Gauge Block Stacks: How Much Error Each Step Actually Adds

You have a 37.425 mm target to set a bore gauge. You select four blocks from the set, clean them, wring them together, and clamp the stack in the comparator. The measurement reads 37.428 mm. Is that the stack, the comparator, or the part?

Gauge blocks are the most accurate length standards on a shop floor — but only if you account for every error source between the box and the measurement. This article walks through the quantified contributions: grade tolerance, wringing film variability, thermal expansion, and stack-building decisions. Each section gives you a number you can use in an uncertainty budget.

Grade K vs 0 vs 1 vs 2: The Tolerance Numbers That Matter

ISO 3650:1998 (adopted in India as IS 2984:2003, identical under dual numbering) defines four grades. The key distinction: Grade K is a calibration grade whose actual length is reported on the certificate — its nominal deviation can be up to ±0.2 µm on a 10 mm block, but its variation in length (flatness/parallelism) is held to 0.05 µm. Grade 0, 1, and 2 are working grades held to tighter nominal deviation but looser variation.

Source: ISO 3650:1998, Table 1 — tolerances for metric gauge blocks (Opus Metrology data sheet).

The trap. A Grade 2 block can be 0.45 µm off nominal and still pass. If you use it as a reference to set a comparator that then inspects a ±5 µm tolerance part, the block consumes nearly 10% of the tolerance before you even wring it. The rule: the block's tolerance should be ≤10% of the smallest tolerance you measure with it. For a ±5 µm feature, that means Grade 0 or better.

The exception. If you have a calibration certificate with actual values for each block, you can use a Grade 1 or even Grade 2 set as a working standard — you correct for each block's known deviation. But that only works if you keep the certificate current and apply the corrections. Most shops do not. They use nominal length and trust the grade.

What a Bad Wring Looks Like — and What It Costs

A proper wring produces a bond that takes 2–5 kgf to separate. The wringing film thickness is included in the calibrated length of each block — the standard defines block length as measured when wrung to a platen. So when you stack blocks, the wringing films are already accounted for in each block's value. No correction is needed.

But that assumption breaks if the wring is bad.

The NPL (National Physical Laboratory, UK) measured variability in wringing film thickness and found that with good technique the variability is below 0.01 µm. With poor technique — excess oil, contamination, or worn surfaces — it can exceed 0.1 µm per interface (NPL Report CLM 3). A four-block stack with bad wrings adds up to 0.4 µm of uncorrected error.

How to diagnose a bad wring visually:

1. No interference fringes when tested with an optical flat. Place the optical flat on the cleaned block face and slide it gently. If no fringes appear, there is a large burr or contaminant. If fringes appear but disappear when you lightly press the flat, the face is clean. If fringes remain locally and move with the flat, a burr exists on the block face (Mitutoyo Quick Guide to Gauge Blocks).

1. The stack separates under light finger pressure. A correctly wrung stack should feel like a single solid block. If it separates when you handle it, one or more faces are not flat enough to wring — typically because of a burr, a scratch, or wear exceeding about 5 microinches (0.125 µm) flatness deviation (Starrett Webber, Wringability and Gage Blocks).

1. Oil film visible at the interface. When you wring a block to an optical flat, no colour or oil should appear on the flat for Grade K, 0, or 1 blocks. For Grade 2, at least 80% of the wrung area must be colourless (Starrett Webber, GB48). Visible oil means excess lubricant is adding a film thickness that the calibration did not account for.

The fix. Lightly stone the suspect face with a clean gauge-block stone (serrated type, not an Arkansas stone). Use 1–1.5 lbf pressure, two or three strokes. Test again. If the block still will not wring after two attempts, it is damaged — recalibrate or replace.

Stack-Building Rules That Minimise Error

Mitutoyo's guide and the NPL good practice guide agree on three rules:

1. Use the minimum number of blocks. Each additional interface adds a wringing film variability term. An 81-piece set can build any dimension up to 100 mm with four blocks or fewer if you select wisely. If you need five or more, reconsider your block combination.

1. Select thick blocks first. Start with the block that covers the least significant digit, then work back to the most significant. For 37.425 mm: select 1.005 mm, then 1.42 mm, then 5 mm, then 30 mm. This minimises the number of thin blocks, which are more prone to bending during wringing.

1. Wring thin blocks to thick blocks first. To prevent thin blocks from deforming, wring a thin block onto a thick block, then wring the next thin block onto the first thin block. Finally, remove the thick block from the stack (Mitutoyo Quick Guide).

The bending risk. A 1.005 mm steel block wrung directly to another thin block can bow by 0.2–0.3 µm under wringing force. Always use the thick-block backing technique for blocks under 3 mm.

Thermal Soak: The Dominant Error Source

Steel expands at 11.5 × 10⁻⁶ /°C (ISO 3650 nominal coefficient). A 100 mm block 1 °C above 20 °C reads 1.15 µm long — larger than the entire tolerance of a Grade 0 block.

Acclimation times from Starrett Webber's experimental data:

Formula for steel on a soaking plate: Time (min) = 4L + 7, where L is in inches. Minimum 8 min, maximum 60 min. In open air: Time = 30L + 50, minimum 60 min, maximum 300 min.

Material multipliers (relative to steel):

• Chrome carbide: ×1.2
• Tungsten carbide: ×1.5
• Ceramic (zirconia): ×2.2

Source: Starrett Webber, Acclimation Time for Gage Blocks (GB51).

Do not soak on a granite surface plate. Granite is a thermal insulator. It will localise heat from the block and slow equilibration, not speed it up.

Handling adds heat. A 100 mm steel block held with five fingers elongates by about 0.5 µm within 30 seconds and takes 20–30 minutes to return to equilibrium (Mitutoyo data). Use gloves — they reduce warming by approximately half (Lipus et al., 2022, Advances in Production Engineering & Management). But even with gloves, the stack must stabilise for at least one hour in controlled conditions before use for micrometre-level measurements.

When a Set Needs Recalibration

ISO 3650 does not specify a recalibration interval. That is left to the user's quality system. Industry practice, based on NPL and Techmaster data:

• Grade K reference sets: 12–24 months, or immediately if a block is dropped or fails a wringability test.
• Grade 0 lab sets: 12–24 months for sets used weekly; 24–36 months for sets used occasionally.
• Grade 1 and 2 shop sets: 6–12 months, because they see more handling and are more likely to develop burrs and wear.

The trigger for early recalibration. Any block that:

• Fails the optical flat interference fringe test
• Will not wring to a clean platen after light stoning
• Shows visible rust, pitting, or scratches on the measuring face
• Has been dropped (internal stress relief can shift length by 0.1–0.3 µm)

The as-found record matters. If a set has been within grade for three consecutive calibrations, you can extend the interval by 50%. If it shows drift approaching the grade limit, shorten the interval. NABL-accredited labs in India typically default to 12 months for working sets and adjust based on as-found history.

The Uncertainty Budget for a Typical Shop-Floor Stack

For a 37.425 mm stack built from four Grade 1 blocks, using nominal lengths (no correction), wrung with good technique, soaked on a steel plate for 30 minutes, measured at 20.5 °C:

That means a stack reading 37.425 mm nominal could actually be anywhere from 37.4241 to 37.4259 mm at 95% confidence. If your part tolerance is ±5 µm, that is acceptable. If your part tolerance is ±2 µm, it is not — you need Grade 0 blocks, a calibration certificate with actual values, and tighter temperature control.

The next decision. If you are setting a comparator to inspect a ±2 µm feature, you need to either upgrade to a Grade 0 set with current calibration, or use a single block (no stack) of the exact dimension. Stacks add uncertainty that may consume your test uncertainty ratio (TUR) below the 4:1 minimum that most quality standards require.

For gauge block calibration services, NABL-traceable reports, or replacement of worn blocks, contact DSN Enterprises: /calibration and /contact.