Robust models for spectrophotometer optimisation

Bruna Falgueras Vallbona

doi:10.1016/j.measurement.2025.120133

2026-01-21

Bruna Falgueras Vallbona *

David P. A. Kilgour

Quentin S. Hanley

DOI：10.1016/j.measurement.2025.120133

原文链接

查看原文

摘要

En 中文

Absorbance precision in spectrophotometers has a well-developed theory with best case precision limited by a Poisson distribution. However, this best case is not generally observed in modern commercial instruments and robust empirical models of their behaviour are unavailable. This study evaluated seven models across eight commercial spectrophotometers using a simple method for quantifying absorbance precision between 220 and 640 nm. Replicate measurements of standards were collected to obtain means and relative standard deviations (RSDs). Models were evaluated using root mean square error (RMSE), residual sum of squares (RSS), Akaike information criterion (AIC), and Bayesian information criterion (BIC), and checked for over fitting and robustness. On balance, the most robust and widely applicable model was a weighted log-quadratic function, fitting data from six of eight instruments. The two remaining instruments apply correction mechanisms which our model identifies and can notify the user. Adopting this model can ensure spectrophotometer precision criteria are evaluated and met at each wavelength and working ranges evaluated for any required precision. Our approach enables users to input a desired precision, or percentage relative standard deviation (%RSD), for a spectrophotometer to obtain the absorbance and wavelength range meeting this requirement. Consequently, standard and sample concentrations can be adjusted without compromising instrument precision. Furthermore, if measurements occur outside the desired range, the approach can alert the user and estimate the resulting precision. This readily automatable method enables reporting accurate precision tailored to individual spectrophotometers and wavelengths of interest.

Keyword:

Spectrophotometer

Model fitting

Precision