Download Area and Power Reduction Techniques for Millimeter-wave Phased-array Transceiver Front-ends Book in PDF, Epub and Kindle
The demands for higher data rates continue to drive research for consumer products to develop new techniques for low-cost and long battery-life wireless radios. This dissertation explores and implements several single-chip CMOS integrated circuits which aim to enable high data rates for the new spectrum opportunity in future mm-Wave 5G radios as well as to improve the spectrum efficiency for current sub-6GHz communication systems.In the mm-Wave design, a 50-58GHz 2x2 phased-array receiver (RX) for 5G communication and radar systems is described. The RX utilizes a Gm-assisted matching network (MN) to reduce the noise figure (NF) of a conventional passive mixer-first RX by placing an extra gain of ~5dB prior to the down-conversion mixers with minimal additional components. In addition to the Gm-assisted MN, this prototype phased-array RX integrates a proposed Gm-assisted MN, mixer-first RX with translational feedback, Cartesian phase shifters as a baseband beamformer, poly-phase filters, and LO generation in TSMC 28nm CMOS, occupying a total area of 0.53mm2. This RX achieves a NF of 7dB, gain of 26dB, input P1-dB of -20dBm with an 8-GHz 3-dB BW and a S11 lower than -10dB over 22-GHz BW. Second, a novel highly-digital 2x2 phased-array receiver implemented in TSMC 28-nm CMOS with several techniques, including a feedforward noise-suppressing front-end, reconfigurable 0-3 SMASH CT-delta-sigma ADC, and digital beamformer. This proposed receiver enables the development of digital multi-beam phased-array with improved front-end NF, reduced silicon area, and low power consumption for future wireless systems. Third, a dual-mode V-band power amplifier (PA) that utilizes a reconfigurable 2/4-way power combiner is introduced to enable two discrete modes of operation and enhance the efficiency at power back-off. The power combiner employs two techniques to further improve the PA efficiency at power back-off: 1) usage of transformers with non-uniform turns ratios to reduce the difference in impedance presented to the PA cores between the two modes and 2) utilize a proposed switching scheme to eliminate the leakage inductance associated with the disabled path in back-off power mode. The 2-stage PA achieves a peak gain of 21.4dB with a fractional BW (fBW) of 22.6% (51-64GHz). At 65GHz, the PA has a Psat of +17.9dBm with an OP1dB of +13.5dBm and a peak PAE of 26.5% in full-power mode. In back-off power mode, the measured Psat, OP1dB, and peak PAE are +13.8dBm, +9.6dBm, and 18.4%, respectively. The PAE is enhanced by 6% points at a 4.5-dB back-off. The PA is capable of amplifying a 6-Gb/s 16-QAM modulated signal with an EVMrms of -20.7dB at an average Pout/PAE of +13dBm/13.6%, respectively. This PA was implemented in 16-nm FinFET, occupies a core area of 0.107mm2, and operates under a 0.95-V supply. Lastly, to improve spectrum usage of the current sub-6GHz systems, a triple-path transmitter (TX) self-interference cancellation (SIC) transceiver for full-duplex (FD) systems which achieves greater than 73dB of SIC is reported. An integrated electrical balance duplexer (EBD) is employed and works in concert with two pre- and post-LNA RF feedforward cancelers implemented as analog finite impulse response (FIR) filters to form a combined three-path cancellation architecture that achieves a deep on-chip TX SIC over a very wide bandwidth (BW). The on-chip self-interference cancellation of 72.8/70.1/65.2dB was measured by comparing integrated channel power difference at the receiver (RX) baseband (BB) output while applying orthogonal frequency-division multiplexing (OFDM) multi-carrier 64-QAM Wi-Fi packages at the transmitter input with a 20/40/80-MHz bandwidth, respectively. The receiver noise figure (NF) degradation due to the TX SI leakage improves from 8dB to 1.6dB after turning on both RF cancelers. TX SI leakage reciprocal mixing with the RX LO phase noise (PN) was also significantly reduced by 11dB up to 1-MHz offset frequency measured at RX BB output. This work integrates the whole transceiver signal path from the analog baseband to RF. The FD prototype chip is fabricated in TSMC 40nm CMOS process with a die size of 4mm2 that consumes 106mW (excluding PA). An Altera Cyclone III FPGA with ADC/DAC daughter board is used to emulate the digital baseband (BB) and executes a nearest neighbor search (NNS) algorithm to close the filter adaptation loop. The RX operates from 1.6GHz to 1.9GHz with a total noise figure of 8.09dB and an IIP3 of -17dBm while operating at the highest gain of 42dB. The TX has an output P-1dB/Psat of 10.6dBm/12.5dBm measured at the antenna port, respectively. This chip also integrates an integer-N synthesizer with a measured locking range from 3.52GHz to 4.28GHz. The synthesizer consumes 10.4mW with a phase noise performance of -117dBc/Hz measured at 1-MHz frequency offset. These four prototypes demonstrate various techniques to reduce power and area for future wireless systems, including the applications in mm-Wave front-ends for new spectrum opportunities with increased data rates as well as improved spectral efficient full-duplex radios for existing sub-6GHz bands.