Why apparently identical gears have different noise profiles
Two gears made to the same drawing, from the same material, with the same tooth count. Why does one train whine and the other not?
A gear pair is supposed to transmit motion conjugately: for every increment of input rotation, an exactly proportional increment of output rotation. Real teeth do not. The difference between the output rotation you get and the output rotation the ratio predicts is transmission error, and transmission error is the excitation that becomes noise.
That is the whole mechanism. Transmission error puts a fluctuating force into the shaft, the shaft puts it into the bearings, the bearings put it into the housing, and the housing radiates it as sound. Two gears identical on a caliper can differ by a factor of five in transmission error, because the deviations that produce it are measured in micrometers and are not what a caliper measures.
The frequencies tell you which deviation you have
The value of the trace above is not its amplitude, it is its composition. Each manufacturing deviation appears at its own frequency, so a spectrum taken with a tachometer reference identifies the cause without a teardown.
fm = z · n / 60
- fm
- mesh frequency, Hz
- z
- number of teeth on the gear being considered
- n
- rotational speed of that gear, rpm
| What you hear or measure | Frequency | Usual cause | Where it is made |
|---|---|---|---|
| Steady whine | Mesh frequency and its harmonics | Profile deviation, poor surface finish, insufficient tip relief under load | Hob condition, cutting parameters, mold surface, post-mold shrinkage |
| Warble or beat on the whine | Mesh frequency with once-per-revolution sidebands | Runout of the gear blank or the bore, eccentric mounting of a molded gear on its shaft | Blank turning, bore concentricity, press fit, gate position on a molded gear |
| Growl at low speed | Broadband, load dependent | Pitch or spacing error, single tooth defect, nick or burr from handling | Indexing, deburring, handling and packaging between operations |
| Rattle in a reversing duty | Impulsive, not tonal, tied to reversals | Excess backlash plus low damping, driveline free to accelerate across the backlash gap | Center distance, tooth thickness, and endplay in the stack |
| Rising noise across a production lot | Mesh frequency, amplitude drifting over time | Tool wear moving the profile, mold degradation, an upstream material change | Tool life management, process control, incoming material control |
The deviations the drawing does not control
A typical fine-pitch gear drawing calls out module, tooth count, pressure angle, tooth thickness or span, material, and an outside diameter. That is enough to make a gear that meshes. It is not enough to make a quiet one. Noise is set by four deviations that a quality grade covers and a bare drawing usually does not.
- Profile deviation. How far the real flank departs from the involute it should follow. Profile deviation puts energy directly into the mesh harmonics.
- Lead or helix deviation. How the flank varies across the face width. A lead error concentrates contact at one end of the face, which raises local stress and changes the noise with load.
- Pitch and spacing deviation. How evenly the teeth are indexed around the blank. Single-pitch error creates an impulse once per revolution of that gear.
- Runout. Eccentricity of the tooth field relative to the bearing bore. Runout is the classic source of a once-per-revolution modulation, and it is often introduced after cutting, by the press fit or the bore, rather than by the gear cutting itself.
This is why two gears from the same drawing differ. A gear cut in one setup on a sharp hob and a gear cut after a re-chuck on a worn hob both satisfy the drawing. Their profile and runout are different, and profile and runout are what the ear hears.
Material and manufacturing route
Molded polymer gears carry a different set of causes than cut metal gears. A molded gear is a copy of a cavity that was itself cut and polished, so cavity condition sets profile quality for the whole population. Shrinkage after molding is anisotropic and continues for some time after the part leaves the tool, so gears measured immediately after molding and gears measured a week later are not the same size. Hygroscopic materials add another shift: several common polyamides absorb moisture and grow measurably, which changes tooth thickness and effective backlash.
Sintered gears carry density variation. A pressed and sintered part has a density gradient set by the compaction, and local density changes elastic modulus and surface finish, which changes both the deflection under load and the way the flank runs in. Cut steel gears are the most controllable and the least forgiving: high stiffness means low damping, so the same transmission error radiates more sound than it would through a polymer.
Tradeoffs when you try to fix it
Every common noise fix costs something else, and the cost is usually paid in a different specification line.
- Tighter tooth quality. Real reduction in mesh-frequency noise, real increase in cost, and it does nothing for runout introduced by the press fit downstream.
- Higher contact ratio, from a finer pitch or a longer addendum. Smoothes the load transfer between teeth and reduces transmission error, but reduces tooth root strength and increases sensitivity to center-distance variation.
- More backlash. Removes the risk of binding and tight-mesh current draw, and directly increases rattle in a reversing duty cycle.
- Polymer stages. Adds damping and cuts radiated noise, and gives up torque capacity, stiffness and dimensional stability with temperature and humidity.
- Softer mounting. Decouples the housing from the chassis and cuts radiated sound at the product level, and changes the alignment the gear train sees under load.
What to measure next
-
Measure with a stated method
Fix the microphone distance, the mounting, the background level, the supply voltage and the load. Noise data taken under varying conditions cannot be compared across a lot.
-
Take a spectrum with a speed reference
An order analysis separates mesh frequency, its harmonics and once-per-revolution content. Without a speed reference, the spectrum moves as the load changes and the components blur together.
-
Measure runout of the assembled train
Runout at the tooth field of each gear, referenced to its own bearing bore. Runout added by assembly is common and invisible in a component inspection.
-
Vary one thing at a time
Swap one gear between a quiet and a loud unit. If the noise follows the gear, it is a component problem. If it stays with the housing, it is a mounting or alignment problem.
A noise investigation that reaches a cause almost always reaches it through the frequency, not through the loudness. The loudness tells you a problem exists. The frequency tells you which operation made it.