Spiral bevel gears are the default choice when a drive must turn a corner under high torque. Their curved teeth engage gradually, so several tooth pairs share the load and motion is smoother than with straight bevel teeth. The same curvature makes the tooth surface difficult to define: the flank is a three-dimensional conjugate surface, the root fillet is produced by a cutter with a finite corner radius, and the tip and root boundaries must close the tooth without intruding into the contact path. In automotive axles, marine transmissions, aerospace actuators and heavy industrial reducers, small errors in that surface appear as a shifted contact pattern, higher contact stress, a noisier transmission-error curve or a shorter bending-fatigue life.
Traditional tooth-surface design follows the generating process. The cutter profile and machine-tool settings define a family of swept surfaces, and the envelope gives the flank. This is accurate when the machine is correctly set, but the result is sensitive to cutter diameter, pressure angle, mounting distance and blank dimensions. The root transition is often approximated and later corrected by surface fitting. The spherical involute route is different: it defines the conjugate curve directly on a sphere centred at the pitch apex, so the involute property is built into the geometry rather than recovered from a simulated cutting process.
A 2024 design study from Shenyang University of Technology combined the spherical involute with a planar logarithmic spiral used as a conjugate directing curve. The aim was not to replace the generating concept but to make the working flank, tip and root boundaries, and root fillet come from one consistent mathematical model. In this context, the term planar conjugate logarithm refers to a logarithmic spiral on the tangent plane of the base cone; it is a geometric directing curve, not a logarithmic scaling of gear coordinates.
Geometric basis: rolling a logarithmic spiral on a base cone
Let the base cone have half-angle \(\delta_b\) and apex \(O\). A plane tangent to this cone rolls without slipping. A point \(P\) on the plane, at a fixed distance from \(O\), traces a spherical involute on the sphere of radius \(OP\). Before rolling, the path of \(P\) in the tangent plane is a logarithmic spiral. In the tangent-plane frame its coordinates can be written as
$$
\mathbf r_{\log} =
\begin{bmatrix}
0\\
e^{\theta\cot\beta}\sin(\phi\sin\delta_b)\\
e^{\theta\cot\beta}\cos(\phi\sin\delta_b)\\
1
\end{bmatrix}
$$
where \(\beta\) is the spiral angle, \(\theta\) is the roll parameter, \(\phi\) is the angular position in the tangent plane and \(\delta_b\) is the base-cone half-angle. The exponential term produces the logarithmic-spiral growth, while the sine and cosine terms place the point around the cone axis. A rolling transformation then maps this planar curve to the sphere. Because the spiral is self-similar, the local normal and tangent directions change smoothly along the contact path. That smoothness is the property needed for a conjugate tooth surface; an arbitrary curve would introduce avoidable curvature discontinuities.
Conjugate contact condition
Let \(S_1\) and \(S_2\) be frames attached to the pinion and gear, with a common apex and shaft angle \(\Sigma\). The rotation angles are \(\phi_1\) and \(\phi_2\), and the ratio is \(i_{21}=\phi_2/\phi_1\). For a contact trace \(\mathbf r_{\Gamma1}\) on the pinion, the corresponding trace on the gear is obtained from the coordinate transformation \(M_{21}\):
$$
\mathbf r_{\Gamma2}=M_{21}\mathbf r_{\Gamma1}
$$
The relative velocity at a contact point is
$$
\mathbf v_{12}=\boldsymbol\omega_1\times\mathbf r_{\Gamma1}
-\boldsymbol\omega_2\times\mathbf r_{\Gamma2}
$$
and the two surfaces remain conjugate only if the relative velocity has no component along the common normal:
$$
\mathbf n_{\Gamma1}\cdot\mathbf v_{12}=0
$$
These equations are solved together with the planar logarithmic trace. The result is a contact path that lies on the sphere and a corresponding path on the gear. The normal-section profile is then attached to this path. A circular arc is used instead of a straight involute generator. In local coordinates the arc has the form
$$
y_{\Gamma c1}=\rho_i(\cos\alpha_i+1),\qquad
z_{\Gamma c1}=\rho_i\sin\alpha_i
$$
where \(\rho_i\) is the arc radius, \(\alpha_i\) is the pressure angle of the normal section and \(i\) labels the convex or concave side. Two conjugate curve pairs, one above and one below the pitch cone, generate the two flanks. The arc radius controls the local curvature, while the conjugate condition controls how that curvature is distributed along the tooth.
Building the closed tooth
The working flank is generated by sweeping the normal-section arc along the conjugate contact trace. The profile orientation changes continuously because the local normal rotates with the spherical involute. A point grid is produced directly from the equations, and the flank is lofted through those points. This avoids recovering the surface from a separate cutter simulation and then fitting a spline to the result.
The outer boundary is set at the outer cone distance \(R\). The tip curve uses the addendum angle \(\delta_{at}\), and the root curve uses the dedendum angle \(\delta_{ft}\):
$$
\rho=R,\quad \theta=\delta_{at}\ \text{(tip)},\qquad
\rho=R,\quad \theta=\delta_{ft}\ \text{(root)}
$$
At the small end, the radial coordinate is reduced by the face width \(B\), and the angular coordinate is shifted by the small-end offset \(\Delta\phi\):
$$
\rho=R-B,\qquad \phi’=\phi\pm\Delta\phi
$$
These boundaries are not cosmetic. They define the usable contact area and the transition from the working flank to the root. The root fillet is generated separately because it is not a conjugate contact surface. Its radius depends on the cutter corner radius, pressure angle \(\alpha\), addendum \(h_a\), pitch radius \(r_p\) and clearance. The transition angle is
$$
\Delta\alpha=
\arccos\left(\frac{r_p\cos\alpha}{r_p+h_a}\right)-\alpha
$$
When the base-cone radius is larger than the root circle, the fillet radius follows from \(\Delta\alpha\) and the cutter corner geometry. If side or top clearance is present, effective pitch radius and addendum replace \(r_p\) and \(h_a\) in the same calculation. When the base-cone radius is smaller than the root circle, the model uses a constant fillet:
$$
r_t=0.3m
$$
where \(m\) is the module. This closes the root surface without adding a separate fitted spline.
Representative geometry and mesh
The evaluation used a 27-tooth left-hand pinion and a 74-tooth right-hand gear, giving a ratio of \(74/27\approx2.74\). The main geometry is listed in Table 1.
| Parameter | Pinion | Gear |
|---|---|---|
| Number of teeth | 27 | 74 |
| Hand | Left | Right |
| Module | 3.85 mm | 3.85 mm |
| Shaft angle | 90° | 90° |
| Spiral angle | 30° | 30° |
| Face width | 38.5 mm | 38.5 mm |
| Pressure angle | 20° | 20° |
| Dedendum (root height) | 4.56 mm | 1.98 mm |
The tooth surfaces were assembled from point grids: curves were lofted into flanks, the tip and root surfaces were closed, and a single tooth was patterned around the blank. The mesh was refined on the working flanks and root fillets and coarsened on the blank. This is the right balance for contact analysis, because the contact ellipse and the fillet carry the stress gradient, while the blank mainly provides stiffness.
Contact stress and shaft-angle error
The finite-element contact model used a 20CrMnTi-type gear steel with density \(7.8\times10^{-9}\ \mathrm{t/mm^3}\), Young’s modulus \(E=205\ \mathrm{GPa}\) and Poisson’s ratio \(\nu=0.3\). The pinion was the driver, and the gear carried a resisting torque. A generating-method model served as the baseline under the same load.
In that comparison, the spherical-involute model with the planar logarithmic directing curve produced a lower peak contact stress on the pinion concave flank and a steadier stress history as the mesh rolled through engagement. The maximum appeared on the working flank, while the explicit root fillet controlled the stress gradient at the root transition. The result is consistent with the design intent: when the two flanks remain conjugate over a larger part of the contact path, the load is distributed more evenly instead of concentrating at the edges of an approximated surface. The comparison is still a numerical contact result, not an absolute material limit; it depends on the applied torque, mesh density and boundary conditions.
For the misalignment case, a small shaft-angle error was introduced after assembly. The pinion was rotated through \(1.3\ \mathrm{rad}\) in \(1\ \mathrm{s}\) with a friction coefficient of \(0.15\). The transmission-error curve remained stable in trend, without a sharp rise or a discontinuous jump. This is the axis-angle compensation property of the spherical involute: a small change in shaft angle shifts the contact path, but the involute geometry can still maintain conjugate motion. The result is numerical and applies to this gear pair and load case; it does not mean that arbitrary misalignment is harmless.
What the approach changes
The practical value is not a new gear ratio or a new material. It is a way to keep the geometric chain consistent from the base cone to the root fillet. The working flank comes from a conjugate spherical curve; the normal-section arc sets the local curvature; the tip and root curves close the tooth; the root fillet follows the cutter geometry. Because the surface is generated as a point grid directly from these equations, it can be exported to CAD and meshed without a separate surface-fitting step. That matters for contact and dynamic simulations, where a small flank error can shift the contact ellipse and change the predicted stress.
The method is relevant to automotive drive axles, marine propulsion, aerospace actuators and industrial reducers, especially where high torque and quiet operation are required at the same time. It is also relevant to small-batch or custom gear sets, where repeated generating trials are expensive. The model still needs accurate base-cone and cutter parameters, and the root fillet calculation assumes a defined cutter corner and clearance. Production use would require cutting or grinding trials and measured transmission error under load, speed, thermal and lubrication variations. The shaft-angle compensation is a small-error property, not a substitute for correct assembly.
Spherical involute geometry and planar conjugate logarithmic directing curves therefore offer a geometry-first route to spiral bevel gear tooth surfaces. The approach ties the working flank, tooth boundaries and root transition to the same kinematic foundation, and it gives contact and transmission-error simulations a surface that is defined by the design equations rather than by a fitted approximation.