应用于高速串行链路的噪声预测部分响应最大似然均衡

发布时间：2018-06-02 04:00

本文选题：信号完整性 + 噪声预测　；参考：《东南大学学报(自然科学版)》2017年01期

【摘要】：为了解决25 Gbit/s及以上的高速串行信号在背板信道传输中的信号完整性问题,研究了能够在高速率和低误码率(BER)之间取得折中效果的部分响应最大似然(PRML)均衡.基于噪声预测法,推导出能够有效降低硬件开销的噪声预测PRML均衡.在具有不同衰减特性的背板信道和不同调制方式下,采用判决反馈均衡(DFE)、PRML均衡和噪声预测PRML均衡3种均衡方式对误码率性能和硬件复杂度进行了研究分析.结果表明,对于25 Gbit/s的非归零(NRZ)信号,噪声预测PRML均衡的误码率性能优于DFE和PRML均衡;而对于12.5 Gbit/s的四电平脉冲振幅调制(4PAM)信号,PRML均衡的误码率性能优于DFE均衡;噪声预测PRML均衡的硬件复杂度优于PRML均衡,更适合应用于高速串行链路中.
[Abstract]:In order to solve the problem of signal integrity of high-speed serial signals of 25 Gbit/s and above in backplane channel transmission, the partial response maximum likelihood (PRML) equalization which can achieve a compromise effect between high rate and low bit error rate (ber) is studied. Based on the noise prediction method, the noise predictive PRML equalization which can effectively reduce the hardware overhead is derived. Under different attenuation characteristics of backplane channel and different modulation modes, the error rate performance and hardware complexity are studied and analyzed by using three equalization methods: decision feedback equalization and noise predictive PRML equalization. The results show that the performance of noise prediction PRML equalization is better than that of DFE and PRML equalization for 25 Gbit/s non-return-to-zero NRZ signals, while the BER performance for 12.5 Gbit/s four-level pulse amplitude modulation (Pam) signal is better than that of DFE equalization. The hardware complexity of noise prediction PRML equalization is better than that of PRML equalization, so it is more suitable for high speed serial link.
【作者单位】：东南大学射频与光电集成电路研究所;
【基金】：国家自然科学基金资助项目(61471119)
【分类号】：TN919

【相似文献】