A Double Link Balanced Scale is a torque-measuring instrument that uses two parallel suspension links to constrain a brake or absorption housing around a rotating shaft, transferring the reaction torque to a balance beam read against a sliding poise weight. The arrangement was refined through 19th-century dynamometer work by engineers like Gaspard de Prony and later William Froude, who needed a stable way to measure shaft power. The double linkage cancels side loads so only pure tangential force reaches the scale, letting you compute brake power from a single weight reading and a known arm length — accuracy typically lands within ±0.5% on engine test stands.
How the Double Link Balanced Scale Works
The mechanism sits around a rotating shaft. A friction band, water brake, or eddy-current housing grips the shaft and tries to rotate with it. Two parallel links — top and bottom, equal length, pinned on knife edges — restrain the housing from rotating but let it float vertically and laterally without binding. The reaction torque from the shaft pushes a torque arm against a balance beam, and you slide a poise weight along the beam until it balances. Multiply the weight by the arm length and you have torque. Multiply torque by angular velocity and you have shaft power.
Why two links instead of one? A single link would let the housing swing in an arc, and any vertical motion would feed a parasitic moment into the scale. The double linkage forms a parallelogram — the housing translates without rotating, so the only force the beam ever sees is the pure tangential reaction. The knife-edge pivots must run sharp and clean. We specify the edges to hold a radius below 0.05 mm; anything blunter and you get hysteresis, where the reading on a rising load differs from the reading on a falling load by 1-2%.
If the link lengths are mismatched by more than 0.1 mm in a 300 mm pair, the housing tilts under load and the torque arm geometry changes with deflection — your calibration drifts as the engine warms up. Common failure modes are dirt on the knife edges (causes sticky, jumpy readings), corrosion on the suspension links (adds tare weight that shifts zero), and bearing drag inside the brake housing itself (reads as phantom torque). A proper tare weight check before each run catches the second one immediately.
Key Components
- Brake or Absorption Housing: The reaction body that surrounds the shaft and absorbs the power. Could be a friction band (Prony brake), water-filled rotor casing (Froude dynamometer), or eddy-current stator. Mass typically 20-200 kg depending on capacity, and it must float on the links with zero contact to ground.
- Upper and Lower Suspension Links: Two parallel rigid links, equal length within 0.1 mm, pinned at both ends on knife edges. They form a parallelogram constraint that allows pure translation of the housing under reaction torque without permitting rotation or introducing side loads.
- Knife-Edge Pivots: Hardened steel edges, typically 60° included angle, ground to a radius under 0.05 mm. They provide near-frictionless rotation about a defined line. Wear or contamination here is the single largest source of measurement hysteresis.
- Torque Arm: Rigid lever fixed to the housing, projecting horizontally to the balance beam. Length is the calibrated moment arm — measured to ±0.05 mm from pivot centreline to load point. Typical lengths run 0.5 to 1.0 m.
- Balance Beam with Poise Weight: Graduated lever with a sliding weight. The operator moves the poise until the beam floats level, indicated by a centre pointer. Resolution is set by the smallest poise increment — typically 0.01 kgf on a 100 kgf scale.
- Tare Weight Adjuster: A small counter-weight or screw on the beam that zeros the system with no shaft torque applied. Must be checked before every run because corrosion, paint chips, or accumulated dust on the housing changes the static balance over time.
Where the Double Link Balanced Scale Is Used
The Double Link Balanced Scale shows up wherever someone needs an honest, traceable measurement of shaft power on a rotating machine. It predates electronic load cells by a century and still earns its place in test labs because the calibration chain is short and obvious — you trace it back to a known mass and a measured length, full stop. Where electronic dynos rely on strain-gauge calibration that drifts with temperature and exciter voltage, the balanced scale just needs gravity, which doesn't drift.
- Marine Engineering: A Royal Navy training establishment at HMS Sultan uses a Heenan & Froude DPX hydraulic dynamometer fitted with a double-link suspension to run student exercises measuring brake power on auxiliary diesel sets up to 500 kW.
- Heritage Power: The Crossness Engines Trust in London uses a portable Prony brake with parallel link suspension to measure indicated versus brake horsepower on the restored 1865 James Watt & Co beam engines during open running days.
- Automotive Education: Coventry University's automotive workshop runs a Plint TE150 chassis dyno teaching rig with a balance-beam reaction frame to demonstrate torque measurement principles to first-year mechanical engineering undergraduates.
- Agricultural Machinery Testing: The Nebraska Tractor Test Laboratory historically used a double-link balanced scale on PTO dynamometers for OECD power certification of tractors up to 200 kW before transitioning to load-cell systems in the 1990s.
- Pump Manufacturing: A Sulzer pump test bay in Winterthur uses a reaction dynamometer with double-link suspension to verify hydraulic shaft power on multistage centrifugal pumps during factory acceptance testing per ISO 9906 grade 1B.
- Small Engine Research: The MIRA powertrain test cells use Froude DPY1 water brakes with double-link reaction frames for endurance testing of motorcycle engines from 10 kW to 80 kW, where mass-based torque calibration is required for traceability.
The Formula Behind the Double Link Balanced Scale
Brake torque comes straight from a force times an arm. The interesting question is how that translates to shaft power across the operating range. At the low end of a typical engine test — say 1000 RPM and 20% load — the scale reads small numbers and the resolution of the poise weight starts to matter. At the nominal mid-range point you get clean, repeatable readings. Push toward the high end where the brake is dissipating its rated capacity and the housing heats up; thermal expansion of the torque arm shifts the effective length by 0.1-0.2%, which is why precision test stands run a temperature correction. The sweet spot for any given balanced scale is roughly 30-80% of rated capacity.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| P | Brake power delivered by the shaft | W | hp |
| W | Net force read on the balance beam (poise weight × g) | N | lbf |
| L | Torque arm length from pivot centreline to load point | m | ft |
| N | Shaft rotational speed | rev/min | rev/min |
| T | Brake torque (intermediate, T = W × L) | N·m | lbf·ft |
Double Link Balanced Scale Interactive Calculator
Vary scale force, torque arm length, and shaft speed to see reaction torque and brake power on an animated double-link dynamometer diagram.
Equation Used
The scale force F acts at arm length L, giving reaction torque T. Shaft speed n is converted to angular speed omega, then brake power is P = T * omega.
- Scale reading is the net tangential force applied at the torque arm.
- The double parallel links cancel side loads so the beam reads pure reaction force.
- Tare, bearing drag, and calibration corrections are not included.
Worked Example: Double Link Balanced Scale in a single-cylinder research engine cell
A university powertrain lab is benchmarking a 500 cm³ single-cylinder Ricardo Hydra research engine on a Froude DPX1 water brake fitted with a double-link balanced scale. The torque arm measures 0.716 m from knife-edge pivot to load point, and the engine is being mapped from idle through peak power. At the rated point the operator reads a poise weight of 5.20 kgf with the engine spinning at 4500 RPM. The lab needs the brake power at this point, and they want to understand how readings will look across the full mapping sweep from 1500 to 6000 RPM.
Given
- L = 0.716 m
- Wpoise = 5.20 kgf
- Nnom = 4500 RPM
- g = 9.81 m/s²
Solution
Step 1 — convert the poise reading from kgf to newtons:
Step 2 — compute brake torque at the load point:
Step 3 — compute nominal brake power at 4500 RPM:
That is the headline number at the rated point. Now look across the operating range. At the low end of the sweep, 1500 RPM, the same engine on a part-load map might pull only 2.0 kgf on the beam — torque drops to roughly 14 N·m and power to about 2.2 kW. At that reading the poise resolution of 0.01 kgf represents 0.5% of full scale, so repeat readings will scatter by ±0.05 kW which is acceptable but visible.
Push to the high-end test point at 6000 RPM with the brake near its rated absorption, the operator may read 6.5 kgf on the beam:
At this point the water brake is dumping nearly 30 kW into the cooling water and the housing skin temperature climbs above 60 °C. The torque arm grows by roughly 0.1 mm from thermal expansion — a 0.014% effect on length, smaller than the poise resolution — but the bearings inside the brake start to drag noticeably, which is why you always re-tare with the engine motoring before recording high-power data.
Result
Nominal brake power at the 4500 RPM rated point is 17. 2 kW. That number means the engine is delivering enough shaft work to run a small domestic generator set at full output, and the scale beam will sit dead-level when the operator parks the poise at the 5.20 kgf graduation — a confident, unambiguous reading. Across the sweep the lab sees roughly 2.2 kW at idle-load 1500 RPM, 17.2 kW at the nominal 4500 RPM point, and around 28.7 kW at the 6000 RPM peak, with the cleanest repeatability inside the 30-80% capacity band as expected. If your measured power runs 3-5% below predicted, suspect three things first: parasitic drag from a stiff cooling-water hose touching the housing (looks like extra torque on the beam), a corroded knife edge that has lost its sharp line contact and now rolls slightly under load, or a tare drift caused by water pooling in the brake casing between runs. Each of these adds 0.3-0.8 N·m of phantom torque that you'll only catch with a proper motoring tare check.
Double Link Balanced Scale vs Alternatives
The Double Link Balanced Scale competes against modern strain-gauge load cells and torque-flange transducers. Each has a clear domain. Pick on accuracy traceability, response speed, calibration cost, and how often you need to certify the instrument.
| Property | Double Link Balanced Scale | Strain Gauge Load Cell | Inline Torque Flange (e.g. HBM T40) |
|---|---|---|---|
| Accuracy at nominal load | ±0.3% to ±0.5% of reading | ±0.1% to ±0.25% of full scale | ±0.05% to ±0.1% of reading |
| Response time | 1-3 seconds (operator balances beam) | <10 ms | <1 ms |
| Calibration traceability | Direct — mass and length, traceable to SI primary standards | Via reference load cell, drifts with temperature | Via reference torque transducer and shunt cal |
| Capital cost (typical mid-range unit) | $3,000-$8,000 | $1,500-$5,000 | $8,000-$25,000 |
| Recalibration interval | 3-5 years (only knife edges and beam wear) | 12 months | 12-24 months |
| Suitable for transient data | No — steady-state only | Yes | Yes — best in class for transient |
| Service lifespan | 30-50 years with knife-edge servicing | 10-15 years | 10-20 years |
| Sensitivity to vibration | Moderate — beam oscillates but mean reads true | High — needs filtering | Low — digital filtering built in |
Frequently Asked Questions About Double Link Balanced Scale
Two thermal effects stack up. First, the brake housing expands and the bearings inside it tighten clearance, adding bearing drag that the scale reads as extra torque. Second, oil viscosity in the brake or shaft seals drops as temperature rises, but the change is non-linear and shows up as a slow walking zero.
Diagnostic check: motor the engine through the brake at zero fuelling and record the tare both cold and at operating temperature. The difference is your thermal tare error — subtract it from your hot readings. On a Froude DPX brake we typically see 0.4-0.8 N·m of warm-up tare creep over 20 minutes.
Match the arm length to the torque range so the poise sits in the middle third of the beam at your most-tested operating point. A 1.0 m arm halves the force at the beam compared to 0.5 m, which is great for resolution at low torque but eats into your high-end capacity. For a research engine running 20-50 N·m most of the time, 0.716 m is a common compromise (it gives a round 1 kgf = 7.02 N·m calibration constant on some Froude designs).
Rule of thumb: pick L so that 50% of expected peak torque places the poise at 60-70% of beam travel. That keeps you out of the noisy bottom end and gives headroom for surprise overloads.
You're seeing the natural frequency of the beam-and-housing assembly excited by torsional vibration in the driveline. Single-cylinder and twin-cylinder engines are the worst offenders because firing pulses come at half engine speed and can land near the 2-5 Hz pendulum frequency of a typical balanced scale.
Fix it by adding a viscous damper (a dashpot of light oil) on the torque arm, not by stiffening the linkage — stiffening kills the measurement. A properly tuned dashpot reduces beam swing to ±0.05 division while leaving the mean reading untouched.
No. The beam has to physically swing and settle, which takes 1-3 seconds. Any data you take during a transient is contaminated by the inertia of the beam itself plus the housing pendulum mode. The reading lags the real torque by hundreds of milliseconds and overshoots when load steps occur.
For transient mapping use a strain-gauge reaction arm or an inline torque flange. Keep the balanced scale for steady-state certification points where its accuracy and traceability genuinely beat the alternatives.
Manufacturer curves are usually corrected to SAE J1349 or DIN 70020 standard atmospheric conditions. Your cell is probably hotter, higher altitude, or more humid than the reference. A 10 °C ambient rise and 50 m elevation easily account for 5-7% of indicated loss before you even look at the brake.
The remaining 1-3% is typically genuine driveline losses — coupling damping, shaft seal drag, and pump losses if there's an accessory drive in line. Apply the relevant atmospheric correction factor before declaring the engine underpowered, and re-check the torque arm length with calipers since a 5 mm error on a 716 mm arm is 0.7% of reading.
The functional spec is a contact line under 0.05 mm wide. A blunter edge rolls under load instead of pivoting, which shows up as hysteresis — load up to a target, back off to zero, and if the beam doesn't return to within one division of zero, the edges are done.
Inspection: shine a torch along the edge and look at the reflected line. A sharp edge gives a thin bright line; a worn edge gives a wide grey band. We regrind to a 60° included angle on a tool-and-cutter grinder, then stone to remove burrs. Done correctly the edges last another 10-20 years of normal lab use.
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.
