Das Projekt "Entwicklung eines On-line Partikeldetektors" wird vom Umweltbundesamt gefördert und von Universität Duisburg, Fachbereich 9 Elektrotechnik, Fachgebiet Prozess- und Aerosolmesstechnik durchgeführt. In semiconductor technology clean process gases must be provided to manufacture large scale integrated chips. To prevent micro contamination of wafers, the process gases must be controlled by on-line detection of suspended particles. In this project the on-line particle detection is realized by means of optical method. Due to the low particle concentration, the entire cross section of the gas pipe has to be controlled. Thus it is necessary to install an illuminating plane of over 32 mm2 perpendicularly to gas flow. In clean gases paricles down to 0,1 my m must be detected. In this size range molecule noise leads to False-Count-Rate. To reduce the F.C.R. an additional illuminating plane is arranged parallel to the original one, and the F.C.R. is decreased via anew Pipeline-Cross-Correlation-Function, which is developed in this project. Using this P.C.C.F. method the influence of molecule noise on particle detection can be strongly reduced, and detection becomes possible without any dead time. To create the illuminatingplanes non-imaging optical elements are used, where beam bend is caused by reflection and diffraction exclusively. Applying computer simulation, the optical characteristics of those non-imaging optics are investigated, and it is shown that the scattered light intensity is nearly independent on local position of detected particles in the plane. The application of the non-imaging optics leads to high precision combined with low manufacturing costs. The results of optical simulation show that the pre-condition to detected particles with non-condition to detected particles with non-imaging optics in clean process gases in fulfilled, i.e. the entire gas flow is controlled, and the low detection limit is expanded down to 0,1 micron and local dependence of detector is smaller than 4 percent within measuring volume 6,4 mm x 6,4 mm. The algorithm of P.C.C.F. is based on classical cross-correlation. The difference between them is that output result from P.C.C.F. is a function of real time, t, but not a function of delaying time, t, which corresponds to the time difference between the two signals. Therefore, it is possible to set P.C.C.F. in on-line measurement. To examine the efficiency of P.C.C.F., a particle signal mixed with two acquired noise are simulated on computer. The results of simulation show that, in output process of P.C.C.F., the noise is reduced and the level of particle signal becomes higher. The signal-noise reatio is improved from one to four.