Torque Sensor

What is a Torque Sensor

A torque sensor is a measurement element that converts the torque, that is the twisting moment, transmitted through a shaft or coupling into a proportional electrical signal. Torque is a vector quantity equal to force times the perpendicular lever arm, expressed in the SI unit newton metre (N·m). The instrument lets engineers measure a quantity that cannot be read directly: how hard a motor is driving, how much a fastener has been tightened, or how much load a gearbox is carrying at speed.

The dominant working principle is the bonded strain gauge. A precisely machined torsion section, the spring element, twists by a tiny angle under load. Four strain gauges are bonded at plus and minus 45 degrees to the shaft axis, the directions of principal shear strain, and wired as a full Wheatstone bridge. Two gauges stretch and two compress, so torque unbalances the bridge and produces an output of roughly 1 to 2 mV per volt of excitation at nominal torque. Arranging all four gauges in one bridge cancels the effect of axial force, bending and temperature to first order, which is why a properly built torque transducer can ignore the parasitic loads that a single gauge could not.

Units matter in procurement because torque sensors are sold into both metric and imperial markets. The SI unit is the newton metre; one pound-force foot (lbf·ft) equals 1.355818 N·m, and one newton metre equals about 141.6 ounce-force inches (ozf·in). Metric industry and European test benches quote N·m and its multiples (mN·m, cN·m, kN·m); North American fastening and automotive work often quotes lbf·ft and lbf·in; small servo, robotics and electronics work uses ozf·in or cN·m. Confirming the unit on a datasheet before comparing ranges avoids a 1.36x sizing error.

Historically, torque was first measured mechanically with prony brakes and spring dynamometers in the nineteenth century. The bonded foil strain gauge, developed in the late 1930s, made electrical torque transduction practical and remains the backbone of the field. From the 1990s onward, non-contact signal transmission by rotary transformer, then surface acoustic wave and magnetoelastic sensing, removed the slip-ring brush as the limiting wear part and pushed rated speeds past 20,000 rpm. Digital on-rotor electronics and bus outputs such as CAN, EtherCAT and PROFINET are now standard on premium test-bench flanges.

Four engineering metrics decide whether a torque sensor is fit for a job: nominal (rated) torque, accuracy class, rated rotational speed, and overload capacity. These set the boundary of safe, accurate operation. A sensor used near its overload limit, spun above its rated speed, or loaded to only a few percent of nominal will not deliver its datasheet accuracy, so selection is fundamentally a matter of matching these four numbers to the application before comparing price.

Structural Types: Reaction and Rotary

The first and most consequential selection decision is structural: does the measuring element rotate or stay still. A reaction torque sensor is fixed in a non-rotating load path and measures the equal and opposite reaction torque; a rotary torque sensor turns with the drive line and measures torque on the spinning shaft. The two share the same strain-gauge sensing core but differ entirely in cost, speed capability and mounting. The table below summarizes the trade-off.

Reaction torque sensors are bolted between two flanges that do not rotate, for example between a motor mount and ground, or between a fixture and a fastener socket. Because no part spins, the bridge connects to the instrument with an ordinary cable, eliminating slip rings and their wear. This makes reaction sensors the lowest-cost and often the most accurate choice, and the natural pick for static or very slow processes: bolt-tightening verification, valve actuator torque, robot wrist joints, hinge and latch testing, and measuring the holding torque of a stalled or stator-mounted motor. Their limit is obvious: they cannot follow torque on a continuously rotating shaft.

Rotary torque sensors are inserted into the drive line, with the spring element turning at shaft speed. They are mandatory whenever torque must be read at operating speed: dynamometer testing of motors and engines, gearbox and transmission validation, pump and fan characterization, and end-of-line drivetrain checks. The engineering challenge is moving power into the rotor and the bridge signal back out across a rotating boundary, which is what separates the slip-ring, transformer and telemetry sub-types described in Chapter 4.

Torque flanges are a compact, hollow rotary form factor with integral digital electronics and an inductive interface. The HBM (HBK) T40B is a representative example rated up to 23,000 rpm in its high-speed option, with an integrated magnetic (magnetoresistive) speed and angle channel, used on high-power engine and electric-motor test benches. Flange transducers minimize axial length and rotating mass, which matters for rotor dynamics at high speed.

A further distinction is static versus dynamic measurement. Strain-gauge sensors of either family read both steady (static) and changing (dynamic) torque, limited at the high-frequency end by their mechanical resonance. Piezoelectric sensors, by contrast, measure only dynamic torque because the generated charge bleeds away under constant load, so they suit shock, impact and transient events rather than steady process monitoring.

Sensing Technologies

Independent of whether the body rotates, four physical principles convert torque into a signal: bonded strain gauge, surface acoustic wave (SAW), magnetoelastic, and piezoelectric. Each occupies a different niche of accuracy, bandwidth, cost and packaging. There is no universal principle, and the right choice depends on whether the priority is laboratory accuracy, high bandwidth, or rugged non-contact volume use. The table compares their key engineering characteristics.

Bonded strain gauge sensing is the industry default. A full Wheatstone bridge of foil gauges on a precisely sized torsion shaft gives excellent linearity, intrinsic compensation of bending and axial load, and a temperature coefficient that can be trimmed to a fraction of a percent of full scale per ten kelvin. Reference-grade transducers from HBM (HBK), Lorenz Messtechnik, Burster and Interface reach accuracy class 0.05 to 0.1 under DIN 51309. The limitation is that the gauges live on a rotating shaft, so high-speed use depends on the signal-transmission method, and the foil bond can creep over many years, which is why high-end units use thin-film or laser-welded gauges.

Surface acoustic wave (SAW) sensors deposit resonators on a quartz substrate bonded to the shaft. Strain shifts the resonant frequency, which a nearby antenna reads by radio with no electrical contact and no rotating power supply. SAW is well suited to compact, sealed, non-contact duties and has been adopted for electric power-assisted steering (EPAS) and other automotive shaft measurements. Sensor Technology Ltd. (TorqSense RWT series) is a recognized supplier of SAW-based transducers.

Magnetoelastic sensing exploits the inverse magnetostriction effect: torque changes the magnetic permeability of a specially treated ferromagnetic shaft region, and stationary coils detect the resulting field change. There are no strain gauges, no bond to creep and no contact, giving a rugged, high-bandwidth sensor that tolerates harsh automotive driveline environments, which is why suppliers such as NCTE target high-volume powertrain and steering applications. Accuracy is typically lower than a laboratory strain-gauge flange, so it trades absolute precision for robustness and cost at scale.

Piezoelectric torque sensors use quartz crystals that generate an electric charge proportional to applied torque, read by a charge amplifier. They offer very high stiffness and bandwidth and survive large overloads, making them the choice for combustion, impact and water-hammer-style transients. Their defining limitation is that the charge leaks away under static load, so they cannot hold a steady-state reading and are unsuitable for slow process monitoring. Kistler is the long-established manufacturer of piezoelectric torque measurement.

Signal Transmission, Materials and Standards

For rotary sensors, getting the bridge signal off a spinning shaft is the central engineering problem and the main driver of speed rating, maintenance and cost. Three methods are established, and a sensor's whole personality follows from which one it uses. The table compares them.

Slip rings press graphite or precious-metal brushes against rotating rings to carry excitation power in and the bridge signal out by direct electrical contact. They are simple, accurate and inexpensive, but the brushes wear and generate contact noise, which caps practical speed at a few thousand rpm and adds a recurring maintenance item. They remain common on low-speed rigs and budget retrofits. Rotary transformers couple power and a modulated signal across a small air gap inductively, with no contact and nothing to wear. This non-contact, bearingless approach is now the standard for in-line test-bench torque sensors, enabling continuous operation past 20,000 rpm with little maintenance, at higher electronics cost. Radio telemetry digitizes the bridge on the rotor and transmits it wirelessly, which suits large shafts and field retrofits where an in-line flange cannot be installed.

The torsion spring element is almost always high-strength stainless or alloy steel, chosen for a stable, repeatable modulus of elasticity and good fatigue life, because the element is cyclically stressed for the life of the sensor. Aluminum is used in some low-capacity or weight-sensitive sensors. The wetted-material concerns of a pressure sensor do not apply, but environmental sealing does: industrial torque sensors are rated to ingress protection grades such as IP54, IP65 or IP67, and the bearing and seal design sets the temperature and washdown limits. Operating temperature ranges of roughly minus 10 to plus 60 degrees Celsius are typical, with compensated ranges and thermal-drift coefficients stated on the datasheet.

Calibration and accuracy classification follow recognized standards. DIN 51309 defines the calibration of static torque measuring devices and assigns accuracy classes (0.05, 0.1, 0.2, 0.5, 1, 2 and 5) from the combined repeatability, reproducibility, reversibility, interpolation and zero deviations, applying stepped loads such as 10 to 100 percent of nominal with the transducer rotated to several mounting positions in both clockwise and counter-clockwise senses. The EURAMET cg-14 guideline gives an internationally harmonized calibration procedure, and VDI/VDE 2646 provides a parallel German framework with minimum requirements for torque calibration. Note that ISO 6789 governs hand torque tools (wrenches and screwdrivers), not transducers, but appears in fastening work because a torque sensor is often the reference that verifies those tools. Where formal traceability is needed, an ISO/IEC 17025 accredited calibration certificate should accompany the device.

Functional-safety and quality frameworks may also apply depending on the industry: automotive driveline sensors may invoke ISO 26262 functional safety, and test-bench measurement chains used for certification need documented measurement uncertainty per the relevant calibration guideline. These are project-specific and should be confirmed against the buyer's quality system, not assumed.

Key Specification Parameters

A torque-sensor datasheet may list dozens of lines, but eight parameters drive the selection decision: nominal torque, accuracy class, sensitivity, nonlinearity and hysteresis, temperature effects, rated speed, overload and breaking torque, and output signal. The Key Specifications comparison below frames the typical values an engineer encounters across precision, industrial and OEM-volume grades, after which each is explained.

Nominal (rated) torque is the full-scale capacity and the anchor of selection. Commercial ranges span from about 0.1 N·m for micro and servo work up to 10 kN·m and beyond for large drivetrains, with separate model families covering each band. Because nearly every error term is quoted as a percentage of full scale, choosing a range so the operating point lands in the upper-middle of the span is essential to real-world accuracy.

Accuracy class under DIN 51309 rolls repeatability, reproducibility, reversibility, interpolation and zero error into one figure, from 0.05 (reference) to 5. As a rule of thumb, class 0.1 suits test-bench and reference duty, class 0.2 to 0.5 covers most production and process work, and class 1 is adequate for monitoring. Each step up in class typically raises price substantially, so over-specifying accuracy is a common cost error.

Sensitivity for an analog strain-bridge output is the signal at nominal torque, typically 1 to 2 mV per volt of excitation; a conditioned sensor instead states its full-scale voltage, current or frequency output directly. Nonlinearity and hysteresis describe how far the response departs from a straight line and how much the reading differs between rising and falling torque; reference flanges hold the combined figure at or below 0.05 percent of full scale, while industrial units sit at 0.1 to 0.2 percent.

Temperature effects are stated as a zero drift and a span (sensitivity) drift per kelvin, valid only inside a defined compensated temperature range. Rated speed caps continuous rotary operation and is set by the signal-transmission method and bearing design, from a few thousand rpm for slip-ring units to 23,000 rpm for premium inductive flanges. Overload is the torque the sensor survives without losing calibration, commonly 150 to 200 percent of nominal, while the higher breaking (limit) torque, often 300 to 400 percent, is the mechanical destruction point and must never be approached. Output signal options run from raw mV/V through plus-or-minus 10 V, 4-20 mA, frequency, and digital buses such as CAN, EtherCAT, PROFINET and RS-485 Modbus.

Selection Decision Factors

To turn the preceding chapters into a specific model, work through the ordered sequence below. As with most instrument selection, errors come less from a single wrong number than from settling lower-level details before the structural decision is fixed. These eight steps double as an RFQ template.

1. Structural type first: Decide reaction versus rotary based on whether torque must be read on a rotating shaft. Reaction is cheaper and more accurate for static or slow work; rotary is mandatory for measurement at operating speed. This choice constrains every later parameter.
2. Nominal torque and overload: Size nominal torque so the operating point sits in the upper-middle of the span, then verify the static overload (150 to 200 percent) covers start-up surge, stalls and seating spikes, and that breaking torque (300 to 400 percent) is never approached.
3. Rated speed and direction: For rotary sensors confirm the maximum continuous rpm and whether torque is unidirectional or bidirectional (clockwise and counter-clockwise), since calibration and overload may differ by direction.
4. Accuracy class: Match the DIN 51309 class to the task: 0.1 for reference and test-bench, 0.2 to 0.5 for production, 1 for monitoring. Avoid paying for reference-grade accuracy on a monitoring loop.
5. Signal transmission: For rotary duty choose slip ring (low cost, low speed), rotary transformer (non-contact, high speed, maintenance-free) or telemetry (large shafts, retrofit), guided by required speed and maintenance tolerance.
6. Mechanical interface and integration: Specify shaft ends (keyed, square drive, cylindrical, flanged), couplings to absorb misalignment, mounting footprint, rotating mass and any integral speed or angle channel needed for power measurement.
7. Output and acquisition: Select mV/V, plus-or-minus 10 V, 4-20 mA, frequency or a digital bus (CAN, EtherCAT, PROFINET, Modbus) to match the data-acquisition or control system, favoring frequency or digital for high-speed rotating rigs and 4-20 mA for plant loops.
8. Environment and protection: Confirm ingress protection (IP54 to IP67), operating and compensated temperature range, vibration and EMC ratings, and any functional-safety or industry framework that applies.
9. Calibration and total cost of ownership: Require an ISO/IEC 17025 or DIN 51309 calibration certificate, then account for recalibration interval and cost, spare-parts and brush replacement (slip-ring units), and downtime, not just purchase price.

One frequently overlooked dimension is serviceability and recalibration logistics: torque references drift and must be recalibrated periodically to remain traceable, so local or regional accredited calibration service, documented recalibration turnaround, and long-term spare availability often matter more over a ten-year bench life than the initial price gap. HBM (HBK), Kistler, Lorenz Messtechnik, Burster, Interface and FUTEK maintain calibration services and broad model ranges for precision work, while Sensor Technology (SAW) and NCTE (magnetoelastic) specialize in non-contact OEM and automotive measurement. Confirm the recalibration path before committing to any vendor.