Filter Bank Multi-Carrier (FBMC) for Future Wireless Systems CD Laboratory Workshop Ronald Nissel November 15, 2016
Motivation Slide 2 / 27
Multicarrier Modulation Frequency index, l 17 0 0 x l,k...transmitted data symbol Time index, k 13 p(t) P(f) 1.5 1 0.5 0 2 1 0 1 2 0 50 t T 0 5 0 5 f F 0 = f T 0 Current generation of wireless systems use OFDM. Can we do better? FBMC OFDM (LTE,WLAN) Slide 3 / 27
5G Design Paradigm: Flexible Time-Frequency Allocation Robust communcation (reduced spectral efficiency) Low latency, high velocity (high efficiency if L >> K) F Frequency Time Required time-frequency resources for one symbol Large area (high efficiency if K >> L) T 1 F L... Number of subcarriers per block K... Number of time-symbols per block F... Frequency-spacing T... Time-spacing Slide 4 / 27
How does FBMC work? Slide 5 / 27
FBMC: Time-Frequency Localized Prototype Filter (Complex) orthogonal for TF = 2 Slide 6 / 27
FBMC Offset-QAM: Transmit Real-Valued Symbols Orthogonal only in the real domain: TF = 0.5 Slide 7 / 27
Comparison of Multicarrier Schemes Pure OFDM Windowed /Filtered OFDM Maximum Spectral Efficiency TF = 1 Time- Localization Frequency- Localization Independent (Bi)- Transmit Orthogonal Symbols yes yes no yes yes no yes yes yes yes FBMC-OQAM yes yes yes Coded FBMC-OQAM yes yes yes real only yes, after despreading yes no Slide 8 / 27
Comparison of Multicarrier Schemes Power Spectral Density (db) 0 20 40 60 80 No OOB reduction CP-OFDM Windowed OFDM Filtered OFDM 21 subcarriers FBMC OQAM T F = 1 (complex) LTE like: T F = 1.07 T F = 1.3 T F = 1.3 100 50 40 30 20 10 0 10 20 30 40 50 Normalized Frequency, f/f Slide 9 / 27
Real World Testbed Measurements Slide 10 / 27
Vienna Wireless Testbed Transmitter: TX antenna RX Receiver: Slide 11 / 27
Channel Estimation in FBMC 1 1 On Pilot-Symbol Aided Channel Estimation in FBMC-OQAM,R. Nissel,et.al.,2016 Slide 12 / 27
Channel Estimation in FBMC 1.5 Channel 1 0.5 0 12 10 Subcarrier-position l 8 Frequency, f 6 4 LS estimation, Data symbols Pilot symbols 2 2 4 6 8 10 12 Time-position k Interpolation Time, t 14 Slide 13 / 27
Channel Estimation in FBMC Cancel the imaginary interference at the pilot positions: 1 Auxiliary 2 Auxiliary Coding Frequency Time Novel Contribution: We suggest the usage of two auxiliary symbols. We extended the coding approach to more than 8 symbols Downloadable MATLAB code: www.nt.tuwien.ac.at/downloads/ Validated by real world measurements Slide 14 / 27
Achievable Rate Improvement: FBMC vs. OFDM (LTE) Achievable Rate Improvement (%) 30 25 20 15 10 5 2 Auxiliary symbols (low complexity) 1 Auxiliary symbol (low complexity) Coding (high complexity) OFDM no cyclic prefix 0 10 5 0 5 10 15 20 25 30 SNR OFDM D (db) FBMC Improvement due to better spectrum utilization Improvement due to no cyclic prefix Slide 15 / 27
Real World Throughput Measurements Throughput Improvement (%) 30 25 20 15 10 5 1 Auxiliary symbol (low complexity) 2 Auxiliary symbols (low complexity) Coding (high complexity) mean FBMC 95 % confidence interval obtained by bootstrapping 0 10 5 0 5 10 15 20 25 30 SNR OFDM D (db) Slide 16 / 27
MIMO in FBMC 2 2 Enabling Low-Complexity MIMO in FBMC-OQAM, R. Nissel and M. Rupp, 2016 Slide 17 / 27
Enabling MIMO in FBMC By Hadamard Spreading (+1,-1) Coded FBMC-OQAM: Code Time Frequency Novel Contribution: Matrix formulation: new analytical insights We investigate the block interference and the influence of time-varying channels Downloadable MATLAB code: www.nt.tuwien.ac.at/downloads/ Validated by real world measurements Slide 18 / 27
Real World MIMO Measurements 10 1 2 2 MIMO (2 bit rate) Zero-Forcing Equalizer Bit Error Ratio 10 2 10 3 16-QAM 95 % confidence interval for FBMC obtained by bootstrapping 2 2 MIMO Maximum-Likelihood Detection (2 bit rate) 2 1 MIMO Alamouti s Space-Time Block Code OFDM FBMC 10 4 5 0 5 10 15 20 25 30 Estimated Signal-to-Noise Ratio for OFDM (db) Slide 19 / 27
Bit Error Probability for FBMC in Doubly Selective Channels 3 3 OFDM and FBMC-OQAM in Doubly-Selective Channels: Calculating the Bit Error Probability, R. Nissel and M. Rupp, 2016, submitted Slide 20 / 27
Closed-Form BEP Expression 10 0 Bit Error Probability, Bit Error Ratio 10 1 10 2 64-QAM Vehicular A, 500km/h Pedestrian A, 10km/h OFDM (nocp) CP-OFDM FBMC CP-OFDM FBMC OFDM (nocp) Theory Gauss approx. Simulation 10 3 Doubly- flat 5 0 5 10 15 20 25 30 35 Signal-to-Noise Ratio [db], see (18) Slide 21 / 27
Outlook Slide 22 / 27
Doubly-Selective Channels in FBMC Questions: Are one-tap equalizers good enough? How good is FBMC compared to OFDM? Which prototype filter Hermite PHYDYAS Fair comparison: Signal-to-Interference Ratio Limiting factor: Doppler spread (time-variant channel) Optimal Subcarrier Spacing Limiting factor: delay spread (multipath delays) Subcarrier Spacing Slide 23 / 27
Doubly-Selective Channels in FBMC LS estimation of the piecewise mean channel MMSE channel estimation ˆ h LS l,k 1.5 1 0.5 A MMSE Ĥ[l, m] 1.5 1 0.5 0 12 10 8 6 4 Subcarrier 2 15 10 5 OFDM Symbol 0 12 10 8 6 Subcarrier 4 2 50 200 150 100 time index Channel estimation using time-frequency correlation Channel equalization Slide 24 / 27
Further Questions in FBMC Bit error probability, including channel estimation Similar to our submitted paper (which assumes perfect channel knowledge), But now also include channel estimation. mmwave transmissions Testbed Jointly with Erich and Martin Further investigate real world hardware effects Quantization Non-linear power amplifier Slide 25 / 27
Conclusions Slide 26 / 27
Conclusions In contrast to most other research groups: Downloadable MATLAB code: www.nt.tuwien.ac.at/downloads/ Theory is validated by real world testbed measurements Main contribution in 2016: Channel estimation in FBMC MIMO in FBMC Bit error probability for FBMC in doubly-selective channels Outlook: A lot of open questions for the next years! Slide 27 / 27
Backup Slides Slide 1 / 9
Ambiguity Function OFDM Slide 2 / 9
Ambiguity Function CP-OFDM Slide 3 / 9
Block Interference No guard-symbols: Transmit Power Block 1 Block 2 Block 3 KT KT KT Time Guard time-slot: Transmit Power Block 1 Block 2 Block 3 KT T KT T KT Set time slot to zero Time Slide 4 / 9
Block Interference 50 Signal-to-Interfernece Ratio (db) 40 30 20 10 One guard slot per block (reduced efficiency) No guard symbols (maximum spectral efficiency) 1 ms for 15 khz subcarrier spacing, (125µs for 120 khz) 0 0 50 100 150 200 250 Spreading Length (Symbols) Slide 5 / 9
Block Interference 20 1ms for 15 khz subcarrier spacing, (125µs for 120 khz) Efficiency Loss (%) 15 10 5 LTE efficiency loss due to the CP One guard slot per block 0 0 50 100 150 200 250 Spreading Length (Symbols) Slide 6 / 9
Time-Variant Channels Signal-to-Interference Ratio (db) 40 30 20 10 Spreading Length K = 32 One guard time slot per block Spreading Length K = 8 Theory Simulation 0 0 10 20 30 40 50 Velocity (km/h) for F=15kHz and f c =2.5GHz 0 50 100 150 200 250 300 350 400 Velocity (km/h) for F=120kHz and f c =2.5GHz Slide 7 / 9
Flexible Subcarrier Spacing, TF = 1 (no CP) Signal-to-Interference Ratio (db) 35 30 25 20 15 10 Channel Model: VehicularA, 250 km/h FBMC-OQAM Hermite OFDM FBMC-OQAM Phydyas 15 khz 5 (LTE) Theory Simulation 0 5 10 15 20 25 30 35 40 45 Subcarrier Spacing (khz) Slide 8 / 9
Peak-To-Power Average Ratio and Signal Power Over Time Pr (PAPR PAPR0) 10 0 10 1 10 2 10 3 OFDM Cod. 2Aux. 1Aux. 1Aux. flipped Signal Power 1.6 1.4 1.2 1 0.8 0.6 0.4 1Aux. flipped 1Aux. Cod. 2Aux. 10 4 6 8 10 12 PAPR 0 (db) 0.2 6T FBMC 0 0 0.1 0.2 Time (ms) Slide 9 / 9