The 1/f~2 noise measurement technique used for determination of electromigration lifetime has had only limited practical success. The correlation of noise magnitude ot lifetime is high but teh results are inconsistent. In this paper the source of the inconsistent results is identified and demonstrated. The presence of voltage transents in noise data, caused by abrupt changes of resistance is a thin-film test structure, is shown to be the primary source of 1/f~2 noise in the 1-10 Hz frequency range. The transients do not appear consistently in the data and are strongly affected by typical data handling techniques such as windowing. The combination of these factors explains the inconsistent results obtained when using the 1/f~2 noise measurement technique.
展开▼
