How this instrument works
Torsional stiffness measured this way skips the material altogether. Instead of starting from a shear modulus, a cross-section, and a length and multiplying them into k = G·J ⁄ L, this instrument starts from a test: apply a known torque T to a real, physical part, measure how far it actually rotates, θ, and divide. The result, k = T ⁄ θ, is the same torque-per-radian quantity, but it comes from the object itself rather than from a drawing of it — which matters for a harmonic-drive gearbox, a bonded rubber engine mount, or a splined coupling, none of which is a plain round bar with a textbook G and J to look up.
The formula is Hooke's law turned sideways: torque plays the role force plays in a linear spring, twist angle plays the role of extension, and their ratio is a spring constant for rotation. A test rig supplies that torque with a calibrated arm and dead weight, a hydraulic rotary actuator, or a torque wrench fitted with a transducer, while the resulting rotation is read off a digital inclinometer, a rotary encoder, or a long pointer sweeping past a protractor. Because the instrument only ever sees T and θ, it never needs to know what the part is made of, how many pieces it is built from, or whether its cross-section even has a name.
One torque-and-angle pair gives one number, and that number is only trustworthy as the part's stiffness if the torque-versus-twist relationship is a straight line through the origin in that region — true for a steel shaft well inside yield, false for a rubber bushing whose stiffness climbs as it is compressed, or for a bolted flange with a soft, slipping phase before its friction grip takes hold. A single reading like the one in the worked example below is really a secant stiffness at that specific load; comparing k from two or more torque levels on the same part is how an engineer checks whether it is behaving linearly at all.
- Enter Applied torque — the known torque you applied during the test, in N·m.
- Enter Measured twist angle — how far the component rotated under that torque; degrees is the default, switch to radians if your instrumentation reads that way.
- Read Measured torsional stiffness, N·m/rad — the torque-per-radian slope implied by that one torque-and-angle pair.
- Repeat the test at a second torque level and compare the two stiffness values: matching results confirm linear, elastic behavior, and a drifting value flags backlash, slip, or a nonlinear part.
Worked example — bench-testing a 15 N·m coupling
A technician clamps one end of a flexible coupling in a fixture, hangs a calibrated weight from an arm on the free end to apply exactly 15 N·m of torque, and reads the resulting rotation off a digital inclinometer: 30°, which is 0.523598775598299 rad. Dividing gives k = T ⁄ θ = 15 ⁄ 0.523598775598299 = 28.6478897565 N·m/rad — the coupling's measured torsional stiffness, a number nobody could have looked up from a materials table, because the coupling is a rubber-and-steel assembly with no single shear modulus to plug into k = G·J ⁄ L.
Read literally, 28.6478897565 N·m/rad says it would take that much torque to twist the coupling a full radian, about 57.3°, if the relationship stayed straight that far — which it will not, since 30° is likely already deep into the rubber element's working range. For comparison, a stiffer coupling that twists only 15° under the same 15 N·m torque computes to 57.2957795131 N·m/rad, exactly double: half the rotation for the same load always means twice the measured stiffness, the direct, inverse read of k = T ⁄ θ.
Questions
What does k = T ⁄ θ actually measure?
It measures the torque needed to twist the tested part by one radian, extrapolated in a straight line from a single measurement. Apply 15 N·m and read 30° (0.5236 rad) of rotation and k works out to 28.65 N·m/rad — not a torque you would ever apply all at once, since 57.3° of twist would likely break or yield the part, but a useful ratio for comparing components or predicting smaller twists under smaller torques.
How is this different from k = GJ ⁄ L?
k = GJ ⁄ L computes stiffness from the shaft's material (shear modulus G), cross-section (polar moment J), and length (L) — it needs a clean round bar and known properties. k = T ⁄ θ, the formula here, needs none of that: it measures stiffness directly from a physical test, which is the only option for a gearbox, a rubber mount, a bolted joint, or any assembly that is not a simple, uniform, isotropic shaft.
Why does the twist angle need to be in radians?
Because a radian is the angle unit that makes k = T ⁄ θ come out directly in newton-metres per radian, the standard SI stiffness unit — arc length divided by radius, a pure ratio with no separate conversion constant hidden inside it. Enter the angle in degrees if that is what your inclinometer shows; the instrument converts to radians internally before dividing, so the displayed torque-per-radian figure is always correct.
Can I trust a stiffness measured from just one torque-and-angle pair?
Only as a secant value at that specific load. A single reading assumes the part's torque-versus-twist line passes straight through the origin, which holds well for a steel shaft below yield but poorly for a rubber bushing or a bolted joint that has not yet seated. Testing at two or three torque levels and checking that k stays roughly constant is how an engineer confirms the part is behaving linearly before trusting any single figure.
What happens if the measured twist angle is zero?
k is undefined — dividing by zero has no answer, and the instrument will reject it, since a torque that produces no measurable rotation cannot be assigned a finite stiffness. In practice a zero reading almost always means the applied torque was too small for the instrumentation's resolution, or that the fixture itself flexed instead of the part, and either way the test needs a larger torque or a finer angular readout before it returns a real number.
Does a higher k mean a stiffer or a more flexible part?
Stiffer. A higher k means more torque is needed to produce the same twist, exactly like a stiffer spring needing more force for the same stretch. The worked example's 28.65 N·m/rad coupling and the comparison figure of 57.30 N·m/rad make the point directly: the second part twists only half as far under an identical 15 N·m load, so its measured stiffness comes out exactly double.