Continuous Phase Modulation: a “New Waveform for 5G ... · Institut Mines-Télécom Reduced State...
Transcript of Continuous Phase Modulation: a “New Waveform for 5G ... · Institut Mines-Télécom Reduced State...
Institut Mines-Télécom
Continuous Phase Modulation: a “New " Waveform for 5G Massive Machine -Type Communications
Malek MESSAI, Karine AMIS, Frédéric GUILLOUD
In collaboration with Giulio COLAVOLPE (Univ. Parma)
ACOSIS 2016,17-19 Oct., Marrakesh
Institut Mines-Télécom
Contents
�Context and Motivation
�CPM basics
�CPM Issues
• Complexity
• Robustness
• Multiple Access
�Conclusions
02/11/20162ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Context and Motivation
02/11/20163ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Machine-Type Communications: The Smart Metering Example (Automatic Meter Reading)
02/11/20164ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
� Meter for (Electricity), Water, Gas, …
Credits : Ondeo Systems
Institut Mines-Télécom
Requirements for Deployment
� Transmissions:
• Short Frames
• Low Throughput
• Rare (idled nearly all the time)
• Mostly Asymmetric Transmissions
� Transceivers:
• Massive deployment
• Low cost devices
• Battery-limited device
02/11/20165ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Battery Powered:How to Increase the Lifetime?
� Re-Charge Battery
• But human technical support is not acceptable
• Energy harvesting may be unavailable (e.g. gas meter
in the cellar, …)
� Reduce the Consumption
• Processing: reduced to minimum
• Transmission
─ Transmitted power
─ Power Amplifier Consumption
02/11/20166ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Power Amplification Basics
02/11/20167ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
APoutPin
Pout
Pin
Variable
envelope
Constant
envelope
Lost energy(heat)
Amplifier
Institut Mines-Télécom
Linear Modulation Envelope
02/11/20168ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Pm
oy=
0.1
25 W
QPSK
Pmax = 0.25 ∆= 3 dB
Institut Mines-Télécom
Linear Modulation Envelope
02/11/20169ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Pm
oy=
0.1
25 W
QPSK 256-QAM
Pmax = 0.5
Pmax = 0.25 ∆= 3 dB∆= 6 dB
Institut Mines-Télécom
Linear Modulation Envelope
02/11/201610ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Pm
oy=
0.1
25 W
QPSK 256-QAM
QPSK – OFDM(256) 256-QAM – OFDM(256)Pmax = 0.86
Pmax = 0.77
Pmax = 0.5
Pmax = 0.25 ∆= 3 dB∆= 6 dB
∆= 8.4 dB∆= 7.9 dB
Institut Mines-Télécom
Modulations with Constant Envelope
02/11/201611ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
sin � � /�
cos � � /��
Constant envelope modulations
(FSK, FM, GMSK, …):
s�� �e�� �
Constant envelope modulations with
higher spectral efficiency:
CPM (Continuous Phase Modulation):
� � 2�� � ��� � � ���
���
http://www.ece.umd.edu/~tretter/commlab/c6713slides/FSKSlides.pdf
Pmoy = Pmax∆= 0 dB
DSP of FSK (switched or filtered)
Low cost modulator:Voltage Control Oscillator (VCO)
Modulation index(VCO gain)
Institut Mines-Télécom
Continuous Phase Modulation(CPM)
02/11/201612ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
CPM is not New!A little bit of history …
� First Waveforms
• 1961: MSK [DH61]
• 1972: Fast FSK [DB72]
� Hide a trellis structure: increase the minimal distance
• 1971: CPFSK [PLD71]
� Generalization: CPM
• 1979 [ARS81]
02/11/201613ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
[DB72] Rudi De Buda. Coherent demodulation of frequency-shift keying with low deviation ratio. Communications, IEEE Transactions on, 20(3):429 435, 1972.
[DH61] M. L. Doelz and E. H. Heald, "Minimum-Shift Data Communications Systems," U. S. Patent No. 2,977,417, (March 28. 1961).
[PDL71] MG Pelchat, RC Davis, and MB Luntz. Coherent demodulation of continuous phase binary FSK signals (coherent demodulation of continuous phase binary fsk signals in additive white noise, determining error probability). In International Telemetering Conference, Washington, D. C, pages 181 190, 1971.
[ARS81] Tor Aulin, Nils Rydbeck, and Carl-Erik W Sundberg. Continuous phase modulation Part i and ii: Full / Partial response signaling. IEEE Transactions on Communications, 29(3):196 225, 1981.
Institut Mines-Télécom
But there’s a re-growth interest for CPM:
02/11/201614ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
DECT
IRIG 106(Airborne telemetry)
Telecom Industry
Industry where measurements matter
Institut Mines-Télécom
CPM Signal Model
� Complex baseband CPM signal: � �, !, � "# e�$ �,!,%
� The information-carrying phase:
& �, !, � 2�� � ��� � � ���
���
02/11/201615ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
GMSK (h=0.5, L=3) MSK (h=0.5, L=1)
Institut Mines-Télécom
CPM Signal Model
� During � �th time interval t ∈ [nT, n + 1�T]:
& �, �, � �� � ���./
���+ 2�� � ��� � � ��
�
���./01
&%,� + &%,���, and �� ∈ ±1, ±3, … , ± 25 � 1
� The CPM signal is completely defined by symbol �� and
state
6%,� &%,�, ��./01, … , ��.1�
02/11/201616ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
CPM Signal Model: Trellis ( � 7/8)
02/11/201617ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
6%,�9:�
&%,�9:�, ��./01
9:� , ��./0;9� , … , ��.1
9�
6%,�019�
&%,�019� , ��./0;
9� , … , ��.19 , ��96%,�
9<�
&%,�9<�, ��./01
9<� , ��./0;9� , … , ��.1
9�
��9
��9
6%,�1� 6%,�01
1�
6%,�=>?@:� 6%,�01
=>?@:�
Institut Mines-Télécom
But CPM offers also some interesting issues …
Issue #1: DECODING COMPLEXITY
02/11/201618ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
[MAG16] M. Messai, K. Amis and F. Guilloud, "On the Optimization of a PSP-Based CPM Detection," in IEEE Transactions on Wireless Communications, vol. 15, no. 3, pp. 2144-2154, March 2016.
Institut Mines-Télécom
Non-Coherent v.s. Coherent Receivers
� Non-coherent
• Bluetooth, AIS
• Low complexity but poor
performance
� Coherent
• GSM …
• Better performance but
more complex (in general)
02/11/201619ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Non-coherent
Coherent
≈ 9 dB
GFSK, M=2, h=0.35, L=3 (Bluetooth)
Institut Mines-Télécom
Coherent Decoding is Based on the Trellis
�� is to be rational
�5 branches per states
• OK if 5 2�But 85/.1 states …
�Ex. GFSK
� 0.35 7/20, 5 2,
E 3• 7 F 2; 28 states!
02/11/201620ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
6%,�9:�
6%,�019�
6%,�9<�
��9
��9
6%,�1� 6%,�01
1�
6%,�=>?@:� 6%,�01
=>?@:�
Institut Mines-Télécom
Reduced State Trellis
� Based on the decomposition of � according to [MAG16]:
� �H + I%
& �, J, � & �, J, �H + I$K + 2�I% � ��� � � ���
���./01
02/11/201621ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Tracked by the Viterbi algorithm
Calculated using the PSP technique
Calculated at the output of the matched filter
[MAG16] M. Messai, K. Amis and F. Guilloud, "On the Optimization of a PSP-Based CPM Detection," in IEEE Transactions on Wireless Communications, vol. 15, no. 3, pp. 2144-2154, March 2016.
Institut Mines-Télécom
Trellis Reduction Example� 5/8, 5 2, E 2
02/11/201622ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
The PSP update of the phase difference
for each state
Δ$KM:9 Δ$K
N∗ + ��./01N∗ � Δ%
Reduced State-trellis with �H 2 3⁄
Institut Mines-Télécom
Simulations results: uncoded case
02/11/201623ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Bluetooth standard (Binary 3GFSK modulation) for � 0,28 Q;R
Full-state receiver Reduced-State receiver
� 7 25⁄ hH 2 3⁄ �H 1 2⁄ �H 2 5⁄200 12 16 20
Institut Mines-Télécom
Simulations results: uncoded case
02/11/201624ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
CPFSK modulation with � 5 8⁄Full-state receiver Reduced-State receiver
� 5/8 hH 2 3⁄ �H 1 2⁄ �H 2 5⁄16 3 4 5
Institut Mines-Télécom
How to avoid bad �Tvalues?
� Use a bound on the minimal distance
• (for uncoded cases)
� Use EXIT charts analysis
• (for coded cases)
� Choose among the remaining �T according to
the lowest complexity
02/11/201625ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Minimum Distance Analysis(Uncoded CPM)
� The normalized squared Euclidean distance:
U; J � V log; 5
� Y 1 � cos & �, J � V, ∆% � & �, J � V, �H�#
�dt
� An upper bound denoted U\ on the minimum distance U]^_can be
found:
02/11/201626ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
(T1): PSP process (T2): reduced trellis
{ } { })( min )( min 2
b)(a,
2
ba f00
badbad −≤−Γ∈≠
2mind 2
Bd
Γa: difference sequence J � V of short duration on the reduced state based upon �H .
Institut Mines-Télécom
Upper Bound on U]^_ Example :Quaternary 2RC
02/11/201627ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Simulations results: Coded caseCC(7,5), 5 4, 2RC CPM � 1 4⁄
02/11/201628ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Iterative decoding
Institut Mines-Télécom
Coded CPM: EXIT Charts Analysis(same parameters)
02/11/201629ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
But CPM offers also some interesting issues …
Issue #2: ROBUSTNESS
02/11/201630ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
[MCAG15a] M.Messai, G.Colavolpe, K. Amis, and Frederic Guilloud, "Robust Detection of Binary CPMs with unknown modulation index" IEEE Commun. Letters, vol. 19, no.3, pp. 3339-342, March 2015.
[MCAG15b] M.Messai, G.Colavolpe, K. Amis, and F. Guilloud, "Binary Continuous Phase Robust to a Modulation Index Mismatch" IEEE Trans. Commun, vol. PP, September 2015.
Institut Mines-Télécom
MLSE (Trellis) Detector is very sensitive to the Modulation Index
02/11/201631ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Modulation Index = 0.5(MSK)
Modulation Index = 0.52
Institut Mines-Télécom
Robustness to Modulation Index Mismatch and Phase Noise for a binary CPM
� Problem: � → � + �d� Solution from State of Art: Non Coherent receiver
� Proposed solutions:
Modulation index error �d induce erroneous Cumulative phase state
02/11/201632ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
∑−
=
=∆Ln
iien ah
0
πφ
Unknown To be cancelled
� Cancelling the phase errorfrom the transmitter side
Adequate precoder
� Tracking the cumulated phase error and phase noise at the receiver side
Factor Graph and SPA
Institut Mines-Télécom
Robustness to Mismatched Modulation Index: Receiver Perspective (Binary CPM)
� We consider a transmission over a channel affected by phase noise [BC07]
� In addition to the phase noise, a modulation index error may occur.
� The cumulated phase error and the phase noise are tracked from the
receiver side by
A Factor Graph and the Sum Product Algorithm
using an a-priori on the modulation index probability density function
� Based on the Laurent decomposition.
• We describe the joint a-posteriori probability of the transmitted symbols,
the channel phase, and the modulation index through a factor graph.
• We apply the sum-product algorithm on the resulting graph to compute-
either exactly or approximately-this a-posteriori probability.
02/11/201633ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
[BC07] A. Barbieri, G. Colavolpe, "Simplified soft-output detection of CPM signals over coherentand phase noise channels," IEEE Trans. Wireless Commun., vol. 6, pp. 2486-2496, July 2007.
Institut Mines-Télécom
Laurent Decomposition (= CPM Linearization)
� Based on Laurent representation, the complex envelope of a CPM signal may be exactly expressed:
� In the binary case, most of the signal power is concentrated in the first component, i.e., that associated with pulse 8_0 (�), which is called principal component and the principal pseudo-symbol satisfies
� A very good trade-off between approximation quality and number of signal components is reached.
02/11/201634ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
)(),,(1
0, nTtphats
Q
m nmnm −=∑∑
−
=
α )12(2 )(log1 2 −= − MLQ
nhajnn e παα 1,0,0 −=
Institut Mines-Télécom
Simulation Results(Bluetooth -BR - GFSK)
02/11/201635ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
+ reduced Trellis
Institut Mines-Télécom
But CPM offers also some interesting issues …
Issue #3: MULTIPLE ACCESS
02/11/201636ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Multiple Accessbased on Frequency Division (FDM -CPM)
� Spectral Efficiency can be increased by reducing the spacing
between two adjacent channels
� But increasing the Interferences
� Factor-Graph based multi-user CPM detector proposed in
[PC10] and extended to the robust case in [MFA15]
02/11/201637ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
[PC10] A. Piemontese and G. Colavolpe, A Novel Graph-Based Suboptimal Multiuser Detector for FDM-CPM Transmissions, IEEE Trans. Wireless Commun., vol. 9, pp. 2812-2819, September 2010.
[MFA15] M. Messai, F. Guilloud, K. Amis, and G.Colavolpe, "Robust Multiuser Binary CPM Detection with Unknown modulation index," Proc. EUSIPCO 2015, Nice.
Institut Mines-Télécom
Simulation Results (MSK, ∆e f/gh,3 or 5 users, CC(7,5))
02/11/201638ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Increase in number of users:Machine-Type Assumptions
02/11/201639ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
� Short frames (isymbols)
� Sporadic
transmission:
• the probability
that a node is
active is very low
8j≪ 1�� Question: estimate
• Active users?
• Data symbols?
Institut Mines-Télécom
Compressed Sensing Based Multi-User Detection
02/11/201640ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
� SoA: Linear modulations
� Matched filter + Nyquist rate
sample at � ��� Solve lm argmin q� � rl ;But Unknown l has length s, and
only one equation!
Using "l is sparse” as an a-priori, can be solved with
CS if iu ≳ sj log wwx
� Introduce diversity iu (antennas or spreading factor for example)
� Then iu equations, but iu y s …
(Joint work also with Prof. Abdeldjalil Aissa El Bey)
Institut Mines-Télécom
Issues for CS -MUD in CPM
� Symbols cannot be sampled at the Nyquist Rate
� Non-linear modulation not fitted to CS
� Laurent Decomposition: Linearization
• But then the waveform does not fulfill the
Nyquist criteria
• Symbols are dependent
• Use Block-wise sparsity
02/11/201641ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Take-Away Messages
� Many multicarrier waveforms are studied for 5G
• But would a single waveform for all use-cases be
reasonable?
� CPM = an interesting alternative for battery powered
transmitter without recharge facility
� CPM issues are now well investigated:
• Complexity of the coherent receiver
• Robustness w.r.t. modulation index / phase noise
• Multiple Access
� Open Issues: Synchronization, Rayleigh Fading, Channel
estimation in Sporadic Communications
02/11/201643ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
Institut Mines-Télécom
Thank you for your attention
Questions Are Welcome!
02/11/201644ACOSIS 2016, 17-19 oct., Marrakesh
CPM for 5G MTC Communications
��را