Suppression of measurement errors in effective-MOSFET-channel-length extraction

The effects of measurement errors in effective-MOSFET-channel-length extraction are estimated, both numerically and experimentally. In the numerical estimation, the errors in the least-squares method, which was used twice in the extraction, are calculated for several reference MOSFET combinations. In the experimental estimation, the errors in the extracted channel lengths were estimated, changing the reference MOSFET combinations. It is found, from these estimations, that choosing both short- and long-channel MOSFETs as reference MOSFETs is effective in suppressing measurement errors.<
>