Calibration and Measurement Error Exercises

These exercises practise calibration evidence as an engineering decision system. The aim is not only to calculate an error. The aim is to decide whether a measurement chain is biased, corrected, traceable, stable enough and accurate enough for the tolerance it supports.

Assume simplified calibration models unless an exercise states otherwise. Real calibration work also needs traceable references, environmental records, method control, fixture checks, software scaling review, uncertainty budgets, as-found/as-left records and a rule for what happens when an instrument was used while out of tolerance.

How to Use These Exercises

For each problem, state the measurand, reference value, indicated value, calibration model, tolerance and decision boundary. Keep sign convention explicit: error is usually indicated value minus reference value, while correction is usually reference value minus indicated value.

Release Evidence Notes

Calibration evidence should connect the measured error to the instrument, range, reference standard, environment, method, software scaling, fixture and decision being supported. A point correction is weak evidence unless the record also states whether the result is as-found or as-left, which tolerance applies and whether the instrument was used in a risk-bearing process before adjustment.

Useful release evidence includes reference traceability, calibration date, environmental limits, load or pressure points tested, zero and span behavior, linearity, hysteresis, resolution, repeatability, applied correction, uncertainty statement, acceptance rule and the action taken when the instrument is outside tolerance.

Engineering Boundary Notes

These exercises use simplified calibration models. They do not replace a complete uncertainty budget, method validation, measurement system analysis, software validation, reference-standard management or out-of-tolerance impact review. A calibration pass at one point does not prove the full range, dynamic response or installed measurement chain.

Common Release Mistakes

• using a correction without preserving the original as-found condition;
• reporting percent error without stating span, range or tolerance basis;
• treating a good zero check as proof of gain, linearity and hysteresis;
• ignoring meter loading, fixture force, temperature and software scaling;
• accepting a value near a limit without uncertainty or guard-band logic;
• returning an adjusted instrument to service without reviewing affected past measurements.

Scenario Map

Scenario	Exercises	Primary check	Engineering decision
Basic calibration error	1, 2, 3, 4	offset, gain, span and correction	Decide how the indicated value should be interpreted.
Model quality	5, 6, 7, 8	linearity, hysteresis, sensitivity and bridge output	Decide whether a simple calibration curve is adequate.
Tolerance evidence	9, 10, 11, 12, 13	percent span, resolution, deadband and TUR	Decide whether the instrument can support the tolerance.
Release decisions	14, 15, 16, 17, 18	combined correction, as-found/as-left state, range transfer and release gate	Accept, adjust, restrict or remove the measurement system.

Validation Package Checklist

• measurand, range, units and sign convention are explicit;
• reference standard, traceability path and uncertainty are identified;
• as-found and as-left states are separated;
• zero, span, linearity, hysteresis and resolution evidence match the intended range;
• corrections, tolerances and guard bands are stated before release;
• out-of-tolerance impact and affected-use window are dispositioned;
• final release decision names allowed range, method and restrictions.

Exercise 1: Offset Error at One Calibration Point

A pressure indicator reads 102.4\ \text{kPa} when a traceable reference applies 100.0\ \text{kPa}. Find the offset error and correction.

Solution

Using error as indicated minus reference:

The correction to apply to the indicated value is:

Engineering Comment

The instrument reads high at this point. A positive error is not automatically conservative; the consequence depends on whether the reading controls protection, process control, acceptance or reporting.

Plausibility Check

The indicated value is higher than the reference, so a positive error and negative correction are consistent.

Exercise 2: Zero Error and Span Error

A 0 to 200\ \text{N} load cell reads 3\ \text{N} at zero and 207\ \text{N} at full-scale load. Estimate zero error and span error after removing zero.

Solution

Zero error:

Full-scale indicated value after zero removal:

Span error:

Engineering Comment

Zero adjustment alone would not fix the span error. The calibration needs both offset and gain review.

Plausibility Check

The raw full-scale reading is 7\ \text{N} high. Removing the 3\ \text{N} zero leaves 4\ \text{N} of span error, which balances.

Exercise 3: Gain from a Two-Point Calibration

A displacement probe follows:

It reads 0.12\ \text{mm} at x=0 and 10.32\ \text{mm} at x=10.00\ \text{mm}. Find a and b.

Solution

Offset:

Gain:

The model is:

Engineering Comment

The probe has both offset and gain error. Using only a single zero correction would leave a scale error across the range.

Plausibility Check

The span reading changes by 10.20\ \text{mm} for a 10.00\ \text{mm} reference change, so the gain should be slightly above one.

Exercise 4: Correcting an Indicated Value

Using the calibration model:

an instrument indicates y=6.24\ \text{mm}. Estimate the corrected reference value x.

Solution

Rearrange:

Substitute:

Engineering Comment

Correction should use the inverse of the calibration model. Subtracting only the offset would give 6.12\ \text{mm} and leave gain error in the result.

Plausibility Check

Because the instrument gain is high, the corrected value should be slightly lower than the offset-corrected indication. 6.00\ \text{mm} fits.

Exercise 5: Linearity Error at Mid-Scale

A linear calibration line predicts 50.0\ \text{N} at mid-scale. The actual indication at the same reference load is 51.4\ \text{N}. Full scale is 100\ \text{N}. Find linearity error as percent full scale.

Solution

Linearity residual:

Percent full scale:

Engineering Comment

Linearity error is a model error. Even if endpoints are adjusted, the middle of the range may still be wrong.

Plausibility Check

A 1.4\ \text{N} residual on a 100\ \text{N} span is 1.4\% full scale.

Exercise 6: Hysteresis Error

At a reference torque of 40\ \text{N m}, a sensor reads 40.6\ \text{N m} on the increasing run and 39.8\ \text{N m} on the decreasing run. Find hysteresis width.

Solution

Hysteresis width:

If full scale is 100\ \text{N m}:

Engineering Comment

Hysteresis means calibration depends on loading history. A single calibration curve may not support bidirectional or cyclic decisions.

Plausibility Check

The two readings straddle the reference by less than one unit, so a 0.8\ \text{N m} width is plausible.

Exercise 7: Sensitivity Error in a Load Cell

A load cell nominal sensitivity is 2.000\ \text{mV/V} at full scale. Calibration finds 1.960\ \text{mV/V}. Find sensitivity error.

Solution

Absolute error:

Relative error:

Engineering Comment

If software uses nominal sensitivity, it will overestimate load because the actual output per unit load is lower than expected.

Plausibility Check

1.960 is 0.040 below 2.000, which is one fiftieth of nominal, or 2\%.

Exercise 8: Bridge Output from Excitation

A strain-gauge load cell has sensitivity 2.0\ \text{mV/V} and excitation 5.0\ \text{V}. Estimate full-scale output.

Solution

Output:

Engineering Comment

Small bridge outputs make offset, amplifier drift, common-mode limits and noise important. Calibration error may be dominated by electronics rather than the mechanical element.

Plausibility Check

A few millivolts per volt at a few volts excitation should produce a millivolt-level signal, so 10\ \text{mV} is credible.

Exercise 9: Percent Span Error of a Transmitter

A temperature transmitter range is -20^\circ\text{C} to 180^\circ\text{C}. At 100^\circ\text{C} it reads 101.8^\circ\text{C}. Find error as percent span.

Solution

Span:

Error:

Percent span:

Engineering Comment

Percent span is useful for transmitter specifications, but acceptance should still consider the process tolerance at the measured point.

Plausibility Check

An error under 2^\circ\text{C} on a 200^\circ\text{C} span should be under 1\% span.

Exercise 10: Tolerance Use by Calibration Error

A dimension tolerance is \pm 0.20\ \text{mm}. The calibrated measurement bias is 0.06\ \text{mm}. What fraction of the tolerance half-width is consumed by bias?

Solution

Fraction:

So bias consumes:

Engineering Comment

Even a repeatable measurement system can be unsuitable if bias consumes too much of the tolerance.

Plausibility Check

0.06 is a little less than one third of 0.20, so 30\% is plausible.

Exercise 11: Deadband in a Switch Measurement

A pressure switch turns on at 505\ \text{kPa} and turns off at 492\ \text{kPa}. Find deadband.

Solution

Deadband:

Engineering Comment

Deadband can be useful for stable control, but it must be included in acceptance tests. The switch has no single threshold unless direction is stated.

Plausibility Check

The on point is above the off point, so a positive 13\ \text{kPa} deadband is consistent.

Exercise 12: Resolution Relative to Tolerance

A digital indicator resolution is 0.01\ \text{mm}. The acceptance tolerance is \pm 0.10\ \text{mm}. How many display counts fit in the tolerance half-width?

Solution

Counts:

Engineering Comment

Ten counts across the half-width is usable for many screening decisions, but resolution alone does not prove low uncertainty.

Plausibility Check

Each count is one tenth of the half-width, so ten counts fit.

Exercise 13: Test Uncertainty Ratio from Error Limit

An instrument tolerance is \pm 1.0\ \text{unit}. The calibration standard expanded uncertainty is \pm 0.20\ \text{unit}. Estimate TUR.

Solution

Using tolerance half-width divided by expanded uncertainty:

Engineering Comment

A TUR of 5:1 is generally strong, but it does not include all measurement contributors unless the calibration procedure says so.

Plausibility Check

The standard uncertainty band is five times smaller than the tolerance half-width.

Exercise 14: Combined Offset and Gain Correction

A force channel uses:

where y is indicated force in newtons and x is reference force. The channel indicates 100\ \text{N}. Estimate corrected force.

Solution

Invert the model:

Engineering Comment

Offset and gain can cancel at one point without making the instrument correct elsewhere. The calibration curve is still needed.

Plausibility Check

The 2\ \text{N} high offset and 2\% low gain cancel at this reading, so 100\ \text{N} is reasonable.

Exercise 15: As-Found and As-Left Status

A gauge tolerance is \pm 0.50\ \text{bar}. As found, the largest error is 0.62\ \text{bar}. After adjustment, the largest error is 0.18\ \text{bar}. Classify the result.

Solution

As-found:

As-left:

The gauge was out of tolerance as found and in tolerance as left.

Engineering Comment

The important follow-up is impact review: measurements made since the previous calibration may be suspect.

Plausibility Check

One value exceeds the limit and the other is below it, so the two statuses differ.

Exercise 16: Calibration Certificate Correction

A certificate reports correction C=-0.08\ \text{mm} at a nominal 20\ \text{mm} point. The instrument reads 20.14\ \text{mm}. Estimate corrected value.

Solution

Apply correction:

Engineering Comment

Certificate corrections must be applied with the sign used by the certificate. Confusing error with correction is a common release error.

Plausibility Check

The correction is negative, so the corrected value should be lower than the indication.

Exercise 17: Range Transfer Risk

A calibration is performed at 0, 50 and 100\ \text{N}. The instrument will be used at 160\ \text{N} after a range change. Is this interpolation or extrapolation?

Solution

The use point is above the highest calibrated point:

So it is extrapolation.

Engineering Comment

Extrapolation should not be treated as calibrated evidence unless the range, linearity, overload history and uncertainty have been justified.

Plausibility Check

The use point lies outside the calibrated interval, so the classification is clear.

Exercise 18: Calibration Release Gate

A measurement system has these findings:

Check	Result	Gate
Offset error	0.04\ \text{mm}	\le 0.05\ \text{mm}
Linearity error	0.08\ \text{mm}	\le 0.05\ \text{mm}
Resolution	0.01\ \text{mm}	\le 0.02\ \text{mm}
TUR	3.2	\ge 4.0
As-left status	in tolerance	in tolerance

Decide whether to release the system.

Solution

Offset passes:

Linearity fails:

Resolution passes:

TUR fails:

The system should not be released without restriction. It needs model revision, tighter method uncertainty, a restricted range or a different instrument.

Engineering Comment

As-left in tolerance is not enough when model quality and calibration capability are weak. The release decision must match the measurement purpose.

Plausibility Check

Two independent gates fail, so a hold or restricted release is the consistent decision.