Matarials can be varied, but the criteria is pretty well determined. The waveguide itself must not have any resonances in it at all. This is NOT the norm by any means. The standard method of mounting does not lend itself to good designs because the greatest mass is right where the cross section is minimum. The situation should be the exact opposite. The material cross section at the junction of the throat to the driver should be the greatest of anywhere along its length and this should be several inches. This means that injection molded plastic or fiberglass is out unless a secondary opperation is performed, just like I used to do - fiberglas skin back filled with dense polyurethane foam. Plastic molded with ribs is not enough and fiberglas just painted with mastic isn't enough either.
My waveguides are 6 inches across at the throat with a 1 inch hole in the center. They are so strong that you can take a sledge hammer to them and while the surface will dent, they will remain a solid structure. The waveguides that I see sold are certainly cheap, but they are also flimsy.
In all the 2" drivers that I have tested the response dies just above about 8-10 kHz. It falls too fast for decent EQ. The power required would put the driver at serious risk of burnout. The same is true of the 1" drivers but at almost double this frequency, i.e. 14-16 kHz. The 1" limit is acceptable, but the 2" is not.
It's the fatal flaw in the 2 way concept, drivers are resonant devices rarely good for usable response for more than 2 1/2 to 3 octaves. To make an effective 2 way system you need to go to extraordinary lengths using equalized arrays. If you add a subwoofer it becomes a 3 way system. With contemporary devices, a 4 way system or more is the only way to cover the full audible spectrum with ease. The typical 2 way ported 8" woofer/1" dome tweeter system has a resonance peak somewhere an octave above the lowest octave and dies below that and the tweeter beames its highest octave in a narrow cone. I'd bet in an anechoic chamber the typical speaker's outpt at 15 khz 180 degrees off axis (behind the speaker) is at least 50 db down or 1/100,000 the energy it propagates on axis. By contrast, walk around most musical instruments being played and tonality hardly changes in any direction at all. It would be a wonder if the sound from a speaker actually sounded like real music. Maybe in an anechoic chamber or outdoors in an open field but not in a real room.
http://www.jblpro.com/pub/components/2380a.pdf
http://www.jblpro.com/pub/components/2447.pdf
http://www.jblpro.com/pub/components/2451.pdf
The JBL 4" driver has done well above 10K ever since the 2441 replaced the 2440 back in the 80s. Not cheap though.
Regarding EQ, even if response had fallen 12-15dB by the top Octave, the mid frequency efficiency is so high and the power requirements above 10k so minor that power handling isn't an issue.
I agree that 1" units have appealing performance, but have no problem using 2" throat units in theater applications. (and they sound good)
David S.
Dave
The reason that those plots appear to have good HF performance is because the horn beams, but the power is falling quite a bit. You can see this in the 2447 at about 8 kHz. On-axis its a small thing but with a truely CD waveguide the response drops like a stone. In the 1" devices this happens at about 16 kHz.
The reason that those plots appear to have good HF performance is because the horn beams, but the power is falling quite a bit. You can see this in the 2447 at about 8 kHz. On-axis its a small thing but with a truely CD waveguide the response drops like a stone. In the 1" devices this happens at about 16 kHz.
Try Urethane Foam , Expanding Marine Polyurethane Foam the 16 lb density at the bottom of the page. You will need a mold to use this however, it can't just be painted on. Even with a mold you have to be careful because of the expansion. Not done correctly the mold can blow appart and you have a big mess on your hands.
One problem with the waveguide as I see it is while its FR above a given frequency is independent of angle and can be equalized to be flat within its power handling capacity, its absolute output level is not, falling off constantly as you go off axis. This is like having a midrange tweeter control that drops relative to the woofer output as a function of angle. I'm not sure if this is better or worse than just tweeter output falling as frequency increases as you move off axis. Overall flat system response of the direct field is only obtained at one offset angle at a time. As soon as you move, overall direct field response of the mid-tweeter changes with respect to the woofer. You must be located at the correct angle and of course there is no independent control over the spectral response of reflections. That can play a significant role in what you hear depending on how reflective room surfaces are. With normal 1/2" sheetrock painted with latex paint in rooms up to several thousand cubic feet this can be considerable unless the walls are largely covered with drapes, tapestries, paintings, or other absorbing wall hangings. Wood or linoleum flooring not covered with heavy rugs or carpeting only makes it worse. These limitations are conceptually the same as for conventional speakers, only the details of how they manifest themselves are different.
Last edited:
Yes, this is a good thing. It allows for crossing the speakers axis and creating a much wider sweat spot than is possible with any other design.
This is not true. The direct field is dominately the response along the listening angle and the reverberant field is dominately the power response. Unless the system is CD these two response cannot be the same. Piston based speaker designs cannot be CD unless it is a full range omni system, but then we know that choice of directivity is not very good.
Last edited:
Actually, I was looking at the terminated tube curves. They hew pretty closely to the 1st order mass breakpoint rolloff. They could be smoother but they don't die at 8kHz.
While we are on the subject of a "truely CD waveguide", should we really apply that term to a unit that has smooth but constantly rising directivity? What variation do you think should be allowed in DI and still deserve the term "constant"? And if units have axial response issues, should that be factored into the term "constant directivity"?
David S.
The TADs that I tested had a failry flat response on axis according to the manufacturers data, but when put on a waveguide that had 90 degree coverage at 10 kHz the response fell dramtically. I have never tested JBL drivers, but the data that I looked at showed similar effects. Plane wave tube data above the cross mode frequency is suspect.
You were not involved some years back when the definition of CD was bantered about. In exact terms nothing has ever been CD, even when they wre called that. As one "relaxes" the deffinition some devices start to be applicable and my waveguides are certainly among those that require the least relaxation of the deffinition. There are examples of so-called CD devices out there that would require a complete relaxation of the deffinition to be valid.
So the answer is that it is a continuum. Some devices fit the ideal better than others and some are completly bogus - they just like the sound of the description. Where you personally choose to put the cutoff point is up to you.
I have tried to do everything possible to achieve CD and what I have is as close as I could get. Personally, I don't know of any other approaches that get closer to the ideal (at least not where the directivity is specified and less than Pi).
Hence, I would say that either the Summa waveguide is CD or the concept is simply not obtainable. Below the Summa the deviations get greater and greater as, of course they will with smaller and smaller waveguide.
In my book there is a section on CD where the problem is worked backwards, starting with the far field and working back to what the source velocity distribution needs to be to achieve that. It is not hard to predict (a spherical section), but it is also obvious that true CD is not obtainable by any finite source.
There was also a PhD thesis done in Australia where the researcher took my same approach but applied linear optimization techniques to find numerically what contour would be required to acheive the closest approximation to CD. Not surprisingly he found that an almost exactly OS shape was optimal and that the handling of the mouth termination was critical. These requirements were both published in my book some five years prior to his work.
You were not involved some years back when the definition of CD was bantered about. In exact terms nothing has ever been CD, even when they wre called that. As one "relaxes" the deffinition some devices start to be applicable and my waveguides are certainly among those that require the least relaxation of the deffinition. There are examples of so-called CD devices out there that would require a complete relaxation of the deffinition to be valid.
So the answer is that it is a continuum. Some devices fit the ideal better than others and some are completly bogus - they just like the sound of the description. Where you personally choose to put the cutoff point is up to you.
I have tried to do everything possible to achieve CD and what I have is as close as I could get. Personally, I don't know of any other approaches that get closer to the ideal (at least not where the directivity is specified and less than Pi).
Hence, I would say that either the Summa waveguide is CD or the concept is simply not obtainable. Below the Summa the deviations get greater and greater as, of course they will with smaller and smaller waveguide.
In my book there is a section on CD where the problem is worked backwards, starting with the far field and working back to what the source velocity distribution needs to be to achieve that. It is not hard to predict (a spherical section), but it is also obvious that true CD is not obtainable by any finite source.
There was also a PhD thesis done in Australia where the researcher took my same approach but applied linear optimization techniques to find numerically what contour would be required to acheive the closest approximation to CD. Not surprisingly he found that an almost exactly OS shape was optimal and that the handling of the mouth termination was critical. These requirements were both published in my book some five years prior to his work.
Last edited:
I agree, My experience with the 2" TAD TD-4001 tells me they can be easily EQed on the top end for on-axis performance. The added performance down to 500Hz makes it all worth while. Im not concerned about any 90degree response since that soundwave is 100% absorbed in room anyways.
I would think that to call something CD you would want the d.i. to fall within a reasonable window, say +-1.5 or +-2dB at the most over the range of interest. Perhaps we need a category of "smooth directivty". There is a valid rational for such units and it would be more honest than terming them "constant" when they are not.
I know you aren't a fan of anything with a diffraction slot but I am very familiar with most of the JBL products and they appear to me to have much flatter directivity than your particular designs.
The 2360, 2365 and 2366 theater horns all have essentially flat directivity in a +-1.5dB window from 500 to 12kHz (1000 to 12kHz for the 2366). The flat front biradials (such as the 2380 and 2385), that tend to be optimized more for horizontal than vertical performance will still hold a d.i. of +-2db and have no significant climb in horizontal directivity. All these units are free from any on-axis difficulties that your units seem to have.
I know your priorities are different than JBL's but you can't dismiss the performance of their products.
Regards,
David S.
Cornucopia
While reading the recent posts in this thread, some general comments regarding horn design come to mind:
1) Use of a horn, implies that the designer is willing to trade driver bandwidth for efficiency. To achieve this mission a horn constrains the driver’s radiation pattern only above the lower limit of its pass-band.
2) Here at this lower limit, driver unloading occurs while diaphragm excursion limits are rapidly approached as well. To avoid driver destruction and the introduction of excessive distortion in driver output, aggressive high pass filtering of the drive signal is required.
3) Under these conditions, a minimum of a 3-way partition of the audible spectrum is needed; e.g., 20-200, 200-2000, 2000-20,000 Hz. This fact is particularly relevant for an all-horn system.
4) To mitigate beaming at the upper limit (determined almost entirely by the perimeter size and geometry of the effective piston radiating area), dispersive elements must be introduced such as an acoustic lens, passage bifurcation, or other beam spreading mechanisms. All drivers, horn loaded or not, exhibit high frequency beaming. This occurs when signal wave length approaches or becomes smaller than radiator dimensions.
5) Conversely, low frequency spreading is restrained only so long as horn overall dimensions (mouth perimeter and body length) exceed signal wave length. Here the effective piston radiating area occurs at the horn mouth instead of at its throat.
This list is not meant to be inclusive of all the numerous design considerations and tradeoffs to be made while developing a loudspeaker system, nor is it directed particularly to authors of this thread. It simply serves as a guide for those readers contemplating the design and use of horns in their DIY loudspeaker projects.
Regards,
WHG
While reading the recent posts in this thread, some general comments regarding horn design come to mind:
1) Use of a horn, implies that the designer is willing to trade driver bandwidth for efficiency. To achieve this mission a horn constrains the driver’s radiation pattern only above the lower limit of its pass-band.
2) Here at this lower limit, driver unloading occurs while diaphragm excursion limits are rapidly approached as well. To avoid driver destruction and the introduction of excessive distortion in driver output, aggressive high pass filtering of the drive signal is required.
3) Under these conditions, a minimum of a 3-way partition of the audible spectrum is needed; e.g., 20-200, 200-2000, 2000-20,000 Hz. This fact is particularly relevant for an all-horn system.
4) To mitigate beaming at the upper limit (determined almost entirely by the perimeter size and geometry of the effective piston radiating area), dispersive elements must be introduced such as an acoustic lens, passage bifurcation, or other beam spreading mechanisms. All drivers, horn loaded or not, exhibit high frequency beaming. This occurs when signal wave length approaches or becomes smaller than radiator dimensions.
5) Conversely, low frequency spreading is restrained only so long as horn overall dimensions (mouth perimeter and body length) exceed signal wave length. Here the effective piston radiating area occurs at the horn mouth instead of at its throat.
This list is not meant to be inclusive of all the numerous design considerations and tradeoffs to be made while developing a loudspeaker system, nor is it directed particularly to authors of this thread. It simply serves as a guide for those readers contemplating the design and use of horns in their DIY loudspeaker projects.
Regards,
WHG
Last edited:
Hello,
The following is more probably what Earl Geddes refers to when it come to shape optimization with directivity as a goal :
http://data.mecheng.adelaide.edu.au...s/2008/preprint_morgans_optimization_2008.pdf
The resulting "waveguide" doesn't look like an OS waveguide IMHO but rather more like a minphase horn...
But when we use another goal , by example the minimisation of the reflection coefficient at different frequencies, the shape is much more different.
See :
http://www.it.uu.se/research/publications/reports/2002-019/2002-019-nc.pdf
The resuling horn looks much more like a Le Cléac'h horn rather than an OS waveguide.
Best regards from Paris, France
Jean-Michel Le Cléac'h
No Defense Required
JMLC,
When it is realized that the curvature of clothoid horn boundary increases with its length (as it should) and that the assumed wave front geometry it supports closely approximates that observed in all horns, there should be no doubt regarding the superiority of a design regimen that implements it.
Regards,
WHG
JMLC,
When it is realized that the curvature of clothoid horn boundary increases with its length (as it should) and that the assumed wave front geometry it supports closely approximates that observed in all horns, there should be no doubt regarding the superiority of a design regimen that implements it.
Regards,
WHG
This isn't always the case. Driver unloading is a myth since the acoustic load is always small enough that changes in it do not alter excursion to any significant degree. And if the device is CD then it needs attenuation at these lower frequencies to match the HF output. This means in essence that the excursion even at the LF limit is not significant, not much more than that at HFs. The problem is directivity control and driver resonance. The response falls below the drivers resonance and this can be a problem if it is too high as the EQ required does increase the excursion. But a typical DE250 has a resonance at about 1.5 kHz and going down to 1 kHz or even 700 Hz is no problem at all. We are talking home situation here, not pro. Thats a different thing altogether, but not the topic here.
To me its more like 100 - 1kHz, and 1 kHz to 10 kHz. > 10 kHz I don't wory about and < 100 Hz is handled by multiple subs.
It is untrue that all devices beam. My waveguides don't, not even at 16 kHz. If they are not carefully design and assembled then, yes, they do beam.
The Summa as shown on my website has a DI within these limits from 900 Hz - 20 kHz. when referenced to 16 degress, which is the design axis. Below 900Hz the DI falls slowely, slow enough that it bis not an issue.
I don't trust older data done on horns and such as I never get the results that the companies posted. But yes, diffraction does a decent job of directivity control at the sake of sound quality. I don't give up either.