Laboratory Analytical Balance Monthly Span, Linearity & Repeatability Verification Log

ContextSpan calibration evaluates the balance's conversion factor between mechanical deflection (or electromagnetic force in EMFC designs) and the displayed mass unit. Testing at or near full capacity matters because span error is proportional — a 0.01 % error at 200 g is only 20 mg, but uncorrected it applies a constant percentage offset to every measurement the instrument makes, regardless of actual sample mass. Use the certified mass value from the weight's calibration certificate, not the stamped nominal value, which may differ by several tenths of a milligram at this grade. Center placement is critical; off-center loading on some balance designs introduces a corner-load artifact that mimics span error and cannot be distinguished from it without further testing.

ContextSpan Error (mg) = Displayed Reading − Certified Weight Value. Span Error (%) = (Span Error in mg / Certified Weight Value in mg) × 100. Most laboratory SOPs set a maximum permissible span error of ±0.1 % for routine verification, but GMP-regulated pharmaceutical environments often tighten this to ±0.05 % or ±0.03 % for primary weighing operations that feed batch release decisions. A critical distinction: running the balance's internal motorized calibration function (AutoCal, FACT, or equivalent) is a self-adjustment, not a verification — this external weight check confirms that the self-adjustment was correct. Record both the raw reading and the calculated error in the log before removing the weight.

ContextThe acceptance criterion must be established in your Standard Operating Procedure or equipment qualification protocol before testing begins — setting it retrospectively after seeing a result constitutes a data integrity violation under all major regulatory frameworks. If the span error exceeds the limit: first repeat the measurement once to rule out a display-settling artifact; if the failure is confirmed, immediately tag the balance Out of Service, quarantine data generated since the last passing verification, and initiate a service call with an accredited metrology provider. Span failure often indicates a contaminated or degraded load cell, a firmware drift, or long-term creep requiring certified external recalibration. Do not simply run the internal calibration function and re-verify without documenting the deviation.

ContextA single span check at full capacity reveals nothing about the balance's behavior at intermediate loads, where most practical weighings occur. Linearity testing maps the full response curve of the instrument. For a 220 g balance with 0.1 mg readability, suitable test points might be 50 g, 100 g, 150 g, and 200 g. Where USP Chapter <41> or OIML metrological regulations apply, a minimum of five test points is required — confirm the specific requirement in your SOP before starting. Stacking calibrated weights to reach each test load is acceptable provided the combined certified mass of each stack is known and documented. Plan weight combinations in advance and calculate their total certified values before the session to avoid arithmetic errors under time pressure.

ContextLinearity Error at each point = Displayed Reading − Certified Mass at that point. A perfectly linear balance would show zero error at every point; in practice, a small and smoothly progressing deviation across the range is expected and acceptable. What requires flagging is non-monotonic behavior — for example, an error of +0.3 mg at 100 g but −0.5 mg at 150 g — which indicates a mechanical or electronic non-ideality rather than simple proportional drift. The maximum acceptable linearity error should not exceed the balance's rated Maximum Permissible Error (MPE), which for Class I instruments is typically ±1 scale division. Enter all results in a tabulated format in the log; do not interpolate, omit inconvenient readings, or round intermediate values before calculating the error.

ContextHealthy balances show a linearity error curve that is essentially flat (excellent) or gently and smoothly bowed — a slight parabolic shape is common in EMFC designs as the coil approaches saturation near full capacity, and this is normal within specifications. A sharp inflection between consecutive test points, a sudden jump, or a curve that reverses direction is abnormal. These patterns suggest mechanical interference (debris on the pan guide, a bent draft shield intermittently contacting the pan), a damaged load cell sector, or a compensation error in the balance firmware. Document the observed profile shape even if every individual point passes its tolerance limit; a profile that progressively worsens across three or four consecutive monthly verifications justifies preventive service before the instrument reaches an outright failure.

ContextRepeatability testing is the most direct measure of the balance's precision under normal operating conditions. Use a single stable OIML E2 or F1 weight (same handling rules as for span and linearity) and load–unload–reload it ten times, recording each displayed reading after stabilization. Do not press the zero or tare button between loads — re-zeroing between readings masks drift and inflates apparent precision by resetting the baseline before each measurement. Ten readings is the minimum sample size for a statistically meaningful standard deviation estimate; USP Chapter <41> explicitly specifies ten repetitions for balance qualification. If ten repetitions are impractical, a minimum of six is defensible, but any departure from the standard protocol must be documented and justified in the SOP.

ContextMean (x̄) = Σxi / n. Standard Deviation (s) = √[Σ(xi − x̄)² / (n−1)]. %RSD = (s / x̄) × 100. Always apply the n−1 denominator (sample standard deviation) rather than n (population standard deviation) — using n systematically underestimates variability in small samples. For a 0.1 mg readability balance, typical acceptance criteria are s ≤ 0.1 mg (one scale division) or %RSD ≤ 0.10 % at a 50 % capacity test load, but verify against your specific SOP. A %RSD approaching 0.2 % on a 0.1 mg instrument at moderate load is a warning-level result; values above 0.5 % indicate a clear failure requiring investigation. Elevated %RSD with stable environmental conditions almost always points to a mechanical cause: a dirty load cell bearing, a worn flexure, or a contaminated pan guide.

ContextA single outlier — one reading far beyond the cluster of others — can inflate the standard deviation and cause a false failure, or if silently discarded, can conceal a genuine intermittent fault. To apply the Grubbs test: G = |x_suspect − x̄| / s; compare the result against the critical value for n=10 at α=0.05, which is G_crit ≈ 2.29. If the calculated G exceeds G_crit, the reading is a statistically confirmed outlier. Investigate the cause (a draft, a disturbance during weight placement, an accidental contact), document the finding and the disposition decision, and follow your SOP guidance on whether to include or exclude the value — never silently delete data. All ten raw readings, including any identified outliers and the disposition rationale, must appear in the verification log.

ContextMeasurement uncertainty is the parameter that tells you whether the balance is fit for a specific weighing task — not merely whether it passed verification. A simplified combined standard uncertainty: u_c = √(u_span² + u_lin² + u_rep²), where each component is in milligrams. The expanded uncertainty U = k × u_c, where k = 2 for approximately 95 % confidence. Example: if span contributes ±0.05 mg, linearity ±0.08 mg, and repeatability (as s) ±0.04 mg, then u_c = √(0.0025 + 0.0064 + 0.0016) ≈ 0.10 mg and U ≈ 0.20 mg. This balance should carry a ±0.20 mg uncertainty on every result it produces. If this expanded uncertainty is larger than the tolerance on the substance being weighed, the balance is not fit for purpose regardless of whether every individual verification parameter passed its criterion.

ContextMinimum weight is defined as the smallest sample mass for which the balance's relative uncertainty remains below a defined threshold — typically 0.1 % for pharmaceutical applications, meaning U / m_min ≤ 0.001. Using the example from the previous step (U = 0.20 mg): m_min = 0.20 mg / 0.001 = 200 mg. This means the balance should not be used to weigh quantities below 200 mg if 0.1 % accuracy is required. If your lab routinely weighs 50 mg samples on this instrument, the verification results indicate a fitness-for-purpose incompatibility even though each individual test passed. Record m_min in the log alongside the actual smallest mass weighed in practice. Flag any discrepancy for review by the responsible analyst or QA before the balance is used for that application again.

ContextData integrity rules under 21 CFR Part 11, EU Annex 11, and GLP/GMP principles require that original records are Attributable, Legible, Contemporaneous, Original, and Accurate — the ALCOA framework. Recording readings on a paper towel and transcribing them later breaks the Contemporaneous and Original requirements simultaneously. For handwritten entries, errors must be corrected by drawing a single line through the mistake, initialing and dating the correction beside it, and writing the correct value clearly — never obscure the original entry. For electronic logs, the system must maintain a complete, time-stamped audit trail linked to the individual user account. Every numerical entry must be made at the instant of measurement, not reconstructed from memory at end of shift.

ContextThe analyst's signature is a legal attestation that the verification was performed as described, on the date and time stated, by the named individual. Backdated signatures — even by a few hours — are a data integrity violation and have led to FDA warning letters and import alerts at pharmaceutical manufacturers. If testing spans a shift change, both analysts must independently sign and timestamp their respective portions at the time of handoff. Electronic signatures in validated systems carry the same legal weight as wet ink and must be individually linked to the user account — shared login credentials are never acceptable under 21 CFR Part 11 or EU Annex 11.

ContextPeer review catches arithmetic errors, unit transpositions (mg vs. g confusion is surprisingly common under time pressure), missing fields, and acceptance criterion misapplications before the record enters the permanent quality system. The reviewer should verify that each Pass/Fail determination is consistent with the raw data and the stated criterion, that the weight certificate lot numbers in the log match the physical weights actually used (cross-check the case label), and that the recorded environmental conditions are plausible for the site. The reviewer must not be the person who performed the verification. Same-day peer review is strongly preferred in GMP environments; next-day review by a supervisor is acceptable if documented but should not become routine practice.

ContextEvery analytical balance in a regulated laboratory must have a unique equipment ID and a dedicated history file containing all calibration records, verification logs, service reports, and deviation investigations. Filing records in a general or shared folder makes it nearly impossible to reconstruct the instrument's performance history during a regulatory inspection — auditors routinely request the last 12 months of verification data for critical instruments, and missing or misfiled records result in audit findings. If your lab uses a Computerized Maintenance Management System such as LabVantage, Blue Mountain RAM, or a validated spreadsheet tracker, update the next due date immediately and attach or reference the completed log by its unique document number.

ContextA passing result on a single re-measurement does not constitute remediation — one confirmed failure triggers an Out of Service label, a formal deviation record, and a retrospective impact assessment. The impact assessment must answer: which samples were weighed on this balance since the last passing verification, and could the failure have materially affected any results that were reported or used in a batch release or study conclusion? If the answer is possibly yes, those results must be flagged, quarantined, and potentially reprocessed. Attempting to clear a failure by running the internal calibration function and re-verifying without creating a deviation record is a data integrity violation under all major regulatory frameworks and has triggered enforcement actions against both individuals and organizations.

ContextSome failures yield to simple, documented corrective actions: a span error may resolve after running the internal motorized calibration cycle followed by re-verification with external weights; a repeatability failure may clear after cleaning the pan guide, removing a debris fragment, or allowing additional warm-up stabilization. If these first-line interventions do not restore compliance, professional intervention is required. Engage a calibration laboratory accredited to ISO/IEC 17025 for mass calibration — their certificate is the only document an auditor will accept as evidence of traceable recalibration. Keep the service provider's current accreditation scope document on file alongside every calibration certificate they issue for your instruments.

ContextThe non-conformance report must document: the nature and magnitude of the failure, the scope of potentially affected data, the root cause investigation findings, the corrective action taken to restore compliance, and the preventive action to reduce recurrence probability. Root cause typically falls into one of four categories: (1) equipment deterioration — load cell wear or electronic component aging; (2) environmental — a temperature excursion or a newly introduced vibration source; (3) process — weights handled improperly, contaminated, or used past their calibration interval; or (4) human error — procedure steps skipped or warm-up time neglected. The completed report must be reviewed and approved by Quality Assurance before the Out of Service label is removed and the balance is released for use.

ContextResidue from samples — particularly hygroscopic, corrosive, or reactive chemicals — accelerates load cell degradation if allowed to accumulate around the pan guide and draft shield interior. Isopropyl alcohol at 70 % (not neat or anhydrous) is effective against most laboratory surface residues without leaving electrically conductive deposits. Do not use acetone — it attacks polycarbonate draft shield panels and some protective instrument coatings. Never spray liquid directly into the balance interior or apply pressure to the pan guide stub or the load cell area. After cleaning, run the balance empty for five minutes to allow any residual solvent vapor to clear the weighing chamber before the next use. Record the cleaning action in the verification log.

ContextAnti-vibration feet are silicone or rubber isolators that harden and lose damping effectiveness over 2–4 years of continuous operation, especially in labs with concrete floors or significant HVAC vibration. A hardened foot looks externally identical to a healthy one but transmits vibration directly into the balance frame — you will detect this as increasing repeatability standard deviation that worsens when nearby centrifuges or shakers are running. Confirm all four feet make firm, even contact with the bench surface with no rocking. Draft shield hinges that stick or bind cause users to apply lateral force during door operation, which can nudge the pan and introduce positional error into a measurement series. Inspect the power cable at both ends for kinking or cracking — damaged insulation near the balance connector can inject 50/60 Hz electrical noise into the load cell signal. Replacement feet typically cost $30–80; addressing wear proactively prevents a $500–1,500 emergency service call.