Auxiliary Beam Pair Enabled AoD Estimation for Large-scale mmwave MIMO Systems Dalin Zhu, Junil Choi and Robert W. Heath Jr. Wireless Networking and Communications Group Department of Electrical and Computer Engineering The University of Texas at Austin This work is funded by a research gift from Huawei Technologies, Inc. www.profheath.org
Hybrid precoding in mmwave MIMO DAC 1-bit ADC RF Chain RF Chain 1-bit ADC ADC Baseband Precoding RF Precoding RF Combining Baseband Combining F BB F RF channel RF W RF W BB 1-bit DAC ADC RF Chain Chain 1-bit ADC ADC Effective channel at baseband A combination of digital and analog beamforming/combining Most prior work assumes a channel estimate is available for configuration 2
Prior work on channel estimation in mmwave MIMO u Compressed sensing based approaches * ª Assumes sparsity in the angular domain of the channel ª Requires a prior knowledge of the number of the propagation paths u Subspace based methods ** ª Echo-based subspace estimation between the transmitter and receiver ª Requires a lot of training between the transmitter and receiver u Exhaustive search based approaches *** ª Searches for the best pairs of analog transmit and receive steering vectors ª Training overhead is high Feasible channel estimation method is required Low training and feedback overhead High-resolution estimate of channel information Contributions: Path-based channel estimator based on auxilary beam pairs (ABPs) * A. Alkhateeb, O. E. Ayach, G. Leus, and R. W. Heath Jr., ``Channel estimation and hybrid precoding for millimeter wave cellular systems,'' in IEEE J. Sel. Top. Signal Process, vol. 8, pp. 831-846, Oct. 2014 ** H. Ghauch, T. Kim, M. Bengtsson, and M. Skoglund, `` Subspace estimation and decomposition for large millimeter-wave MIMO systems,'' submitted to IEEE J. Sel. Top. Signal Process., arxiv preprint arxiv:1507.00287, Jul. 2015 *** C. Kim, T. Kim, and J.-Y. Seol, `` Muti-beam transmission diversity with hybrid beamforming for MIMO-OFDM systems,'' in IEEE Global Telecomm. Conf. (GLOBECOM 13), pp. 61-65, Dec. 2013. 3
Exhuastive search via grid of beams True AoD Each beam is probed Quantization error Quantized AoD Receiver determines which beams have the strongest signals Index of preferred beams are sent to the transmitter Transmitter Receiver The resolution for AoD angle is limited by grid size 4
Angle estimation in monopulse radar systems Projections of on and Echo signal Custom designed analog receive beams Beams distinguished by sending with different polarization, code, or time Receive beam pair Angle estimation Amplitude comparison Signal detection S. M. Sherman, ``Monopulse principles and techniques,'' Artech House, 1984 5
Basic design principle of auxiliary beam pairs (1/2) Pairs of analog transmit beams are formed to cover a given angular range Exact AoD can be extracted from the ABP that covers the AoD...... System setup : boresight angle of n-th ABP : steering angle of : steering angle of scatter : dominant path s AoD : n-th ABP Transmit array response vector: Auxiliary beam pair that covers the AoD: 6
Basic design principle of auxiliary beam pairs (2/2)...... Quantized version of Amplitude comparison Calculate received signal strengths Angle estimation If is invertible with respect to 7
Quantization options Distribution of the ratio metric Distribution of the estimated AoD More codewords can be allocated in densely distributed portions Uniformly distributed codewords Quantizing the ratio metric provides more quantization resolution 8
Procedure for estimating the AoD using ABP Auxiliary beam pairs are probed A set of ratio metrics are calculated, each corresponds to a ABP The ratio metric with the highest received signal strength is quantized Transmitter Receiver The AoD is estimated using the quantized ratio metric 9
Deployment scenarios using ABP-based method (1/3) Beamforming range for the next layer is optimized using the estimated AoD Beam finding for control channels UE UE UE UE UE UE Base station Base station Base station Layer-1: system-specific information Layer-2: cell-specific information Layer-3: user-specific information Beam finding process for control channels is facilitated via ABP design 10
Deployment scenarios using ABP-based method (2/3) An example of control channels beamforming using ABP Beam finding for control channels Layer-I beam-b The UE decodes system-specific control information from beam-a 1 UE Base station beam-a UE The UE calculates a ratio metric with respect to the ABP formed by beam-a and beam-b Base station The UE quantizes the ratio metric and sends it back to the base station 2 3 4 beam-b beam-a UE UE UE Base station Using the quantized ratio metric, the base station estimates the AoD Base station The probing range for layer-ii beamforming is determined around the estimated AoD Base station Layer-II 11
Deployment scenarios using ABP-based method (3/3) Hybrid precoding for data channels PA... Baseband Precoder DAC... DAC Mixers Phase shifters Antenna array... Antenna array Multi-path s AoDs/AoAs can be estimated via ABP Phase shifters... Mixers ADC... ADC Baseband Combiner... Analog and digital precoder and combiner are optimized using the high resolution estimates of AoDs/AoAs via ABP design 12
Numerical results (1/2) MSE of AoD estimation ª ª Channel assumption AoDs/AoAs take continuous values AoDs/AoAs are uniformly distributed Parameters : total # of transmit antennas : total # of receive antennas : total # of transmit RF chains : total # of receive RF chains : total # of data streams : total # of channel paths (6) Performance metric: MSE of AoD estimation In radians With increase in the number of antennas, more and narrower auxiliary beams are probed, which improves the estimation performance Promising MSE performance of AoD estimation can be obtained 13
Numerical results (2/2) Spectral efficiency performance ª ª Channel assumption AoDs/AoAs take continuous values AoDs/AoAs are uniformly distributed Parameters : total # of transmit antennas : total # of receive antennas : total # of transmit RF chains : total # of receive RF chains : total # of data streams : total # of channel paths (6) Multi-path s AoDs/AoAs estimation performance via ABP design is promising Assume, the performance gap is marginal 14
Conclusions u High-resolution AoD/AoA estimation method via ABP is possible ª Suited for mmwave MIMO with directional beamforming ª Quantizing the ratio metric results in better performance ª Applicable to beamformed control channels and data channels design in practice u Future work ª ABP-based AoD/AoA estimation assuming arbitrary antenna array geometry ª Exploiting TDD channel s reciprocity to enable mmwave MIMO precoding ª Simultaneously employing multiple RF chains to facilitate the estimation 15
Questions? 16
Backup Slides 17
System model for hybrid precoding baseband combiner, baseband precoder, transmit signal vector, Complex Gaussian noise vector analog precoder: similarly defined to analog combiner Steering angle Analog precoder and combiners are simple spatial matched filters 18
Spatial channel model Total number of propagation paths channel matrix Path-r s gain AoD (spatial frequency) to be estimated Assuming ULA AoA (spatial frequency) to be estimated 19
Detailed derivations of ABP design principle u Received signal with respect to each beam in the ABP (noiseless) u Received signal strength with respect to each beam in the ABP u A ratio metric via the difference and sum of and If is invertible with respect to 20
Performance evaluation of quantization options MSE of quantizing the estimated AoD: MSE of quantizing the ratio metric: In radians The performance gap between two quantization options is marginal The receiver requires knowledge of ABP parameters from the transmitter to estimate and quantize the AoD 21
Estimation of multiple angles using multiple RF chains (1/2) An example of multiple AoDs estimation Beam finding for control channels u Multiple analog beams can be simultaneously probed by the TX and RX Total number of transmit probings Total number of receive probings u Consider noiseless reception and a given receive probing, u Assume that path-r s AoD, and Desired transmit ABP that covers path-r s AoD Their positions in 22
Estimation of multiple angles using multiple RF chains (2/2) An example of multiple AoDs estimation Beam finding for control channels u The ratio metric that characterizes path-r s AoD is calculated as u The quantized version of the ratio metric is fed back to the transmitter u Path-r s AoD is estimated by the transmitter using the ratio metric u The above process iterates until all paths AoDshave been estimated 23