Eccentricity and Linearity Check for Balance Calibration

Eccentricity and Linearity Check is a paired balance assessment of position-dependent weighing error and the proportionality of indication to applied load across a selected range. The two checks ask independent questions: does moving an unchanged load change the indication, and does changing the load reveal a changing response error? A satisfactory result for one does not establish a satisfactory result for the other.

In operational terminology, an eccentricity test is also called an off-center load test or corner load test. “Corner” describes a possible placement pattern, not every load receptor: select positions appropriate to its geometry and intended loading rather than imposing a corner-based pattern universally.

For interpreting load-response results, the technical definition in OIML R 76-1, T.5.5.1 treats indication error as the displayed value less the corresponding conventional true mass. A routine multi-point error check examines these differences at selected loads. That is not necessarily a formal nonlinearity calculation, which evaluates departures from a specified reference line. The procedure needs to identify the quantity evaluated and its acceptance criteria; a table of errors alone does not identify the reference line.

In the worked measurement tables, example masses and indications are expressed in grams; other examples state their units where used. Positions, loads, intervals, and limits belong to hypothetical procedures. Arithmetic omits uncertainty and rounding corrections and does not reproduce a complete OIML verification.

Conclusions remain bounded by the tested instrument, configuration, positions, load range, and procedure. Passing does not establish suitability for every weighing task; assess the balance capability relative to process tolerance separately. Keep these check results distinct from the broader as-found and as-left calibration record.

Select the activity by the question it must answer. These are operational distinctions for planning balance work, not formal legal definitions.

A successful single-load routine check answers a narrow question: whether that result meets its criterion. It cannot substitute for the paired assessment above because neither position sensitivity nor response across multiple loads has necessarily been examined. Define the applicable weighing tolerance band before deciding acceptance.

Keep observing response separate from changing it. The lifecycle section addresses pre-adjustment and post-adjustment results and return-to-use decisions.

Start with a position map, not an assumed number of corners. Describe the load-receptor segments to be tested and how the load will occupy each segment. Keep the segment layout, chosen load and comparison rule explicit; do not treat a particular position count or fraction of capacity as universal.

OIML segments versus a local centre reference

The technical test method in OIML R 76-1, A.4.7 distinguishes two arrangements: place a single weight centrally within the segment, or distribute several small weights uniformly across it. That provision permits testing only the eccentric segments, without measuring at the centre of the load receptor, and calls for a sketch showing the load location in the test report.

A site's centre-referenced routine check is a separately defined method, not a restatement of that OIML segment test. Before collecting readings, define the comparison as the off-centre indication minus the centre indication, and specify whether acceptance applies to each absolute difference. Document the local tolerance band alongside that metric.

Worked case: five locally specified locations

Assume a local five-location procedure, an unchanged 100.000 g load and a displayed increment of 0.001 g. Its hypothetical criterion is that every off-centre absolute difference must be no greater than 0.004 g.

For example, the front-left difference is 100.006 − 100.001 = +0.005 g. The maximum absolute difference is therefore 0.005 g. Front-left fails the local criterion; rear-right is exactly at the accepted boundary. Do not replace this calculation with indication error against 100.000 g: that would give +0.006 g at front-left and answer a different question.

The result establishes that one tested off-centre position exceeded the permitted departure from the recorded centre reading under these test conditions. It does not establish performance at untested positions or loads, nor identify a mechanical cause. Preserve the location map and individual readings for the site's nonconformance or deviation assessment, rather than recording only the maximum difference.

A multi-load check examines indication error at several loads; a one-point check establishes a result only at that tested load. For a local operating check, define the load points in the procedure across the intended operating range, including loads relevant to actual weighing work. Do not assume that equal spacing alone provides useful coverage.

For the scope and sequence of the formal weighing-test procedure, check OIML R 76-1, A.4.4.1. Do not label every local multi-point check as full execution of that procedure. Specify the local loading sequence, any comparison between increasing and decreasing loads, and the acceptance criteria before testing; no universal point count is prescribed here.

Worked case 2: an intermediate load fails

Assume the assigned reference loads and recorded indications below, all in grams. This hypothetical procedure applies an inclusive ±0.005 g criterion independently to each result, with no additional uncertainty allowance in its decision rule. Indication error = indication − assigned load. The four points are example choices, not a prescribed test design.

At 100.000 g on increasing load, 100.006 − 100.000 = +0.006 g: this exceeds the positive limit by 0.001 g and fails. The other seven listed results pass. The largest absolute difference between loading directions is also at 100.000 g: |100.006 − 100.004| = 0.002 g. That difference is a separate comparison, not another pointwise error; no direction-difference acceptance limit is assumed in this example.

These calculations are pointwise errors, not a fitted-line estimate of nonlinearity. A fitted-line assessment would need a defined fitting method and a separate interpretation of residuals. Here, neither averaging the errors nor passing the highest tested load removes the failed intermediate result. Retain each indication with its assigned load and loading direction, including when using a data logger, so reviewers can reconstruct the comparisons rather than seeing only a summary pass rate.

As site operating practice, distinguish three scheduling routes for eccentricity and linearity checks:

• Scheduled routine checks: recurring checks intended to monitor performance between calibration events.
• Calibration-associated checks: checks included in the calibration procedure, with their timing relative to adjustment identified in the as-found/as-left record.
• Event-triggered checks: checks selected after a relevant change or event, with scope justified by the potential effect on weighing.

Document interval selection as a site decision, subject to applicable requirements. Consider intended use, consequences of weighing error, operating range, handling, environment, and observed performance. Daily, weekly, and monthly are possible schedule labels—not recommended defaults or regulatory frequencies.

Service, adjustment, altered operating conditions, or an adverse trend should prompt review of whether the existing schedule remains appropriate. Use out-of-trend handling to connect performance observations with that reassessment; do not assume every event warrants the same tests.

Name a schedule owner, an authorised interval-change approver, and the person responsible for executing due checks. Define how due dates or triggering events are logged, escalated when missed, and linked to a documented equipment-use decision.

Support interval changes with dated results and trends, service history, changes in use or environment, a risk rationale, approval, and a review date. Handle actual failed results through the test-event lifecycle and recovery decision, not by substituting additional scheduled testing.

Use the following as a practical workflow, not a universal regulatory checklist. Build one event record connecting the balance’s condition, measurements, acceptance decision, and subsequent actions.

Prepare and preserve the starting condition

Identify the instrument, intended weighing use, procedure revision, reference standards, operator, and event date. Record configuration and setup observations before intervention: location, displayed settings, level condition, pan condition, and any observed disturbance. Define the planned measurements, calculations, acceptance rule, and reviewer. Keep pre-adjustment results separate from results following adjustment or service, using an explicit as-found/as-left record.

For every planned position or load point, capture its identity, reference load, indication, unit, sequence, and relevant zero or tare state. Mark omitted or interrupted measurements rather than silently closing gaps. Link each calculation to its observations and the acceptance rule actually applied, so review can reconstruct why the event passed or failed.

Worked case 3: recover from the failed intermediate point

Assume the preceding example’s intermediate point failed the approved acceptance rule and the balance has supported production weighing. Retain that result without repeating its arithmetic. Place the instrument in a controlled unavailable or restricted-use status under the local procedure, identifying who imposed the status and which uses it prevents.

Assess test validity before choosing a remedy. A documented execution problem, such as a confirmed mismatch between the planned and applied reference weight, may justify invalidating that measurement with a recorded reason. An unexplained failure remains an adverse result; a later pass does not erase it. Follow the applicable out-of-specification handling process, investigate the cause, justify corrective action, and repeat testing under an approved scope that addresses the findings—not merely the point most likely to pass.

Review potentially affected weighing uses proportionately. Use actual instrument-use records and linked batch execution history to identify loads, materials, batches, and dates. Bound the review using the investigation’s findings and relevant prior performance evidence, documenting uncertainty rather than assuming a fixed retrospective period.

Assign authorised people responsibility for return-to-use approval and any related material or batch disposition. Record their rationale and restrictions. Give software a supporting role in routing and status control, not decision authority.

Build the applicability map before approving the balance procedure. The table below is an operating-practice framework: it identifies questions to resolve, not requirements that apply to every balance. Record the product, jurisdiction, instrument type, and weighing use so each procedural control has a traceable basis.

Use the 21 CFR 211 overview and 21 CFR 111 overview for product-context navigation, not as interchangeable rule sets. Capture software needs separately in the user requirements specification. Review the map when the product scope, jurisdiction, weighing use, or referenced edition changes.

Treat the following as proposed implementation requirements to evaluate against the approved procedure and local record controls, not as regulatory requirements or claims about any software product.

Connect observations to their interpretation

Give each test event a unique identifier linking the balance asset, procedure revision and reference standards used. Link each observation to that event, its load point or pan position, and the relevant reference standard. Attach the authorised reviewer’s decision to the specific observation set and evaluation version reviewed.

Keep original indications separate from derived errors and pass/fail evaluations. Each calculation should identify its input observations, reference value and unit, comparison metric, acceptance rule and procedure revision. Record whether the metric compares an indication with a reference value or with another indication. Link the applicable weighing tolerance band to the evaluation rather than embedding an unexplained threshold in an interface.

Specify both interface boundaries

For instrument capture, specify explicit instrument identity, raw value and unit, capture time, observation sequence, and stable-reading information where the instrument makes it available. Identify unavailable stability information explicitly rather than inferring stability from receipt.

For record-system transfers, evaluate controlled unit conversion, duplicate detection, reconciliation of expected and received observations, and preservation of the original payload. Retain converted values alongside their conversion rule; do not overwrite the source value. Define acknowledgements and exception ownership within the integration design.

Recover without obscuring provenance

After interruption, distinguish an unreceived record from an unperformed measurement by reconciling source records with destination acknowledgements before retransmission. Do not treat a missing destination row as proof that weighing never occurred. Retain a visible link between the original record and any corrected interpretation, including the reason and review decision.

Identify controlled manual entry as a separate capture route, with operator identity and a documented checking step. Preserve that distinction during exports and review.

Use these recommended procedural controls to make execution weaknesses visible before accepting an eccentricity or linearity result. They are practical review checks, not a regulatory checklist; acceptance remains tied to the applicable approved procedure.

Keep suspected causes separate from established investigation findings. A position-dependent result alone does not establish a particular component fault. Record the hypothesis, the observations used to test it and the basis for the authorised reviewer’s conclusion; do not relabel a passing repeat as proof of the original cause.

Use these practice-based metrics for review, not as proof of continued suitability.

• First-attempt acceptance rate = events accepted on their original complete run ÷ events with a complete original run × 100. A successful repeat does not replace the original outcome.
• On-time completion rate = events completed by their due time ÷ events due × 100. Define the reporting period and completion milestone consistently.
• Exclude cancelled or superseded events from the on-time denominator only when authorised before their due time; retain later cancellations in it. Report exclusions separately and preserve links to replacement events.
• Show failed, incomplete, invalidated, and repeated events separately in the process event log, alongside numerator and denominator counts; identify overlapping statuses.
• Trend actual worst-position differences or pointwise errors only among comparable procedures and configurations, retaining units, load points, and position identities.

Use exception age and recurring failure locations or load points as trend-review signals, not automatic evidence that a check interval can be extended.

Use these questions as a practical readiness review, not as a substitute for the approved procedure. Refer to the method, recovery, regulatory and evidence discussions above for the detailed controls.

• Is the intended weighing use defined, including the operating range and relevant loading conditions?
• Is the chosen metric named, with its units and calculation reference?
• Are approved positions or load points unambiguous and appropriate to that scope?
• Are acceptance criteria and decision rules approved before execution?
• Is reference-standard suitability established for the planned measurements?
• Does a named procedure owner control revisions and resolve method questions?
• Do event records identify the instrument, procedure revision, operator and test sequence?
• Are original observations preserved alongside derived results and subsequent corrections?
• Are interface reconciliation, interrupted transfers and incomplete events assigned defined handling paths?
• Can invalidated readings and repeat measurements be traced to their reasons and original event?
• Can authorised reviewers distinguish execution validity, measurement acceptance, investigation status and the return-to-use decision?
• Is regulatory applicability recorded for the actual jurisdiction, product and weighing use rather than inferred from this glossary term?

Resolve unanswered questions before adopting the workflow. Where the review identifies an unresolved exception, assign an owner and disposition rather than treating a completed checklist as permission to resume weighing.

Does OIML specify one position pattern for every instrument?

No. The technical recommendation distinguishes load-receptor arrangements. For instruments with no more than four support points, OIML R 76-1 A.4.7.1 specifies loading four quarter segments in turn. For more than four support points, A.4.7.2 specifies loading over each support. These are arrangement-specific instructions, not a universal corner-count rule. When adopting a local method, identify the applicable instrument arrangement before approving its position map. Keep that technical-method choice separate from the jurisdiction and product applicability assessment discussed above.

Can two procedures use the same numerical acceptance limit?

Compare the definitions before comparing the numbers. A procedure might evaluate departure from a reference value, departure from a centre-position reading, or the spread between observations. Treat those as separate calculations unless the approved method explicitly connects them. During procedure transfer, check the units, sign convention, rounding and treatment of equality at the limit. Copying a number alone leaves the decision rule ambiguous. For changed calculations, use a controlled review approach such as recalculation review and sign-off, retaining both the original result and the revised interpretation.

Should the calculation use nominal or assigned reference-weight values?

Make that choice explicit in the procedure rather than leaving it to the operator or spreadsheet designer. For the local method, identify the reference value to enter, its units, and any approved treatment of certificate information. Check that the worksheet and instrument interface use the same convention. When several weights form a load, preserve their identities and the calculation connecting them to the entered reference value. If the convention changes, review affected calculations through change control instead of silently replacing historical reference values in completed events.

Can a rarely used balance have a different check schedule?

Consider a different schedule through the approved scheduling review, not through an informal exemption. Document the intended use, storage conditions, relevant history and events that prompt reassessment before use. Define who reviews the schedule and what information would justify changing it. Low usage alone should not serve as the entire rationale. If observations remain acceptable but develop a concerning pattern, the out-of-trend handling process can structure the review. Keep the scheduling decision separate from the acceptance decision for any individual event.