Software defined radio (SDR) has been proposed as a transformative solution to increase the longevity of wireless radios and allow them to be reprogrammed to support new emerging wireless standards and services. A key hardware requirement is to eliminate the use of bulky and costly filters which cannot be reprogrammed anyway. However, by removing the external filters and using broadband power amplifiers and upconverters, spurious signals are introduced by the nonlinearities which interfere with other channels and bands and introduce inband distortion. Thus predistortion linearization techniques need to be introduced in filter-less SDR architecture to ensure compliance with the stringent linearity requirement, especially for the multi-carrier signals exhibiting high envelop fluctuation.
In this work, a FPGA-controlled parallel wireless filter-less transmitter architecture for multi-carrier signals is proposed to address the three key factors modern SDR transmitters challenge: the generation of baseband signal with multiple wireless standards, the linearization of broadband power amplifier, and the linearization of broadband upconverter.
Firstly OFDM signals and multisine signals are implemented using FPGA. BPSK and QPSK modulation are also implemented with adjustable symbol durations. For the experimental demonstration of the linearization techniques, 64-tone OFDM signals of 8 MHz and 10 MHz bandwidth with typical OFDM communication PAPR are adopted.
Then a fully orthogonal frequency-selective baseband predistortion linearization system for RF multi-carrier power amplifiers affected by strong differential memory is proposed. This new scheme implemented in FPGA establishes full orthogonality between the interband and inband predistortion, such that multiple iterative steps between the interband and inband linearizations are no longer required in the optimization. Adjacent channel leakage ratio of up to -45 dBc for inband and interband are achieved. This orthogonal implementation will facilitate the auto-adaptation of the linearization algorithm with a receiver with reduced bandwidth owing to its frequency selective and orthogonal nature.
We also present a new model-based predistortion linearization scheme for single sideband mixers called poly-harmonic predistortion linearization. It relies on an orthogonal expansion in the frequency domain of the nonlinearities for the mixer modeling. It takes into account memory effects which are piece-wise quasi-memoryless and enables the independent cancellation of unwanted spurious sidebands of the digital IF harmonics. An augmented multipath version of the poly-harmonic modulator model is also shown to have the capability to predict the transition from weak to hard nonlinear regimes observed in modulators under high power drive. The ability of the poly-harmonic predistortion algorithm to linearize the 4-path polyphase mixer for input signals with high envelope fluctuation is also experimentally demonstrated.
Combined with the polyphase multipath technique, the capability of this novel transmitter architecture to meet the joint requirements of wide-bandwidth, linearity, power efficiency and frequency agile filtering is demonstrated in this work. The proposed transmitter architecture offers an attractive filter-less approach for the development of multi-mode multi-standard broadband SDR.