Designing a speaker for narrow dispersion (& Atmos ceiling reflection)

Words from the man, himself! Thanks for chiming in here. I had come across your HyperboLine post. And I appreciate how you are trying to do something different in audio. Refreshing, interesting, and fun. šŸ˜€

You could fill the hole with absorptive material, the speaker pointing at the dish blocks (and reflects...) the center parabolic reflections anyway.
Ah, good point. I liked the idea of using the center hole to, basically, put some speakers up on a stick. But the ideas may be changing to a single driver...

I used a six foot diameter dish driven by a 13 x13 degree conical horn, and later copied it's profile to a 45" x 45" (truck box integer) reflector. Reflecting off a wall about 20 feet away, the high frequencies appeared to be coming from it.
Well, there may be some hope. šŸ˜€

Below ~2kHz, the 4" drivers would be near omnidirectional, most of their sound would spill out of the dish.
The speaker's polar pattern determines the near field response, most of it's radiated sound will bypass the reflector dish.
And from your other thread you said:
ā€œThe hyperbola should have been deeper to avoid side spill, reduce cabinet depth and width, and reduce acoustic center to center horizontal driver distance.ā€
the near field sound will be minimized to the extent of the horn's directivity, but the -6dB near field spill will be louder than the reflected ceiling sound. The Haas effect will place much of the perceived location at the point of first arrival.

So trying to reduce spill of sound from the dish would be good, if achievable. What if there was a tube around the parabola/hyperbola? To reduce side spill. So make a tube a little wider than the parabola, put the parabola at the bottom. Line the tube with some sound absorbing material.

The directional sound goes straight out of the tube. The side spill goes up the tube, decreasing in intensity as it hits more absorbing material. Probably not great and there are wavelength/frequency limits of sound absorption and dimensions, but it blocks and reduces the sound that would go directly to the listener straight from the driver. Making the directional sound a little more prominent. This is probably too simplistic and there is some reason it would not work?

Of course there would be physics limits. One lower frequency limit of directional sound would be the radius of the parabola. Another is from the driver, itself. So go for as large of a driver as possible that would still be full range. At least 4ā€ would work, maybe larger if there are any suitable ones. Maybe try to add a horn of some sort for a little more directivity. Combine things and the best case might be directional sound down to ~1,000 Hz. Or not. šŸ˜€
 
I was a telescope optics nut long before I was a diyaudio nut. First a sanity check: paraboloid surface is very close to spheroid surface (correction done during mirror polishing not grinding); and its focal point is about halfway between center of mirror surface and center of the sphere, meaning inside the "bowl" unless nearly/uselessly flat. And, the speaker's acoustic center (to be located at the focal point) is well back of the membrane. (Sketch it.)

That said, there are tricks (some I came up with) to go a bit off-axis "unobstructed" etc.

My reflector-coaxial point-source series of speaker experiments (PrimeRadiant) are the opposite of your stated goals. But there might be something relevent in the discussion that began here:

(analogue of Newtonian or Off-Axis-Newtonian telescope.)
 
A point source "sprays sound everywhere" - it is its main feature and superior advantage i.e. to create an infinite amount of identical images of the input signal emitted in all directions. Try make one šŸ™‚

A bounce (or 2... many) creates image / duplication - this is one way of trying to mimic a point source and its diffuse sound field characteristics.

//
 
If you still want to build something practical in size and form, why don't you build a small MTM with a narrow waveguide tweeter and two, maybe 4" drivers, tilted to the tweeter a bit. Like a mini speaker array. Do a matching 12dB crossover, between 2 an 3 kHz. Just like building a usual TMT speaker. With reasonable affort, this will give you the best directional sound you can get in a small space. From 300-400 Hz up this will give you quite high SPL if needed. You may even find some DIYS plan you can copy and modify a little, to reduce development time.

Then place these new speakers inside your room, into the position you want and tollerate, maybe on top of your L+R floor standers or even better, on some adjustable stand. Just at the position they would be, to have better options in directing them.
Now do two things:
First play some music, only from these speakers. Try to find a position/ direction that gives you the illusion of sound comming from the ceiling, while sitting in the listening spot.
If you can get a result that you like and somehow meets your expectations, start to measure. You should be able to find some kind of a system in how they work and be able to optimize the effect using data from a microphone.

All these mentioned array and reflective speaker ideas have shown not to work in conventional speakers and will not work in what you are aiming for as well.
Also you will not be able to design a multi amp system which DSP's each single driver to do some crazy phase bending stuff and to steer a beam in some direction. Leave that to the multi billion companys.

With your specific room, you can do two things. One is the practical test I promote, with a speaker well known in it's principle. The other one would be a professional measurement with resulting simulation. Which, IMO, would be a litttle too expensive and also not give a directly useable result for what to do...
 
The directional sound goes straight out of the tube. The side spill goes up the tube, decreasing in intensity as it hits more absorbing material. Probably not great and there are wavelength/frequency limits of sound absorption and dimensions, but it blocks and reduces the sound that would go directly to the listener straight from the driver. Making the directional sound a little more prominent. This is probably too simplistic and there is some reason it would not work?
The concept is OK, but the question is what is the goal for the "work" the speaker system accomplishes ?

Overhead speakers with a wide dispersion pattern pointing downward at 2-3 times the listening ear height will accomplish the closest replication of the Dolby Atmos system cinematic environment.

If parabolic reflectors in absorber tubes something the size of a few stacked 55 gallon drums worked "perfectly" directing a column of sound to the surface it reflected off, the area of reflection would mirror the dish size.
The huge parabolic speakers wouldn't work as well as small pendant speakers, if the goal is replication of the Dolby Atmos system cinematic environment in a vaulted ceiling home environment.

Art
 
  • Like
Reactions: wchang
A point source "sprays sound everywhere" - it is its main feature and superior advantage i.e. to create an infinite amount of identical images of the input signal emitted in all directions. Try make one šŸ™‚

A bounce (or 2... many) creates image / duplication - this is one way of trying to mimic a point source and its diffuse sound field characteristics.
My understanding is that you can, in theory, use regular MT speakers for ambient sound. And point them up at the ceiling for some reflections. But, in practice, it's too much sound all over the place. Doesn't work well or at all. And that you do really want some directivity. Which would be either an actual speaker up in the ceiling. Or a reflection that has a lot of directivity. As I am learning here, the reflection case is not easy. šŸ˜€
 
  • Like
Reactions: aslepekis
If you still want to build something practical in size and form, why don't you build a small MTM with a narrow waveguide tweeter and two, maybe 4" drivers, tilted to the tweeter a bit. Like a mini speaker array.
Right. So let's follow the idea through...
An MTM speaker can have vertical directivity. But wide horizontal dispersion. Put the MTM on it's side and you get the opposite case. So to get both horizontal and vertical directivity, you end up with a mini 2X2 array of W's.

You can put a tweeter in the middle. So it's a square of 4 M's with a T in the middle. Restrictions are that you would need a Linkwitz-Riley second-order crossover (LR2) to hit the 12 dB/octave needed to preserve the directional lobing. Hitting real LR2's is possible, but challenging. Not a given that the end result would be a good frequency response. Specific driver choice for this goal would help. Another issue is that larger diameter drivers are better at staying directional to lower frequencies.

So maybe a better option is an array of full range drivers. The larger the diameter the lower the frequency that they will stay directional. But the drivers cannot get too big or they won't play the high frequencies. I still need to do some modeling on this one to compare different arrangements of drivers, how, say, 4x4 of 4" drivers compares to 2x2 of 8" drivers, etc.

At least this is how I understand things. An alternative option is the whole parabolic reflector that we are also talking about. So there may be two routes worth more exploration here.
 
Yes, I found some interesting things. Most notably this interesting read:
You are in there. šŸ˜€ Not exactly what I’m after, but definitely interesting and relevant. Basically how polar plots change for linear arrays. There are a lot of insights, it goes in a different direction, and I’ve not yet digested it all. This quote jumped out at me:

ā€œI've noticed with arrays in general, that the more drivers you use, the less sever the peaks and the dips are. If you have two drivers in an array you're going to get a really strong null at 30 degrees off axis. (https://www.diyaudio.com/community/threads/what-causes-off-axis-nulls.342494/) Add two more drivers, and the depth of the null is reduced. Add four more, it's even less of an issue. Etc etc.ā€



Ah, OK. Did you see the .pdf file within that link ā€œManipulating Directivity?ā€ It’s excellent. And has some of the very basic, general info that I’ve been after. The Horbach-Keele filters are even mentioned.

I still need to think things through more. But the file implies that we could get enhanced directivity by using drivers with as large of diameters as practical (such that you still get high frequency output). mattstat mentioned this earlier. Then add a waveguide or horn for further directivity. Then, from other sources, make an array of these multiple large, horned/waveguide drivers. Finding large, full range drivers with waveguides/horns may be challenging. But it's maybe an interesting direction to think about.

Did you find the Fractal array thread? https://www.diyaudio.com/community/...raight-cbt-with-passive-xos-and-no-eq.330031/
 
  • Like
Reactions: aslepekis
I was a telescope optics nut long before I was a diyaudio nut.
I can totally respect such a nerdy hobby like that. šŸ‘

First a sanity check: paraboloid surface is very close to spheroid surface (correction done during mirror polishing not grinding); and its focal point is about halfway between center of mirror surface and center of the sphere, meaning inside the "bowl" unless nearly/uselessly flat.
Good point! You got me excited for a few moments. Spheres should be easier to source than parabolas. I was looking around and came upon things like large stainless steel mixing bowls. Very cheap, available in many size, robust. But then I realized that they tend to have flat bottoms. I have not found anythings quite suitable for a half sphere. So I might be back to the parabolic reflectors that you can buy. Open to other ideas, though. It's more a matter of sourcing half spheres versus parabolas in various sizes and for prices commensurate with a hobby project.
salad.jpeg


And, the speaker's acoustic center (to be located at the focal point) is well back of the membrane. (Sketch it.)
Excellent point. I had not thought about that. I don't think that most drivers say where, exactly, the acoustic center is. But it's usually around the bottom of the cone, meeting at the spider, I think. At any rate, estimating the acoustic center about there and placing it in the reflector accordingly will definitely help.
 
The concept is OK, but the question is what is the goal for the "work" the speaker system accomplishes ?

Overhead speakers with a wide dispersion pattern pointing downward at 2-3 times the listening ear height will accomplish the closest replication of the Dolby Atmos system cinematic environment.

If parabolic reflectors in absorber tubes something the size of a few stacked 55 gallon drums worked "perfectly" directing a column of sound to the surface it reflected off, the area of reflection would mirror the dish size.
The huge parabolic speakers wouldn't work as well as small pendant speakers, if the goal is replication of the Dolby Atmos system cinematic environment in a vaulted ceiling home environment.

Art
Sure enough. Anything less than speakers up in the ceiling is going to be suboptimal. I'm not expecting perfection. But it might be good enough to try. Better than nothing. I like thinking about this topic. There is a lot of creativity here in speaker design. Kind of like your HyperboLine project. It's great to see different ideas toward a give goal and even different goals. So maybe the most important aim here is to have a fun speaker project that does something different. šŸ˜€
 
An MTM speaker
... has a non flat power response "by design" (flaw). This means that if you collect all energy out from a MTM speaker, the resulting "FR" will not be flat - there are phase suckouts in certain angels that will never be recovered. Therefore, a MTM is extra bad as an effect speaker in a multi channels system, and completely not wanted for this task as these should ideally be diffuse and power flat - as per specification.

//
 
@ it is not about diffuse, but exactly the opposite. Maybe read the whole thread. No need for any flat response at listener space, as this is impossible anyway. It is for an "effect", not realistic fullrange.

The longer you think about this ceiling reflecting speaker, you will realize that an MTM with a wave guided tweeter is the only realistic construction giving at least some directivity. In the useable, small package. This will be worth experimenting in your room and could get a maybe better result than a MT or simple wideband speaker.
Next thing realistic would be a horn speaker, which will increase size and cost to get more of a directional source.
You initial idea is to have a height source that is not up in the air and direct radiating as optimal, but near the floor has aesthetic reasons. A huge construction sitting on your main L+R speakers will not fit that bill. So all these parabolic and huge baffle stuff is not realistic at all.

The industry has given up on this ceiling reflection thing (if they ever tried) and offers simple, cheap (build) one or two way speakers, just turned upwards to the ceiling. You can take any conventional speaker and do just the same. Only difference you don't get such cheap and cheesy driver as the industry uses for profit optimization.
 
I read through this whole patent and it really is interesting. More so than many of the patents that I read for work. šŸ˜€ It provides something for consideration here.
Patent link: https://patents.google.com/patent/US6801631B1/en
Here is what such a speaker looks like:
66ceKu0.jpg


Here is the driver arrangement. Maybe it’s something like I said earlier in the thread that you want to have different distances between drivers:

lUITmm8.png


Here is the on and off axis response, with a tweeter:
300_Good.jpeg


And here is a more conventional 2x2 array of drivers, no tweeter:
Standard4.jpeg


Then the on and off-axis responses:
100_Bad.jpeg


He gets more directivity in that unique driver configuration. Interesting! He is using 6.5ā€ woofers and a tweeter. (Not sure why there is not much output above 7,0000 Hz with a tweeter, but anyways.)

There is a little more directivity in his 4 W + T configuration than the same arrangement of 4 W's but without the T. Interesting.

From the classic MTM configuration, it looks like we need to restrict crossover topology to an LR2 if we want to preserve the loving and directivity. But the patent here says that several different crossover types can be used. So I guess use a tweeter and don’t worry about the crossover type. Just get the frequency response plots to look go.

The patent mentions that you want physical alignment of the tweeter and woofer acoustics centers. So recess the tweeter into the baffle. And/or use a tweeter with a horn or waveguide.
 
Last edited:
OK, how might we reduce this to practice? Well, horn/waveguide tweeters happen to be good at playing to lower frequencies than a normal tweeter. And larger diameter woofers are better at being more directional off-axis, which is what we want here.

Let me pick an extreme case as a thought experiment. Start with a waveguide tweeter that plays low. This one goes down to 1,000 Hz nicely. $95.
https://www.madisoundspeakerstore.c...6stwgc-4-fabric-dome-tweeter-with-wave-guide/
$95 SB26STW-GRAPH.jpg


For the woofer, we want to go as large of a diameter as possible, still reach up to the ~1,000 of the tweeter, and see good off axis measurements showing it going directive at as low frequency as possible. There are likely other suitable options, but from a quick look this one seems good. At least some directivity all the way down to ~500 Hz, which is great. Huge at 13", but bigger may be better here.
https://www.madisoundspeakerstore.com/approx-12-woofers/scan-speak-32w/4878t11-revelator-13-woofer/
32w-11-fr1.jpg


Maybe this driver combination would do well in the patent’s configuration? Hmmm.

I wonder if some of the directivity in the ~1,000-3,000 Hz range might be lost to the tweeter. But if the crossover point is low at ~1,000 Hz that might be a little less of an issue.

For the moment, we can ignore the fact that the woofers cost $750 each and we’d need 8 of them. Plus we’d end up with a baffle around 30ā€ wide or larger. There are smaller, cheaper woofers that also look good, but give up some directivity. Like, say, the 10ā€ Dayton RS270 at $131. We’re still in theory land now and will let practicality come later. Just trying to think some things through and see if any other ideas are out there.
 
To summarize for the moment, getting good directivity might come from 3 approaches:
1. An array of some type.
2. The unique driver arrangement shown above.
Once I get to modeling different driver arrangements, I wonder if these two approaches will converge. It’s likely, I’m guessing.
3. Parabolic reflectors.

In theory, it might be good to combine approaches. Say an array of drivers into the middle of a parabolic reflector. But then the overall size gets out of hand quickly. The driver array could be almost 3 feet wide. So then the reflector would end up being, say, 12 feet wide. Good directivity, I bet. However if the directed sound ends up being the same size as the entire room, well, it kind of defeats the purpose. šŸ˜‰ Nonetheless, this is fun to think through.