Christian Doppler Laboratory for Three-Dimensional Beamforming Fjolla Ademaj 15.11.216
Studying 3D channel models Channel models on system-level tools commonly 2-dimensional (2D) 3GPP Spatial Channel Model (SCM) A geometric stochastic model Only linear antenna arrays can be inspected with 2D channel models Three-Dimensional (3D) channel model enable investigations on Full-dimension MIMO 3D Beamforming Slide 2 / 19
Studying 3D channel models Channel models on system-level tools commonly 2-dimensional (2D) 3GPP Spatial Channel Model (SCM) A geometric stochastic model Only linear antenna arrays can be inspected with 2D channel models Three-Dimensional (3D) channel model enable investigations on Full-dimension MIMO 3D Beamforming Slide 2 / 19
Studying 3D channel models Channel models on system-level tools commonly 2-dimensional (2D) 3GPP Spatial Channel Model (SCM) A geometric stochastic model Only linear antenna arrays can be inspected with 2D channel models Three-Dimensional (3D) channel model enable investigations on Full-dimension MIMO 3D Beamforming Slide 2 / 19
Studying 3D channel models Channel models on system-level tools commonly 2-dimensional (2D) 3GPP Spatial Channel Model (SCM) A geometric stochastic model Only linear antenna arrays can be inspected with 2D channel models Three-Dimensional (3D) channel model enable investigations on Full-dimension MIMO 3D Beamforming Slide 2 / 19
Studying 3D channel models Channel models on system-level tools commonly 2-dimensional (2D) 3GPP Spatial Channel Model (SCM) A geometric stochastic model Only linear antenna arrays can be inspected with 2D channel models Three-Dimensional (3D) channel model enable investigations on Full-dimension MIMO 3D Beamforming Slide 2 / 19
Studying 3D channel models Channel models on system-level tools commonly 2-dimensional (2D) 3GPP Spatial Channel Model (SCM) A geometric stochastic model Only linear antenna arrays can be inspected with 2D channel models Three-Dimensional (3D) channel model enable investigations on Full-dimension MIMO 3D Beamforming Antenna array Slide 2 / 19
Contents Modeling the wireless channel in 3-dimensions Complexity- Simulation run times Massive MIMO: Spatial resolution of planar antenna arrays Conclusion and Outlook Slide 3 / 19
Overview Modeling the wireless channel in 3-dimensions Complexity- Simulation run times Massive MIMO: Spatial resolution of planar antenna arrays Conclusion and Outlook
Modeling the wireless channel in 3-dimensions The 3GPP 3D channel model in Vienna LTE-A system-level simulator Considers elevation and azimuth Incorporates planar antenna arrays z BS Elevation Paths z UE θtx,m y Cluster θrx,m ϕtx,m Azimuth ϕrx,m y x x Slide 4 / 19
Reduction in complexity The 3GPP 3D model imposes high complexity Complexity reduction by: Partition the scenario into equally sized cubes Within a cube UE experiences the same propagation conditions Spatial resolution 1 m Temporal resolution 1 ms UE speed v = [7.5,, ] km/h Change the cube each 36 sub-frames Temporal resolution refers to the length of one LTE sub-frame Slide 5 / 19
Reduction in complexity The 3GPP 3D model imposes high complexity Complexity reduction by: Partition the scenario into equally sized cubes Within a cube UE experiences the same propagation conditions Spatial resolution 1 m Temporal resolution 1 ms UE speed v = [7.5,, ] km/h Change the cube each 36 sub-frames Temporal resolution refers to the length of one LTE sub-frame Slide 5 / 19
Reduction in complexity The 3GPP 3D model imposes high complexity Complexity reduction by: Partition the scenario into equally sized cubes Within a cube UE experiences the same propagation conditions Spatial resolution 1 m Temporal resolution 1 ms UE speed v = [7.5,, ] km/h Change the cube each 36 sub-frames Temporal resolution refers to the length of one LTE sub-frame Slide 5 / 19
Reduction in complexity The 3GPP 3D model imposes high complexity Complexity reduction by: Partition the scenario into equally sized cubes Within a cube UE experiences the same propagation conditions Spatial resolution 1 m Temporal resolution 1 ms UE speed v = [7.5,, ] km/h Change the cube each 36 sub-frames Temporal resolution refers to the length of one LTE sub-frame Slide 5 / 19
Reduction in complexity The 3GPP 3D model imposes high complexity Complexity reduction by: Partition the scenario into equally sized cubes Within a cube UE experiences the same propagation conditions z Spatial resolution 1 m 1 m Temporal resolution 1 ms UE speed v = [7.5,, ] km/h Change the cube each 36 sub-frames 1 m UE v x y 1 m Temporal resolution refers to the length of one LTE sub-frame Slide 5 / 19
Reduction in complexity The 3GPP 3D model imposes high complexity Complexity reduction by: Partition the scenario into equally sized cubes Within a cube UE experiences the same propagation conditions z Spatial resolution 1 m 1 m Temporal resolution 1 ms UE speed v = [7.5,, ] km/h Change the cube each 36 sub-frames 1 m UE v x y 1 m Temporal resolution refers to the length of one LTE sub-frame Slide 5 / 19
Reduction in complexity The 3GPP 3D model imposes high complexity Complexity reduction by: Partition the scenario into equally sized cubes Within a cube UE experiences the same propagation conditions z Spatial resolution 1 m 1 m Temporal resolution 1 ms UE speed v = [7.5,, ] km/h Change the cube each 36 sub-frames 1 m UE v x y 1 m Temporal resolution refers to the length of one LTE sub-frame Slide 5 / 19
Calibration: Zenith spread Comparing system-level simulations and 3GPP TR 36.873 results 1 1.9.9.8.7.6 3GPP - data from 21 sources 3GPP - median from 21 sources.8.7.6 Vienna LTE-A System-Level simulator ECDF.5.4 Vienna LTE-A System-Level simulator ECDF.5.4 3GPP - median from 21 sources.3.3 3GPP - data from 21 sources.2.2.1.1 1 2 3 4 5 6 Zenith spread of departure [ ] 1 2 3 4 5 6 Zenith spread of arrival [ ] Slide 6 / 19
Overview Modeling the wireless channel in 3-dimensions Complexity- Simulation run times Massive MIMO: Spatial resolution of planar antenna arrays Conclusion and Outlook
Simulation run times A network of seven hexagonally arranged macro-sites, each employing three enodeb sectors Evaluate the simulation complexity [Ademaj et al., a]: Number of interfering links N sector = {2, 8, 14, 2} Number of UEs per sector K = {2, 2, 5} Number of antenna elements M = {8, 24, 4, 8} Simulation length N TTI = {1, 5, 1} Slide 7 / 19
Simulation run times A network of seven hexagonally arranged macro-sites, each employing three enodeb sectors Evaluate the simulation complexity [Ademaj et al., a]: Number of interfering links N sector = {2, 8, 14, 2} Number of UEs per sector K = {2, 2, 5} Number of antenna elements M = {8, 24, 4, 8} Simulation length N TTI = {1, 5, 1} Slide 7 / 19
Simulation run times A network of seven hexagonally arranged macro-sites, each employing three enodeb sectors Evaluate the simulation complexity [Ademaj et al., a]: Number of interfering links N sector = {2, 8, 14, 2} Number of UEs per sector K = {2, 2, 5} Number of antenna elements M = {8, 24, 4, 8} Simulation length N TTI = {1, 5, 1} Slide 7 / 19
Simulation run times A network of seven hexagonally arranged macro-sites, each employing three enodeb sectors Evaluate the simulation complexity [Ademaj et al., a]: Number of interfering links N sector = {2, 8, 14, 2} Number of UEs per sector K = {2, 2, 5} Number of antenna elements M = {8, 24, 4, 8} Simulation length N TTI = {1, 5, 1} Slide 7 / 19
Simulation run times A network of seven hexagonally arranged macro-sites, each employing three enodeb sectors Evaluate the simulation complexity [Ademaj et al., a]: Number of interfering links N sector = {2, 8, 14, 2} Number of UEs per sector K = {2, 2, 5} Number of antenna elements M = {8, 24, 4, 8} Simulation length N TTI = {1, 5, 1} Co-pol λ/2 Q={2, 6, 1, 2} ω Q (θ tilt )......... λ/2 ω 1 (θ tilt ) p p 2 p 1 p 3 Slide 7 / 19
Simulation run times A network of seven hexagonally arranged macro-sites, each employing three enodeb sectors Evaluate the simulation complexity [Ademaj et al., a]: Number of interfering links N sector = {2, 8, 14, 2} Number of UEs per sector K = {2, 2, 5} Number of antenna elements M = {8, 24, 4, 8} Simulation length N TTI = {1, 5, 1} Co-pol λ/2 Q={2, 6, 1, 2} ω Q (θ tilt )......... λ/2 ω 1 (θ tilt ) p p 2 p 1 p 3 Slide 7 / 19
Simulation run times: Varying the size of antenna array K N sector N TTI M Number of UEs Number of interferers Simulation length in TTIs Number of antenna elements in array M = 8 M = 24 M = 4 M = 8 1 5 Simulation runtime [s] 6 5 4 3 2 1 1 4 Simulation runtime [s] 2 1.5 1.5 1 2 Number K of UEs 3 4 5 25 5 75 Simulation length N TTI 1 1 2 Number K of UEs 3 4 5 5 1 15 Interfering enodebs N sector 2 Slide 8 / 19
Simulation run times: Varying the size of antenna array K N sector N TTI M Number of UEs Number of interferers Simulation length in TTIs Number of antenna elements in array M = 8 M = 24 M = 4 M = 8 1 5 Simulation runtime [s] 6 5 4 3 2 1 1 4 Simulation runtime [s] 2 1.5 1.5 1 2 Number K of UEs 3 4 5 25 5 75 Simulation length N TTI 1 1 2 Number K of UEs 3 4 5 5 1 15 Interfering enodebs N sector 2 Slide 8 / 19
Simulation run times: Varying the size of antenna array K N sector N TTI M Number of UEs Number of interferers Simulation length in TTIs Number of antenna elements in array M = 8 M = 24 M = 4 M = 8 1 5 Simulation runtime [s] 6 5 4 3 2 1 1 4 Simulation runtime [s] 2 1.5 1.5 1 2 Number K of UEs 3 4 5 25 5 75 Simulation length N TTI 1 1 2 Number K of UEs 3 4 5 5 1 15 Interfering enodebs N sector 2 Slide 8 / 19
Simulation run times: Varying the size of antenna array K N sector N TTI M Number of UEs Number of interferers Simulation length in TTIs Number of antenna elements in array M = 8 M = 24 M = 4 M = 8 1 5 Simulation runtime [s] 6 5 4 3 2 1 1 4 Simulation runtime [s] 2 1.5 1.5 1 2 Number K of UEs 3 4 5 25 5 75 Simulation length N TTI 1 1 2 Number K of UEs 3 4 5 5 1 15 Interfering enodebs N sector 2 Slide 8 / 19
Simulation run times: Varying the size of antenna array K N sector N TTI M Number of UEs Number of interferers Simulation length in TTIs Number of antenna elements in array M = 8 M = 24 M = 4 M = 8 1 5 Simulation runtime [s] 6 5 4 3 2 1 1 4 Simulation runtime [s] 2 1.5 1.5 1 2 Number K of UEs 3 4 5 25 5 75 Simulation length N TTI 1 1 2 Number K of UEs 3 4 5 5 1 15 Interfering enodebs N sector 2 Slide 8 / 19
Simulation run times: Varying the size of antenna array K N sector N TTI M Number of UEs Number of interferers Simulation length in TTIs Number of antenna elements in array M = 8 M = 24 M = 4 M = 8 1 5 Simulation runtime [s] 6 5 4 3 2 1 1 4 Simulation runtime [s] 2 1.5 1.5 1 2 Number K of UEs 3 4 5 25 5 75 Simulation length N TTI 1 1 2 Number K of UEs 3 4 5 5 1 15 Interfering enodebs N sector 2 Slide 8 / 19
Simulation run times: Varying the size of antenna array K N sector N TTI M Number of UEs Number of interferers Simulation length in TTIs Number of antenna elements in array M = 8 M = 24 M = 4 M = 8 1 5 Simulation runtime [s] 6 5 4 3 2 1 1 4 Simulation runtime [s] 2 1.5 1.5 1 2 Number K of UEs 3 4 5 25 5 75 Simulation length N TTI 1 1 2 Number K of UEs 3 4 5 5 1 15 Interfering enodebs N sector 2 Slide 8 / 19
Simulation run times: Varying the size of antenna array K N sector N TTI M Number of UEs Number of interferers Simulation length in TTIs Number of antenna elements in array M = 8 M = 24 M = 4 M = 8 1 5 Simulation runtime [s] 6 5 4 3 2 1 1 4 Simulation runtime [s] 2 1.5 1.5 1 2 Number K of UEs 3 4 5 25 5 75 Simulation length N TTI 1 1 2 Number K of UEs 3 4 5 5 1 15 Interfering enodebs N sector 2 Slide 8 / 19
Throughput performance: Rayleigh versus 3GPP 3D model Desired channel modeled by 3GPP 3D model Interfering channels modeled as A noise-limited network Rayleigh fading The 3GPP 3D model Slide 9 / 19
Throughput performance: Rayleigh versus 3GPP 3D model Desired channel modeled by 3GPP 3D model Interfering channels modeled as A noise-limited network Rayleigh fading The 3GPP 3D model Slide 9 / 19
Throughput performance: Rayleigh versus 3GPP 3D model Desired channel modeled by 3GPP 3D model Interfering channels modeled as A noise-limited network Rayleigh fading The 3GPP 3D model 1.9.8.7 ECDF.6.5.4 Noise only.3.2.1.5 1 1.5 2 2.5 3 3.5 Average UE throughput [Mbit/s] Slide 9 / 19
Throughput performance: Rayleigh versus 3GPP 3D model Desired channel modeled by 3GPP 3D model Interfering channels modeled as A noise-limited network Rayleigh fading The 3GPP 3D model 1.9.8 Rayleigh fading.7 ECDF.6.5.4 Noise only.3.2.1.5 1 1.5 2 2.5 3 3.5 Average UE throughput [Mbit/s] Slide 9 / 19
Throughput performance: Rayleigh versus 3GPP 3D model Desired channel modeled by 3GPP 3D model Interfering channels modeled as A noise-limited network Rayleigh fading The 3GPP 3D model 1 ECDF.9.8.7.6.5.4 Rayleigh fading 3GPP 3D model Noise only.3.2.1.5 1 1.5 2 2.5 3 3.5 Average UE throughput [Mbit/s] Slide 9 / 19
Overview Modeling the wireless channel in 3-dimensions Complexity- Simulation run times Massive MIMO: Spatial resolution of planar antenna arrays Conclusion and Outlook
Spatial resolution of planar antenna arrays How does the channel impacts the spatial resolution of an antenna array? What is the spatial separation of narrow beams in LOS and NLOS, outdoors and indoors [Ademaj et al., b]? 9 θ s h UE h BS d =25 m UE d UE=15 m d UE=5 m Slide 1 / 19
Resolution of a uniform vertical linear array Antenna array radiation pattern in elevation w M(q s ) q =9 s w (q ) 3 s M w (q ) 2 s w (q ) 1 s Slide 11 / 19
Resolution of a uniform vertical linear array Antenna array radiation pattern in elevation w M(q s ) q =9 s 12 25 2 3 15 1 6 w (q ) 3 s M M=1 M=1 M=1 5 9 w (q ) 2 s 12 w (q ) 1 s 24 18 15 Slide 11 / 19
Resolution of a uniform vertical linear array Antenna array radiation pattern in elevation w M(q s ) q =9 s 12 25 2 3 5 15 1 6 w (q ) 3 s M M=1 M=1 M=1 5 9 w (q ) 2 s 12 w (q ) 1 s 24 18 15 Slide 11 / 19
2D Antenna array structure Antenna port configuration: N Tx N Rx = 4 2 Antenna array geometry at enodeb: N Tx M enodeb UE M={1, 1, 1} ω M (θ s ) ω 2 (θ s )............ λ/2 ω 1 (θ s ) p={1, 2,..., Ν Tx } r={1,..., Ν Rx } Slide 12 / 19
Indoor UE: Channel energy and UE throughput UE height h UE = 1.5 m UE distance d UE = 15 m M = {1, 1, 1} Steering angle θ s = {, 1, 2,, 18 } Target angle θ s = 98.9 Slide 13 / 19
Indoor UE: Channel energy and UE throughput UE height h UE = 1.5 m UE distance d UE = 15 m M = {1, 1, 1} Steering angle θ s = {, 1, 2,, 18 } Target angle θ s = 98.9 Channel energy [db] -8-9 -1-11 -12-13 -14-15 2 4 6 8 Steering angle [ ] 1 12 14 16 18 Average UE throughput [Mbit/s] 8 7 6 5 4 3 2 1 2 4 6 8 1 12 14 16 18 Steering angle [ ] Slide 13 / 19
Indoor UE: Channel energy and UE throughput UE height h UE = 1.5 m UE distance d UE = 15 m M = {1, 1, 1} Steering angle θ s = {, 1, 2,, 18 } Target angle θ s = 98.9 LOS NLOS M=1 M=1 M=1 Channel energy [db] -8-9 -1-11 -12-13 -14-15 2 4 6 8 Steering angle [ ] 1 12 14 16 18 Average UE throughput [Mbit/s] 8 7 6 5 4 3 2 1 2 4 6 8 1 12 14 16 18 Steering angle [ ] Slide 13 / 19
Indoor UE: Channel energy and UE throughput UE height h UE = 1.5 m UE distance d UE = 15 m M = {1, 1, 1} Steering angle θ s = {, 1, 2,, 18 } Target angle θ s = 98.9 LOS NLOS M=1 M=1 M=1 Channel energy [db] -8-9 -1-11 -12-13 -14-15 2 4 6 8 Steering angle [ ] 1 12 14 16 18 Average UE throughput [Mbit/s] 8 7 6 5 4 3 2 1 2 4 6 8 1 12 14 16 18 Steering angle [ ] Slide 13 / 19
Overview Modeling the wireless channel in 3-dimensions Complexity- Simulation run times Massive MIMO: Spatial resolution of planar antenna arrays Conclusion and Outlook
Conclusion Incorporation of the elevation dimension increases the computational complexity by more than three times The complexity grows roughly linearly with the number of antenna elements per antenna array A more optimistic view on performance observed by 3D model: simple channel models may underestimate the achievable performance In terms of spatial resolution, doubling the number of antenna elements in elevation does not double the received channel energy The optimal number of antenna elements -? Slide 14 / 19
Outlook: 5G System Level The 3GPP 3D channel model Reduce the model complexity by pruning the multipath components Adapt the model for moving scenarios: spatial correlation and channel transitions BS z y Clusters x Clusters z UE y x Slide 15 / 19
Outlook: 5G System Level 3D Channel models > 6 GHz 3GPP TR 38.9: Channel models for carrier frequencies 6 GHz 1 GHz Covering many scenarios (rural, urban, street canyons, open area, indoor scenarios, D2D and V2V) Three-dimensional beamforming Slide 16 / 19
Outlook: Full-dimension MIMO Antenna configurations for 2-dimensional antenna arrays New transmitter architectures: TXRU modeling and two-layer mapping New precoding strategies to support the element-level antenna structure Slide 17 / 19
Outlook: Full-dimension MIMO Antenna configurations for 2-dimensional antenna arrays New transmitter architectures: TXRU modeling and two-layer mapping New precoding strategies to support the element-level antenna structure Slide 17 / 19
Outlook: Full-dimension MIMO Antenna configurations for 2-dimensional antenna arrays New transmitter architectures: TXRU modeling and two-layer mapping New precoding strategies to support the element-level antenna structure Slide 17 / 19
Outlook: Full-dimension MIMO Antenna configurations for 2-dimensional antenna arrays New transmitter architectures: TXRU modeling and two-layer mapping New precoding strategies to support the element-level antenna structure TXRU Transceiver Unit Slide 17 / 19
Outlook: Full-dimension MIMO in moving scenarios Beam alignment schemes for efficient UE tracking Cars and high speed trains http://www.profheath.org/research/millimeter-wave-cellular-systems/mmwavefor-vehicular-communications-2/ Slide 18 / 19
Christian Doppler Laboratory for Three-Dimensional Beamforming Fjolla Ademaj 15.11.216
References I Ademaj, F., Taranetz, M., and Rupp, M. 3GPP 3D MIMO Channel Model: A Holistic Implementation Guideline for Open Source SImulation Tools. EURASIP Journal on Wireless Communication Networks 216. doi:1.1186/s13638-16-549-9. Ademaj, F., Taranetz, M., and Rupp, M. Evaluating the spatial resolution of 2d antenna arrays for massive mimo transmissions. The 24th European Signal Processing Conference, EUSIPCO 216. Slide 1 / 1