基于低采样率模数转换器的延时复用频分多址无源光网络

基于正交频分复用技术的无源光网络中,光网络单元为了获得其所属小部分下行数据,需高采样率模数转换器将所有频宽的信号恢复才能分出其所需要数据.同时正交频分信号峰均比很高,传输中容易引起非线性效应.为此,本文提出一种基于低采样模数转换器的延时复用频分多址无源光网络.在光线路终端将数据序列交错排序并在时域映射为正交幅度调制信号;再通过离散傅里叶变换扩频技术,将信号转换为频域信号并映射到子载波上.通过预先发送和回传训练信号,估测包括延时采样和低采样接收在内的信道频响;再将频域信号利用估测信息在光线路终端做预处理,从而使信号传输中的失真得到有效预补偿.本文实验演示了含有多个光网络单元的系统,对于含有M个光网络单元的无源光网络,模数转换器的采样率可以降低到1/M Nyquist采样率,实验中模数转换器的采样率可以降低到1/32 Nyquist采样率;由于下行信号通过光线路终端预处理实现失真预补偿,光网络单元接收到的信号不需要均衡,不需要傅里叶变换和傅里叶逆变换,避免了与之对应的相关计算量,降低了光网络单元的计算复杂度;由于使用了扩频技术,信号波形具有更低的峰均比,从而降低了非线性对信号的影响,增加了功率预算.此外,随着光网络单元的增加,信号的误码率几乎没有增加,光网络单元个数增加到32时,向前纠错极限为10~(-3)的功率代价小于0.5 dB;系统对光网络单元采样时刻偏离具有一定容限;25 km光纤传输的功率代价大约0.5 dB.理论和实验均证明本方案能够简化光网络单元,降低无源光网络的成本;与传统的无源光网络相比具有明显优势.

Delay division multiplexing orthogonal frequency-division multiple access passive optical networks using low-sampling-rate analog-to-digital converter

In traditional orthogonal frequency-division multiple access passive optical networks (OFDMA PON) or time-division multiplexing access (TDMA) based OFDM PONs, analog-to-digital converters (ADCs) with a high sampling rate are required to demodulate high-speed aggregated OFDM data in order to receive a small portion of the downstream data at optical network users (ONUs). Meanwhile, OFDM signal has a higher peak-to-average power ratio (PAPR) than the single carrier signal, which can result in the nonlinear effect. The resulting nonlinearity reduces the received signal performance. To enhance practicability of the present PONs, according to the sub-Nyquist sampling theory, we propose and detail a delay-division-multiplexing (DDM) scheme to enable a FDMA PON with low-sampling-rate ADCs. Based on pre-allocated relative time delays among the ONUs and discrete Fourier transform spread (DFT-S) technique, pre-processed signals sent from an optical line terminal (OLT) can be detected as different downstream signals following spectral aliasing caused by ADCs operating at a sub-Nyquist sampling rate. In the proposed scheme, as the signal distortion introduced by the propagation, aliasing and time shifted sampling is pre-compensated, the DFT and inverse discrete Fourier transform (IDFT) are unnecessary for de-mapping and picking out the signal at ONUs. Therefore, the proposed DDM scheme greatly enhances cost efficiency and enables a reduction in computational complexity. Meanwhile, DFT-S FDMA signal has low PAPR, which relieves the nonlinear effect in signal E/O conversion and transmission. As a result, the proposed scheme benefits the power budget of the OLT and power consumption of the ONUs. In experiment, we demonstrate that each ONU with an ADC operating at 1/2-1/32 of the Nyquist sampling rate is able to receive 1/2-1/32 of the downstream data, with an insignificant performance penalty. Furthermore, the details of the matrices that include channel response, aliasing and time delay are first analyzed. In addition, training symbol is very important for estimating the channel response, and how to derive and design training symbols is the first study to outline the details of this issue. The effects of fiber dispersion and the sampling instant of an ADC on signal performance are also studied. The results show that the signal performance has some degree of tolerance to sampling instant deviation and the power penalty is less than 0.5 dB to achieve a forward error correction limit of 10^-3 after 25 km fiber transmission. The theoretical analysis and experimental results indicate that the proposed scheme can simplify the ONU and reduce the cost of the PON.