Servaas Vandenberghe

Straightforward and accurate nonlinear device model parameter-estimation method based on vectorial large-signal measurements

To model nonlinear device behavior at microwave frequencies, accurate large-signal models are required. However, the standard procedure to estimate model parameters is often cumbersome, as it involves several measurement systems (DC, vector network analyzer, etc.). Therefore, we propose a new nonlinear modeling technique, which reduces the complexity of the model generation tremendously and only requires full two-port vectorial large-signal measurements. This paper reports on the results obtained with this new modeling technique applied to both empirical and artificial-neural-network device models. Experimental results are given for high electron-mobility transistors and MOSFETs. We also show that realistic signal excitations can easily be included in the optimization process.

Easy and accurate empirical transistor model parameter estimation from vectorial large-signal measurements

The standard empirical nonlinear model parameter estimation is often cumbersome as several measurement systems are involved. We show that the model generation complexity can be reduced tremendously by only using full two-port vectorial large-signal measurements. Furthermore, realistic operating conditions can easily be included in the optimisation procedure, as we illustrate on GaAs PHEMTs.

New analytic expressions for mutual inductance and resistance of coupled interconnects on lossy silicon substrate

A new analytic model for series mutual impedance of coupled interconnects on lossy silicon substrate is presented. The model includes the frequency-dependent distribution of the current on the silicon substrate (the substrate skin effect). From this model, simple formulas for accurate calculation of the frequency dependent distributed mutual inductance and the associated series mutual resistance of coupled interconnects on a silicon substrate are derived. The validity of the proposed formulas has been checked by comparison with equivalent circuit model data and corresponding full wave solutions. Through this work, it is found that the effect of the semiconducting substrate return path on the transmission behaviour of the interconnects must be well modeled for accurate prediction of the resistance and inductance over the whole frequency range.

Characteristic impedance extraction using calibration comparison

A robust line impedance identification method is presented in this paper. It determines the characteristic impedance of on-wafer thru-line-reflect (TLR) standards measured after an initial off-wafer line-reflect-match or TLR calibration. The only assumption made is that the obtained trans-wafer error boxes are a cascade of a symmetric probe-related disturbance and a change in reference impedance. The proposed method yields an unbiased estimate of the complex characteristic impedance. Results from coplanar lines on a medium resistivity silicon substrate support the made assumption.

Accurate analytic expressions for frequency-dependent inductance and resistance of single on-chip interconnects on conductive silicon substrate

Simple and accurate closed-form expressions to calculate frequency-dependent distributed inductance and associated distributed series resistance per-unit-length of single on-chip interconnects on a lossy silicon substrate are presented. The analytic formulas for the frequency-dependent series impedance parameters are obtained using a closed-form integration method and the vector magnetic potential equation. It is shown that the calculated frequency-dependent inductance L(f) and resistance R(f) per-unit-length are in good agreement with the results obtained from rigorous full wave solutions and CAD-oriented equivalent-circuit modeling approach.

Capabilities of Vectorial Large-Signal Measurements to Validate RF Large-Signal Device Models

The trend towards system-on-chip realisation tightens the design specifications and consequently imposes high accuracy requirements on device models. This paper presents an overview of the surplus value of using vectorial large-signal measurements to validate the large-signal accuracy of RF MOSFET models. We show that these models can be evaluated at operating conditions close to real applications, such as intermodulation characterisation combined with loadpull. The large-signal model verification is not limited to analogue applications, because also the RF large-signal performance of digital circuits, such as inverters, can be examined. In this paper, we focus to the results obtained for the BSIM3v3 compact model and for the in-house developed large-signal look-up table model.

Applicability of non-linear modelling methods based on vectorial large-signal measurements to MOSFETs

Non-linear MOSFET models are mostly indirectly derived from DC, C-V and S-parameter measurements. We present two accurate modelling techniques that directly determine the MOSFET state-functions from high-frequency vectorial large-signal measurements. The parameter optimisation method quickly generates models for specific applications, while the extraction method is preferred to obtain general models.

Compensating Differences Between Measurement and Calibation Wafer in Probe Tip Calibrations - Deembedding of Line Parameters

Differences in probe-tip-to-line geometry and substrate permittivity between measurement and calibration wafer deteriorate measurement accuracy. A technique is presented which characterises the discontinuities near the probe-tip based on the measurement of two lines with different length. Unlike previous methods, the equivalent elements representing the discontinuity are extracted at each frequency point together with the propagation constant and the characteristic impedance of the line. The method has been tested by characterising the differences between coplanar lines on an Alumina calibration and a GaAs measurement substrate. In most cases, a frequency-independent shunt capacitance to ground was found to be the dominating parasitic.

Admittance matrix calculations of on-chip interconnects on lossy silicon substrate using multilayer Green's function

In this paper, a simple method for computation of the shunt admittance matrix of multiconductor interconnects on a general lossy multilayer substrate at high bit rates is presented. The analysis is based on the semi-analytical Greens function approach and the recurrence relation between the coefficients of potential in n and n+1 layers. The electromagnetic concept of free charge density is applied. It allows us to obtain integral equations between electric scalar potential and charge density distributions. These equations are solved by the Galerkin procedure of the method of moments. The new approach is especially suited to modeling 2D layered structures with planar boundaries for frequencies up to 20 GHz (quasi-stationary field approach). The transmission line parameters (capacitance and conductance per unit length) for the given interconnect multilayer geometry are computed. A discussion of the calculated line admittance matrix in terms of technological and geometrical parameters of the structure is given. A comparison of the numerical results from the new procedure with the techniques presented in the previous publications are also provided.

Automatically Controlled Coverage of the Voltage Plane of Quasi-Unilateral Devices

We developed a systematic procedure to efficiently cover the (V1, V2) voltage plane of two-port microwave devices. The method is restricted to (quasi-)unilateral devices, because we assume that the V1(t) does not change when applying an additional a2 travelling voltage wave. By choosing the adequate magnitude and phase of this a2 signal, the V2 of interest can be constructed. We illustrate this method on a HEMT and show that, for the used experimental conditions, only 27 vectorial large-signal measurements are sufficient to cover the (V1, V2) operating region of the device. This is a significant reduction in the number of required measurements for non-linear model generation, in comparison to the classical approach based on multi-bias broadband S-parameter measurements.