The ELPI Response and Data Reduction II: Properties of Kernels and Data Inversion

The first part of this article reports the analytical form of the Electrical Low Pressure Impactor (ELPI) kernel functions. In this latter part, the numerical quality of ELPI response matrices is studied and an example of an inversion algorithm is given. The ELPI assemblies with and without an electrical filter stage and with smooth or sintered impaction plates are studied and compared with basic impactor kernels and the kernels for the calculation of the aerosol mass distribution. It is shown that the ELPI assembly with the electrical filter stage and smooth impaction plates should be the best choice for the inversion of data if no bounce occurs. The comparison to a mass impactor shows that the devices are on par in data inversion. The inversion ELPI data is studied with a Bayesian algorithm assuming a bimodal lognormal size distribution of the aerosol. The algorithm includes a novel procedure for obtaining an initial guess of the distribution parameters. To our knowledge, it is also the first algorithm to use ELPI current readings as its input. Simulations and diesel emission measurements show that the proposed algorithm is a useful tool in the study of ELPI data.