改进型并行双非线性模型的ROF系统非线性建模

An Augmented Parallel Two-nonlinear Model Based Modeling for ROF Systems

包含光载无线通信（ROF）链路和射频功放的ROF前传系统存在严重的静态非线性和记忆效应,提出了改进型并行双非线性广义记忆效应模型（APTGMEM）,以准确建模ROF前传系统的强静态非线性和记忆效应。所提出的APTGMEM包含无记忆和广义记忆效应两个独立部分,其中无记忆部分只考虑高阶静态非线性,广义记忆效应部分只考虑低阶次超前广义非线性。文中采用带宽为20 MHz的4G LTE（Long Term Evolution长期演进）信号为测试信号,光模块和峰值输出功率为51d Bm的Doherty功放被用于模型有效性验证。测试结果表明,APTGMEM相比广义记忆多项式模型（GMPM）,模型系数减少40%,而建模精度相似,同时比并行双非线性两厢模型（PDNTBM）建模精度提高了3.6 d B,其在建模精度、系数数量和训练复杂度上取得了良好折中。

An Augment Parallel Two-nonlinear General Memory Effect Model（ APTGMEM） is proposed in this paper for modeling a wideband ROF（ radio-over-fiber） Front Haul Systems with strong memory effects and static nonlinearity. The APTGMEM contains two independent parts as non-memory and generalized memory polynomial sub-models. The non-memory part just considers high-order static nonlinearity,while the generalized memory polynomial part considers the lower order leading general memory effect. To evaluate the performance of APTGMEM,a 20 MHz LTE test signal,an optical module and a Doherty power amplifier with 51 d Bm output peak power are used in modeling measurement. Compared to the Generalized Memory Polynomial Model（ GMPM）,APTGMEM reduces the model coefficient by 40%,remaining the similar modeling accuracy,while compared to the Parallel Dual-nonlinear Two-box Model（ PDNTBM）,APTGMEM leads to 3.6 d B improvements in normalized mean square error（ NMSE）. APTGMEM has a good compromise in modeling accuracy,coefficient quantity and training complexity.