Why stall torque specifications are routinely misused

Why is the stall torque on a data sheet almost never a usable design number, and what should be used in its place?

Stall torque is the torque a motor produces at zero speed with rated voltage applied. At that condition there is no back-EMF, so current is limited only by the terminal resistance, and the shaft is not moving, so the mechanical output power is exactly zero. Every watt drawn from the supply is going into heat.

That makes stall torque a useful number for exactly two purposes: predicting the current a jam will draw, and bounding the torque a mechanism might see if it stops. It is used, instead, for sizing. "Stall torque is 42 mN·m and I need 8, so I have five times margin" is the most common sizing error in compact motion, and it is wrong in three separate ways.

Reason one: the motor cannot hold it

The limit on continuous torque is thermal, not magnetic. Current heats the winding as the square of current, and the winding has an insulation temperature limit. The torque a motor can produce forever is the torque whose current produces a steady-state temperature rise the insulation survives, in the actual ambient temperature, with the actual mounting and airflow.

For the illustrative machine used across these articles, the continuous limit is a small fraction of stall. That is not unusual. Continuous torque in the range of ten to twenty percent of stall is ordinary for small brushed motors, and it depends on mounting: a motor bolted to a metal chassis will hold more than the same motor in a plastic housing with still air around it.

Reason two: stall torque falls as the motor heats

The data sheet value is a cold value. Two effects reduce it as the machine warms, and they compound.

R(T) = R25 · [1 + 0.00393 · (T − 25)]

R(T)
winding resistance at temperature T, Ω
R25
winding resistance at 25 °C, Ω
0.00393
temperature coefficient of resistance of copper, per K
A 100 K rise raises copper resistance by about 39 percent, which drops stall current, and therefore stall torque, to about 72 percent of its cold value.

The magnet contributes the second effect. Remanence falls with temperature at roughly 0.11 to 0.12 percent per kelvin for common neodymium grades and roughly twice that for ferrite, and the torque constant falls with it. A 100 K rise therefore costs a further ten percent or so on a neodymium machine. Taken together the hot stall torque can be near sixty percent of the value printed on the sheet, and the value on the sheet is the one people divide by to get their margin.

Reason three: something else is the fuse

In a gearmotor, motor stall torque multiplied by the reduction ratio usually exceeds the gearhead's torque rating by a wide margin. A motor that stalls at 42 mN·m behind a 100:1 reduction presents several newton meters at the output, and a small gearhead is not built for that. The first thing to fail in a jam is frequently a gear tooth, an output bearing or a press fit, not the motor.

  • Check the gearhead momentary torque rating against motor stall torque times ratio, not against the application torque.
  • Check the shaft and the coupling. A stalled motor applies its stall torque through the shaft into whatever the mechanism has hit.
  • Check the drive. If the controller current limit is below stall current, the drive rather than the motor sets the real jam torque, and the calculation should use the limited current.
  • Decide deliberately which element is intended to be the weak point, and state it in the specification.

What to use instead

071421283542025507510002468Shaft torque (mN·m)Efficiency (%)Output power (W)efficiencyoutput powerpeak efficiency, 4.6 mN·mpeak power, half of stallcontinuous limitstall
Efficiency and output power for the illustrative machine, computed from its four constants. Peak efficiency sits near 11 percent of stall torque. Peak power sits at exactly half of stall, where efficiency has already fallen to about 48 percent and the motor is drawing more than a thousand milliamps.

The peak efficiency point is a genuinely useful design target, and it has a closed form. Maximizing output over input for the linear machine gives a current of the geometric mean of no-load and stall current, and an efficiency that depends only on their ratio.

Imax η = √(I0 · IS), ηmax = (1 − √(I0 / IS))²

I0
no-load current, A
IS
stall current, A
ηmax
peak efficiency of the ideal linear machine
For the illustrative machine, 40 mA and 2667 mA give a peak-efficiency current of 327 mA and a peak efficiency of 77 percent. The formula is also a quick sanity check on a data sheet: a motor whose no-load current is a large fraction of its stall current cannot be efficient, whatever else it claims.
The four torque figures, and what each one is for
Figure What it is Valid use Invalid use
Stall torqueTorque at zero speed, cold, at rated voltagePredicting jam current and bounding mechanism loadsSizing. Dividing it by a factor to get a working torque
Torque at peak powerHalf of stall torque, by constructionUnderstanding where the machine stops gaining outputA continuous operating point. Efficiency there is about 50 percent
Torque at peak efficiencyThe geometric mean current point, roughly a tenth of stallChoosing a running point for a continuously operating mechanismA peak or acceleration rating
Continuous torqueThe torque whose current produces an acceptable steady temperatureSizing, once ambient and mounting are statedQuoting without stating the ambient temperature and mounting it assumes

What to ask for, and what to measure

  1. Ask for continuous torque with its conditions

    Ambient temperature, mounting, and the insulation limit it is referenced to. Without those three, a continuous torque figure is not comparable between suppliers.

  2. Measure current in the running mechanism

    Current converts to torque through the torque constant, so one meter reading places the mechanism on the curve above and settles most sizing arguments.

  3. Measure the temperature rise at the real duty cycle

    A thermocouple on the housing plus a resistance measurement of the winding before and after a run gives the winding rise directly, since copper resistance is a thermometer.

  4. Decide and test the jam case

    If the mechanism can stall, either limit the current in the drive, or verify by test what fails first and how long it takes.

Winding resistance as a thermometer is worth using. Measure the cold resistance, run the duty cycle, stop and measure again quickly, and the ratio gives the average winding temperature through the coefficient above. It is the most accessible measurement of the quantity that actually limits the motor.

How the number is produced, and why sources disagree

There are two ways a stall torque figure reaches a data sheet. It can be measured, by locking the shaft against a torque transducer and applying rated voltage. It can also be computed, as the torque constant multiplied by stall current, where stall current itself is computed from voltage over terminal resistance. The two do not agree, and neither is wrong.

A measured value depends on how long the measurement window is. Current begins heating the winding immediately, resistance climbs, and the reading falls while it is being taken. Read at 100 milliseconds and read at 5 seconds are different numbers from the same motor. A computed value avoids that by describing a motor that is always at 25 °C, which no stalled motor is.

The practical consequences are worth stating. Comparing a measured stall torque from one supplier against a computed one from another is not a comparison of motors. And a published stall torque, however it was produced, is a room-temperature figure that the machine holds for a fraction of a second, which is the point of this article.

Who wrote it, who checked it, and what has been corrected since.

Every article on this site carries this block. A correction is recorded here with its date rather than edited silently into the text.

Written by
Micro Motion application engineering
Reviewed by
Micro Motion manufacturing engineering
First published
Last reviewed
Reading time
7 minutes

Corrections

No corrections have been issued for this article. When a figure, a number or a statement here is found to be wrong, the correction is recorded in this block with its date rather than edited silently into the text.

Report an error in this article