Spectrum Sharing between Radar and Communication Systems

Transcription

1 Spectrum Sharing between Radar and Communication Systems Mo Ghorbanzadeh and Awais Khawar Hume Center for National Security and Technology Bradley Department of Electrical and Computer Engineering Virginia Tech, Arlington Student Seminar Series November 21, 2014

2 Outline Radar to LTE Base Stations Interference Simulations Background Radar Basics Prior Coexistence Work Simulations Conclusions Radar interference mitigation Prior Work Spatial Spectrum Sharing Performance Analysis Radar Waveform Design for Spectrum Sharing Conclusions

3 Disclaimer: Results presented hereafter are preliminary and will continue to be refined. LTE Base Station Receiver non-linear effects of saturation and front-end overload from radar signals are for further study.

4 Background The proposed CBS band Band 42 TDD ( ) Band 43 TDD ( ) 3400 National Telecommunications and Information Administration (NTIA) identified existing 3.5 GHz Federal operations. - Radiolocation systems: Includes Dept. of Defense (DoD) ground-based (GB), shipborne (SB), & airborne (AB) surveillance/tracking radars. High-power surveillance measure targets altitude, range, and bearing at ranges as great as 300 nmiles. Air Force assisting pilots in formation flying, drop-zone training Weapon control systems (e.g. data update communications to missiles, gunfire control) in MHz, air defense in MHz. - Radionavigation systems: Includes Air Traffic Control (ATC), air marshalling, and short-range air-search radar systems. Navy ship-borne radars operate in 21 channels throughout this band. Frequency relocating the above may require new technology and significant redesign. Radars increasingly operate over larger bandwidths to improve image resolutions as targets grow complex. - NTIA focused on geographic sharing leading to geographically-limited licensing. Adjacent band radars must be considered as they may pose an interference to the deployment of 3.5 GHz wireless systems. - Potential interference from in-band and adjacent radars might significantly limit how much spectrum is fully usable

5 Radar Block Diagram Radars transmit high power pulses into outer space. Pulses undergo atmospheric, propagation, and system attenuation. Hitting an object, the pulse radiates in all directions (target acts an omni source). The received pulses are at the same frequency as the radar pulses plus a doppler shift. The received echo reveals a target is detected. Targets can be as far as km away (ballistic missile early warning system).

6 Operating Frequency: Carrier frequency within each pulse. Radar Operating Parameters Pulse Repetition Interval (PRI): Time between pulse arrivals. Peak Power: Maximum pulse power. Pulse energy is peak power multiplied by the pulse width (PW). Rotation Speed: Complete horizontal scans in a minute. Gives the horizontal scan time. Average power is peak power amplitude (PA) multiplied by duty cycle (pw/pri). Average energy in the pulse is average power multiplied by the duty cycle. Azimuth/Elevation (θa/ θe) beamwidths relate to antenna diameters D as typically 75λ/D an gain as 33000/(θA θe). Dwell time: Time antenna beam spends on a target & relates to θ A and antenna rotation under non-track mode.

7 Interference Scenario 1-10 GW peak effective isotropic power (EIRP) via tube type transmitters (TX)S MW EIRP via solid state TXs with longer PWs. More effect on Base Station (BS) 1 than user equipments (UEs). 1. F. Sanders, J. Carrol, G. Sanders, R. Sole, Effects of Radar Interference on LTE Base Station Receiver Performance, NTIA, U.S. Dept. of Commerce, 2013.

8 Propagation Models Airborne Radars: Free Space propagation path loss LFS (db) for central frequency f (MHz) & distance d (km) at line-of-sight (LoS) for interferer and victim antenna heights hi and hv in meters. L FS 20log( f ) 20log( d) 32.45; d 4.1( hi hv ) In the non-los (NLoS) region, a diffraction loss 0.62/λ0.33dB/km is added (λ~10cm for S band radars). Ground Based and ShipBorne Radars: Irregular Terrain Model (ITM), an improved version of Longley-Rice, predicts signal NLoS loss (median attenuation) in 20 MHz-20 GHz as a function of distance and signal temporal, location, and situation variability. Temporal variability accounts for attenuation variability and can be short (due to fading) or long term (due to atmospheric and seasonal conditions such as snow, soil moisture, vegetation, and foliage) and expressed as %, time duration when received field strength is expected not to be less than the hourly median field. Location variability describes the full range of signal levels over the small area and is based on antenna location changes, and expressed as %, fraction of locations where field strength is expected not be less than the median field strength. Situation variability is a probability measure imposed on collection of all propagation paths. Expressed as %, it gives fraction of identical paths on which field strength is expected not be less than field computed by the program.

9 WiMAX/Radar Exclusion Zones

10 WiMAX-Shipborne Radar Exclusion Zones

11 Comments The exclusion zones are extracted from link budget analysis of radar-wimax systems. NTIA: Any changes to the involved system can change the obtained exclusion zones. LTE will be one of the predominant cellular technologies in the 3.5 GHz band. New exclusion zones with respect to LTE deployments are needed.

12 Radar Interference Effect on LTE Base Stations in the 3.5 GHz Band

13 Radar Simulation Radar parameters in the table are adopted from NTIA s Fast Track Evaluation 1. Radar radiates on LTE at 50, 100, 150, and 200 km away. 83 dbm without antenna, and = 128 dbm radiation. 360 deg horizontal scan. Parameters Value Operating Frequency 3.5 GHz* Peak Power 83 dbm Antenna Gain 45 dbi Antenna Pattern Cosine Antenna Height 50 m Insertion Loss 2 db Pulse Repetition Interval 0.5 ms Pulse-Width 78 µs Rotation Speed 30 rpm* Azimuth Beam-Width 0.81 deg* Elevation Beam-Width 0.81 deg* Azimuth Scan 360 deg Distance to LTE 50, 100, 150, 200 km 1 An Assessment of the Near-Term Viability of Accommodating Wireless Broadband Systems in the MHz, MHz, MHz, MHz and MHz Bands NTIA, U.S. Dept. of Commerce, Nov. 2010

14 Radar Simulation Radar circulates at 30 rotation per minute (rpm). Horizontal scan time becomes 2 s. 360/0.81 = 445 beam positions for the search fence. Antenna dwell time becomes 2/445 = 4.5 ms. PRI = 0.5 ms gives 9 pulses during the dwell time. An BS under radiation is hit by 9 pulses pulses (each dbm) are radiated in a rotation of the antenna. At distance R radar radiation diameter becomes: d 2Rtan( a ) 0. 03R 1.5, 3.0, 4.5, 6.0 km radiation diameter when radar is 50, 100, 150, and 200 km away.

15 Antenna back-lobe -50 db vs. the main lobe. Radar Simulation Based on ITU-R M.1851 (mathematical model for radar antenna used in NTIA Fast Track Evaluation). G( ) 68.8 sin( ) cos( ) 3dB ( sin( ) ( ) ( ) 2 3dB loge ( ) 3dB 50 db 2 )

16 Antenna Pattern for Macro LTE BS LTE Simulation (Macros, Outdoor Small Cells) i i, t 2 Gi ( i ) min{12( ), Am }, i { A, E} G min{ ( GA( A) GE ( E ), Am } 3dB Parameters Operating Frequency Layout Mode Value 3.5 GHz Hexagonal macro cell grid, clustered small cells TDD Macro/Small Cells BS TX Power 46/30 dbm UE Transmit (TX) Power 23 dbm Macro-cell sites/cells 7/21 (3 cells per site) Small cells 84 (4 per macro cell) Indoor UE ratio for Macro / Small cells 80% / 20% Bandwidth for Macro / Small cells 20 MHz BS Antenna Gain for Macro / Small cells 17/ 5 dbi UE Antenna Gain 0 dbi Macro Inter-site Distance (ISD) 500 m Minimum UE-BS Distance for Macro / Small cells 25 / 5 m BS Antenna Downtilt for Macro 12 deg BS Antenna for Small Cells Omni-directional BS Antenna Height 25 (macro), 10 (outdoor small cells) UE Antenna Height 1.5m UE Distribution for Macro / Small cells Uniform/Clustered UE Mobility 3 km/h, uniform direction BS/UE Noise Figure (NF) 5/9 db Thermal Noise -174 dbm/hz Service Profile Full buffer best effort UEs per Cell for Macro / Small cells 10 / 30 Channel Model for Macro / Small cells UMa / UMi [1] 1 3GPP TR V9.0.0 ( ), Further advancements for E-UTRA physical layer aspects, Release 9.

17 LTE Simulation (Indoor Small Cells) Parameters Value Operating Frequency 3.5 GHz Layout Indoor hall Mode TDD BS TX Power 30 dbm UE Transmit (TX) Power 23 dbm Indoor Small cells 2 Indoor UE 100% Bandwidth 20 MHz BS Antenna Gain 5 dbi UE Antenna Gain 0 dbi BS Antenna Omni-directional BS Antenna Height 6 m UE Antenna Height 1.5m UE Distribution Uniform UE Mobility 3 km/h, uniform direction BS/UE Noise Figure (NF) 5/9 db Thermal Noise -174 dbm/hz Service Profile Full buffer best effort UEs 20 (10 per indoor small cell) Channel Model for Small cells InH[1] 1 3GPP TR V9.0.0 ( ), Further advancements for E-UTRA physical layer aspects, Release 9.

18 LTE System Simulation Model 3GPP-compliant system-level simulator. 3GPP-defined macro, small-cell and indoor scenarios. Utilizes proportional-fair scheduler in both time and frequency domains. Detailed UL air interface modeling, UL MIMO, and receiver diversity. Non-ideal link adaptation with Hybrid ARQ and EESM linkto-system mapping.

19 LTE Model Enhancements From Nokia s FNPRM Filings Modeling RF receiver saturation threshold More precisely modeling Turbo decoder saturation Updated SC-OFDMA SINR calculation with radar interference present Explicitly using pilot symbols for Base Stations Interference Measurements

20 Macro Cells Layout Macro cell layout for 7 sites. 500 m ISD Macro Cell Layout Macro Site Positions UE Positions 400 m m

21 Outdoor Small Cells Layout 800 Macro and Small Cell Layout 600 Pico cell UE Macro cell m m

22 Indoor Small Cells Layout For an UE at a distance R (km) from the LTE BS: PL LoS = log (R) PL NLoS = log(R) The LoS probability is: Indoor Small Cell Layout UE Indoor Pico 1 R exp( ( R 0.018) / 0.027), R R m m

23 Propagation Models (radar-lte path) In LoS, FSPL represents the loss radar signal undergoes. L db, FSPL ( r) 20log( f ) 20log( r) 32.45, r r LoS r LoS 4.1( hradar hlte ) In NLoS region, ITM model represents the loss. Parameters Value Operation Mode Area Prediction Mode Small, Macro cells LTE/Radar Antenna Height 10, 25/50 m Dielectric Constant 15 Conductivity S/m Refractivity 301 N-units Climate Continental Temperate Variability Mode Single Message Surface Refractivity 15 Sitting Criteria Random

24 Simulation Results (Macro Cells) Signal-to-interference-to-noise ratio (SINR) of an LTE macro BS versus LTE symbol and subcarrier indices. Even when radar is present, SINR recovers until next pulse. Radar pulse is centered in the LTE band, so most energy is concentrated around subcarrier 300 (middle of the LTE channel). 78 µs wide pulses exceed the duration of the LTE symbol (71.4 µs). Energy is mostly concentrated in symbols 1 and 8, with some remaining pulse energy also present in symbols 2, 9 and 14.

25 Normalized Throughput. Simulation Results (Macro Cells-UMa)

26 Normalized Throughput. Simulation Results (Outdoor Small Cells-UMi)

27 Normalized Throughput. Simulation Results (Indoor Small Cells-InH)

28 Normalized Throughput Simulation Results (Comparison of indoor/outdoor/macros) Macro Cells Outdoor Small Cells Indoor Small Cells km 100 km 150 km 200 km Baseline

29 Simulation Results (Out of Band) Radar at 50 km. Radar operating frequency offsets from the LTE frequency.

30 Conclusions The exclusion distances between radars and LTE in NTIA Fast Track Report are overly conservative Need to accurately model the radar and LTE systems. Need to better characterize the propagation characteristics between radars and LTE On-going/Future work Downlink (radar to LTE UEs interference) RF saturation and burnout Interference mitigation and avoidance LTE to radar interference Premature to lock in exclusion zones What is FCC s timeline? Will there be opportunities for reconsideration even after initial rules (e.g., to decrease zones further, allow LTE deployments inside exclusion zones, i.e., convert to coordination zones, etc)?

31 Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary Spectrum Sharing between Radar and Communication Systems Awais Khawar Hume Center for National Security and Technology Bradley Department of Electrical and Computer Engineering Virginia Tech, Arlington Student Seminar Series November 21, 2014

32 Table of contents Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary 1 Introduction Prior Work Contributions 2 Models 3 NSP 4 Performance 5 Waveform Design 6 Summary

33 Motivation Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary Motivation: Lack of available frequency spectrum for communications is a limiting factor in the use of wireless systems. One solution is to use Radar bands for communication systems. Traditional Radar and Communications systems cannot occupy same frequency bands due to interference between the two systems. Solution: MIMO Radar transmits in the Null Space of the MIMO Communication system Expectations: Small degradation in radar performance Big gain in communication capacity

34 Prior Work Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary Radar spectrum sharing in the past: Wireless Local Area Network (WLAN) devices U-NII Frequency Bands: and GHz Incumbent example: weather surveillance radar (TDWR) Opportunistic sharing using dynamic frequency selection (DFS) mechanism Transmit power control Limitations of DFS: No co-channel sharing Non-occupancy period of at least 30 minutes Double checking a false alarm is not allowed DFS not always reliable due to errors in sensing

35 Contributions Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary 1 Waveform Shaping Single MIMO communication system Multi-cell MIMO communication system Performance analysis Cramér Rao bound (CRB) Maximum likelihood (ML) estimate Variations in beampattern Probability of target detection 2 Waveform Design BPSK waveforms QPSK waveforms Constant envelope QPSK waveforms

36 Radar and Communication System Models Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary Colocated MIMO Radar M T transmit and M R receive antennas MIMO Communication Systems N T transmit and N R receive antennas Single Cell Model: r(t) = H I x(t) + Hs(t) + w(t) Multiple Cell Model: r i (t) = H i x(t) + j H js UE j (t) + w(t)

37 Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary NULL SPACE PROJECTION

38 Null Space Projection Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary NSP is a spatial approach for coexistence Utilizes spatial dimensions for spectrum sharing Also known as spatial division multiplexing (SDM) To avoid interference, make [H i ]x(t) = 0 Array Configurations for NSP Case M T > N R M T N R Solution Conventional NSP NSP with threshold NSP requires rank of channel to be known Cooperative Sharing: CSI exchange between radar and comms (e.g. military radar and military comms) Non-cooperative Sharing: CSI estimation by radar (e.g. military radar and commercial comms)

39 Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary RADAR METRICS

40 Performance Metrics Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary NSP radar waveform x(t) = Px(t) results in R R = I MT, also rank( R) < rank(r). Target s Angle of Arrival Estimation Cramér Rao Bound (CRB) Maximum Likelihood (ML) Estimate Transmit/Receive Beampattern Variations in main beam Variations in sidelobes Probability of Target Detection Additional gain required

41 Simulation Results Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary RMSE (degree) 10 4 Original Waveform NSP Waveform onto H Best NSP Waveform onto H Worst SNR Power (db) 40 Original Waveform NSP Waveform onto small null space 20 NSP Waveform onto large null space θ (deg) 25 1 P D for P FA = dB 5dB theta estimate(deg) P D P D for NSP Waveforms to BS 1 P D for NSP Waveforms to BS 2 P D for NSP Waveforms to BS 3 P for NSP Waveforms to BS 4 D P D for NSP Waveforms to BS 5 P D for Orthogonal Waveforms Original Waveform NSP Waveform onto H Best NSP Waveform onto H Worst theta (deg) SNR

42 Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary BEAM-PATTERN MATCHING OPTIMIZATION

43 Goals and Problem Description Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary MIMO radar waveform design with sharing constraints Power received from the target at location θ k : P(θ k ) = E{a H (θ k ) x(n) x H (n) a(θ k )} = a H (θ k ) R a(θ k ) Waveform design problem (previous approaches) 1 min ψ ij K K ( 2 2 π ah (θ k ) sin 1 (R)a(θ k ) φ(θ k )) k=1 Waveform design problem (proposed) 1 min ψ ij K K ( 2 2 π ah (θ k )P sin 1 (R)P H a(θ k ) φ(θ k )) k=1 BPSK: R UU H QPSK: R R(R) + j I(R)

44 BPSK Waveform Results Introduction Prior Work Contributions Models NSP Optimization Problem BPSK Waveform Z Desired Beampattern Projection Z NSP Block P V H Best MIMO Radar Tx/Rx H i... Optimization Problem BPSK Waveform Desired Beampattern Projection Block Z opt P V Z opt NSP H Best MIMO Radar Tx/Rx H i... Performance Channel Selector Channel Estimator Channel Selector Channel Estimator Waveform Design Summary X = Χ Δ 1 2 W H Gaussian RV Generation Interference Channel Selection Block dim [N(H i)] H i SVD Block X = Χ Δ 1 2 W H Gaussian RV Generation Interference Channel Selection Block dim [N(H i)] H i SVD Block

45 QPSK Waveform Results Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary

46 Summary Introduction Prior Work Contributions Models NSP Performance Waveform Design Summary Null space projection algorithm Conventional NSP (M T > N R ) NSP with threshold (M T N R ) Radar performance metrics Cramer-Rao bound Angle estimation (azimuth or elevation) Beampattern side-lobe suppression in db Probability of Detection Waveform design with spectrum sharing constraints BPSK waveform QPSK waveform