Servo Torque Calculator
The rating is stall torque, about five times what you can design to — and static torque grows as arm length while dynamic grows as its square.
How the two terms scale with arm length
Demand against arm angle
The number on the box is about five times what you can design to
A rating is a moment, not a capacity: "10 kg·cm" means ten kilograms at one centimetre, one kilogram at ten, or a third of a kilogram at thirty. It's a hyperbola, and arm length is exactly the term people leave out when comparing two servos on price.
And the published figure is stall torque at the top voltage. Stall is where the motor has stopped turning and is drawing everything it can — survivable for seconds, not minutes. Anything that holds a position wants 20–25% of it, so a "10 kg·cm" servo is really a 2.5 kg·cm servo — and 2.0 at 4.8 V rather than 6.
Demand varies as the cosine of the angle from horizontal, so a level arm is the worst case and everything above it is cheaper. A bench test at 45° understates the requirement by 29% and one at 60° by half — which is why a mechanism that seems fine folds the first time it swings out flat.
The one that catches people out: static torque grows as L, dynamic torque grows as L². Doubling the arm doubles the holding load and quadruples the acceleration load, because inertia goes as the square of the radius. For a 90° slew the two cross over at about 6 cm of arm at 0.2 s, 39 cm at 0.5 s, and 156 cm at 1 s — the crossover moves as the square of the move time. If you're anywhere near it, slowing the move down is a far cheaper fix than a bigger servo: halving the time quadruples the demand, and doubling it quarters it.
- The arm's own mass acts at its centre, so it costs half what the same mass would at the tip. Minor on a short arm, real past about 20 cm — and the reason a longer arm in a lighter material often beats a shorter one in steel.
- Budget the supply for stall even though you'll never design to it. A servo pulls stall current every time it moves quickly or fights a jam, and a supply that sags resets the microcontroller sharing it. That reads as a random glitch rather than a power problem.
- A large capacitor across the servo supply fixes more mysterious behaviour than any code change. It's the first thing to try when a servo project misbehaves intermittently.
- Micro servos strip their gears before the motor stalls. The failure is silent and total, so anything that might take a knock wants metal gears regardless of the torque arithmetic.
How to use
- Design to about a quarter of the stall figure, not to the rating.
- Size for a level arm, since that is the worst case.
- Add the slew term — it grows as the square of the arm length.
- Budget the power supply for stall current even so.
Frequently asked questions
What does a servo torque rating in kg-cm mean?
It is a moment rather than a capacity: ten kilogram-centimetres means ten kilograms held at one centimetre from the shaft, one kilogram at ten centimetres, or a third of a kilogram at thirty. It is a hyperbola, not a single load figure, and arm length is exactly the term people leave out when they compare two servos on price alone.
How much of a servo rated torque can I actually use?
Around twenty to twenty-five per cent for anything that holds a position. The published figure is stall torque, meaning the motor has stopped turning and is drawing everything it can, which it survives for a few seconds before overheating. So a servo sold as ten kilogram-centimetres is really about two and a half for a static arm, and two at 4.8 volts rather than six.
Does supply voltage change servo torque?
Substantially, and roughly in proportion over the usable range. A servo rated at six volts produces about eighty per cent of that torque at 4.8, and rather more at 7.4 if it is rated for it. That is a large enough swing that a project working from four AA cells behaves quite differently as they discharge, which is a common source of intermittent stalling.
At what angle does a servo arm need the most torque?
Level. Demand varies as the cosine of the angle from horizontal, so a level arm is the worst case, forty-five degrees needs seventy-one per cent, sixty degrees needs half, and vertical needs nothing at all. A bench test at an angle understates the requirement, which is why a mechanism that seems fine folds the first time it swings out flat.
Why does a longer servo arm need so much more torque?
Because two different terms grow at different rates. Holding torque grows in proportion to the arm length, but the torque to accelerate it grows as the square, since rotational inertia goes as the square of the radius. Doubling the arm doubles the holding load and quadruples the acceleration load, which is why a design that scales up gently on paper often fails in practice.
When does inertia matter more than weight for a servo?
At a length that depends sharply on how fast you move. For a ninety-degree slew the two are equal at about six centimetres of arm if the move takes a fifth of a second, thirty-nine centimetres at half a second, and a metre and a half at a full second. The crossover moves as the square of the move time, so fast movements make inertia the design case.
How can I reduce the torque a servo needs?
Slow the movement down first, because it is free and the effect is quadratic — doubling the move time quarters the acceleration demand while leaving the holding demand untouched. After that, shorten the arm, lighten the payload, or counterbalance it. Buying a bigger servo is the expensive option and often the last one worth trying.
Does the arm own weight matter in servo sizing?
It acts at the centre of the arm rather than at the tip, so it costs half what the same mass would at the end. That makes it a minor term on a short arm and a real one past about twenty centimetres, where it can be a fifth of the static load. It is also why a longer arm in a light material often beats a shorter one in steel.
Why does my microcontroller reset when the servo moves?
The servo is pulling something close to its stall current during acceleration and dragging the shared supply down with it. It reads as a random glitch rather than as a power problem, which is what makes it hard to find. Budget the supply for stall current even though you will never design the mechanism to it, and put a large capacitor across the servo supply.
Are metal gear servos worth the extra cost?
For anything that might take a knock, yes — and the reason is not torque. Plastic gear trains strip before the motor ever stalls, and the failure is silent and total: the servo still buzzes and no longer moves anything. Metal gears shift the failure back to the motor and the electronics, which at least give some warning by getting hot.
What is the difference between stall torque and continuous torque?
Stall is the maximum the motor can produce with the output held still, drawing peak current and converting nearly all of it to heat. Continuous is what it can sustain without cooking, which is a small fraction of that. Hobby servos are almost always specified at stall because it is the larger number, and datasheets rarely give a continuous figure at all.
How do I work out the torque for a robot arm joint?
Sum the moment of every mass about that joint, each at its own distance, with the arm level. That means the payload at full reach, every link beyond the joint at its own centre, and every servo further out along the arm counted as a mass too. Then add the acceleration term for the whole assembly, which for a multi-link arm is usually larger than people expect.
🔒 This tool runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.