Beyond the Ariel

"I built something as a direct result of this thread. The Tone Tubby 12 Alnico/PM6A that I showed at RMAF and VSAC are from this information. Plans are in public domain. Not the monster giant-killer that Lynn has planned, but a lot of people liked it and have built it."

Lynn has been pretty hard on Lowthers, but this is additional proof that they can and do sound outstanding on open baffles!! I think that some spend too much time agonizing over the waterfall plots. The Lowther waterfall plots don't look so good, but they sound great to my ears and the ears of others.

It is great that you left off wings as the wings tend to create resonances in the midrange. Is it the width of the baffle which causes you to crossover at 200 hz?

Were the drivers enABLEd? If not, the treatment would have made the sound even better!!!

I am still listening to my modified Basszilla speakers. I have lots of modifications to make, but I am too involved in a house remodeling project to get to them.

Retsel
 
Lowthers, AERs, and Feastrex are interesting special cases. All three of these whizzer-cone drivers have pretty rough upper-midrange measurements, but sound better than they have any right to.

I still don't know why, except that the coaxial mechanical crossover sidesteps the complex vertical polar-patterns of conventional multiway electrical crossovers. But then again, the Lowther/AER/Feastrex fullrange trio don't sound like Tannoys, either - they sound like themselves - a distinctive, very coherent sound.

The gap operating near saturation may be responsible for some of that sound as well.
 
Lynn Olson said:
Lowthers, AERs, and Feastrex are interesting special cases. All three of these whizzer-cone drivers have pretty rough upper-midrange measurements, but sound better than they have any right to.

I still don't know why, ...


Even with the incredible possibilities of advanced response shaping by DSP or PC XO it takes considerable effort in merging speakers to create absolutely homogenious presentation.

Extreme low overlapping did not work for me - wide overlapping sounds better to my ears in this regard but obviously has its own set of issues to deal with.

Bottom line - we seem to be extremly sensitive to any sonic disjointment.

Michael
 
Lynn Olson said:
Lowthers, AERs, and Feastrex are interesting special cases. All three of these whizzer-cone drivers have pretty rough upper-midrange measurements, but sound better than they have any right to.

I still don't know why, except that the coaxial mechanical crossover sidesteps the complex vertical polar-patterns of conventional multiway electrical crossovers. But then again, the Lowther/AER/Feastrex fullrange trio don't sound like Tannoys, either - they sound like themselves - a distinctive, very coherent sound.

The gap operating near saturation may be responsible for some of that sound as well.

Great efforts to operate the gap near saturation have been made in the better JBL drivers, and is a major reason for undercut poles. It's also quite possibly the reason they're known for sounding very dynamic.
 
I apologize for chiming in one month too late on the reflection/diffraction question (see the early June posts for the discussion), but I thought I'd recommend some good reading on the subject. Richard Feynman's great little book "QED" is about quantum electrodynamics but it has a truly insightful description of the difference between a mirror and a diffraction grating. (The book is written for laymen so don't be put off by the topic.) The former is all about reflection, the latter is all about diffraction. The only difference in the construction is that a mirror has a large, flat, uniformly reflective surface while a diffraction grating is a mirror with evenly spaced (very) narrow nonreflective stripes. The difference between the two boils down to the nature of the interference between the waves scattered by neighboring points on the surface.

One interesting feature of Feynman's description is that every point on the surface scatters waves, independent of whether there is a discontinuity or change of slope or change in impedance or whatever. He was talking about light, but the same applies to sound. I find his way of looking at the problem helpful when tackling the interaction of sound with surfaces; perhaps you will too.

Few
 
hi

Just have finished :

"Dipole Horn - introducing "Dipole Directivity Control Device” (DDCD) to the audio community


http://www.kinotechnik.edis.at/pages/diyaudio/DDCD/DDCD_dipole_horn.html

Should give an "all in one" overview about the basic lines that have lead to and are related to my development – instead of needing to jump through several threads.

It later on will be linked from my main page "Audio and Loudspeaker Design Guidelines"

Michael
 
Hi Michael

I saw your write up on your speaker project and horn discussion.
You have highlighted a number of things about conical horns and details, which are worth further discussion and are often misunderstood.

When the conditions allow them to produce a spherical patch of radiation, conical horns work well.

In your “Impacts” portion you mention a discontinuity between the driver exit wave front and that required by the horn. In number two you mention when you do not drive a conical horn with a “breathing” source (I like that reference by the way) there is a problem.

I think these are the same things really. What I have seen is that when you have a conical horn and you drive it from a dimension where the throat / apex is less than a quarter wavelength across, a planar wavefront will assume a “breathing” source profile without producing higher order modes. At this acoustic size or smaller, one can drive the horn from several locations on the horn walls and that still acts like a point (breathing) source.
There is an “acoustic size” issue strongly at play here, below about ¼ wave length in size, sound pressure propagates like a fluid, even 20KHz can bend around a 90 degree angle without reflection IF the dimension is small enough.
The one inch exit of a compression driver is already a wave and a half across at 20KHz, it’s size is already partly defining the radiation angle and not all compression drivers produce a plane wave at the exit either.
If one assumes the speed of sound and examines many compression drivers, one finds some are likely producing a converging wavefront, not a plane wave or what is ideal for a real horn, a diverging (breathing) wavefront.

I have mentioned a driver a few times here which I use a ton of and works particularly well in conical horns. It has a small internal origin, has a small conical horn leading to the exit (produces an expanding wavefront) and has only one path length to the radiator. That is the BMS 4550, it is excellent in conical horns..

I think the shape in fig 13 is a good one.
You will find Don Keele’s rule of thumb for pattern loss and pattern waist banding issues fixed adding a second flare angle will apply here.
If you extended that profile to a point at the apex by using a “breathing source”, extend the mouth to a larger dimension and make the front baffle / curve in three steps, that is pretty much the same shape we use at work. In our case, with a diverging source, I try to transition into the conical as fast as possible where the dimension is as small as possible.

Also, something I observed too was that the larger one made the mouth, the lower the hf energy was everywhere outside the pattern (like at 90 degrees off axis or behind).
This suggests that the larger the mouth is, the less important the curve at the mouth is as these horns all had the same profile (a flat baffle and two angles), but showed lower hf energy diffracted at the edge. It appears a horn with a mouth break / termination at say 50 wavelengths across will have greater forward directivity (in pattern energy vs out of pattern energy) than one 10 wavelengths across..

A last though on the mouth reflection, this is a real and strong thing in acoustically small bass horns but I am concerned that your higher frequency models may not be an acoustic size which shows the radiation resistance in scale if you still see strong moth reflections..
I mean that by the time the mouth is approaching the ”right size”(approaching 1wl in circumference ) , there is little reflection back. At higher frequencies, it is not hard to get to that size.

I had a thought on your cool hf driver experiment too, particularly fig7 and 16

A concern is that you don’t want to have horns with parallel walls of a significant acoustic dimension unless you really have a plane wave.
Where the walls are about a half wavelength across, there is a strong chance of having a cancellation notch. Energy is transferred into an “across duct mode” and that produces the notch. You can avoid this by having an angle in the Vertical part of the horn as well.
You have a source, which is several inches tall, large enough to be controlling the radiation angle in that plane but doesn’t radiate a plane wave, and when the size of a source is controlling directivity, what it does is frequency dependent.

Your source will then traverse a region where it has little pattern control and the horn will do it and then where it is controlling by virtue of its length.
In commercial line arrays they reduce the severity of the frequency dependence of this effect by curving them physically or electronically, to in effect make them a sort of point source.


Consider what is needed to reach that “breathing source” condition in a rectangular slot throat?, that requires that you retard the ends of your drivers radiation to impart a curvature consistent with the angle of non-parallel walls in the tall dimension, when the driver itself is acoustically large enough to be defining the angle.
A rough rule of thumb for this transition point was that when a piston source is about a wavelength across, it’s pattern had closed down to about 90 degrees.
One might measure the vertical radiation angle without the horn at say midband, then make the vertical angle of the horn equal to that.
Above that frequency, the driver will normally have a progressively narrower vertical angle
If the driver were sitting normally, one might put a wedge of “Earl foam” in the top and bottom of the horn so that the sound has a progressively longer path to travel through as you move towards the ends.
If you have a way to measure time, you could measure how much time an inch adds and calculate a “shape” for the lens.
In any case, what you want is the horn to define the angle and as your wonderful electrostatic concept shows, when the driver’s acoustic size takes over, it has to produce the same shape wave front “or else” all that “fancy HOM stuff” starts happening.
By the time the wave front is large enough across to have directivity, you better be done bending it in a normal horn or bend it ever so slowly and carefully as in Earls horns..

Your general shape would allow you to avoid another common horn problem not discussed here, which is “pattern flip”.
When you make a normal shape horn that is asymmetric, say 30 by 90 degrees, one finds the plane conical shape losses pattern control in the 30 degree plane long before the 90 and then radiation pattern “flips” 90 degrees.

The old electrovoice T-35 tweeter (a small asymmetric horn) works better mounted “up and down” because it’s pattern flip dominated nearly it’s entire usable range.
Anyway, for a narrow vertical angle horn and wide horizontal angle, the ideal proportion is tall and narrow, like yours. If you dig up Don Keele’s paper, I think it was called “what’s so sacred about exponential horns” and another one on “constant directivity horns” these have an approximation for pattern loss frequency vs size vs horn wall angle.
Ideally you want to loose pattern in both planes at the same frequency, then no pattern flip.
Anyway, a great w rite up and visuals too, on a subject near to my heart, good luck with your driver project, very cool!.
Best regards,
Tom Danley
 
Tom Danley said:
Your general shape would allow you to avoid another common horn problem not discussed here, which is “pattern flip”.
When you make a normal shape horn that is asymmetric, say 30 by 90 degrees, one finds the plane conical shape losses pattern control in the 30 degree plane long before the 90 and then radiation pattern “flips” 90 degrees.

The old electrovoice T-35 tweeter (a small asymmetric horn) works better mounted “up and down” because it’s pattern flip dominated nearly it’s entire usable range.
Anyway, for a narrow vertical angle horn and wide horizontal angle, the ideal proportion is tall and narrow, like yours. If you dig up Don Keele’s paper, I think it was called “what’s so sacred about exponential horns” and another one on “constant directivity horns” these have an approximation for pattern loss frequency vs size vs horn wall angle.
Ideally you want to loose pattern in both planes at the same frequency, then no pattern flip.

Great info Tom! This is the first I've heard of this pattern flip. Cam you explain a bit more? Are you saying for a typical high aspect rectangular horn in a typical position, when you go below the cutoff frequency in the vertical direction the vertical directivity now behaves like the wider horizontal directivity? And at the same time the horizontal directivity narrows to be more like the vertical was above the cutoff?
 
The Altec Tech Letters section is worth visiting. In particular, view Tech Letter 201-A, which has vertical and horizontal directivity information for a variety of Altec horns and loudspeaker systems.

Scroll down to the middle portion of the PDF and look for the dashed (horizontal) and dotted-dashed (vertical) lines (ignore the solid line). You'll see several of the sectoral horns exhibit the well-known "pattern flip", becoming more directive in the horizontal (and less directive in the vertical) plane at lower frequencies.

The frequency where the vertical and horizontal lines cross is the actual "pattern flip" frequency, which read critically, implies something about the suitability of that model horn for high-quality applications below that frequency. There's a reason that Altec reserved the more-expensive multicells for the most demanding applications - compare the multicells to the sectorals for smoothness of pattern control over a wide frequency range. The discerning customer could read through the corporate PR-speak and see for themselves which horns had the smoothest pattern control over the broadest frequency range.

Tech Letter 255 is also worthy of very close reading - compare the pattern control of the multicells to the sectorals and you can see where they pull ahead. It also shows the patterns of the one of the earliest constant-directivity horns, the Mantarays.

Tech Letter 211 has a crystal-clear explanation of "Q" and directivity. Tech writing at its best. Tech Letter 183, by Don Davis, is also a good read.
 
When looking at Tech Letter 201-A, compare the 511 and 811 sectoral horns to the 1003 and 1203 multicell horns. Note the 2.2 kHz pattern flip of the sectorals to the 550 and 750 Hz pattern flip of the multicells. Guess which horns will sound best in the 800 Hz to 2.5 kHz range?

It's bad enough to wrangle with crossover lobing (and nulls) in the vertical plane without also having to fight a horn that is operating below its pattern flip as well - any wonder that designing a crossover for a sectoral horn is a non-trivial exercise? Looking at this data, I can see why Stephens stayed with multicells and avoided sectorals in the Tru-Sonic product line.
 
Administrator
Joined 2004
Paid Member
Funny! About the time Tom was posting - I was over reading some stuff at the Renkus-Heinz site about their "COMPLEX CONIC HORNS". They claim no pattern flip. Patten flip was news to me, too.

(One doesn't see R-H mentioned here much, hard to buy the raw drivers, but they are nice in my experience)

Lynn, thanks a million for the Altec links. Should help me with my 803B horns.
 
I think fig 13 in Michael's writeup is only the best I could come up for now. Ideally I would like better Constant Directivity in the horizontal direction. The next step would be figuring out how to do that while still maintaining a uniform roll off after the CD angle. This is going to be tough.
 
When the mouth of a waveguide is greater than a few wavelengths then its polar pattern is governed by a geometric "ray" model wherein the response is almost a direct one-to one relationship with the mouth velocity. But at a wavelength or lower mouth dimnesion, the mouth becomes diffraction limited and is governed only by its dimensions - wall shape being irrelavent. So a wide waveguide with a narrow vertical dimension will have a wide and narrow polar pattern above the point where the dimensions are greater than a few wavelengths, but below this point the polar response will go wider in the vertical than the horizontal because a narrow dimension diffracts wider than a wider one. For a square device the polar angle cross the diagonal will be the narrowest before the mouth size limitation and the widest above the mouth size. This is why square is not good.

This complication is also why I say that it is not at all clear that an elliptical waveguide is a better option that a circular one. That all depends on the crossover point and the waveguides dimnesions. I suspect that an elliptical waveuide would need to be as large as the circular one in its narrowest dimension for it to work comparably, but then this makes for a pretty big device.
 
Hi Guys

I know this will sound odd but trust me it’s true, speaker companies or most any company selling something, is not likely to talk about short comings, especially if there isn’t an obvious fix..

I have been told I have to be careful about what I say about other manufacturers so instead I can show you on one of our products what pattern flip looks like.
From then on, when the needed information is supplied, you can see it too.

In the screen shot below, I have three panes of the CLF viewer open (CLF is common loudspeaker data format). This is a large 60 by 90 degree conical horn used in large spaces like theaters.
This is about as asymmetric as I feel comfortable making a horn because this problem is more pronounced, the more asymmetric the coverage angle is.

Our full range speakers are measured by a third party, in a full spherical pattern every 5 degrees, in a large acoustic space which puts the microphone at 21 feet away.
That data is crunched into EASE models for system designers and with a power test, included in the CLF model. The “Max input voltage” is defined as the level where at any point the response shape has changed by 3 dB from the 1Watt shape. Sort of a practical or usable loudness rating rather than failure..

What you see are three instances of the viewer, it is showing the 3d radiation balloon, with the balloon (speaker) facing you.
On the left, you see the balloon at 315Hz and pattern flip is visible in that the pattern goes up and down where the horn is positioned 90 degrees horizontal.
In the lower left, one has the V and H polar plots also at 315Hz.
The middle set is the horn at 1.6KHz when the pattern is established, note it is now the intuitive shape given the horn.
The right panel is at 16KHz with the balloon turned to the side a little.
What one can also see is that the lower right panel I selected the –6dB beam width graph which shows a reasonably constant directivity, which gets narrow out at 16KHz.
This is because I had to use a 1.4 inch driver on this cabinet and 90 degrees is a little wider than a source that size can drive that high (source size directivity).

If you download the model and viewer on the download page, you can see much more.
Also one can see this way of constructing the CD horn as a multi way system, avoids the nulls and lobes at crossover which a larger spacing causes. At each “seam” the ranges are added at a quarter wavelength or less separation so that the horn defines the pattern.
The front to back separation offsets the phase shift of the crossover in most cases leaving the appearance of no crossover in the system phase..
Hope that helps, got to run.
Tom Danley
 

Attachments

  • clf data sh-96.jpg
    clf data sh-96.jpg
    81 KB · Views: 980