Indoor mmWave Wearable Networks: mmWave for 5G and ...users.ece.utexas.edu/~rheath/presentations/2015/Indoormm...[1] T. Bai and R. W. Heath Jr., “Coverage and rate analysis for millimeter
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
*Vutha Va, and Robert W. Heath, Jr, "Basic Relationship between Channel Coherence Time and Beamwidth in Vehicular Channels,'' Proc. of the IEEE Vehicular Technology Conference (VTC 2015-Fall), 2015.
Beams should be narrow but not too “pointy”
• Derived coherence time with multipath, pointing error, and directional antennas • Optimum beamwidth is a tradeoff between pointing error and Doppler
• Modeled blockages using random shape theory • LOS probability is an exponential function of the link length R and β • β depends on the density of buildings and their average perimeter
[1] T. Bai and R. W. Heath Jr., “Coverage and rate analysis for millimeter wave cellular networks", IEEE Trans. Wireless Commun, Feb. 2015. [2] T. Bai, A. Alkhateeb, and R. W. Heath, Jr., ``Coverage and capacity of millimeter wave cellular networks", IEEE Commun. Mag., Sep. 2014. [3] T. Bai, R. Vaze, and R. W. Heath, Jr., ``Analysis of blockage effects on urban cellular networks", IEEE Trans. Wireless Commun., Sep. 2014.
Building
NLOS link
LOS link
Apply different path loss to LOS and NLOS links
SINR sensitive to BS density with blockage effects
Dense mmWave achieves better coverage than lower frequency
SIR invariant with BS density assuming no blockages
Cone-blocking model for self-body blocking SINR with different penetration losses
Body blockage increases low SINR outage
Body blocked BS
[1] T. Bai and R. W. Heath Jr., “Coverage and rate analysis for millimeter wave cellular networks", IEEE Trans. Wireless Commun, Feb. 2015. [2] T. Bai and R. W. Heath Jr., “Analysis of self-body blocking effects in millimeter wave cellular systems“, in Proc. of Asilomar Conf., Nov. 2014
• Derived SINR and rate distribution with building and self-body blockage effects
• Small form factor • Different types of array geometries: linear, planar, circular, … [1] Cudak, M. et. al., "Experimental mm wave 5G cellular system," in Globecom Workshops (GC Wkshps), 2014 , vol., no., pp.377-381, 8-12 Dec. 2014 [2] W. Hong; K. Baek; Y. Lee; Y. Kim; S. Ko, "Study and prototyping of practically large-scale mmWave antenna systems for 5G cellular devices," in Communications Magazine, IEEE , vol.52, no.9, pp.63-69, September 2014 [3] W. Roh et al. "Millimeter-wave beamforming as an enabling technology for 5G cellular communications: theoretical feasibility and prototype results," in Communications Magazine, IEEE , vol.52, no.2, pp.106-113, February 2014 [4] G. M. Rebeiz et. al. “Millimeter-wave large-scale phased-arrays for 5G systems” Proc. IEEE MTT-S International Microwave Symposium, 2015.
IEEE Communications Magazine • February 2014 109
presented in [5], where link- and system-levelsimulation results are provided with variousnumbers of transmit/receive antennas and RFchains. Using a 500 MHz bandwidth at 28 GHz,[5] presents some notable results for the hybridbeamforming system including an 8 dB gain overthe conventional spatial multiplexing schemeand 8 Gb/s average sector throughput with 16antennas with 4 RF chains at the base stationand 8 antennas with a single RF chain at themobile station.
MMWAVE BEAMFORMINGPROTOTYPE
In this section, we present a detailed descriptionof the mmWave beamforming prototype devel-oped and tested at the DMC R&D Center, Sam-sung Electronics, Korea, including systemconfiguration, key parameters, and capabilities.The main purposes of the mmWave prototypeare to check the feasibility of mmWave bandsfor sufficiently large geographical coverage forcellular services and support for mobility even inNLoS environments. As a result, an mmWaveadaptive beamforming prototype was developedincluding RF units, array antennas, basebandmodems, and a diagnostic monitor (DM), asshown in Fig. 3.
Both transmit and receive array antennashave two channels and each comprises 32 anten-na elements arranged in the form of a uniformplanar array (UPA) with 8 horizontal and 4 ver-tical elements, confined within an area of 60 mm× 30 mm. This small footprint was made possi-ble by the short wavelength of the carrier fre-quency at 27.925 GHz. Two channels at thetransmit and receive array antennas are designedto support various multi-antenna schemes suchas MIMO and diversity. The array antenna isconnected to the RF unit, which contains a setof phase shifters, mixers, and related RF circuit-ry. The set of phase shifters control the phasesof the signals sent to the antennas to form adesired beam pattern. Therefore, by setting thephase shifter values to a particular set, transmitand receive array antennas are capable of form-ing a sharp beam pattern in the intended hori-zontal (azimuth) and vertical (elevation) angles.
In order to reduce the hardware complexity,a sub-array architecture was employed to group
8 antennas into a sub-array, thus requiring only4 RF units per channel instead of 32. The reduc-tion in the number of RF paths results in areduction of antenna gain at the desired angle(except antenna boresight), a reduction of beamscanning ranges, and an increase in side lobelevels, but still meets the overall beamformingrequirements. The resulting full width at halfmaximum (FWHM) of the beam at the antennaboresight is approximately 10° horizontally and20° vertically with an overall beamforming gainof 18 dBi. In addition, a set of beam patterns ispredefined to reduce the feedback overheadrequired for the adaptive beamforming opera-tion between the transmitter and the receiver,where the overlapped beam patterns cover theintended service area with a unique beam identi-fier (ID) for each beam. These beam IDs areused by the baseband modem to control thephase shifter weights and to feed back the pre-ferred transmission beam information to thetransmitter. Table 1 lists key system parametersof the implemented prototype.
The baseband modem shown in Fig. 3 wasdesigned and implemented for real-time opera-tion with commercial off-the-shelf signal pro-cessing units including Xilinx Virtex-6 fieldprogrammable gate arrays (FPGAs), and anADC and a DAC each with up to 1 Gs/s conver-
Figure 2. Block diagram of a hybrid beamforming architecture.
IFFT DACMixer
RF beamformer RF beamformer
MIMOchannel
H
...
P/S
MIM
O encoder
Baseband precoder
Transmitter ReceiverBaseband channel
IFFT FFT
RF chains
DACPA
LNA
ADC
Array ant.
Phase shiftersRF chains
... ...
......
......
P/S
S/P Baseband combiner
MIM
O decoder
Ntc
Nta Nr
aNr
c
FFTADC...S/P
Figure 3. Configuration of the mmWave beamforming prototype.
RF/antenna
Modem
Array antenna Diagnostic monitor
ROH_LAYOUT.qxp_Layout 1/30/14 1:20 PM Page 109
Prototype phased arrays by Samsung [2],[3]
IEEE Communications Magazine • September 2014 67
The conformal topology further maximizes therange of the beamsteering scanning angles in theazimuth plane. Moreover, the slanted topologyconforms to the cellular device and enables thedesigned mmWave antenna array to appear asan extremely low profile metallic trace line thatencompasses the edges of the PCB. The width ofthe trace lines is less than 0.2 mm, which is evenless than the 1 mm spacing required from thePCB edges for conventional surface mount tech-nologies (SMTs). From the vantage point of thehardware layout, the inclusion of a total of 32mmWave antenna elements requires a negligibleantenna footprint. Based on this antenna solu-tion, a truly massive MIMO antenna system mayactually be realizable for mmWave 5G in thelong term.
The beam patterns of each set of phasedarray antennas are synthesized by the 28 GHzRF unit, composed of 32 6-bit phase shifters,power amplifiers, and low noise amplifiers forthe transmit and receive paths, respectfully.The phases of the 28 GHz RF signal are indi-vidually controlled to form a beam in theintended direction along the azimuth plane.Each mesh grid antenna element within the twosets of antenna arrays is connected with the 28GHz RF unit through K type coaxial connec-tors. The required RF signal phase informationrequired to steer the main lobe beam are storedand retrieved from the in-house designed base-band modem.
The modem analog front-end (AFE) is con-nected to the RF port of the RF unit to trans-mit and receive the complex analog basebandsignal. The analog beamforming algorithm usedin this work is designed to search for and identi-fy the strongest transmit and receive beamdirection within 45 ms. The current size of theRF unit and baseband modem prohibits fullimplementation inside the cellular phone proto-type in this research. We are exploring a num-ber of different approaches to completelyintegrate the mmWave antenna array, RF unit,and baseband modem in the foreseeable future.In the meantime, the mmWave cellular phoneprototype containing two sets of 16-elementmesh grid antenna arrays is tested and mea-sured in conjunction with a reference mmWavebase station prototype as illustrated in Fig. 4.The measurement scenario is confined to anLOS environment inside a laboratory located inthe headquarters of Samsung Electronics,Suwon, South Korea. The cellular phone proto-type is fixed at a distance of 6 m away from thebase station prototype. Afterward, both meshgrid antenna arrays are connected to the twoavailable downlink channels of the RF unit anddesignated as the device under test (DUT). A16-QAM signal with 528 Mb/s data rate is trans-mitted from the mmWave base station proto-type. Each of the antenna arrays are activatedalternately to separately measure the error vec-tor magnitude (EVM) for each discrete beamsteering angles and confirm 10–6 block errorrate (BLER). Based on this measurement, thenormalized radiation patterns of all the antennaarrays are retrieved. The identical procedure isrepeated for scenarios when the mesh gridantenna array is exposed to free space condi-
Figure 3. Photographs of the mmWave 5G antenna system prototype: a)standalone view of the antenna array with K type coaxial connectors; b)integrated inside a Samsung cellular phone and zoomed in views of themmWave antenna region.
<0.2 mm
16-element array 1
16-element array 2
(a)
(b)
Figure 4. Measurement configuration of the mmWave 5G cellular deviceprototype.
Carrier frequency
Bandwidth/duplexing
TX/RX configuration
Channel coding
Modulation
27.925 GHz
520 MHz / TDD
TX: base station (BS) ant.RX: mesh-grid array (MS)
LDPC
QPSK/16-QAM
256 elements (16 ×16) array(~24 dBi antenna gain)
MSBS
8 cm
8 cm
HONG_LAYOUT_Layout 8/28/14 4:45 PM Page 67
blocks are grouped together to form a TTI or slot. The payload burst is composed of 140 NSC-CP blocks containing 1 pilot block, 1 control block and 138 data blocks. 5 slots form a TDM frame and 40 frames form a TDM superframe. At a 1.536 GHz sampling rate, one slot period is exactly 100 us, one tenth of 4G and one superframe is 20 ms identical to 4G. Limitations on the available ADC rate have the system running at slightly slower sampling rate of 1.5 GHz. The numerology of the experimental system is captured in Table 1.
Figure 5 Frame Structure
The experimental system uses a standard LTE turbo decod-
er implementation for error correction. Multiple LTE physical resource blocks are mapped into three consecutive NCP-SC blocks and46 codewords occupy 138 NCP-SC blocks. As enumerated in Table 2, the experimental system has 4 modula-tion and coding levels providing data rates from 295 Mbps at cell edge to a peak of 2.3 Gbps for a single stream. For the commercial system it is envisioned that the system will support greater than 10 Gbps peak rate using 2x2 SU-MIMO by ex-ploiting polarization diversity of two RF antenna arrays.
Table 1 NCP-SC Frame & Slot Timing
Table 2 Modulation and Coding Levels
IV. EXPERIMENTAL SYSTEM
The experimental system was constructed using a dielectric lens antenna as a substitute for integrated mmWave panel as described in Section II. Much R&D is required to achieve the mmWave system vision and the lens acts as opportune proxy to prove many of the system concepts for access. A picture of the BS is shown in Figure 6.
Figure 6. Experimental mmWave BS
A dielectric lens focuses the mmWave energy like an opti-cal lens focuses light. The size and curvature of the lens de-termines the gain and beamwidth of the antenna. Figure 7 shows a block diagram of the dielectric lens system. In this case, the gain of the antenna is 28 dB and the corresponding half-power beamwidth (HPBW) is 3 degrees in both azimuth and elevation. The direction of the beam can be selected by moving the position of the focal point at the base of the lenses. This selection process is accomplished by using a 4 row by 16 column array of patch antennas to feed the lens. These 64 patch antennas are switched by 3 levels of SP4T switches that determine which one of the 64 elements is excited for trans-mission or selected for reception. The feeding array is de-signed such that the HPBWs slightly overlap, as seen in Figure 8, so that a gain within 3dB can be maintained over the steering range of the lens. The combination of the lens and feeder array may be steered +/- 4 degrees in elevation and +/- 17 degrees in azimuth. The 3-level switching matrix can be switched with 1 us settling time and driven by the baseband processing unit and switched in synchronization with the TDM slot structure.
Choice of carriers depends on base station density
• Compared SINR and rate performance using a common stochastic geometry model
Sub-6 GHz BW: 100 MHz Blockages statistics as in [3]
[1] T. Bai and R. W. Heath, Jr., “Comparing massive MIMO: millimeter wave or lower frequency?”, Preprint, 2015 [2] T. Bai and R. W. Heath, Jr., “Uplink massive MIMO SIR analysis: how do antennas scale with users?”, To appear in Proc. of Globecom 2015. [3] M. R. Akdeniz et al, “Millimeter wave channel modeling and cellular capacity evaluation”, JSAC, 2014.
Interference is still an important consideration in mmWave
• Transmission capacity INR can be large, especially for short D2D links
[1] Andrew Thornburg, T. Bai, and R. W. Heath, Jr., Interference Statistics in a Random mmWave Ad Hoc Network, in Proc. of the IEEE Int. Conf. on on Acoustics, Speech, and Signal Processing, Brisbane, AUS, April 19-24, 2015.
Interference can be strong in dense D2D networks Noise limited Interference limited
[1] Jianhua Mo and R. W. Heath, Jr., ``Capacity Analysis of One-Bit Quantized MIMO Systems with Transmitter Channel State Information, '’ IEEE TSP, 2015.
Reasonable capacity is possible at moderate SNR values
10mW 1 bit, 240 Gs/s
much less at 4 Gs/s
Baseband Precoding
Baseband Processing
1-bit ADC
1-bit ADC
1-bit ADC
1-bit ADC
RF Chain
Nr
RF Chain
Recent result from UT
• Use coarse but power efficient one-bit ADCs o Exact capacity characterization in some cases, high SNR bounds in other cases o Small loss at low SNR when channel state information is used at the transmitter
Millimeter wave can also be used for RF energy harvesting
• Studied the potential of energy harvesting and wireless power transfer @ mmWave
[1] T. Khan, A. Alkhateeb, and R. Heath,``Energy Coverage in Millimeter Wave Energy Harvesting Networks,” accepted in IEEE GLOBECOM Workshops, San Diego, 2015. [2] T. Khan, A. Alkhateeb, and R. Heath,``Millimeter Wave Energy Harvesting,”submitted to IEEE Trans. Wireless Commun., avaiable online at arXiv:1509.01653 [cs.IT].
Smart bridge
Smart building
mmWave BS Smart house
UHF
mmWave
Energy coverage prob. with different beam patterns.
• Approach to reduce the number of transceivers and ADCs/DCAs • Digital precoder/combiner can correct for lack of precision in the analog • Analog processing can be implemented with phase shifters, switches or lenses
Combine analog and digital beamforming
Baseband Precoding
1-bit ADC DAC
1-bit ADC DAC
RF Chain
RF Precoding
1-bit ADC DAC
1-bit ADC DAC
Baseband Combining
Nt Nr Lt Lr Ns Ns
RF Combining
FBB FRF WBB WRF
RF Chain
RF Chain
RF Chain
>= 1
>= 1
Different hardware leads to different tradeoffs performance-power consumption
Good tradeoff spectral efficiency-power consumption
• Switch different antennas to obtain phase differences • Lower power solution due to switch efficiency • Reduced complexity compared to phased shifters
[1] Roi Méndez-Rial, Cristian Rusu, Ahmed Alkhateeb, Nuria González-Prelcic and Robert W. Heath Jr., “Channel Estimation and Hybrid Combining for mmWave: Phase Shifters or Switches?”, ITA 2015
Baseband Precoding
Baseband Combining
1-bit ADC ADC
1-bit ADC ADC
RF Chain
RF Chain
Mr
LNA
LNA
LNA
Lr
Mr
LNA
LNA
LNA
A1: fully connected phase shifters A2: fully connected switches A3: subsets of phase shifters A4: subsets of switches
• Near-optimal performance compared with single-user rate • Reasonable hybrid precoding gain over analog-only beamsteering
[1] A. Alkhateeb, G. Leus, and R. W. Heath Jr., “Limited Feedback Hybrid Precoding for Multi-User Millimeter Wave Systems,” accepted in IEEE Transactions on Wireless Communications, arXiv preprint arXiv:1409.5162, 2014.
Sparse channel estimation with hybrid architectures
Recent result from UT
• Hybrid architectures with switches and phase shifters achieve comparable performance[1]
*
Compressed sensing tools can be used for efficient channel estimation [1] Roi Méndez-Rial, Cristian Rusu, Ahmed Alkhateeb, Nuria González-Prelcic and Robert W. Heath Jr., “Antenna selection for hybrid mmWave MIMO architectures”, submitted September 2015
Leveraging channel sparsity leads to efficient multi-user channel estimation [1] A. Alkhateeb, G. Leus, and R. W. Heath Jr, “Compressed-Sensing Based Multi-User Millimeter Wave Systems: How Many Measurements Are Needed?” in Proc. of the international conference on Acoustics, Speech, and Signal Processing (ICASSP), Brisbane, Australia, April 2015
+
+
+
FRF
Beamformerswu
RFcombiner
RF Chain
NBSNRF NMS
Base station uth mobile station
RF Chain
RF Chain
s1
sNRF
• Compressed sensing based multi-user channel estimation o All users channels are trained at the same time o Trade-off between training overhead & estimation quality
BS’s has 64-ant. ULA – 4 MS’s with 32-ant. ULA each
u High-end wearables used by people in public transport systems ª Commuting between home and office ª Highly interfering environment, 1-2 persons/meter2
u Support heterogeneous devices ª Rates from ~100 kbps to ~8Gbps
Modeling the dense wearable setting and performance analysis
27
For more information, see: 1. Kiran Venugopal, Matthew Valenti, and Robert Heath Jr, “Interference in finite-sized highly dense millimeter wave networks,”
Proc. of ITA, Feb 2015.
2. Kiran Venugopal, Matthew Valenti, and Robert Heath Jr, “Device-to-Device Millimeter Wave Communications: Interference, Coverage, Rate, and Finite Topologies,” submitted to Trans. Wireless Comm., available online arXiv:1506.07158 [cs.IT].
u MmWave can provide Gbps data rates to wearables ª Applicable to both fixed and random network scenarios ª Substantial variation as a function of location
u UT + UPF + TUT are creating foundation for mmWave wearables ª Understanding of fundamental limits of wearable mmWave networks ª Protocols that support multi-band and heterogeneous devices ª Channel models including self-body blocking and surface reflections