162 CHAPTER 6 CONCLUSION AND FUTURE SCOPE 6.1 Conclusion Today's 3G wireless systems require both high linearity and high power amplifier efficiency. The high peak-to-average ratios of the digital modulation schemes used in 3G wireless systems require that the RF power amplifier should maintain high linearity over a large range while maintaining its high efficiency. These two requirements are often at odds with each other. Thus, to maintain high efficiency and to comply with regulatory specifications regarding the maximum allowed disturbance in the adjacent channels, mitigation of the non-linear distortions is of great significance. Non-linear equalization at the receiver can be a possible solution but is complicated due to the unknown effects of the channel. It is therefore easy to reduce the non-linear distortions at the transmitter side. Linearization techniques can be used to mitigate the nonlinear distortions in a power amplifier. But the linearization task can t be accomplished without properly characterizing and modeling a practical power amplifier. Modeling of power amplifier has been a subject of intense research during the last few years. For understanding the characteristics of power amplifier different parameters like inter-modulation products, Adjacent Channel Leakage Ratio, efficiency, CF, Error Vector Magnitude and PCDE can be used. Power amplifier can be better modeled using behavioral modeling. Non-linear behavioral models can be divided into three types; memory-less, quasi memory-less and models with memory. Majority of the memory models available in literature are based on Volterra series, which is used to represent the non-linear systems. Memory polynomial, which is reduced
163 form of Volterra series, can be more suitable to represent the actual behavior of the power amplifier with memory. But available memory polynomial models are not easy to implement. A simplified and easy to implement memory polynomial model of power amplifier called Complex Memory Polynomial model has been proposed and implemented. Although this proposed model can be employed to characterize power amplifier with reasonable accuracy under certain conditions, yet there is no systematic way to check the validity of this model. The effect of memory depth and the order of memory polynomial have also been studied for a W-CDMA signal. It has been concluded that non-linearity does not show much effect as the memory depth and the order of memory polynomial is increased. A novel approach, called Black Box has been proposed in this thesis, which directly derives the power amplifier model from its circuit topology. In this approach complex S-Parameters of power amplifier are derived for Motorola LDMOS power amplifier circuit available in Agilent library. Comparison of the proposed power amplifier model, with the actual power amplifier circuit topology, show that the proposed model exactly models the behavior of the actual power amplifier. From the comparison of existing linearization techniques like Boot Up Bias, Dynamic Bias, Baseband Envelope Feedback, Polar Feedback, Cartesian Feedback, Envelope Elimination and Restoration, Adaptive Feedforward, RF/IF Predistortion, Digital Predistortion, it has been concluded that Digital Predistortion technique can be of main concern. It has also been concluded that although Digital Predistortion shows moderate complexity and high flexibility, Feedforward linearization technique shows more Intermodulation distortion reduction and greater bandwidths over Digital Pre-distortion technique.
164 A novel linearization technique named Adaptive Feedforward Linearizer combined with Adaptive Digital Predistorter has been proposed. To develop this technique a Complex Memory Polynomial based Adaptive Digital Pre-distortion technique and Adaptive Feedforward Linearization technique has also been proposed and presented. For optimization of the parameters, a new adaptation technique with much fast convergence rate, based on the conjugate-gradient method developed by Fletcher and Reeves has been proposed. A new step by step optimization methodology has been proposed to achieve optimized parameters of complex gain adjusters in Digital Pre-distorter and complex gain adjusters in signal cancellation and error cancellation loops in Adaptive Feedforward Linearizer. As the Field Programmable Gate Arrays have been gaining considerable attraction in high performance digital signal processing applications, the proposed linearization techniques have also been implemented on a Virtex 4 Field Programmable Gate Array for comparison of resource utilization and power consumption. Performances of all proposed linearization techniques have also been compared for single, two and three carrier W-CDMA signal by evaluating inter-modulation distortion reduction, Error Vector Magnitude reduction and Adjacent Channel Leakage Ratio improvement. Comparisons show that although Adaptive Feedforward Linearizer combined with Adaptive Digital Predistorter results in only10 db reduction in 3 rd order IMD as compared to 30 db reduction by Complex Memory Polynomial based Adaptive Digital Pre-distortion technique and 17 db by Adaptive Feedforward Linearization technique, yet it results in 45 db reduction in 5 th order IMD, 60 db reduction in 7 th order IMD, 30 db reduction in 9 th order IMD and 7dB reduction in 11 th order IMD as compared to 36 db reduction in 5 th order IMD, 53 db reduction in 7 th order IMD, 27 db reduction in 9 th order IMD and 0dB reduction in 11 th order IMD by Adaptive Feedforward Linearization technique and 10 db
165 reduction in 5 th order IMD, 36 db reduction in 7 th order IMD, 19 db reduction in 9 th order IMD and 0dB reduction in 11 th order IMD by Complex Memory Polynomial based Adaptive Digital Pre-distortion technique. Comparisons also show that among the proposed linearization techniques, Adaptive Feedforward Linearizer combined with Adaptive Digital Predistorter outperforms in terms of Error Vector Magnitude reduction and Adjacent Channel Leakage Ratio improvement. Adaptive Feedforward Linearizer combined with Adaptive Digital Predistorter reduces Error Vector Magnitude to 16.81% as compared to 17.20% by Complex Memory Polynomial based Adaptive Digital Pre-distortion technique and 17.46% by Adaptive Feedforward Linearizer. Also Adaptive Feedforward Linearizer combined with Adaptive Digital Predistorter improves Adjacent Channel Leakage Ratio by 56.98 db as compared to 44.19 db by Adaptive Feedforward Linearizer and 42.33 db by Complex Memory Polynomial based Adaptive Digital Pre-distortion technique. Comparisons also show that among the proposed linearization techniques, Adaptive Feedforward Linearizer combined with Adaptive Digital Predistorter seems to be costly in terms of resource utilization. It consumes 4568 Registers, 4213 Look up Tables, 3754 Slices, 44 BRAMs and 31 DSP48s as compared to 3652 Registers, 3654 Look up Tables, 2987 Slices, 38 BRAMs and 29 DSP48s by Adaptive Feedforward Linearizer, 3516 Registers, 3758 Look up Tables, 3431 Slices, 60 BRAMs and 19 DSP48s by Look up Table based Digital Pre-distorter and 2547 Registers, 2695 Look up Tables, 2862 Slices, 32 BRAMs and 22 DSP48s by Complex Memory Polynomial based Adaptive Digital Predistortion technique. Also Adaptive Feedforward Linearizer combined with Adaptive Digital Predistorter consumes 1052 mw of power as compared to 982 mw by Adaptive Feedforward Linearizer, 845 mw by Complex Memory Polynomial based Adaptive Digital Pre-distortion technique and 799 mw by Look up Table based Digital Pre-distorter.
166 Thus form the results presented it has been concluded that selection of a particular linearization technique requires a compromise between improvement in inter-modulation distortion, Error Vector Magnitude, Adjacent Channel Leakage Ratio and Field Programmable Gate Array resource utilization. The proposed Adaptive Feedforward Linearizer combined with Adaptive Digital Predistorter linearization technique can be used in wireless base station transmitters where high degree of improvement is required in terms higher order inter-modulation distortion, Error Vector Magnitude and Adjacent Channel Leakage Ratio. The proposed Complex Memory Polynomial based Adaptive Digital Predistortion technique can be used in wireless base station transmitters where Field Programmable Gate Array resource utilization is of main concern and Adaptive Feedforward Linearizer can be used in wireless base station transmitters where both moderate degree of improvement in inter-modulation distortion, Error Vector Magnitude, Adjacent Channel Leakage Ratio and moderate Field Programmable Gate Array resource utilization is required. 6.2 Scope for Future Work Future work can focus on introducing some kind of parallelism into the algorithm. An approach would be to convert the modules into hardware description modules which can process in parallel while the rest of the operations can be done in soft processor. This may speed up the whole algorithm, but can take up some FPGA resources as well. This tradeoff requires more investigation. The results from these investigations could be used to develop any general algorithm or a class of algorithms. Further the implementation of adaptation algorithm using genetic algorithm or neural networks may be a future area of concern.