Transmissions
Choose the ratio from chapter 4, then choose the mechanism that can carry the load for the required life. For every transmission, a designer must do these.
- Multiply output torque by efficiency:
tau_output = eta * G * tau_motor[1]. - Check torque margin on both sides of the reducer [2].
- Put rotor inertia in the model as
G^2 * J_rotor[3]. - Check life and heat, as well as strength.
- Add the backlash, hysteresis and lost motion to the position error budget.
Price rows are thin. The catalog holds a few quantity 1 prices for timing belts, bevel gears and lead screws, but they come from stores whose terms do not allow copying, so this site does not show them. Strain wave, planetary and cycloidal gearboxes, ball screws and roller screws have no price rows. Ask the maker.
Families
| Mechanism | Ratio in the database | Backlash | Efficiency | Choose it for | Main costs |
|---|---|---|---|---|---|
| Harmonic gearbox | not shown | Backlash-free mesh per the maker; typical hysteresis 3 arc min at ratio 30 [4] | At rated torque; correct below it [5] | High ratio in one short coaxial stage | Wind-up, flexspline fatigue, ratcheting |
| Cycloidal and RV gearbox | not on file | 1 to 1.5 arcmin upper bound | not on file | Shock and overload, integrated output bearing | Eccentric vibration; mass; cost |
| Planetary gearbox | not shown | Stated as an upper bound per grade; values not shown | Falls with each added stage; values not shown | Power density, many vendors | Backlash unless a reduced option is bought |
| Worm gearbox | none | none | Low at the mesh; no number on file | Large ratio, right angle, possible self-lock | Heat; self-lock is not guaranteed; not backdrivable |
| Capstan and cable | none | Very low: no gear mesh | not shown | Backdrivable light joints | Tension loss; few catalog parts |
| Synchronous belt | not on file | Low, depends on tension [6] | not shown | Motor placed away from the joint | Rating falls below six teeth in mesh |
| Lead screw | Lead 2 to 8 mm | Grows with wear | 30 to 50 percent [7] | Simple, can self-lock | Sliding wear; not for continuous power |
| Ball screw | Lead 2 to 64 mm | Preload can remove it | About 95 percent [8] | Efficiency and life | Bulky nut; back-drives, so a vertical axis may need a brake |
| Roller screw | not shown | Option to eliminate | Roughly 75 to 90 percent (class note, Wikipedia figure) | Highest load in a short nut | Up to an order of magnitude over a ball screw (class note, Wikipedia figure) |
General gear guidelines
Friction often rises with ratio [9]. Pressure angle spreads the gear centres, and that spread can deform the structure until teeth skip [10]. Do not carry current through gears [11].
| Level | Statement | Check | Source |
|---|---|---|---|
| should | Multiply gear output torque by efficiency: tau_gear = eta G tau_motor, with eta at or below 1. | calc: tau_output = eta * G * tau_motor | [1] |
| guide | Expect gearhead friction to often rise with ratio. | measure: friction torque per ratio option | [9] |
| must | Check gear trains for tooth pitting, brinelling, bending fatigue and strength at nominal and peak load, plus backlash, temperature and misalignment, throughout service life. | calc: gear stress and wear analysis has one line per condition | [12] |
| must | Include impact tooth loads from maximum combined axial, radial and moment loads in gear analysis. | calc: tooth stress at impact load | [13] |
| must | Do not carry electrical current through gears. | review: no gear in a ground or signal path | [11] |
| must | Account for the spreading force of the pressure angle on gear centres; often overlooked, it deforms the structure so teeth skip. | calc: F_spread = F_t tan(phi); centre distance stiffness check | [10] |
| guide | Standard low contact ratio spur gears near 1.3 raise dynamic loads. | calc: contact ratio noted per mesh | [14] |
| guide | Contact ratio above 2 shares load between two or three teeth and should lower dynamic load and noise, but such gears tend to have lower bending strength and more tooth sliding. | calc: contact ratio > 2 if tooth load sharing is claimed | [15] |
Strain wave gears
Strain wave catalogs print four torque levels. Each has its own guideline. Read all four from the maker's sheet for your size and ratio, for example the CSG-25, 100:1. The catalog lists the Harmonic Drive and Leaderdrive gears by name only, without their values.
| Level | When it applies | Source |
|---|---|---|
| Rated | Continuous load at 2000 rpm input; use the life equation for other speeds | [16] |
| Average torque limit | Cube-weighted average over the cycle must stay below it | [17] |
| Repeated peak | Acceleration and deceleration torque | [18] |
| Momentary peak | Collision or emergency stop only, never in a normal cycle | [19], [20] |
Leaderdrive uses its own torque names and rates at a higher input speed than the 2000 rpm basis of the strain wave rated-torque guideline [16], so rated values do not compare directly. The strain wave gear guidelines come from the Harmonic Drive CSF and CSG manual. Read Leaderdrive's own sheet for its limits.
Under normal operating conditions, check life with Lh = Ln (Tr/Tav)^3 (Nr/Nav) [21]. Rated L10 life is 7,000 hours for CSF and 10,000 hours for CSG [22]. The gear procedure does not cover the output bearing, so check it separately ([23], chapter 7).
Model the gear as a three-segment spring, not one stiffness [24]. Backdriving torque is not a holding torque [25].
| Level | Statement | Check | Source |
|---|---|---|---|
| must | Use strain wave gear rated torque as allowable continuous load at 2000 rpm input. | calc: rated torque applied at its 2000 rpm input basis; apply life equation for other speeds | [16] |
| must | Calculate average torque when load and speed vary, and keep it below the table limit. | calc: T_av = cube_root(sum(n_i t_i abs(T_i)^3)/sum(n_i t_i)) <= limit for average torque | [17] |
| must | Keep acceleration and deceleration torque below repeated peak torque. | calc: T1, T3 <= repeated peak torque | [18] |
| must | Do not let momentary peak torque occur during normal operating cycles; allow it only in a collision or an emergency stop. | sim: normal trajectories stay under repeated peak; momentary peak only for E-stop | [19] |
| must | Limit momentary peak events to N = 1.0e4 / (2 (n/60) t), with n input speed before collision and t collision time; beyond this the flexspline may fatigue. | calc: N_events <= 1.0e4 / (2 * (n/60) * t) | [20] |
| must | Stay under ratcheting torque; ratcheting can leave the flexspline non-concentric and shorten life. | calc: max torque <= ratcheting torque table | [26] |
| should | Under normal operating conditions, estimate wave generator life by Lh = Ln (Tr/Tav)^3 (Nr/Nav). | calc: L_h >= required hours | [21] |
| should | Use rated L10 life of 7,000 hours for CSF and 10,000 hours for CSG. | calc: L_n in life equation by series | [22] |
| must | Apply a compensation factor when strain wave gear load is below rated torque, because published efficiency is at rated torque. | calc: eta = eta_graph * compensation for T < T_rated | [5] |
| should | Do not rely on backdriving torque to hold load; fit a failsafe brake. | review: holding-critical joints have a brake | [25] |
| should | Model strain wave torsional stiffness as three segments K1, K2, K3 across T1 and T2, not as one constant. | sim: three slope stiffness model around the operating torque | [24] |
| should | Typical hysteresis loss is 3 arc min at ratio 30; use table 44 for the applicable size and ratio. | calc: positional error budget includes hysteresis | [4] |
| should | Note that an Oldham coupling adds small backlash even with a zero backlash gear element; a direct drive version avoids it. | calc: backlash budget includes Oldham table value | [27] |
| must | Keep strain wave gear oil at or below 90 C, since it quickly loses lubricating ability above this. | measure: oil temperature <= 90 C | [28] |
| should | Read grease temperature limits as grease temperature during operation, not ambient. | measure: grease temperature during operation, not ambient | [29] |
| should | Check output bearing life separately for housed units: the gear size procedure does not cover it. | calc: output bearing L10 from axial, radial and moment loads | [23] |
Cycloidal and RV gearboxes
Data in the database come from Nabtesco RV-E and Sumitomo Fine Cyclo catalogs. This site shows the Nabtesco rows that it may copy. Over those rows, rated torque runs 58 / 333 / 784 N.m, torsional stiffness 20 / 147 / 980 Nm/arc-min, and intermittent output speed 25 / 55 / 100 rpm (minimum / median / maximum, gearbox category). Part examples: RV-6E at 58 Nm and 1.5 arc-min, RV-80E at 784 Nm and 1.0 arc-min. The larger RV-1500E and the Sumitomo series are listed by name only. No transmission guideline covers this family specifically. The general gear guidelines and the margin guidelines apply.
Planetary gearboxes
Each added stage raises the ratio, and costs efficiency, length and mass. Compare the stage options of one frame on the maker's sheet before you add a stage. The catalog lists the maxon GP 32 HP in 2-stage, 3-stage and 4-stage versions by name only: its source does not allow copying the values.
In the Neugart PLE line, backlash falls and allowed radial force rises with frame size, from PLE040 to PLE160. These parts are listed by name only. The database has no gear ratio row for these parts, so reflect the load with the ratio you order. For a QDD, see the quasi-direct-drive ratio guidelines [30], [31].
The Apex AF tables list ratio, torque, mass, efficiency and a backlash upper bound for each precision grade. The catalog lists these gearboxes by name only: the values are not shown.
Spur gear contact ratio matters too. Standard spur gears near a contact ratio of 1.3 raise dynamic loads [14]. Above 2, load is shared between two or three teeth, and dynamic load and noise should fall, but such gears tend to have lower bending strength and more tooth sliding [15].
Belts
Belts are rated at six or more teeth in mesh on the smaller sprocket. Below that, derate 20 percent per tooth [32], [33]. At three teeth, the factor is 0.40 [34]. Parts on file: GT3 2MGT, 3MGT, 5MGT, 8MGT, 14MGT and HTD 8M. Pitch is on file for all of them. The database also holds a rated torque for 2MGT, 3MGT and 5MGT, but from a source this site cannot show. Take power ratings from the Gates manual cited on each guideline, and apply the width and length correction factors.
On a fixed-centre drive, belt length and sprocket tolerances alone can change the tension threefold or more [35]. Design for it.
| Level | Statement | Check | Source |
|---|---|---|---|
| should | Use a service factor of 1.5 to 2.0 for small pitch synchronous drives, with the upper end for high peaks, high speed or severe conditions. | calc: peak design load = load * service factor (1.5 to 2.0) | [36] |
| must | Make belt rating equal or exceed peak design load. | calc: rating * length factor * width multiplier >= peak design load | [37] |
| must | Belt ratings assume at least six teeth in mesh on the smaller sprocket; with fewer, derate the drive. | calc: teeth in mesh >= 6, else derate | [32] |
| must | Subtract 20 percent of rating for each tooth below six in full contact. | calc: rating * K_TM for fewer than 6 teeth | [33] |
| must | Use special sprocket material and dynamic balancing above belt speed 33.02 m/s. | calc: V = 0.0000524 * PD * RPM <= 33.02 m/s | [38] |
| should | Review with the maker any belt running outside -54 to 85 C. | measure: belt environment temperature between -54 and 85 C | [39] |
| should | On fixed-centre belt drives, control tension tolerance: with fibreglass cord, belt length and sprocket tolerances can change the tension threefold or more. | calc: tolerance stack of belt tension across length and sprocket tolerances | [35] |
| should | Use large sprockets, tight belts, a rigid frame and low-runout sprockets to reach low synchronous drive error. | review: registration drive design list | [40] |
| should | Avoid belt misalignment that forces the belt edge against a flange. | measure: belt tracking and flange contact | [41] |
| guide | Do not assume low backlash at low tension and light load: the GT3 advantage depends on tension, load and sprocket size. | measure: backlash at working tension and load | [6] |
| must | At three teeth in mesh, multiply the belt rating by K_TM = 0.40. | calc: rating * 0.40 at three teeth in mesh | [34] |
Screws
Check buckling on any long screw, assess critical speed over the full travel, and check the lead against the critical value before you claim self-lock [42], [43], [44].
| Family | Part examples | Dynamic rating Ca, kN unless stated | Static rating C0a, kN unless stated | Other |
|---|---|---|---|---|
| Ball screw, HIWIN | 1923 rows: 1521 precision-ground, 314 rolled, 88 heavy-load | 100 / 3395 / 124340 kgf | 178 / 9750 / 538690 kgf | Nominal diameter 6 / 40 / 125 mm, lead 2 / 10 / 64 mm |
| Lead screw | Lead nuts LSN8x2 and LSN8x8 | none | none | Diameter 8 mm, lead 2 and 8 mm |
Ranges are minimum / median / maximum from the screw category. The catalog also lists ball screws from THK, roller screws from SKF and Rollvis, and an igus drylin lead screw, by name only, without their ratings. Many contact lines give the roller screw its high load rating in a short nut.
| Level | Statement | Check | Source |
|---|---|---|---|
| guide | Assume sliding leadscrew system efficiency around 30 to 50 percent. | calc: motor power check using the right screw efficiency | [7] |
| guide | Assume about 95 percent efficiency for a recirculating ball screw. | calc: eta = 0.95 for early sizing | [8] |
| should | A screw is not backdrivable when its lead is below the critical value pi mu D / cos(alpha). | calc: lead <= pi * mu * D_pitch / cos(alpha) for self-locking | [42] |
| should | Assess position-dependent screw critical speed over the full travel; shaft bending natural frequency falls with length squared. | calc: rpm < critical speed at max unsupported length | [43] |
| must | Check long screws in compression for Euler buckling. | calc: P_cr > max compressive screw force | [44] |
| must | Use pulley guards that extend to cable tangent points. | review: each cable pulley has a guard to the tangent points | [45] |
Sources
Each guideline in this chapter with its source, the date it was read and the exact words of the source. Level: must means breaking it causes failure or harm, should is good practice, guide is a rule of thumb.
- Multiply gear output torque by efficiency: tau_gear = eta G tau_motor, with eta at or below 1. 1 Level: should. hades.mech.northwestern.edu 2026-11-08
- Meet torque margin on both sides of gearboxes and lever arms, since reducers are not fully efficient. Level: must. standards.nasa.gov 2026-11-08“wave gear drives, lever arms) shall meet torque margin requirements at both the input and output”
- Compute the rotor's reflected inertia at the joint as G squared times rotor inertia. Level: guide. hades.mech.northwestern.edu 2026-11-08
- Typical hysteresis loss is 3 arc min at ratio 30; use table 44 for the applicable size and ratio. 3 arc min Level: should. harmonicdrive.net 2026-11-08
- Apply a compensation factor when strain wave gear load is below rated torque, because published efficiency is at rated torque. Level: must. harmonicdrive.net 2026-11-08
- Do not assume low backlash at low tension and light load: the GT3 advantage depends on tension, load and sprocket size. Level: guide. maxiparts.com.au 2026-11-08
- Assume sliding leadscrew system efficiency around 30 to 50 percent. Level: guide. pergatory.mit.edu 2026-11-08
- Assume about 95 percent efficiency for a recirculating ball screw. 95 percent Level: guide. pergatory.mit.edu 2026-11-08
- Expect gearhead friction to often rise with ratio. Level: guide. hades.mech.northwestern.edu 2026-11-08
- Account for the spreading force of the pressure angle on gear centres; often overlooked, it deforms the structure so teeth skip. Level: must. pergatory.mit.edu 2026-11-08
- Do not carry electrical current through gears. Level: must. standards.nasa.gov 2026-11-08“[DDMR 21] Gears shall not be used to carry electrical current.”
- Check gear trains for tooth pitting, brinelling, bending fatigue and strength at nominal and peak load, plus backlash, temperature and misalignment, throughout service life. Level: must. standards.nasa.gov 2026-11-08“a. Tooth pitting, brinelling, bending fatigue, and bending strength under nominal and peak”
- Include impact tooth loads from maximum combined axial, radial and moment loads in gear analysis. Level: must. standards.nasa.gov 2026-11-08“b. Impact tooth loads from maximum combined axial, radial, and moment loads sustained”
- Standard low contact ratio spur gears near 1.3 raise dynamic loads. 1.3 Level: guide. ntrs.nasa.gov 2026-11-08“contact ratio of approximately 1 . 3 . This low contact ratio causes increased dynamic”
- Contact ratio above 2 shares load between two or three teeth and should lower dynamic load and noise, but such gears tend to have lower bending strength and more tooth sliding. 2 Level: guide. ntrs.nasa.gov 2026-11-08“High-contact-ratio gears (contact ratio greater than 2) have load sharing between”
- Use strain wave gear rated torque as allowable continuous load at 2000 rpm input. 2000 rpm Level: must. harmonicdrive.net 2026-11-08
- Calculate average torque when load and speed vary, and keep it below the table limit. Level: must. harmonicdrive.net 2026-11-08
- Keep acceleration and deceleration torque below repeated peak torque. Level: must. harmonicdrive.net 2026-11-08
- Do not let momentary peak torque occur during normal operating cycles; allow it only in a collision or an emergency stop. Level: must. harmonicdrive.net 2026-11-08
- Limit momentary peak events to N = 1.0e4 / (2 (n/60) t), with n input speed before collision and t collision time; beyond this the flexspline may fatigue. Level: must. harmonicdrive.net 2026-11-08
- Under normal operating conditions, estimate wave generator life by Lh = Ln (Tr/Tav)^3 (Nr/Nav). Level: should. harmonicdrive.net 2026-11-08
- Use rated L10 life of 7,000 hours for CSF and 10,000 hours for CSG. 7000 h Level: should. harmonicdrive.net 2026-11-08
- Check output bearing life separately for housed units: the gear size procedure does not cover it. Level: should. harmonicdrive.net 2026-11-08
- Model strain wave torsional stiffness as three segments K1, K2, K3 across T1 and T2, not as one constant. 3 Level: should. harmonicdrive.net 2026-11-08
- Do not rely on backdriving torque to hold load; fit a failsafe brake. Level: should. harmonicdrive.net 2026-11-08
- Stay under ratcheting torque; ratcheting can leave the flexspline non-concentric and shorten life. Level: must. harmonicdrive.net 2026-11-08
- Note that an Oldham coupling adds small backlash even with a zero backlash gear element; a direct drive version avoids it. Level: should. harmonicdrive.net 2026-11-08
- Keep strain wave gear oil at or below 90 C, since it quickly loses lubricating ability above this. 90 C Level: must. harmonicdrive.net 2026-11-08
- Read grease temperature limits as grease temperature during operation, not ambient. Level: should. harmonicdrive.net 2026-11-08
- Quasi-direct drive uses a single-stage reduction with a ratio under 10:1. It raises torque at the expense of control bandwidth but keeps the motor backdrivable. Level: guide. arxiv.org 2026-11-08
- For the T-motor U12, an internal single-stage planetary gearbox is infeasible above 7:1, and an external one is the better option from 7:1 to 11:1. 7 Level: guide. arxiv.org 2026-11-08“7:1, ISSPG becomes infeasible, making ESSPG the better option ratios, and spring stiffness while overlooking gearbox design. in the 7:1 to 11:1 range.”
- Belt ratings are based on 6 or more teeth in mesh on the smaller sprocket; with fewer than 6, the drive must be derated. 6 Level: must. maxiparts.com.au 2026-11-08
- Subtract 20 percent of rating for each tooth below six in full contact. 20 percent per tooth Level: must. gates.com 2026-11-08
- At three teeth in mesh, multiply the belt rating by K_TM = 0.40. 0.4 Level: must. gates.com 2026-11-08
- On fixed-centre belt drives, control tension tolerance: with fibreglass cord it can vary by a factor of 3 or more across length and sprocket tolerances. 3 Level: should. gates.com 2026-11-08
- Use a service factor of 1.5 to 2.0 for small pitch synchronous drives, with the upper end for high peaks, high speed or severe conditions. 1.5 Level: should. maxiparts.com.au 2026-11-08
- Make belt rating equal or exceed peak design load. Level: must. maxiparts.com.au 2026-11-08
- Use special sprocket material and dynamic balancing above belt speed 33.02 m/s. 33.02 m/s Level: must. maxiparts.com.au 2026-11-08
- Review with the maker any belt running outside -54 to 85 C. 85 C Level: should. gates.com 2026-11-08
- Use large sprockets, tight belts, a rigid frame and low-runout sprockets to reach low synchronous drive error. Level: should. gates.com 2026-11-08
- Avoid belt misalignment that forces the belt edge against a flange. Level: should. gates.com 2026-11-08
- A screw is not backdrivable when its lead is below the critical value pi mu D / cos(alpha). Level: should. pergatory.mit.edu 2026-11-08
- Assess position-dependent screw critical speed over the full travel; shaft bending natural frequency falls with length squared. Level: should. pergatory.mit.edu 2026-11-08
- Check long screws in compression for Euler buckling. Level: must. pergatory.mit.edu 2026-11-08
- Use pulley guards that extend to cable tangent points. Level: must. standards.nasa.gov 2026-11-08“[DDMR 60] All pulleys shall use pulley guards that extend to the tangency points of the”