Sensor Calibration Process

Compare raw sensor readings with trusted references, fit a correction model, and verify the error that remains.

Animation Beginner Sensor accuracy Reference points

Sensor Calibration Process

Watch a sensor compare itself with trusted reference points, fit a correction equation, and prove whether the remaining error is small enough for deployment.

Two-pointcalibration method

0.00 Cmaximum calibrated error

+/-0.50 Callowed tolerance

Checkingdeployment decision

Reference First

Calibration starts with known values from a trusted standard, not with the sensor under test.

Correct The Pattern

Offset shifts the line; gain changes the slope; non-linearity bends the response.

Verify Separately

A useful calibration checks points that were not used to create the equation.

Record Limits

A calibration record should state method, references, residual error, tolerance, and next check date.

1

Known Reference

Stable physical value is prepared.

2

Raw Sensor

The device reports its uncorrected value.

3

Fit Equation

Reference and raw values produce a correction model.

4

Apply Correction

New readings are translated through the model.

5

Verify Result

Remaining error is compared with tolerance.

Offset0.00 CEstimated correction at zero or midpoint.

Gain1.000Linear scale factor from raw value to true value.

RMSE0.00 CAverage size of remaining calibrated error.

Points2 reference pointsData used to fit the selected method.

Controls

Cold chain

Sensor scenarioraw behavior

Calibration methodoffset + gain

Viewfit plot

PlaybackReference

Field conditionsstress test

Drift offset +0.40 C Aged sensors often shift after deployment.

Repeatability noise 0.05 C Noise affects the trust you can place in each point.

Allowed tolerance +/-0.50 C Pass or fail is judged against the application, not against perfection.

Verification value 5.0 C Move the blue point to test a value not necessarily used for fitting.

Fit raw and calibrated response

displayed value vs true value

trusted reference raw sensor after calibration calibration points

Reference valueThe trusted standard defines what correct means.

Raw readingThe uncalibrated device can be shifted, scaled, noisy, or curved.

Corrected outputThe fitted equation converts future raw values into estimated true values.

Diagnosis

within tolerance

Ready to use The calibrated error is below the selected tolerance.

True verification5.0 C

Raw sensor says5.0 C

Corrected reading5.0 C

Remaining error0.0 C

Equationy = x

Next actionRecord result

Two-point calibration is usually the minimum useful field method for a linear sensor because it corrects both offset and gain.

Single-Point Rule

Single-point calibration only shifts readings by one offset.

calibrated = raw + offset

Two-Point Rule

Two-point calibration fits a straight line through low and high references.

gain = (ref2 - ref1) / (raw2 - raw1)

Multi-Point Rule

This animation uses quadratic least-squares for a curved sensor response.

true = a raw^2 + b raw + c

Calibration Quick Reference

Calibration is a comparison process. The reference instrument or reference condition is treated as the trusted value; the sensor under test is adjusted mathematically.

• Offset error means the whole response is shifted.
• Gain error means the slope is too steep or too shallow.
• Non-linearity means one straight line cannot fit the whole range well.
• Noise and repeatability limit how much precision the correction can prove.

Method Selection

Choose the method to match the error pattern and the risk of the application.

• Use single-point only for quick field checks near one operating value.
• Use two-point for most linear sensors across a known range.
• Use multi-point when the residual plot still bends after a straight-line fit.
• Verify with points that were not used to fit the model.

Technical Accuracy Notes

The plotted sensor model is a teaching model, not a datasheet model. Raw readings are generated with offset, gain, curvature, drift, and deterministic repeatability noise so learners can see why different calibration methods behave differently.

• Single-point: corrected = raw + offset, using one reference point.
• Two-point: corrected = gain * raw + offset, fitted through two reference points.
• Multi-point: corrected = a * raw^2 + b * raw + c, fitted by least-squares to five reference points.
• RMSE and maximum error are calculated over the displayed operating range, in the scenario units.
• A real traceable calibration also needs uncertainty budget, standards traceability, environmental conditions, and controlled procedures.

Calibration Record Checklist

A useful calibration record should let another person understand what was calibrated, how it was checked, and when it must be checked again.

• Sensor ID, date, operator, and firmware or hardware revision.
• Reference standard ID, certificate date, and stated uncertainty.
• Reference points, raw readings, correction equation, tolerance, and pass/fail result.
• Environmental conditions and next calibration interval.

Compare Methods

Select the pressure scenario, increase drift, and compare single-point against two-point. Notice where the error grows.

Find Curvature

Select the CO2 room scenario and switch to the residual view. A curved residual pattern explains why multi-point calibration helps.

Set Tolerance

Tighten the tolerance until a method fails. The same sensor can be acceptable for one job and unacceptable for another.

ADC Resolution Visualizer

See how quantization and noise limit the raw data before calibration.

Signal Conditioning Pipeline

Follow filtering and scaling before the sensor value reaches software.

IoT Formula Calculator

Practice unit conversions and sensor equations used in measurement systems.