Device modeling using non-parametric statistical determination of boundary data vectors

摘要

In accordance with the preferred embodiment of the present invention, an electronic component is modeled. Performance of the electronic component is measured to produce a plurality of data vectors representing actual performance of the electronic component. Boundary vectors are selected from the plurality of data vectors. A density estimate is calculated for each data vector in the plurality of data vectors. A first selected number of the data vectors are partitioned from the plurality of data vectors into a partitioned group. The partitioned group includes the first selected number of data vectors from the plurality of data vectors with greatest density estimates. A second selected number of the partitioned group is selected as candidates for boundary vectors. The candidates for boundary vectors include the second selected number of the partitioned group with lowest density estimates. The boundary vectors are selected from the candidates for boundary vectors. From the boundary vectors, values are calculated for boundary model parameter vectors which model the electronic component. The nominal vector is chosen from the plurality of data vectors as the one with the highest density estimate.