Modern mobile communication systems strive to maximize capacity within the system. The capacity of the system to handle multiple subscribers is adversely affected by adjacent channel interference (ACI). ACI results from nonlinear distortion created by the transmitter power amplifier (PA). This is especially problematic for high power base station PAs within the cellular infrastructure. Such distortion produces out-of-band signals that interfere with other subscribers' reception. Thus ACI, effectively raises the noise floor forcing lower bit rates and/or increased transmit power for those subscribers. As medium rate data services for internet access are being introduced into the cellular systems, higher demands are being placed on both capacity and quality of service. For these reasons, minimization of ACI, which requires PA linearization in basestation infrastructure, is of great importance.
Commercially available state-of-the-art base station PAs use feed-forward linearization. While this method is effective in improving ACI, the architecture needs additional components as well as a high power combining. These added components, and the losses they introduce in both DC and RF power, tend to make PAs based on the feed-forward method both costly and inefficient.
Predistortion linearization is another method of linearization. This approach is more cost effective than feed-forward linearization, because the signal processing is done digitally, and there are no components following the PA that waste power. However, the performance of predistortion is generally not as good as feed-forward linearization, because a near perfect model of the amplifier must be extracted and a predistorter model formulated to counteract the distortion mechanisms. One area in which current models are deficient is in characterizing memory effects. Memoryless predistortion has been thoroughly investigated in the literature, but for high peak-to-average signals, such as multicarrier CDMA or OFDM signals, self-heating effects arise, creating delay responses in the distortion that manifest themselves as memory effects. In general, higher power amplifiers, such as those used in basestations, are more susceptible to memory effects than those used in lower power mobile equipment.
This proposal will outline a research program that will investigate modeling of PA memory effects, as well as linearization of PAs that exhibit such effects. The authors propose a research program that will develop new methods in PA characterization and modeling, as well as advance the state-of-the-art in predistortion linearization to achieve or exceed performance levels previously seen only using costly and inefficient feed-forward techniques.