WG simulation tools I have available are Axi-Driver and Hornresp. In other words, I can sim only round WGs.

Any opinions on whether a 6:5 H:V aspect elliptical mouth would be enough to break up the on-axis dip apparently inherent in axi-symmetric waveguides? If not, what ratio would do the job? I would like to make the aspect ratio as tight as possible, but eliminating the dip has priority over aspect ratio.

This is my starting point. You can see an on-axis dip of about 4db centered around 5k...the dip decreases to about 2db at 10 degrees off-axis and is completely gone before 20 degrees.
...
Does this provide any clues on the amount of asymmetry needed? I don't picture myself building "scores" of molds since it took me weeks to get something usable for the first set of WGs :dead:.

Hello Paul

I guess you missed the discussion on that topic some time ago, so -
as for the axial dip and round over issue of OS horns you may find my work documented here :

http://www.diyaudio.com/forums/multi-way/140190-jean-michel-lecleach-horns-9.html#post1891535

and here :

http://www.diyaudio.com/forums/multi-way/140190-jean-michel-lecleach-horns-10.html#post1924924

applying.


You will find the on-axis dip to be kinda inherent to oblade sheroide contour – though my simus agree with Earl (or vice versa 😉 ), that it can be smoothed out to some degree.
If you want to get rid of the on-axis dip completely as its possibly not bearable at all in your app – I agree with Soongsc, to better look out for alternative contours.



Yup, I've gained a better overall understanding by using both Axi-Driver and Hornresp. However, since they both sim only axi-symmetric devices, they weren't particularly helpful with the asymmetrical question (at least not to me).



As for non-axisymmetric contours I think you possibly could perform two simus on AxiDriver with the x-axis and the y-axis contour respectively and simply interpolate the two results with some weighting applied for any polars in between.

This means – if you do a text book OS in the x-axis and some other contour in the y-axis you nevertheless get the on-axis dip of OS - unless possibly the y-axis contour provides the exact inverse peak ....

Just my 2 cent.

Michael
 
Is it safe then to say that the end of the flare would be far less sensitive to roughness, and that a moderately inconsistent surface there, particularly if damped with wool felt or something absorbtive, would likely be desirable to help damp this destructive interference?

I have thought of this approach as well. I have to say that I am not sure what would happen, but I would tend to believe that what is wanted is something that is not uniform around the mouth. You want to break-up the coherence of the diffracted wave from the direct wave. With the approach you noted, the diffraction may (or may not) be made smaller, but it is still coherent because its still circular. But the idea of "damping" a diffracted wave is probably incorrect since this wave is not damped by an absorptive material, its the incoming wave that is damped, resulting in a lessor to to be diffracted. Incorrectly done therfor the diffraction could increase due to a greater boundary impedance mismatch from the added material. Lots of variables to consider.
 
It is possible to eliminate on axis dip with axisymmetric device. My past posts show this. The issue is related with how the waves expand. However, if you think that asymmetric is the only way to go, then it's a different issue. maybe ABEC can do it, but the modeling takes some effort.

For the record I do not accept all the posts on waveguides that you presented because of the modeling limitations that I described at the time. You need to build and test a device and show that this is true and not just rely on the models. They can be right but they can also be wrong.

Alternate contours may reduce the dip by introducing a less than ideal spherical wavefront at the mouth, but then the polar response will suffer. It thus becomes a tradeoff - good polar response versus good axial response. However, since it is the mouth that causes the axial response problems and not the contour, it is only logical to deal with the mouth and use the better contour for the polar response. It is less than optimal to change the contour to correct for a problem with the mouth.
 
Last edited:
For the record I do not accept all the posts on waveguides that you presented because of the modeling limitations that I described at the time. You need to build and test a device and show that this is true and not just rely on the models. They can be right but they can also be wrong.

Alternate contours may reduce the dip by introducing a less than ideal spherical wavefront at the mouth, but then the polar response will suffer. It thus becomes a tradeoff - good polar response versus good axial response. However, since it is the mouth that causes the axial response problems and not the contour, it is only logical to deal with the mouth and use the better contour for the polar response. It is less than optimal to change the contour to correct for a problem with the mouth.
I think you forget that I have posted measurements on at least four waveguides of different configurations. Directivity is well behaved with no on-axis dip. Three of these wave guides have been simulated and posted comparisons against the measurements. This is much more variety of configurations than you have posted in this thread. Perhaps you'd want to review the data again? Also, feel free to recomment on each of them as well.
 
Last edited:
Michael & Frank(not sure???)
Yes, I've followed both your work with great interest. Very good.

The only axi-symmetric devices I've found (so far) required narrowed HF response to break up the on-axis dip. Since the HF narrowing is a matter of degree (none is perfect) it seems a matter of preference. Earl won't agree but, to me, narrowed HF is not necessarily good or bad, just not what I want to try at the moment.

Two separate HxV CD profiles simply have different resonances, so I do not trust my ability to extrapolate two resonances into a resonance free elliptical.

Earl's suggestion of altering the mouth around the perimeter seems interesting...gotta think about that one before breaking out the felt!
 
I think you forget that I have posted measurements on at least four waveguides of different configurations. Directivity is well behaved with no on-axis dip. Three of these wave guides have been simulated and posted comparisons against the measurements. This is much more variety of configurations than you have posted in this thread. Perhaps you'd want to review the data again? Also, feel free to recomment on each of them as well.

I suppose that I would like to see this data again as nothing sticks in my mind as being resolved. Please post the simulation and a comparable measurement on the same device. Not just axial curves please, that's easy, but show where the polar map is the same for the two devices - on axis and off axis.

I will look at two things 1) is the directivity narrow (<= 90 degress) and controlled (no beaming or narrowing within the bandwidth) and 2) is the axial hole actually gone. If you have both of those features then I will be impressed. I expect that you can do one or the other, but not both. Also please make sure that the physical size is defined as all of these problems go away in a larger device.

And I certainly do not agree that HF beaming is neither positive or negative. The incredibly smooth subjective impression of the HF response in my devices is due in no small way to the fact that the directivity does not beam. This means that you get the feeling and ambiance of HFs without the beaming into your ears that is so offensive. The control of beaming is not only not an irrelavent factor, it is a critical one.
 
I suppose that I would like to see this data again as nothing sticks in my mind as being resolved. Please post the simulation and a comparable measurement on the same device. Not just axial curves please, that's easy, but show where the polar map is the same for the two devices - on axis and off axis.

I will look at two things 1) is the directivity narrow (<= 90 degress) and controlled (no beaming or narrowing within the bandwidth) and 2) is the axial hole actually gone. If you have both of those features then I will be impressed. I expect that you can do one or the other, but not both. Also please make sure that the physical size is defined as all of these problems go away in a larger device.

And I certainly do not agree that HF beaming is neither positive or negative. The incredibly smooth subjective impression of the HF response in my devices is due in no small way to the fact that the directivity does not beam. This means that you get the feeling and ambiance of HFs without the beaming into your ears that is so offensive. The control of beaming is not only not an irrelavent factor, it is a critical one.
Well, here is a link I coould find for now.
http://www.diyaudio.com/forums/multi-way/103872-geddes-waveguides-320.html#post1887771
These are the first four waveguides I have ever done in my life.
 
Soongsc, you don't have any larger pictures do you? I can see the trend well, but specifically how fast they are dropping off is beyond my visual capabilities as are finer details. Looks like some rescaling would be nice as well. In any case from what I can tell they look great.
 
Last edited:
Well, here is a link I coould find for now.
http://www.diyaudio.com/forums/multi-way/103872-geddes-waveguides-320.html#post1887771
These are the first four waveguides I have ever done in my life.

Where are the simulations for comparison?

That data is so compressed its hard to tell much of anything - as someone said. And why only up to 45 degrees? Basically the curves should not change much up to that point and then drop somewhat rapidly from there. Looks like yours might be down as much as 15 dB at 45 degrees. Do you have complete data or is this all that you have?

If you had raw data at 7.5 degree increments up to 90 degrees, I would plot it as a contour plot for you and then we could really tell what was going on.
 
Last edited:
Probably we can start with this one.
http://www.diyaudio.com/forums/multi-way/103872-geddes-waveguides-319.html#post1887491

Soongsc, you don't have any larger pictures do you? I can see the trend well, but specifically how fast they are dropping off is beyond my visual capabilities as are finer details. Looks like some rescaling would be nice as well. In any case from what I can tell they look great.
I'm afraid it won't work since the impulse data generated from these graphs have a reference impulse. There is also a mic calibration file involved.
It's been so long that I can't remember if I went to 90deg because I rotate about as close to the AC point as possible. 90deg would be behind the lip. My one meter measurement is measured from the diaphragm.

Where are the simulations for comparison?

That data is so compressed its hard to tell much of anything - as someone said. And why only up to 45 degrees? Basically the curves should not change much up to that point and then drop somewhat rapidly from there. Looks like yours might be down as much as 15 dB at 45 degrees. Do you have complete data or is this all that you have?

If you had raw data at 7.5 degree increments up to 90 degrees, I would plot it as a contour plot for you and then we could really tell what was going on.
 
Probably we can start with this one.

Well those curves completely contradict your statements. There is a hole on-axis, and the sim does NOT show this. It is hard to say if it is CD since the usual deffinition of beam width is at the -6 dB points and this is very hard to read on those plots. The scale of the plots makes everything look good. The region above 4 kHz is not well represented in the simulations, being almost flat in the data and falling in the sims. Its not even close above 10 kHz, but thats not very important.
 
Well those curves completely contradict your statements. There is a hole on-axis, and the sim does NOT show this. It is hard to say if it is CD since the usual deffinition of beam width is at the -6 dB points and this is very hard to read on those plots. The scale of the plots makes everything look good. The region above 4 kHz is not well represented in the simulations, being almost flat in the data and falling in the sims. Its not even close above 10 kHz, but thats not very important.
Well, the notch above 10KHz goes through 30deg, not a typical on-axis situation. I'm not saying this is good performance, but it certainly is not an on-axis hole. To be honest, if the hole were only on axis, it would probably be preferrable than this one.

Concerning CD, the difference between 0 and 45 deg is about 11db, quite comparible with the Nathan wave guide. Viewing data at this scale should really not have any visual error over 1db. Looking at the Nathan data, the lines sort of not parallel with each other, so it's ever more difficult to define beam width.

I agree that I have never seen any simulation of anything on any software that presents accurate data above 10KHz. However, once famliar with the charaqteristics of the software, it presents enough to show where the problem is. I know this because once I did better modeling of the driver section, the results were better.

For the record, none of data I presented in this thread provides satisfactory performance to me technicaly or listening. It's just a phase that I've been through. Driver loading seems far more important.
 
Well, the notch above 10KHz goes through 30deg, ...

For the record, none of data I presented in this thread provides satisfactory performance to me technicaly or listening.

I was talking about 4 kHz where a clear hole is evident in the measurements. Looks like several dB, at least as bad as any one my waveguides. The scale is a real problem, makes accurate comparisons impossible.

Attached is the Summa on the same scale.

You made claim to such success earlier and now, when pressed, the story seems to have changed.
 

Attachments

  • Summa.GIF
    Summa.GIF
    3.8 KB · Views: 282
Last edited: