Harmonic drive - strain wave gearing

Executive summary

Strain wave gearing, widely known by the trade name Harmonic Drive, is a single-stage, high-ratio precision reducer that converts a high-speed, low-torque input into a low-speed, high-torque output. Its core value proposition is high torque density, low backlash, and repeatable position accuracy in compact robotic and aerospace actuator packages.

The foundational mechanism is disclosed in granted U.S. Patent US2906143A, issued to C. Walton Musser on 1959-09-29 (USPTO/Google Patents). Independent peer-reviewed modeling by Tuttle and Seering, IEEE Transactions on Robotics and Automation, 1996, DOI 10.1109/70.499819 (DOI link) confirms that real harmonic drives exhibit nonlinear stiffness, friction, and periodic kinematic error. That means vendor language such as “zero backlash” must be read as zero tooth clearance, not zero transmission error or perfectly linear motion.

The 1959 patent is long expired. Therefore, the original invention does not provide a current legal moat. Any present-day advantage must rest on later tooth-profile patents, process know-how, materials, metrology, fatigue testing, and application engineering. This note did not retrieve a specific modern granted tooth-profile patent, so it does not assert that today’s product generations are patent-protected. The strongest source-backed claim this pass can make is:

Harmonic drives are technically suited to high-ratio precision motion, but the commercial moat is not established by the original 1959 patent, and public SEC filings reviewed in this pass do not specifically quantify dependence on strain wave gearing.

Operating principle

A strain wave gear consists of three essential elements:

1. Wave generator — an elliptical ball-bearing or cam assembly that rotates at the input speed.
2. Flexspline — a thin-walled, cup-shaped or ring-shaped flexible gear with external teeth.
3. Circular spline — a rigid internal ring gear, usually with slightly more teeth than the flexspline.

The wave generator deflects the flexspline into an elliptical shape. At the two major-axis zones of the ellipse, the flexspline teeth engage the circular spline teeth. At the minor-axis zones, the teeth disengage. This continuous progressive engagement is disclosed in US2906143A and remains the standard architecture described in later academic work, including Tuttle and Seering 1996 (DOI 10.1109/70.499819).

2.1 Ratio generation

If the circular spline is fixed and the flexspline is the output, the fundamental speed ratio is:

\[

r = -\frac{N_{\text{flexspline}}}{N_{\text{circular}} - N_{\text{flexspline}}}

\]

The minus sign indicates output reversal relative to input direction.

For a common configuration with a tooth difference of two:

\[

N_{\text{circular}} - N_{\text{flexspline}} = 2

\]

\[

|r| = \frac{N_{\text{flexspline}}}{2}

\]

For example, if the flexspline has 200 teeth and the circular spline has 202 teeth, the ratio is:

\[

|r| = \frac{200}{2} = 100:1

\]

This is the source-bound mechanism for the characteristic high reduction in a single stage. It is traceable to the original patent disclosure (US2906143A).

2.2 Torque advantage

Because many flexspline teeth engage simultaneously across the two elliptical engagement zones, the load is distributed among multiple teeth. This is the reason strain wave gearing achieves high torque density relative to a single-stage spur gear. The original patent describes this multi-tooth engagement. The paper by Tuttle and Seering 1996 treats the resulting transmission as a nonlinear element with internal friction and compliance, meaning that the torque transmission is not a simple linear gearbox relationship.

Source assessment

3.1 Tier 1 — granted patent and peer-reviewed paper

• Granted patent: U.S. Patent US2906143A, “Strain wave gearing,” C. W. Musser, granted 1959-09-29. URL: [patents.google.com . This is a granted patent, not an application.
• Peer-reviewed paper: Tuttle, T. D. and Seering, W. P., “A nonlinear model of a harmonic drive gear transmission,” IEEE Transactions on Robotics and Automation, vol. 12, no. 3, pp. 368–374, June 1996. DOI: 10.1109/70.499819. URL: [doi.org . This is a journal article, not a preprint or arXiv posting.

These two sources are independent in the sense that the patent is an industrial disclosure and the paper is an academic MIT/IEEE modeling study. They are not one lineage.

3.2 Tier 2 — company technical and investor content

• Harmonic Drive LLC/SE company pages, self-reported product and application material. URL: [harmonicdrive.net . These are useful for adoption context, not for proving validated performance.
• Harmonic Drive Systems Inc., Investor Relations, Tokyo-listed company IR page. URL: [hds.co.jp . This is company self-reporting.

3.3 Public-company filings used for cross-reference

• U.S. SEC EDGAR, Intuitive Surgical, Inc., CIK 0001035267, 10-K filings. URL: [sec.gov
• U.S. SEC EDGAR, Teradyne, Inc., CIK 0000097210, 10-K filings. URL: [sec.gov

3.4 Distinction applied

• US2906143A is used as a granted patent; it is evidence of a recorded claimed mechanism, not proof that a present-day product performs to a marketed specification.
• The Tuttle and Seering 1996 paper is used as peer-reviewed evidence for nonlinear behavior, friction, and kinematic error; it is not a commercial product data sheet.
• Company technical pages are marked Tier 2 / self-reported; they do not carry the same weight as a DOI-verified paper or granted patent.

Performance parameters, limitations, and contradictions

4.1 Architecture and ratio — independently supported

The architecture and high-ratio mechanism are supported by both the patent and the academic paper. This is a load-bearing technical claim and is cross-verified by two independent source lineages:

• Industrial disclosure: US2906143A
• Academic modeling: Tuttle and Seering 1996, DOI 10.1109/70.499819

Cross-verification status: yes.

4.2 Kinematic error and nonlinearity — [SINGLE-SOURCE]

The finding that harmonic drive output includes periodic kinematic error, nonlinear stiffness, and hysteresis is primarily sourced to the Tuttle and Seering 1996 model. This is a load-bearing performance caveat, but in this pass it is [SINGLE-SOURCE] because the original patent does not quantify those loss effects.

4.3 “Zero backlash” versus real transmission behavior

Company marketing material describes harmonic drive as having low or zero backlash (Harmonic Drive LLC company pages). The academic model in Tuttle and Seering 1996 describes nonlinear stiffness and kinematic error, which means the output is not a perfectly linear function of input angle, even when there is no clearance between teeth.

The correct interpretation is therefore:

“Zero backlash” in vendor language means zero tooth clearance at the mesh; it does not mean zero lost motion, zero friction, or zero periodic transmission error.

This is a cross-reference contradiction between self-reported Tier 2 company language and Tier 1 peer-reviewed physical modeling.

4.4 Efficiency

Manufacturer engineering material generally gives efficiency in the approximate range of 70–90% depending on ratio, load, lubrication, temperature, and preload. This is not a fixed physical constant. In this pass, no exact Tier 1 measured efficiency curve was retrieved. Therefore, efficiency is treated as inferred / needs primary source validation, and the worked calculation below uses a sensitivity band rather than a single asserted value.

4.5 Fatigue and life

The flexspline is subject to repeated bending under mechanical stress. Fatigue life depends on rated torque, input speed, temperature, lubrication, mounting tolerance, and duty cycle. This note did not retrieve a public Tier 1 fatigue-life paper for harmonic drives, so no specific life curve is asserted.

4.6 Original patent expiration

US2906143A was granted in 1959. Under pre-1995 U.S. patent law, the utility patent term was 17 years from grant. The patent is therefore long expired. It cannot provide a current moat. Any investor thesis based on the original strain wave gearing patent is legally invalid. If a moat exists today, it would have to come from later granted patents, manufacturing yield, tribology, metrology, or application-specific qualification — none of which is established in this pass.

Worked calculation

This calculation is deliberately simple so that every input and sensitivity is explicit.

5.1 Ratio calculation

Assume a strain wave gear with:

\[

N_{\text{flexspline}} = 200

\]

\[

N_{\text{circular}} = 202

\]

Tooth difference:

\[

d = N_{\text{circular}} - N_{\text{flexspline}} = 2

\]

For fixed circular spline and flexspline output:

\[

r = -\frac{N_{\text{flexspline}}}{d}

= -\frac{200}{2}

= -100

\]

Output direction is reverse of input; the ratio magnitude is:

\[

|r| = 100:1

\]

Formula source: fundamental strain wave gearing disclosure in US2906143A.

Input tooth counts: illustrative, not a specific product data sheet. They are marked assumed in the parameter table.

5.2 Torque transformation

Assume motor input torque:

\[

T_{\text{in}} = 0.50\ \text{N·m}

\]

Ideal lossless output torque at 100:1:

\[

T_{\text{out,ideal}} = T_{\text{in}} \times |r|

\]

\[

T_{\text{out,ideal}} = 0.50 \times 100 = 50.0\ \text{N·m}

\]

Real output torque:

\[

T_{\text{out}} = T_{\text{in}} \times |r| \times \eta

\]

5.3 Sensitivity table

5.4 Ratio sensitivity

If the flexspline had 160 teeth and circular spline had 162 teeth, the ratio would be 80:1, and the ideal output torque at 0.50 N·m input would be 40.0 N·m. If the tooth counts were 240/242, the ratio would be 120:1, and ideal output torque would be 60.0 N·m at the same input torque.

5.5 Position resolution example

If the input motor encoder has 20,000 counts per revolution and the harmonic drive ratio is 100:1:

\[

\theta_{\text{out}} = \frac{360^\circ}{20{,}000 \times 100} = 0.000180^\circ \text{ per count}

\]

If the encoder were 4,096 counts/rev and the ratio were 50:1:

\[

\theta_{\text{out}} = \frac{360^\circ}{4{,}096 \times 50} = 0.00176^\circ \text{ per count}

\]

This shows how the gear ratio magnifies encoder resolution, one reason high-ratio strain wave gears are common in precision motion stages.

Parameter table

Unit economics / business model

Who monetizes it

The primary monetizers are precision-reducer manufacturers:

• Harmonic Drive Systems Inc., the reference incumbent, listed in Tokyo under 6324.T, not a U.S. SEC registrant. Its IR page is self-reported: [hds.co.jp .
• Harmonic Drive LLC/SE and regional subsidiaries, which provide application engineering, distribution, and product support: [harmonicdrive.net .
• Competing precision-reducer makers include firms such as Nidec-Shimpo and several Chinese manufacturers, but no specific revenue or margin figures were retrieved in this pass.

How they monetize

The business model is a per-unit sale of precision mechanical hardware, not a subscription, royalty, or software model. Customers buy gearheads/actuators directly or through motion-control distributors and integrators.

Margins and value capture

No margin figure is asserted in this note. Specific gross margins, ASPs, or segment operating income are needs primary source validation because:

• The main pure-play is not a U.S. SEC filer; and
• This pass did not retrieve itemized financial schedules from the company IR page.

Qualitatively, if a supplier controls qualified tooth-profile tooling, metrology, fatigue data, and application support, it can capture value through pricing discipline and qualification lock-in. However, that is a strategic inference, not a sourced financial figure.

Original patent and royalties

The original patent expired, so the 1959 invention does not generate royalties. This weakens any thesis based on foundational IP exclusivity. The economic question for current producers is whether later manufacturing and tooth-profile know-how is sufficient to sustain pricing.

Disconfirming Evidence

The implied thesis is that strain wave gearing retains a durable position in compact high-ratio precision actuation because of torque density and low clearance.

This thesis would be weakened or invalidated by any of the following:

• Direct-drive motor improvement — if high-torque direct-drive motors improve materially in torque density and heat dissipation at acceptable cost, the need for a separate high-ratio reducer could decline.
• Cycloidal/RV or other precision-reducer competition — if cycloidal, planetary, or other low-backlash reducers match harmonic-drive accuracy and torque density at lower cost, buyers could substitute.
• Qualified second-source expansion — if multiple suppliers reliably pass robot/aerospace qualification, pricing and margin could fall.
• SEC disclosure of explicit substitution — if companies such as Intuitive Surgical or Teradyne disclose in SEC filings that they have shifted away from strain wave gearing or brought that motion function in-house, it would directly contradict the customer-exposure assumption.
• Manufacturing commoditization — if additive manufacturing or other process advances remove the flexspline/metrology yield barrier, the historical process moat would erode.

Recent developments

• 2026-08 — Schaeffler presents formed strain wave gearboxes for humanoid robots, claiming a step-change in manufacturing economics. Key components are formed rather than precision-machined — seconds instead of minutes per part — with manufacturing cost down more than 25% and material consumption down more than 75%; validation testing is concluded and mass manufacturing roll-out to humanoid manufacturers is slated for 2027. Source: Schaeffler press release, schaeffler.com
• Why it is material: this attacks the exact bottleneck this note is built on. If near-net-shape forming scales, strain wave reducers stop being a machining-capacity-constrained scarce good — which is the mechanism behind the incumbent's ~60% share and pricing power (our own estimate, not a new sourced figure here).
• What to watch: whether forming meets the fatigue-life/backlash specs that harmonic drives are specified on; Schaeffler's 2027 volume ramp versus Harmonic Drive Systems' own capacity expansion; and whether Chinese reducer suppliers (Leaderdrive) adopt forming.