Artificial neural network based modeling of GaAsHBT and power amplifier design for wireless communication system

The power amplifier (PA) used in modern wireless system needs to perform better from point of view linearity and efficiency. A crucial perquisite for the design of PA is the availability of suitable device models. Artificial neural networks (ANN) recently gained attention as a fast and flexible way to develop HBT device model when compared to the conventional modeling approach based on empirical equations and can demonstrate better accuracy. The framework for this ANN based model is a common-emitter large-signal equivalent circuit model, which has been implemented in Agilent Advance Design System (ADS) simulation environment. An excellent agreement between modeled and bias dependent DC and S -parameters and harmonic power in non-linear mode of operation was obtained. Using developed ANN based HBT model PA design was carried out. At collector voltage V-c of 34V, the power amplifier shows an excellent linearity (first ACPR <-42.2dBc) up to 28dBm of rated output power for CDMA applications. At the rated output power level, PAE was found to be more than 35%. All the CDMA wireless PA specifications like gain, power-added efficiency (PAE) and adjacent channel power rejection (ACPR) are achieved over nominal and extreme temperature conditions.