Gear Surface Waviness: How It Drives Gear NVH

• 1
  What gear surface waviness does to noise
• 2
  How aggressive gear sharpening removes waviness
• 3
  When modifying waviness helps and when it does not
• 4
  Measuring waviness before and after sharpening
• 5
  Cutting parameters and setup for aggressive sharpening
• 6
  Waviness control methods compared

Waviness is the mid-spatial-frequency error left on a tooth flank after hobbing, shaping or grinding. It sits between form error, which is long wavelength, and surface roughness, which is short wavelength. A flank can measure Ra 0.4 μm and still whine, because roughness averages the profile and hides the periodic component underneath.

When two flanks mesh, every waviness peak contacts its mate at a slightly different roll angle. That shifts the effective transmission error along the path of contact. The shift repeats once per wave, so a flank with 30 waves per tooth generates an excitation near the 30th mesh harmonic. That is the tone you hear.

This is why polishing alone rarely fixes gear whine. Reducing Ra lowers friction and scuffing risk, but it leaves the wave pattern intact. The mesh still sees the same periodic error, and the same order line stands out on the spectrum.

The useful lever is the wave itself: amplitude, wavelength and phase across the face width. Change those and you change which harmonic gets energy. That is the whole basis for modifying gear surface waviness instead of just making it smoother.

• 1
  RoughnessShort wavelength, affects friction and film formation
• 2
  WavinessMid wavelength, affects mesh excitation and tone
• 3
  Form errorLong wavelength, affects load distribution

Aggressive gear sharpening is a stock-removal pass that cuts deeper than a normal finish pass, with a dressed wheel or cutter that stays open and free-cutting. The intent is not to chase a lower Ra number. The intent is to cut through the wave layer and leave a new, more random flank texture.

Typical stock removal runs 10–30 μm per flank on a ground gear, and more on a shaped gear with visible feed marks. Depth per pass is kept in the 5–15 μm range so the wheel does not load or burn. More passes at moderate depth beat one heavy pass, because a single deep pass pushes the wave pattern around instead of erasing it.

Wheel or cutter condition matters more than most shops admit. A glazed wheel rubs the flank and generates heat, which can temper the surface and leave a smeared layer. Dressing every 8–15 parts keeps the abrasive sharp and the cut cool. Coolant should be delivered at the mesh point, not aimed at the top of the blank.

After sharpening, the flank carries a shorter, more randomized wave pattern. Total waviness amplitude drops, and the dominant wavelength shifts. On the spectrum, the order line usually drops and, in some cases, spreads across neighboring orders. That spread is audible as a softer, less tonal sound.

This process pays off when the noise is tonal and tracks mesh order. If the spectrum shows a sharp order line that moves with speed, waviness is a strong suspect. If the noise is broadband and does not track mesh order, look at bearings, shaft stiffness or housing resonance before touching the flanks.

It also pays off on low-hardness gears and on gears where the wave layer is shallow. A 20CrMnTi gear at 58–62 HRC after case hardening has limited stock to give. Cutting 30 μm per flank may break through the case or leave a soft spot. Measure the case depth before you commit.

It does not help when the excitation is dominated by lead error, tooth-to-tooth spacing or runout. Those are form errors, and no finish pass fixes them. If the single-flank roll chart shows large low-order swings, correct the cutting setup first.

It also does not help if the mating gear is untouched and carries the same wave pattern. Waviness on one flank still meshes against waviness on the other. In practice, both members of a pair should be finished with the same strategy, otherwise the pair can generate a new beat frequency.

• 1
  Good candidateTonal whine that tracks mesh order, shallow wave layer
• 2
  Poor candidateBroadband noise, dominant lead or runout error
• 3
  Watch the caseCheck case depth before removing 20 μm or more

Waviness needs a profile measurement, not a roughness measurement. A contact profilometer with a 2 μm stylus and a 0.8 mm cutoff gives a usable trace, but the evaluation length must span at least 30 waves to see the pattern. On a typical 3 mm active flank, that means a longer trace than most roughness programs use.

On a gear measuring center, the same data comes from a flank trace in the profile direction and a lead trace across the face. Run both. Waviness in the profile direction excites mesh harmonics; waviness in the lead direction modulates contact and can create a second, lower-frequency tone.

Take baseline traces on at least three teeth, spaced 120° apart, before sharpening. After the pass, repeat on the same teeth. Comparing the same flank positions removes tooth-to-tooth variation from the result and makes the change visible.

Filtering choices matter. A Gaussian filter at 0.8 mm cutoff separates roughness from waviness in a predictable way. Switch to 0.25 mm and part of the wave pattern disappears into the roughness band. Keep the same filter settings across all traces, or the comparison means nothing.

On a gear grinder, a starting point for case-hardened steel is a wheel speed of 30–35 m/s, work speed 20–30 rpm and infeed 5–10 μm per pass. On softer, unhardened steel, work speed can go higher, up to 40 rpm, because the material clears the wheel more easily.

Coolant pressure should be high enough to penetrate the contact zone, typically 8–12 bar for grinding. Low-pressure flood coolant tends to ride on top of the flank and lets the wheel rub. That rub is what produces the smeared layer that later shows up as a false roughness reading.

For shaped gears, a sharpened cutter with proper rake geometry does the same job. Feed per stroke and cutter sharpness control the wave pattern. A dull shaper cutter leaves a periodic mark that mirrors the cutter index, which then appears as a high-order line in the spectrum.

Deburr and edge-break after sharpening. A sharp flank edge on the tooth tip can raise the contact pattern and create a new noise source. A 0.05–0.1 mm edge break at the tip and ends is enough. Do not round the active flank.