摘要
受隧道磁阻比(Tunnel Magneto Resistance,TMR)下降、工艺偏差和热波动等因素影响,先进工艺节点(亚25 nm)下的自旋转移力矩磁随机存储器(Spin Transfer Torque Magnetoresistive Random Access Memory,STT-MRAM)的读取裕度降低,读写正确率下降。纠错码(Error Correction Codes,ECC)技术方案能有效提升STT-MRAM读取可靠性。但是,常见的一位硬判决量化器无法发挥多位量化软判决ECC编解码算法的优势,而基于模数转换器的软判决量化器在实现上又要以牺牲面积和功耗为代价。提出的面向STT-MRAM的两位量化器采用全裕度灵敏放大器(Sensitive Amplifier,SA)和电阻量化判决门限,显著降低了量化器的电路复杂度和读写错误率。结果表明,在TMR和低读取裕度条件下,基于提出的两位量化器和极化码(Polar)编码的ECC算法能实现优于基于一位量化器的里德-索洛蒙博斯-乔赫里-霍克文黑姆码(Bose,Chaudhuri&Hocquenghem,BCH)编解码方案的输出帧错误率(Frame Error Rate,FER)。此外,本文提出在STT-MRAM级联信道模型中考虑实际SA引入的读判据错误,用于准确评估读取裕度对读正确率造成的影响。结果表明,基于全裕度SA的两位电阻量化方案能够显著提升判决器的有效读取裕度,减小SA引入的读判据错误,在TMR≤90%时,实现比采用一位硬判决量化器的BCH码平均低47%的输出FER。
Due to the process and thermal fluctuations as well as the reducing of the Tunnel Magneto Resistance(TMR),the read margin of the Spin Transfer Torque Magnetic Random-Access Memory(STT-MRAM)at sub-25 nm process node narrows,which prevents getting desire reading accuracy.To take advantage of the multi-bit soft-decision Error Correction Codes(ECC)coding and decoding algorithm,in this paper,we proposed a two-bit quantizer for the STT-MRAM by using the full-margin Sensitive Amplifier(SA)and Magnetic Tunnel Junction(MTJ)-based threshold.The results show that in the case of small TMR and low read margin,the polar code decoder based on the proposed two-bit quantizer can achieve better output Frame Error Rate(FER)than the Bose,Chaudhuri&Hocquenghem(BCH)scheme based on the one-bit quantizer.We also considered the read decision errors of SA in the STT-MRAM cascaded channel model to accurately evaluate the effect of read margin.The results indicates that the full-margin SA can significantly broaden the effective read margin of the second digit verdict.Moreover,since the induced read decision error decreases,the average output FER becomes 47%lower than that using BCH code with one-bit hard decision quantizer when TMR≤90%.
作者
赖怡桐
唐慧琴
陈平平
王少昊
LAI Yitong;TANG Huiqin;CHEN Pingping;WANG Shaohao(Jinjiang Joint Institute of Microelectronics,Fuzhou University,Jinjiang 362200,China)
出处
《微电子学与计算机》
2024年第2期108-114,共7页
Microelectronics & Computer
基金
国家重点研发计划(2018YFB0407603)。
关键词
自旋转移力矩磁随机存储器
灵敏放大器
信道模型
信道量化器
纠错编码
spin transfer torque magnetic random-access memory
sensitive amplifier
channel model
channel quantizer
error correction coding
