Iterative De-Embedding and Extracted Maximum Oscillation Frequency $f_{\text{MAX}}$ in mm-Wave InP DHBTs: Impact of Device Dimensions on Extraction Errors

InP/GaAsSb DHBTs with varying emitter lengths $\pmb{L}_{\text{E}}$, emitter widths WE, and base metal widths $\pmb{W}_{\mathbf{B}}$ are characterized to determine their maximum oscillation cutoff frequency $f_{\text{MAX}}$ using iterative de-embedding in comparison to the standard OPEN-SHORT and SHORT-OPEN de-embedding schemes. With increasing measurement frequencies the OPEN-SHORT and SHORT-OPEN methods become increasingly subject to pathologies in the Mason Unilateral power gain which corrupt the extracted $f_{\mathbf{MAX}}$. In contrast, iterative de-embedding is free of such complications. In the present work, we characterize the $f_{\text{MAX}}$ extraction error introduced by the usual de-embedding techniques with respect to iterative de-embedding as a function of DHBT dimensions $\pmb{L}_{\mathbf{E}},\pmb{W}_{\mathbf{E}}$, and $\pmb{W}_{\mathbf{B}}$ to reveal that extracted $\pmb{f}_{\mathbf{MAX}}$ values are especially sensitive to the emitter width $\pmb{W}_{\mathbf{E}}$. De-embedding errors appear to carry over to extrinsic collector area which sensitively affects $f_{\text{MAX}}$.