Quieting Gears by Deliberately Making Teeth Differ: Gleason's MicroForm (Psychoacoustic Tooth Flank Modification) Explained

Why ground gears whine

As a gear turns, each tooth entering mesh produces a small impact and a rotational unevenness (transmission error). A 13-tooth pinion at 3,000 rpm does this 650 times a second. This repetition produces a sound at the gear mesh frequency.

A ground gear has teeth that are nearly identical, with very small indexing (pitch) error. As a result, the impacts repeat with exactly the same timing and strength every time, so the sound energy concentrates at the gear mesh frequency and its integer multiples, and the frequency peaks become sharp. Gleason's researchers likewise explain that ground bevel gears, despite very small indexing error, are often noisy because the impacts as teeth enter mesh are too uniform in timing^[1].

Tonality: a psychoacoustic metric of how much a pure tone at one particular pitch stands out within a sound. Even at the same sound level (decibels, loudness), people hear a sharp peak as an irritating, piercing whine. Conversely, a gear that measures as loud may not bother the human ear at all^[2].

In combustion-engine cars, engine noise masked gear noise, but in electric vehicles the motor is quiet, so the noise of reducer and differential gears is heard as it is. Truck and agricultural machinery makers have also been asking more and more for quieter gears recently^[1].

The "scatter the teeth" idea: failure and success

To break up the sharp peaks, it should be enough to turn an impact that is the same every time into a slightly different impact each time. However, that approach once failed badly.

1. HintFans and tires
   Automotive cooling fans deliberately space their blades unevenly, and tires vary the spacing of their tread pattern. The idea is to split a single-frequency tone into small surrounding sounds (sidebands), turning a harsh sound into a smooth one^[3].
2. FailureDeliberately disturbing the tooth spacing
   To do the same with gears, gears with randomly shifted tooth positions (a spacing error of 5 microradians) were tested. The 1st mesh order peak actually grew, and lower-order noise increased too, so tonality did not drop in any of the tests. The accuracy grade drops, making quality harder to monitor, and under load the contact ratio also drops, which affects strength. The difference from tires is that a vehicle's speed stays constant, whereas with gears, uneven spacing makes the rotation itself speed up and slow down.^[3].
3. SuccessTopography scattering (RWTH Aachen)
   Professor Brecher and colleagues at the Laboratory for Machine Tools and Production Engineering (WZL) at RWTH Aachen University in Germany studied a method that varies, instead of tooth position, the flank shape (helix angle, pressure angle, etc.) from tooth to tooth. It was the first method to actually reduce tonality, and test vehicles were rated as quieter^[2][4][10].
4. IssueBut it is not suited to mass production as is
   Varying the helix and pressure angles tooth by tooth shifts tooth thickness and indexing. In an example with an 11-tooth pinion, the result was flank corner deviations of around ±10µm and pitch variation of about ±5µm. The accuracy grade drops by several grades, and because the contact pattern changes tooth by tooth, load may concentrate at the corners. Moreover, one machine setup is needed per tooth space, so data processing and correction from measurements are multiplied by 11^[4].
5. SolutionMicroForm (Gleason)
   It leaves tooth position and tooth thickness untouched and applies only a sine-wave form that is zero at the center, with a different magnitude on each tooth. The next section looks at it in detail^[4][7].

How MicroForm works: a two-stage modification

MicroForm divides the tooth flank modification into two stages.^[4][7].

Why accuracy does not drop with this shape

• Center is always 0: tooth indexing (pitch) and tooth thickness are normally measured at the center point of the flank. A sine wave is zero at the center, so every tooth's center point stays as the original. Hence there is no indexing error or tooth thickness error^[5].
• Zero slope where the impact occurs: at light load, the contact actually occurs between a sine-wave peak and valley (about 50% of the flank). The point where mesh passes to the next tooth falls exactly on a peak or valley, where the slope is 0, so the impact is no worse than before and only its timing changes from tooth to tooth. With a linear modification, a slope remains at this point and the impact gets worse^[5].
• Corners barely move: even with a 10µm sine amplitude, the flank corner deviation stays within 0–3µm, about half that of earlier methods. "Edge contact", where contact occurs at a corner, leads to chipped teeth, so this matters^[5].
• The mean equals the original flank: averaged over all teeth, the flank is calculated to equal that of a gear with no modification^[6].
• Robust to axis misalignment: a sine wave along the path of contact makes corner contact unlikely even when the pinion and gear mounting is slightly misaligned^[5].

What the machine does

MicroForm is realized through the axis motions of Gleason's free-form bevel gear grinding and cutting machines (Phoenix series). A small rotation that depends on roll position is added to the workpiece rotation axis (A axis), and the X, Y and Z axis positions are corrected to match. Changing the A-axis rotation amounts to slightly changing the helix angle and the pressure angle at the same time^[5].

• Machine settings and the machining program do not change: the MicroForm software calculates the tooth-by-tooth modification in parallel with machining. Just by switching it on, grinding time does not increase^[6][1].
• Closed-loop correction from measurement works as usual: since no separate setting is needed per tooth space, correcting the machine settings from CMM results also takes just one set^[6].
• Wave length and height can differ between toe and heel: for example, a long wave (0.8× frequency) at the toe and a short wave (1.2×) at the heel^[6].
• Applicable only to gears made by generating: if both gear and pinion are generated, apply it to both; if the gear is made by the non-generated (Formate) method, apply it to the pinion only^[5].

A pitfall found in mass production: the center of the measurement gridThe zero point of the sine wave (the roll center) and the center of the CMM measurement grid are often slightly offset in practice, and once mass production started, this sometimes caused indexing error to look larger in the measurement report. So around the center, a dwell section (0–4° of roll angle) with no modification was introduced. With the dwell section, tonality reportedly dropped even further^[6].

How much quieter does it get?

Third-party study: topography scattering cut tonality by 20–30%

The MicroForm figures themselves come only from the manufacturer, but the idea that varying the flank tooth by tooth lowers tonality has been confirmed by university research. In a study by WZL at RWTH Aachen University with Daimler and Klingelnberg, ground bevel gears (passenger-car rear axle) had crowning, pressure angle, helix angle and more varied tooth by tooth (random scattering of up to 2µm / 0.02°, and normally distributed scattering of up to 10µm / 0.10°). The results were as follows^[10].

The same study also reports that the peaks of mesh orders 2–8 dropped substantially, but that sounds at orders below the 1st mesh order increased, and that the effects on gear strength and efficiency have not yet been investigated. It says two to four flank variants are enough^[10].

Manufacturer's report: mass production at many axle plants

Gleason reports that MicroForm is already used in mass production at many automotive axle plants, that tonality has dropped, that indexing error is small, and that no failures due to edge contact have been recorded. In an example of single flank test results analyzed by frequency, switching MicroForm on lowered the peaks at all mesh orders except the 11th. A test with the pinion position shifted in 9 ways also showed reductions at many orders, but some combinations did not drop, and in those cases the MicroForm input values are changed to adjust^[6].

The inventor explains that the gears come out quieter than gears finished by fine honing, polishing or lapping, and that the cost per part is the same as ordinary grinding^[1]. However, this is the manufacturer's own assessment, and no third-party comparison test of MicroForm itself was found in this research.

Comparison with earlier noise-reduction methods

Sources: Gleason's technical book^[3][4][6], Gear Technology interview^[1], studies by WZL and others^[10].

What it means on the shop floor

Spreading to cylindrical gears

In 2025, WZL presented a method of applying a different crowning to each tooth in continuous generating grinding of cylindrical gears (threaded grinding wheel). It is a design method that builds the tooth-specific shapes into a single threaded grinding wheel and reproduces them with a dressing gear, and it is said to apply the scattering without slowing down generating grinding. However, confirmation by machining experiments is still to come^[11]. If it also becomes usable on the cylindrical gears common in EV reducers, the impact could be considerable.

Watch out for patentsMicroForm's approach is patented in Europe (EP 4208307 B1), Japan (Japanese Patent No. 7753346), China and elsewhere, and is pending in the US^[7][8][9]. If you apply "tooth-by-tooth sine-wave modification" on your own machines or software, check the scope of the rights.

FAQ

Doesn't the accuracy grade drop if the teeth have different shapes?

Since MicroForm does not change the center point at which indexing and tooth thickness are measured, the indexing and tooth thickness grades are said not to drop. The flank corner deviation is also only about 0–3µm^[5]. However, profile and lead charts look different from tooth to tooth, so how to set the inspection criteria needs to be discussed with the customer.

How much is changed?

In the book's example the sine amplitude is 10µm. WZL's study tried scattering of up to 2µm or 10µm^[5][10]. It is a µm-scale world that you cannot see or feel by eye or by hand.

How is it different from lapping?

Lapping is a finishing process in which the gear and pinion are run together as a set with abrasive to blend the contact. MicroForm builds the flank shapes as calculated at the grinding stage. Its advantages are said to be no need to finish as a pair and no increase in grinding time^[1].

Can it be used on cylindrical gears (spur and helical)?

MicroForm is a technology for bevel and hypoid gears. A similar method of applying scattering in generating grinding of cylindrical gears has only reached the stage where RWTH Aachen presented a design method in 2025^[11].

Does all the noise get quieter?

No. The aim is not loudness but reducing the piercing "sharpness" that catches the ear (tonality). In WZL's tests too, loudness (the perceived sound level) hardly changed^[10].

For gear basics (module, pressure angle, accuracy grade), see Gear study; for machining methods such as grinding and honing, see Gear machining methods; for accuracy grades and error items, see Gear calculation and standards.