A multi-roller traction drive can replace the helical gear reducer in a high-speed EV drive unit. It runs as quietly as a bearing, handles input speeds over 100,000 RPM, and gets a 14.7:1 ratio in a single planetary stage.
EV motors already run above 20,000 RPM, and they keep getting faster. At those speeds a helical gear reducer gets harder to keep quiet, tooth loads go up, and so do losses. A traction drive has no teeth, so there is no gear mesh to deal with.
High-speed ball bearings already run at these speeds, quietly, for thousands of hours. A traction drive puts that same rolling contact to work carrying torque.
A traction drive sounds like a bearing, not a gearbox. There is no tooth impact to excite the housing at any speed. One documented case measured a traction drive 24 dB quieter than a comparable gear drive (NASA TP 1710).
Nasvytis drives have been run to 480,000 RPM, and one ran 43 continuous hours at 360,000 RPM. A 100,000 RPM EV motor is well below speeds these drives have already reached.
The stepped planet layout gets to about 15:1 in one compact stage.
Every ball bearing is a traction drive. The inner race is a sun, the outer race is the ring, and the balls are the planets. A traction drive uses those same rolling contacts, sized and loaded to carry real power.
Torque passes through a thin elastohydrodynamic (EHL) traction film. Under contact pressure the traction fluid briefly behaves like a solid, and shearing it carries the tangential force. The metal surfaces never touch. Life is calculated with the same Lundberg-Palmgren L10 methods used for rolling bearings.
The Nasvytis architecture adds compound stepped planets. Each planet set carries a large roller (P1) contacting the sun and a smaller coaxial roller (P2) driving an outer row of P3 rollers. Compounding two stages inside one ring gets to 14 or 15:1, with the load shared across 25 EHL contacts.
The layout is flexible. One, two, or three planet rows cover practical ratios from 3:1 to 300:1, and changing roller sizes moves the ratio around within that range. Any member (sun, ring, or carrier) can be held as the reaction member, or all three can rotate in a split-power configuration.
Table I — Traction drive vs. helical gear reducer
| Attribute | Helical reducer | Multi-roller traction drive |
|---|---|---|
| Noise, vibration | Gear mesh excites housing | Bearing-grade quiet |
| Max input speed | Limited by tooth loading | 100,000+ RPM |
| Ratio per stage | Typically 4–6:1 | 14–15:1 |
| Peak efficiency | ~98% | ~98% |
| Life prediction | Gear tooth fatigue methods | L10 bearing methods |
| Packaging | Offset layouts | Coaxial, cylindrical |
| Lubricant | Standard gear oil | Requires traction fluid |
| Supply base | Mature, global | Bearing industry processes apply |
■ marks the stronger option. Gears still win on supply chain maturity and use ordinary gear oil. The traction drive wins where motor speed is high.
This is an example sizing for a 20,000 RPM EV motor. The drive is a cylinder that mounts coaxially on the motor output shaft. The 3D model shows the drive assembly with simplified roller surfaces; the second row rollers run on rolling element bearings.
Table II — Rated Operating Point
This is not a new idea. Algirdas Nasvytis developed and patented the stepped planet design at TRW in the 1960s. NASA Lewis (now Glenn Research Center) tested it in the late 1970s and early 1980s, at speeds well above what EV motors run today.
NASA reports a 3.7 kW, 120:1 three-row Nasvytis speed increaser that ran 15 minutes at 480,000 RPM, and 43 consecutive hours at 360,000 RPM without liquid lubrication (TP 1378, RP 1154).
The first Nasvytis drive tested was a 373 kW torpedo drive with three planet rows, a 48.2:1 reduction ratio, and a 53,000 RPM input.
NASA Lewis ran its own tests to 73,000 RPM sun speed and 127 kW (170 hp), using 1970s traction fluids. Reported in NASA Technical Papers 1378, 1710 and 2027.
Efficiencies up to 97 percent were recorded in parametric tests of improved Nasvytrac drives (NASA RP 1154). The NASA Lewis hardware reached 95 percent at the highest torque tested, and efficiency was still climbing when the test stand hit its torque limit (TP 1378).
Traction contact life is calculated with the same Lundberg-Palmgren methods bearing engineers already use (NASA TP 1710).
The Nasvytis drive worked. It was shelved because the application it needed didn't exist yet.
Modern traction fluids reach a peak traction coefficient near 0.12, about twice what was available during the NASA tests. A higher coefficient carries more torque at the same contact load.
EV motors keep getting faster to save size and weight. A faster motor needs a higher reduction ratio, and that is where a traction drive has the edge over gears.
An electric motor doesn't need a variable ratio. The CVT complexity that held traction drives back in the past isn't needed in an EV drive unit.
Joe Kliewer is a named inventor on eleven U.S. patents, granted from 2014 to 2025, covering traction CVT design, hydraulic ratio control, preload and backlash management, and electrified propulsion. They are assigned to Orbital Traction and GM, not to Traction Drive LLC, but they are the background this work builds on.
Table III — Issued U.S. Patents, J. Kliewer, most recent first
| Patent No. | Title | Assignee | Granted |
|---|---|---|---|
| US 12,420,632 | All-wheel drive system for electric vehicle | GM Global Technology Operations | 2025 |
| US 12,194,827 | Integrated electrified propulsion system for a vehicle | GM Global Technology Operations | 2025 |
| US 10,704,688 | Preload and torsional backlash management for a CVT device | Orbital Traction, Ltd. | 2020 |
| US 10,443,692 | Fluid movement systems including a continuously variable transmission | Orbital Traction, Ltd. | 2019 |
| US 10,302,178 | Hydraulic race shifter for a CVT device | Orbital Traction, Ltd. | 2019 |
| US 10,006,526 | Fluid movement systems including a continuously variable transmission | Orbital Traction, Ltd. | 2018 |
| US 9,995,375 | Symmetry control for a CVT device | Orbital Traction, Ltd. | 2018 |
| US 9,765,863 | Hydraulic race shifter for a CVT device | Orbital Traction, Ltd. | 2017 |
| US 9,702,459 | Preload and torsional backlash management for a CVT device | Orbital Traction, Ltd. | 2017 |
| US 9,464,701 | Fluid movement systems including a continuously variable transmission | Orbital Traction, Ltd. | 2016 |
| US 8,740,743 | Continuously variable transmission device and method | Orbital Traction, Ltd. | 2014 |
CVT: continuously variable transmission. Each link opens the full patent at Google Patents.
Traction Drive LLC is led by Joe Kliewer, a bearing and traction drive engineer with over 20 years designing rolling element contacts.
Before this he worked on traction CVTs and EV propulsion systems, including hydraulic ratio shifting and torsional damping. A fixed ratio multi-roller drive is a simpler machine than a CVT, and most of that experience carries straight over.
Also Available — Consulting
Bearing design for advanced reactor programs: design basis for bearings running in sodium, friction estimates, and material selection (Stellite, Inconel, ceramics) for high temperature, corrosive service.
Hertz contact stress, EHL film thickness, and Lundberg-Palmgren life calculations, for traction drives and standard rolling bearings.
If you are looking at a traction drive for an EV program, or need help with a bearing design, send a note and Joe will get back to you.