Covariance Matrix Evaluation in Correlated Type B Measurement Uncertainty Propagation
Evaluating input covariance matrices in Type B propagation prevents underestimating multi-channel variance and ensures valid system uncertainty limits.
Partial Derivatives
Differential Sensitivity
Mathematical calculus tools measure the local rate of change of a multivariable function with respect to one isolated variable. In uncertainty propagation algorithms, partial derivatives act as sensitivity coefficients that weight the contribution of each input variable to output variance. Metrologists calculate these coefficients analytically or evaluate them numerically by perturbing input values around nominal calibration operating points.
Applicability breaks down at points where the underlying physical transfer function exhibits step changes, non-differentiable elbows or numerical discontinuities.
Local Linearization
Linear approximations replace complex non-linear curves within a narrow neighborhood around the operating point. Evaluating partial derivatives at mean input values establishes local slope factors for variance propagation equations. Small parameter variations remain valid under this linear model, maintaining accurate uncertainty estimates across standard calibration ranges.
Cross Derivative
Higher order derivatives capture curvature effects when input variations expand beyond micro-scale ranges. Mixed partial derivatives quantify interaction terms where one input variable alters the sensitivity slope of another variable. Including second-order derivative terms prevents underestimating measurement uncertainty in highly non-linear sensor channels.
Discontinuity Bound
Discontinuous transfer functions prevent direct evaluation of differential slope values. Digital quantization steps and threshold switching mechanisms introduce zero or infinite derivative values at transition boundaries. Numerical finite difference approximations replace analytical derivatives when processing step-response sensor data.