BOBO Hifi Tools
FDTD Directivity Solver 1.1.0 (BETA)
A full 3D acoustic simulation of the horn you designed in the Horn Design Tool. Rather than approximating the mouth as a radiating aperture, it solves the actual wave equation on a 3D grid — so diffraction around the horn, and its true horizontal and vertical directivity, come out of the physics rather than an assumed pattern.
Method One broadband pulse is simulated and Fourier-transformed, so a single run yields every frequency in the range at once.
Counter-intuitively, a narrower range costs more time, not less. Frequency resolution is 1/T — to separate features 10 Hz apart you need 0.1 s of recorded sound, so zooming in demands a longer recording. Raising the top bound is expensive for a different reason: it forces a finer grid, and it pushes the microphone arc further out (the far-field distance grows with frequency), which enlarges the grid in all three directions at once.
Lowering the bottom bound is cheap. Below the frequency where the mouth becomes acoustically significant the answer is known in closed form (see Low frequencies), so the recording only has to be long enough to resolve the part that is actually simulated.
Method Virtual microphones on an arc around the horn, in both the horizontal and vertical planes.
This is the biggest single lever on run time. Up to ±90° the simulation only needs space in front of the horn; going beyond that requires modelling the space behind it too, which substantially enlarges the grid.
Method Sets how finely the horn's throat is resolved, which is what governs the cell size across most of the usable band.
Cost rises as the fourth power of cell size, so the three settings span roughly ten times faster to five times slower. Draft is for shaping a design — fast enough to iterate. Fine is for a final check. Running the same range at Draft and Fine and comparing is also the honest way to tell whether a feature is real or an artefact of the grid: anything that moves between them is numerical.
Method Below the frequency where the mouth becomes acoustically significant, the rows are the exact analytic result rather than simulation output.
A source that is small compared to a wavelength radiates equally in all directions whatever its shape — at ka < 0.5 a horn mouth is omnidirectional to within 0.3 dB. There is nothing to compute there, and a simulated answer would be worse: the microphone arc is closer than a wavelength at those frequencies, so it reads the horn's near field instead of its radiation pattern. Using the closed form is both more accurate and much faster. The status line says when this applies.
Level relative to on-axis, angle against frequency, in 3 dB bands. Dark maroon is 0 dB (loudest); deep blue is 30 dB down. The same banded scale as the Horn Design Tool, so the two can be read side by side.
Axes are always the full 20 Hz–20 kHz and ±180°, so what you calculated appears as a band within the whole space and two runs can be compared directly. Anything outside the calculated range is left blank — never extrapolated.
The pale wash underneath is the Horn Design Tool's own fast prediction, carried across with the geometry. It covers the whole chart, so you can see what the quick model expects everywhere you have not solved yet — and, where you have, exactly how far the simulation departs from it. Solved results are drawn at full strength on top; the wash is never simulated data.
Below roughly ka < 0.5 — a mouth small compared with a wavelength — a horn is omnidirectional to within 0.3 dB whatever its shape. Those rows are filled in from that exact analytic result rather than simulated: the mic arc is inside the near field down there, so simulating them would only add noise. For a 240 mm mouth that crossover is about 227 Hz.
Where ranges overlap, the finer run wins, not the most recent — so a quick Draft sweep will not overwrite a region you have already solved at Fine.
Runs accumulate. Solve one range, then another, and both stay on the map, so you can build up coverage a piece at a time instead of paying for the whole span at once. They are kept against the horn's geometry and survive leaving the page, so returning from the Horn Design Tool does not lose them. Changing the horn, or pressing Clear accumulated results, starts over.
Method The level at one microphone, swept by angle. The map above shows level relative to on-axis, which divides the horn's own response straight out of it — a dip present at every angle is invisible there, because it is not a difference between angles. This is the response itself.
The quantity is the horn's gain over a free-space monopole of the same volume velocity: flat where the horn loads properly, falling away below cutoff. It is not calibrated SPL, so read differences rather than absolute numbers.
Drag the slider, or type an angle, to walk the microphone from on-axis out to whatever angle the run measured; on-axis stays on screen in grey as a reference, since off-axis level only means anything against the level it fell from. The vertical scale is fixed across every angle, so the curve moves and the axis does not.
Typed values snap to the nearest angle — or, on the polar plot, the nearest frequency — that was actually solved. A value between two computed bins was never calculated.
Method A binary STL is voxelised straight onto the solver's grid, so the simulated shape is the file rather than a fitted approximation of it.
Either a solid body or a surface with no thickness — whichever your modeller produced; which one it is gets worked out from whether the mesh is closed. It must be open at the throat and mouth and closed everywhere else. A hole anywhere else is refused rather than warned about: sound escapes through it, and the run does not fail, it comes back looking normal and wrong.
The horn only — no cabinet, no driver body, no mounting hardware, since all of it would be simulated as part of the horn. Axis, throat end, a quarter turn about the axis and the file's units can all be corrected on the upload page before handing it over.
The horn is simulated free-standing, with the driver's rear sealed as a real compression driver's is — there is no cabinet, baffle or room.
That matters most behind the horn. With nothing back there, sound diffracting around the mouth rim travels the same distance to every point on the rear axis and arrives in phase, producing a real lobe pointing directly backwards at 180°. It is genuine physics for a bare, perfectly symmetrical horn — but a real one has a magnet, a frame and usually a cabinet, any of which breaks that symmetry and largely removes it. Treat anything past about 120° as indicative.
On-axis frequency response is deliberately not shown; use the Horn Design Tool for that. This is an experimental research tool — validate any final design against a real measurement.