A dynamometer is an instrument that measures force, torque, or mechanical power produced by a rotating or translating machine. It solves the problem of quantifying real shaft output under load — nameplate ratings rarely match what an engine, motor, or actuator actually delivers. The device absorbs or transmits the load through a calibrated reaction element — a torque arm on a load cell, a strain-gauged shaft, or a hydraulic brake — and reads the reaction force against a known lever arm. Outputs typically span 0.1 Nm bench rigs to 50,000 Nm marine test stands.
Dynamometer Interactive Calculator
Vary load-cell force, lever arm length, shaft speed, and arm-length error to see torque, power, horsepower, and measurement sensitivity.
Equation Used
Torque is the measured load-cell force multiplied by the calibrated lever arm length. Power is found by multiplying torque by shaft angular speed. The arm-error output mirrors the article note that lever length uncertainty directly creates the same percentage torque reading error.
FIRGELLI Automations - Interactive Mechanism Calculators.
- Load-cell force is perpendicular to the torque arm.
- Torque arm is rigid and measured from shaft centerline to load-cell line of action.
- Bearing stiction and absorber losses are not included.
- Power is calculated from measured shaft speed and reaction torque.
Operating Principle of the Dynamometer (form)
Every dynamometer works on the same principle — Newton's third law on a calibrated lever. You connect the device under test to an absorber or transmitter that resists motion, then measure the reaction force the absorber tries to push back into its mount. Multiply that force by the lever arm length and you have torque. Multiply torque by angular velocity and you have power. The mechanism splits into two families. An absorption dyno (water brake, eddy current, hydraulic, fan brake) dissipates the input power as heat — the Froude DPX series and Land & Sea water brakes are textbook examples. A transmission or in-line dyno (strain-gauged shaft, torque flange) lets the power pass through to a real load while reading torque on the way. Hand-held spring or hydraulic dynamometers, like the Dillon AP and Chatillon DPP, do the same job in pure tension or compression.
Why the calibrated lever arm matters — if your torque arm length is off by 1 mm on a 250 mm arm, your reading is off by 0.4%. That sounds small until you're certifying a motor against a ±0.25% spec sheet. The load cell must sit perpendicular to the arm at the calibration position, the trunnion bearings must be free of stiction, and the housing must be free to rotate within ±2° without rubbing harness lines or coolant hoses. Stiction in the trunnion bearings is the single most common reason a dyno reads low at light load and correct at heavy load — the housing simply doesn't break free until you've put real torque into it.
Failure modes are predictable. Eddy current units overheat when coolant flow drops below the rated 40 L/min and the rotor warps, throwing eccentric load into the load cell. Water brakes cavitate at low water flow and the reading goes noisy — typically ±5% jitter where you'd expect ±0.5%. Strain-gauged torque flanges drift with temperature if the bridge isn't temperature-compensated, and you'll see the zero crawl 0.1% per °C across a 30°C cell warm-up.
Key Components
- Absorber or Transmitter: The element that resists or measures the input torque. Water brakes, eddy current rotors, hysteresis brakes, and AC regen dynos sit in the absorber camp. Strain-gauged shafts and torque flanges sit in the transmitter camp. Sizing rule — the absorber's continuous power rating must exceed the test article's peak by at least 20%.
- Trunnion Bearings: Two low-friction bearings that let the absorber housing rotate freely about the shaft axis so the reaction torque transfers cleanly to the load cell. Bearing friction torque should stay below 0.05% of full-scale rated torque, otherwise low-load readings drift.
- Torque Arm: A rigid lever bolted to the housing, ending at the load cell. Length must be known to ±0.05 mm and stay rigid under full-scale load — deflection above 0.1 mm at the load cell tip introduces measurable cosine error.
- Load Cell: A strain-gauged or piezoelectric force sensor reading the reaction force at the end of the torque arm. Typical accuracy classes are 0.05% to 0.25% of full scale. The cell must sit perpendicular to the arm at zero deflection — off-axis loading kills accuracy fast.
- Speed Sensor: An optical or magnetic pickup measuring shaft RPM. You need this to compute power. A 60-tooth gear with a Hall pickup is the workshop standard, giving 1 RPM resolution at 100 Hz update.
- Cooling System: Water or air loop that removes the absorbed power as heat. A 200 kW eddy current dyno needs roughly 5 L/s of coolant at a 30°C rise. Insufficient flow warps rotors and trips thermal protection.
- Data Acquisition: Reads load cell, speed, fuel flow, exhaust temps, and emissions. Modern systems like Sierra CP CADET and HORIBA STARS sample at 1 kHz minimum so transient torque steps are captured cleanly.
Real-World Applications of the Dynamometer (form)
Dynamometers turn up anywhere a shaft has to deliver predictable power. The reader Googling 'how do you measure engine horsepower' lands on a chassis dyno; the reader specifying a hand winch lands on a Dillon spring dyno. Every torque arm and load cell pairing follows the same physics — only the scale and the absorber type change. Below are the realistic application slots, with named hardware and named end users.
- Automotive performance tuning: Mustang Dynamometer MD-AWD-500 chassis dyno measuring 500 kW at the wheels on tuned Subaru WRX builds, with eddy current absorbers under each roller pair.
- Marine propulsion certification: Froude AG800 water brake dyno used at MAN Energy Solutions' Augsburg works to certify medium-speed marine diesels at 8 MW shaft power.
- Electric motor R&D: Magtrol HD-825 hysteresis dynamometer benchmarking 5 kW BLDC traction motors for e-bike OEMs, reading torque to 0.1 Nm resolution at speeds up to 25,000 RPM.
- Aerospace engine test: GE Aerospace's Peebles, Ohio test site uses water brake and air brake dynos rated to 100,000 lb-ft for turboshaft and propeller-shaft engines like the T901.
- Lifting and rigging certification: Dillon AP-50000 mechanical force gauge in-line on a crane hook to proof-test 25-tonne shackles per ASME B30.26 before deployment offshore.
- Pharmaceutical mixing: In-line torque flange (HBM T40B) on a 250 L jacketed reactor's agitator shaft at a Pfizer process pilot plant, tracking viscosity rise during a polymerisation batch.
- Cycling biomechanics labs: SRM PowerMeter cranksets used on lab ergometers at the Australian Institute of Sport, reading rider torque at the spider arms with a 0.5% accuracy class.
The Formula Behind the Dynamometer (form)
The fundamental equation converts a measured reaction force at the end of a known torque arm into shaft torque, then combines that torque with measured speed to give shaft power. At the low end of typical operation — say 10% of full-scale torque on a 1000 Nm dyno — you're reading 100 Nm and the load cell signal is well above the noise floor but trunnion stiction starts to matter. At nominal full-scale you get the cleanest signal-to-noise. Push beyond rated capacity and load cell linearity falls off, the torque arm starts to deflect measurably, and cosine error creeps in. The sweet spot for accurate work sits between 20% and 80% of rated full scale.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| T | Shaft torque | N·m | lb·ft |
| F | Reaction force at the load cell | N | lbf |
| L | Torque arm length, load cell centre to shaft centre | m | ft |
| P | Shaft power | W | hp |
| ω | Angular velocity | rad/s | rad/s |
| N | Shaft speed | RPM | RPM |
Worked Example: Dynamometer (form) in a wind turbine gearbox test rig
An offshore wind OEM is qualifying a 3 MW gearbox on a back-to-back test stand at Fraunhofer IWES in Bremerhaven. The gearbox couples to an ABB AC regen dynamometer through a 2.5 m torque arm reacting against a 500 kN HBM C6A load cell. The test point is the gearbox high-speed shaft turning at 1500 RPM under partial load, and the engineer needs the measured shaft power at three operating points to confirm the dyno reads correctly across the certification envelope.
Given
- L = 2.5 m
- Fnom = 8000 N
- N = 1500 RPM
- Flow = 2000 N (25% load)
- Fhigh = 12000 N (150% load — overload check)
Solution
Step 1 — at the nominal test point, compute torque from reaction force and arm length:
Step 2 — convert 1500 RPM to rad/s and compute nominal shaft power:
Step 3 — at the low end of the typical certification envelope, 25% load with Flow = 2000 N:
At 25% load the load cell signal is still 5× above the noise floor of an HBM C6A, and the torque arm sees only 0.05 mm deflection — clean, accurate, well within the cell's 0.03% linearity class. This is where you do most of your part-load efficiency mapping.
Step 4 — at 150% load, the overload check, Fhigh = 12,000 N:
At 150% you're past the load cell's calibrated range — linearity error doubles to roughly 0.1%, the torque arm deflects ~0.3 mm and introduces measurable cosine error, and you'd only run this point briefly to verify the gearbox survives a grid-fault torque spike. The sweet spot is clearly the 20–80% band where the C6A is both linear and stiff.
Result
The nominal shaft power is 3. 14 MW at 20,000 N·m and 1500 RPM. That figure tells the test engineer the gearbox is converting input mechanical energy at the expected efficiency for a 3 MW class — dropping 60 kW lower than predicted would flag a bearing drag or oil-shear loss anywhere in the train. Across the operating range, the dyno reads cleanly at 0.785 MW (25% load), 3.14 MW nominal, and 4.71 MW at the overload check, with accuracy degrading visibly only above the 80% mark. If the measured power comes back 2-3% low, suspect three things in this order: trunnion bearing stiction breaking the housing free late under load (a 0.5% friction-torque loss looks like a 0.5% power loss across the band), torque arm length error from a re-assembly with non-original spacers (each 1 mm of arm error on a 2.5 m arm is 0.04% straight off the answer), and load cell zero drift from a cold cell that was tared at 25°C and is now reading at 8°C ambient (HBM C6A drifts roughly 0.02% per 10°C if not compensated).
Choosing the Dynamometer (form): Pros and Cons
The right dyno depends on what you're measuring, how fast you need to change load, and how much heat you can dissipate. The three families below cover roughly 95% of real-world test work — water brake for big iron, eddy current for fast transients on medium-power gear, and AC regen for anything that needs four-quadrant operation or energy recovery.
| Property | Eddy Current Dynamometer | Water Brake Dynamometer | AC Regenerative Dynamometer |
|---|---|---|---|
| Power range | 1–2,000 kW | 50 kW–10 MW | 1 kW–5 MW |
| Torque accuracy class | ±0.25% FS | ±0.5% FS | ±0.1% FS |
| Maximum shaft speed | 12,000 RPM (Magtrol WB series) | 8,000 RPM (Froude AG) | 25,000 RPM (Sierra AC) |
| Load step response time | 50–200 ms | 500 ms–2 s | 5–20 ms |
| Four-quadrant motoring capability | No (absorption only) | No (absorption only) | Yes — drives and absorbs |
| Cooling requirement | Water-cooled, 5 L/s per 200 kW | Process water, 10 L/s per 500 kW | Air-cooled drive, water-cooled motor |
| Capital cost (typical) | $40k–$300k | $80k–$1.5M | $150k–$2M |
| Best fit application | Engine and motor R&D, transient testing | Marine, large diesel, turboshaft certification | EV powertrain, e-axle, regen energy recovery |
Frequently Asked Questions About Dynamometer (form)
This is almost always trunnion bearing stiction. The housing must rotate freely against the load cell to register reaction torque, but a sticky bearing — even one that feels fine by hand — needs a threshold torque to break free. At idle that threshold consumes a measurable fraction of the small reaction force, so your reading sits low. At peak torque the stiction torque is a tiny fraction of the total and disappears in the noise.
Diagnostic check — with the shaft stationary, put a calibration weight on the torque arm and watch the load cell. If you have to tap the housing to get the reading to settle, your bearings need attention. Replace with sealed angular-contact bearings rated for oscillating motion, not deep-groove ball bearings.
AC regen for transient work, every time. A water brake's load step response is bottlenecked by how fast you can change water flow into the rotor cavity — typically 500 ms minimum, often 1-2 seconds for a clean settle. An AC regen unit changes load in 5-20 ms because it's just a torque-controlled inverter command. For a marine genset where you're simulating sudden hotel-load changes or grid faults, you need that bandwidth.
The other reason is energy recovery. A 500 kW water brake dumps 500 kW into your cooling tower as waste heat. An AC regen unit pushes that power back to the grid at 90%+ efficiency, which over a multi-week endurance test is a five-figure electricity bill saved.
Run a static torque check first. With the shaft locked, hang calibrated weights on the torque arm and confirm the load cell reads correctly across 10%, 50%, and 90% of full scale. If those three points are within spec, your torque chain is fine and the error lives in the speed measurement.
For speed, compare the dyno's RPM channel against an independent optical tachometer aimed at a piece of reflective tape on the shaft. Common speed-sensor faults are missing teeth on the trigger wheel (gives a periodic 1.7% low reading on a 60-tooth wheel with one chipped tooth) and Hall sensor air-gap drift after thermal cycling — anything above 1.5 mm gap and you start dropping pulses at high RPM.
That's drivetrain loss and it's mostly real, not a measurement error. Power leaves the crankshaft and gets eaten by the clutch or torque converter, gearbox gear mesh, propshaft U-joints, differential, axle bearings, and finally tyre-roller interface losses. A typical RWD passenger car loses 12-18%, AWD loses 18-25%, FWD with a longitudinal layout sits around 10-14%.
The tyre-to-roller interface alone can swallow 3-5% — soft sidewalls flex, heat up, and lose energy. If your delta is much larger than 18% on a RWD, check tyre pressure (run cold to 40 psi for dyno work), confirm the rollers are clean and dry, and verify the dyno's parasitic-loss coast-down compensation is actually being applied to the result.
Yes, if you treat the winch drum as a torque arm. Attach the Dillon in-line with the cable at a known drum radius, pull a steady load, and torque equals force times drum radius. This is exactly how field winch certifications get done when nobody wants to ship a winch to a test house.
Two gotchas. First, the cable must leave the drum tangent to the radius you measured — if it spirals up the drum across the test, your effective radius changes by the cable diameter per layer, typically 6-12 mm per wrap. Second, the Dillon reads peak-hold by default, not average, so a jerky pull gives you the worst-case spike, not the working torque. Use the smoothed average mode for real torque numbers.
Roughly 5-10 N·m before noise and stiction make the reading useless — about 0.5-1% of full scale. The limit comes from three sources stacking up: load cell noise floor (typically 0.02% FS for a good strain-gauge cell), trunnion friction torque (0.05-0.1% FS even on good bearings), and electrical noise on the data acquisition (around 0.05% FS on a 16-bit system in a noisy test cell).
If you regularly need to read low torques, don't oversize the dyno — fit a smaller absorber or use a lower-range torque flange in series. A 100 N·m flange reading 5 N·m is at 5% FS and well within its accuracy class. A 1000 N·m flange reading 5 N·m is at 0.5% FS and below it.
References & Further Reading
- Wikipedia contributors. Dynamometer. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
