Spiral bevel gears are used in rotorcraft transmissions, turboprop drives, and other high-power systems where intersecting shafts must be connected with high torque density and smooth motion. Their flanks are generated as conjugate surfaces and then deliberately modified so that contact is localized rather than spread over the full tooth. The design problem is therefore not only to place a contact pattern at a reference point, but to control how the contact path and transmission error evolve across the whole flank. Classical second-order contact analysis fixes principal curvatures, principal directions, contact ellipse orientation, and the first derivative of the transmission ratio at one point. Those quantities are local. They do not determine how the curvature of the contact trace or the transmission ratio changes along the path. A higher-order description adds the short-range curvature of the contact trace and the second derivative of the transmission ratio. Ease-off topography provides a practical route to both quantities because it represents the normal mismatch field from which they can be differentiated.
Why second-order contact parameters are not sufficient
Second-order design practice is associated with Gleason’s local conjugate method and Litvin’s local synthesis. At a selected reference point, usually near the midpoint of the pitch cone, the analysis determines the principal curvatures and directions of the two flanks, the contact ellipse, the direction of the contact trace, and the first-order variation of the transmission ratio. These parameters are necessary for a well-localized contact pattern, but they are insufficient for global flank control. A surface can satisfy the prescribed second-order data at the reference point and still produce nonuniform curvature along the contact trace and nonconstant high-order transmission error elsewhere.
The high-order contact parameters used in the present formulation are the second derivative of the transmission ratio, \(d^2\phi_2/d\phi_1^2\), and the short-range curvature of the contact trace, \(\kappa_g\). The first measures the rate at which the transmission ratio changes within a small motion cycle and is therefore related to motion irregularity. The second measures how the contact trace bends in the common tangent plane. Both are accumulated effects of the flank topography along the contact path, so they provide global rather than pointwise information. The computational difficulty of a direct third-order contact theory is avoided by extracting these quantities from the ease-off surface.
Ease-off construction for a generated spiral bevel gear pair
Let \(S_1\) and \(S_2\) be frames fixed to the pinion and wheel, and let \(S_h\) be an auxiliary fixed frame. The generated wheel surface and its unit normal are \(\mathbf r_2(u_2,\theta_2)\) and \(\mathbf n_2(u_2,\theta_2)\). After transformation to \(S_h\), the meshing condition is
\mathbf r_h(u_2,\theta_2,\phi_2)=M_{hf}M_{f2}\mathbf r_2(u_2,\theta_2),\qquad
\mathbf n_h(u_2,\theta_2,\phi_2)=L_{hf}L_{f2}\mathbf n_2(u_2,\theta_2),
$$
f(u_2,\theta_2,\phi_2)=\mathbf n_h\cdot \mathbf v_h^{(21)}=0,
$$
where \(\mathbf v_h^{(21)}\) is the relative velocity between the two members. Solving the meshing equation for \(u_2\) and transforming the result back to \(S_1\) gives the pinion surface conjugate to the wheel:
\mathbf r_{1c}=M_{1h}\mathbf r_h,\qquad
\mathbf n_{1c}=L_{1h}\mathbf n_h.
$$
The pinion design surface \(\mathbf r_1(\theta_1,\phi_1)\) is obtained from the cutting model. Before the two surfaces can be compared, they must be angularly aligned so that they touch at the reference point \(C\). Rotation about the pinion axis gives the alignment conditions
y_1^{(C)}\cos\lambda_1+z_1^{(C)}\sin\lambda_1=0,\qquad
y_{1c}^{(C)}\cos\lambda_{1c}+z_{1c}^{(C)}\sin\lambda_{1c}=0.
$$
After alignment, the ease-off is the normal separation between the design surface and the conjugate surface:
\Delta\delta(\theta_1,\phi_1)=
\left[\tilde{\mathbf r}_1(\theta_1,\phi_1)-\tilde{\mathbf r}_{1c}(\theta_2,\phi_2)\right]\cdot
\tilde{\mathbf n}_1(\theta_1,\phi_1).
$$
The flank is projected onto a plane using \(X_m=x_1\) and \(Y_m=(y_1^2+z_1^2)^{1/2}\), then rotated by the pitch angle to coordinates \(X\) along the face width and \(Y\) along the tooth height. In these coordinates, a fourth-order ease-off surface is written as
\Delta\delta=F(X,Y)=a_{11}X^2Y^2+a_{12}X^3Y+a_{13}XY^3+a_{14}X^4+a_{15}Y^4.
$$
The coefficients have distinct roles. The term \(a_{11}\) acts together with \(a_{14}\) and \(a_{15}\) to produce coupled modification, while \(a_{12}\) and \(a_{13}\) represent high-order torsional effects. The coefficients \(a_{14}\) and \(a_{15}\) are fourth-order crowning terms along the face width and tooth height, respectively. Evaluating \(F(X,Y)\) on the projected grid produces the discrete ease-off matrix used in the subsequent finite-difference calculation.
High-order parameters from the ease-off field
For a normal mismatch \(F\), the additional wheel rotation is proportional to the mismatch and inversely proportional to the gear ratio. With \(N_1\) and \(N_2\) the pinion and wheel tooth numbers, and \(K=N_2/N_1\), the transmission error is
\Delta\phi_2=\frac{N_1}{N_2}F(X,Y)=\frac{F(X,Y)}{K}.
$$
Along the contact path, \(X\) and \(Y\) are functions of the pinion rotation \(\phi_1\). Differentiating twice gives the high-order transmission-ratio parameter:
\frac{d^2\phi_2}{d\phi_1^2}
=\frac{1}{K}\frac{d^2F}{d\phi_1^2},
$$
\frac{d^2F}{d\phi_1^2}
=F_{XX}\dot X^2+2F_{XY}\dot X\dot Y+F_{YY}\dot Y^2
+F_X\ddot X+F_Y\ddot Y,
$$
where a dot denotes differentiation with respect to \(\phi_1\). The second high-order parameter is obtained from the projected contact trace \(R_e=[X(\phi_1),Y(\phi_1),0]\). Its short-range curvature in the tangent plane is
\kappa_g=
\frac{\left|X’Y”-Y’X”\right|}
{\left(X’^2+Y’^2\right)^{3/2}},
$$
with primes denoting \(d/d\phi_1\). At the reference point, the design surface, the conjugate surface, and the ease-off surface share a common tangent plane. The curvature of the projected ease-off trace therefore equals the absolute short-range curvature of the corresponding contact traces on the two tooth surfaces.
Finite-difference evaluation on the discrete grid
The discrete tooth-surface points are mapped to a grid, and the derivatives of \(F\) with respect to \(X\) and \(Y\) are evaluated analytically from the quartic. The derivatives with respect to \(\phi_1\) are evaluated by central differences along the grid direction that follows the contact path. With step \(h=2\pi/(NN_1)\), where \(N\) is the number of contact points, the first and second derivatives of the projected coordinates are
\left(\frac{dX}{d\phi_1}\right)_{ij}
=\frac{X_{i+1,j+1}-X_{i-1,j-1}}{2h},\qquad
\left(\frac{d^2X}{d\phi_1^2}\right)_{ij}
=\frac{X_{i+1,j+1}-2X_{ij}+X_{i-1,j-1}}{h^2},
$$
and similarly for \(Y\). Let \(A=F_{XX}\), \(B=F_{YY}\), \(C_m=F_{XY}\), \(D=F_X\), and \(E=F_Y\). From the quartic ease-off surface,
\begin{aligned}
A &= 2a_{11}Y^2+6a_{12}XY+12a_{14}X^2,\\
B &= 2a_{11}X^2+6a_{13}XY+12a_{15}Y^2,\\
C_m &= 4a_{11}XY+3a_{12}X^2+3a_{13}Y^2,\\
D &= 2a_{11}XY^2+3a_{12}X^2Y+a_{13}Y^3+4a_{14}X^3,\\
E &= 2a_{11}X^2Y+a_{12}X^3+3a_{13}XY^2+4a_{15}Y^3.
\end{aligned}
$$
The discretized transmission-ratio second derivative is then
\left(\frac{d^2\phi_2}{d\phi_1^2}\right)_{ij}
=\frac{1}{K}\left[
A_{ij}\dot X_{ij}^2
+2C_{m,ij}\dot X_{ij}\dot Y_{ij}
+B_{ij}\dot Y_{ij}^2
+D_{ij}\ddot X_{ij}
+E_{ij}\ddot Y_{ij}
\right].
$$
Substituting the same finite-difference expressions into the curvature formula gives the short-range curvature at each contact point. The calculation therefore reduces high-order contact analysis to post-processing of a discrete ease-off grid: fit the quartic coefficients, differentiate analytically with respect to \(X\) and \(Y\), and differentiate numerically along the contact path.
Example: a 34/43-tooth helicopter transmission pair
The formulation was evaluated on a spiral bevel gear pair intended for helicopter transmission service. The basic geometry is listed in the accompanying table.
| Parameter | Pinion | Wheel |
|---|---|---|
| Number of teeth | 34 | 43 |
| Large-end module (mm) | 2.0 | 2.0 |
| Face width (mm) | 15 | 15 |
| Pressure angle (°) | 20 | 20 |
| Spiral angle (°) | 20 | 20 |
| Shaft angle (°) | 108 | |
| Pitch cone angle (°) | 44.8608 | 63.1392 |
| Face cone angle (°) | 49.4637 | 66.0355 |
| Root cone angle (°) | 41.9645 | 58.5362 |
| Hand | Right | Left |
The pinion working-surface cutting parameters are given below. These define the generating motion and cutter position used to produce the design flank before ease-off modification.
| Parameter | Value |
|---|---|
| Cutter tip radius \(R_c\) (mm) | 44.4713 |
| Angular cutter position \(q_0\) (rad) | 1.0827 |
| Radial cutter position \(S_k\) (mm) | 51.5697 |
| Roll ratio \(R_a\) | 1.4391 |
| Vertical offset \(E\) (mm) | −3.2839 |
| Axial offset \(X_p\) (mm) | −0.2389 |
| Bed position \(X_b\) (mm) | 0.1598 |
| Mounting angle \(\delta_m\) (rad) | 0.7324 |
The initial flank was a second-order design with a maximum normal modification of 11.21 μm. Its transmission-error curve looked approximately quadratic, as expected from a second-order surface. The computed second derivative of the transmission ratio, however, was not constant: it varied by 0.0009 along the contact sequence. The contact trace was also not a uniform circular arc; its short-range curvature varied by 0.0002. These fluctuations are small in absolute terms, but they show that a visually quadratic transmission-error curve does not guarantee a second-order contact process. The contact path and the transmission ratio retain higher-order variation that the reference-point parameters do not capture.
Modified topographies and comparative behavior
Three modified ease-off surfaces were specified by the quartic coefficients in the table below. All three produced a maximum transmission-ratio second derivative of zero in the central contact region, where the flanks are fully engaged. The nonzero values appeared as fluctuations concentrated toward the tooth edges, consistent with the stronger edge relief in the modified topographies. The fluctuation of the transmission-ratio second derivative was 0.0031, 0.0019, and 0.0010 for the three cases, all larger than the 0.0009 fluctuation of the initial surface.
| Surface | \(a_{11}\) | \(a_{12}\) | \(a_{13}\) | \(a_{14}\) | \(a_{15}\) | Ratio second-derivative fluctuation | Contact-trace rotation (°) | Short-range curvature fluctuation |
|---|---|---|---|---|---|---|---|---|
| Initial | — | — | — | — | — | 0.0009 | 0 (reference) | 0.0002 |
| Modified 1 | 0 | 0.000005 | 0 | 0.0000012 | 0.0002063 | 0.0031 | 2.40 | 0.0000769 |
| Modified 2 | 0 | −0.000005 | 0 | 0.0000014 | 0.0002063 | 0.0019 | 5.63 | 0.000586 |
| Modified 3 | 0.00001 | 0 | 0.000006 | 0.0000014 | 0.0002063 | 0.0010 | 23.76 | 0.000127 |
The contact traces of the three modified surfaces were rotated by 2.40°, 5.63°, and 23.76° relative to the initial trace. Their short-range curvature fluctuations were \(7.69\times10^{-5}\), \(5.86\times10^{-4}\), and \(1.27\times10^{-4}\). These values are numerically small, but they are not zero, and they quantify how the contact trace bends as it crosses the modified flank. The case with the largest trace rotation, modified 3, did not produce the largest curvature fluctuation; modified 2 did. The relationship between trace orientation and high-order curvature is therefore not monotonic, which is a reason to evaluate both metrics rather than rely on a single contact-pattern descriptor.
What the metrics add to flank design
The ease-off formulation converts high-order contact analysis into a sequence of well-defined numerical operations. The quartic surface supplies analytic first and second partial derivatives with respect to the flank coordinates. The contact path supplies the total derivatives with respect to pinion rotation. The result is a pair of global metrics that can be evaluated on the same grid used for tooth contact analysis. This is considerably simpler than deriving and solving the full third-order contact equations for every design iteration.
The example also shows why the distinction matters. A second-order design can produce a transmission-error curve that appears smooth and quadratic, while the second derivative of the transmission ratio and the short-range curvature of the contact trace still vary along the path. Those variations are relevant to motion smoothness, sensitivity to misalignment, and the comparison of alternative flank modifications. Because the metrics are obtained from the ease-off coefficients, they can be used directly in flank optimization: change the coefficients, recompute the contact path, and compare the resulting high-order parameters before a more expensive loaded contact or stress analysis is performed.
The scope of the calculation should be kept in mind. The quantities are geometric and kinematic; they describe unloaded contact and do not by themselves predict contact stress, load distribution, or dynamic response. The finite-difference estimates depend on grid resolution and on the smoothness of the fitted ease-off surface. Measured flank data would require filtering before the derivatives are evaluated. The numerical values reported here are also specific to the 34/43-tooth helicopter example and should not be treated as universal design targets.
Conclusion
High-order contact parameters of spiral bevel gears can be extracted from a discrete ease-off surface without solving a full third-order contact theory. The quartic ease-off representation provides analytic derivatives with respect to the face-width and tooth-height coordinates, while central differences along the contact path provide the total derivatives with respect to pinion rotation. The two resulting metrics are the second derivative of the transmission ratio and the short-range curvature of the contact trace. In the 34/43-tooth example, the initial surface showed fluctuations of 0.0009 and 0.0002, while three modified surfaces produced transmission-ratio fluctuations of 0.0031, 0.0019, and 0.0010 and curvature fluctuations of \(7.69\times10^{-5}\), \(5.86\times10^{-4}\), and \(1.27\times10^{-4}\), with contact-trace rotations of 2.40°, 5.63°, and 23.76°. The approach provides a practical post-processing route for global flank design and for comparing modifications before loaded contact analysis.