Accurate prediction of PHEMT intermodulation distortion using the nonlinear discrete convolution model

Abstract

A k i a d -A general-purpose, technology-independenl behavioral model is adopted for the intermodulation performance prediction of PHEMT devices. The model can be easily identified since its nonlinear functions are directly related to conventional DC and small-signal differential parameter measurements. Experimental resulta which confirm the model accuracy at high operating frequencies are provided in the paper.

Accurate pHEMT nonlinear modeling in the presence of low-frequency dispersive effects

“…Moreover, the wellknown and largely adopted pulsed -techniques [14]- [18] are based on the assumption that the state , which is associated with the device quiescent point ( , ) before the application of the superimposed voltage pulses, is not perturbed by either the pulse "shape" or amplitudes (i.e., by second and higher moments of the pulse waveforms) during the measurement of the device currents. Thus, (1) can be replaced by (2) According to conventional approaches, will be considered in the following as being linearly dependent on the average dissipated power under dynamic conditions , i.e., (3) where is the case (or substrate) temperature and is the device thermal resistance.…”

“…The LF model presented in the previous sections was embedded into a nonquasi-static large-signal device model for microwave and millimeter-wave applications, i.e., the NDC model [2]. To this aim, the NDC model was identified on the basis of bias-and frequency-dependent -parameters carried out by means of a 110-GHz VNA (HP8510XF).…”

“…in the third-order IMD as defined in (22) considering: Z = 48.9 + j 3 54.7 and Z = 14.4 + j 3 9.7. Measurements (fulfilled bar) are compared to predictions based on the NDC model [2] with the new empirical model embedded (dashed bar) and with the purely static dc I-V characteristic (dotted bar).…”

“…12 have been taken into account; the corresponding values are shown for different output power levels. In particular, for each output power condition, measurements (fulfilled bar) are compared to predictions based on the NDC model [2] with the presented LF model (dashed bar) embedded. In the same figure also, the poor prediction capability due to the lack of the LF dispersion modeling, obtained with the purely static -characteristics, is shown (dotted bar).…”

“…Two slightly different formulations are proposed, according to the degree of accuracy pursued in the description of above-cutoff thermal dispersion, while the discussion on the extraction procedure points out how different experimental large-signal techniques can be exploited. In Section III, the LF -model is identified for a 0.25-m Triquint GaAs pseudomorphic high electron-mobility transistor (pHEMT) 600 m and is then embedded into a largesignal device model for microwave and millimeter-wave applications (namely, the nonlinear discrete convolution (NDC) model [2]), and the experimental validation at both LFs (2 MHz) and high frequencies (microwaves) is presented. Finally, intermodulation distortion (IMD) predictions of a pHEMT-based highly linear power amplifier (PA) at -band, designed and realized by Ericsson Laboratory Italy, are shown.…”