Comparative study of 5G waveformcandidates for below 6GHz air interface
Dimitri Ktenas
R. Gerzaguet & D. Ktenas & N. Cassiau & J-B. Dore
CEA-Leti
01/28/2016
Table of Contents
1. Context and issues2. Waveforms2.1 CP-OFDM & SC-FDMA2.2 Filter bank multicarrier (FBMC)2.3 Universal Filtered Multicarrier (UFMC)2.4 Generalized Frequency Division Multiplexing
(GFDM)3. Comparizon between waveforms4. Multi-user access scenario5. Conclusion
5G ContextI 4G massively rolled out but will soon reaches its limits
I RAN 5G Workshop 09/15 : New non backward compatible Radio AccessTechnology as part of 5G
I Aggregation of non contiguous network is consideredI Spectrum agility : need to study alternative multicarrier waveforms
I Sporadic access & MTCI Strong traffic overhead (fast dormancy)I Massive number of devices : Use relaxed synchronism
I Several candidates have been independently introduced in the past few yearsI Classic CP-OFDM shows its limits : Spectral efficiency, frequency leakage, need
of tight synchronisationI We propose a comparative study of 5G waveform candidates for below
6GHz air interface
ETSI presentation Dimitri Ktenas 01/28/2016 3 - 16
Context and objectives
Considering several candidates for 5G physical layerI Baseline for comparison : OFDM and SC-FDMAI Filter bank multicarrier (FBMC) [3]I Universal Filtered Multicarrier : UFMC (or UF-OFDM) [9]I Generalized Frequency Division Multiplexing (GFDM) [6]
A fair comparison between waveforms in literature is lacking
Considering several metrics for comparisonI Spectral Efficiency (SE)I Peak To Average Power Ratio (PAPR)I Power spectral Density (PSD)
Also consider the asynchronous multi-user scenario [1, 11]
ETSI presentation Dimitri Ktenas 01/28/2016 4 - 16
CP-OFDM & SC-FDMA
Modulateddata
stream
Serialto
parallel
DFT& pilot
insertionIFFT
Parallelto
Serial
CPInsertion
ak
x(n)
Basebandto RF Channel
RF tobaseband
CPremoval
Serialto
ParallelFFT
CHEST&
Equa.IDFT
Parallelto
serial
modulateddecodedstream
OFDM & SC-FDMA (additional stages in dash) transceiver scheme
I Multicarrier modulations, serves as physical layers for 3GPP-LTE or802.11.a/g/n
I Efficient implementation (IFFT/FFT), simple equalization schemesI Spectral efficiency loss due to Cyclic Prefix (CP) insertion to handle
multipath channelI For SC-FDMA : DFT/IDFT precoding stages to reduce PAPR (3GPP-LTE
uplink : DFT-spread OFDM)
ETSI presentation Dimitri Ktenas 01/28/2016 5 - 16
Filter bank multicarrier (FBMC)I Set of parallel data through bank of modulated filters
I Good spectral location,orthogonality and spectralefficiency kept with OQAMmodulation
I Prototype filter in frequencydomain (FS) [2]
I Overlapping factor KI Filter defined in frequency
domain (K=4)H0 = 1H1 = H−1 = 0.971960
H2 = H−2 =√
22
H3 = H−3 =√
1−H21
Modulateddata stream
S/P
OQAMmodu-lation
Spreading –Frequency
filteringIFFT Overlap
and sum
FBMC Tx stage
CP-OFDM (top) and FBMC(bottom) frames
ETSI presentation Dimitri Ktenas 01/28/2016 6 - 16
Universal Filtered Multicarrier (UFMC)
I Derivative of OFDM wherea group of subcarriers (RB)is filtered with aDolph-Chebyshev filter withlength L and attenuationfactor [9]
I B subbands are generatedand combined
I On Rx side, Zero padding isapplied before a 2N FFT
I Possibility to add awindowing process on theRx side (asynchronousmulti-user scenario)
+xk
Modulateddata
stream
Serialto
parallel
IDFTspreader& PS
FilterF1k withlength L
a1kx1k
Modulateddata
stream
Serialto
parallel
IDFTspreader& PS
FilterF2k withlength L
a2kx2k···
Modulateddata
stream
Serialto
parallel
IDFTspreader& PS
FilterFBk withlength L
aBkxBk
Basebandto RF
Channel
RF tobaseband
Windowing& Serial
to parallel2 NFFT
pointsFFT
···Zero padding
FrequencyDomain
processing(subcarrier
equ.)
······
UFMC transceiver
ETSI presentation Dimitri Ktenas 01/28/2016 7 - 16
Generalized Frequency Division Multiplexing (GFDM)
I Based on time-frequencyfiltering of data blocks of sizeP ×M
I Shaping filter : Root RaisedCosine filter (RRC)
I Non orthogonal waveform :interference in time andfrequency domains
I A CP is added at eachsymbols (P subsymbols)
I Possibility to add awindowing process to reducethe ACL
I Parametrized by P,M androll-off factor α
Modulateddatastream
Reshapingin blockK × M
TimeFrequencyfiltering
Parallelto
Serial
CPInsertion& win-dowingak
x(n)
Basebandto RF
Channel
RF tobaseband
CPremoval
Serialto
Parallel
Rx stageZF - MF
Parallelto
serial
modulateddecodedstream
ak
GFDM transceiver
I On Rx side, different architectures :MF, ZF, MMSE [7]
I With MF, need to add InterferenceCancellation (IC) scheme
I With ZF, no self-interference butnoise enhancement
ETSI presentation Dimitri Ktenas 01/28/2016 8 - 16
Simulation parameters & Spectral EfficiencyOverall parameters
FFT size NF F T 1024Bit per Symbol m 2
Resource block size NRB 12
Number of active RBs N1Re 3 for User 1
N2Re 9 for User 2
Sampling frequency Fe 15.36 MHzOFDM and SC-FDMA parametersCyclic prefix NCP 72 samples
UFMC parametersFilter length L 73
Stop band attenuation 40 dBGFDM parameters
Number of subsymbols P 15FFT size M 1024
Roll Off factor α 0.1FBMC parameters
Spreading factor K 4Asynchronous access parameters
Guard carriers [0, 1, 2, 5]Timing Offset [-0.25 :0.25]
Carrier Frequency Offset 0 ; 10%
I We consider 2 users forasynchronous multi-user accessscheme [1]
I 3 RBs for user 1 (12 carriers)I 9 RBs for user 2 (36 carriers)
I Same FFT size for all users : 1024
I Parameters are based on LTE10MHz
I Length of UFMC filter has been setto have same Spectral Efficiency forUFMC and OFDM : L = NCP + 1
ETSI presentation Dimitri Ktenas 01/28/2016 9 - 16
Power Spectral Density
Power Spectral Density of wave-forms :
I OFDM : high ACL due tosinc in freq. domain
I UFMC has lower ACL thanGFDM (circular convolution)
I GFDM with windowing :Better OOB than GFDM andcomparable to UFMC
I Best frequency location isobtained with FBMC
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2−80
−70
−60
−50
−40
−30
−20
−10
0
Frequency [MHz]
Power
Spec
tral
dens
ity[d
Bc/
Hz]
OFDMUFMCFBMCwGFDMGFDM
Power spectral density of waveforms
I 2 users of 3 RBs with 1 RB of guardcarriers to better stress ACL impact
ETSI presentation Dimitri Ktenas 01/28/2016 10 - 16
Spectral Efficiency
Spectral Efficiency for eachwaveform [bit/s/Hz]
I ηOF DM = m×NF F TNF F T +NCP
I ηUF MC = m×NF F TNF F T +L−1
I ηGF DM = m×P×MP×M+NCP
I ηF BMC = m×SS+K− 1
2
UFMC and OFDM have sameSEGFDM SE depends on sizeblock
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 300.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Duration of a burst [ms]
Spectral
efficiency
[b/s/H
z]form=
2
Spectral Efficiency for 5G candidates
OFDMFBMCGFDMUFMC
Comparaison between SE of waveforms
FBMC SE depends on burst duration : If burstduration > 3ms, better SE than UFMC andOFDM
ETSI presentation Dimitri Ktenas 01/28/2016 11 - 16
Peak to Average Ratio
PAPR computed on a 3ms burst :
I PAPR = max[|y[k]|2]E[|y[k]|2]
I We computeComplementary CumulativeDensity ProbabilityFunction (CCDF)
I Low PAPR only obtainedwith SC-FDMA
I All multicarrier modulationshave a comparable PAPR(gap around ∼ 0.5 dB)
3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 10.5 1110−3
10−2
10−1
100
PAPR [dB]
CCD
Fof
PAPR
PAPR for 5G waveforms : Burst = 3 ms
OFDMUFMCSC-FDMAFBMCGFDM
PAPR measured on 3ms burst
ETSI presentation Dimitri Ktenas 01/28/2016 12 - 16
Multi-user access scenarioComparison in a multi-user asynchronous access scenario between 2 users [1]
I First user is perfectly synchronised and second user interferes with the firstone (due to time delay error and CFO)
I Performance measured in terms of Mean Squared Error (MSE), with differentnumber of guard carriers (0, 1, 2 and 5)
Time
Frequency
User 1 : per-fectly sync
User 2 : coarsesync : CFO andtiming offset
Timing OffsetCFO
Frequency
PSD User 1 GC User 2
I Several 5G candidates with specific parametrisation (best case) :1. CP-OFDM (SC-FDMA has the same MSE)2. UFMC with windowing approach [10]3. GFDM with windowing [8] ; with MF receiver and IC [4]4. And FBMC
ETSI presentation Dimitri Ktenas 01/28/2016 13 - 16
Multi-user access scenario : No CFOI 0 < Delay error NCP : no
interference for OFDM
I No GC, GFDM withwindowing has betterperformance
I No GC, Small delay value :UFMC with windowing hasgood performance
I wGFDM > wUFMC if atleast one GC
I At least one GC inserted :FBMC has the bestperformance : no interference(Phydyas filter + OQAM[5]) !
−0.2 −0.1 0 0.1 0.2
−60
−40
−20
Time delay error (n/nFFT)
MSE
[dB
]
Guard carrier = 0
−0.2 −0.1 0 0.1 0.2
−60
−40
−20
Time delay error (n/nFFT)
MSE
[dB
]
Guard carrier = 1
−0.2 −0.1 0 0.1 0.2
−60
−40
−20
Time delay error (n/nFFT)
MSE
[dB
]
Guard carrier = 5
OFDMFBMCwUFMCwGFDM
−0.2 −0.1 0 0.1 0.2
−60
−40
−20
Time delay error (n/nFFT)
MSE
[dB
]
Guard carrier = 2
ETSI presentation Dimitri Ktenas 01/28/2016 14 - 16
Multi-user access scenario : 10% CFOI CFO breaks OFDM
orthogonality and lowersperformance for all waveforms
I No GC, wGFDM has sameperformance as FBMC
I No GC, Small delay value :UFMC with windowing hasthe best performance
I wGFDM > wUFMC if atleast one GC is inserted butimpact of CFO
I At least one GC inserted :FBMC has the bestperformance : no interference(Phydyas filter + OQAM) !
−0.2 −0.1 0 0.1 0.2
−60
−40
−20
Time delay error (n/nFFT)
MSE
[dB
]
Guard carrier = 0
−0.2 −0.1 0 0.1 0.2
−60
−40
−20
Time delay error (n/nFFT)
MSE
[dB
]
Guard carrier = 1
−0.2 −0.1 0 0.1 0.2
−60
−40
−20
Time delay error (n/nFFT)
MSE
[dB
]
Guard carrier = 5
OFDMFBMCwUFMCwGFDM
−0.2 −0.1 0 0.1 0.2
−60
−40
−20
Time delay error (n/nFFT)
MSE
[dB
]
Guard carrier = 2
ETSI presentation Dimitri Ktenas 01/28/2016 15 - 16
ConclusionFair comparison for several repre-sentative criteria :
I Spectral Efficiency, PAPR,PSD comparison
I Mean Square Error inmulti-user access scenario
Comparison between 5G wave-form candidates that outperformCP-OFDM :
I UFMC offers LTE backwardcompatibility
I GFDM and FBMC gofurtherBUT still open questions :short packet, MIMO, . . .
0
0,5
1
1,5
2
2,5
3
3,5
4
4,5
5
Spectral Efficiency
Spectral Efficiency ShortPacket
ACLR
PAPRMUAC (no GC)
MUAC (1 GC)
Complexity
CP-OFDM SC-FDMA UFMC GFDM FBMC
Comparison between waveforms
ETSI presentation Dimitri Ktenas 01/28/2016 16 - 16
References I
[1] 5G-Now. Deliverable D3.2 : 5G waveform candidate selection. Technical report, 5G Now(5th Generation Non-Orthogonal Waveforms for Asynchronous Signalling), 2015.
[2] M. Bellanger. FS-FBMC : An alternative scheme for filter bank based multicarriertransmission. In 5th International Symposium on Communications Control and SignalProcessing (ISCCSP), pages 1–4, May 2012.
[3] M. Bellanger and al. FBMC physical layer : a primer, 06 2010.
[4] R. Datta, N. Michailow, M. Lentmaier, and G. Fettweis. GFDM interference cancellation forflexible cognitive radio PHY design. In Proc. IEEE Vehicular Technology Conference (VTCFall), pages 1–5, Sept 2012.
[5] J.-B. Dore, V. Berg, N. Cassiau, and D. Ktenas. FBMC receiver for multi-user asynchronoustransmission on fragmented spectrum. EURASIP Journal on Advances in Signal Processing,2014(1) :41, 2014.
[6] G. Fettweis, M. Krondorf, and S. Bittner. GFDM - generalized frequency divisionmultiplexing. In Proc. IEEE 69th Vehicular Technology Conference (VTC), pages 1–4, April2009.
ETSI presentation Dimitri Ktenas 01/28/2016 17 - 16
References II
[7] N. Michailow, S. Krone, M. Lentmaier, and G. Fettweis. Bit error rate performance ofgeneralized frequency division multiplexing. In Proc. IEEE Vehicular Technology Conference(VTC Fall), pages 1–5, Sept 2012.
[8] N. Michailow, M. Matthe, I. Gaspar, A. Caldevilla, L. Mendes, A. Festag, and G. Fettweis.Generalized frequency division multiplexing for 5th generation cellular networks. IEEETransactions on communications,, 62(9) :3045–3061, Sept 2014.
[9] V. Vakilian, T. Wild, F. Schaich, S. ten Brink, and J.-F. Frigon. Universal-filteredmulti-carrier technique for wireless systems beyond LTE. In Proc. IEEE GlobecomWorkshops (GC Wkshps), pages 223–228, Dec 2013.
[10] T. Wild, F. Schaich, and Y. Chen. 5G air interface design based on universal filtered(UF-)OFDM. In 19th International Conference on Digital Signal Processing (DSP), pages699–704, Aug 2014.
[11] G. Wunder, P. Jung, M. Kasparick, T. Wild, F. Schaich, Y. Chen, S. Brink, I. Gaspar,N. Michailow, A. Festag, L. Mendes, N. Cassiau, D. Ktenas, M. Dryjanski, S. Pietrzyk,B. Eged, P. Vago, and F. Wiedmann. 5GNOW : non-orthogonal, asynchronous waveforms forfuture mobile applications. Communications Magazine, IEEE, 52(2) :97–105, February 2014.
ETSI presentation Dimitri Ktenas 01/28/2016 18 - 16