Simulation of 1/fα noise for analytical measurements

• DOI: 10.1002/cem.3137
• Publication Date: 2019-06-20T11:49:18Z
• Authors: Stephen Driscoll; Michael Dowd; Peter D. Wentzell
• Keywords: Gradient noise; Noise spectral density; Value noise; Noise power; Gaussian Noise; Impulse noise

A simple procedure is described that can be used to generate 1/ f α noise, also known as power law in simulated analytical measurement vectors. Certain types of such pink noise ( =1), dominate many signals, so its simulation important optimizing data processing strategies. In this work, error sequences are created directly from white via the theoretical covariance matrix (ECM) by rotation and scaling. The ECM obtained coefficients a finite impulse response filter easily adapted multiplicative probably more common for systems exhibiting proportional characteristics. Simulating offers two main advantages. First, presence other errors additive combination ECMs. Second, model real experimental noise. It shown spectral density function generated proposed method closely approximates behaviour To demonstrate utility evaluating methods, (pink) analyzed maximum likelihood principal component analysis (MLPCA) takes structure into account, baseline using brown test fitting asymmetric least squares (AsLS).