Günter Kompa’s book, Parameter Extraction and Complex Nonlinear Transistor Models, has a unique perspective that fills a gap in the market today. Read on to find out more:
Unique Determination of Complex FET Small-Signal Equivalent Circuits
It seems that the written book closes an essential gap in the field of FET modeling concerning more complex parasitic networks. Referring to the excellent book edited by Matthias Rudolph, Christian Fager and David E. Root in 2011 titled Nonlinear Transistor Model Parameter Extraction Techniques, it is pointed out that “depending on the layout of GaN high-power HEMTs, distributed effects have to be added to a classical equivalent circuit”. And further on, it is said: “Another issue … is related to the more complex layout of multifinger power GaN HEMTs. In some cases, the source air bridges used are in close proximity to the drain and gate fingers and result in significant additional extrinsic capacitances. These distributed capacitances form along the gate and drain fingers to the source. An alternative equivalent circuit … accounts for the interelectrode capacitances …”.
Direct model parameter extraction methods have proven so far concerning, e.g., the widely-used 15-element FET model. However equivalent circuits including additional distributed parasitic capacitive effects comprise an increased number of model parameters that are commonly determined by optimization-based extraction methods. And, as the experience shows, this is a big problem for any modeler with view to the local minimum problem, and thus “creating” reliable model parameter values.
Manifold optimization approaches have been applied to the parameter extraction. Most important methods are discussed in the book commenting pros and cons based on own experiences and/or published results. These comprise localizers, such as gradient search, Levenberg-Marquardt, and simplex to name a few, as well as globalizers, such as multistart, genetic algorithm, simulated and tree annealing. Furthermore, it becomes evident that the combination of local and global search methods finds increasing acceptance.
There is no doubt that global optimization is a necessity with a huge number of optimization variables, and it needs full control of the processing parameters to get satisfying results. Nevertheless, it must be remembered that it does not guarantee to provide unique equivalent circuit element values.
The book follows the parameter extraction in a consistent way. Regarding cold pinch-off operation, the complex small-signal equivalent circuit consists of three capacitive shells. At low frequencies the distributed gate-source, drain-source and gate-drain capacitances can be combined in each case. And these effective gate-source, drain-source and gate-drain capacitances can be accurately, uniquely, and physically-relevant determined incorporating prior physics- and experimental-based knowledge of the device. Note that means, that under given restrictive starting conditions the optimizer approaches the global minimum. Excellent agreement between simulation and measurement at low, medium and moderate high frequencies verify this approach. With increasing frequency model mismatch becomes obvious: i.e., the distribution of the capacitances must be taken into account. As pointed out in the book, this can be done in different ways, e.g. using model parameter scanning method or reliable top view device analysis. In any case, as this extraction step provides only a refinement of the “low-frequency” model parameter process, the determination of the model elements is unique, accurate and physically significant.
The greatest benefit of the proposed extraction algorithm seems to be that the modeler can rely on the extracted results without having any doubt on the acceptability. This feature is not only important for the device modeler interested in reliable circuit design but in particular also for the device designer interested in technology-relevant feedback knowledge of the fabricated device for further device optimization.
For more information or to order, click here.
