Multivariate Analysis of a-SiGe:H Multispectral Photodiodes for Proper Band Selection

The demand for low cost and disposable devices has driven the development of intelligent photodiodes, especially multispectral photodiodes that were first manufactured by Rieve et al. in 2000 [1]. The most remarkable feature of these devices is the voltage-controlled spectral response. In this paper we present a-SiGe:H based bias sensitive ni 3 p photodiodes that have been fabricated successfully in a low temperature PECVD process. Multispectral diodes based on ni 3 p structures make use of an intrinsic layer divided into three regions. Further subdivision of the i-layer into different collection regions is the key to an unequivocal optical detection. In addition to the a-SiGe:H multispectral photodiodes developed at the IMT, there is an inestimable wealth of different sensor approaches that has been optimized for different applications. To optimize the separation of secondary colors (i.e. spectral signatures) it is necessary to separate only those response curves (i.e. bands) with a very high information density [2, 3]. In our case, samples of whitish powder suspected to be dangerous or illegal must be unequivocally characterized. There is need for an application specific band selection method by which various sensors can be compared and evaluated. Unfortunately it is not sufficient to optimize the spectral response only as discussed in [4]. The requirements for measuring surroundings and the mechanical handling of all parameters together form a multivariate data set. Examined were in fact more than 5,000 measurement setup combinations, fictional and real, each considering 10 parameters that partially influence each other. The influence of all these parameters has to be examined, with the awareness of multivariate analysis [5].