Page 1
1
POPS-OFDM:
Ping-Pong Optimized Pulse Shaping OFDM for
5G Cellular Systems and Beyond
Mohamed Siala
MEDIATRON Laboratory
Higher School of Communication of Tunis (SUP’COM)
The 2015 International Symposium on Networks, Computers and
Communications (ISNCC 2015)
May 13-15 2015, Yasmine Hammamet, Tunisia
Page 2
Outline
Motivation of Research Activities on Pulse Shaping for 5G
OFDM/OFDMA
Background on Multi-Carrier Systems
5G Challenges and Requirements
POPS-OFDM to Systematically Respond to 5G Radio Interface
Challenges
Conclusion and Perspectives for Future Research Work on 5G
2
Page 3
Motivation of Research Activities on Pulse
Shaping for 5G OFDM/OFDMA
3
Page 4
5G Challenges and Requirements 1/5
45GNOW
Page 5
5G Challenges and Requirements 2/5
55GNOW
Page 6
5G Challenges and Requirements 3/5
65GNOW
Page 7
5G Challenges and Requirements 4/5
75GNOW
Page 8
5G Challenges and Requirements 5/5
The main drivers for 5G are:
Tactile Internet: The human tactile sense distinguishes latencies in the order of 1ms 1ms round trip time requires a time budget on
PHY of maximum 100 µs.
Internet of Things (IoT): A scalability problem (>100k MTC nodes
in a cell) under cost, coverage, energy (life time) and privacy
constraints.
Gigabit Wireless Connectivity: Quick downloads (streaming content with data rates in the order of ~100 Mbit/s) Download
times in the order of ~ 10 Gbit/s.
Fragmented Spectrum and the Spectrum Paradox: Spectrum is
scarce and expensive but underutilized. With White Spaces
Communication, a 100x better localization is expected.
8
Page 9
4G (LTE-A) Pitfalls
LTE is tailored to maximize performance by enforcing strict synchronism and perfect orthogonality
Machine-type communication (MTC) requires bulky procedures to
ensure strict synchronism
Collaborative schemes (e.g. CoMP) use tremendous efforts to
collect gains under strict synchronism and orthogonality
Digital Agenda/Carrier aggregation forces systems to deal with
fragmented spectrum
9
Page 10
Need for Non-Orthogonal Waveforms
Non-orthogonal waveforms on the physical layer will enable:
Asynchronous MTC traffic with drastically reduced signalling and
increased life time
The provision of asynchronous coordinated multi-point (CoMP) /
Heterogeneous Networking (HetNet)
Implementation of asynchronous carrier aggregation concepts with
well frequency localization
A (filtered) multicarrier approach will enable:
The mix of synchronous / asynchronous and orthogonal / non-
orthogonal traffic types
The aggregation of non-contiguous spectrum thanks to low out-of-
band emissions of the non-orthogonal waveforms
10
Page 11
Workload of Current Mobiles
11
Outer receiver consists of channel decoder and de-interleaver
Page 12
Projects on 5G
From 2007 to 2013, the European Union set aside €700 million of
funding (FP7) for research on future networks, half of which was
reserved for wireless technologies and the development of 4G and
beyond-4G technologies.
METIS, 5GNOW, iJOIN, TROPIC, Mobile Cloud Networking,
COMBO, MOTO and PHYLAWS are some of the latest EU research
projects that address the architecture and functionality needs of 5G
networks, representing some €50 million EU investment.
European Union’s FP7 projects, PHYDYAS (Duration: 30 months,
Start: January 2008, End: October 2010, Total Cost: 4 093 483€),
investigated Filter Bank Multi-Carrier (FBMC) and corresponding
transceiver functionalities.
12
Page 13
5GNOW Candidate Waveforms
European Union’s FP7 projects, 5GNOW (5th Generation Non-
Orthogonal Waveforms for Asynchronous Signaling), (Start:
September 2012, End: February 2015, Total Cost: 3 526 991 €),
investigated 4 candidate waveforms:
Generalized frequency division multiplexing (GFDM)
Universal Filtered Multicarrier (UFMC): UFMC applies filtering
to subsets of the complete band instead of single subcarriers
(GFDM) or the complete band (Filtered OFDM)
Filter Bank Multi-Carrier (FBMC)
Bi-orthogonal Frequency Division Multiplexing (BFDM)
13
Page 14
GFDM: Generalized Frequency Division
Multiplexing
14
Page 15
UFMC: Universal Filtered MultiCarrier
15
Spectral behavior within a single sub-band
Single PRB compared to OFDM
Page 16
Background on Multi-Carrier Systems
16
Page 17
17
History of OFDM
Late 50’s: Concept of multicarrier without overlapping in frequency
Late 60’s: Orthogonal multicarrier [Chang66, Salzberg67]
Early 70’s: Use of the Fast Fourier Transform (FFT) [Weinstein &
Ebert71] - Concept of Guard Interval (GI)
Early 80’s: Concept of Cyclic Prefix (CP) [Peled & Ruiz80]
Early 90’s: DAB standardization
Late 90’s: Standardization of ADSL, DBV-T and WIFI
Early and mid 00’s: Standardization of WiMAX, DVB-H, PLC and
LTE
Page 18
18
Applications of OFDM
Wireless applications :
Broadcasting for digital terrestrial television (DVB- T, DVB- H)
Digital Audio Broadcasting (DAB) and Digital Radio Mondiale
(DRM)
802.11a wireless networks (WIFI5) , 802.16 (WiMAX) and
HiperLAN/2
New generation radio mobile networks (LTE, LTE-A)
Wireline applications:
ADSL
PLC (Power-Line Communications)
Page 19
Classification of Multi-Carrier Systems
19
Multi-carrier systems
Filter Lattice Symbol
Ort
hogonal
Hex
agonal
Rec
tangu
lar
Non-o
rthogonal
Rea
l
OF
DM
w/o
GI
OF
DM
/OQ
AM
OF
DM
w/
ZP
Pulse
Mult
i-puls
eU
FM
C
Bio
rthogonal
Com
ple
x
OF
DM
w/
CP
OF
DM
/QA
M
Mono-p
uls
eO
FD
MOQAM: Offset Quadrature Amplitude Modulation
CP: Cyclic Prefix, ZP: Zero Padding, GI: Guard Interval
TimeC
onti
nuous
Dis
cret
e
Quin
cunx
Sta
gger
ed
Page 20
20
Pulse Shaping for OFDM without Guard Interval
Rectangular pulse shaping filter with duration T :
1/T
t
[0, [( ) rect ( ) /Tt t T
1
ˆ( ) sinc( )f fT
1/T
T
… …f
Fourier Transform
Page 21
21
OFDM without Guard Interval
t
f
nT ( 1)n T0 0f
/mf m T
1 ( 1) /Nf N T
… …
0 ( )n t
1, ( )N n t
tnT ( 1)n T
… …
( 1)n T ( 2)n T
OFDM symbol between nT and (n+1)T
Modulated Signal ee(t)
Orthogonal
sinusoidal
functions
( )mn t
Page 22
Time-Frequency Lattice Layout: OFDM without
Guard Interval (GI)
22
Frequency
Time
1/ uF T
uT TArea (1/ ) 1u uFT T T
Lattice density 1/ 1FT Critical density
0 & π/2 0 & π/20 & π/20 & π/2
0 & π/2 0 & π/20 & π/20 & π/2
0 & π/2 0 & π/20 & π/20 & π/2
0 & π/2 0 & π/20 & π/20 & π/2
Inphase and quadrature
components
Page 23
23
Power Spectral Density of OFDM without Guard
Interval 1/2
Subcarrier spacing : f = 1/T
1/f T
f
Page 24
Power Spectral Density of OFDM without Guard
Interval 2/2
24
16N
0gT
32N
64N 128N
Page 25
25
No Interference in the Gaussian Channel –
Perfect Orthogonality
Modulated signal:
Symbol amn transported by function ( ):
Absence of Inter-Symbol Interference (ISI) and Inter-Carrier Interference (ICI) equivalent to orthogonality conditions:
Functions {mn(t)} form un orthonormal base of the space of modulated signals
1
0( ) ( )
N
e mn mnn me t a t
( ) ( ) exp( 2 )mn mt t nT j f t
*( ), ( ) ( ) ( )kl mn kl mn km lnt t t t dt
/mf m T
Page 26
Balian-Low Theorem for a Time-Frequency
Critical Lattice Density
26
2 22 2 ˆ( ) ( )t t dt or f f df
Fourier transform of ( )t
Page 27
27
Behavior of an OFDM System without GI in the
Presence of Time Dispersive Channel
t
f
nT ( 1)n T0 0f
/mf m T
1 ( 1) /Nf N T
… …
tnT ( 1)n T
… …
( 1)n T ( 2)n T
Noiseless received
signal
( )c
Channel
Inter-Symbol Interference
(ISI)
Loss
of orthogonality!
mT0
0 ( )n t
1, ( )N n t
( )mn t
Page 28
28
Interference Suppression: Guard Interval Insertion with
Zero Padding (ZP)
t
f
… …
tnT ( 1)n T
… …
( 1)n T
1, ( )N n t
( )mn t
0 ( )n t
T
gTuT
( )c Channel
mT0
g mT T
Guard Interval
Zero Padding
Page 29
29
Interference Suppression: Guard Interval Insertion with
Zero Padding (ZP)
t
… …
No Inter-Symbol
Interference (ISI)
Persistent loss of
orthogonality!
t
f
… …
g mT TT
Page 30
30
Interference Suppression: Guard Interval Insertion with
Cyclic Prefix (CP)
t
… …
tnT ( 1)n T
… …
( 1)n T
0 ( )n t
T
gT uT
( )c Channel
mT0
f
Cyclic Prefix
1, ( )N n t
( )mn t
nT ( 1)n T( 1)n T
Page 31
31
Interference Suppression: Guard Interval Insertion with
Cyclic Prefix (CP)
t
… …
tnT ( 1)n T
… …
( 1)n T
T
gT uT
f
Overlapping restricted
to the Guard Interval
0 ( )n t
1, ( )N n t
( )mn t
Page 32
32
Interference Suppression: Guard Interval Insertion with
Cyclic Prefix (CP)
t
… …
t
… …
fRecovered
orthogonality
t
… …
( 1)n T( 1)n T
Cyclic Prefix Suppression
1, ( )N n t
( )mn t
0 ( )n t
nT
Page 33
Power Spectrum Density of Conventional OFDM
33
16N
/ 4g uT T
32N
64N 128N
Page 34
Time-Frequency Lattice Layout: OFDM with
Cyclic Prefix (CP)
34
Time
1/ uF T
u gT T T Area (1/ )( ) 1 / 1u u g g uFT T T T T T
Lattice density 1/ 1FT
Frequency
uTgT
Cyclic Prefix Useful part
Page 35
35
OFDM/OQAM with Square Lattice
The used shaping filter filtre is generally a root-raised cosine filter with
roll-off (typically equal to 1)
OFDM with Offset QAM (OQAM) alternately transmit phase and
quadrature
t
1/f T QAM
Symbol
f
T/ 2T
: In-phase component : Quadrature component
Page 36
Time-Frequency Lattice Layout: OFDM/OQAM
with Square Lattice
36
Frequency
Time
1/ uF T
/ 2uT TSurface (1/ )( / 2) 1/ 2u uFT T T
Lattice density 1/ 2FT Critical density
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
π/2
Quadrature componentInphase component
Page 37
Time-Frequency Lattice Layout: OFDM/OQAM
with Hexagonal Lattice
37
Frequency
Time
3π/4
π/2 0
π/4 3π/4
π/2 0
π/43π/4
π/2 0
π/4
3π/4
π/2 0
π/4 3π/4
π/2 0
π/43π/4
π/2 0
π/4
Lattice density 1 Critical density
[1] M. Siala, “Novel OFDM/OQAM system with hexagonal time-frequency lattice,” Third International Symposium on
Image/Video Communications over fixed and mobile networks (ISIVC’06), Hammamet, Tunisia, September 2006.
[2] M. Siala and A. Yongaçoglu, “Prototype waveform optimization for an OFDM/OQAM system with hexagonal
time-frequency lattice structure,” 9th International Symposium on Signal Processing and its Applications (ISSPA’07),
Sharjah, United Arab Emirates, February 2007.
Page 38
Waveforms for OFDM/OQAM
38
Linear decreasing in the
logarithmic scale
Exponential decrease in
time and frequency
Gaussian waveform
Hexagonal lattice waveform
[1] M. Siala, “Novel OFDM/OQAM system with hexagonal time-frequency lattice,” Third International Symposium
on Image/Video Communications over fixed and mobile networks (ISIVC’06), Hammamet, Tunisia, September 2006.
Page 39
Small-Scale Propagation: Multipath Rayleigh
Fading 1/2
39
Time
F
T
Frequency
Doppler shift
Time delay
minmin Df
max Df
max
: Doppler spread
DB
mT
: Delay spread
DB
mT
Scattering function
Page 40
Small-Scale Propagation: Multipath Rayleigh
Fading 2/2
40
Time
F
T
Frequency
Doppler shift
Time delay
: Channel spread
DB
mT
D mB T
(Diffuse) Scattering function
ISIISI
ICI
ICI
ISI: Inter-Symbol Interference
ICI: Inter-Carrier Interference
Page 41
Doppler Spread-Delay Spread Balancing 1/3
41
Time
F
T
Frequency
Doppler shift
Time delayDB
mT
Reduction in F & Increase in T
Substantial increase in ICI Global increase in ICI+ISI
ISIISI
ICI
ICI
Page 42
Doppler Spread-Delay Spread Balancing 2/3
42
Time
F
T
Frequency
Doppler shift
Time delayDB
mT
ISI
ICI
ICI
ISI
Reduction in T & Increase in F
Substantial increase in ISI Global increase in ICI+ISI
Page 43
Doppler Spread-Delay Spread Balancing 3/3
43
Time
F
T
Frequency
Doppler shift
Time delayDB
mT
ISIISI
ICI
ICI
Good balancing between T and F
Global reduction in ICI+ISI
mDTB
F T
Page 44
5G Challenges and Requirements
44
Page 45
Requirements for 5G: Coordinated MultiPoint
(CoMP)
Joint Processing (JP):
Coordination between multiple BSs
MSs are simultaneously transmitting or receiving to or from
multiple BSs
Coordinated Scheduling/Coordinated Beamforming (CS/CB):
Coordination between multiple BSs
MSs are transmitting or receiving to or from a single transmission
or reception BS
45
Page 46
Requirements for 5G: Coordinated MultiPoint
(CoMP) – Overlapping in Time
46
timeAt the BSs
MS2MS1
time
TDOA Overlapping in time
Artificial delay spread
Inter-Symbol Interference
At MS2
time
At MS1TDOA: Time Difference of Arrival
Applicable even for fully
time synchronous BSs
Page 47
Requirements for 5G: Coordinated MultiPoint
(CoMP) – Overlapping in Frequency
47
MS
frequency
Carrier Frequency Offset Overlapping in frequency Artificial Doppler spread
Inter-Carrier Interference (ICI)
At MS
From BS1
frequency
frequency
From BS2
From BS3Applicable only for not fully
frequency Synchronous BSs
Page 48
Requirements for 4G, 5G and DVB-T: MBMS
and SFN
48Overlapping replicas Artificial delay spread Interference
time
At the BSs/DVB-T TV Station
time
At the TV Set
(SFN)
At the MS
(MBMS)
SFN: Single Frequency
Network
MBMS: Multimedia Broadcast
Multicast Service
Page 49
Requirements for 5G: Sporadic Traffic and Fast
Dormancy 1/4
2, 3 and 4G systems continuously transmit reference signals and
broadcast system information that is used by terminals as they move
across cells
The more signaling the cellular standard requires the more complex
and power-hungry will be the devices
With denser deployment and more network nodes (MTC), such
“always-on” transmissions are not attractive from an interference and
energy consumption perspective
Maximizing the devices’ sleep opportunities, through sporadic
access, can minimize energy consumption, leading to long battery life
49
Page 50
Requirements for 5G: Sporadic Traffic and Fast
Dormancy 2/4
Sporadic access poses a significant challenge to mobile access
networks due to fast dormancy:
Fast dormancy is used to save battery power: The mobile breaks
ties to the network as soon as a data piece is delivered
When the mobile has to deliver more pieces of data it will always
go through the complete synchronization procedure again
This can happen several hundred times a day, resulting in
significant control signaling growth and network congestion threat
It is desirable to achieve “zero-overhead” communications by
providing channel access with minimal signaling
50
Page 51
Requirements for 5G: Sporadic Traffic and Fast
Dormancy 3/4
Get rid of closed-loop timing control (which costs energy and
signaling overhead, being undesirable for MTC) and use open loop
timing control mechanisms: The device uses the downlink pilot signals
by the BS for a rough synchronization (RSSI: Received Signal
Strength Indication)
51
Page 52
Requirements for 5G: Sporadic Traffic and Fast
Dormancy 4/4
52Nokia Siemens Networks, Understanding Smartphone Behavior in the Network,
White Paper, 2011, [Available: http://www.nokiasiemensnetworks.com/sites/default/files
Comparisons of Data and Signaling Traffic
Page 53
Requirements for 5G: Sporadic Traffic and Fast
Dormancy – Relaxed Frequency Synchronization
53
MS2
Reduced synchronization overhead Relaxed frequency synchronization
Carrier Frequency Offset Overlapping in frequency
Inter-user interference in frequency
From MS1
frequency
MS1MS3
frequencyFrom MS2
frequencyFrom MS3
At BS
frequency
Inter-user interference
Unaligned carrier frequencies
Page 54
Requirements for 5G: Sporadic Traffic and Fast
Dormancy – Relaxed Time Synchronization
54
MS2
Reduced synchronization overhead Relaxed time synchronization
Overlapping in time
Inter-user interference in time
MS1
From MS1
time
timeFrom MS2
At BS
time
Inter-user interference
Page 55
Requirements for 5G: Asynchronous Signaling in
the Uplink – RACH 1/2
55
MS2MS1
RACH random access
Page 56
Requirements for 5G: Asynchronous Signaling in
the Uplink – RACH 2/2
56
time
No synchronization overhead Strong overlapping in time
Inter-user interference in time
To/from BS
time
To/from MS1
To/from MS2
Inter-Burst interference
time
Synchronization
channel
RACH burst
from MS2
RACH burst
from MS1Propagation
delay to MS1
Propagation
delay to MS2
Page 57
Requirements for 5G: Spectrum Agility and
Carrier Aggregation 1/2
TV White Spaces (TVWS) exploration can represent a new niche
markets if it overcomes, with spectrum agility, the rigorous
implementation requirements of low out of band radiations for
protection of legacy systems
The LTE-A waveform imposes generous guard bands to satisfy
spectral mask requirements which either severely deteriorate spectral
efficiency or even prevent band usage at all
5G will address carrier aggregation by implementing non-orthogonal
waveforms, with low out-of-band emissions, in order not to interfere
with other legacy systems and tight spectral masks
57
Page 58
Requirements for 5G: Spectrum Agility and
Carrier Aggregation 2/2
58
OFDM+CP vs. ESM: Loss of efficient of traditional OFDM with CP to fit in
an ESM (Emission Spectrum Mask) due to its non-negligible side lobes
Page 59
Requirements for 5G: Low Latency 1/2
4G offers latencies of multiple 10 ms between terminal and BS that
originate from resource scheduling, frame processing, retransmission
procedures, and so on.
The access latency offered by LTE is not sufficient for latency-critical
applications, such as tactile internet (motivated by the tactile sense of
the human body, which can distinguish latencies on the order of 1 ms
accuracy), traffic safety and infrastructure protection.
To ensure support for such mission-critical MTC applications, next-
generation wireless access should allow for latencies on the order of 1
ms or less.
59
Page 60
Requirements for 5G: Low Latency 2/2
A 1 ms round-trip time for a typical tactile interaction requires a time
budget of maximum 100 µs on the physical layer
Far shorter than LTE-A allows, missing the target by nearly two
orders of magnitude
Clear motivation for an innovative and disruptive redesign of
the PHY layer
Lower latency over the radio link can be achieved by reducing
transmission-time intervals and widening the bandwidth of radio
resource blocks in which a specific amount of data is transmitted
60
Page 61
Requirements for 5G: Lower Latency vs Doppler
Spread-Delay Spread Balancing 1/2
61
Time
F
T
Frequency
Doppler shift
Time delayDB
mT
Reduced global ICI+ISI Good balancing between T and F
Increased Latency
ICI
ICI
ISIISI
Processing
Time at the Rx2
mTmin
Contribution of the PHY to the latency
Page 62
Requirements for 5G: Lower Latency vs Doppler
Spread-Delay Spread Balancing 2/2
62
Time
F
T
Frequency
Doppler shift
Time
delayDB
mT
Decreased Latency
Bad balancing between
T and F
Increased global
ICI+ISI
ISIISI
Processing
Time at the Rx2
mTmin
Contribution of the PHY to the latency
ICI
ICI
Page 63
POPS-OFDM to Systematically Respond to 5G
Radio Interface Challenges
63
Page 64
POPS-OFDM Categories
64
POPS-OFDM
Continuous DiscreteTime
Optimum
exploration space2 ( ) 2 ( )
Practical
exploration space
1
0Vect({ ( )} )N
k kt
1
0Vect({ ( )} )N
k kt
0{ ( )}k kt
0{ ( )}k kt
: Hermite functions
: Prolate Spheroidal Wave Functions (PSWF)
To be explored next
2 ( )I
Page 65
33φ32φ31φ30φ
23φ22φ21φ20φ
13φ12φ11φ10φ
OFDM Time-Frequency Lattice: Transmitter
Side
Time
Frequency
Signal
00 φ φ 01φ 02φ 03φ
Time Shift by TTime Shift by 2TTime Shift by 3T
Frequency
Shift by F
Frequency
Shift by 2F
Frequency
Shift by 3F
Symbol Period T
=
Symbol Spacing
Symbol Bandwidth F = Subcarrier Spacing
: Transmitter Prototype Waveform (Vector)φ
: (OFDM) Symbol PeriodT
: Subcarrier SpacingFmnφ
Subcarrier Index Symbol Index
Frequency shift of by mF Time shift of by nT65
Page 66
30 30a φ
20 20a φ
10 10a φ
00 00a φ
1
0 0
0
Q
m m
m
a
φ
OFDM Transmitted Signal
Time
Frequency
Signal21 21a φ
11 11a φ
01 01a φ
31 31a φ
1
1 1
0
Q
m m
m
a
φ
1
0
: Sampled Version of the Transmitted OFDM SignalQ
mn mn
n m
a
e φ
1
2 2
0
Q
m m
m
a
φ
32 32a φ
22 22a φ
12 12a φ
02 02a φ
1
3 3
0
Q
m m
m
a
φ
33 33a φ
23 23a φ
13 13a φ
03 03a φ
SubcarriersQ
66
Page 67
Propagation Channel Characteristics: Delay and
Doppler Spreads
Mobile speed
( , )S p
p
dB : Doppler spread
Doppler spread spectrum
: Discrete time delay
: Doppler frequency shift
( , )S p : Channel scattering function
: Discrete time delay spreadmT 67
Page 68
30 30a φ
20 20a φ
10 10a φ
00 00a φ
1
0 0
0
Q
m m
m
a
φ
OFDM Received Signal
Time
Frequency
Signal
1
0
: Sampled Version of the Received OFDM SignalQ
mn mn
n m
a
r φ n
21 21a φ
11 11a φ
01 01a φ
31 31a φ
1
1 1
0
Q
m m
m
a
φ
: Additive White Gaussian Noisen
: Channel distorted version of mn mnφ φ
68
ISIICI
Page 69
Decision variables
: Receiver Prototype Waveform (Vector)ψ
klψ
Subcarrier Index Symbol Index
Frequency shift of by kF Time shift of by lT
H
kl kl ψ r : Decision variable on kla
( , ) ( , )Noise TermUseful Term
Interference Term
H H H
kl kl kl kl mn kl mn kl
m n k l
a a
ψ φ ψ φ ψ n
69
Page 70
Signal-to-Interference and Noise Ratio (SINR)
S
I N
PSINR
P P
: Average power of the Useful Term
: Average power of the Interference Term
: Average power of the Noise Term
S
I
N
P
P
P
( , )
( , )
1
H
S p
H
S p
SINR
SNR
φ
φ
ψ KS ψ
ψ KI I ψ
: Ratio of two definite positive quadratic
forms on for a given
( , )
( , )
1
H
S p
H
S p
SINR
SNR
ψ
ψ
φ KS φ
φ KI I φ
: Ratio of two definite positive quadratic
forms on for a given
0
: Signal to Noise RatioE
SNRN
70
Page 71
Optimization Philosophy
Transmitter Side Receiver Side(0)φ
(0 )
(0)
( , )
( , )
Maximize 1
H
S p
H
S p
SINR
SNR
φ
φ
ψ KS ψ
ψ KI I ψ (0)ψ
(0 )
(0 )
( , )
( , )
Maximize 1
H
S p
H
S p
SINR
SNR
ψ
ψ
φ KS φ
φ KI I φ(1)φ
(1)
(1)
( , )
( , )
Maximize 1
H
S p
H
S p
SINR
SNR
φ
φ
ψ KS ψ
ψ KI I ψ (1)ψ
(1)
(1)
( , )
( , )
Maximize 1
H
S p
H
S p
SINR
SNR
ψ
ψ
φ KS φ
φ KI I φ(2)
φ 71
Page 72
Optimization Philosophy
φ
ψ
(0)φ
(0)ψ
(1)φ
(1)ψ
(2)φ
72SINR
Equal-SINR curves
(Contour plot of SINR)
SINR maximum
Page 73
First Optimization Technique
SINR
0
ψ
( , ) ( , )
1S p S p
SINR φ φ
KI ψ KS ψ
Generalized Eigenvalue Problem
Find the eigenvector with the smallest eigenvalue
SINR
0
φ
( , ) ( , )
1S p S p
SINR ψ ψ
KI φ KS φ
73
Page 74
Second Optimization Technique
( , )
( , )
H
S p
H
S p
SINR
φ
φ
ψ KS ψ
ψ KIN ψ
( , )
H
S p φKIN UΛU
: Unitary Matrix
: Diagonal Positive Matrix
U
Λ
( , )
H H H H
S p φψ KIN ψ ψ UΛU ψ u u
1/2 Hu Λ U ψ
H
HSINR
u Φu
u u
1/2 1/2
( , )
H
S p
φΦ Λ U KS UΛ
maxFind the eigenvector of with maximum eigenvalueu Φ
1/ 2
max
1/ 2
max
opt
UΛ uψ
UΛ u74
Page 75
Signal and Interference Kernel Computation
1/3
1
( , )
0
( )k
KH
S p nN k p
n k
φ
K Σ Σ φφ Ω
0 ( ( )) if ( )mod 0
0 else
D s
pq
QJ B T p q p q Q
1
0 ( , )
0
( )k
KH
S p k p
k
φ
K Σ φφ Π
0( ( ))pq D sJ B T p q
( , ) 0 ( , )S p S p φ φKS K ( , ) ( , ) 0 ( , )S p S p S p φ φ φ
KI K K
Π Q Ω
Dependence on channel Doppler; Computed once
DN Q
75
Page 76
Signal and Interference Kernel Computation
2/3
φ H φφ
1
0
( )k
KH
k p
k
Σ φφ
Duration: DT
DN samples
76
Matrix shifts according to
The multipath power profile
Page 77
Signal and Interference Kernel Computation
3/3
Matrix shifts according to
the normalized symbol duration N
77
1
0
( )k
KH
nN k p
n k
Σ Σ φφ
Page 78
Numerical Results: Impact of Initialization and
Existence of Local Maxima
78
Local maxima
Conjecture to
be the global
maximum
Page 79
Numerical Results: Evolution of Transmit and
Receive Pulse Shapes Through the Iterations
79
Iterations: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,…,20,…,30,…,100
φ ψ
Initialization: Gaussian pulse
Page 80
Numerical Results: Doppler Spread-Delay Spread
Balancing
80
Best balancing
Page 81
Numerical Results: Doppler Spread-Delay Spread
Balancing
81
Page 82
Numerical Results: Optimized Waveforms
82
Page 83
Numerical Results: Optimized Waveforms
83
Page 84
Numerical Results: Performance and Gain in
SINR – Identical Pulse Shape Durations
84
Gain > 5dB
Page 85
Numerical Results: Performance and Gain in
SINR – Different Pulse Shape Durations
85
Page 86
Numerical Results: Uneven Distribution of PHY
Delay/Complexity Between Transmitter and
Receiver 1/2
86
PHY delay = (D+D)/2 = 3T
Page 87
Numerical Results: Uneven Distribution of PHY
Delay/Complexity Between Transmitter and
Receiver 2/2
87
PHY delay = (D+D)/2 = 5T
Page 88
Numerical Results: Spectrum of One Subcarrier
88
~ 60 dB
Page 89
Numerical Results: Spectrum of 64 Subcarriers
89
~ 60 dB
Page 90
Numerical Results: Sensitivity to an Estimation
Error on BdTm
90
Page 91
Numerical Results: Sensitivity to Synchronization
Errors in Frequency
91
Tolerence margin > 10%
Page 92
Numerical Results: Sensitivity to Synchronization
Errors in Time
92
38-sample error tolerence
34-sample error tolerence
Page 93
Conclusion and Perspectives for Future Research
Work on 5G
93
Page 94
Conclusion
We proposed a new and straightforward technique for the
systematic optimization of transmit and receive waveforms
for OFDM/FBMC/GFDM systems
Increased SINR
6 orders of magnitude reduction in out-of-band emissions
Robustness to synchronization errors
94
Page 95
Perspectives
Extension to OFDM/OQAM
Extension to multi-pulse OFDM/QAM, to OFDM/OQAM and to
Staggered OFDM (quincunx and hexagonal time frequency lattices)
Extension to single-carrier communications
Extension to underwater acoustic communications
OFDM pulse shapes optimized for partial equalization
OFDM tolerant to bursty communications with relaxed
synchronization
OFDM pulse shapes optimized for carrier aggregation and reduced
out-of band emissions
OFDM pulse shapes optimized very low latencies
Optimization of RADAR pulses
95
Page 96
96
Thank You for Your Attention!
Mohamed Siala
MEDIATRON Laboratory
Higher School of Communication of Tunis (SUP’COM)
The 2015 International Symposium on Networks, Computers
and Communications (ISNCC 2015)
May 13-15 2015, Yasmine Hammamet, Tunisia
Page 97
References 1/4
M. Siala, T. Kurt, and A. Yongaçoglu, “Orthonormalization for Multi-Carrier Transmission,”
Canadian Workshop on Information Theory 2005 (CWIT’05), Montreal, Quebec, Canada, June 2005.
T. Kurt, M. Siala, and A. Yongaçoglu, “Multi-Carrier Signal Shaping Employing Hermite
Functions,” European Signal Processing Conference 2005 (EUSIPCO’05), Antalya, Turkey,
September 2005.
N. Debbabi, M. Siala, and H. Boujemâa, “Optimization of the OFDM Prototype Waveform for
Highly Time and Frequency Dispersive Channels Through a Maximization of the SIR,” 12th
IEEE International Conference on Electronics, Circuits and Systems 2005 (ICECS’05), Gammarth,
Tunisia, December 2005.
A. Ben Salem, M. Siala, and H. Boujemâa, “Performance Comparison of OFDM and
OFDM/OQAM Systems Operating in Highly Time and Frequency Dispersive Radio-Mobile
Channels,” 12th IEEE International Conference on Electronics, Circuits and Systems 2005
(ICECS’05), Gammarth, Tunisia, December 2005.
M. Siala, T. Kurt, and A. Yongaçoglu, “A Unified Framework for the Construction of
OFDM/OQAM Systems,” 12th IEEE International Conference on Electronics, Circuits and Systems
2005 (ICECS’05), Gammarth, Tunisia, December 2005.
97
Page 98
References 2/4
A. Ben Salem, M. Siala, and H. Boujemâa, “OFDM systems with hexagonal time-frequency
lattices and well time frequency localized prototype functions,” Third International Symposium
on Image/Video Communications over fixed and mobile networks 2006 (ISIVC’06), Hammamet,
Tunisia, September 2006.
M. Siala, “Novel OFDM/OQAM system with hexagonal time-frequency lattice,” Third
International Symposium on Image/Video Communications over fixed and mobile networks
(ISIVC’06), Hammamet, Tunisia, September 2006.
I. Trigui, M. Siala, and H. Boujemâa, “Optimized pulse shaping for OFDM multi-user
communications over doubly dispersive channels,” 9th International Symposium on Signal
Processing and its Applications (ISSPA’07), Sharjah, United Arab Emirates, February 2007.
M. Siala and A. Yongaçoglu, “Prototype waveform optimization for an OFDM/OQAM system
with hexagonal time-frequency lattice structure,” 9th International Symposium on Signal
Processing and its Applications (ISSPA’07), Sharjah, United Arab Emirates, February 2007.
I. Trigui, M. Siala, S. Affes and A. Stephenne, “SIR Optimized Hermite-Based Pulses for BFDM
Systems in Doubly Dispersive Channels,” International Symposium on Signals, Systems and
Electronics (ISSSE’07), Montreal, Quebec, Canada, July 2007.
98
Page 99
References 3/4
R. Ayadi, I. Kammoun, and M. Siala, “Optimization of the pulse shape of OFDM systems Using
the Arrow-Hurwicz Algorithm,” 4th International Symposium on Wireless Communication
Systems (ISWCS’07), Trondheim, Norway, October 2007.
R. Ayadi, M. Siala, and I. Kammoun, “Transmit/receive pulse-shaping design in BFDM systems
over time-frequency dispersive AWGN channel,” IEEE International Conference on Signal
Processing and Communications (ICSPC’07), Dubai, United Arab Emirates, November 2007.
I. Trigui, M. Siala, S. Affes, A. Stephenne, and H. Boujemaa, “Optimum Pulse Shaping for
OFDM/BFDM Systems Operating in Time Varying Multi-Path Channels,” IEEE Global
Telecommunications Conference (GLOBECOM’07), Washington DC, USA, November 2007.
M. Bellili, M. Siala, and L. Ben Hadj Slama, “Pulse design for maximizing SIR in partially
equalized OFDM/BFDM systems,” IEEE 19th International Symposium on Personal, Indoor and
Mobile Radio Communications (PIMRC’08), Cannes, France, September 2008.
M. Bellili, L. Ben Hadj Slama, and M. Siala, “Multi-pulse/single-pulse design for maximizing SIR
in partially equalized OFDM systems over highly dispersive channels,” 16th IEEE International
Conference on Electronics, Circuits, and Systems, 2009 (ICECS 2009), Hammamet, Tunisia,
December 2009.
99
Page 100
References 4/4
R. Ayadi, I. Kammoun, and M. Siala, “Optimal OFDM Pulse Design, Analysis and
Implementation Over Doubly Dispersive Channel,” 21st European Signal Processing Conference
(EUSIPCO 2013), Marrakech, Morocco, September 9-13, 2013.
M. Siala, F. Abdelkefi and Z. Hraiech, “Novel Algorithms for Optimal Waveforms Design in
Multicarrier Systems,” IEEE Wireless Communications and Networking Conference
(WCNC’2014), Istanbul, Turkey, April 2014.
Z. Hraiech, M. Siala, and F. Abdelkefi, “Numerical Characterization for Optimal Designed
Waveform to Multicarrier Systems in 5G,” 22nd European Signal Processing Conference 2014
(EUSIPCO 2014), Lisbon, Portugal, 1-5 September 2014.
Z. Hraiech, F. Abdelkefi, and M. Siala, “POPS-OFDM: Ping-pong Optimized Pulse Shaping-
OFDM for 5G systems,” accepted at IEEE International Conference on Communications (ICC’15),
London, UK, June 2015.
Z. Hraiech, F. Abdelkefi, and M. Siala, “POPS-OFDM: Ping-pong Optimized Pulse Shaping-
OFDM for 5G systems,” Accepted at IEEE Vehicular Technology Conference – Spring 2015
(VTC’S15), Glasgow, Scotland, May 2015.
100
Page 101
101
Analog Transmitter for OFDM
Complex implementation: Use of a battery of N costly analog filters
ka
0na
mna
1,N na
S
/
P
( )t
( )t
( )t
0exp( 2 )j f t
exp( 2 )mj f t
1exp( 2 )Nj f t exp( 2 )cj f t
{} ( )ee t ( )e t
S/P : Serial-to-Parallel Converter
0m
mf f
T
0 0f in general
Page 102
102
Vectorial Equivalent of the OFDM Transmitter
ka
0na
mna
1,N na
S
/
P
exp( 2 )cj f t
{} ( )ee t ( )e t
0 ( )n t
( )mn t
1, ( )N n t
Page 103
103
Analog Receiver for OFDM
Complex implementation: Use of a battery of N costly analog filters
ˆkaP
/
S
( )t
0exp( 2 )j f t
exp( 2 )mj f t
1exp( 2 )Nj f t exp( 2 )cj f t
( )x t
P/S : Parallel to Serial Converter
0ˆ
na
ˆmna
1,ˆ
N na
( )t
( )t
*( ) ( )t T t
Page 104
104
Vectorial Equivalent of the OFDM Receiver
ˆkaP
/
S
exp( 2 )cj f t
( )r t
0ˆ
na
ˆmna
1,ˆ
N na
*( ), ( ) ( ) ( )u t v t u t v t dt
0 ( ),n t
( ),mn t
1, ( ),N n t