In this study, the local variation of the effective channel reduction parameter ( ${\Delta }{L}={L}_{\text {m}}-{L}_{\text {eff}}$ ) of a MOSFET is extracted by means of the traditional shift-and-ratio (SAR) method. ${\Delta }{L}$ is then correlated with the threshold voltage difference ( ${\Delta }{V}_{\text {TH}}$ ) between the device under test (DUT) and the reference device. It is demonstrated that there exists an optimal ${V}_{\text {G}}$ range for extracting reliable values of ${\Delta } {L}$ through the SAR method. Statistical data analysis shows that for ${R}\approx {(}{L}_{\text {long}}/{L}_{\text {short}}{)}\approx {{25}}$ , better results are achieved since the value of $\sigma {(} {\Delta } {L}{)}$ varies typically as the reciprocal 1/ $\sqrt {W}$ . The test structure used in this work is a Kelvin-based 2-D addressable MOSFET matrix implemented in 180-nm bulk CMOS technology. The sample space is of 2304 devices divided into nine subgroups of 256 same size closely placed nMOSFETs.

中文翻译：

---

位移比法提取的有效通道长度的局部变异性评估

在本研究中，有效信道减少参数的局部变化（ ${\Delta {L}={L}_{\text {m}}-{L}_{\text {eff}}$ ) 的 MOSFET 是通过传统的移位和比率 (SAR) 方法提取的。 ${\Delta {L}$ 然后与阈值电压差相关（ ${\Delta {V}_{\text {TH}}$ ) 在被测设备 (DUT) 和参考设备之间。证明存在一个最优 ${V}_{\text {G}}$ 提取可靠值的范围 ${\Delta } {L}$ 通过 SAR 方法。统计数据分析表明，对于 ${R}\approx {(}{L}_{\text {long}}/{L}_{\text {short}}{)}\approx {{25}}$ , 获得更好的结果，因为值 $\sigma {(} {\Delta } {L}{)}$ 通常变化为倒数 1/ $\sqrt {W}$ . 本工作中使用的测试结构是基于开尔文的 2-D 可寻址 MOSFET 矩阵，采用 180-nm 体 CMOS 技术实现。样本空间由 2304 个器件组成，分为 9 个子组，每组 256 个相同大小且紧密放置的 nMOSFET。