Nonlinear Modeling and Precision Analysis of Neural Recording System in Simulink

In recent years, multi-channel neural recording systems have become a popular research focus. The challenge in their design lies in balancing system performance, such as spike detection accuracy, with resource efficiency, including power consumption and area. To ensure accuracy while minimising design costs, it is crucial to investigate the specific impact of factors like nonlinearity on system accuracy. Therefore, we propose a nonlinear modeling method for neural recording systems using Simulink and a precision analysis method based on spike detection and sorting. Initially, we derive linear models for each system stage through transfer functions and introduce adjustable nonlinear gradients. Spike detection and sorting are then applied to the system output. We propose metrics to achieve this balance and analyze the relative impact of nonlinearities at each stage on the system. To achieve a $\pm 10 \%$ spike number error and $90 \%$ accuracy range, a total harmonic distortion as low as ${- 3 4. 3 2 d B},-33.73 \mathrm{~dB}$, and -44.42 dB is required for the low-noise amplifier (LNA), programmable gain amplifier (PGA), and analog-to-digital converter (ADC), respectively. This finding can guide efficient and performance-balanced hardware design for neural recording systems and is expected to be significantly less expensive to develop.