Applications
What to look at when picking a servo for each kind of job. Torque and speed are the easy part — the figures that actually decide service life are duty cycle, holding load and gear material.
RC models and UAV
Aerodynamic loads are cyclic and often reversing. What kills a servo here is backlash and gear wear, not peak torque.
- Fixed-wing control surfaces — aileron, elevator, rudder
- Helicopter swashplate linkages
- RC car steering
- Rudder servos on model boats
What to check
Size the torque from hinge moment, then double it
Control surface hinge moment scales with airspeed squared and surface area. A servo running near stall torque overheats and drifts. Leave at least 100% headroom over the calculated load.
Speed matters more than you think for stabilised flight
A flight controller correcting at 50 Hz needs the servo to settle faster than the correction interval. Below 0.10 s/60° is the practical floor for a stabilised airframe; 0.08 s/60° or better for aerobatics.
Metal gears are not optional on anything that lands hard
A nylon gear train strips on the first hard landing or ground strike. Metal gears cost more but survive the shock loads that a belly landing puts through the output shaft.
Check the voltage your receiver actually delivers
A 2S LiPo sags under load. A servo rated 3.6-10 V runs directly off 2S without a regulator, but the torque figure you get is the one at the sagged voltage, not the nominal one.
Robotics
Joints hold position against gravity for long stretches. Holding torque and thermal behaviour under continuous load matter more than peak figures.
- Desktop robot arm joints
- Small quadruped hip and knee joints
- Finger joints in dexterous hands
- Pan/tilt heads for sensors and cameras
What to check
Holding a load is harder than moving it
A joint that holds an outstretched arm draws current continuously with no airflow over the case. Derate the stall torque figure by half for anything that holds position for minutes at a time.
Weight compounds down the kinematic chain
A servo at the wrist is dead weight that every joint upstream has to lift and hold. Saving 10 g at the end effector often saves more than 10 g of torque budget at the shoulder.
Bus control pays off past three or four joints
PWM needs one signal line per servo and gives no feedback. A serial, CAN or RS485 bus daisy-chains on one pair and reads back position, load and temperature — which is what you need to detect a stalled joint before it cooks.
Backlash accumulates
Each joint contributes its own gear play to end-effector error. Tolerable slop in a single servo becomes visible wobble when five of them stack up in series.
Industrial automation
Duty cycle is the constraint. A machine that indexes every four seconds, sixteen hours a day, puts millions of cycles through the gear train per year.
- Conveyor diverters and sorting gates
- Valve and damper actuation
- Positioning stages on test fixtures
- Pick-and-place end effectors
What to check
Work out the cycle count before anything else
One actuation every 4 s over a 16-hour shift is 14,400 cycles a day — around 3.7 million a year. That number, not the torque figure, decides whether a hobby-grade servo is appropriate at all.
Bus feedback is how you catch a fault before it stops the line
Reading back position and load lets the controller detect a jam, a slipping gear or a rising temperature and flag it during a planned stop instead of mid-shift.
Electrical noise is worse than you expect near VFDs
RS485 uses a differential pair and tolerates the noise a motor drive puts on a cabinet. Single-ended PWM over a long unshielded run in the same trunking will glitch.
Plan the failure mode
Decide what the mechanism should do when the servo loses power or signal mid-travel. If an unpowered joint back-driving is unacceptable, you need a mechanical detent or a self-locking gear ratio — not a firmware setting.
No model listed for this application yet. Tell us the requirement and we will say whether an existing model fits or a custom build is needed.
Education and research
Servos in a teaching lab get miswired, stalled and driven past their limits. What matters is surviving mistakes and being documented well enough to debug.
- STEM robotics kits
- University control-theory lab rigs
- Maker-space prototypes
- Competition robots
What to check
Documentation is the actual deliverable
A student debugging a control loop needs the register map and packet format, not a marketing sheet. Every protocol document and control table on this site downloads without registration for exactly this reason.
Expect stalls, so care about overload behaviour
A miswired rig will drive a joint into its end stop and hold it there. What happens next — current limit, thermal cutback, or a burnt winding — is worth checking before you buy thirty of them.
Buy one and test it first
There is no minimum order here. Get a single piece, run it against your actual control stack, and only then decide on the class-set quantity.
Standard interfaces save teaching time
PWM works from any Arduino or Raspberry Pi with no library beyond the built-in one. Bus protocols teach more but cost a session or two of setup — decide which one the course is actually about.
No model listed for this application yet. Tell us the requirement and we will say whether an existing model fits or a custom build is needed.
Not sure which figures matter for your case?
Send the load, the travel, the duty cycle and a target price. That is enough for us to name two or three models and send the datasheets — usually within a business day.