An X-Band Multifrequency Difference Frequency-Shift Keying CMOS Radar for Range Tracking and AI-Based Human Gesture Recognition

This article presents an
${X}$
-band single-chip fully integrated multifrequency difference frequency-shift keying (FSK) in-phase and quadrature (I/Q) direct-conversion radar transceiver implemented in a 130-nm CMOS process. Combined with the artificial intelligence (AI)-based algorithm, accurate gesture recognition is realized. The phase-locked loop (PLL)-based dual-frequency difference FSK waveform and ranging algorithm are proposed to solve the tradeoff between the maximum unambiguous range and the ranging accuracy. The loop bandwidth optimized integer-
${N}$
PLL is integrated to generate standard-compliant FSK signals with predictable frequency differences. The relationship between instantaneous ranging bias and PLL loop parameters is analyzed in detail. The proposed bias noise filter fully suppresses the flicker noise and improves the receiver noise figure (NF). The quadrature signal generator (QSG) adopts a fully differential amplifier buffered resistance compensation quadrature all-pass filter (QAF) to improve the I/Q balance and insertion loss. The chip consumes a current of 56 mA and provides a conversion gain (CG) of 22–102 dB. The single-ended transmit power is 1.5 dBm. To the best of our knowledge, the implemented transceiver is the first recorded
${X}$
-band FSK radar transceiver and provides a low-cost solution for indoor human detection.