TL;DR: Choose a stepper when the axis is light, slow, low duty, and a lost step costs you nothing worse than a re-home. Choose a servo when the axis carries inertia, moves fast, must settle accurately, or when the machine has to prove where it is. The deciding factor is not resolution — steppers often win on paper resolution. It is feedback, torque at speed, and whether the machine can detect its own failure. And before you pick either, check whether the move actually has two ends: a pneumatic cylinder from SMC or Festo often beats both motors on clamps, gates and ejectors.
Motionwell Automation designs custom production machines in Singapore, and this question comes up in nearly every concept review. We specify servos on most axes of our special purpose machines, steppers on a small number of auxiliary axes, and pneumatics on everything that only needs “in” and “out”. This article explains the reasoning behind those defaults, whether you are building a machine, buying one, or modernising one you already run.
What Actually Separates a Servo Motor From a Stepper Motor?
Open loop versus closed loop
A hybrid stepper has many magnetic poles. The drive energises phases in sequence, the rotor snaps to each new magnetic position, and a standard 1.8-degree motor gives 200 full steps per revolution. The drive counts pulses out. It does not know whether the rotor arrived. That’s open loop control, and it works remarkably well right up to the moment it doesn’t.
A servo motor carries an encoder on the back of the shaft. Every control cycle, the drive compares commanded position against measured position and injects the current needed to close the gap. Load increases, current increases. Something jams, the position error grows past a threshold, and the drive faults the machine rather than quietly falling behind.
The practical consequence matters more than the physics. A stepper failure is silent. A servo failure is an alarm on the HMI with a fault code and a following-error value you can look up.
How each one makes torque
The stepper holds its rated torque at zero speed and loses it as speed climbs. Winding inductance limits how quickly current can build in each phase, and back-EMF grows with rotor speed, so there is less current available to make torque the faster you spin. Torque falls off steeply past a few hundred rpm, and a typical hybrid stepper has little usable torque left by the time it reaches 1,000 rpm. A higher drive bus voltage pushes the curve out but does not change its shape.
A servo motor holds roughly flat torque from zero to rated speed, then moves into a constant-power region above it. It will also deliver short bursts well above its continuous rating — around three times continuous is a common figure on datasheets — which is exactly what a fast index-and-settle profile needs, since acceleration torque is transient while heating is governed by the RMS torque across the whole cycle.
Why Do Stepper Motors Lose Steps, and What Happens Next?
Three causes cover almost every case in the field:
- Torque demand exceeds torque available at that speed. Someone sized the motor from the holding torque figure on the front page of the datasheet instead of reading the torque-speed curve at the operating point.
- Acceleration is too aggressive. Load inertia demands more torque during ramp than during the constant-velocity phase. A profile that runs fine empty misses steps once a real payload is fitted.
- Resonance. A stepper and its load form a spring-mass system. In the low-speed region the oscillation can be violent enough to slip a pole. Microstepping and mechanical damping help; passing through the band quickly helps more.
What happens next is the part that costs money. The machine has no idea. The axis reports the commanded position, the HMI shows green, and every downstream operation referenced to that axis is now wrong by one or more steps. On an assembly machine with a dispensing head or a camera trigger, that shows up as scrap nobody can explain, days later, with no fault log to point at.
Closed-loop steppers close this hole. An encoder on the rear shaft lets the drive correct accumulated error, alarm on genuine stall, and drop idle current to cut heat. It is a good middle option on light axes. It does not change the torque-speed curve, so it will not rescue an axis that was undersized for speed.
Resolution, Accuracy and Repeatability Are Not the Same Thing
This is where most specifications go wrong, because steppers look excellent on the spec that matters least.
Resolution is the smallest increment you can command. A 1.8-degree stepper at 1/16 microstepping gives 3,200 increments per revolution, which sounds better than it is. Torque follows the sine of the electrical angle error, so a single 1/16 microstep only commands about eleven percent of holding torque. Under friction the rotor may simply not move until several microsteps have accumulated. Microstepping buys smoothness and reduced resonance. It does not buy accuracy.
Accuracy is how close the axis gets to the true target. Stepper datasheets typically quote a non-cumulative step angle error of around ±5% of a full step, and at the machine level that error is usually dwarfed by ballscrew lead error, coupling wind-up, backlash and thermal growth. A 23-bit servo encoder resolves the shaft to a level no mechanical transmission can honour.
Repeatability is the spread when the axis returns to the same target. This is the number your quality engineer actually cares about, and it lives in the mechanics — preload, guide quality, homing method, thermal state — far more than in the motor.
The engineering conclusion: pick the motor for torque, dynamics and feedback. Then get your accuracy from the transmission, the homing strategy, and inline measurement. On the machines where dimensional truth matters, we verify with an external measurement rather than trusting the axis. Our machine vision inspection guide covers that side, and a 2026 copper-plate thickness measurement machine we delivered used a Keyence displacement system for exactly this reason: the axis positions, the sensor judges.
What Does Heat and Duty Cycle Do to the Decision?
A stepper draws close to rated current whenever it is energised, including standstill. A hot stepper case is normal behaviour, not a fault. That constant thermal load matters in three places: inside a sealed enclosure where there is nowhere for the heat to go, in a cleanroom where you are already fighting for airflow budget against the fan filter units, and near any precision fixture or camera where thermal drift moves your datum.
A servo draws current in proportion to load. An unloaded servo at rest draws almost nothing. On a vertical axis holding against gravity it draws real current, which is why we specify a motor brake on vertical axes rather than paying for continuous holding current — and why that brake matters for safety when power drops.
For any axis running continuously, size on RMS torque across the full duty cycle, not on the peak of the move.
What Does Each Option Really Cost to Own?
Hardware cost is the part everyone compares, and it is the part that matters least over a machine’s life. Cost drivers worth putting in the spreadsheet:
| Cost driver | Stepper | Servo |
|---|---|---|
| Motor and drive hardware | Lower | Higher, and it scales with power |
| Feedback cable and connector | None on open loop | Encoder cable, correct routing, correct shielding |
| Commissioning time | Set current and go | Gain tuning; autotune gets most of the way, the last part is manual |
| Sizing margin | Oversized deliberately, because there is no feedback to warn you | Sized to duty, with peak torque available for ramps |
| Diagnostics | None | Torque, following error, and cycle data available to the PLC |
| Spares strategy | Cheap units, easy to stock | Drive plus motor plus parameter backup |
Two hidden costs decide it more often than the hardware price. The first is undetected scrap from silent step loss, which shows up as a quality investigation rather than a maintenance ticket. The second is the value of the data. A servo’s torque signal is a free process signature: force-monitored insertion on our 12-station rotary syringe assembly machine (P22068, 15-second cycle) uses exactly this to log every press and reject out-of-band assemblies. You cannot get that from an open-loop axis.
One more cost that has nothing to do with the motor: plant standardisation. If your factory runs Allen-Bradley, a Kinetix servo axis costs less to own than a cheaper unfamiliar brand, because spares, training and support already exist. We build on Allen-Bradley, Siemens, Mitsubishi, Omron, Beckhoff and Inovance depending on the customer’s existing standard, and we ask which one you run before we start selecting drives.
When to Use a Servo Motor: A Decision Table by Application
| Application | What we specify | Reasoning |
|---|---|---|
| Rotary indexing table, multi-station assembly | Servo | High inertia, short index time, and the settle must be verified before the next station fires |
| High-rate pick and place, variable pitch | Servo (or a SCARA robot) | Path and settling time drive the cycle; we run Yamaha SCARA units on this duty |
| Piston dosing pump, filling head | Servo | Fill volume is a direct function of stroke position; ±0.5% of target volume needs per-head closed-loop calibration |
| Long gantry or robot linear track | Servo, absolute encoder | Removes the homing routine, holds position after power loss, and supports dual-motor gantry sync |
| Press, insertion or crimp station | Servo | Force and position logged together as the process record |
| Fixed-speed conveyor or pump | Induction motor and VFD | No positioning requirement; a PowerFlex-class drive is the cheaper right answer |
| Label unwind, screw feeder, low-duty adjust axis | Stepper or closed-loop stepper | Light, slow, forgiving, and re-homed each cycle |
| Clamp, gate, ejector, two-position stop | Pneumatic cylinder | Two ends, high force, no tuning |
The filling column above comes from delivered work: our GMP filling and sealing machines (P23005, P25026) run one servo per fill head so each nozzle calibrates independently. Details are on the custom filling machines page and in the pharmaceutical filling case study.
Pneumatic vs Electric Actuator: When Does Air Beat Both Motors?
Air wins when the move has two ends and nothing in between. A cylinder gives you very high force from a small, light package, tolerates shock and side load abuse that would wreck a ballscrew, survives washdown, and needs no tuning or parameter backup. On the food-grade tray filling platform, where the enclosure is IP65 and the line gets hosed down, pneumatics do the clamping and transfer work for good reasons. We buy heavily from SMC and Festo across nearly every machine we build.
Air stops paying at four specific points:
- You need intermediate positions. Standard cylinders have two, plus whatever you can achieve with mechanical stops. Beyond that you are into proportional valves or servo-pneumatics, and an electric rod actuator is usually simpler.
- You need controlled velocity. Flow restrictors are open-loop speed control. The setting drifts with temperature, supply pressure, seal wear and lubrication, so the profile you commissioned in January is not the one running in June.
- You need force accuracy or force data. Cylinder force is pressure times area, and your supply pressure moves during the shift. When a process genuinely needs constant force, it needs closed-loop control — our force-controlled robotic grinding cell holds 100 N through an active force head, not through a regulator.
- You need position confirmation for the PLC. Reed switches confirm the ends only. Anything in between is unknown.
There’s also a plant-level cost that rarely reaches the machine specification: compressed air is one of the most expensive utilities to produce per unit of delivered energy, and every leak runs around the clock. That does not overturn a clamp cylinder, but it argues against air on dozens of high-cycle axes.
On most Motionwell machines all three appear on the same frame: servos on the axes that need a profile, a stepper or two where duty is light, air on the clamps, ejectors and gates. Mixing them is sizing each actuator to its job.
How This Plays Out on Retrofit Projects
Control-system modernisation is now the largest single stream of work we run: replacing worn mechanical cams, obsolete DC drives and unsupported controllers on legacy production machines with a current PLC, servo and VFD platform. Our Allen-Bradley spend for this work grew roughly fourfold in a single year across PowerFlex 755 drives, Kinetix servos and CompactLogix and ControlLogix controllers.
The pattern repeats. A machine built around mechanical cams and pneumatics gets a new product variant, and the cam is not adjustable. Replacing the cam with a servo and an electronic cam profile turns a machining job into a recipe change. That is usually the whole business case, ahead of any speed gain. The same job also adds safety circuits, interlocks and guarding to current standards, which is a separate conversation covered in our notes on cobot and machine safety standards.
If you are scoping a new machine rather than a retrofit, the motion architecture is settled during concept design, and changing it later is expensive. Our special purpose machine design guide explains where that decision sits in the project timeline.
Getting the Specification Right
Bring these six numbers to a motion review and the servo-versus-stepper question usually answers itself in the meeting:
- Stroke or travel, and the move distance per cycle
- Move time and settle time you need, plus cycles per hour
- Moving mass, and whether the axis is horizontal or vertical
- Required repeatability at the tool point, not at the motor shaft
- Duty cycle: percentage of time under load, and ambient conditions (cleanroom class, washdown, temperature)
- Your plant’s control platform standard and preferred spares