A4988 Stepper Driver

What it is

An A4988 carrier is a postage-stamp board that sits between a microcontroller and a bipolar stepper motor. You give it two signals — a pulse and a direction — and it handles everything else: energising the coils in the right sequence, and, crucially, regulating how much current goes into them.

That second job is the one that matters and the one people skip. A stepper winding is a low-resistance inductor, typically 2 to 4 ohms. Connect 24 V across 2.8 Ω and Ohm’s law gives 8.6 A, which will destroy a 1.7 A motor in seconds. The A4988 avoids that by chopping the supply on and off tens of thousands of times a second, holding the average current at whatever you set.

So the board has exactly one adjustment: a trim pot that sets the current limit. Get it right and the driver is invisible. Get it wrong and you have either a weak motor that loses steps or a hot one that cooks.

Pinout

Two pins account for most “it does nothing” reports: ENABLE left floating or high (it is active low, so it must be pulled down), and RESET left floating (it must be high, which is why every wiring guide tells you to jumper RESET to SLEEP).

Wiring to an Arduino

Arduino 5V  ──── VDD
Arduino GND ──── GND (logic)  ──┬── GND (motor)
                                └── PSU −
PSU + (12–24 V) ──── VMOT ──┬── 100 µF electrolytic ──┘
                            
Pin 3  ──── STEP            RESET ──┬── SLEEP
Pin 4  ──── DIR                     
Pin 5  ──── ENABLE          MS1, MS2, MS3 ──── 5 V  (1/16 stepping)

Motor coil A ──── 1A, 1B
Motor coil B ──── 2A, 2B

Non-negotiables:

• 100 µF electrolytic across VMOT and GND, physically close to the board. This is the single most common cause of a board that dies on power-up.
• Common ground between the Arduino and the motor supply. Without it the STEP signal has no reference and the driver sees noise.
• Never unplug the motor while powered. An interrupted energised coil produces a spike with nowhere to go.
• Coil pairing matters; polarity within a pair does not. If the motor buzzes without turning, the pairs are wrong. If it turns the wrong way, swap two wires of one pair.

Setting the current limit — the numbers

Vref = I_phase × R_sense × 8

Measure Vref between the trim pot’s metal screw and a ground pin, with VMOT powered and the motor disconnected. Aim for 70–85% of the motor’s rated current: the last fifth buys very little torque and a lot of heat.

Read the marking on the two black SMD resistors near the bottom edge before you calculate. Setting 1.2 V on an R050 board asks for 3 A through a motor rated for 1.7 A.

Use a ceramic screwdriver. A metal one that bridges the pot to something else on a live board is a routine way to destroy a driver.

What you are setting, physically, is the peak torque the motor can produce — the T_peak in the sine law that governs everything a stepper does. Too low and the rotor lags further behind for the same load and slips sooner; too high and you get heat you did not need.

Microstepping

All three high is the normal choice. It is worth being clear about why: microstepping does not make the machine more accurate under load — the error a load causes is unchanged by how finely you divide the step. What it does is stop the rotor ringing at every step, which is the difference between a machine that screams and one that hums, and between a surface with visible step ripple and one without.

The cost is pulse rate. At 1/16 on an 80 steps/mm axis, 100 mm/s needs 8 000 pulses per second. An Arduino Uno bit-banging steps runs out somewhere around 10 000–15 000 per second, which is a real ceiling on a lead-screw axis at 400 steps/mm.

Current, heat and the power budget

The A4988’s own dissipation is modest — the pain is thermal density, not total watts. Roughly 0.5 W per phase at 1 A in a package with almost no copper to spread it into. The limits in practice:

Thermal shutdown does not announce itself. The driver stops, the motor stops mid-move, the driver cools and restarts, and what you observe is an axis that “randomly” loses a chunk of position. An axis that fails more often late in a long job, or in a warm room, is thermal until proven otherwise.

On the supply side, remember the chopper behaves as a buck converter: the current drawn from a 24 V supply is well below the phase current, so budget on power rather than adding up phase currents. Each motor at a 1.2 A limit on 2.8 Ω windings dissipates about 8 W, and it does that whenever it is energised — including standing still. See the robot power budget.

Minimal working code

const int STEP_PIN = 3, DIR_PIN = 4, EN_PIN = 5;

void setup() {
  pinMode(STEP_PIN, OUTPUT);
  pinMode(DIR_PIN, OUTPUT);
  pinMode(EN_PIN, OUTPUT);
  digitalWrite(EN_PIN, LOW);        // active low — energise the coils
  digitalWrite(DIR_PIN, HIGH);
}

void loop() {
  for (int i = 0; i < 3200; i++) {  // one revolution at 1/16
    digitalWrite(STEP_PIN, HIGH);
    delayMicroseconds(2);           // datasheet minimum is 1 us
    digitalWrite(STEP_PIN, LOW);
    delayMicroseconds(300);         // this delay is the speed
  }
  delay(1000);
}

Good for a bench test, wrong for a machine: the step rate jumps straight to full, which above the motor’s pull-in rate produces a buzz and no motion. Use AccelStepper or your own motion profile for anything that has to arrive somewhere. You can watch exactly what that unramped command does to the rotor in the stepper simulator.

What current to actually set

The rated current on a motor’s label assumes the cooling conditions of a test rig — bolted to a large aluminium plate, in still air at 25 °C. Your machine is a printed bracket in a warm room.

A working rule, with the reasoning rather than just the number:

The reason the low end costs less than it looks is in the second column. Torque is proportional to current, but heat is proportional to current squared. Dropping from 1.7 A to 1.2 A gives up 29% of the torque and removes half the heat:

1.7 A: P = 2 × 1.7² × 2.8 = 16.2 W per motor
1.2 A: P = 2 × 1.2² × 2.8 =  8.1 W per motor

If the sizing arithmetic says you are using 12% of the available torque, the 29% you gave up was never doing anything.

De-energising, and the pin most builds ignore

ENABLE exists because a stepper burns its full set current standing perfectly still. That is unlike every other motor a beginner has met, and it has consequences:

// Two energised NEMA 17s at 1.2 A dissipate ~16 W doing nothing at all.
digitalWrite(EN_PIN, HIGH);   // active low, so HIGH = coils off

When to use it:

• Between jobs on X and Y. A plotter or laser gantry has no reason to hold position while idle, and 16 W of heat into a printed frame over an hour is enough to matter.
• Never on a Z axis holding weight, unless the mechanism is self-locking. A T8 lead screw will not backdrive, so de-energising is safe. A belt-driven Z will drop the moment you cut the current.
• Never mid-move. The rotor keeps going, the field does not, and you have thrown away the position with no way to know by how much. Stop, then disable.

The catch that surprises people: after de-energising, the rotor settles onto the nearest detent — the magnetic rest position it has with no current at all — which may be up to a full step away from where it was being held. On a belt axis at 80 steps/mm that is 0.2 mm of unannounced movement, so either re-home after every disable, or accept that idle-disable and sub-tenth accuracy do not go together.

SLEEP is the deeper version: it shuts down the internal regulator too, drawing microamps instead of milliamps, but it needs about 1 ms to wake and it resets the driver’s internal step position to the home microstep. That last part is the trap — waking from SLEEP can shift the motor by up to a full step. For almost every hobby machine, ENABLE is the pin you want and SLEEP is the pin you jumper high and forget.

Troubleshooting

A4988, or the alternatives?

The L298N and TB6612FNG rows are there because both datasheets mention steppers, and both will technically turn one. Neither regulates current, so you get a fraction of the torque and a lot of waste heat. They are DC motor drivers; use them for driving wheels.

The TMC drivers are worth the extra money the moment the machine lives somewhere you can hear it. Same job, smoother current waveform, and a printer fitted with them makes fan noise instead of motor noise.

Where it is used

This driver powers both axes of the pen plotter build, and setting it up correctly is a checkpoint on the precision motion roadmap. The full wiring and Vref walkthrough covers the procedure step by step.