A Fully Integrated 1.13 NEF 32-Channel Neural Recording SoC With 12.5 pJ/Pulse IR-UWB Wireless Transmission for Brain Machine Interfaces
A Fully Integrated 1.13 NEF 32-Channel Neural Recording SoC With 12.5 pJ/Pulse IR-UWB Wireless Transmission for Brain Machine Interfaces
复制标题
DOI：
10.1109/access.2023.3321796
复制
发表时间：
2023
期刊：
影响因子：
3.9
通讯作者：
N. Tasneem;D. Biswas;Sakib Reza;Ifana Mahbub
中科院分区：
文献类型：
--
作者：
N. Tasneem;D. Biswas;Sakib Reza;Ifana Mahbub
在高通道计数神经记录系统的帮助下，监测大量神经元的电活动对于神经科学研究和临床治疗的最新方法至关重要。在神经信号采集系统中，需要优化通道数、总功耗、噪声、面积和鲁棒性等性能权衡。本研究提出了一种可植入的32通道神经信号采集片上系统（SoC），其中包括32个神经放大器，一个模拟多路复用器（MUX），一个模数转换器（ADC）和一个脉冲无线电超宽带（IR-UWB）发射机，采用180 nm CMOS工艺设计。神经放大器允许在可编程增益带宽设置中分别检测局部场电位（LFP）和动作电位（AP）。lfp在0.2-345 Hz范围内获得48 dB增益，ap在310 - 20.7 khz频率范围内获得59.7 dB增益。设计了一个10位SAR-ADC，可重构采样率为10 - 40 ksps，实现了629-fJ/步的FOM（品质图）。设计的IR-UWB发射机数据传输速率为100mbps，能量效率为12.5 pj /脉冲。该系统为神经科学研究和脑机接口（bmi）应用提供了一种有前途的可重构多通道神经信号记录范式。
Monitoring electrical activity from numerous neurons with the help of a high-channel count neural recording system is crucial for the state-of-the-art approaches in neuroscience research and clinical treatment. The performance trade-offs such as channel count, total power consumption, noise, area, and robustness need to be optimized for compatibility in neural signal acquisition system. This work proposes an implantable 32-channel neural signal acquisition system-on-chip (SoC), which includes 32 neural amplifiers, an analog multiplexer (MUX), an analog-to-digital converter (ADC), and an impulse radio ultra-wide band (IR-UWB) transmitter designed using 180 nm CMOS process. The neural amplifier allows to detect local field potentials (LFP) and action potentials (AP) separately in a programmable gain-bandwidth setup. It achieves 48 dB gain within the 0.2–345 Hz for LFPs and a gain of 59.7-dB within the 310–20.7-kHz frequency range for APs. A 10 bit SAR-ADC is designed with reconfigurable sampling rate of 10–40-kSps achieving an FOM (Figure-of-Merit) of 629-fJ/step. The designed IR-UWB transmitter can transmit data at 100-Mbps data rate with an energy efficiency of 12.5-pJ/pulse. The proposed system shows a promising reconfigurable multi-channel neural signal recording paradigm for neuroscience research and Brain Machine Interfaces (BMIs) applications.