Testing 5G: Evolution or Revolution?

Transcription

1 ? Moray Rumney Lead Technologist October 3 rd 2016

2 Wireless evolution: 1990 to 2020 Increasing efficiency, bandwidth and data rates 2G 2.5G 3G 3.5G 3.9G 4G 5G PDC (Japan) imode W-CDMA (FDD & TDD) HSDPA HSUPA HSPA+ / E-HSPA embb HSCSD GSM (Europe) TD-SCDMA (China) EDGE Evolution LTE (R8/9 FDD/TDD) LTE-Adv. (R10 and beyond) mmtc GPRS IS-136 (US TDMA) E-GPRS (EDGE) 1x EV-DO 0 A B e (Mobile WiMAX) m / [WiMAX2] cdma2000 (1x RTT) Page Keysight Technologies 3 October 2016 Page 1 Market evolution MCC (UR/LL) Cellular IS-95A (US CDMA) IS-95B (US CDMA) WLAN d (Fixed WiMAX) WiBRO (Korea) ax ay b a/g h/n ac ad

3 Phases of Technology Adoption 1. Standardization The Committee Room 2. Regulation The Test House 3. Physics Maxwell s Place 4. Commerce The Shopping Mall And then there was light Where engineers consume lots of coffee while creating wireless law Where products are put on trial to prove their conformance to man-made laws Where electromagnetic law determines if it actually works What role can early trials play in the successful development of 5G standards? Where commercial law determines whether anyone actually buys it Page Keysight Technologies 3 October 2016 Page 2

4 Enabling future capacity and performance growth in cellular wireless communications There are many technical domains impacting 5G research including new modulation and coding schemes, carrier aggregation, higher order MIMO, network evolution etc. but two areas clearly dominate: 1. The move to mmwave frequencies and wider bandwidths Potential for >10x capacity growth with similar increase in peak data rates 2. The further exploitation of the spatial domain within cells through massive multi-user MIMO (MU-MIMO) Potential for >10x capacity growth over traditional sectored cells How will these two domains - new to cellular communications - impact system design, performance and testing? Page Keysight Technologies 3 October 2016 Page 3

5 Cellular communications at mmwave frequencies The possibility of mmwave cellular communications depends on two very important principles: 1. The problem - Path loss at mmwave frequencies increases with frequency 35 db more loss at 60 GHz than 1 GHz from Friis equation 2. The solution - Antenna size decreases with frequency With much smaller antennas it becomes feasible to create high gain arrays to overcome the path loss Hence: usable mmwave signals have a much narrower beamwidth than at RF -> rethink much of what we know about radio design/test Page Keysight Technologies 3 October 2016 Page 4

6 Evolution of radio channel models Correctly understanding the mmwave propagation is key to 5G Page Keysight Technologies 3 October 2016 Page 5

7 3GPP 5G Timeline 2020 may be closer than you think! Rel. 14 Rel. 15 Rel. 16 Rel. 17 & beyond SI: Channel Model SI: Scenarios and Requirements SI: 5G new RAT WI: 5G new RAT (Phase 1) WI: 5G new RAT (Phase 2) WI: LTE Evolution Page Keysight Technologies 3 October 2016 Page 6

8 mmwave channel modelling at 3GPP 3GPP completed its first study of mmwave propagation (6 100 GHz) in June 2016 captured in TR Many are concerned this work is rushed and will be incomplete The current TR extends the existing stochastic models developed over many years for <6 GHz There is concern such models may not sufficiently characterize the mmwave environment Keysight is also promoting a hybrid channel modelling approach which aims to more fully characterize the mmwave environment for the identified 5G use cases Page Keysight Technologies 3 October 2016 Page 7

9 Map-based Hybrid model basic idea Hybrid channel modelling calculates deterministic paths from a map up to a certain point. This increases the accuracy of the simulations because the environment is taken into account. When something is better modelled statistically, the hybrid methodology switches to stochastic blocks. For example, the human beings, trees, cars, etc. are difficult generally to be modelled deterministically but, instead, a probability distribution approach is more reasonable. Open issues being studied in mmmagic include: Spatial consistency, clustering, ground reflection, frequency dependency, channels for link level simulation, blocking (soft line of sight) Page Keysight Technologies 3 October 2016 Page 8

10 Stochastic modelling vs. deterministic and field trials All models attempt to predict reality The hybrid model claims to better represent reality with some higher complexity But ultimately real results from the field are what matter This is why experience from early field trials at 28 GHz is so crucial as it is the task of standards to predict reality since reality cannot be altered to match incomplete theory If the channel models are not accurate the NR design and subsequent test cases may not reflect reality Page Keysight Technologies 3 October 2016 Page 9

11 Keysight 60 GHz channel sounder with 2 GHz real-time bandwidth at University of Bristol in mmmagic project Page Keysight Technologies 3 October 2016 Page 10

12 Corner diffraction study ftp.3gpp.org/tsg_ran/wg1_rl1/tsgr1_84b/docs/r zip How well do 60 GHz signals bend round corners? Page Keysight Technologies 3 October 2016 Page 11

13 Corner diffraction measurements They don t! 25 db signal loss in just 10 cm of travel At these frequencies propagation is quasioptical Now you see me, now you don t Page Keysight Technologies 3 October 2016 Page 12

14 Diffraction is a strong function of frequency Simulated vs. measured at 3.5 GHz and 60 GHz At 3.5 GHz the shadow effect is much less pronounced Even at 2m distance with 40cm of travel: 60 GHz is at -25 db 3.5 GHz is much better at -8 db Page Keysight Technologies 3 October 2016 Page 13

15 mmwave surface scattering The built environment uses many material types, how do these behave at mmwave frequencies? Page Keysight Technologies 3 October 2016 Page 14

16 Rician K-factor Ratio of line of sight power to diffuse component power A pure line of sight environment has k = A pure non LoS environment (Rayleigh) has k=0 What would the K factor of a bowl of Special K be? That depends on the frequency and angle of incidence Page Keysight Technologies 3 October 2016 Page 15

17 Diffuse Scattering in mmwave Small-scale fluctuations in the received signal strength of each scattered path from a given surface can be modelled along a route as a function of a K-factor and a coherence distance (i.e. the distance where the spacedistance autocorrelation function falls to 0.9). The values of these parameters depend on the surface material (i.e. roughness) and on the incident angle (0 ο defined as the normal from the wall). Example for a rough wall: K-factor (db) Coherence distance (cm) Incident angle (= reflected 75 ο 45 ο 30 ο 75 ο 45 ο 30 ο angle) Rough wall Page Keysight Technologies 3 October 2016 Page 16

18 Preliminary K-factor Analysis of Wall Scatter 10 2 Cumulative Distribution Function (CDF) Estimated 5.9 db K-Factor Measured CDF 10 2 Cumulative Distribution Function (CDF) Estimated 5.7 db K-Factor Measured CDF Probability Envelope < Abscissa Signal Strength relative to Mean (dbm) Probability Envelope < Abscissa , K = 5.9 db 45, K = 5.7 db Signal Strength relative to Mean (dbm) 10 2 Cumulative Distribution Function (CDF) Estimated 5.0 db K-Factor Measured CDF 10 2 Cumulative Distribution Function (CDF) Estimated 3.6 db K-Factor Measured CDF Probability Envelope < Abscissa Signal Strength relative to Mean (dbm) Signal Strength relative to Mean (dbm) Page Keysight Technologies 3 October 2016 Page 17 Probability Envelope < Abscissa , K = 5.0 db 45, K = 3.6 db

19 Surface scattering measurement setup Page Keysight Technologies 3 October 2016 Page 18

20 Specular Reflection and Diffuse Scattering in mmwave Concrete Wall (Rough) Plasterboard Wall (Smooth) The LoS is blocked and the user moves 2 meters away from the AP; Analogue beamforming: exhaustive search is exploited AP: 32 antenna elements and forms 64 beams; User: 8 antenna elements and forms 16 beams; Specular reflection: the signal power of the reflection path from a given surface is calculated using the Fresnel reflection formula; Diffuse scattering: the small-scale fluctuations on top of the mean signal power is modelled along a route as a function of a K-factor and a coherence distance; Rough wall: K-factor = -3dB and coherence distance < 1cm; Smooth wall: K-factor = 5dB and coherence distance = 5cm; Page Keysight Technologies 3 October 2016 Page 19

21 Specular Reflection and Diffuse Scattering in mmwave Specular reflection: the reflected signal from the concrete wall (rough surface) has ~6 db higher power than the plasterboard (smooth surface) due to the dielectric constant Diffuse scattering: the diffuse signal power from the concrete wall has higher average power but much larger dynamic range as well as much shorter coherence than from the plasterboard Page Keysight Technologies 3 October 2016 Page 20

22 Specular Reflection and Diffuse Scattering in mmwave The available power using isotropic Tx and Rx antennas is ~ -89 dbm for the reflected signal and ~ -95m dbm for the diffuse signals But with beamforming and optimal beamsteering the available power rises to ~ -70 dbm for the reflected signal and ~ -75 dbm for the diffuse signals Page Keysight Technologies 3 October 2016 Page 21

23 Specular Reflection and Diffuse Scattering in mmwave But to get close to the theoretical 24 db beamforming gain possible with the BS and UE antenna arrays it is necessary to point the beams in the optimal direction. The pointing angle for the specular reflection (red) is nearly constant at the concrete wall but the optimal diffuse pointing angle (blue) varies rapidly. Page Keysight Technologies 3 October 2016 Page 22

24 Specular Reflection and Diffuse Scattering in mmwave Page Keysight Technologies 3 October 2016 Page 23

25 Specular Reflection and Diffuse Scattering in mmwave The green trace shows the available power using optimal beamsteering while the red and blue curves show the power when pointing only at the concrete or the plasterboard. The cdf curves predict the loss of power when the BS and UE arrays are not pointing in the optimal direction. At 90% probability there is a 5 db loss pointing only at the concrete wall which increases to 17 db when pointing only at the plasterboard Page Keysight Technologies 3 October 2016 Page 24

26 Mixed wall scattering Power statistics This setup emulates what a user might experience walking past a window in a wall with the UE blocked by body shadowing from a mmwave transmitter following the UE from across the street Page Keysight Technologies 3 October 2016 Page 25

27 Mixed wall scattering In-channel analysis For rough stone: The signal has lost its polarization diversity and has a 10 db slope For glass: The received signal shows 25 db polarization diversity and flat frequency response Page Keysight Technologies 3 October 2016 Page 26

28 Mixed wall scattering In-channel analysis at transition At transition from wood to glass: A few ms later and the null has moved across the channel making this a hard demanding signal to equalize 1 ns At transition from wood to glass: A strong reflection at 1 ns causes serious 20 db fade mid-channel Page Keysight Technologies 3 October 2016 Page 27

29 The challenge of acquiring and tracking narrow beamwidth mmwave signals Search Strategies High Gain Large Volume to search Low Probability of both stations pointing in the same direction Low-Gain: Higher Probability of looking in the right direction, but much less energy to detect Connected High Gain Tracking Requires multiple antennas for coverage Page Keysight Technologies 3 October 2016 Page 28

30 Blocking: At traditional frequencies used for cellular communications blocking is not a major issue Only 2.5 db of performance variation is seen in one 360 rotation for a directional UMi channel 2.6 GHz Page Keysight Technologies 3 October 2016 Page 29

31 Sources of mmwave Hand blocking Body blocking Corner diffraction Moving vehicles Outdoor to indoor transition Page Keysight Technologies 3 October 2016 Page 30

32 Solutions for blocking To mitigate hand/body blocking, devices will need to be carefully designed with multiple active antenna arrays to account for the proximity of different hand/body parts To mitigate other causes of blocking it will be necessary to design the network with spatial redundancy such that the device can simultaneously track and connect with more than one base station with near zero handover time This is similar to the soft handover concept in CDMA systems and will require much denser base station deployment than at low frequencies Page Keysight Technologies 3 October 2016 Page 31

33 Why getting the standard right matters Consider requirements and testing for the ideal teapot We could defining: Capacity Water tightness Thickness of wall Pouring ability Aesthetic appeal Cost And of course speed of pouring (Teaput!) Tests could then be defined for all the above Page Keysight Technologies 3 October 2016 Page 32

34 But what if A teapot is presented for test made entirely of chocolate! It would pass all the tests because the standard failed to predict a critical environment (hot tea) But such a device passing all tests would clearly be useless for a primary environment This can t be allowed to happen for 5G! Page Keysight Technologies 3 October 2016 Page 33

35 Looking ahead: The evolution of cellular radio performance testing. From 0D to dynamic 3D Spatial Domain Analysis (SDA) 1G G G G G 2023? 0D Cabled 1D SISO OTA Took 7 years to develop! 3D SDA is MUCH harder than MIMO OTA 2D MIMO OTA 3D SDA Page Keysight Technologies 3 October 2016 Page 34

36 Evolution of cabled vs. radiated over the air (OTA) test needs To exploit the spatial domain it is necessary to develop spatial requirements and spatial test techniques to match Test needs History Today Future mmwave Design verification Cabled Predominantly cabled, OTA in ascendance RF/baseband conformance Predominantly radiated Cabled Cabled Radiated (cable replacement) Radiated performance Limited to SAR/EMC SISO established, MIMO emerging Radiated 3D SDA Production Cabled Predominantly cabled Radiated (cable replacement) Problem: Existing OTA test techniques may not scale to mmwave 3D SDA Page Keysight Technologies 3 October 2016 Page 35

37 Reverberation chamber methods BS Emulator/ Channel emulator Reverberation chambers create complex uncontrolled spatial signals and usability at mmwave with dynamic antennas is to be studied Page Keysight Technologies 3 October 2016 Page 36

38 Radiated Two-Stage (RTS) antenna pattern method Stage 1 Antenna pattern measurement BS Emulator Measured pattern Stage 2 Throughput measurement BS Emulator Channel Emulator Load measured or simulated antenna patterns Connection to UE can be a conducted link (benchtop R&D) or a radiated link (anechoic chamber for conformance) RTS relies on static antenna patterns, and extension to active antennas is possible for R&D but unclear for conformance Page Keysight Technologies 3 October 2016 Page 37

39 Multi-probe anechoic (MPAC) (boundary array) method Example 8x1 antenna configuration shown Likely solutions are 8x2 cross polarized or 16x2 BS Emulator Channel Emulator The MPAC method has the potential to test active antennas but at mmwave there may be issues with how large the DUT can be Page Keysight Technologies 3 October 2016 Page 38

40 MPAC scalability to mmwave The quiet zone is the area of controlled power, the test zone is the area of controlled correlation within the quiet zone into which the radiating elements must fit. Does the test zone matter for mmwave? Quiet zone (power) Dual polarized Probes The quiet zone scales with chamber size not an 60 GHz The test zone is a function of the angular spacing of the probes. Test zone (correlation) An 8x2 system has a test zone in the 0.7λ to 1 λ range and 16x2 is around 1.6 λ which is 8 60 GHz Enough for just one array? Page Keysight Technologies 3 October 2016 Page 39

41 Alternative ideas Limited scope 3D Spatial Domain Analysis RRH RRH RRH L1/MAC Control & Channel Emulation Such an approach can handle devices with multiple spaced antenna arrays Page Keysight Technologies 3 October 2016 Page 40

42 4G vs. 5G testing Five reasons to be concerned about 5G performance testing: What How 4G RF Cabled test plus 2D spatial channel models with fixed geometry Cabled test plus Lots of choice, MPAC, RTS, Reverb 5G mmwave Yet undefined, much more dynamic and 3D No obvious solutions Importance Not critical Critical, no cabled fallback Difficulty Hard (OTA part) Very hard! Timescales OTA Approaching 8 years Needed urgently! The need and challenge of 5G OTA test is not yet widely understood but will become a showstopper in the near future so is an urgent research topic Page Keysight Technologies 3 October 2016 Page 41

43 The future is already here, it s just not evenly distributed. William Gibson Thank you for listening! Page Keysight Technologies 3 October 2016 Page 42