Professional Documents
Culture Documents
Abstract—In this paper, spatial modulation (SM) is introduced delivers high data rate service for fixed receivers. Therefore,
to layered division multiplexing (LDM) systems for enlarging the the UL and the LL are also referred to as mobile layer (ML)
arXiv:1712.00243v4 [cs.IT] 14 Jan 2018
spectral efficiency over broadcasting transmission. Firstly, the SM and fixed layer (FL), respectively. When detecting the ML
aided LDM (SM-LDM) system is proposed, in which different
layered services utilize SM for terrestrial broadcasting transmis- service, the FL service is treated as additional interference,
sion with different power levels. Then a spectral efficiency (SE) and when detecting the FL service, the ML service need to be
analysis framework for SM-LDM systems is proposed, which firstly cancelled [6].
is suitable for the SM-LDM systems with linear combining. Spatial modulation (SM) is proposed as a novel architecture
Moreover, the closed-form SE lower bound of SM-LDM systems of multiple-input multiple-output (MIMO) systems, which
with maximum ratio combining (MRC) is derived, which is based
on this framework. Since the theoretical SE analysis of single only activates one transmit antenna (TA) for delivering the
transmit antenna (TA) LDM systems with MRC and spatial constellation symbol in each time slot with only one radio
multiplexing (SMX) aided LDM systems with MRC lacks a frequency (RF) chain [7] [8]. Therefore, the information can
closed-form expression, the closed-form SE is also derived for be transmitted from both the spatial domain and constellation
these systems. Monte Carlo simulations are provided to verify domain, and SM systems can achieve a better energy efficiency
the tightness of our proposed SE lower bound. Furthermore,
it can be shown via simulations that our proposed SM-LDM (EE) than traditional MIMO systems. In addition, with only
systems always have a better SE performance than single-TA one TA active in each time slot, SM has a more relaxed
LDM systems, which can even outperform the SE of SMX aided inter-antenna-synchronization (IAS) than traditional MIMO
LDM (SMX-LDM) systems. systems, and SM has no inter-channel interference (ICI) [8].
Index Terms—Layered division multiplexing (LDM); Spatial SM systems can also be combined with other schemes,
modulation (SM); terrestrial broadcasting transmission; spectral such as massive SM MIMO systems [9] [10], non-orthogonal
efficiency (SE). multiple access aided SM (SM-NOMA) systems [11] and gen-
eralized spatial modulation (GenSM) aided millimeter wave
I. I NTRODUCTION (mmWave) systems [12] [13]. In broadcasting transmission
scenarios, SM systems are also introduced to obtain a bet-
L AYERED division multiplexing (LDM) technology is re-
cently proposed to satisfy the rapidly increasing spectral
efficiency (SE) demand of digital terrestrial television (DTT)
ter trade-off of SE and EE [14] [15]. More specifically, in
[14], SM is combined with massive MIMO and orthogonal
frequency division multiplexing (OFDM) in high speed train
transmission, which has been accepted in the Advanced Tele-
systems, and in [15], a block-sparse compressive sensing (BS-
vision Systems Committee (ATSC) 3.0 standard [1]–[6]. As a
CS) based method is proposed for detection of GenSM with
non-orthogonal multiplexing technology, LDM simultaneously
NOMA in terrestrial return channel.
transmits different layered services at different power levels.
However, to the best of our knowledge, there are no research
Comparing with traditional time division multiplexing (TDM)
about the SM system combined with LDM system. Therefore,
and frequency division multiplexing (FDM), LDM has a higher
in this paper, we combine the SM system with a two-layer
SE, which is benefited from power allocation of different
LDM system, which is denoted as the SM aided LDM (SM-
services [5]. Since different layers share the main part of
LDM) system. In this SM-LDM system, both the ML service
physical layer modules, the LDM system only has a slightly
and FL service utilize SM for terrestrial broadcasting trans-
higher complexity than the FDM or TDM system [3].
mission. The SE analysis framework of SM-LDM systems
For LDM systems, in most instances there are two layers,
with linear combining is also proposed, in which the signal-
i.e., the upper layer (UL) and the lower layer (LL), and the UL
to-interference-plus-noise-ratio (SINR) determined by specific
is allocated with a higher power level than the LL [4]. The UL
combining schemes is the only variable of mutual information
delivers low data rate service for mobile receivers, and the FL
(MI). Moreover, the closed-form SE lower bound of SM-LDM
Y. Sun, J. Wang, C. Pan and L. He are with the Department of Electronic systems with maximum ratio combining (MRC) is derived
Engineering and Tsinghua National Laboratory for Information Science and by calculating out the SINR value. In addition, since the
Technology (TNList), Tsinghua University, Beijing, 100084, China (email:
suny15@tsinghua.edu.cn). derived SE of single-TA LDM systems with MRC and spatial
B. Ai is with the State Key Laboratory of Rail Traffic Control and Safety, multiplexing (SMX) aided LDM systems with MRC lack the
Beijing Jiaotong University, Beijing, 100044, China (e-mail: aibo@ieee.org). closed-form expressions [4] [5], we also derive the closed-form
This work was supported in part by the National Key R&D Program of
China (Grant No. 2016YFB1200102-04) and the National Natural Science SE of these systems.
Foundation of China (Grant No. 61471221 and No. 61471219). The organization of this paper is summarized as follows.
2
Eh gml,n hml,n 2
ρml
H
Nt
SINRml,n = Nt n Nt
, (7)
o n o
ρml gH hml,n′ 2 − ρml Eh gH hml,n 2 + ρfl gH hml,m 2 + σ 2 Eh kgml,n k2
P P
Nt
Eh ml,n Nt ml,n Nt
E h ml,n ml
n′ =1 m=1
( Nt Nt
" Nt
#)
1 X X X det (Σml,n )
I (ŷml ; xml ) = log 2 (Nt ) − Nt + log2 (1 + Nt SINRml,n ) − log2 , (8)
Nt n=1 n=1
det (Σml,n + Σml,n′ )
n′ =1
ρfl
H 2
Nt
Eh gfl,m hfl,m
SINRfl,m = P n 2 o 2 , (18)
Nt ρfl ρfl
hfl,m + σfl2 Eh kgfl,m k2
H
H
m′ =1 Nt h
E gfl,m hfl,m′ − N t
Eh gfl,m
( Nt Nt
" Nt
#)
1 X X X det (Σfl,m )
I (ŷfl ; xfl ) = log2 (Nt ) − Nt + log2 (1 + Nt SINRfl,m ) − log2 , (19)
Nt m=1 m=1
det (Σfl,m + Σfl,m′ )
m′ =1
is applied in (20), and the asymptotically unbiased spatial- Theorem 2: The downlink SE of FL in SM-LDM systems
domain MI lower bound can be derived as follows: with linear combining can be lower bounded as (19), where
Σfl,m can be denoted as follows:
I(ŷml ; aml ) log2 (Nt ) − Nt
1 1
Nt
" N #
1 X Xt det(Σml,n ) (21)
− log . Σfl,m = diag , ..., + Nt diag{âfl,m },
Nt n=1 2 ′ det(Σml,n + Σml,n′ ) SINRfl,1 SINRfl,Nt
n =1 (23)
and âfl,m denotes the m-th column of an Nt -by-Nt identity
Therefore, by substituting (14) and (21) into (12), the SE matrix INt .
lower bound of ML can be formulated as (8), which completes Proof: The proof of Theorem 2 follows from a direct
this proof. application of the proof of Theorem 1.
From (8), it can be illustrated that the SE of ML is composed From Theorem 1 and Theorem 2, it can be shown that SE
of the constellation-domain part and the spatial-domain part, lower bound expressions for both ML and FL are almost the
and both parts are increased with the increasing of SINR. same, but the SINR of ML differs from the SINR of FL. More
Therefore, a higher SINR leads to a larger SE of ML. With specifically, the SE of ML is influenced from both transmit
a specific linear combining algorithm, the closed-form SINR symbols in FL and transmit symbols in ML, but the SE of FL
can be derived, and then aided by Theorem 1, the theoretical is only influenced by transmit symbols in FL. This is because
value of SE in ML can be formulated. the perfect ML signal cancellation and perfect CE are assumed
in this paper, and the CLI is not explicitly considered.
B. Analysis for FL
Aided by (6), the received symbol of FL can also be C. Analysis for Single-TA LDM and SMX-LDM
transformed as a vector form as follows: For conventional single-TA LDM systems, the SE of ML
Nt can be obtained by substituting Nt = 1 into (8), and the SE
X √ of FL can be obtained by substituting Nt = 1 into (19). Thus
yfl = ρfl sfl,m γfl,m hfl,m + nfl . (22)
m=1 the SE of both ML and FL for single-TA LDM systems can
be derived as follows:
For FL, gfl,m represents the linear combining vector for
ST
= log2 1 + SINRST ST ST
the m-th TA, and the SINR of the m-th TA can be lowered Rml ml , Rfl = log2 1 + SINRfl ,
bounded as (18). From (18), it can be seen that the numerator (24)
ST
represents the received power of the transmit symbol of the where Rml and RflST represent the SE of ML and FL in
m-th TA in FL, the first two terms of the denominator denote single-TA LDM systems, respectively. In addition, SINRST ml
the ISI introduced by other transmit symbols in FL, and the and SINRST fl denote the SINR of ML and FL in single-TA
last term of the denominator denotes the influence of AWGN. LDM systems, respectively. The SINRST ST
ml and SINRfl can be
Different from the SINR of ML in (7), for SINR of FL, only obtained by substituting Nt = 1 into (7) and (18), respectively.
the transmit symbols of FL introduce the ISI, and the transmit In single-TA LDM systems, only the constellation symbols
symbols of ML have no influence on the SINR of FL assuming transmit information, so (24) represents the exact value of
perfect cancellation. SE. The approximation is only conducted when deriving the
With respect to the SE lower bound of FL, from a direct spatial-domain MI.
application of the SE analysis for ML, Theorem 2 can be For SMX-LDM systems, since all transmit antennas are
introduced based on (6). active to transmit constellation symbols, the SE of ML and
5
FL can be quantified by CMCC capacity as follows: By substituting (27), (28) and (29) into (7), the SINR of the
Nt
n-th TA in ML with MRC can be formulated as follows:
X
SMX
log2 1 + SINRSMX
Rml = ml,n , ρml Nrm
n=1 SINRml,n = 2 . (30)
(25) ρml Nt + ρfl (Nt + Nrm ) + Nt σml
XNt
RflSMX = log2 1 + SINRSMX
fl,m ,
In (30), the numerator denotes the power of the targeted
m=1
received symbol of ML, the first term of the denominator
SMX
where Rml and RflSMX denote the SE of ML and FL in SMX- denotes the ISI caused by other symbols of ML, the second
LDM systems, respectively. Besides, SINRSMX ml,n represents the term of the denominator denotes the ISI cause by symbols
SINR of the n-th TA in ML of SMX-LDM systems, and of FL, and the last term of the denominator represents the
SINRSMX
fl,m represents the SINR of the m-th TA in FL of AWGN. In addition, from (30), it can be seen that increasing
SMX-LDM systems. Similarly, for SMX-LDM systems, (25) the number of RAs in ML or decreasing the number of TAs can
is the exact value rather than the lower bound, since only the bring a larger SINR for ML. Besides, although enlarging the
constellation domain transmits information. transmit power of ML can also increase the SINR of ML, the
SINR cannot increase indefinitely because of the ISI caused
IV. C LOSED -F ORM SE L OWER B OUND WITH MRC by symbols from both ML and FL.
Following from a similar application of SINR derivation in
In our proposed SE analysis framework, the SINR values ML, the SINR corresponding to the m-th TA of FL with MRC
are related to specific combining algorithms. In this section, can be derived too, which can be denoted as follows:
MRC is considered for SM-LDM systems, single-TA LDM
systems and SMX-LDM systems. In addition, the closed-form ρfl Nrf
SE lower bound for SM-TDM/FDM systems with MRC is also SINRfl,m = . (31)
ρfl Nt + Nt σfl2
formulated.
In (31), the numerator represents the power of the targeted
received symbol in FL, the first term and the second term of
A. SM-LDM
the denominator represent the ISI caused by other symbols in
In this subsection, MRC is considered for both ML and FL and AWGN, respectively. Similarly, increasing the transmit
FL, and the SINR values of these two layers are derived as power of FL can enlarge the SINR of FL. More RAs in FL
closed forms. Then substituting the closed-form SINR values or less TAs can also increase the SINR of FL.
into Theorem 1 and Theorem 2, the closed-form SE lower Aided by the SINR of ML in (30), the closed-form SE lower
bound of SM-LDM systems with MRC can be formulated. bound for ML in SM-LDM systems with MRC can be derived
For MRC, the combining vector of the n-th TA for ML is by substituting (30) into (7) and (8). The closed-form SE lower
the estimated n-th column of Hml , and the combining vector bound for FL in SM-LDM systems with MRC can also be
of the m-th TA for FL is the estimated m-th column of Hfl . derived by substituting (31) into (18) and (19).
Since the perfect CE is assumed, we have:
For the ISI terms in ML, if n′ 6= n, gml,n and hml,n′ ρml Nrm
SINRST
ml = 2 ,
are independent. When m 6= n, gml,n and hml,m are also ρml + ρfl (1 + Nrm ) + σml
(32)
independent. Therefore, in these cases we have: ρfl Nrf
SINRST
fl = .
n
H
2 o n
2
o ρfl + σfl2
Eh gml,n hml,n′ = Eh kgml,n k = Nrm ,
n 2 o n o (28) By substituting (32) into (24), the SE exact value of single-
H 2
Eh gml,n hml,m = Eh kgml,n k = Nrm .
TA LDM systems with MRC is derived. Comparing (32) with
In addition, if n′ = n, gml,n and hml,n′ are correlated. When (30) and (31), the SINR of ML and FL in single-TA LDM
m = n, gml,n and hml,m are also correlated. In these cases, systems is larger than the SINR of ML and FL for SM-LDM
aided by the property of the central complex-valued Wishart systems, respectively. Therefore, the constellation-domain MI
distribution [18], we have: of single-TA LDM systems is larger than that of SM-LDM
n systems. However, the spatial domain can also be utilized
2 o n 2 o
H
Eh gml,n hml,n′ = Eh gml,nH
hml,m for information transmission in SM-LDM systems, so the SE
n o (29) comparison between SM-LDM systems and single-TA LDM
4
= Eh khml,n k = Nrm (Nrm + 1). systems is conducted in the section of simulation results.
6
Nt Nt Nt
det ΣTF
Lml 1 X
X
ml,n
TF TF
X
Sml = log2 (Nt ) − Nt + log 2 1 + Nt SINRml,n − log2 , (34)
Lml + Lfl Nt n=1 TF TF
n′ =1 det Σml,n + Σml,n′
n=1
Nt Nt Nt
det ΣTF
Lfl 1 X X X fl,m
SflTF = log 2 (Nt ) − Nt + TF
log2 1 + Nt SINRfl,m − log 2 , (35)
Lml + Lfl Nt m=1 m=1 ′ det ΣTF + ΣTF
m =1 fl,m fl,m′
2.5
Simulation, IL=20dB Simulation, IL=20dB Single-TA LDM, IL=20dB Single-TA LDM, IL=20dB
4 1.6 1.6
SMX-LDM, IL=20dB SMX-LDM, IL=20dB
IL=5dB
IL=20dB
2.0 IL=20dB
3 1.2 1.2
1.5
2 0.8 0.8
IL=5dB
1.0 IL=20dB IL=5dB
IL=5dB
1 0.4 0.4
0.5
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 -10 -5 0 5 10 -10 -5 0 5 10
Number of RAs in ML: Nrm Number of RAs in FL: Nrf SNR (dB) SNR (dB)
(a) (b)
2.0
Bound, IL=5dB
Simulation, IL=5dB IL=20dB 2.5
SM-LDM, IL=5dB
3.0
SM-LDM, IL=5dB
1.6
Bound, IL=20dB Single-TA LDM, IL=5dB Single-TA LDM, IL=5dB
Simulation, IL=20dB SMX-LDM, IL=5dB SMX-LDM, IL=5dB
Spectral Efficiency: SE (bits/s/Hz)
2.5
Single-TA LDM, IL=20dB Single-TA LDM, IL=20dB
2.0
0.8 IL=20dB IL=20dB
IL=5dB SMX-LDM, SMX-LDM,
0.4
IL=5dB
2.0
0.0 1.5
-10 -5 0 5 10
0.5 IL=20dB
0.0
5 10 15 20 25 30 0.0 0.0
5 10 15 20 25 30 5 10 15 20 25 30
(b) SNR (dB)
(a) SNR (dB) (b) SNR (dB)
Fig. 3. The SE performance versus SNR based on simulation results and our Fig. 5. The SE performance of FL in SM-LDM, single-TA LDM and SMX-
proposed SE lower bound for ML in (a) and for FL in (b). LDM systems versus SNR with Nrm = 2, Nrf = 2 and IL ∈ {5 dB, 20 dB}.
Nt = 2 in (a) and Nt = 4 in (b).
1.0
form SE of traditional single-TA LDM systems and SMX-
LDM systems is also formulated. Our proposed SE analysis
0.8
N t
= 1,2,4
scheme can also be easily extended to the multi-layer SM-
LDM systems. Simulation results are provided to validate
N t
= 1,2,4
the tightness of our proposed SE lower bound for SM-LDM
0.6
systems, and SM-LDM systems can outperform the SE of SM-
TDM/FDM systems and single-TA LDM systems. The SM-
0.4
SM-LDM, N =1
t
LDM systems can even have a higher SE than SMX-LDM
SM-LDM, N =2
t
systems in low SNR region via simulations.
SM-LDM, N =4
t
0.2
SM-TDM/FDM, N =1
t R EFERENCES
SM-TDM/FDM, N =2
t
SM-TDM/FDM, N =4
[1] L. Fay, L. Michael, D. Gómez-Barquero, N. Ammar, and M. W. Caldwell,
t
0.0 “An overview of the ATSC 3.0 physical layer specification,” IEEE Trans.
0.0 0.4 0.8 1.2 1.6 2.0 2.4 Broadcast., vol. 62, no. 1, pp. 233-243, Mar. 2016.
[2] T. Shitomi, E. Garro, K. Murayama, and D. Gómez-Barquero, “MIMO
Spectral Efficiency of FL (bits/s/Hz)
scattered pilot performance and optimization for ATSC 3.0,” IEEE Trans.
Broadcast., to be published.
Fig. 6. The SE performance of ML and FL in SM-LDM and SM-TDM/FDM [3] S. I. Park, J. Y. Lee, S. Myoung, L. Zhang, Y. Wu, J. Montalbán, S.
systems with Nt ∈ {1, 2, 4}, Nrm = 2, Nrf = 2, SNRml = 0 dB and Kwon, B. M. Lim, P. Angueira, H. M. Kim, N. Hur, and J. Kim, “Low
SNRfl = 20 dB. complexity layered division multiplexing system for ATSC 3.0,” IEEE
Trans. Broadcast., vol. 62, no. 1, pp. 233-243, Mar. 2016.
[4] L. Zhang, W. Li, Y. Wu, X. Wang, S. I. Park, H. M. Kim, J. Y. Lee,
P. Angueira, and J. Montalban, “Layered-division-multiplexing: Theory
the FL SE of the SM-LDM system. This is because with a and practice,” IEEE Trans. Broadcast., vol. 62, no. 1, pp. 216-232, Mar.
relatively high SNR for FL transmission, the ISI rather than 2016.
the AWGN mostly influences the SINR. From (33) and (31), [5] D. Gómez-Barquero and O. Simeone, “LDM vs. FDM/TDM for unequal
error protection in terrestrial broadcasting systems: An informationtheo-
in FL, the ISI terms for both the SM-LDM systems and SMX- retic view,” IEEE Trans. Broadcast., vol. 61, no. 4, pp. 571-579, Dec.
LDM systems are similar, and in high SNR region, the AWGN 2015.
terms for SM-LDM systems and SMX-LDM systems almost [6] L. Zhang, Y. Wu, W. Li, H. M. Kim, S. I. Park, P. Angueira, J.
Montalban, and M. Velez, “Channel capacity distribution of layer-
have the same influence on SINR. Therefore, with a high division-multiplexing system for next generation digital broadcasting
SNR in FL, the SINR of SM-LDM systems are almost the transmission,” in Proc. IEEE BMSB, pp. 1-6, Jun. 2014.
same as the SINR of SMX-LDM systems. In this case the [7] R. Mesleh, H. Haas, S. Sinanovic, C. W. Ahn, and S. Yun, “Spatial
modulation,” IEEE Trans. Veh. Technol., vol. 57, no. 4, pp. 2228-2241,
constellation-domain MI in SMX-LDM systems is higher than Jul. 2008.
the spatial-domain MI in SM-LDM systems, so with a high [8] M. Di Renzo, H. Haas, A. Ghrayeb, S. Sugiura, and L. Hanzo, “Spatial
SNR the FL SE of SMX-LDM systems is higher than the FL modulation for generalized MIMO: challenges, opportunities and imple-
mentation,” Proceedings of the IEEE, vol. 102, no. 1, pp. 56-103, Jan.
SE of SM-LDM systems. In addition, for SMX-LDM systems, 2014.
in low SNR region, the FL SE with Nt = 4 is lower than that [9] T. Narasimhan, P. Raviteja, and A. Chockalingam, “Large-scale multiuser
with Nt = 2, but in high SNR region, the FL SE with Nt = 4 SM-MIMO versus massive MIMO,” in Proc. ITA, pp. 1-9, Feb. 2014.
[10] P. Patcharamaneepakorn, S. Wu, C. Wang, H. M. Aggoune, M. M.
is higher than that with Nt = 2. This is because in low SNR Alwakeel, X. Ge, and M. Di Renzo, “Spectral, energy and economic
region, this system is a power-limited system, but in high SNR efficiency of 5G multi-cell massive MIMO systems with generalized
region, this system is an interference-limited system. spatial modulation,” IEEE Trans. Veh. Technol., vol. 65, no. 12, pp. 9715-
9731, Dec. 2016.
In Fig. 6, we compare the SE between SM-LDM systems [11] X. Wang, J. Wang, L. He, and J. Song, “Spectral efficiency analysis for
and SM-TDM/FDM systems with different Nt . It can be downlink NOMA aided spatial modulation with finite alphabet inputs,”
observed that a larger Nt leads to a higher SE, but the SE IEEE Trans. Veh. Technol., vol. 66, no. 11, pp. 10562-10566, Aug. 2017.
[12] L. He, J. Wang, and J. Song, “On generalized spatial modulation aided
gain between Nt = 4 and Nt = 2 is lower than that between millimeter wave MIMO: spectral efficiency analysis and hybrid precoder
Nt = 2 and Nt = 1. This is because although increasing Nt design,” IEEE Trans. Wireless Commun., vol. 16, no. 11, pp. 7658-7671,
leads to a larger spatial-domain MI, a larger Nt also brings a Nov. 2017.
[13] L. He, J. Wang, and J. Song, “Spatial modulation for more spatial multi-
larger ISI in (30) and (31). It can also be illustrated that the plexing: RF-chain-limited generalized spatial modulation aided MmWave
SM-LDM systems can outperform the SE of SM-TDM/FDM MIMO with hybrid precoding,” IEEE Trans. Commun., to be publised.
systems. In addition, for FL in SM-LDM systems, when IL is [14] B. Gong, L. Gui, Q. Qin, and X. Ren, “Compressive sensing-based
detector design for SM-OFDM massive MIMO high speed train systems,”
small enough, decreasing IL can hardly increase the FL SE, IEEE Trans. Broadcast., vol. 63, no. 4, pp. 714-726, Aug. 2017.
which is because in a small IL, it is a interference-limited [15] T. Wang, S. Liu, F. Yang, J. Wang, J. Song, and Z. Han, “Generalized
system for FL. Similarly, when IL is large enough, increasing spatial modulation-based multi-user and signal detection scheme for
terrestrial return channel With NOMA,” IEEE Trans. Broadcast., pp. 1-9,
IL can also barely increase the SE of ML. Oct. 2017.
[16] E. Björnson, E. G. Larsson, and M. Debbah, “Massive MIMO for
maximal spectral efficiency: how many users and pilots should be
VI. C ONCLUSION allocated?” IEEE Trans. Wireless Commun., vol. 15, no. 2, pp. 1293-
In this paper, a SM-LDM system is proposed to increase the 1308, Feb. 2016.
[17] Y. Yang and B. Jiao, “Information-guided channel-hopping for high data
SE for terrestrial broadcasting transmission. The SE analysis rate wireless communication,” IEEE Commun. Lett., vol. 12, no. 4, pp.
framework is proposed with linear combining algorithms, and 225-227, Apr. 2008.
9
[18] A. M. Tulino and S. Verdú, “Random matrix theory and wireless 1, pp. 1-182, Jun. 2004.
communications,” Foundations Trends Commun. Inf. Theory, vol. 1, no.