A 428.4mW 6.78Mbps/3.61Mbps Up/Down Data Rate Wireless Chipset with TinyAI Regulation
Yunfang Zhang, Zhijie Lin, Peiyi Zhou, Jie Zhang, Zhezhi He, Hong Zhang, Yongfu Li, Yong Lian, Yang Zhao · 2024
Wireless power and data transmission (WPDT) systems that possess high power capability and efficiency as well as fast uplink/downlink data transmission (UDT/DDT) rate via single inductive link are in high demand for compact implants, such as cochlear, retinal prostheses and pacemaker for their increasingly advanced functionality and thermal safety. Conventionally, the load of the implanted RX-end needs to be monitored to either intermittently short the excessive power at the RX input for local regulation ($R X-L R$) or report the condition to the TX-end via UDT to correct the TX power for global regulation (TX-GR) [1] as in Fig. 1. $R X-L R$ is simple and fast but deteriorates the overall power efficiency which might cause rapid temperature rise while TX-GR improves the efficiency at the cost of UDT bandwidth. This design leverages a trained tiny artificial neural network (TinyAI) to extract the RX load condition solely from $V_{L}$ at the TX-end to timely adjust the TX power without UDT occupation nor excessive power shorting. In addition, a cyclic on-off-keying (COOK) with auto time tracking (ATT) and an adaptive encoded binary phase-shift keying (A-BPSK) are respectively proposed for UDT and DDT over the same inductive link at 13.56 MHz carrier frequency. The resulting $T X / R X$ chipset achieves 0.56% target regulation error, 3.7 X higher transmitted power and 1.4X higher DDT speed than prior arts.