Adaptive noise cancellation on inductively coupled plasma spectroscopy

Signals from inductively coupled plasma spectroscopy (ICP) are commonly subjected to uncontrollable fluctuations, as a direct result of shot noise and plasma fluctuations. Since the first is random from nature, the signal-to-noise ratio can be improved by enlarging the number of photon counts. Plasma fluctuations exhibit correlation in time and can therefore be approached by adaptive signal processing methods. The concept of adaptive noise cancellation (ANC) has been applied on ICP spectroscopy in order to improve the signal-to-noise ratio and to eliminate the influence of plasma fluctuations. ANC is based on the assumption that the colored noise residing in the measured signal can be estimated using a correlated reference signal, processed by an adaptive filter. Since noise is canceled out rather than filtered out, ANC generally outperforms conventional filtering methods. It is demonstrated by this feasibility study that artificial neural networks (ANN) can successfully be applied as noise cancelers. In this study, colored noise has artificially been imposed by computer controlled temporary power interruptions and pulsed sample injections. The results of neural networks (an Adaline network and a multi-layer feed-forward network (MLF)) are compared to the more conventional Kalman filter. Additionally, the sensitivity to deviations due to changing experimental conditions is investigated by means of an experimental design.