Compliance and audits
Load cell calibration: procedure, classes and intervals
Load cell calibration applies known forces to the cell, from a deadweight machine or against a reference force transducer, in steps up and down its range and records the output at each step, along with hysteresis, creep and the return to zero. ISO 376 grades force-proving instruments as class 00, 0.5, 1 or 2 and limits the calibration interval to 26 months, while ASTM E74 sets Class AA and Class A loading ranges and a limit of 24 months.
On this page
What a load cell calibration checks
A calibration applies known forces to the force measuring device and records the output at each one. The device may be the load cell alone, the cell with its amplifier or the whole chain through to the display. DKD-R 3-3 treats these as three different calibration items (3.1), because a certificate for a transducer calibrated without its amplifier only applies directly with an identical amplifier (clause 8).
The result depends on what was calibrated. A device with its own indicator gets an error of indication in force units. A transducer alone gets a transmission coefficient, such as mV/V per newton (DKD-R 3-3, clause 7). The same run also shows how well readings repeat, how far falling forces differ from rising ones, how the output drifts under a steady force and whether the cell comes back to zero.
Comparing those results with a limit and recording a pass or fail is a verification against a specification. A certificate can carry both the measured values and that statement.
The standards: ISO 376, ASTM E74, OIML R 60 and DKD-R 3-3
ISO 376:2011 is the international standard for force-proving instruments used to verify uniaxial testing machines. It applies to instruments that determine force from the elastic deformation of a loaded member and sets out how to classify them. ISO lists the 2011 edition, edition 4, as current.
ASTM E74-18(2026) is the US practice for calibrating force-measuring instruments used to verify testing machines. Most US force laboratories work to it, and NIST says it does too. NIST's NVLAP accreditation handbook (HB 150-2e2024, D.1.2) requires force labs to follow ASTM E74, ISO 376 or both.
DKD-R 3-3 is the calibration guideline of the German Calibration Service, published by PTB. Edition 03/2024, revision 1, is free. It covers devices that measure force through elastic deformation and that can be loaded up and down, with four calibration sequences graded by effort. The fullest is based on ISO 376. It no longer sets classes of its own.
OIML R 60-1:2021 is the legal metrology recommendation for load cells used in weighing instruments. It sets the accuracy classes and error limits a cell design has to meet for type approval, so it is not a routine calibration procedure.
EURAMET Calibration Guide No. 4, version 3.0 (2022), covers uncertainty in ISO 376 and other force calibrations. The Guide to the Measurement of Force, written by NPL and the Institute of Measurement and Control and re-issued in 2025, is a free practical reference.
What the ISO 376 classes mean
ISO 376 ranks a force-proving instrument as class 00, 0.5, 1 or 2, with class 00 the best. The class rests on several measured quantities together, among them reproducibility, repeatability, interpolation, zero error and creep or reversibility. To hold a class, an instrument has to meet all of that class's criteria.
The analysis can go two ways (EURAMET Calibration Guide No. 4, 6.1). An instrument classified for interpolation can be used anywhere in its range, with the force worked out from a fitted equation. An instrument classified for specific forces can only be used at the forces it was calibrated at. Check which one your certificate states before you read the cell between calibration points.
EURAMET works out the worst case for an instrument that only just meets a class. The table shows the limits behind that figure and the result, leaving out the temperature contribution. A real instrument can do better than this limit.
| Class | Expanded uncertainty of the applied force (k = 2) | Zero error | Interpolation error | Maximum expanded uncertainty of the result |
|---|---|---|---|---|
| 00 | 0.01% | 0.012% | 0.025% | 0.08% |
| 0.5 | 0.02% | 0.025% | 0.05% | 0.16% |
| 1 | 0.05% | 0.05% | 0.10% | 0.32% |
| 2 | 0.10% | 0.10% | 0.20% | 0.64% |
ASTM E74 loading ranges: Class AA and Class A
ASTM E74 works differently. It fits a calibration equation to the data and calculates a lower limit factor, or LLF, a statistical estimate of the error in forces worked out from that equation. EURAMET's guide (6.2) sets out the two standard loading ranges: Class AA starts at 2,000 times the LLF and Class A at 400 times the LLF.
For a cell with an LLF of 2 N, the Class A range starts at 800 N and the Class AA range at 4,000 N. Forces below that lower limit are outside the loading range for that class.
Deadweight or comparison: choosing the reference
A deadweight machine hangs known masses in the earth's gravitational field. The force is F = Mg(1 - ρa/ρw), where M is the true mass, g is the local acceleration due to gravity and ρa and ρw are the densities of air and of the weights (NIST HB 150-2e2024, D.1.4.1.6). A primary standard under ASTM E74 has no levers, hydraulic multipliers or counterbalance between the weights and the instrument (D.1.4.1.2).
A comparison calibration puts the cell in series with one or more reference force transducers and loads them together, usually in a testing machine or a comparator frame. It is quicker and reaches forces that a deadweight stack cannot, at a higher uncertainty. The reference needs its own current calibration. NIST uses deadweight machines up to 4.45 MN, the largest of its six, and comparison above that.
How good does the reference need to be? The Guide to the Measurement of Force (7.5.1) suggests a calibration machine uncertainty three times better than the accuracy needed from the system. ISO 376 goes further and ties each class to a maximum uncertainty of the applied force, shown in the class table above.
| Machine type | Typical expanded relative uncertainty |
|---|---|
| Deadweight | 0.005% to 0.01% |
| Hydraulic amplification | 0.01% to 0.05% |
| Lever amplification | 0.01% to 0.05% |
| Comparator with one or more reference transducers | 0.02% to 0.5% |
Set up: the whole chain, temperature, preloads and alignment
Calibrate the whole chain where you can. The Guide to the Measurement of Force (7.5.1) advises using the same cables, instruments and power supplies as in service. If the lab uses its own indicator, the certificate has to say so and interchangeability is up to the owner (DKD-R 3-3, clause 5).
Let everything reach thermal equilibrium. DKD-R 3-3 (clause 4) follows ISO 376: a stable temperature within ±1 K somewhere in the range 18 °C to 28 °C, with 19 °C to 23 °C recommended in Germany. NIST's NVLAP handbook recommends 23 °C (D.1.3.2). Keep the chain powered until it is stable, then read the zero signal of the unloaded transducer in a defined position. A change from the last calibration can reveal an overload (clause 5).
Calibrate each direction of force separately. After each mounting or change of direction, preload to the maximum force for at least 1 minute and read the zero about 30 seconds after the transducer is fully unloaded. DKD-R 3-3 asks for three preloads if the cell is designed for tension and compression, and two if it loads in one direction or the owner can state the direction of its last use (clause 5).
Load the cell along its axis and with the fittings used in service. The mounting parts count as part of the calibration item and should follow ISO 376 Annex A or copy the conditions of use (DKD-R 3-3, 3.3). NIST lists accurate alignment of the transducer axis first among its basic force calibration procedures.
Force steps: how many and in what order
Spread the force steps evenly over the range. DKD-R 3-3 (clause 5) calls for at least 3 steps per series, zero not counted, or at least 5 if a third-degree equation is wanted, such as 20, 40, 60, 80 and 100%. Wait the same time at each step, at least 30 seconds, and read no earlier than 30 seconds after the start of the force change. Record the unloaded signal before and after each series.
ISO 376 is the fullest method. EURAMET's guide (6.1) bases an interpolation classification on at least 32 deflection values, which is four increasing series at eight force steps. Its worked example in Annex A uses 10 steps and six series with the instrument at 0°, 120° and 240°.
Fewer series means leaning more on other knowledge, such as a previous ISO 376 calibration or the data sheet (DKD-R 3-3, clause 5). DKD-R 3-3 is clear that the reduced sequences cannot give a smaller uncertainty than the fuller procedure on the same item (clause 1). Approach each force from the same direction and let it settle. If you overshoot a step, go back past it and approach it again (Guide to the Measurement of Force, 7.5.2).
| Sequence | Mounting positions | Rising series | Falling series | Minimum steps (1st degree / 3rd degree equation) | Taken from prior knowledge |
|---|---|---|---|---|---|
| A | 3 | 4 | 2 | 3 / 5 | Nothing |
| B | 1 | 2 | 2 | 3 / 5 | Reproducibility |
| C | 1 | 1 | 1 | 3 / 5 | Repeatability and reproducibility |
| D | 1 | 2 | 0 | 3 / 5 | Reproducibility and hysteresis |
Hysteresis, creep and zero return
Hysteresis is the difference in output between rising and falling force at the same force. It is usually largest near mid-range (Guide to the Measurement of Force, 3.5). ISO 376 calls this quantity reversibility. A cell used only on rising forces needs no reversibility term in its uncertainty, but one that reads on the way down does (EURAMET Calibration Guide No. 4, 7.1).
Creep is the change in output with time after a step in force. Makers usually quote the largest change over a set time after going from zero to the rated force, for example 0.03% of rated output over 30 minutes. Creep recovery is the same effect after the force comes off (Guide to the Measurement of Force, 3.5). EURAMET's creep measure for ISO 376 is the change in output between 30 seconds and 300 seconds after the maximum force goes on or comes off, relative to the deflection at maximum force.
Zero return compares the signal at no load before and after a force is applied. ISO 376 reports it as a zero error, the difference between those readings relative to the deflection at maximum force (EURAMET Calibration Guide No. 4, 6.1). Compare it with earlier calibrations. Under NVLAP, an overload that changes the zero-force output by 1% or more means recalibration before further use (HB 150-2e2024, D.1.4.4.2).
- OIML R 60-1:2021 limits creep for load cells in scales: over a 30-minute test at maximum load the change may not exceed 0.7 times the maximum permissible error for that load (5.5.1).
- The change between 20 and 30 minutes may not exceed 0.15 times that error (5.5.1).
- The minimum load output has to return to within 0.5 v of its initial reading after the creep test, where v is the load cell verification interval (5.5.2).
Load cells in scales: OIML R 60 classes
OIML R 60-1:2021 ranks load cells for weighing instruments in four accuracy classes, A, B, C and D, by the number of verification intervals they can be divided into (5.1.1, 5.1.2). The common designation joins the class with that number in thousands, so C3 means class C with 3,000 intervals.
The error limits are in verification intervals and widen as the load rises. Under Table 4, a class C cell may be off by 0.5 v up to a load of 500 v, 1.0 v up to 2,000 v and 1.5 v up to 10,000 v, each multiplied by an apportioning factor that is 0.7 unless the maker declares another value. These limits include hysteresis, so they apply to rising and falling load alike (5.3).
R 60 approves a cell's design. Once the cell is inside a scale, the scale as a whole is calibrated or verified with test weights.
| Class | Lower limit | Upper limit |
|---|---|---|
| A | 50,000 | Unlimited |
| B | 5,000 | 100,000 |
| C | 500 | 10,000 |
| D | 100 | 1,000 |
In the machine or on the bench
The Guide to the Measurement of Force (7.4) describes three options. The preferred one for low uncertainty keeps the transducer permanently installed and compares it at intervals with a transfer standard placed in series. A removable transducer can go to a lab instead, which gives an accurate calibration of the cell but misses effects from the installation such as friction, lever ratios and alignment.
The third option, a permanently installed cell with no periodic calibration, relies on a large margin at the start. Periodic shunt calibration and a check of one or more points including zero give some confidence, but any traceability claim has to account for the uncertainty of the shunt calibration.
A laboratory gives stable reference conditions. In situ work is chosen for convenience or cost, or to calibrate under the exact conditions of use (7.5.1). Whichever you choose, match the end-loading conditions of the application as closely as you can.
As-found readings, adjustment and a load cell out of tolerance
Calibration and adjustment are separate operations. Agree first whether the device may be adjusted and ask for results as found as well as after adjustment, since the Guide to the Measurement of Force (7.5.1) warns that as-found information may be vital for the traceability of the process.
If an as-found result is outside its limit, the cell was out of tolerance for an unknown part of the time since its last good calibration. ISO 9001 clause 7.1.5.2 expects you to look back at what it measured and deal with anything that may be wrong. For a load cell, look at the testing machines verified with it, the test results those machines produced and any weighing or product decision that relied on its readings.
How often to calibrate a load cell
ISO 376:2011 allows up to 26 months once stability has been shown. ASTM E74-18 limits the interval to 24 months, using stability criteria of 0.032% for Class AA and 0.16% for Class A, as force metrologists at Egypt's National Institute of Standards set out in a 2026 paper on the validity of force measurements. Those are ceilings, not targets.
The Guide to the Measurement of Force (7.5.1) says to start from the maker's long-term stability specification. Lengthen the period if several calibrations show no appreciable change in sensitivity and shorten it if they show significant change. Harsh handling, frequent or long use and the environment can all call for a shorter interval, and the choice rests with the user.
Recalibrate sooner, whatever the date, after a repair or modification that may change the calibration curve and after an overload that shifts the zero output by 1% or more (NIST HB 150-2e2024, D.1.4.4). Between calibrations, the Guide suggests periodic checks of the zero and of a well-defined point. Record the reason for any change to the interval. ILAC-G24:2022 / OIML D 10:2022 lists the factors to weigh, among them drift and the severity of use (5.1).
What the certificate and record should show
A force calibration certificate follows ISO/IEC 17025:2017 clause 7.8 and adds force-specific content, listed in DKD-R 3-3 clause 8 and NVLAP HB 150-2e2024 D.1.6. Keep the same details for an in-house calibration so that someone else can repeat it.
- The identity of the transducer or chain, and the force introduction fittings used.
- The standard or sequence followed, such as ISO 376 or DKD-R 3-3 sequence B, and the source of any contribution taken from prior knowledge.
- The force calibration machine used and its uncertainty.
- The direction of force and the temperature, with its limits during the calibration.
- The supply voltage, the wiring (4-wire or 6-wire), the cable length and the amplifier settings.
- The forces applied, the deflections, the fitted equation and its deviations, and the uncertainty.
- The zero signal of the unloaded transducer.
- Where a class or conformity statement is given, the rule used and the origin of the limits.
- The as-found and as-left results, the date, who did the work and the next due date.
How to calibrate a load cell, step by step
- Identify the cell or chain, its capacity, the force direction and the standard you need, such as ISO 376, ASTM E74 or a DKD-R 3-3 sequence.
- Choose a deadweight machine or reference transducer whose uncertainty suits the class or limit you need.
- Inspect the cell and check its nameplate, electrical function and zero signal.
- Mount the cell with the fittings used in service and align the force with its axis. Let it reach thermal equilibrium.
- Warm up the chain, then preload to the maximum force at least twice, for at least 1 minute each time, and read the unloaded zero.
- Apply the force steps evenly up the range, waiting the same time at each and at least 30 seconds.
- If the cell reads on falling forces, repeat the steps coming down. Repeat series and mounting positions as your sequence requires.
- Work out the error or coefficient at each step, the hysteresis, the creep and the zero return. Compare them with the limit and record these as-found results before any adjustment.
- If the cell fails, adjust or repair it, repeat the steps as left and assess what it measured since its last good calibration. Then set the next due date.
Where Axiospec fits
Axiospec is calibration management software. It keeps each load cell and force transducer with its calibration history, certificates and next due date, and each calibration names the reference standard used. One calibration record holds up to 100 test points. Each point has its nominal value, its tolerance as an absolute band, a percentage of nominal or a percentage of span, whether it was read up-scale or down-scale, and the as-found and as-left readings. The certificate Axiospec produces lists those points.
On an in-house calibration, a final reading outside its tolerance sets the result to Fail, and recording a pass anyway takes a written reason. A Fail puts the load cell in quarantine. A Fail, or a Pass with an as-found reading out of tolerance, asks you to write down what the cell measured since its last good calibration, and any not yet written show in Needs attention. Axiospec does not run the calibration, fit the curve or pick the class or the interval. Those stay with you or your lab.
The free plan covers up to 50 active instruments. A new workspace starts with a 45-day trial of Professional, with room for 1,500 active instruments and no credit card. We load your existing records for you at no charge on every plan.
Common questions
- How do you calibrate a load cell?
- Apply known forces from a deadweight machine or against a reference force transducer, in evenly spaced steps up the range and back down if the cell reads on falling forces. Record the output at each step and the unloaded zero before and after. Compare the error, the hysteresis and the zero return with the limit, and record the as-found results before any adjustment.
- How often should load cells be calibrated?
- ISO 376 allows up to 26 months once stability has been shown and ASTM E74 up to 24 months against its stability criteria. Start from the maker's long-term stability figure and adjust from the cell's own as-found history. Recalibrate sooner after a repair or an overload that shifts the zero output by 1% or more.
- What is the difference between ISO 376 and ASTM E74?
- ISO 376 assigns class 00, 0.5, 1 or 2 from several measured criteria across mounting positions. ASTM E74 fits a calibration equation and calculates a lower limit factor, and the Class AA and Class A loading ranges start at 2,000 and 400 times that factor. NIST's NVLAP handbook notes that the two standards have different requirements.
- What do ISO 376 classes 00, 0.5, 1 and 2 mean?
- They grade force-proving instruments from best to worst on criteria that include repeatability, reproducibility, interpolation, zero error and creep or reversibility. EURAMET's worst-case figures for an instrument that just meets each class are 0.08% for class 00, 0.16% for 0.5, 0.32% for 1 and 0.64% for 2, as expanded uncertainty.
- Is deadweight or comparison calibration better for a load cell?
- Deadweight machines give the lowest uncertainty, typically 0.005% to 0.01%, because the force comes directly from known masses. Comparison machines use reference transducers, typically reach 0.02% to 0.5%, and can go to higher forces. Choose the method whose uncertainty fits the class or limit you need.
- What is creep in a load cell?
- Creep is the change in output over time while the force stays constant after a step. Makers usually quote it as the largest change over a set time, such as 0.03% of rated output over 30 minutes. OIML R 60 limits it for load cells in scales with a 30-minute test at maximum load.
Put it into practice
Got a gage list? Send it over and we load it for you, usually in a couple of business days. Free on every plan. Then log every calibration to a tamper-evident audit trail and produce records on demand.
Axiospec is a documentation and workflow tool. It helps you keep clean, traceable, audit-ready records. Certification depends on your own processes, scope and assessor.
Take the audit-prep checklist with you
First the checklist, with each item phrased the way an auditor asks it. Then four short emails with practical tips. You can unsubscribe in one click.
First we email you a link to confirm your address. Nothing else comes until you confirm, and you can unsubscribe from any email after that. See our Privacy Policy.