Phase errors in accelerometer arrays: An analysis based on collocated sensors and FEM

• Analysis of an array of identical accelerometers on a shaking table.
• Errors derived from comparison of FEM-predictions and observations.
• Acceleration measurements characterized by non-Gaussian dynamic noise.
• Evidence of significant phase noise (jitter) during strong motions.
• Significant differences between identical instruments.

The error properties of arrays of accelerometers were studied on the basis of systematic experiments and analysis of the differences in the recordings of collocated sensors. Four identical tri-axial force-balance accelerometers with common recordings and time stamping, were set at symmetrical positions on a prototype oscillator, a T-type inverted pendulum, and were subject to controlled excitations. Based on Finite Element Modeling (FEM) the response of the oscillator was modeled and there were identified time series of acceleration of different collocated sensors predicted to be identical (“comparable”). The differences of the corresponding “comparable” measurement time series, which are functions only of noise, were found important, indicating that the accelerometer arrays are characterized by significant dynamic noise and not by low-amplitude Gaussian noise, as is widely assumed.Further analysis of time series of accelerometer noise was based on spectral analysis techniques in the frequency and the time domain and on coherence, on filtering and on cross-correlation, as well as on examination of the impact of noise in the drift (errors computation of displacements using double numerical integration).The output of this study is that phase noise (jitter, or phase instability) characterizes accelerometers, as in the case with all other array sensors and sensor networks and this is a main source of dynamic noise. The latter becomes significant especially at the level of sampling interval (0.005 sec in our experiments), characterizes strong motions, is not constant and may be regarded as instrument-specific noise. Strategies to overcome this dynamic noise important for the modeling of structural response are discussed.