NEMA 17 Stepper Motor

What it is

A NEMA 17 is a hybrid stepper motor in a 42 mm square frame. It is the motor in essentially every desktop 3D printer, most small CNC routers, laser engravers, camera sliders and pen plotters — not because it is special, but because the frame size hit a sweet spot and the entire hobby ecosystem standardised on it.

What a stepper offers that a DC gearmotor does not is position without feedback. Tell it to advance 3200 microsteps and, if you have not overloaded it, it has advanced 3200 microsteps — no encoder, no PID loop, no tuning. That is a genuinely large simplification, and it is the reason a 3D printer can cost a hundred pounds and still place a nozzle within a tenth of a millimetre.

The catch is in “if you have not overloaded it”, and it is a sharp catch rather than a gradual one. A stepper asked for more torque than it has does not slow down — it slips and loses position permanently, silently, in jumps of four full steps.

How it works

Inside is a permanent-magnet rotor with 50 teeth, and two coils arranged so that energising them in different combinations points a magnetic field in different directions. The driver rotates that field; the rotor follows it.

The 1.8° figure comes straight out of the tooth count: 50 rotor teeth × 4 field positions per magnetic cycle = 200 stable positions per revolution, so 360 ÷ 200 = 1.8°. Every “1.8° motor” you will meet has this same 50-tooth geometry, which is why one full step is always 90 electrical degrees regardless of the motor’s size or brand.

The rotor is never exactly where the field is — it trails behind by the lag angle, and the torque it produces is proportional to the sine of that angle. That single relationship is the whole of stepper behaviour, and it is worth reading once properly because it explains lost steps, microstepping, resonance and motor sizing all at the same time.

Reading the numbers

The two most misread rows are holding torque and inductance.

Holding torque is a stationary measurement. It is real, it is what stops the axis being pushed out of position at rest, and it is almost irrelevant to a moving machine. Available torque falls as speed rises because the winding fights back — and inductance is how hard it fights. A low-inductance motor of the same torque rating will reach a much higher speed on the same supply, which is why serious machine builders read the inductance and casual ones read the torque.

The practical consequence is one line: more amps buy torque, more volts buy speed. If a machine stalls under load at slow speed, raise the current limit. If it stalls only when you speed it up, raise the supply voltage — turning the current up will not help, because past the corner speed the driver was never reaching the set current anyway.

Current, heat and the power budget

A stepper draws its full set current whenever it is energised, whether or not it is moving. This surprises people coming from DC motors, where a stalled-but-unloaded motor draws almost nothing.

Per motor, at a 1.2 A limit on a 2.8 Ω winding, both phases energised:

P = 2 × I² × R = 2 × 1.2² × 2.8 ≈ 8.1 W

Eight watts, continuously, per motor, at standstill. A three-axis machine with three NEMA 17s is dissipating around 24 W before it does anything. That drives three decisions:

• Supply sizing. The motor supply must carry every axis at once. Note that this is not current limit × motors on the supply side — the chopper is a buck converter, so at 24 V the supply current is well under the phase current. Budget on power, not amps, and leave headroom; see the robot power budget.
• Heat. 60–80 °C on the case is normal. Mount the motor to something metal where you can.
• Disabling. Pulling ENABLE high de-energises the coils and stops the heat, but the axis is then free to be pushed. Fine for X and Y on a plotter between jobs; dangerous for a Z axis that is holding weight up, unless the screw is self-locking.

Wiring it to a driver

Four wires, two coils. Find the pairs with a multimeter on ohms — a pair reads a few ohms, a mismatch reads open circuit — then:

Order within a pair only reverses the direction, so if the axis runs backwards you can swap two wires instead of editing firmware. Order between pairs is the one that matters: get it wrong and the motor buzzes and vibrates without turning, because the field is no longer rotating in a coherent sequence.

Never disconnect a motor while the driver is powered. An energised coil interrupted mid-chop produces a voltage spike with nowhere to go, and that spike is what kills drivers.

Some motors ship with six or eight wires — those are unipolar or dual-bipolar windings. For a bipolar driver like the A4988, use the four wires that form the two full coils and tape off the centre taps; you get more torque that way than wiring them as unipolar.

Minimal working code

Pulses on one pin, direction on another. There is no library in this and none is needed:

const int STEP_PIN = 3, DIR_PIN = 4, EN_PIN = 5;
const float STEPS_PER_MM = 80.0;     // 20T GT2 pulley at 1/16 stepping

void setup() {
  pinMode(STEP_PIN, OUTPUT);
  pinMode(DIR_PIN, OUTPUT);
  pinMode(EN_PIN, OUTPUT);
  digitalWrite(EN_PIN, LOW);          // active low: energise
}

// Blocking, unramped. Fine for a bench test, wrong for a machine.
void moveMillimetres(float mm, unsigned stepDelayUs) {
  digitalWrite(DIR_PIN, mm >= 0 ? HIGH : LOW);
  long steps = lround(fabs(mm) * STEPS_PER_MM);
  for (long i = 0; i < steps; i++) {
    digitalWrite(STEP_PIN, HIGH);
    delayMicroseconds(2);
    digitalWrite(STEP_PIN, LOW);
    delayMicroseconds(stepDelayUs);
  }
}

void loop() {
  moveMillimetres(50, 300);
  delay(500);
  moveMillimetres(-50, 300);
  delay(500);
}

That will run a motor on the bench. It is also the version to move away from immediately, because the step rate jumps from zero to full instantly — above the motor’s pull-in rate that produces a loud buzz and no motion at all. For anything real, add an acceleration profile or hand it to AccelStepper, and use the non-blocking timing pattern so the endstops still get read.

Sizing one: a worked example

Pick the motor from the load, not from a forum post. The arithmetic is short.

The machine: a belt-driven X axis, 20-tooth GT2 pulley (40 mm per revolution), a 300 g carriage, running up to 200 mm/s with 2000 mm/s² of acceleration.

Step 1 — what torque does the motion need? Convert the linear world to the shaft. The pulley’s effective radius is 40 mm ÷ 2π = 6.37 mm:

accelerating the carriage:  F = m·a = 0.3 kg × 2.0 m/s² = 0.6 N
                            T = F × r = 0.6 × 0.00637 = 0.004 N·m
friction (rails, belt, a dragging cable) ≈ 5 N
                            T = 5 × 0.00637 = 0.032 N·m
rotor's own inertia:        α = a ÷ r = 2.0 ÷ 0.00637 = 314 rad/s²
                            T = J·α = 5.4e-6 × 314 = 0.002 N·m
                                                     ─────────────
                                              total ≈ 0.038 N·m

The result is worth sitting with: friction is eight times the accelerating torque. On a light machine the load is almost entirely friction, which means a binding rail costs you far more margin than a heavier carriage would. Measure the friction before buying a bigger motor.

Step 2 — what speed does it need? 200 mm/s ÷ 40 mm/rev × 60 = 300 rpm.

Step 3 — read the torque curve at that speed, not the headline. A 0.44 N·m motor at a 1.2 A limit on 24 V is still flat at 300 rpm, so about 0.31 N·m is available.

Step 4 — check the margin. 0.038 ÷ 0.31 = 12%. In lag-angle terms, sin⁻¹(0.12) = 7° of the 90° available. That is a very comfortable margin, and it should be — the number you compute is always optimistic, because it does not include the bearing that stiffens in winter, the cable chain that snags, or the belt you tensioned harder than you meant to.

The usual “size it at half the available torque” advice lands in the second row, which is the right place to be. Anything past the fourth row produces the most frustrating class of fault there is: a machine that works on the bench, works in a demo, and fails once an hour in real use.

The same axis on a lead screw tells you why the mechanism matters. A T8’s effective radius for this purpose is 8 mm ÷ 2π = 1.27 mm, five times smaller — so the same 5 N of friction is only 0.0064 N·m at the shaft, a 2% margin. The trade is speed: the screw needs five times the step rate for the same travel, which is the ceiling in steps per millimetre.

Six and eight-wire motors

Most NEMA 17s sold today have four wires and are unambiguously bipolar. Older and industrial stock often has six or eight, and the extra wires are taps on the windings:

The six-wire case is the one to get right. Using the centre tap and one end — driving it as a unipolar motor — energises half of each winding and costs you roughly 30% of the torque, for no benefit at all with a bipolar driver. Find the two ends of each winding (the pair with the highest resistance between them; the centre tap reads about half that to either end), use those, and insulate the taps.

For eight-wire motors, series wiring doubles both the inductance and the resistance, which raises torque at low speed and lowers the corner speed. Parallel does the opposite. Series is the right default for a machine that carries load; parallel is for one that has to move quickly.

Troubleshooting

NEMA 17, or the alternatives?

The honest summary is that a stepper is the right answer when you need repeatable position along an axis and can guarantee the torque margin. It is the wrong answer for a drive wheel: wheels slip, slipping is unbounded, and a motor that loses position silently is the last thing you want under a robot that is trying to do odometry.

Where it is used

This motor is the actuator behind the pen plotter build, and it is the part the whole precision motion roadmap is about. You can see how it behaves — lag angle, step loss, microstepping — without buying one, in the stepper simulator.