A 24–30 GHz 4-Stream CMOS Transceiver Based on Dual-LO Phase-Shifting Fully Connected Architecture

This paper presents a wideband low-complexity multi-stream beamforming transceiver (TRX) based on multiple local-oscillator (multi-LO) phase-shifting fully-connected (FC) architecture for MIMO communication. In supporting the same beams, the proposed architecture halves the complexity of the FC combiners as well as the ADCs/DACs overhead. Meanwhile, the FC combining network of each stream is implemented at IF-domain, which benefits from avoiding the mutual overlap between the combiners in the RF-domain, but also reducing combiner loss. To verify the proposed architecture, a 24–30 GHz 4-stream dual-LO phase-shifting FC TRX chip is fabricated in 65-nm CMOS process. Each single-stream path integrates a TX and RX channel with independent up/down-mixer, 0.4-dB gain and 6-bit phase control, where the whole TRX IC only has one PA and LNA to amplify 4-stream incoherent signals saving chip area and power consumption. The measured peak conversion gain (CG) of the TX is 31.4 dB, with a 3-dB bandwidth of 24.4-30.1 GHz. In the RX mode, the measured CG versus IF is 18.3 dB, with a 3-dB bandwidth of 3.3-11.8 GHz (112.6% fractional bandwidth). Besides, the TRX IC achieves a maximum OP1dBof 15.4 dBm and minimum NF of 4.6 dB including T/R switch. Additionally, it demonstrates a 26 dB gain control and 360otuning range, where the maximum gain variation of only ±0.1 dB during 360ophase tuning at 27.5 GHz without any calibration.

I. Introduction

Phased-array beamformers with multi-stream capability at millimeter-wave (mm-wave) frequencies have emerged as the preferred to further boost the data rates and channel capacity. Compared with the multi-stream partially-connected (PC) beamforming architectures, the fully-connected (FC) ones have better spectral efficiency, superior spatial filtering and higher energy efficiency due to each stream FC to all available antenna elements [1]–[3]. However, FC architectures increase hardware design complexity significantly in massive arrays, which is difficult to scale to more streams.