面向GSM-R/LTE-R双模基站功放的通用双带Volterra预失真系统模型及算法

General Dual-Band Volterra DPD System Model and Algorithm for Power Amplifiers of GSM-R/LTE-R Dual-Mode Base Stations

在GSM-R/LTE-R双模基站中,双带射频功放的非线性使得发射信号中2个频带的信号在时域产生混叠,将有可能威胁通信基站的安全。因此,基于通用通带Volterra非线性系统模型,理论推导得到通用双带Volterra预失真系统模型,对其分别进行简化得到双带记忆多项式模型和双带无记忆多项式模型;同时给出双带Volterra预失真系统模型的参数估计算法。以GSM-R信号(带宽200kHz)和LTE信号(带宽5MHz)作为双带的基带输入信号,信号的长度均为2×104个样本,用于预失真参数估计的样本长度为2 000个,分别采用3种模型进行预失真仿真试验。实验结果表明:双带Volterra预失真系统模型不但具有较好的通用性,而且其预失真效果最优,带外频谱抑制达到约20dB,但实现也最难。因此,在实际功放预失真系统应用中,可根据应用场景的需求,选用预失真系统性能与实现复杂度平衡的模型。同时也验证了通用双带Volterra预失真系统模型及其参数估计算法对预失真系统具有重要的实用价值。

In GSM-R/LTE-R dual-mode base stations, the nonlinearity of dual-band RF power amplifier causes the time domain aliasing between the two signals in two different carrier frequencies. The aliasing may threaten the safety of the dual-mode base stations. Based on the general Volterra pass-band nonlinear system model, the general dual-band Volterra digital predistortion （DPD） system model is derived theoretically and simplified into dual-band memory polynomial model and memoryless polynomial model respectively. Moreover, the algorithm for estimating the parameters of dual-band DPD system model is presented correspondingly. Three models are applied in DPD system performance tests. The input signals for dual-band baseband are GSM-R signal with 200 kHz bandwidth and LTE signal with 5 MHz band- width, and the lengths of input signals are both 2 × 10^4 samples. The numbers of samples applied in parameter estimation are 2 000. Simulation results show that dual-band Volterra DPD system model is not only of better universality but also with the best DPD performance. The out-of-band spectrum suppression is about 20 dB, yet it is hard to achieve. Thus in practical DPD system implementation, the model can be determined by the balance between DPD system performance and implementation complexity. The results also demonstrate the important practical value of the general dual-band Volterra DPD system model and the corresponding parameter estimation algorithm.