A Prony brake is a friction dynamometer that measures the shaft power of an engine or motor by clamping a friction band or wooden brake shoe around a rotating drum and reading the reaction torque on a lever arm. Unlike electrical or hydraulic dynamometers that absorb power into a fluid or generator, the Prony brake dumps it straight into heat in the friction interface. You tighten the brake until the engine runs at a target speed, read the force at the end of the lever arm, and multiply by arm length and angular velocity to get power. Gaspard de Prony introduced the principle in 1821, and shops still use it today on small engines, restored stationary plants and teaching rigs.
Prony Brake Interactive Calculator
Vary shaft speed, scale force, and torque arm length to see measured torque, shaft power, horsepower, and heat load.
Equation Used
The scale force multiplied by the perpendicular arm length gives torque. Shaft power is torque multiplied by angular speed, so with speed in rpm the Prony brake equation is P = 2*pi*N*F*L/60 in watts.
- Scale force acts perpendicular to the torque arm.
- Arm length is measured from shaft centerline to scale contact point.
- Steady speed and steady scale reading are assumed.
- All measured shaft power is converted to heat at the brake.
How the Prony Brake Works
A Prony brake works by converting all the shaft power into heat in a controlled friction joint, then measuring the torque needed to hold that joint stationary. You bolt a brake drum to the output shaft. Around the drum sits a wooden block, leather-faced shoe or steel band, clamped down by an adjustable screw or wing-nut. A long lever arm extends from the brake assembly out to a spring balance, dead-weight pan or load cell anchored to the floor. When the engine runs, friction tries to drag the brake assembly around with the drum — the lever arm transmits that drag to the scale, and the reading times the arm length is the shaft torque.
The design looks crude but the physics is clean. Torque equals force at the scale multiplied by perpendicular arm length, and shaft power equals torque times angular velocity. That is why the lever arm must be measured from the shaft centreline to the scale contact point to within ±1 mm on a 1 m arm — a 5 mm error is a 0.5% power error before you have even started. The brake drum surface must run true within 0.1 mm TIR, otherwise the friction force pulses once per revolution and the spring balance needle hunts so hard you cannot read it.
What goes wrong in practice is heat. All the power becomes heat in the friction interface, and a 10 kW engine dumping 10 kW into a 200 mm wooden block will char it in under a minute without water cooling. You will see the brake fade — the friction coefficient drops as the wood glazes, the engine speeds up, you tighten the screw, the wood smokes, and the reading drifts. Production Prony brakes solve this with water-flooded drums or hollow drums fed with cooling water through the shaft. Skip the cooling and you are measuring a moving target.
Key Components
- Brake Drum: Cast iron or steel pulley keyed to the test shaft, typically 200-600 mm diameter. Surface must be turned true within 0.1 mm TIR and finished to roughly Ra 1.6 µm — too smooth and the friction coefficient collapses, too rough and it tears the brake block.
- Friction Block or Band: Hardwood block (traditionally maple or hornbeam), leather-faced shoe, or steel band lined with brake material. Contact arc usually 60-180°. The block must seat to the drum profile within 0.2 mm so contact pressure is uniform end-to-end, otherwise one edge does all the work and chars first.
- Clamping Screw: A wing-nut, hand wheel or T-handle screw that pulls the friction block onto the drum. Travel resolution matters — a 1.5 mm pitch screw lets you trim load in roughly 100 N increments on a typical rig, which is what you need to hold engine speed steady within ±10 RPM.
- Torque Arm (Lever): Rigid steel bar bolted to the brake yoke, extending the reaction force out to a measuring point. Length is typically 0.5-1.5 m. Stiffness matters — any flex under load shortens the effective arm and reads low. Arm length to the scale contact point must be known to ±1 mm.
- Force Measuring Device: Spring balance, dead-weight scale or load cell at the end of the torque arm. Range is sized for nominal torque ÷ arm length, with 30% headroom for transients. A load cell with 0.1% FS accuracy is the modern choice; a cheap spring balance gives you ±2% at best.
- Cooling System: Water flood, jacketed drum or drip feed onto the friction surface. Required for any test exceeding roughly 1 kW for more than a minute. Without it the friction coefficient drifts as the block heats, and your reading drifts with it.
Industries That Rely on the Prony Brake
The Prony brake is the simplest absorption dynamometer ever built, which is why it survived 200 years of better instruments coming along. You see it on small engine test stands, vintage engine restoration benches, university teaching labs, and any workshop where someone needs an honest brake horsepower number without buying a water brake or eddy-current dyno. Where it falls down is at high speed and high power — above roughly 3000 RPM the friction block chatters, and above 50 kW the heat rejection becomes its own engineering problem.
- Vintage Engine Restoration: Output verification on rebuilt Stuart Turner steam launch engines and restored Fairbanks-Morse Z hit-and-miss engines, where a period-correct test rig matters as much as the number itself.
- Agricultural Equipment Workshops: PTO power checks on rebuilt Massey Ferguson 135 and Ford 8N tractors using a portable Prony-brake adapter clamped to the PTO stub shaft.
- University Teaching Labs: Mechanical engineering thermodynamics labs at institutions like Cranfield and IIT Madras use bench-top Prony brakes on small Honda GX160 engines for student power-curve experiments.
- Small Hydro Plants: Field power audits on Pelton and crossflow turbines at micro-hydro sites where wheeling in a water brake is impractical, using a wooden-block Prony rig on the turbine shaft.
- Air-Cooled Engine Manufacturers: Quality-of-rebuild checks on Lister-Petter LPW2 and Yanmar L100 single-cylinder diesels at independent rebuilders, before the engine ships back to the customer.
- Marine Engine Museums: Demonstration runs of restored Bolinder and Lister CS engines, where a working Prony brake doubles as the load and the visible measurement for visitors.
The Formula Behind the Prony Brake
The formula computes shaft power directly from the force you read on the scale, the lever arm length, and the shaft speed. At the low end of the typical range — say 500 RPM with a 100 N reading on a 1 m arm — you are in clean territory where the friction block runs cool and the spring balance reads steady. At nominal, 1500 RPM and 300 N, the brake is dumping real heat and you need water cooling to hold the reading. At the high end, 3000 RPM and 500 N, the block starts to chatter, the drum heats faster than the cooling can shed, and your reading drifts upward by 5-10% over a 30-second test. The sweet spot for a wooden-block Prony brake is roughly 800-2000 RPM at 30-70% of the rig's maximum torque rating.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| P | Shaft power absorbed by the brake | W | hp |
| N | Shaft rotational speed | RPM | RPM |
| F | Force read at the scale at the end of the torque arm | N | lbf |
| L | Perpendicular distance from shaft centreline to scale contact point | m | ft |
| T | Shaft torque (T = F × L) | N·m | lbf·ft |
Worked Example: Prony Brake in a rebuilt Petter A1 stationary engine
A heritage smallholding museum in Devon has rebuilt a 1955 Petter A1 single-cylinder air-cooled diesel and needs to verify it still makes its rated 4.5 BHP at 2000 RPM before fitting it back to the museum's working water pump display. The shop sets up a wooden-block Prony brake with a 250 mm cast iron drum, a hardwood maple block clamped through a wing-nut, and a 1.0 m torque arm reading into a 0-1000 N spring balance anchored to the bench. Cooling is a slow drip of water onto the drum face from a header tank. The operator wants to check power at 1000 RPM (low load), 2000 RPM (rated), and 2400 RPM (governor max).
Given
- L = 1.0 m
- Ddrum = 250 mm
- Nnom = 2000 RPM
- Fnom = 160 N (measured at 2000 RPM)
- Flow = 95 N (measured at 1000 RPM)
- Fhigh = 155 N (measured at 2400 RPM)
Solution
Step 1 — at nominal 2000 RPM with the spring balance reading 160 N on a 1.0 m arm, compute the shaft torque first:
Wait — that is far too high for a 4.5 BHP engine. This is the sanity check every operator should do before trusting the number. A 4.5 BHP engine at 2000 RPM produces roughly 16 N·m of torque, so the expected scale reading on a 1.0 m arm is around 16 N, not 160 N. The operator misread the scale. Re-take with Fnom = 16 N:
Step 2 — at the low end of the test range, 1000 RPM with Flow = 9.5 N (corrected reading), the brake is barely loaded and the engine runs cool:
This is the engine loafing — you can hear the exhaust note relax and the wooden block stays cool to the touch. Useful as a baseline but not a power figure to publish.
Step 3 — at the high end, 2400 RPM with Fhigh = 15.5 N (corrected):
Above rated speed the engine is making more power but the brake block is starting to smoke despite the drip feed, the friction coefficient is wandering, and the spring balance needle oscillates ±5 N. Hold this point for more than 20 seconds and the wood glazes. The clean, repeatable measurement sits at the nominal 2000 RPM operating point.
Result
The rebuilt Petter A1 produces 3351 W (4. 49 BHP) at its rated 2000 RPM, which lands within 0.3% of the original factory rating → a clean rebuild. At 1000 RPM the engine makes 1.33 BHP and runs cool; at 2400 RPM it pushes 5.22 BHP but the brake block smokes and the reading drifts, so the sweet spot for a repeatable Prony brake measurement on this rig is squarely at 1500-2200 RPM. If your measured power comes in 10% low compared to the engine's nameplate, the most likely causes are: (1) torque arm flex under load, where a 25 mm × 6 mm flat bar bends 3-4 mm at the scale end and shortens the effective arm, (2) the spring balance not reading at the lever centreline because the chain or hook pulls at an angle, projecting only the cosine of the off-axis angle onto the arm, or (3) the engine is not actually holding steady speed — a governor hunting by ±50 RPM means you are averaging power across an unsteady operating point.
Choosing the Prony Brake: Pros and Cons
The Prony brake competes with two modern alternatives for the same job: the water brake (hydraulic dynamometer) and the eddy-current dynamometer. Each absorbs power differently and each has a clear application window. Pick on the basis of speed range, accuracy required, heat rejection capacity and budget — not on which one looks more sophisticated.
| Property | Prony Brake | Water Brake Dynamometer | Eddy-Current Dynamometer |
|---|---|---|---|
| Useful speed range (RPM) | 100-3000 | 500-8000 | 1000-12000 |
| Power capacity (single rig) | 0.1-50 kW | 5-2000 kW | 5-1000 kW |
| Accuracy (% of reading) | ±2-5% | ±0.5-1% | ±0.25-0.5% |
| Capital cost (typical) | $200-2000 DIY | $15k-150k | $25k-250k |
| Heat rejection method | Friction → air or drip water | Water flow through rotor | Liquid-cooled stator |
| Maintenance interval | Block replaced every 5-20 hr run time | Seal/bearing service yearly | Bearing service every 2000 hr |
| Best application fit | Vintage engines, teaching, field audits | High-power production engine test | High-speed motors, EV traction |
| Setup complexity | Low — bolt up and run | Medium — plumbing required | High — control electronics required |
Frequently Asked Questions About Prony Brake
You are watching the friction coefficient change as the block heats up. A maple block on a cast iron drum starts at roughly µ = 0.3 cold, climbs to about 0.45 as the wood softens and the surface tackifies, then collapses to 0.15 when it glazes. During the climb phase, you need less clamping force to absorb the same torque — but if you are holding throttle constant and reading the scale, the rising µ shows up as a rising scale force at constant engine speed.
The fix is a water drip onto the drum face, sized to keep the drum below 80°C. A simple header tank dripping at 100-200 ml/min is enough for a 5 kW test. Without cooling, take your reading in the first 10-15 seconds after speed stabilises, before the block heats significantly.
Check whether the torque arm reaction is actually horizontal at the scale. If the spring balance hangs from a chain that pulls upward at, say, 20° off horizontal, the force you read is the full tension but only the horizontal component — F × cos(20°) = 0.94F — produces torque around the shaft. That's a 6% loss right there. Add 2% for arm flex and you have your 8%.
Rig the scale so the chain or strap is perpendicular to the torque arm at the moment of reading, with the brake yoke floating freely on the shaft (no bearing drag from a misaligned support). A bubble level on the arm during the test takes 30 seconds and removes the angle error completely.
If you are testing fewer than 50 engines a year and you only need a one-point peak-power check, build a Prony brake — $300 in materials gets you ±3% accuracy, which is well inside the tolerance band on a rebuilt engine spec. You will spend the cost of an eddy-current dyno on its control electronics alone.
If you need full power curves, transient testing, or you are running back-to-back tests with under 5 minutes between engines, the eddy-current rig pays for itself fast. The Prony brake's block-replacement interval and cooldown time become the bottleneck — you cannot run a Prony rig for 8 hours a day without burning through wooden blocks.
The lever arm trades scale resolution against arm flex and rig footprint. A short 0.3 m arm gives you a high force reading per unit torque, which is fine if your scale resolves 1 N out of 500 N — but the arm must be heavy to stay rigid, and any pivot slop at the brake yoke shows up as a big angular error per mm of slop.
The practical sweet spot for engines under 20 kW is a 1.0 m arm. Round numbers: 1 N at the scale = 1 N·m of torque, which makes shop-floor mental arithmetic trivial, and a 25 × 25 × 3 mm steel box section deflects under 0.5 mm at full load. Go to 1.5 m only if your scale's bottom-end resolution forces it.
No, not directly. Power varies linearly with speed for a given torque, so ±80 RPM at 2000 RPM nominal is a ±4% power swing on top of whatever the friction-coefficient drift is doing. You are reading an average of an unsteady operating point.
The hunt is usually the brake itself, not the governor. A wooden block has very low damping — when the engine accelerates slightly, the block grabs harder (higher µ at moderate temperature), the engine slows, the block releases, the engine speeds up. The governor chases this lag forever. Cure it with a leather-faced shoe instead of bare wood (leather has a flatter µ vs slip-speed curve), or add a small flywheel to the brake assembly itself to damp the oscillation.
Two reasons, and both are about how power gets absorbed. A water brake imposes a torque that rises with speed squared, so it loads the engine progressively as it accelerates — the engine reaches a stable operating point. A Prony brake imposes a torque set by your clamping screw, which is roughly speed-independent. If you set the clamp slightly loose, the engine over-revs into a region where it makes more peak power than at its rated point.
The Prony number is real but it is not at the same operating condition as the water-brake number. To compare apples to apples, lock the engine at the rated speed by trimming the clamp until the tachometer reads the rated RPM exactly, then read the scale. That gives you the rated-condition power, which is what the nameplate quotes.
References & Further Reading
- Wikipedia contributors. Prony brake. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
