Why the contact path direction governs the design limit
Spiral bevel gears transmit motion between intersecting shafts with gradual, overlapping tooth engagement, which is why they dominate the input stage of helicopter drivetrains and other high-speed, high-load reductions where smoothness, low noise and load capacity must be satisfied at the same time. The parameter that most strongly controls how well a pair performs is the contact ratio, and in a bevel pair this quantity is not fixed by tooth height alone. It is set primarily by the direction of the contact path across the flank relative to the root cone, because that direction determines how much pinion rotation elapses between the first and the last instant of contact of a single tooth pair.
Early spiral bevel geometry placed the contact path roughly perpendicular to the root cone, in the manner of a spur gear. The resulting contact ratio is small: few teeth share the load at any instant, bending stress in the tooth root is high, and the mesh generates strong vibration excitation. Three routes have been used to raise the contact ratio. The first is to increase tooth height, but practical limits on blank size and the risk of a sharpened tooth tip cap the achievable gain, even when the cutting parameters are designed through local synthesis to avoid edge contact. The second is to exploit non-zero profile shifts, typically negative, chosen jointly with the cutting parameters so that the geometric and the manufacturing contact ratios both rise and the pinion root becomes stronger. The third, and the most effective for bevel gears, is to reduce the angle between the contact path and the root cone, producing an internal diagonal contact pattern. Measured and simulated results for such designs show a clear reduction in contact stress and in mesh noise, and curvature-corrected variants improve load distribution further.
All three routes share one penalty. A path that sweeps across the tooth height carries the contact point through the regions of highest sliding velocity, near the tip and the root, and the sliding direction reverses as the point crosses the pitch line. High sliding speed combined with reversal is the classical mechanism behind scuffing and scoring, and it limits how far the contact ratio can be pushed on a heavily loaded pair. The alternative is to orient the path along the tooth length, in the manner of a helical gear whose contact line runs near the pitch line. Along such a path the instantaneous contact point moves with almost no relative sliding, friction approaches zero, and the load is carried by a long, narrow band that spreads across the face. Historically this geometry was avoided because the resulting pattern is sensitive to assembly error; with the manufacturing and mounting accuracy now achievable, a tooth-length path becomes the natural target for contact pattern design.
Prescribing a symmetric parabolic transmission error
The design starts from a transmission error curve that is deliberately shaped rather than left to chance. Transmission error is defined as the difference between the actual and the ideal wheel rotation angle when the pinion turns through a given angle. A symmetric parabola is prescribed, anchored at the entry point A, the exit point B and a reference point M, and bounded by the mesh transition value $\delta_{TE}$ at the ends of the mesh cycle $T_z = 2\pi/z_1$, where $z_1$ is the pinion tooth count. The curve has three properties: the transmission error is equal at entry and exit, the entry and exit rotation angles are symmetric about the reference point, and the entry branch AM and exit branch MB are mirror images.
$$\delta\varphi_2=\tfrac{1}{2}\,m’_{21}\left(\varphi_1-\varphi_1^{(0)}\right)^{2}$$
where $m’_{21}$ is the first derivative of the reciprocal transmission ratio, $\varphi_1$ the pinion rotation angle and $\varphi_1^{(0)}$ its initial value. The mesh transition value and this derivative are linked by the mesh period:
$$m’_{21}=-\frac{8\,\delta_{TE}}{T_z^{2}},\qquad T_z=\frac{2\pi}{z_1}$$
Substituting the period gives the transmission error in closed form and, through its definition, the relation between pinion and wheel rotation angles:
$$\delta\varphi_2=-\frac{\delta_{TE}\,z_1^{2}}{\pi^{2}}\left(\varphi_1-\varphi_1^{(0)}\right)^{2}$$
$$\varphi_2=\varphi_2^{(0)}+m_{21}\left(\varphi_1-\varphi_1^{(0)}\right)-\frac{\delta_{TE}\,z_1^{2}}{\pi^{2}}\left(\varphi_1-\varphi_1^{(0)}\right)^{2}$$
with $m_{21}=z_1/z_2$ the reciprocal transmission ratio and $z_2$ the wheel tooth count. The parabola therefore fixes both the amplitude and the symmetry of the kinematic error before any flank geometry is generated.
Generating the pinion auxiliary surface from the wheel
The wheel is treated as an imaginary shaping cutter. Its flank $\Sigma_2$, described by position vector $\mathbf{r}_2(\theta_g,\phi_g)$ and unit normal $\mathbf{n}_2(\theta_g,\phi_g)$ in the wheel coordinate frame $S_2$, is known from the cutter geometry and the machine settings, where $\theta_g$ is the cutter rotation angle and $\phi_g$ the cradle angle. Rolling the imaginary cutter against the pinion blank with the ratio prescribed by the transmission error function produces a pinion auxiliary surface $\Sigma’_1$ that is in line contact with the wheel flank by construction. With $S_h$ as the fixed frame and $S_1$ attached to the pinion, the surface is obtained by homogeneous coordinate transformation together with the spatial meshing condition:
$$\mathbf{r}’_1=M_{1h}(\varphi_1)\,M_{h2}(\varphi_2)\,\mathbf{r}_2,\qquad \mathbf{n}’_1=L_{1h}(\varphi_1)\,L_{h2}(\varphi_2)\,\mathbf{n}_2,\qquad f_{12}(\theta_g,\phi_g,\varphi_1)=0$$
Here $M_{h2}$ contains the shaft angle $\Sigma$ and the wheel rotation angle, $M_{1h}$ contains the pinion rotation angle, and $L_{1h}$, $L_{h2}$ are the corresponding rotation matrices obtained by deleting the last row and column of the transformation matrices. Because the parabola of the previous section enters through $\varphi_2(\varphi_1)$, the auxiliary surface already carries the intended kinematic error; what it does not yet carry is point contact.
Placing the contact path along the tooth length
On the auxiliary surface, a straight contact path $\Gamma$ is prescribed, parallel to the pitch cone line but displaced from it by $\Delta y$ in the local frame $xoy$. The entry point A lies on the inner cone at radius $R_i$ and the exit point B on the outer cone at radius $R_e$; both positions are expressed in terms of the pinion pitch cone angle $\delta_1$. For each of the two points, a nonlinear system is solved for the pinion rotation angle that brings the corresponding meshing line through that position while satisfying the meshing equation and the surface definition:
$$x’_1(\theta_g,\phi_g,\varphi_1)=R_{i}\cos\delta_1+\Delta y\sin\delta_1,\qquad \sqrt{y’^{\,2}_1+z’^{\,2}_1}=R_{i}\sin\delta_1+\Delta y\cos\delta_1,\qquad f_{12}=0$$
with an analogous system written at $R_e$ for $\varphi_{1B}$. Neglecting edge contact, the design contact ratio follows directly as $\varepsilon_r=\left|\varphi_{1B}-\varphi_{1A}\right|/T_z$. Because the path runs from the inner to the outer cone, this ratio is governed by the face width and is essentially independent of tooth height, which is precisely the property that makes the tooth-length layout attractive: contact ratio becomes a face-width design decision rather than a compromise with tip sharpening or root clearance.
The path is described by $y=\tan(\delta_1+\Delta)\,x+\Delta y/\cos\delta_1$, where $\Delta$ is the angle between the path and the pinion pitch cone angle. Setting the path on the mid-line of the face width corresponds to a specific offset $\Delta y$. When the mean spiral angle is small, a small positive $\Delta$, typically between 0 and the pinion addendum and dedendum angles, tilts the path into an internal diagonal and recovers part of the contact ratio; for $\Delta \neq 0$ the entry and exit positions above must be adjusted accordingly. Keeping the path straight also reduces the sensitivity of the pattern to mounting error, so that acceptable meshing quality is preserved over a range of realistic misalignments.
Modification applied on the tangent plane
Line contact must be converted into controlled point contact. Conventional practice modifies the auxiliary surface on its rotational projection plane, which introduces approximation. A more direct route applies the modification in the tangent plane at each contact point. For an instantaneous pinion rotation angle $\varphi_{1i}$, two points $P_1$ and $P_2$ are taken on the meshing line $L_i$ of the auxiliary surface, and $P$ is the intersection of $L_i$ with the contact path $\Gamma$. Projecting $P_1$ and $P_2$ onto the tangent plane $T_i$ through $P$ gives $P’_1$ and $P’_2$; the distance $d$ between $P’_1$ and $P$ is used as the modification length, and the modified points $P^{*}_1$, $P^{*}_2$ are displaced along the normal at $P$ by
$$\zeta_i=\frac{e}{a^{2}}\,d^{2}$$
where $e$ is the elastic deformation between the contacting flanks, taken at the light-load value of 0.00635 mm established empirically by Gleason Works, and $a$ is the prescribed semi-major axis of the contact ellipse. The quadratic form reproduces a parabolic profile across the contact line while the depth is set by the elastic approach, so the unloaded contact ellipse matches the intended size under light load.
The auxiliary surface is discretized into $k=m\times n$ grid points, with $m=5$ across the tooth height and $n=9$ along the tooth length, giving 45 points. Each point carries a position vector and unit normal, and the target pinion surface $\Sigma^{*}_1$ is obtained by superposing the modification on the auxiliary surface:
$$\mathbf{p}^{*}_i=\mathbf{p}’_i+\zeta_i\,\mathbf{n}’_i,\qquad i=1,2,\dots,k$$
Back-solving the machine settings
What remains is to find the cutting settings that actually produce the target surface on a conventional cradle-type hypoid generator. Given a set of pinion cutter parameters and machine settings, the generated pinion flank $\Sigma_1$ has position $\mathbf{r}_1(\theta_p,\phi_p,\phi_1)$ and unit normal $\mathbf{n}_1(\theta_p,\phi_p,\phi_1)$, where $\theta_p$ is the cutter rotation angle and $\phi_p$ the cradle angle. Writing $\mathbf{p}_i$ and $\mathbf{n}_i$ for the corresponding discrete quantities, the normal deviation between generated and target surface at point $i$ is
$$h_i(\mathbf{d})=\left(\mathbf{p}^{*}_i-\mathbf{p}_i\right)\cdot\mathbf{n}_i,\qquad i=1,\dots,k$$
with the design vector $\mathbf{d}=\left[\alpha_1,R_c,S_{r1},q_1,C_{r1},E_{m1},X_{G1},X_{B1},\gamma_m,2C,6D\right]$ collecting the profile angle, cutter tip radius, radial and angular cutter settings, roll ratio, vertical and horizontal offsets, axial base setting, machine mounting angle, and the second- and third-order modified-roll coefficients. The settings are then found by minimizing the sum of squared normal deviations,
$$\min f(\mathbf{d})=\mathbf{h}^{T}(\mathbf{d})\,\mathbf{h}(\mathbf{d}),\qquad \mathbf{d}\in\left[\chi_1,\chi_2\right]$$
subject to the limits $\chi_1$, $\chi_2$ imposed by the working range of the generator. The problem is a highly nonlinear least-squares one with many variables and multiple local minima, and each evaluation requires solving a system of nonlinear meshing equations, so a genetic algorithm is used: population 5, 100 generations, crossover probability 0.5 and mutation probability 0.1, settings obtained by repeated trials on this problem size. Of the eleven entries in $\mathbf{d}$, the profile angle $\alpha_1$ and the blank mounting angle $\gamma_m$ are held fixed, leaving nine optimization variables acting on 45 surface points.
Worked example
The procedure is demonstrated on the working flanks of a 23-tooth pinion meshing with a 65-tooth wheel at a 90° shaft angle, with a module of 3.9 mm, a normal pressure angle of 25°, a mean spiral angle of 35°, a face width of 37 mm and an outer cone distance of 134.451 mm. The mesh transition value is set to $\delta_{TE}=0.00278°$. Table 1 lists the blank geometry and Table 2 the cutting data for the two working flanks; both were derived from a conventional Gleason summary, which gives a design contact ratio of 2.3159 before optimization and 2.8242 after the flank is reshaped by the procedure above.
| Parameter | Pinion | Wheel |
|---|---|---|
| Number of teeth | 23 | 65 |
| Module (mm) | 3.9 | 3.9 |
| Normal pressure angle (°) | 25 | 25 |
| Mean spiral angle (°) | 35 | 35 |
| Hand of spiral | right | left |
| Shaft angle (°) | 90.0 | 90.0 |
| Pitch cone angle (°) | 19.4861 | 70.5138 |
| Face cone angle (°) | 21.7770 | 71.6718 |
| Root cone angle (°) | 18.3281 | 68.2229 |
| Outer cone distance (mm) | 134.451 | 134.451 |
| Face width (mm) | 37 | 37 |
| Addendum (mm) | 4.6455 | 1.9844 |
| Dedendum (mm) | 2.7176 | 5.3787 |
| Parameter | Pinion concave | Wheel convex |
|---|---|---|
| Cutter tip radius (mm) | 93.0329 | 94.2350 |
| Profile angle (°) | 22.5 | 22.5 |
| Radial cutter setting (mm) | 109.9071 | 114.7853 |
| Angular cutter setting (°) | 48.6637 | 48.7694 |
| Roll ratio | 2.87862 | 1.05991 |
| Vertical offset (mm) | 2.6534 | 0.0 |
| Axial setting (mm) | −1.8929 | 0.0 |
| Sliding base (mm) | 0.5952 | 0.0 |
| Blank mounting angle (°) | 18.3281 | 68.2229 |
| Second-order modified roll | −0.042582 | 0.0 |
| Third-order modified roll | 0.091392 | 0.0 |
Tooth contact analysis is carried out on the two flank surfaces transformed into a common fixed frame, using the condition of continuous tangency, with the pinion rotation stepped at fixed increments to yield the contact path, the transmission error and, through differential geometry, the major axis of the contact ellipse at each instant; the envelope of these ellipses forms the contact pattern. Deviations of the pinion flank are evaluated against a fully conjugate pinion surface, projected along the normal.
Pitch line versus mid-face height
Two placements of the same straight path were compared: along the pitch cone line, and along the mid-line of the tooth height. Placing the path on the pitch line gives a remarkably quiet mesh in kinematic terms, but the pattern sits near the wheel addendum, because height modification moves the wheel pitch cone towards the tip. Under light or moderate load the contact zone therefore approaches the tooth edge, which reduces strength and can excite strong vibration; the computed transmission error at entry and exit is −0.02507° and −0.02524°, closely symmetric. Placing the path on the mid-line of the tooth height keeps the pattern in the central band of the face, which is far better for load capacity, with entry and exit values of −0.02499° and −0.02482° and equally symmetric behaviour about the reference point. The flank deviation maps confirm that the pitch-line layout removes more material near both ends of the tooth than the mid-height layout does.
Relative sliding velocity separates the two options more sharply. Taking the pinion angular velocity as 1 rad/s, the pitch-line path shows small velocities, but they are not equal on the two sides of the contact and change sign: the fifteenth contact point falls close to the base cone, so a contact point on the opposite side of the pitch line reverses the friction force. The mid-height path produces larger sliding velocities, all of the same sign, so no friction reversal occurs and the risk of a reversing tangential load on the flank is eliminated.
Sensitivity to mounting error and spiral angle
Four single mounting errors were applied at grade 5 of GB 11365—89 for bevel and hypoid gears: shaft angle deviation 0.01 rad, shaft centre distance deviation 0.2 mm, pinion axial displacement 0.2 mm and wheel axial displacement 0.35 mm. With the path on the mid-line of the tooth height, every one of these perturbations shifts the pattern only up or down along the tooth height, so neither an internal nor an external diagonal develops. The sensitivity ranking, from largest to smallest effect, is shaft angle deviation, shaft centre distance, wheel axial displacement, then pinion axial displacement.
Mean spiral angle changes the geometry of the contact line itself. Four variants at 25°, 30°, 35° and 40° were generated with the tooth-length path, and a fifth at 25° with the path tilted into an internal diagonal by $\Delta = 3.0°$. Larger spiral angle increases the angle between the contact line and the pitch cone, which improves lubricant entrainment, and enlarges the pattern area, which raises load capacity. Maximum relative sliding velocity rises with spiral angle at the same time, and the diagonal variant is markedly worse on that measure.
| Mean spiral angle | Path orientation | Design contact ratio | Maximum relative sliding velocity (m/s) |
|---|---|---|---|
| 25° | along tooth length | 1.8684 | 0.65 |
| 30° | along tooth length | 2.3273 | 0.67 |
| 35° | along tooth length | 2.8242 | 0.71 |
| 40° | along tooth length | 3.2261 | 0.89 |
| 25° | internal diagonal, Δ = 3.0° | 2.4426 | 1.3975 |
The 40° case exposes a limit of the present prescription. Because only the transmission error at the mesh transition point is controlled, a longer contact path at large spiral angle forces both the shape and the magnitude of the error curve to change, and the resulting mismatch and wheel angular velocity variation raise the sliding velocity disproportionately; further increase of spiral angle amplifies the effect. The diagonal variant at 25° raises the contact ratio from 1.8684 to 2.4426 but concentrates the highest sliding velocity at the wheel root and the pinion tip, where scuffing resistance is weakest.
Design rules that follow
For mean spiral angles above about 30°, the contact path should be laid along the tooth length. Doing so keeps the relative sliding velocity low and unidirectional, and it makes the design contact ratio a function of face width alone, so contact ratio can be raised by widening the face rather than by stretching the tooth height. The path should sit on the mid-line of the tooth height rather than on the pitch cone: the pitch-line version comes closest to zero relative velocity at the contact point, but height modification pushes the pattern towards the pinion root and the wheel tip, and edge contact then appears under light load or mounting error, with a direct loss of load capacity.
For small mean spiral angles, the tooth-length path yields too little contact ratio to be useful, and the path should be tilted into an internal diagonal, which restores contact ratio and load capacity at the cost of a higher probability of scoring and scuffing. The trade-off is explicit and quantifiable: in the 25° example the diagonal raises the contact ratio by roughly 30% while more than doubling the peak sliding velocity. Between these two regimes, the mid-height tooth-length layout gives the best combination of small sliding velocity, a broad contact pattern and low sensitivity to misalignment.