Selection guide
Stepper Gearbox Encoder vs Servo Gearhead for Laser Aiming
Short answer
Convert the allowed target displacement into an angular budget before choosing hardware: displacement divided by distance, so 0.1 ft at 300 ft allows about 0.33 mrad. Backlash and loaded repeatability then decide the axis: a preloaded worm or strain-wave gearhead with an encoder on the output shaft meets it, microstepping and frame size do not.
On this page
- Convert the target displacement into an angular budget
- Split the budget across the real error sources
- The four candidate axes
- Backlash, not step count, sets the floor
- Size torque and current from the load
- Put the encoder on the output shaft
- Wind and temperature move a tight budget
- The decision
- Questions asked
Size the axis from the permitted displacement at the target, not from step count or frame size. Divide the allowed spot movement by the distance to get the angular budget in radians: 0.1 ft of allowed movement at 300 ft is 0.33 mrad. Backlash and loaded repeatability then decide the hardware — a preloaded or strain-wave transmission with an encoder on the output shaft meets that budget, and neither microstepping nor a bigger motor frame rescues a loose gear train. Stepper with a self-locking worm gearbox versus servo with a gearhead is a holding-and-cost question once the transmission is right, not an accuracy question. Keep the laser unpowered while you prove out the mechanics; when it runs, beam-path control and access follow the laser class under IEC 60825-1 and the safe-use practice of ANSI Z136.1, which set the controls rather than this article.
Convert the target displacement into an angular budget #
For a beam aimed straight at its target, spot movement follows the geometry of the line of sight. Work the requirement backwards before opening a motor catalog:
- Spot displacement = distance × tan(angular error)
- Permitted angular error ≈ permitted displacement ÷ distance, in radians, valid for the small angles involved here
Keep distance and displacement in the same length unit. A calculated example for one requirement: allowed spot movement of 0.1 ft (about 1.2 in) at 300 ft gives 0.1 ÷ 300 = 0.00033 rad = 0.33 mrad, which is 0.019° or 1.15 arc-minutes. That single number is the entire accuracy specification for the axis.
The distance is what makes this unforgiving. The same 300 ft turns a 2° aiming error into 300 × tan(2°) = 10.5 ft of spot movement, so "a couple of degrees off" and "several feet at the target" are the same statement. Write the spec as a displacement and convert it; "fine resolution" is not a requirement.
Split the budget across the real error sources #
Three words get confused in this selection. Command resolution is the smallest movement you can ask for. Repeatability is the spread when the axis returns to the same aim. Accuracy is the difference between where the beam points and where you meant it to point. A stepper drive accepts microstep commands finer than a milliradian while the output flange sits off by several milliradians of backlash — resolution is cheap, accuracy is not.
Allocate the angular budget across everything that can eat it, and keep any one contributor under about half. One worked allocation for the 0.33 mrad case:
| Contributor | Example share | Controlled by |
|---|---|---|
| Feedback quantization | 0.10 mrad | output encoder resolution; 16 bits gives 0.096 mrad per count |
| Lost motion: backlash and wind-up | 0.15 mrad | preloaded worm or strain-wave gearhead |
| Bearing and mount deflection | 0.05 mrad | bearing stiffness, load overhang, frame |
| Thermal drift | 0.03 mrad | shading and symmetric structure |
The shares are a worked example, not a rule — reweight them for your mechanics. The decision they force is structural: with roughly a third of a milliradian to spend, you cannot hand any of it to a gear train with ordinary catalog backlash, whatever motor turns it.
The four candidate axes #
Compare complete axes, not motors. Each row is a motor, transmission and feedback package:
| Axis | Where it wins | What limits it |
|---|---|---|
| Stepper, self-locking worm gearbox, encoder on the output | holds aim unpowered; a small motor suffices once the head is balanced | worm lost motion unless preloaded; very low efficiency raises motor torque; drive must close the loop on the output encoder |
| Stepper or servo, strain-wave gearhead, output encoder | preloaded tooth mesh keeps lost motion near zero in a small package | capacity per size and cost; the lost-motion figure must be read at your torque |
| Servo, precision gearhead, motor feedback plus output encoder | smooth velocity and acceleration control, tuning headroom for dynamic aiming | closed loop does not cancel gearhead backlash; needs dual-loop support in the drive |
| Fixed laser, 45° mirror on the rotation axis, small tilt stage for elevation | moving inertia and gravity torque drop by orders of magnitude | optical layout, beam clearance, cable wrap; travel is what the mechanics allow, not 360° by right |
The stepper-versus-servo split is smaller than it looks on a pointing axis: at low slew speeds either motor holds a setpoint equally well through the same output encoder. The choice becomes decisive when the axis must regulate speed continuously rather than settle on positions, the tradeoff laid out for a speed-regulated pinch wheel.
Backlash, not step count, sets the floor #
Convert every transmission spec into milliradians before comparing it to the budget. One arc-minute is 0.29 mrad. If a gearhead datasheet lists 15 arc-minutes of lost motion, that is 4.4 mrad at the output — thirteen times the example budget — and no motor or drive setting changes it. Check also the torque at which the datasheet quotes the figure; lost motion grows with load.
Two transmission families can stay inside a sub-mrad budget. A worm stage can be spring-loaded into mesh with the wheel, the practice precision telescope drives use to minimize backlash, and preloaded or duplex worm designs pursue the same goal. Strain-wave gearheads hold near-zero lost motion by preloaded tooth engagement — take the number at your torque from the datasheet of the exact size. Everything else belongs on axes with milliradian budgets.
Microstepping does not buy loaded accuracy. One full step of a 1.8° motor is 31 mrad at its own shaft; dividing it into 256 microstep commands asks for 0.12 mrad increments, but between full-step positions the restoring torque is a small fraction of holding torque, so a loaded rotor settles wherever load torque balances it. The same lost-motion arithmetic that picks a ballscrew versus a zero-backlash leadscrew on a milling axis picks your gearhead: the transmission spec, read in the beam's units, either fits the budget or it does not.
Closed-loop control does not remove gearhead backlash either — that is a feedback-location question, covered next.
Size torque and current from the load #
Four relationships do the sizing:
- Gravity torque = mass × 9.81 m/s² × perpendicular distance from the axis to the center of gravity, in newton-meters for kilograms and meters
- Motor torque = output torque ÷ (gear ratio × efficiency)
- Reflected load inertia = load inertia ÷ ratio²
- Acceleration torque = inertia × angular acceleration
Calculated example for a compact head. A 2.0 kg laser and mount with its center of gravity 80 mm from the elevation axis demands 2.0 × 9.81 × 0.080 = 1.57 N·m to hold. Balance the head to a 5 mm residual offset and the demand falls to 2.0 × 9.81 × 0.005 = 0.098 N·m — sixteen times less, for free.
Through a 30:1 worm stage at 40% efficiency (an assumed input for this example; high-ratio worms run at very low efficiency, so read the real figure from the datasheet), the motor sees 1.57 ÷ (30 × 0.40) = 0.13 N·m unbalanced, or 0.098 ÷ 12 = 0.008 N·m balanced. The point-mass inertia of that head is about 2.0 × 0.080² = 0.013 kg·m², reflected to the motor as 0.013 ÷ 900 = 1.4 × 10⁻⁵ kg·m². Inertia matching is not what binds this axis; lost motion is.
Current ratings do not size the axis either. Motor heating runs close to current squared times winding resistance, so a motor holding full current at standstill runs hot by design, and a drive's standstill current-reduction setting cuts the heat and the holding torque together — know where it is set. Heatsinks cool the driver, fans cool the motor, and neither fixes an undersized torque margin or a backlash problem.
NEMA 17 and NEMA 24 name the mounting face, not the torque. Torque follows ampere-turns and stack length, so two motors in the same frame with 0.4 A and 1.5 A windings are not comparable until you read each one's torque-versus-speed curve at your slew speed and confirm both drives use the same current convention. Holding torque at zero speed is not running torque.
Put the encoder on the output shaft #
Feedback belongs where the beam is. An encoder on the motor shaft cannot see backlash, wind-up or compliance between the gearbox output and the laser mount, so the control loop closes on a position the beam never visits. On the output shaft, the encoder's own quantization has to fit its share of the budget: a 12-bit absolute encoder resolves 360° ÷ 4096 = 0.088° per count, which is 1.5 mrad — nearly five times the whole example budget. A 16-bit device resolves 360° ÷ 65,536 = 0.0055° per count, or 0.096 mrad, under a third of it.
Then confirm the drive can use it. A servo needs dual-loop support: motor feedback for commutation, load encoder for position. A stepper needs a closed-loop mode that accepts the output encoder; without it the output encoder is only a monitor and the axis is open-loop to everything the gearbox does. Verify loaded repeatability against an independent angular reference at the final aiming element before calling the axis done.
Wind and temperature move a tight budget #
Outdoor conditions add torque the catalog sizing never sees. Drag = ½ × 1.225 kg/m³ × wind speed² × drag coefficient × projected area. With a 10 m/s wind (about 22 mph), a drag coefficient of 1.2 on 0.02 m² of projected area, the load is 0.5 × 1.225 × 100 × 1.2 × 0.02 = 1.5 N; at the 80 mm lever of the example head that is 0.12 N·m — more than its balanced gravity torque of 0.098 N·m. A cover or a streamlined enclosure is worth more than a frame size up.
Temperature bends the structure through the same budget. Aluminum grows about 23 µm per meter per kelvin, a handbook value for its alloys; if one side of a 150 mm support heats 20 K more than the other, the differential is 23 × 10⁻⁶ × 0.15 × 20 = 69 µm, tilting the head 69 µm ÷ 0.15 m = 0.46 mrad — more than the entire example budget. Shade the machine, build the mount symmetrically so both sides move together, and keep the aiming reference close to the rotation axis. Steppers also lose torque as windings heat, so leave margin between the torque you calculated and the curve you read.
The decision #
- Budget at or above about 2 mrad (0.6 ft of spot movement at 300 ft): build the stepper axis — self-locking worm gearbox, 16-bit output encoder, and the smallest frame that meets the calculated torque on a balanced head. Unpowered holding comes free, which is the whole argument for the worm.
- Budget below about 0.5 mrad, as in the worked example: the transmission is a strain-wave gearhead or a spring-preloaded worm, the encoder is 16-bit on the output, and the motor is whichever type your drive closes the outer loop on. A stepper is equally accurate at pointing speeds; a servo pays off when you need controlled velocity and acceleration between aims.
- Coverage near 360° of azimuth and 90° of elevation on a tight budget: steer with a mirror before you swing the laser. Fire the beam along the rotation axis at a 45° mirror; rotation sweeps azimuth, a small second stage tilts elevation. A reflected beam turns twice the mirror angle, so the mirror's own budget is half the beam's, and the fixed laser removes nearly all gravity torque and inertia. Verify the travel geometrically — aperture, cable wrap and supports decide it.
- If you were about to jump to a NEMA 24 for torque: balance the head first. The example above trades a 16× reduction in torque demand for a counterweight, which beats a frame-size increase every time it is geometrically possible.
Questions asked
Should the encoder go on the motor shaft or the gearbox output?
On the output whenever lost motion matters. A motor-side encoder cannot see gearbox backlash or elastic wind-up, so the loop closes on a shaft position the beam never reaches. An output encoder also shows you wind-up live during acceleration, which is how you confirm the transmission, not the motor, sets loaded repeatability.
How do I convert a gearbox backlash rating in arc-minutes into target displacement?
One arc-minute is 0.29 mrad, so multiply the rating by the distance: 5 arc-minutes at 300 ft moves the spot about 0.44 ft. Convert budget and rating to the same unit before comparing, and note the torque at which the datasheet quotes lost motion, because the number applies at that torque, not necessarily at yours.
Can I rely on a self-locking worm gearbox to hold position with power off?
For a static, balanced load, yes. High-ratio worm stages resist backdriving precisely because their efficiency is very low, so the axis holds aim with the drives off. Shock, vibration and any sustained reversed torque can still creep it, so verify unpowered holding on the real assembly, and never count on it where a person could be exposed: that job belongs to an independent brake or a mechanical stop.
Does microstepping give finer pointing than a servo?
No. Microstepping buys command resolution and smoothness; loaded accuracy comes from where the rotor settles against load torque, and between full-step positions the restoring torque is only a fraction of holding torque. One full step of a 1.8° motor is 31 mrad at its own shaft before any gearbox, so the transmission and the output encoder set the floor, not the microstep count.
Why do two motors with the same NEMA frame and similar current behave so differently?
The NEMA number sizes the mounting face, 17 meaning 1.7 inches, and says nothing about torque, which follows ampere-turns and stack length, so two same-frame motors can differ widely. Compare the torque-versus-speed curves from each datasheet at your slew speed, and set the driver current to the motor's rated current rather than guessing from the winding's amp label.