Home
Hydraulic dynamometer
Measurement integrity / torque reference
Torque calibration establishes the relationship between the torque a dynamometer reports and a torque that can be traced to a reference. Without that relationship, a torque number is a reading, not a measurement, and every power figure derived from it inherits the same uncertainty. The method is old and unglamorous: apply a known force at a known distance, and compare what the system says.
This guide walks through the static calibration principle, the checks worth performing on each point, the mechanical and electrical error sources that survive a good procedure, and the records that make a calibration defensible a year later. It does not state a Hance calibration interval, an acceptance tolerance for Hance equipment or a procedure approved for a specific machine. Those belong to the equipment documentation, the applicable standard and the facility's quality plan.
01 / WHY IT MATTERSWhy torque calibration decides what the report means
Torque is the anchor value
Speed is easy to verify independently, and temperature and pressure can be cross-checked against a second instrument without much trouble. Torque is different: in most test cells it is the one quantity with no natural second opinion. If the torque channel is offset, the power curve is offset with it, the specific fuel consumption is distorted, and every comparison against a previous build carries the same error.
That is why calibration is not an annual formality attached to an instrument list. It is the step that turns a transducer output into a statement a customer or an authority can examine. EPA's engine-testing regulation makes the intent explicit at 40 CFR 1065.310: calibrate all torque-measurement systems, including dynamometer torque transducers, on initial installation and after major maintenance, and repeat the calibration using good engineering judgement.
Static calibration is a defined scope, not a full guarantee
ASTM E2428-22, the practice for elastic torque measurement standards, states its own limit plainly in its scope: the practice covers static elastic torque standards, it is not applicable to dynamic or high-speed torque calibration, and results obtained under it cannot be assumed valid for dynamic or high-speed measurement.
That sentence should be pinned above the calibration bench. A dynamometer spends most of its life measuring a torque that is changing, often with torsional oscillation on top of the mean value. Static calibration fixes the scale; it does not by itself prove the behaviour of the whole chain under acceleration or oscillation.
02 / REFERENCE TORQUEStatic torque calibration: force times lever arm
The reference torque comes from a force and a length
The classical method applies a known force at a measured distance from the rotation axis. EPA's recommended procedure quantifies the lever arm from the centreline of the dynamometer to the point where the reference force is measured, requires the arm to be perpendicular to gravity and to the dynamometer's rotational axis, and requires the arm length to be traceable within ±0.5% uncertainty.
The arm's own unbalance matters. The regulation asks for the lever arm's torque to be balanced or for its net hanging torque to be quantified and then accounted for as part of the reference torque, traceable within ±1%. A perfectly measured arm that is left unbalanced produces a systematic offset at every point.
Dead weights, or a reference transducer
With dead weights, the reference force is the mass multiplied by the local acceleration of gravity, and the regulation requires at least six weight combinations for each applicable torque range, spaced about equally across it, with the dynamometer oscillated or rotated during calibration to reduce frictional static hysteresis.
The alternative is a strain-gauge load transducer or proving ring. Here the reference torque is the reference meter's force multiplied by its effective lever arm, measured from the point where the force is measured to the rotational axis, perpendicular to both the meter's measurement axis and the rotational axis. The same six-point and hysteresis guidance applies.
| Element | What it contributes | Where it usually goes wrong |
|---|---|---|
| Lever arm length | Directly proportional to reference torque | Measured to the wrong reference point, or assumed from a drawing |
| Perpendicularity | Cosine error in the applied torque | Arm not horizontal, or not square to the rotation axis |
| Tare / unbalance torque | Constant offset across all points | Ignored because the arm looks balanced |
| Weight value | Local gravity instead of nominal g | Using standard gravity at a site where it differs |
| Friction and hysteresis | Different reading on load and unload | Reading only one direction of travel |
| Point spacing | Ability to detect non-linearity | Clustering points near full scale |
Point spacing
The six-point requirement is not a formality. Geometry and mechanism behaviour rarely produce a perfectly straight line, and equally spaced points are what let a reader see the shape instead of inferring it from two ends.
03 / CHECKSZero, span, linearity and direction checks
Zero is a measurement, not a reset
Zero is checked with no applied torque, after the system has reached its operating condition. It drifts for ordinary reasons: bridge and amplifier thermal effects, cable and connector resistance, and mechanical settling after the arm is mounted. HBK's torque sensor guidance lists temperature effect at zero as one of six components of a sensor's uncertainty, which is a reminder that zero is a quantity with its own behaviour rather than a button.
The practical discipline is to record zero before and after the loaded points, with the machine in the same state. A zero captured cold and compared with a span captured warm describes two different instruments.
Span fixes the slope
Span sets the output change per unit of torque. A full-scale point alone cannot separate a gain error from a linearity error, which is why the equally spaced points in the reference procedure matter.
Linearity and hysteresis need both directions of travel
Linearity compares the measured output with the ideal straight line between zero and full scale. Hysteresis compares the reading at the same torque reached from below and from above. Both are invisible in a single ascending sweep, and both are exactly what a dynamometer with bearing drag or a sticky linkage will show.
EPA's procedure addresses this at the source by oscillating or rotating the dynamometer during calibration, so the mechanism settles instead of storing a static friction offset. A calibration that skips that step can look excellent and still misread small torques in service.
Direction and sign
A torque channel has a sign convention. Reversed sign, or a magnitude that is correct while the sign is wrong, produces a plausible curve with the power flowing the wrong way. Check the sign at a known load in each direction, and record the convention on the certificate rather than in someone's memory.
04 / ERROR SOURCESMechanical and electrical error sources
Mechanical sources
The lever arm is the largest single mechanical influence, because its length enters the reference torque directly. Perpendicularity failures add a cosine error that grows quickly with angle. Unbalanced arm torque adds a constant. Friction in bearings and linkages adds hysteresis. Off-axis loading, coupling forces and thermal expansion of the mounting structure add smaller, harder-to-see terms.
A reaction-type torque measurement deserves separate mention. It senses the torque reacted by the absorber housing, filtered through the machine's mass, rather than the true shaft torque. Calibrating that sensor is still meaningful, but the certificate describes the sensor and its mount, not the shaft.
Electrical sources
The electrical chain contributes its own offsets: bridge excitation drift, contact resistance in connectors and slip rings, amplifier zero and gain, and the acquisition channel's own scaling. Screening and grounding matter in a cell with a large absorber and variable-frequency drives, because induced noise on a low-level bridge signal is easily mistaken for torque ripple.
Filtering introduces a subtler error. A filter that steadies the display also delays the value, so a filtered torque channel compared against an unfiltered speed channel carries a timing error as well as an amplitude one. Record the filter setting with the calibration; otherwise the next engineer cannot reproduce the comparison.
Uncertainty is a set of terms, not one percentage
HBK's six components, output and sensitivity, linearity and hysteresis, temperature effect at zero, temperature effect at full scale, repeatability and off-axis loading, describe the shape of a realistic uncertainty statement. A single datasheet percentage hides which term dominates in a particular installation.
05 / RECORDSCalibration records and acceptance criteria
As-found and as-left are different records
NIST's guidance on calibration intervals makes a practical point that applies directly here: recalibration reports should carry the as-submitted and post-calibration accuracy data, because those two records are what allow an interval to be validated or refined. An as-left number alone cannot tell anyone whether the instrument drifted or was always wrong.
For a torque system, the as-found record should state the condition in which it was taken, including zero behaviour, and the as-left record should state what was adjusted. If nothing was adjusted because the result was inside requirements, that is also a fact worth writing down.
Acceptance criteria come from requirements
There is no universal acceptable torque error. The criterion belongs to the test programme: what the customer asked for, what a regulation requires in that application, what the facility's quality plan states for the equipment class. A calibration certificate reports results and uncertainty; the decision that a result is acceptable is made against a stated requirement.
This is the point where a well-meaning checklist does damage. Writing a tolerance into a procedure because a similar machine elsewhere used it produces a document that looks authoritative and has no basis for this installation.
Intervals are argued from evidence
NIST does not require or recommend a single recalibration interval and points to accuracy requirements, contractual or regulatory requirements, the instrument's inherent stability and environmental factors as the inputs. In practice a facility sets an initial interval, then refines it from as-found history.
Event triggers sit alongside the calendar. A suspected overload, a repair, a relocation, a change of amplifier or acquisition hardware, or a set of results that conflicts with a known-good reference all justify an earlier assessment. Waiting for a due date in those cases protects the schedule, not the data.
| Record element | Why a later reader needs it |
|---|---|
| Transducer and system identification | Ties the certificate to one installed chain |
| Reference standard and its traceability | Establishes what the values were compared against |
| Method and point schedule | Shows whether non-linearity could have been detected |
| Direction, filter and gain settings | Reproduces the condition of the measurement |
| As-found and as-left results with uncertainty | Supports both the disposition and the interval |
| Disposition and any restriction | States whether the system may be used, and how |
Close the loop with the tests that used the instrument
A calibration certificate filed away from the test records helps nobody. The calibration date and status belong with the results that depended on it, so a questionable number stays traceable to the state of the instrument at the time of the run. The engine dyno test report guide applies the same discipline to reporting, and the data acquisition channel and timing guide applies it to timing; calibration is the third leg of the same traceability. Add the internal link to that guide once it is published.
Where the installation itself is the variable, the hydraulic dynamometer maintenance checklist covers the condition and release side, and the engine dyno testing procedure covers the run setup that the calibration should accompany.
Share the torque sensor, reference standard and the accuracy your test programme needs, and Hance can help define a calibration and verification plan that fits the machine you actually run.
Discuss your torque calibration plan
Questions and answers
How often should torque be calibrated?
There is no single correct interval. NIST declines to prescribe one and lists accuracy requirements, contract or regulation, instrument stability and environment as the inputs. The practical answer is an initial interval from the manufacturer and the test programme, refined from as-found history, plus an immediate reassessment after overload, repair, relocation or unexplained results.
Why can zero drift after warm-up?
Because the zero is affected by temperature. Bridge and amplifier thermal effects, connector resistance and mechanical settling all move the zero as the machine reaches operating condition, which is why temperature effect at zero is listed as its own uncertainty component. Record zero cold and warm in the same state, and do not treat an early reading as the final one.
How is calibration uncertainty reported?
As a combination of terms rather than one percentage: sensitivity, linearity and hysteresis, temperature effects at zero and at full scale, repeatability and off-axis loading are the usual components. The reported uncertainty should say which reference standard and method produced it, so a reader can judge whether it suits the application.
Can a static calibration be used for dynamic measurements?
Only with care. ASTM E2428-22 limits itself to static elastic torque standards and states that its results cannot be assumed valid for dynamic or high-speed torque measurement. Static calibration fixes the scale and the traceability; dynamic behaviour, including torsional oscillation and filter delay, needs separate consideration.
What should a calibration record contain?
Identification of the transducer and the whole measurement chain, the reference standard and its traceability, the method and point schedule, direction and filter settings, as-found and as-left results with uncertainty, and a clear disposition stating whether and how the system may be used. Without the as-found record, no later interval decision can be supported.
Technical references
EPA — 40 CFR 1065.310, Torque calibration: lever-arm and reference-force requirements, six-point scheduling, hysteresis reduction.
ASTM E2428-22 — Calibration and Verification of Elastic Torque Measurement Standards: scope limited to static elastic torque standards; not applicable to dynamic or high-speed measurement.
NIST — Recommended Calibration Interval: interval inputs, as-submitted and post-calibration data.
HBK — How to select a rotating torque sensor: six uncertainty components, reaction versus in-line sensing.
This article describes general torque-calibration practice. It is not a Hance calibration procedure, not an approved interval or tolerance, and not evidence that any specific machine has been calibrated. Use the equipment documentation, the applicable standard and the facility's quality plan, with competent personnel performing the work.