Site Links
Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Categories
What do you think of this . . .
Link: 
. . . as a speaker cabinet construction technique? A large sphere would be the ideal shape for a baffle, at least in terms of diffraction. So, just as long as you do not cut large circular flat spots on the sphere to mount the drivers, you should be OK. No bracing required, as the pressure build up inside the sphere would be equally distributed.
The author makes this 19" sphere out of 3/4" birch plywood, so the internal volume would be 1.6 cu. ft. That would be just about right for a couple small 6 inch woofers mounted on the sides. I'm thinking MTM with a small 1 inch hole drilled for a press-fit tweeter. Then flank the small tweeter with a couple small 3" flat membrane type mids. Thoughts?
Comments
Sourcing a suitable sphere is a problem. Building those tools would be even tougher.
If you just wanted the spheres it would be fairly easy to cut on a lathe or router jig. Cutting two half sections would make it a lot easier.
Or I could line them internally with lead roll roof flashing and then call them "The Lead Balloons"
Yes, really pricy. But this would be strictly a DIY affair, if I decide to try it. I have a nice table saw and a dual bevel sliding miter saw, so making up the basic ball would be a time consuming, trial & error process; but doable. I don't have a lathe, but once I get the sphere put together, I could hand sand the giant ball into a perfect sphere using my palm and belt sanders. Time consuming, but doable.
I don't have a lathe, so two half sections would not work for me. A router jig would work; but only if I could figure out how to build one to produce something this big. I know Nick S. (uglywoofer) made a special router fixture to create his "Stink eyes" speakers, but that was much smaller.
Building a giant plaster tractrix horn seems like more fun.
There is no doubt that if I decide to go ahead with a project like this, it is going to be a real challenge. Probably a year from start to finish.
The author gives some tips in youtube comments as to how to zero in on the proper angles. You start with 1/8" thick tempered hardboard and then cut up a bunch of small triangles without bevels. Then you layout an experimental hemisphere by hot gluing intermediate sections together. Once a perfect hemisphere is obtained, angle measurements can be taken and then transferred to the table and miter saws.
I was looking at the exclamations speakers and noticed that this sphere has a fairly large flat spot cut on the front to mount the drivers. Even though the driver cones are not equidistant to the edges, the edges are still there and will diffract and create phantom images. On my speaker, I want to attempt mounting my drivers onto the curved surface of the sphere so as to avoid an edge transition. To do this, I would have to carefully grind off a portion of the plastic driver flanges to match the curvature of the sphere. There would still be a small amount of flat (or curved) surface created by the driver itself, but that is unavoidable.
I was also thinking, audio is perceived in the horizontal plane. Would a vertical cylinder measure the same as a sphere?
I am looking at a 19" diameter sphere to allow enough room for a good woofer alignment with a couple 6 inch woofers mounted on the sides.
Based on my understanding, a cylinder will not measure the same as a sphere. Olson actually compared a cylinder type shape to a sphere in his original paper. The sphere produced the smoothest overall responce is his tests.
http://www.aes.org/aeshc/pdf/how.the.aes.began/olson_direct-radiator-loudspeaker-enclosures.pdf
Here is what I come up with using Ken's trans laminated ring idea. This is a 2D cross section showing a total of 13 rings per hemisphere. Only the exterior bevels are shown, as the interior of each ring does not need a bevel. Note that the bevels start at 2.3 degrees, then 6.6 degrees, 11.1 degrees, etc. The bevel angle increases as the ring diameter decreases. The first 7 ring layers could probably be rough cut with my jig saw. But the bevel angles above 30 degrees would have to be cut by CNC or some other method. If I did this, I would probably cut ring layers 7 to 13 all at 30 degrees and then grind off the excess later with my belt sander.
I did some prototype testing on a makeshift 12" diameter styro sphere to see how it would go (or not go, as the case may be). I wanted to see what kind of internal box resonance problems would be created by an internal spherical enclosure shape.
I created my test sphere from two FloraCraft 12" diameter styrofoam domes, available at Michaels for about $12 each. The domes are about 7/8" thick, so the inside diameter is approximately 10.25" On-line sphere calculators yield an internal volume of 0.326 cubic feet (9.23 liters). I used Mortite window caulking cord to "glue" the two domes together to create a sphere.
I cut holes for a pair of Eminence Alpha 3-8 three inch full range drivers, placing one on each side, 180 degrees apart. I "glued" them in position with Mortite. I cut a small 2-3/8" diameter hole on one side for a 2" ID PVC BR port. I varied the length of the port from 1.5" up to 6" to change the tuning from roughly 100Hz (fairly flat alignment) down to 60Hz or so (very droopy extended bass shelf alignment).
When I scale things up for the 19" diameter sphere (17.5"ID) , the peaks and valleys above 300Hz will move to lower frequencies based on the change in wavelength distances involved. Based on half-wavelengths, I am estimating 1130*12 = 13560 / 17.5 = 774Hz verses 13560/10.25 = 1323Hz. So, as a guesstimate of what my bigger sphere will do, I need to drop all the peaks and dips by roughly one octave. So the peak at 875Hz will become 437Hz in the larger sphere. The peak at 1.8kHz will become 900Hz, etc.
In the final system, the crossover will be 350Hz at 12dB/octave electrical. So this should help to keep these peaks under control. The nice thing about using two woofers on opposite sides of the sphere is that the wavelengths are essentially cut in half, boosting the peaks and valleys to higher frequencies. This also solves a problem with constructing the sphere, allowing me to leave large 6 inch diameter flat spots on the top and bottom during construction.
Also, placing the port entrance along the outside circumference of the sphere (instead of at the center of the sphere) helps quite a bit to smooth things out. This probably has something to due with the fact that a sphere concentrates a great deal of its acoustical energy directly towards the center.
I messed up a little bit with my previously posted NF measurements. In OmniMic, you must check the "ALL" button when making NF measurements. I accidently left the button set to "blended" which then applied a 5ms gate. This gate chopped off the NF measurement before the internal box echo had a chance to die off completely on it own. This gave the peaks and dips more of a rounded off type look. I re-did the measurements and posted the corrected graphs below. They use the "ALL" setting, which produces sharper looking NF plots with more well defined peaks and dips. I also fixed a couple air leaks on the spheres which helped to clean up a few strange ripples on the previous graphs.
For comparison purposes, I cobbled together a test box using the same volume as my test sphere and ran some more measurements. The test box is 6.5 x 7.25 x 12" internally, which comes out to roughly 0.326 cubic feet. I was going to use golden mean ratio dimensions, but this is what I had available on the scrap pile. I mounted the drivers and port in the typical places that you would normally put them if building a small two way speaker (see pics). As you can see in the graphs, the rectangular box has a nasty internal resonance at about 580Hz, which is the same whether I used one woofer or two woofers on opposite sides of the box. This resonance comes from the boxes 12" top to bottom dimension. Using 13560 inches/second as the speed of sound divided by 12 yields a full wavelength of 1130Hz and half wavelength of 565Hz (fairly close to 580).
The sphere, on the other hand, does not develop a peak until 830Hz, in the case of a single woofer; and about 1.3KHz, in the case of dual woofers (see graphs). There was no damping material of any kind used inside either the sphere or rectangular box.
This is very good news, because it indicates that it will probably not be necessary for me to add additional flat partitions inside the sphere to break up low frequency resonant modes. The larger 19" diameter sphere will have peaks roughly 1 octave lower, but, they should still be high enough so that I can significantly dampen them with the woofer's low pass crossover parts. In addition, there will be an internal circular partition over the midrange drivers which will tend to mix up the reflective distances a bit more. I will also be lining the inside of the sphere with 1/2" or so of damping material.
I took my "styro-sphere" and re-purposed it for a tweeter test. I have a pair of CSS LD22F's and my 2.85" woofer cutout was an almost perfect fit to the LD22F's flange. So, I mounted one tweeter into one of the two woofer holes. After pressing it in, it flattened out to a perfectly flush and level mount using a few strips of masking tape. (see pic)
I think this is the same tweeter that Ani used in his Penguins and JR used in his Vermillions. Correct me if I am wrong, but I think Ani used the flat flange LD22F version and JR used the LD22C curved flange version.
I ran up a set of OmniMic horizontals. Very smooth response. The ripple notches at 1.3kHz, etc., are a function of my 12" diameter "styro-sphere" unloading into 4 pi space. Notice how the ripples disappear as you move off axis. When I build the larger 19" sphere, these ripples should reduce in size and frequency, similar to Olson's original AES paper.