Publication
Over the past years, implantable closed-loop neural interfaces have received increasing attention for their potential to enhance the understanding of the brain, treat neurological disorders, and provide unprecedented opportunities for individuals with paralysis or limb loss. The untethered, autonomous neural interfaces require high-density recording, artificial intelligent-driven brain-state classification, therapeutic stimulation, and communication capabilities with external hubs for closed-loop operation. In addition, they rely on batteries or wireless power transfer for their energy needs. However, integrating these intricate functions into a single system presents significant challenges, primarily due to the stringent power and size constraints of the neural implants. Some of these design challenges include: 1) high voltages required by most neurostimulation applications exceed the voltage limits of advanced, scaled CMOS technologies, which is most suitable for the design of recording circuits, digital signal processing, and machine learning hardware; 2) increasing demand for high-density neural interfaces requires a large data bandwidth; 3) a small footprint is necessary to minimize invasiveness and enable chronic implantation; 4) embedding duplex uplink and downlink functionality is essential in wireless neural devices. This dissertation focuses on the optimization of two critical blocks of the implantable closed-loop neural interface, i.e., the neurostimulator and uplink transmitter (TX), paving the way for the next-generation high-performance neural interface systems-on-chip (SoCs).\ In the first part of the thesis, to address the voltage compliance challenges at the electrode-tissue interface, two high-voltage-compliant area-efficient stimulators are proposed in a standard 65-nm CMOS process using transistor stacking. These two neurostimulator ICs are integrated within two intelligent closed-loop neural interface SoCs and verified in close-loop electrical and in-vivo experiments.\ In the second part of the thesis, two miniaturized impulse-radio ultra-wideband TXs for implantable high-density and wireless neural recording systems are proposed to address the trade-offs between data rate, transmission range, wireless module size, and power consumption. Through a co-designed antenna and PA interface, the first TX design achieves a meter-level transmission range and up to 500-Mb/s data rate at different implantation depths within a small, 49.8-mm\textsuperscript{2} footprint and under strict safety limits. To support high-density recording with versatile channel counts and multiple modalities of neural signals, the second TX design is further optimized to reduce power consumption across various data rates (50 Mb/s--1.2 Gb/s). Furthermore, a novel carrier and modulation generation scheme is proposed. Additionally, a combined TDMA and FDMA approach is incorporated to enable an extended network of untethered, distributed, free-floating b