DIY Omni Dodecahedron Speaker

[Naughtius]

As discussed here I've been considering building an omni-ish speaker, being curious what they sound like, as well as looking for a fun speaker to build to listen to podcasts from, playing from the center of my room.

Looking for omni speakers I've found three categories: The MBL style (with 'omni transducers'), speakers using normal transducers with circular waveguides, and 12 drivered dodecahedrons.

The dodecahedrons only seem to be used as noise sources for acoustical measurements, and not for playing actual content. Being stubborn and curious as I am, that's exactly what I'm going to attempt. It'll probably suck. I'm going to try anyway.

The main problem with these seems to be comb filtering between the individual drivers. At low frequencies, where the wavelength is significantly larger than the dodecahedron, the drivers will simply add. However, when the distance between two drivers moves near half a wavelength, there will be lots of cancelation.

This means the smaller the dodecahedron, the higher the frequency at which it's still fully omnidirectional. However, the smaller, the less low end.

To see if there's a practical optimum somewhere I've run some simulations - continued in the next post here.

 

[Naughtius]

Here's a 3D polar plot of a 100mm dodecahedron, with 12 2"drivers:

As we see it's pretty good up to 4kHz, okay at 6kHz, but pretty damn crappy at 8kHz. And a 100mm speaker already isn't going to have too much low end.

So, next try, here's a 150mm truncated icosahedron with 32 2" drivers:

It's somewhat more balanced throughout, but goes awkward before 8kHz already.

Then, should we just go smaller? A 50mm dodecahedron with 12 1" drivers:

This looks quite okay! But... how much low end would one get from a 50mm sphere of 1" drivers?

Looks like there's a clear reason no one seems to use this for playing actual audio.

 

[Naughtius]

That said, I'd love to build one anyway, and see how (bad) it actually sounds! I'm planning on using 12 Dayton DMA45 1.5" drivers. I'll try to get them into an actual design, run some more simulations, and update this thread when I know more.

 

[ai1]

Well, you have much more options with almost arbitrary number of drivers :
https://en.wikipedia.org/wiki/Semiregular_polyhedron

 

[Naughtius]

Well, you have much more options with almost arbitrary number of drivers :
https://en.wikipedia.org/wiki/Semiregular_polyhedron

I wouldn't expect that to solve the comb filtering problem (and still give an omnidirectional radiation pattern). I'd imagine the platonic solids - with most equal distribution of drivers - having the highest change of achieving this. The icosahedron had trouble fitting large drivers due to the triangle faces, instead of hexagons and pentagons.

Another option would be simply taking a sphere, and as uniformly as possible placing an arbitrary number of drivers on there according to the Fibonacci lattice. But if 32 drivers don't give a comb-filtering-less radiation pattern, I'm not sure a different realistic number would?

 

[ppataki]

Love these simulations! Really cool!!
Which software did you use to make it happen?
How do you plan to build the cabinet? I assume it will be 3D printed

 

[Mnyb]

You could try a two way the ball of drivers for midbass and a Tweeter with a lense on top ?

 

[Salt]

Omnidirectional is not that difficult:
https://www.visatonshop.at/en/Kits/Omnidirectional-Speakers/
https://lautsprecherselberbauen.de/Bausatz-U-Do-51-CoRound-Passiv

and so on.

And there was the 360° ribbon I can't remember the name of that looks like Data's viewing aids.

 

[Duke]

... should we just go smaller? A 50mm dodecahedron with 12 1" drivers:

Maybe the small dodecahedron could sit atop a "stand" that houses a down-firing woofer which would cover the bottom end of the spectrum.

 

[Naughtius]

Love these simulations! Really cool!!
Which software did you use to make it happen?
How do you plan to build the cabinet? I assume it will be 3D printed

Thanks! The simulations are done in COMSOL - and indeed I'll most likely print it (and / or in the end get it printed in SLA or SLS)

 

[Naughtius]

You could try a two way the ball of drivers for midbass and a Tweeter with a lense on top ?

I could! Though I think the other way around makes more sense - get a ball of tweeters / small full range drivers for a fully omnidirectional high end, then add a woofer nearby for the low end, where the longer wavelengths make it easier to somewhat cleanly cross it over when the sub and tweeter all aren't super close to each other - as Duke also mentioned

 

[Naughtius]

Omnidirectional is not that difficult:
https://www.visatonshop.at/en/Kits/Omnidirectional-Speakers/
https://lautsprecherselberbauen.de/Bausatz-U-Do-51-CoRound-Passiv

and so on.

And there was the 360° ribbon I can't remember the name of that looks like Data's viewing aids.

Omnidirectional in two dimensions - sure - point any speaker upwards and you have an 'omni' in that sense. What I'm aiming at is a speaker that also has omnidirectional radiation in the vertical plane, across (ideally) the entire frequency band.

 

[Naughtius]

A bunch of ugly designs later I've settled on making extra covers to hide the square baskets, and mounting the cable gland inside out, to simply give me a uniform ball:

Inner volume is about 600ml (excluding the drivers), meaning I'll need about 180W to reach the 2mm maximum excursion on the drivers (according to VituixCAD), which would give me this response (if all drivers were aimed the same way). Should be plenty for my purposes (and will be DSP'ed flat-ish later).

I did do some testing at home, tuning giant dips around 8kHz in my current speakers, and it's not too noticeable in voice - so increased the ball size a bit to allow for this inner volume, pushing the 'radiation pattern breakup frequency' down to 6.5kHz.

I'll start ordering parts, and pick a printing shop, and post updates when I have them.

 

[ppataki]

This looks supercool! Looking forward to the end result

 

[Hayabusa]

Omnidirectional is not that difficult:
https://www.visatonshop.at/en/Kits/Omnidirectional-Speakers/
https://lautsprecherselberbauen.de/Bausatz-U-Do-51-CoRound-Passiv

and so on.

And there was the 360° ribbon I can't remember the name of that looks like Data's viewing aids.

these? https://www.hifinext.com/veddan-origin-cylindrical-planar-and-omnidirectional-acoustics/

 

[D!sco]

Is a full sphere really necessary for a wall or ceiling mounted device? Wouldn't a hemisphere or half sphere accomplish the same without ceiling bounce?

I'm very impressed, but this seems like a standmount omni design to me.

 

[Salt]

these? https://www.hifinext.com/veddan-origin-cylindrical-planar-and-omnidirectional-acoustics/

No. Found the item now: Elac 4Pi plus
https://www.elac.com/wp-content/uploads/2021/09/ELAC-4Pi-Plus-2.pdf
Though it's a set-on-top tweeter, no midrange.

 

[Naughtius]

Is a full sphere really necessary for a wall or ceiling mounted device? Wouldn't a hemisphere or half sphere accomplish the same without ceiling bounce?

I'm very impressed, but this seems like a standmount omni design to me.

That would indeed probably accomplish the same - it'll hang near the dining table at about head height, in the middle of the room, so the ceiling / floor portions (top 3 and bottom 3 speaker) might not constructively contribute to the sound. Making it fully spherical is mainly driven by my curiosity towards what the reflections of a full omni would sound like in a room. Would playing drums sound more like real drums, since they also radiate spherically?

If it's indeed the case, I could cross those 6 drivers over at the frequency where the drivers become more directional - as to only send the low end to the drivers, where they radiate omnidirectionally anyway. Good that you mention - I optionally could work on a passive crossover for that later.

 

[Naughtius]

Parts have been ordered! This is what the inside looks like:

A last minute change was closing the threaded insert holes, worrying they might not have been air tight otherwise.

The final simulated radiation pattern is worse than I hoped. It looks like adding the covers shifted the funk down somewhat:

Super curious what it will sound like!

 

[D!sco]

Being good until 4khz isn't the end of the world. You could inset the drivers deeper and use a flare to tighten their directivity, but I can't imagine that would be easy.