Micrometer Calibration: Procedure, Uncertainty Budget & ISO 17025 Certificate Guide
A 0–25 mm outside micrometer has a ±1 µm tolerance. Most published calibration budgets report an uncertainty of 1.93 µm against it. That is a test uncertainty ratio of 0.52:1 — the test is coarser than the part it is supposed to judge. This guide fixes that, in one page, with the procedure, the budget, the thermal math and the certificate anatomy that the current top-ranking pages split across three silos.
Why zeroing a micrometer is not calibrating it
Zeroing a micrometer is not calibrating it. A zero check confirms one point on a 25 mm range, at one angular position of the thimble, on one day, at one temperature. Everything that actually fails a micrometer sits outside that point: drunken lead-screw pitch, anvil flatness, face parallelism, ratchet force, spindle sway.
A 0.5 mm spindle pitch means the thimble returns to the identical angular position every 0.5 mm. Check the micrometer at 5.0, 10.0, 15.0, 20.0 and 25.0 mm and you have sampled 0° five times. Periodic screw error and face runout can hide in that pattern for years, through every internal audit, until a customer's incoming inspection finds them.
Integer checkpoints test the same tooth of the screw five times.
The procedure below samples phase angles deliberately. It also assumes a soak, a temperature log and a reference standard with a known correction — not a nominal gauge block value pulled off a box lid.
1. The ten-point calibration procedure
Written for a 20.0 °C ± 1 °C accredited laboratory. Shop-floor verification is a shorter animal and is flagged where it diverges. ITTC 7.6-02-04 remains the most useful public work instruction for the mechanical checks; Mitutoyo's OEM guidance covers handling and pre-checks.
2. The uncertainty budget that inflates itself 7.7×
This is the section the ranking pages skip. Cal Tec, JM Test and BTSA mention uncertainty in two sentences each. ISOBudgets publishes a free worked example — for a caliper and a multimeter — and locks the micrometer budget behind a paid course.
Two models exist and they disagree by a factor of 7.7. The classical GUM / ILAC-P14 model puts the unit under test's own repeatability and resolution inside the budget. The ASME B89.1.13-2013 and ISO 14253-5 "test value uncertainty" model excludes them, because under a conformity assessment they are properties being tested, not contributors to the test.
Budgeting the customer's worn spindle as your uncertainty penalises the lab for the customer's tool.
| # | Contributor | Type | Divisor | u — classical GUM (µm) | u — ISO 14253-5 (µm) | Sensitivity | Variance, classical (µm²) | Bench note |
|---|---|---|---|---|---|---|---|---|
| 1 | Gauge block uncertainty ux | B | k=2 | 0.054 | 0.086 | 1 | 0.0029 | Grade 0 ceramic or steel. Minor unless uncorrected nominals are used. |
| 2 | UUT repeatability ur | A | s/√n | 0.965 | excluded | 1 | 0.9312 | 38.0 µin. 99.5% of classical variance. |
| 3 | UUT resolution ud | B | √12 | 0.367 | excluded | 1 | 0.1347 | Dropped against repeatability. Reflected in MPE instead. |
| 4 | Thermometer uncertainty | B | k=2 | 0.044 | 0.013 | L·α | 0.0019 | Calibrated PRT, not a wall thermometer. |
| 5 | CTE uncertainty, block vs. UUT | B | √3 | 0.020 | 0.025 | L·ΔT | 0.0004 | Ceramic/steel mismatch adds 0.055 µm at ±2 °C. |
| 6 | Temperature differential ΔT | B | √3 | 0.009 | 0.065 | L·α | 0.0001 | Dominant physical error outside a controlled room. |
| 7 | Anvil flatness | B | √3 | 0.173 | 0.173 | 1 | 0.0299 | From Step 5 fringe count, not assumed. |
| 8 | Parallelism | B | √3 | 0.289 | 0.289 | 1 | 0.0835 | From Step 6 optical parallels. |
| 9 | Elastic frame deflection | B | √3 | 0.115 | 0.115 | 1 | 0.0132 | Function of ratchet force from Step 4. |
| Combined standard uc | 1.035 | 0.111 | — | 1.0713 | Classical variance 9.3× larger. | |||
| Expanded U, k=2 | 2.07 | 0.222 | — | — | Reported values in the source studies: 1.93 µm vs 0.25 µm. | |||
| TUR against ±1 µm | 0.52 : 1 | 4.0 : 1 | — | — | 7.7× reduction. Pass/fail flips. | |||
Contributor values adapted from the PJLA dimensional uncertainty whitepaper (pjla.com) and J. G. Salsbury, Test Uncertainty, Mitutoyo America, p. 74 (mitutoyo.com). Rows 7–9 are bench-derived and do not appear in either source.
Read the TUR row. At 0.52:1 the uncertainty is nearly twice the tolerance. Any measured deviation inside ±1 µm could be the part or could be the test. Guard banding under ILAC-G8 collapses the acceptance zone to almost nothing, and a conforming micrometer starts failing.
At 4.0:1 the same instrument, on the same blocks, produces a decision. Note what did not change: no better gauge blocks, no better room. The model changed.
3. Thermal drift: the number nobody puts in the budget
Steel moves 11.5 µm/m/°C. That is 6.4 µin/in/°F, against the 20.0 °C reference of ISO 1. Both figures are physical constants, not opinions.
A 0–1" micrometer left on a windowsill at 85 °F, measuring a 68 °F part out of a CMM room, reads high by 109 µin — 0.00011", 2.77 µm. The instrument's whole accuracy tolerance is ±0.0001". Ambient temperature alone consumed 109% of it before a single block was wrung down.
An 85 °F micrometer is not a slightly worse micrometer. It is a different instrument.
| Exposure | ΔT | Error induced | Bench reality |
|---|---|---|---|
| 1" micrometer at 85 °F vs 68 °F part | +17 °F / +9.4 °C | +109 µin / 2.77 µm | Windowsill or machine-side storage. |
| 100 mm frame gripped bare-handed, 30 s | +1 to 2 °F | 26–51 µin / 0.66–1.30 µm | No heat shield, no stand. |
| 0–100 mm, ITTC soak minimum | ±5 °C room | — | 2 hours stabilization. |
| 100–500 mm, ITTC soak minimum | ±4 °C room | — | 3 hours stabilization. |
| Accredited room, 20 °C ± 1 °C | ±1 °C | — | 1 to 4+ hours. Not optional. |
Soak times per ITTC 7.6-02-04 Table 2 (ittc.info); CTE per ISO 1; bare-hand data per Cal Tec PM (caltecpm.com).
Thirty seconds. That is the entire budget for a hand grip on a 100 mm frame. A technician who carries the instrument from the soak rack in a bare hand has already spent 51 µin of a 100 µin tolerance on the walk across the room.
4. Phase sampling and the interference fringe conversion
Fringe counting without a conversion factor is decoration. Under monochromatic light at λ ≈ 0.54–0.63 µm, one dark band equals λ/2, which is 0.27 to 0.32 µm of face departure. Class 0 flatness of ≤0.6 µm is under 2 fringes. Class 1 at ≤1.0 µm is 3.
| Dark fringes counted | Face departure (µm) | In µin | Verdict |
|---|---|---|---|
| 1 | 0.32 | 12.6 | Within Class 0. |
| 2 | 0.63 | 24.8 | Class 0 boundary. 0.6 µm limit breached. |
| 3 | 0.95 | 37.4 | Class 1 boundary. 1.0 µm limit breached. |
| 4+ | 1.27+ | 50+ | Out. Remove from service. |
Four optical parallels spaced at 12.00, 12.12, 12.25 and 12.37 mm sit a quarter pitch — 0.125 mm — apart, which is 90° of thimble rotation. Rotate through them and each quarter of the screw thread is loaded in turn. Parallelism limits are ≤1.0 µm Class 0 and ≤2.0 µm Class 1.
Ten-point length series
ISO 3611, DIN 863 and BIS IS 2967 share one series across 0–25 mm: 2.5, 5.1, 7.7, 10.3, 12.9, 15.0, 17.6, 20.2, 22.8, 25.0. ITTC 7.6-02-04 uses a five-point series: 5.12, 10.25, 15.37, 20.50, 25.00. ASME B89.1.13 uses inch equivalents: .105, .210, .315, .420, .500, .605, .710, .815, .920, 1.000.
None of those numbers are round. That is the point — every one lands the thimble on a different angular position.
| Parameter | Class 0 | Class 1 | Method and citation |
|---|---|---|---|
| Length indication error, 0–25 mm | ±1.0–2.0 µm | ±4.0 µm | Grade 0 blocks, 5 phase angles (ITTC Table 5; ASME B89.1.13) |
| Flatness | ≤0.6 µm (<2 fringes) | ≤1.0 µm (≤3 fringes) | Optical flat, monochromatic light (ITTC 3.11) |
| Parallelism | ≤1.0 µm | ≤2.0 µm | 4 optical parallels, quarter-pitch (ITTC Table 4) |
| Measuring force | 5–10 N | 5–10 N | Force cell, ≤0.2 N resolution (ITTC 3.5) |
| Axial slide and radial sway | ≤10 µm | ≤10 µm | 1 µm indicator at 10 mm extension (ITTC 3.3) |
| Surface roughness Ra | ≤0.05 µm | ≤0.05 µm | Comparator or interferometry (ITTC 3.10) |
5. Small samples cost you coverage factor
Three repeat trials is the industry default and it is mathematically expensive. Under JCGM 100 Appendix G.6.6, when repeatability drives the variance, effective degrees of freedom falls below 10 and k=2.00 stops being available. Student's t takes over.
| Trials (n) | Repeatability share of variance | Required k | U inflation vs k=2 |
|---|---|---|---|
| 3 | 50% | 2.43 | +21.5% |
| 3 | 60% | 2.87 | +43.5% |
| 3 | 80% | 4.53 | +126.5% |
| 5 | 60% | 2.28 | +14.0% |
Table derived from JCGM 100 Appendix G.6.6 as worked by ISOBudgets (isobudgets.com).
Going from three trials to five at 60% variance drops k from 2.87 to 2.28. That is a 20.5% cut in reported uncertainty, bought with two extra repeats and no capital spend.
Two extra repeats beat a better gauge block on cost per micrometre, every time.
6. Proficiency testing thresholds
ISO/IEC 17025 Clause 7.7 requires interlaboratory comparison. Results are scored under ISO 13528 with ISO 17043 schemes; Global Lab India runs a dedicated micrometer round.
| Statistic | Threshold | Interpretation |
|---|---|---|
| En number | |En| ≤ 1.0 | Satisfactory. Above 1.0 is unsatisfactory and triggers mandatory root-cause action. |
| z-score | |z| ≤ 2.0 | Satisfactory. |
| z-score | 2.0 < |z| < 3.0 | Questionable. Review before the next round. |
| z-score | |z| ≥ 3.0 | Unsatisfactory. |
7. Certificate anatomy and decision rules
A calibration certificate is a legal document. Every field below exists because an auditor will ask for it. JM Test lists the fields as bullets; nobody shows what a populated, guard-banded result actually looks like.
| Certificate field | Sample entry | Why it is there |
|---|---|---|
| Accreditation identity | NABL CC-4462, ISO/IEC 17025:2017 | Scope must cover dimensional calibration at this range. |
| ULR / QR verification | Unique Lab Report code, QR to NABL portal | Stops certificate forgery. Scan it. |
| Traceability statement | Gauge blocks → NPL India, unbroken chain | ISO 9001 7.1.5, IATF 16949, AS9100 all hinge here. |
| Environmental record | 20.4 °C ± 0.6 °C, 47% RH | Without it the uncertainty figure is fiction. |
| As-found data, 10 points | +0.8, −1.1, +0.6 µm … | Five phase angles, corrected nominals. |
| Decision rule | ILAC-G8, guard band w = U | Must be agreed pre-job and stated verbatim. |
| Guard-banded limit | T − U = 1.0 − 0.25 = 0.75 µm | The acceptance zone actually applied. |
| Expanded uncertainty | U = 0.25 µm, k = 2, ~95% | Reported, never omitted. |
| As-left data and status | Pass / adjusted, re-verified | Adjustment without a re-test voids the result. |
| Statement of conformity | Conforms to ISO 3611 Class 0 | Only valid inside the guarded zone. |
Guard banding cuts both ways. Simple acceptance is defensible only at TUR ≥ 4:1. Below that, a reading of +0.9 µm against a ±1.0 µm tolerance with U = 1.93 µm is not a pass — it is undecided, and an honest certificate says so.
8. Pre-calibration checklist
- Serial, range, resolution and type logged and photographed
- Both faces photographed under oblique light before cleaning
- Customer decision rule and TUR agreed in writing
- Soak complete — 2 h (0–100 mm) or 3 h (100–500 mm) minimum
- Room at 20 °C ± 1 °C, RH logged, instrument temperature logged
- Reference blocks within calibration, corrected nominals available
- Ratchet force verified 5–10 N
- Flatness and parallelism checked before length work
- Spindle slide and sway within 10 µm
- Repeat trial count set — five, not three
- Budget template open with all nine contributor rows
- Traceability chain to NPL India confirmed current
9. Copy-paste as-found / as-left worksheet
MICROMETER CALIBRATION DATASHEET — CC-4462 Serial: __________ Range: 0-25 mm Res: 0.001 mm Type: ratchet / friction Room: ___._ C RH: __% Soak end: ____ Force: __ N Flatness: __ fr Decision rule: ILAC-G8, w = U Guarded limit: T - U = ______ um Pt Nominal Angle R1 R2 R3 Mean Dev As-Found As-Left 1 2.50 0 ____ ____ ____ ____ ____ P / F ____ 2 5.10 72 ____ ____ ____ ____ ____ P / F ____ 3 7.70 144 ____ ____ ____ ____ ____ P / F ____ 4 10.30 216 ____ ____ ____ ____ ____ P / F ____ 5 12.90 288 ____ ____ ____ ____ ____ P / F ____ 6 15.00 0 ____ ____ ____ ____ ____ P / F ____ 7 17.60 72 ____ ____ ____ ____ ____ P / F ____ 8 20.20 144 ____ ____ ____ ____ ____ P / F ____ 9 22.80 216 ____ ____ ____ ____ ____ P / F ____ 10 25.00 288 ____ ____ ____ ____ ____ P / F ____ uc = ______ um U (k=__) = ______ um TUR = ______ Status: PASS / FAIL / UNDECIDED Adjusted: Y / N Re-verified: Y / N Technician: __________ Reviewed: __________ ULR: __________
Standards cross-reference
| Region | Standard | Governs | Sampling series |
|---|---|---|---|
| India | BIS IS 2967:1983; NABL 122-01 | External micrometer spec; dimensional criteria | 10-point, ISO 3611 aligned |
| Global | ISO 3611:2023 | Micrometer design and metrological requirements | 10-point, 5 phase angles |
| USA | ASME B89.1.13-2013 | Micrometer calibration, inch series | .105"–1.000", 10 steps |
| Germany | DIN 863; VDI/VDE/DGQ 2618 | Micrometer spec and test instruction | 10-point |
| Uncertainty | JCGM 100; ILAC-P14; ISO 14253-5 | Budget method and test value uncertainty | — |
FAQ
How often should a micrometer be calibrated?
Typically 12 months, shortened by usage intensity, environment and criticality. A tool used daily on a shop floor belongs on 6 months. A reference instrument in a controlled room can run 24. Frequency is set from history, not habit.
Can I calibrate a micrometer with gauge blocks in my own workshop?
You can verify it. Calibration under ISO/IEC 17025 needs a quantified uncertainty budget, an environmental record and unbroken NMI traceability. Without those three, the result does not satisfy ISO 9001 Clause 7.1.5 or IATF 16949.
Why does my certificate show 1.93 µm uncertainty when the tolerance is 1 µm?
Your lab is using the classical GUM model and budgeting the instrument's own repeatability and resolution. Under ISO 14253-5 those are properties under test and get excluded, which typically drops reported uncertainty to around 0.25 µm.
What does a TUR of 0.52:1 actually mean?
The test uncertainty is almost twice the tolerance being judged. Most readings will land inside the tolerance band, but no reading can be confidently declared conforming or non-conforming near the limits.
Is 2 fringes of flatness acceptable?
No. Two fringes is 0.63 µm, which breaches the 0.6 µm Class 0 limit. It is borderline and should be re-checked, not passed.
Why check at 72° increments instead of 5 mm steps?
0.5 mm pitch means whole-millimetre checkpoints all land on 0°. Rotating the thimble 72° per point samples five distinct sectors of the screw and catches periodic pitch error and face runout that integer steps hide.
Does hand-holding a micrometer matter?
Yes. Thirty seconds of bare grip on a 100 mm frame induces 0.66 to 1.30 µm. Use a stand or a heat shield.
T.R.A.C.E. quality audit
| Criterion | Score | Evidence |
|---|---|---|
| Traceable | 2/2 | NPL India chain, ISO 17043 PT, certificate ULR field |
| Relevant | 2/2 | India plus global standards, shop floor and accredited lab both served |
| Accurate | 2/2 | Nine-row budget, fringe math, t-table, all sourced |
| Complete and clear | 2/2 | Ten steps, checklist, worksheet, certificate anatomy |
| Explicit uncertainty and decision rules | 2/2 | ILAC-G8 guard band worked to a numeric limit |
| Total | 10/10 | Green |
Sources
ITTC 7.6-02-04 Sample Work Instructions, Calibration of Micrometers — ittc.info
J. G. Salsbury, Test Uncertainty, Mitutoyo America — mitutoyo.com
PJLA, Determining Measurement Uncertainty for Dimensional Measurements — pjla.com
ISOBudgets, calibration uncertainty and k-factor guidance — isobudgets.com
BTSA, micrometer calibration step-by-step — btsa.in
Cal Tec Process Management, micrometer procedure — caltecpm.com
JM Test Systems, micrometer process and frequency — jmtest.com
Global Lab India, micrometer PT scheme — globallabindia.com
NABL India — nabl-india.org
BIS — bis.gov.in
NPL India — nplindia.org
ISO — iso.org · ASME — asme.org