2015 08-31 kofidis

Channel Estimation in
Filter Bank-based Multicarrier Systems:
Fundamentals and Recent Advances
Eleftherios Kofidis
Computer Technology Institute, Greece
University of Piraeus, Greece

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)
 …

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
 …

FBMC research and applications
Filter bank-based multi-carrier modulation:
• FBMC/OQAM
• FMT
• GFDM
• UFMC
• …

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

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

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

Fundamentals of FBMC/OQAM

FBMC/OQAM vs. OFDM/QAM
 1 2F 
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 2F 
 1/ 2F 
Phase space

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

FBMC/OQAM Transmitter
IFFT
2
0 ( )A z
2
1( )A z
2
1( )MA z
2M
2M
2M


1
z
1
z
  

0,n

0,n

1,n

1,M n 

1,M n 

1,n
0,nd
1,nd
1,M nd 
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.

FBMC/OQAM Receiver
2
0 ( )B z
2
1( )B z
2
1( )MB z
FFT
1
z
1
z
2M
2M
2M
  



Subchannel
processing
Subchannel
processing
Subchannel
processing
*
0,n
*
1,n
*
1,M n 

*
0,n

*
1,n
Re

*
1,M n 
0,nd
1,nd
1,M nd 
Re
Re
 
R2C
R2C
R2C
S/P
conversion
Polyphase
filtering
Transform block OQAM demodulation
AFB:

System model

System model (1)
 M: #subcarriers
 K: overlapping factor
 g: prototype filter (length )
C2R SFB h + AFB
Intrinsic interference:

13 March 2014 Patras (ENDECON) 15
Intrinsic interference in FBMC/OQAM
1,1,11,1
1,1,11,1
1,,1,
1,1,11,1
1,0,01,0





nMnMnM
nknknk
nknknk
nknknk
nnn
ddd
ddd
ddd
ddd
ddd



With good TF localization,
contributions to intrinsic
interference only come from the
first-order neighboring TF
points
1,1,11,1
1,1,11,1
1,,1,
1,1,11,1
1,0,01,0





nMnMnM
nknknk
nknknk
nknknk
nnn
ddd
ddd
ddd
ddd
ddd



With good TF localization,
contributions to intrinsic
interference only come from the
first-order neighboring TF
points




 

 
   
1,1,11,1
1,1,11,1
1,,1,
1,1,11,1
1,0,01,0





nMnMnM
nknknk
nknknk
nknknk
nnn
ddd
ddd
ddd
ddd
ddd



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 nd 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).

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

System model (2)
 Common assumptions (locally freq./time-invariant channel):

System model (2)

System model (2)
OFDM-like

System model (2)
OFDM-like
colored
virtual Tx symbol
(pseudo-symbol)

FBMC/OQAM channel estimation:
Fundamentals

Preamble-based channel estimation (1)
Control / Data
Preamble
Frame: SFB
non-zeropart
0 0
prevents interference
from previous frame
(often unnecessary!)
prevents interference
from control/data
channel time invariant

Preamble-based channel estimation (2)
Control/Data
Preamble
Full
(block-type):
Control/Data
Sparse
(comb-type):
0
0
protect from ICI

Scattered pilot-based channel estimation
 Help (auxiliary) pilot
J.-P. Javaudin, D. Lacroix, and A. Rouxel, VTC-2003 (Spring).
1,1,11,1
1,1,11,1
1,,1,
1,1,11,1
1,0,01,0





nMnMnM
nknknk
nknknk
nknknk
nnn
ddd
ddd
ddd
ddd
ddd



Preamble-based channel estimation

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):
0,0 0,1 0,2
1,0 1,1 1,2
2,0 2,1 2,2
1,0 1,1 1,2M 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

Example: IAM-R
 Null side symbols ( base design on middle
symbol only)
 Carefully choose signs so as to maximize
pseudo-pilots’ magnitude
010
010
010
010
010
010
010
010




31
d
d
d
Idea:
Example:
M=8, OQPSK

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!
• C. Lélé et al., ICC-2008.
• J. Du and S. Signell, ICC-2009.
• PHYDYAS deliverable D3.1
• E. Kofidis and D. Katselis, EUSIPCO-2011.
00
00
00
00
00
00
00
00
0
0
1
1
1
0
0
0
jd
d
d
jd
d
d
jd
d




00
010
00
010
00
010
00
010
j
j
j
j




11
1
11
1
11
1
11
1








j
jj
j
jj
j
jj
j
jj
IAM-I IAM-C E-IAM-C
1/3 of the subcarriers:

Price for good performance: high PAPR!
SFB-modulated preambles (magnitudes squared)
Sample no.
• M=256, K=4
• OQPSK
Interf. from data part

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
• 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).

Highly frequency selective channels
 No simplifying assumptions:
D. Kong et al., IEEE TSP, Jan. 2014
E. Kofidis, ICASSP-2014.

 Optimization problem:
 Problem structure:
E. Kofidis, ICASSP-2014

 Block-type preamble:
 Complex-valued:
 Real-valued:
 Simple estimation procedure (for real preamble):
 Take the first terms of
 Divide them by

 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

Simulation example: Block-type
error floor
?

Simulation example: Comb-type
ITU-VehA channel model error floor

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).
…

2015 08-31 kofidis

More Related Content

What's hot (20)

Similar to 2015 08-31 kofidis (20)

More from SCEE Team (9)

Recently uploaded (20)

2015 08-31 kofidis