Eleftherios Kofidis - Channel Estimation in Filter Bank-based Multicarrier Systems: Fundamentals and Recent Advances

Transcript

1. Channel Estimation in Filter Bank-based Multicarrier Systems: Fundamentals and Recent Advances Eleftherios Kofidis Computer Technology Institute, Greece University of Piraeus, Greece
2. 31 Aug. 2015 CentraleSupelec, Rennes 2 Future mobile networks – Vision and needs  High  data rate  reliability  QoS in demanding transmission scenarios  Increased flexibility  Efficient use of fragmented spectrum  Robustness to asynchronism  Co-existence of different systems (HetNets)  …
3. 31 Aug. 2015 CentraleSupelec, Rennes 3 Toward a new PHY – Modulation  Is OFDM an adequate solution?  Poor spectral containment  Bandwidth/power inefficiency  Challenging synch in multi-access  Sensitivity to severe dispersions  …  FBMC: an attractive alternative  Good spectral (/time) containment  High spectral (/power) efficiency  Flexibility (e.g., for multi-mode comms)  Relaxed synch requirements  Able to cope with severe multipath (e.g., large cells) and high mobility  …
4. 31 Aug. 2015 CentraleSupelec, Rennes 4 FBMC research and applications Filter bank-based multi-carrier modulation: • FBMC/OQAM • FMT • GFDM • UFMC • …
5. 31 Aug. 2015 CentraleSupelec, Rennes 5 FBMC research and applications Filter bank-based multi-carrier modulation: • FBMC/OQAM • FMT • GFDM • UFMC • … • Max. spectral efficiency • Time-freq. localization • Robust to lack of synch • But: Intrinsic interference
6. 31 Aug. 2015 CentraleSupelec, Rennes 6 FBMC/OQAM challenges - Solutions  Intrinsic ISI/ICI  Frequency / time selective subchannels  Challenges in Channel Estimation (CE)  Classical assumption: channel of low freq./time selectivity  CE analogous (similar) to OFDM  Preamble/pilots design for increased accuracy  However: in many realistic scenarios  Severe performance error floors  outperformed by OFDM at higher SNRs  More recently: CE training and techniques for demanding channels
7. 31 Aug. 2015 CentraleSupelec, Rennes 7 Outline  Fundamentals of FBMC/OQAM  System model  Intrinsic interference effect  FBMC/OQAM CE fundamentals  Preamble-based  Pilot-based  Preamble-based CE  Low frequency selective channels  Highly frequency selective channels  Simulation examples  Additional results - on-going/future work
8. Fundamentals of FBMC/OQAM 31 Aug. 2015 CentraleSupelec, Rennes 8
9. 31 Aug. 2015 CentraleSupelec, Rennes 9 FBMC/OQAM vs. OFDM/QAM   1 2 F   complex QAM real imaginary F=1/T: sub-carrier spacing T: OFDM/QAM symbol duration T-F density: OFDM/QAM (without CP): 1/(TF)=1 OFDM/OQAM: Spectral efficiency (e.g., (O)QPSK): OFDM/QAM (without CP): 2/(TF)=2 OFDM/OQAM:   1 2 F     1/ 2 F   Phase space
10. 31 Aug. 2015 CentraleSupelec, Rennes 10 Offset-QAM Modulation (staggering) Re Im 2 2 z-1 + d2k,n c2k,m Im Re 2 2 z-1 + d2k+1,n c2k+1,m even sub-carriers odd sub-carriers
11. 31 Aug. 2015 CentraleSupelec, Rennes 11 FBMC/OQAM Transmitter IFFT 2 0 ( ) A z 2 1 ( ) A z 2 1 ( ) M A z  2 M  2 M  2 M    1  z 1  z       0,n   0,n   1,n   1, M n    1, M n    1,n  0,n d 1,n d 1, M n d  C2R C2R C2R OQAM modulation Transform block Polyphase filtering P/S conversion SFB: P. Siohan et al., “Analysis and design of OFDM/OQAM systems based on filterbank theory,” IEEE Trans. SP, May 2002.
12. 31 Aug. 2015 CentraleSupelec, Rennes 12 FBMC/OQAM Receiver 2 0 ( ) B z 2 1 ( ) B z 2 1 ( ) M B z  FFT 1  z 1  z 2 M  2 M  2 M        Subchannel processing Subchannel processing Subchannel processing * 0,n  * 1,n  * 1, M n    * 0,n   * 1,n  Re  * 1, M n   0,n d 1,n d 1, M n d  Re Re   R2C R2C R2C S/P conversion Polyphase filtering Transform block OQAM demodulation AFB:
13. System model 31 Aug. 2015 CentraleSupelec, Rennes 13
14. 31 Aug. 2015 CentraleSupelec, Rennes 14 System model (1)  M: #subcarriers  K: overlapping factor  g: prototype filter (length ) C2R SFB h + AFB Intrinsic interference:
15. 13 March 2014 Patras (ENDECON) 15 Intrinsic interference in FBMC/OQAM 1 , 1 , 1 1 , 1 1 , 1 , 1 1 , 1 1 , , 1 , 1 , 1 , 1 1 , 1 1 , 0 , 0 1 , 0                    n M n M n M n k n k n k n k n k n k n k n k n k n n n d d d d d d d d d d d d d d d      
16. 13 March 2014 Patras (ENDECON) 16 Intrinsic interference in FBMC/OQAM With good TF localization, contributions to intrinsic interference only come from the first-order neighboring TF points 1 , 1 , 1 1 , 1 1 , 1 , 1 1 , 1 1 , , 1 , 1 , 1 , 1 1 , 1 1 , 0 , 0 1 , 0                    n M n M n M n k n k n k n k n k n k n k n k n k n n n d d d d d d d d d d d d d d d      
17. 13 March 2014 Patras (ENDECON) 17 Intrinsic interference in FBMC/OQAM With good TF localization, contributions to intrinsic interference only come from the first-order neighboring TF points                    1 , 1 , 1 1 , 1 1 , 1 , 1 1 , 1 1 , , 1 , 1 , 1 , 1 1 , 1 1 , 0 , 0 1 , 0                    n M n M n M n k n k n k n k n k n k n k n k n k n n n d d d d d d d d d d d d d d d      
18. 13 March 2014 Patras (ENDECON) 18 Example – “PHYDYAS filter” FBMC/OQAM TMUX transfer function (interference function): ( - Even k - after “de-phasing” ( ) to bring into the form - before that: green real, brown  imaginary  OQAM ! ) time freq. n-4 n-3 n-2 n-1 n n+1 n+2 n+3 n+4 k-1 j0.005 -j 0.043 j0.125 -j0.206 j0.239 -j 0.206 j0.125 -j0.043 j0.005 k 0 j0.067 0 j0.5644 1 -j0.5644 0 -j0.067 0 k+1 -j0.005 -j0.043 -j0.125 -j 0.206 - j0.239 -j0.206 -j0.125 -j 0.043 -j0.005   * , k n k n j      , , , k n k n d ju k   • N. J. Fliege, “DFT polyphase transmultiplexer filter banks with effective reconstruction,” EUSIPCO 1992. • C. S. Lee and K. Y. Yoo, “Polyphase filter-based OFDM transmission system,” VTC-2004 (Fall).
19. 13 March 2014 Patras (ENDECON) 19 More examples IOTA filter Bregović-Saramäki filter P. Siohan and C. Roche, IEEE Trans. SP, Dec. 2000. M. G. Bellanger, ICASSP-2001. R. Bregović and T. Saramäki, IEEE Trans. SP, Aug. 2005 PHYDYAS filter
20. 31 Aug. 2015 CentraleSupelec, Rennes 20 System model (2)  Common assumptions (locally freq./time-invariant channel):
21. 31 Aug. 2015 CentraleSupelec, Rennes 21 System model (2)  Common assumptions (locally freq./time-invariant channel):
22. 31 Aug. 2015 CentraleSupelec, Rennes 22 System model (2)  Common assumptions (locally freq./time-invariant channel):
23. 31 Aug. 2015 CentraleSupelec, Rennes 23 System model (2)  Common assumptions (locally freq./time-invariant channel): OFDM-like
24. 31 Aug. 2015 CentraleSupelec, Rennes 24 System model (2)  Common assumptions (locally freq./time-invariant channel): OFDM-like colored virtual Tx symbol (pseudo-symbol)
25. FBMC/OQAM channel estimation: Fundamentals 31 Aug. 2015 CentraleSupelec, Rennes 25
26. 31 Aug. 2015 CentraleSupelec, Rennes 26 Preamble-based channel estimation (1) Control / Data Preamble Frame: SFB non-zero part 0 0 prevents interference from previous frame (often unnecessary!) prevents interference from control/data channel time invariant
27. 31 Aug. 2015 CentraleSupelec, Rennes 27 Preamble-based channel estimation (2) Control/Data Preamble Full (block-type): Control/Data Sparse (comb-type): 0 0 protect from ICI
28. Scattered pilot-based channel estimation  Help (auxiliary) pilot 31 Aug. 2015 CentraleSupelec, Rennes 28 J.-P. Javaudin, D. Lacroix, and A. Rouxel, VTC-2003 (Spring). 1 , 1 , 1 1 , 1 1 , 1 , 1 1 , 1 1 , , 1 , 1 , 1 , 1 1 , 1 1 , 0 , 0 1 , 0                    n M n M n M n k n k n k n k n k n k n k n k n k n n n d d d d d d d d d d d d d d d      
29. Preamble-based channel estimation 31 Aug. 2015 CentraleSupelec, Rennes 29
30. Interference Approximation Method (IAM): Interference in a positive role!  Known input  interference approximation possible  pseudo-pilots  Choose input so as to maximize pseudo-pilot magnitude  Compute channel estimate (as in OFDM): 31 Aug. 2015 CentraleSupelec, Rennes 30 0,0 0,1 0,2 1,0 1,1 1,2 2,0 2,1 2,2 1,0 1,1 1,2 M M M d d d d d d d d d d d d    C. Lélé et al., “Channel estimation methods for preamble-based OFDM/OQAM modulations,” European.Trans. Telecomm., 2008. estimation error
31. Example: IAM-R  Null side symbols ( base design on middle symbol only)  Carefully choose signs so as to maximize pseudo-pilots’ magnitude 31 Aug. 2015 CentraleSupelec, Rennes 31 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0     31  d d  d Idea: Example: M=8, OQPSK
32. More IAM variants – Using imaginary pilots  Idea: Use Imaginary pilots to generate imaginary- or real-valued pseudo-pilots (of even larger magnitude)  Not a strictly OQAM input! 31 Aug. 2015 CentraleSupelec, Rennes 32 • C. Lélé et al., ICC-2008. • J. Du and S. Signell, ICC-2009. • PHYDYAS deliverable D3.1 • E. Kofidis and D. Katselis, EUSIPCO-2011. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 jd d d jd d d jd d     0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 j j j j     1 1 1 1 1 1 1 1 1 1 1 1             j j j j j j j j j j j j IAM-I IAM-C E-IAM-C 1/3 of the subcarriers:
33. Price for good performance: high PAPR! 31 Aug. 2015 CentraleSupelec, Rennes 33 SFB-modulated preambles (magnitudes squared) Sample no. • M=256, K=4 • OQPSK Interf. from data part
34. Optimal preambles (1)  Preamble optimization: Minimize MSE subject to transmit power/energy constraint  For low frequency selective channels:  FBMC/OQAM  Block-type: equal pilot tones  Comb-type: equispaced & equipowered  OFDM/QAM (no account for CP energy):  Block-type: DFT matrix column  Comb-type: equispaced & equipowered 31 Aug. 2015 CentraleSupelec, Rennes 34 • D. Katselis et al., IEEE Trans. SP, May 2010. • E. Kofidis et al., Signal Processing, July 2013. • C. Mavrokefalidis et al., EURASIP JASP, May 2014 (for relaying networks).
35. 31 Aug. 2015 CentraleSupelec, Rennes 35 Highly frequency selective channels  No simplifying assumptions: D. Kong et al., IEEE TSP, Jan. 2014 E. Kofidis, ICASSP-2014.
36. 31 Aug. 2015 CentraleSupelec, Rennes 36 Optimal preambles (2)  Optimization problem:  Problem structure: E. Kofidis, ICASSP-2014
37. 31 Aug. 2015 CentraleSupelec, Rennes 37 Optimal preambles (3)  Block-type preamble:  Complex-valued:  Real-valued:  Simple estimation procedure (for real preamble):  Take the first terms of  Divide them by
38. 31 Aug. 2015 CentraleSupelec, Rennes 38 Optimal preambles (4)  Comb-type preamble ( pilot tones):  Equipowered and equispaced  Estimation procedure:  Prototype filter autocorrelation:  Compute the “weighted” freq. response first:  Compute the “weighted” impulse response via IFFT and divide by the weights to arrive at the impulse response estimate: E. Kofidis, ISCCSP-2014
39. 31 Aug. 2015 CentraleSupelec, Rennes 39 Simulation example: Block-type error floor ?
40. 31 Aug. 2015 CentraleSupelec, Rennes 40 Simulation example: Comb-type ITU-VehA channel model error floor
41. 31 Aug. 2015 CentraleSupelec, Rennes 41 More and on-going  Preamble-based CE:  POP etc. [1,3]  MIMO case [2,3,4]  Multiuser case [7]  Longer preambles [5,8]  LMMSE channel estimation [10]  Scattered pilot-based CE:  Extend help pilot idea to highly selective channels  Take into account  virtual (edge) subcarriers [6]  interference from data [6] 1. C. Lélé et al., EW-2007. 2. E. Kofidis and D. Katselis, ICSIPA-2011. 3. E. Kofidis et al., Signal Process., July 2013. 4. E. Kofidis, EW-2015. 5. M. Newinger et al., VTC-2013 (Spring). 6. L. Baltar et al., EUSIPCO-2014. 7. F. Rottenberg et al., ISWCS-2015. 8. E. Kofidis, ISWCS-2015. 9. EMPhAtiC deliverable D3.1 10. L. Caro et al., VTC-2015 (Spring). …
42. Thank you! Questions?