Blog · 2026-08-15 · Units
Marketplace motor listings are frequently optimistic and occasionally fictional. A kitchen scale, a strip of aluminium and twenty minutes will tell you what your motor actually produces — and the method is simple enough that there's no excuse for guessing.
Buy a gearmotor from a marketplace listing and you'll get a torque figure. Sometimes it's the manufacturer's honest stall torque. Sometimes it's the figure for a different ratio in the same family. Sometimes it appears to have been chosen for its aesthetic qualities. Since torque is the specification your whole drivetrain design rests on, and since measuring it needs no specialist equipment, it's worth an afternoon to find out the truth about the motors on your bench.
Torque is force multiplied by distance. Clamp a rigid arm of known length to the motor shaft, let the end of the arm press down on a scale, power the motor so it pushes against the scale, and read the force. Multiply and you have torque.
torque (N·m) = force (N) × arm length (m)
force (N) = scale reading (kg) × 9.81
shortcut: torque (kg·cm) = scale reading (kg) × arm length (cm)
That shortcut is worth noting — because hobby motors are usually specified in kg·cm, and a kitchen scale reads in kilograms, the measurement comes out directly in the listing's own units with no conversion at all. A 10 cm arm reading 1.2 kg means 12 kg·cm. If you need N·m or oz·in, the Torque Unit Converter does the rest.
A digital kitchen scale (1 g resolution, 5 kg range covers most hobby motors). A rigid arm — a strip of aluminium, a steel rule, or a printed lever with a shaft coupler; anything that won't flex under load. A way to clamp the arm to the shaft that won't slip: a shaft coupler, a printed hub with a grub screw on the shaft flat, or a bolted-on wheel hub. A bench power supply, or your battery with an inline ammeter — current measurement is what turns a single reading into a full characterisation. And a means of mounting the motor firmly, because the motor will try very hard to rotate itself instead.
1. Mount the motor rigidly above or beside the scale, with the shaft horizontal and the arm able to press down on the scale pan. Clamp the motor body to something solid — it will apply exactly as much torque to its mount as to the arm.
2. Zero the scale with the arm resting on it, motor unpowered. This subtracts the arm's own weight, which is otherwise a systematic error in every reading.
3. Apply power briefly and read the scale. The motor pushes the arm down; the scale reads the force. Read quickly — a stalled motor heats up fast — and cut power. Two or three seconds is plenty.
4. Record the current at the same moment if you have an ammeter inline. Stall current is the specification your motor driver, fuse and wiring must be sized around, via the Power Budget Calculator, and it's just as commonly misquoted as torque.
5. Repeat three times and average. Repeat at the shaft's other rotational position too — gearbox torque varies slightly through a revolution, and a single reading can catch a favourable spot.
6. Multiply. Scale reading in kg × arm length in cm = torque in kg·cm.
The measurement above is stall torque, and stalling a motor is genuinely stressful for it: full current, no cooling from rotation, all of it becoming heat in the windings. Keep each test under about three seconds, let the motor cool between attempts, and stop immediately if you smell anything. Small motors survive this comfortably; large ones at high voltage can draw tens of amps, so make sure your supply and wiring are up to it — and use the fuse you'd use on the robot.
If you'd rather not stall the motor at all, measure at a partial load instead: hang known weights from the arm at a known radius and find the heaviest weight the motor can still lift. That gives you a point on the curve below stall, which is arguably more useful anyway since it's closer to real operating conditions.
Stall torque is one end of a line, and the full torque-speed curve tells you far more (the theory is in Torque vs Power vs Speed). To get it, you need torque and speed simultaneously, which means loading the motor while it turns.
The simple approach: a friction brake. Wrap a strip of leather or webbing around a pulley on the motor shaft with a spring scale on each end — the difference in the two readings, times the pulley radius, is the torque. Tighten the wrap to increase load. Measure shaft speed at the same time with an encoder or a cheap optical tachometer, and record current from your ammeter. A handful of points from lightly loaded to nearly stalled gives you the line.
Plot torque on one axis and speed on the other; it should be close to straight. Compute power at each point (torque × RPM ÷ 9.55 for N·m) and you'll see it peak in the middle, at roughly half stall torque and half no-load speed — which is a satisfying confirmation that the theory describes your actual motor.
Three patterns come up repeatedly. Listings quote the family's best figure: a seller offering one gearmotor in eight ratios sometimes copies the highest-ratio torque into every listing, so a low-ratio motor tests far below its claim. Torque is quoted at a voltage you're not using: torque scales roughly with voltage, so a 12 V spec run at 7.4 V delivers around 60% of the figure — not a lie, but not what you'll get. Efficiency is worse than assumed: comparing measured output power against measured electrical input (volts × amps) often reveals a gearbox eating more than expected, especially in cheap plastic-geared units.
All three are arguments for measuring the motors you'll actually use, at the voltage you'll actually use, before designing a drivetrain around a number from a product page. Feed the measured figure into the Motor Sizing Calculator rather than the claimed one, and the sizing margin you think you have will be the margin you really have.
Yes, and it's just as revealing. Clamp a horn-mounted arm over a scale, command the servo to push down, and read the force. Do it at the voltage you'll run — servo torque at 6 V is meaningfully higher than at 5 V. The arm geometry and joint loads for a real arm come from the Servo Torque Calculator.
Within a few percent, which is far better than the listings it's checking. The main error sources are arm length measurement, arm flex, and the scale reading off-centre — all controllable with a rigid arm and a careful measurement to the contact point.
Reverse the motor's polarity, or flip the arm to the other side of the shaft. The magnitude is the same either way; brushed DC motors produce essentially equal torque in both directions.
The measurement is worth more if you write it down somewhere permanent. A simple table — motor model, supplier, test voltage, measured stall torque, measured stall current, no-load current, date — turns a one-off afternoon into a reference you'll consult for years. Buying a second batch of "the same" motors and finding they test 30% differently is common enough that having the original numbers is genuinely valuable.
Two extra columns repay themselves. Measured no-load current is an excellent health indicator: it rises as a gearbox wears or dries out, so a motor that used to draw 90 mA free-running and now draws 200 mA is telling you something before it fails. And the date lets you re-test an old motor against its own younger self, which is the only honest way to know whether a robot's declining performance is the motors, the battery, or the mechanism binding.
Tape a label with the key figures to each motor, too. Six months later, when three visually identical gearmotors are sitting in a drawer at three different ratios, the version with a label on it is the one that ends up in the robot.
An arm, a scale, and a multiplication. Convert whatever comes out into the units your project uses with the Torque Unit Converter, and read the measurement in context with Torque vs Power vs Speed.