OmniDirectional - work in progress

Hi DonVK,

Interesting work - please go on...

This is a conical reflector, mine is actually curved. So next step is to look at importing CAD files.

Just as a note: I did some ABEC simulations of an Omni within the ABEC-Thread in a German DIY-forum. Please indicate if I should tranlslate anything...
It shows as well the import of a CAD file derived from FreeCAD. I now tend to use Autodesk Fusion 360 though.

BW,
Christoph
 
Thanks Christoph, but the link you provided to your ABEC reflector sim is broken. I am interested in seeing that work, every bit helps.

My lower bass reflector is also curved (see previous photos). The last grpah was just a to see what a pure conical would look like (I used to have one, before curving it).
 
Thanks, I can read the update.

That's using "symmetry" in the control solver, to reduce computation time. The computation time is not very long for this design (so far ;)). Unfortunately all model elements would also need to share this same symmetry.

I've been able to use "shells" to experiment with shapes without having to use CAD (would be better). Like this pseudo curve (spline).
 

Attachments

  • curved reflector.jpg
    curved reflector.jpg
    232.9 KB · Views: 336
Shells seem to be a quite powerful tool in ABEC.

That's using "symmetry" in the control solver, to reduce computation time. The computation time is not very long for this design (so far ). Unfortunately all model elements would also need to share this same symmetry.

True. However, you could simply rotate your 2D-scetch full 360° instead of 90°, then export the respective *.stp or *.step file and mesh in gmesh.

Mesh files then can be imported (and placed) in ABEC.

Christoph
 
Finally,.. a full ABEC working simulation

Well,.. that was more painful than I expected.

This ABEC model is a full representation of the Omni speaker with all components. It has the electrical LR4 filters, woofer, compression driver, waveguide and reflectors. It still needs some tuning for correct wall impedance and more precise compression driver model (it was a guess). You can see the woofer (LF), both drivers (MF) and only the CD (HF). The VACS (graphing package) auto ranges and solarizes the image so you can't directly compare SPL colors, but they are close.
 

Attachments

  • All working @43Hz.jpg
    All working @43Hz.jpg
    199.9 KB · Views: 289
  • All working @1363Hz.jpg
    All working @1363Hz.jpg
    212.3 KB · Views: 272
  • All working @2000Hz.jpg
    All working @2000Hz.jpg
    214.3 KB · Views: 275
Why are the pressure streams not torroidal ?

I was expecting, like in the lower LF case, for the pressure to radiate in a torroidal fashion from the tower. It does this at LF but for higher frequencies the pressure "bends" upwards. It would be easy to dismiss it as due to an upward firing woofer, but its not the pattern I want. I think it may also be contributing to an audio defect that I can hear w.r.t the fine details.

There is some very good insight that I garnered from Patrick Bateman's many waveguide posts, in particular his attempt to bend sound from small form factor waveguides for cars. Do both sides of a waveguide need to be equal length for a waveguide to interface properly to the external air? Mine aren't.

The discs are Rd=20cm, from woofer center to the edge. The LF path above is a curved surface, lets call it a 1/4 circle for convenience. So to be equal paths Rd=1/4 * (2*Pi*Rr), or solving Rr=12.7cm or Dr=25.4 much small diameter than the lower disc. It sounded wrong, a smaller reflector would cause more bending up, shouldn't it?.

Well, the woofer Dw=18cm, and starts getting directional at 1/4 wavelength, or in this case (344m/s / 4*18cm)= 477Hz and above, so we should see a change at that point. Its easy to try both small reflector and large reflector to see the difference.

The first pic (meshed) shows a large reflector and you can see the bending (the yellow nose shape waves ) and compare it to the smaller reflector (spherical, symmetric). Well it looks like it was true. Next improvement shapes.
 

Attachments

  • 458 - large.jpg
    458 - large.jpg
    232.3 KB · Views: 537
  • 458 - small.jpg
    458 - small.jpg
    220.6 KB · Views: 504
Waveguides - recalculated

I thought I would go back to basics and calculate what shape the waveguides should be rather than use best guess and experiment. I have noticed some reduced clarity and lack of "smoothness" in the freq response and wanted to refine the design.

There is an outstanding paper "Horn Theory, by Bjorn Kolbrek" that provides the theory in a concise and very readable format. The original at http://kolbrek.hoyttalerdesign.no/images/misc/horntheoryintroarticle.pdf (all figures), and its also in AudioExpress http://www.audioxpress.com/assets/upload/files/kolbrek2884.pdf as a publication. I used this paper to work out some theory and the numbers for exponential horns. No reason for exponential other than the math is easier and they seem to be established and well understood. In all cases I'm assuming a spherical wavefront.

My HF design uses coaxial waveguides to form the inner and outer walls. It turns out my earlier experiments determined that a 6-8cm is a good spacing. It makes the co-axial design response similar to a single waveguide axial response. After running the math, it seems I have a some adjusting to do, but the experimental spacing is good. The tip shape of the inner wall is wrong (I used round, but it should be pointed). Due to the offsets, the inner wall needs to have a smaller exit diameter that the outer wall in order to make the wall path lengths equal. Easy enough.

My LF design has the woofer mounted to a flat disc and a reflector above it. The shape required to make this an exponential expansion, assuming spherical wavefront is odd, but the math works. In an earlier experiment, the low freq does not need a reflector at all, but as we move into mid freq the woofer becomes directional and it does need it. Again the reflector diameter needs to be smaller in order to maintain equal inner and outer wall lengths.

The pics below are the design calcs. The "yellow" is the Dayton waveguide shape and you can see the same shape, shifted. The "green" is the wall path length. The "blue" is the range of suitable solutions to make all the paths nearly equal. The left side graphs shows the path cross sections and the right graphs show the spherical wavefront area and its exponential curve fit.
 

Attachments

  • HF waveguide design reduced.jpg
    HF waveguide design reduced.jpg
    326.8 KB · Views: 453
  • LF waveguide design reduced.jpg
    LF waveguide design reduced.jpg
    313.5 KB · Views: 451
Last edited:
Omni V3 - new cones, new test results

The new waveguides / reflectors were reshaped and tested. The first pic shows the new setup, Omni V3.

I previously tested the axial response of the woofer and tweeter without the disc + cone structure and I know there are bumps at 500Hz, 2.3Khz, and 17KHz. These are inherent for the driver and LR4 crossover combination used. If you look at the FR graphs you can clearly see these 3 bumps. This means the polar response is now very close to the axial response. You can also see the rolloff at 100Hz from reducing the area above the woofer. I have also converted the woofer cab back to a 16L BR with 5cmx13cm port because its behaviour is simpler than the ABC. As an added bonus I got an MF+HF gain with the new shapes, that makes the LF look lower by comparison.

The "bumps" can be easily EQ'd out but the ripple would require something like a 1K tap FIR to improve any further. The ripples could be from mechanical issues, or disc diffraction, or room interaction.
 

Attachments

  • 1 - baseline BR @1m mid disc.jpg
    1 - baseline BR @1m mid disc.jpg
    138.9 KB · Views: 282
  • DSCN1859 reduced.jpg
    DSCN1859 reduced.jpg
    378.5 KB · Views: 453
Vibrations, reflections, resonances.....

I thought I'd try the obvious first. Make the entire enclosure free (or reduced) of vibrations, reflections, and resonances. Unfortunately, energy tends to go everywhere and multiple fixes were required. My initial "test" was to tap it, and if it sounded dead, it was good.

1) stiffen upper 2 discs with glued 1" thick polystyrene discs
2) damp residual vibrations from top 2 discs by compressing polyester fill between the discs.
3) damp reflections between woofer disc and first cone by adding acoustic tile CAC35 to woofer disc.
4) damp reflection from woofer BR chamber by adding acoustic tile to chamber floor.
5) add loose polyester fill to woofer chamber

This reduced some of the ripple, but not all of it.

Graph#1 is the baseline (new cones), graph#2 is after mechanical treatment (above), graph#3 is EQ to remove 3 driver bumps, graph#4 is adding a 10" subwoofer to fill in the LF. All measurements are at 1m with room acoustics.
 

Attachments

  • 1 - baseline BR @1m mid disc.jpg
    1 - baseline BR @1m mid disc.jpg
    138.9 KB · Views: 140
  • 12 - acoustic tile CAC33 bottom of woofer chamber damping.jpg
    12 - acoustic tile CAC33 bottom of woofer chamber damping.jpg
    138.7 KB · Views: 103
  • 13 - acoustic tile CAC33 bottom of woofer chamber damping corrected.jpg
    13 - acoustic tile CAC33 bottom of woofer chamber damping corrected.jpg
    138.9 KB · Views: 106
  • 13 - all damping + EQ + sub.jpg
    13 - all damping + EQ + sub.jpg
    139.2 KB · Views: 138
An improvement over the earlier version. They sound very good. I've put in several hours of listening to various type of music so far. I did manage to improve the clarity and detail with the latest changes.

The sound stage is wider than a conventional speaker and it sounds more natural to my ears. The tone is constant as I move around and good imaging occurs in a larger "sweet" area.
 
They are placed 1.3m from the wall so 7.5ms delay from the incident wave at my listening position.

These speaker FR measurements were taken at a spot where the room opens and there are minimal reflections (other than room modes) so I can isolate (mostly) the speaker response from the room.

There are other measurements of the room interaction with conventional speakers and the Omni V2 at http://www.diyaudio.com/forums/multi-way/306460-speaker-room-interaction-pics-your-opinion-3.html . These were all taken at my listening position.
 
Diffraction

Something is causing those ripples. Actually they look like a series of peaks and nulls.

S. Linkwitz did some excellent work characterizing the effect of baffle size on diffraction due to the change from 2pi to 4pi space. The experiment and results are at Diffraction from baffle edges . My setup is a little different but the effect should still be there and it would be easy to calc it.

I took these measurements after the driver bumps were EQ'd out, leaving just the ripple behind. The new FR graph is at 1/24th octave using swept sine to grow the grass so I can find narrow peaks and nulls. I extracted 20 peaks and 20 nulls from 100Hz to 1800Hz. This is the range that the woofer operates in, and according to Linkwitz, the magnitude of the diffraction decreases with frequency so diffraction should be largest here.

The spreadsheet shows a very strong indication of 5 sources of diffraction extracted from the repeating patterns of nulls and peaks. The peaks caused by multiples of even fractions and the null caused my multiples of odd fractions. This maps very well to Linkwitz's experiments. There is a column for peaks and nulls with each diffraction group source mapped to a common colour.

Next,..... try to reduce a diffraction source and see if I can measure it, and more importantly can I hear it.
 

Attachments

  • Omni V3 - Diffraction issues - reduced.jpg
    Omni V3 - Diffraction issues - reduced.jpg
    168.3 KB · Views: 118
  • REW FR 24th octave.jpg
    REW FR 24th octave.jpg
    250.3 KB · Views: 121
Interesting analysis of the diffraction. Nice to see you have got REW working. If you want to just get a screenshot of the window you are working on press the camera/capture icon this will take a screenshot of the window only and you can add a label and choose the size. Ends up being much easier to see :)