Millimeter-Wave FET Nonlinear Modelling Based on the Dynamic-Bias Measurement Technique

Abstract

In the paper, the nonlinear model of a microwave transistor is extracted from large-signal measurements acquired under “dynamic-bias” operation. Specifically, the transistor is driven by low-frequency large signals while a high-frequency tickle is applied on top of them. The low-frequency large signals, along with the dc bias voltages, set the large-signal operating point which represents a dynamic-bias condition for the device under test. Thanks to this technique, one can get at once and separately the nonlinear currents and charges of the transistor as a result of a very few nonlinear measurements. Additionally, the proposed technique allows one to accurately reconstruct the time-domain waveforms at the device-under-test terminals while the frequency of the tickle can be set as high as the bandwidth of today’s vector calibrated nonlinear measurement systems (i.e., 67 GHz). The approach, which is general and independent of device technology, is applied on a 0.15- m GaAs pHEMT specifically designed for resistive cold-FET mixer applications

Hybrid Nonlinear Modeling Using Adaptive Sampling

“…Specifically, initial values of the parasitic elements can be extracted from 'cold' FET measurements with the channel pinched and the channel open [19]. Next, multibias Y-parameters, derived from multibias S-parameters, can be used to derive initial estimate for the nonlinear capacitances model [20]. Finally, the main parameters of the nonlinear currents can be extracted by few dc measurements in linear and saturation region.…”

A Non-Quasi-Static FET Model Extraction Procedure Using the Dynamic-Bias Technique

“…DOUBLE-TICKLE DYNAMIC-BIAS TECHNIQUE The dynamic-bias technique [6] exploits the possibility of setting the traps-occupation and thermal states of the transistor by a low-frequency (LF) multi-harmonic load-pull system [5]. At LF the transistor operation, from mixing [6] to high-efficiency power amplification [7], can be simply forced avoiding the use of expensive microwave instrumentation and overcoming LSNA bandwidth limitations. Then, a high-frequency (HF) tickle is superimposed on the LF large-signal excitation for modelling the strictly-nonlinear dynamic effects (i.e., nonlinear capacitances).…”

“…It is worth noticing that, by definition [6] and due to the small-signal HF regime, the spectral content around the tickle harmonic frequencies can be neglected as well as every intermodulation product arising from the interactions between the two tickles. 1531-1309 © 2015 IEEE.…”

“…Since, at the beginning, the parasitic elements have to be identified, the intrinsic device is not accessible. Nevertheless, the tickle frequencies can be selected using the initial values of the model parameters, which are identified starting from a small set of dc and S-parameter measurements [6]. At the end of the extraction procedure, when the final values of the parasitic elements are available, the choice can be definitely validated or possibly adjusted.…”

“…In conclusion, differently from [6], [7], the proposed approach requires one additional HF tickle in the measurement phase ( Fig. 1) and one additional optimization step in the extraction phase.…”