The mm-wave frequency band has been a fascinating abundant alternative band compared to the already populated sub-6 GHz for mobile and fixed communication due to its still available wideband spectrum such as 28GHz, 39Ghz, 60GHz, and even higher bands. Therefore, the mm-wave is a critical frequency band for the 5th and 6th generations of mobile communication. However, we still need to overcome challenges to fully utilize the mm-wave bands. Over the decades, we have seen dramatic technological development, such as the beamforming RFIC (Radio Frequency Integrated Circuits). The recent RFIC commonly integrates 4 ~16 channels with a phase shifter, attenuator, and even memory component in a single bulk CMOS (Complementary Metal Oxide Semiconductor) or SOI (Silicon on Insulator) CMOS technology. Now we can design mm-wave beamforming frontend within a small single PCB (Printed Circuit Board), while it should be a large and heavy system before the advancement in the component. However, even though the integration of the multifunction into a single chip becomes dramatically improved, the individual performance of the function block still needs to catch up to that of sub-6 GHz components, especially for power amplifier performance.
Furthermore, mm-wave inherently has large free space path loss. In order to overcome those challenges in mm-wave frequency application, a large array of antennas and corresponding power amplifiers have been used for high antenna gain and EIRP (Equivalent Isotropic Radiated Power). Correspondingly, the beamwidth becomes narrow, and the beamforming is required to cover the service area in mobile communication.
A large array and beamforming in the mm-wave frequency require an accurate alignment of the phase and amplitude between the array elements. Therefore, the phase and amplitude of every array element need to be accurately calibrated in the factory and the field. The in-field calibration is essential because the active array elements, such as the power amplifier, low noise amplifier, and phase shifter, change their performance in the field due to the environment change, and device aging. Another critical but underdeveloped area is linearizing the large array in mm-wave. Digital predistortion technology is a matured area, but array DPD (Digital Predistortion) in mm-wave is still premature. The two different and essential areas for mm-wave array application have one important common component. The loopback path is essential to perform effective in-field calibration and array DPD. However, there are only limited applications and studies due to the complexity of the loopback path implementation on the large-scale mm-wave array. Therefore, the large 2D mm-wave array does not incorporate loopback and shows limitations on the in-field calibration and the array DPD operation.
This thesis proposes a novel 2D scalable loopback path structure in 4 by 4 LTCC (Low Temperature Co-Fired Ceramic) AiP (Antenna in Package) and verifies its in-field calibration and closed-loop DPD ability by designing, implementing, and measuring. The four-by-four arrayed AiP is 21.6 mm by 21.6 mm by 2.3mm. Aperture-coupled, cavity-backed, ground-posted dual pole patch antennas are placed on the top side of AiP, and four beamforming IC with peripheral capacitors are assembled on the bottom side. A commercial beamforming IC, summit2629 from Sivers, has four arrays of each H-pole and V-pole with phase shifters, attenuators, and even memory for the beam table. The unit coupler comprises four gap couplers combined with half-wavelength transmission lines. The inherently large impedance is matched to 50 Ohm with the half-wave transmission line combiner and deQ resistor so that the unit couplers can be easily duplicated and combined to form a 2D large array coupler. The proposed coupler has a symmetric structure and is scalable to a large array. The loopback path with this scalable coupler will enhance the in-field calibration capability and enable the real closed-loop DPD in the mm-wave array. Eventually, it can contribute to implementing the robust and green mm-wave array.
Four metal posts are proposed in the antenna feeding area to suppress resonance to the coupler. Three combiners are designed: a modified impedance microstrip combiner, modified strip combiner, and simple T-type strip combiner for each V-pole, H-pole, and coupler signal combining. Microstrip, stripline transmission, single patch antenna, and 28GHz cavity filter are designed and fabricated to validate the LTCC process. The sensitivity analysis is done by simulating the coupling coefficient with respect to the layer-to-layer misalignment. The return loss of every H-pole and V-pole antenna and coupler is measured with a probe station for the process validation before the component’s assembly. After the measurement, the AiP is assembled through its ball ports on the test PCB. The test PCB with AiP is installed with the aluminum heat sink. As for the measurement, CATR (Compact Antenna Test Range) chamber with VNA (Vector Network Analyzer) is used for AiP calibration and beam test. The bench top mm-wave anechoic chambers with the vector signal generator and signal analyzer are used for DPD verification.
The loopback port response shows narrow bandwidth, such as 500MHz. The cause and improvement methods are presented in the Appendix with simulation results. The beam patterns are measured in CATR with calibration in OTA (Over the air) set-up, and some of them were compared with those with calibration in loopback set-up. There is a negligible difference found in the comparison. The closed-loop DPD was verified with commercial equipment of vector signal generator and signal analyzer from Rohde & Schwarz. The direct DPD function provided by the equipment is used for the verification. By the closed-loop DPD with a subset of the 16 channels, the OTA test of AiP with the full channels shows about 3 % improvement of EVM (Error Vector Magnitude).