A Prony brake is a friction dynamometer that measures the shaft power of a rotating engine or motor by clamping a friction band or wood block onto a brake drum and reading the reaction torque on a lever arm. The drum drags against the friction faces, the lever tries to rotate, and a scale or load cell at the end of the arm captures the force. Multiply that force by the arm length and the shaft speed and you get brake power directly. Gaspard de Prony invented it in 1821 and engine builders still use it today for verifying rated output on rebuilt diesels, vintage tractors, and small turbines.
The Prony Brake Rule in Action
The principle is friction absorbing energy. You clamp two wooden blocks, a fabric band, or a pair of brake shoes onto a drum that turns with the engine shaft. As the drum spins, friction tries to drag the whole brake assembly around with it. A lever arm extends out from the brake body, and the far end of that arm rests on a scale, a spring balance, or a strain-gauge load cell. The shaft delivers torque, the friction converts that torque into heat in the drum, and the lever transmits the reaction force to the scale. Read the force, multiply by the arm length to get torque, multiply torque by angular velocity and you have shaft power.
Why build it this way? Because it is brutally simple and self-calibrating. There are only three things you measure — force at the end of the arm, lever length, and shaft RPM. You do not need to know the friction coefficient of the brake faces, the drum diameter, or the clamping force. They cancel out of the torque equation. That is the whole point. A workshop with a tape measure, a spring scale, and a tachometer can measure brake horsepower to within 2-3% if the operator keeps the speed steady.
The failure modes are all thermal. The drum gets hot — sometimes glowing-hot on a sustained run — because every watt of shaft power ends up as heat in the friction faces. If you run a 20 kW engine for 10 minutes against a dry Prony brake with no water cooling, the wooden blocks char and the friction coefficient collapses, the lever arm reading drifts downward, and you read low power. Stick-slip is the second killer. If the clamping force is too low or the faces glaze, the brake grabs and releases at audible frequency, the scale reading jumps ±15%, and you cannot get a clean number. The fix is either water-cooling the drum (cast a hollow drum and circulate water through it) or limiting test runs to 60-90 seconds at full load.
Key Components
- Brake Drum: Cast iron or steel cylinder bolted to the engine output shaft, typically 200-400 mm diameter for engines under 50 kW. The drum surface must be ground true to within 0.05 mm runout — anything more and the clamping force pulses once per revolution, which shows up as oscillation on the load reading.
- Friction Blocks or Band: Two hardwood blocks (traditionally maple or hornbeam) or a steel band lined with brake material, clamped onto the drum. The friction faces wear in over the first 10-20 minutes of use, and the coefficient stabilises around 0.25-0.35 once burnished. Glazed or oily faces drop µ below 0.15 and you cannot load the engine fully.
- Clamping Mechanism: A bolt or hand-wheel that tightens the two halves of the brake around the drum. You adjust this to set the load — more clamp, more torque absorbed, lower engine speed. On a well-built brake the clamp moves smoothly without grabbing, otherwise you get stick-slip oscillation.
- Lever Arm: Rigid steel arm extending radially from the brake body, typically 0.5-1.5 m long. The effective length is measured from the shaft centreline to the point of contact on the scale, and that dimension must be known to ±1 mm — every millimetre of error becomes a direct percentage error in torque.
- Force Measurement: Spring balance, beam scale, or strain-gauge load cell at the end of the arm. Modern setups use a 0-500 N or 0-2 kN load cell logged to a data acquisition system at 100 Hz so you can see oscillation and average it out.
- Tachometer: Measures shaft RPM directly off the brake drum or engine flywheel. Optical tachometers reading a reflective strip give ±1 RPM accuracy. RPM drift during a load run is the single biggest source of test-to-test scatter, so the operator must trim the throttle to hold speed within ±2%.
- Cooling Provision: On any brake absorbing more than about 5 kW continuously, the drum needs water cooling — either a hollow drum with internal flow, or a drip-feed onto the outer surface. Without cooling, friction faces char and µ drops, and your power reading walks downward through the test.
Who Uses the Prony Brake Rule
The Prony brake is the oldest absorption dynamometer still in regular use. It survives because it is cheap to build, needs no electronics, and gives a directly traceable torque measurement from a tape measure and a scale. You see them today in engine restoration shops, agricultural extension labs, small hydro stations, and university teaching benches — anywhere a workshop needs a defensible shaft-power number without buying a 50,000 USD eddy-current dyno.
- Vintage Engine Restoration: Stationary engine clubs verifying rated output of rebuilt Lister, Petter, and Ruston Hornsby single-cylinder oil engines before they go into traction or boat service.
- Agricultural Equipment Testing: Nebraska Tractor Test Laboratory historically used Prony brakes through the 1920s-1940s to certify drawbar and PTO horsepower for John Deere, Farmall, and Allis-Chalmers tractors.
- University Teaching Labs: Mechanical engineering thermodynamics courses at institutions like IIT Madras run Prony brake experiments on small Kirloskar diesel engines to teach brake power, indicated power, and mechanical efficiency calculations.
- Small Hydro Power: Field commissioning of Pelton and Turgo wheels under 30 kW at remote micro-hydro sites in Nepal and Peru where eddy-current dynos are not transportable.
- Industrial Motor Audit: Verifying actual shaft output of rebuilt 5-30 kW induction motors against nameplate before reinstallation in pump or compressor service.
- Marine Engine Workshops: Bench-testing rebuilt Gardner, Lister, and Yanmar marine diesels before fitting into fishing boats and canal barges, where the rated output must match the propeller pitch.
The Formula Behind the Prony Brake Rule
The Prony brake formula computes brake power from three measured quantities: the force at the end of the lever arm, the arm length, and the shaft speed. At the low end of a typical operating range — say 25% of rated load — the force reading on the scale is small, often near the resolution limit of a mechanical balance, and noise dominates. At the nominal operating point (rated speed, rated load) the force sits comfortably mid-scale and the reading is clean. Push to the high end, beyond about 120% of rated, and the brake drum overheats within minutes, friction faces glaze, and the reading drifts downward. The sweet spot for a Prony test is a brief 60-90 second hold at 80-100% of rated load, with the drum cool enough to keep your hand on (under 100°C).
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| Pbrake | Brake power delivered by the shaft | W | hp |
| N | Shaft rotational speed | RPM | RPM |
| F | Net force on the lever arm at the scale | N | lbf |
| L | Effective lever arm length, shaft centreline to scale contact | m | ft |
| T | Shaft torque (= F × L) | N·m | lbf·ft |
Worked Example: Prony Brake Rule in a rebuilt single-cylinder stationary engine
A heritage engine workshop in Saskatchewan is testing a freshly rebuilt 1928 Fairbanks-Morse Z 3 hp single-cylinder hit-and-miss engine before it goes back to a prairie museum. The engine drives a Prony brake with a 1.0 m lever arm and a 300 mm cast iron drum. The crew runs the engine at its rated 500 RPM and reads the spring balance at three throttle settings to map the load curve.
Given
- L = 1.0 m
- N = 500 RPM
- Fnom = 44.7 N
- Flow = 22.4 N
- Fhigh = 53.6 N
Solution
Step 1 — at the nominal operating point (rated 500 RPM, rated load), compute torque from the lever reading:
Step 2 — convert to brake power at the nominal point:
That matches the original Fairbanks-Morse rating of 3 hp at 500 RPM almost exactly — the rebuild is on spec.
Step 3 — at the low end of the typical operating range (50% load, Flow = 22.4 N):
The engine is loafing at this point. The hit-and-miss governor fires roughly every third or fourth stroke, the exhaust note is lazy, and the brake drum stays cool to the touch even after 5 minutes. This is the safe zone for extended running.
Step 4 — at the high end (120% of rated, Fhigh = 53.6 N):
You can pull this number for about 60 seconds before the wooden brake blocks start smoking. The friction coefficient drops as the wood chars, and if you hold the load any longer the scale reading walks downward by 5-10% even though the actual engine output is unchanged. Above this point you are testing the brake, not the engine.
Result
Nominal brake power is 2,340 W or 3. 14 hp at 500 RPM — within 5% of the original Fairbanks-Morse Z nameplate rating, which means the rebuild's compression, ignition timing, and valve lash are all sound. The 50% load point gives 1.57 hp with the drum cool, the 120% overload point gives 3.76 hp but only for a one-minute pull before thermal drift contaminates the reading — so the usable test window sits between the nominal and 110% load points. If your measured power comes in 15-20% below predicted, suspect three things in this order: (1) lever arm length measured to the wrong point — make sure L runs from shaft centreline to the exact scale contact, not to the brake body face, because a 50 mm error on a 1.0 m arm is a direct 5% power error; (2) RPM drift during the read — hit-and-miss engines hunt by nature, and a drop from 500 to 460 RPM costs you 8% on the power figure; (3) glazed brake blocks from a previous overheated run, which read low because µ has collapsed and the engine throttle is actually further open than the load suggests.
Choosing the Prony Brake Rule: Pros and Cons
The Prony brake competes with two other absorption dynamometers in the small-engine test space: the water brake (hydraulic dynamometer) and the eddy-current brake. Each has a clear application zone, and the choice usually comes down to how much power you need to absorb, how long you need to hold the load, and what the test budget is.
| Property | Prony Brake | Water Brake Dynamometer | Eddy Current Dynamometer |
|---|---|---|---|
| Typical power range | 0.5 - 30 kW | 10 kW - 5 MW | 1 - 500 kW |
| Accuracy (well-calibrated) | ±2-3% | ±1-2% | ±0.5-1% |
| Sustained run time at full load | 60-90 seconds (dry) / continuous (water-cooled) | Continuous | Continuous |
| Build cost (small unit) | 100-500 USD shop-built | 3,000-10,000 USD | 8,000-50,000 USD |
| Speed control / load stability | Manual hand-wheel, prone to stick-slip | Valve-controlled, smooth | Electronic, very smooth |
| Cooling requirement | Water drip or hollow drum above 5 kW | Open water circuit, large flow | Forced air or water jacket |
| Maintenance interval | Replace friction blocks every 20-50 hours | Inspect rotor pockets yearly | Bearings every 2,000 hours |
| Best application fit | Vintage engine shops, teaching labs, field tests | Marine and industrial engine cells | Production test cells, certification work |
Frequently Asked Questions About Prony Brake Rule
Thermal degradation of the friction faces. As the wooden blocks or fabric band heat up, the friction coefficient µ drops — a maple block at 20°C has µ ≈ 0.30, but at 200°C the surface chars and µ falls to 0.15-0.18. Your operator compensates by tightening the clamp to hold engine speed, but the lever arm only measures torque reaction, and that reaction reflects the actual µ × clamp force. The reading walks down because µ is collapsing.
Fix it by water-cooling the drum, or limit individual load points to 60 seconds with cool-down between. If you must run dry, watch the scale during the first 10 seconds of the hold — that initial reading, before the heat soaks in, is your honest number.
Length of test matters more than peak power. If you only need to verify rated output on a 90-second pull, a water-cooled Prony brake will do it for under 500 USD in materials. If you need to run extended endurance tests, fuel-burn maps, or governor stability runs that hold full load for 30+ minutes, the Prony brake will overheat and you need a water brake.
15 kW is right at the edge. A shop-built Prony brake with a hollow water-cooled drum (cast iron, 350 mm diameter, 60 mm wide friction face) handles 15 kW continuously if you pump 10-15 L/min of cold water through it. Below that flow rate, get a Froude or Heenan & Froude DPX-series water brake instead.
Check the lever arm length first. The single most common error is measuring L from the brake body face instead of from the shaft centreline. On a 1.0 m arm with a 300 mm drum, that mistake adds 150 mm to the effective length and gives you a 15% high reading.
Second suspect: scale zero offset. A spring balance or beam scale that reads 5 N with no load (because the lever arm itself is unbalanced) inflates every reading by 5 N. Always tare the scale with the brake disengaged but the lever resting on the scale, before you start the engine. Third — check that you are reading net force, not gross. If the lever arm has its own weight pressing on the scale, that has to be subtracted.
Size it so the expected force at rated power lands in the middle of your scale's range. Work backward from the engine's rated torque: T = P / (2π × N / 60). For a 5 hp engine at 600 RPM, rated torque is about 59 N·m. If your scale reads 0-100 N cleanly, pick L = 0.6 m so the rated reading is 98 N — near full scale, with good resolution. If your scale is 0-500 N, pick L = 0.12 m, but that short arm magnifies any pivot friction.
Practical sweet spot for shop-built brakes is L between 0.5 and 1.0 m. Shorter than 0.5 m and the arm becomes sensitive to brake-body bearing friction. Longer than 1.0 m and the arm flexes under load, which adds non-linearity.
Stick-slip oscillation, almost always caused by a static friction coefficient much higher than the dynamic coefficient. Glazed wood blocks, oil contamination on the drum surface, or a clamping pressure too low for the speed all produce this. The brake locks momentarily, the engine torque builds, the brake breaks free, the load drops, the cycle repeats — typically at 5-30 Hz.
Roughen the friction faces with coarse sandpaper to break the glaze, degrease the drum with brake cleaner, then re-bed the blocks under light load for 5 minutes. If it still oscillates, increase clamp force — stick-slip dies out once you load the brake into its stable friction regime, usually above 40-50% of rated torque.
A well-built Prony brake gets you ±2-3% on a steady-state point if you control RPM drift and thermal drift. An eddy-current dyno with a calibrated load cell and modern instrumentation hits ±0.5-1%. For agricultural extension testing, vintage engine verification, or teaching labs, ±2-3% is fine — the engine itself varies more than that day-to-day. For EPA emissions certification or OEM development work, the Prony brake is not accepted because you cannot hold a transient cycle, you cannot do motoring (driving the engine), and the thermal drift contaminates any run longer than a couple of minutes.
Yes, and several modern shop-built Prony brakes do exactly this. A vented cast iron rotor from a mid-size car (280-320 mm diameter) handles thermal load far better than wood, and a hydraulic caliper gives you fine clamping control through a master cylinder and pressure gauge. The catch is that pad µ varies with temperature too — semi-metallic pads drop from µ ≈ 0.40 cold to 0.30 hot, so you still see drift.
The bigger advantage is that you can dunk the rotor in a water bath for cooling without ruining anything, and replacement pads cost 30 USD instead of finding seasoned hardwood blocks. Just make sure the lever arm pivots on the caliper bracket assembly and not on the rotor hub bearing, or you will read shaft bearing drag plus brake torque.
References & Further Reading
- Wikipedia contributors. De 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.
