MatroVal A Calibration Software With Advanced Uncertainty and Prediction Capabilities Including Complex Numbers Alex Lepek Newton Metrology Ltd

Presentation Outline Calibration Procedure Calibration Procedure linked to uncertainty analysis and prediction Types of uncertainty analysis Prediction of aging and instability as an uncertainty component

The calibration procedure is a table that is read line by line and cell by cell. Each cell Contains a parameter or instruction. The table can be prepared and linked automatically. The Calibration Procedure Overview

Goes to Certificate Instrument adjustment? Uncertainty file and page History file used for prediction Uncertainty method links Measurement repetitions Instructions Compliance Calibration Procedure

Read Procedure Table How does it work User setup Instructions measure Send measurements and other data (e.g. past calibrations) to uncertainty calculator Past calibrations of the system used for its prediction Certificate / Report Predefined uncertainty budget template or equation and method Measurand & Measurement System link

Linking cal procedure, uncertainty and prediction a.his Unc. components

Description or instructions to user Components reference values Correlated components measured simultaneously and processed (e.g. sigma of difference) Measurements Entry fields Manual Calibration

Using alternative template } Unc. Components Result Compliance

Calibration results report generated by a template Can indicate k factor for each line Decision per ILAC G8

Available uncertainty methods GUM (first order Taylor expansion). Expanded uncertainty computation includes correlations. C-sensitivities may be complex numbers. Simplified Monte Carlo (assuming RSS equation). C-sensitivities may be complex numbers. Monte-Carlo simulation directly from the measurement equation. Equation may include complex numbers. Distributions and reference values are given in the budget table. Best precision of results.

Deviation Uncertainty budget Methods: GUM and Monte- Carlo can be complex c-sensitivity can be complex

Uncertainty from measurement equation j Imaginary unit

Adjustment info k, df, confidence Past Calibrations used for prediction and interval analysis

Prediction Method Past calibrations differences consider adj. Graphic presentation of a prediction Prediction result

Available Prediction Methods Weighted Linear Regression, takes into consideration both regression statistics and measurement uncertainties. Best for white noise instabilities. Weighted Linear Regression of Differences. Advantage when instrument passes occasional adjustments (“as received” and “as left”). Best for random walk instabilities. The program can chose the best method and use it.

Hands-on experience: download ‘MetroVal from download ‘MetroVal from