Uplink Contention Based SCMA for

5G Radio Access
Kelvin Au, Liqing Zhang, Hosein Nikopour, Eric Yi, Alireza Bayesteh,
Usa Vilaipornsawai, Jianglei Ma, and Peiying Zhu
Huawei Technologies Canada Co., LTD.
Ottawa, Ontario, Canada
{kelvin.au, liqing.zhang, hosein.nikopour, zhihang.yi, alireza.bayestech,
usa.vilaipornsawai, jianglei.ma, peiying.zhu}@huawei.com

Abstract—Fifth generation (5G) wireless networks are Sparse code multiple access (SCMA) is introduced in [3] as
expected to support very diverse applications and terminals. a new multiple access scheme. SCMA is a codebook-based
Massive connectivity with a large number of devices is an non-orthogonal access mode allowing overloading of the
important requirement for 5G networks. Current LTE system is system with a large number of SCMA layers to enable massive
not able to efficiently support massive connectivity, especially on connectivity. Sparsity of SCMA codewords makes the near-
the uplink (UL). Among the issues arise due to massive optimal detection of over-laid SCMA layers practically
connectivity is the cost of signaling overhead and latency. In this feasible. In this paper, we employ SCMA with a contention
paper, an uplink contention-based sparse code multiple access based mechanism to address the problems described in terms of
(SCMA) design is proposed as a solution. First, the system design
massive connectivity. Section II overviews the problems in
aspects of the proposed multiple-access scheme are described.
The SCMA parameters can be adjusted to provide different
current LTE systems to support uplink massive connectivity
levels of overloading, thus suitable to meet the diverse traffic and describes the benefits of applying SCMA as a solution to
connectivity requirements. In addition, the system-level existing issues. The system model for contention-based
evaluations of a small packet application scenario are provided transmission is provided in Section III. Uplink contention
for contention-based UL SCMA. SCMA is compared to OFDMA based SCMA PHY and MAC layer designs are discussed in
in terms of connectivity and drop rate under a tight latency Section IV. Numerical results and analysis for a small packet
requirement. The simulation results demonstrate that transmission scenario are presented in Section V. The
contention-based SCMA can provide around 2.8 times gain over conclusion is given in Section VI.
contention-based OFDMA in terms of supported active users.
The uplink contention-based SCMA scheme can be a promising II. MOTIVATION AND OVERVIEW OF UL SCMA
technology for 5G wireless networks for data transmission with CONTENTION ACCESS
low signaling overhead, low delay, and support of massive
connectivity. A. Limitations of Existing LTE Technology
In LTE, uplink transmission is scheduled by a serving
Keywords—SCMA; OFDMA; contention access; 5G; multiple basestation with a request-grant procedure. A UE sends a
access scheme scheduling request (SR) to the network in UL dedicated
resources that occur periodically. The periodicity of such
I. INTRODUCTION opportunity is typically every 5 or 10 ms. In the best case
The advent of smartphones and tablets over the past several scenario, if data arrive at the UE right before the SR
years has resulted in an explosive growth of data traffic over opportunity, it needs to wait for 7 ms between the request and
the cellular network not seen in previous generations. With the uplink data transmission [2]. The delay is even larger if data
proliferation of more smart terminals communicating with arrive in between SR opportunities or the periodicity of SR is
servers and each other via broadband wireless networks, new configured to be longer. This is disadvantageous for delay
applications emerge to take advantage of wireless connectivity. sensitive traffic of future applications.
Fifth generation (5G) wireless networks are expected to
The basestation, upon receiving an SR, performs
support very diverse services – from very low latency to very
scheduling for uplink data transmission. An uplink grant is
high delay tolerant, and from very small to very large packets
then sent in the downlink control channel (PDCCH) [4]. The
in different applications [1]. An important 5G requirement is
uplink grants, together with downlink resource assignments,
that it should support massive connectivity [1] with a large
occupy the downlink control region. The cost of such dynamic
number of devices such as smart-phones, tablets and machines.
signalling for uplink is higher for small packets since the ratio
As a result, multiple access schemes play a critical role in
of signalling overhead to useful payload is high. As an
providing the increasing demand in services for future
example, if the average number of control channel elements
terminals and applications. The current Long Term Evolution
(CCEs) used for an uplink grant is 2, this is equivalent to 72
(LTE) [2] system is not able to efficiently support massive
resource elements (REs) [5]. To transmit a small amount of
connectivity, especially in the uplink. Among the issues arise
data (e.g. 20 bytes of data using QPSK ½ occupying 160 REs),
due to massive connectivity is the cost of signaling overhead
the UL grant signalling overhead (i.e. 72 REs) to the total
and latency.
resources occupied by the UL grant and UL payload (i.e. 72
REs + 160 REs) can be around 30%. Semi-persistent The above problem is addressed by the proposed contention
scheduling can be an option in order to reduce the dynamic based SCMA which takes the overloading and non-
signalling overhead [2]. However, such mechanism is more orthogonality into account. The properties of an uplink
suitable for traffic arrival that exhibits some form of periodicity contention based SCMA system are listed as follows:
such as VoIP. Hence, it cannot efficiently support bursty
traffic.  Same as OFDMA/SC-FDMA, one or a few
predefined time-frequency regions are defined for the
Conventional contention-based transmission mechanism, SCMA contention based transmission mode.
such as the schemes used in IEEE 802.11 standard [6], allows  Multiple-access is achievable through multiplexing of
terminals to attempt to transmit data immediately after packet SCMA layers within a contention region. A layer is
arrival without the need to wait for a transmission grant. spread across the entire resources of a contention
However, it can only support one user data transmission in the region.
wireless medium at any given time in the non-MIMO scenario.  Every layer represents a user. A layer or user is
The adoption of multi-user MIMO schemes in 802.11 increases characterized by a specific SCMA codebook.
the simultaneous data connections in the network. However, Furthermore, a user is represented with its
more work is needed to support massive connectivity of 5G corresponding contention region, SCMA codebook,
wireless networks. and pilot sequence.
To combat the problems described above, a new multiple-  SCMA codewords are sparse (in the sense of number
access technology is desired with the following features: of non-zero elements) such that a moderate-
complexity multi-user detection based on the idea of
 Removal or reduction of the resource allocation and message passing algorithm (MPA) is made possible
control signalling overhead, [3]. MPA provides a near-optimal joint detection and
 reduction of transmission latency, and cross-layer interference cancellation for SCMA.
 support of large amount of users to enable massive  The system can be overloaded where the number of
connectivity. multiplexed layers is more than the length of
codewords. The overloading feature of SCMA can
B. UL Contention Based Data Transmission Overview provide massive connectivity with a limited
Physical random access channel (PRACH) [4] of LTE is complexity of detection thanks to the sparsity of
used to setup a connection with the serving basestation and SCMA codewords.
prepare for data transmission. Once the connection is
established, the data transmission is carried out through a III. SYSTEM MODEL
request and grant procedure described previously.
A. SCMA Multiple Access Transmission and Detection
To reduce the latency and signaling overhead in UL, An SCMA encoder is defined as a map from bits
contention based data transmission can be used. UEs transmit to a -dimensional complex codebook of size [3]. The
data in such a way that they can contend for one or more -dimensional complex codewords of a codebook are sparse
resources without the need to go through the request and grant vectors with non-zero entries. In the uplink contention
procedure. Such a scheme is a solution to some future based multiple access, a user is configured with a codebook.
applications with very stringent latency requirements. The user’s data bits are mapped to a -dimensional codeword
Orthogonal frequency division multiple access (OFDMA) selected from the codebook and transmitted on radio
or single-carrier frequency division multiple access resources (e.g. OFDMA subcarriers). is the length of an
(SC-FDMA) can be used for either grant or grant-free UL data SCMA codeword, or equivalently it represents the spreading
transmission. In the case of grant transmission mode, factor of the system. Every SCMA block is carried over
OFDMA/SC-FDMA resources are orthogonally split among OFDMA tones. Depending on the size of a contention region,
the scheduled users. Users are informed about the scheduling multiple non-overlapped SCMA blocks may fit within the
decision through the control signaling at the cost of the extra assigned time-frequency resources.
overhead. In the case of the grant-less transmission mode, one Multiple access is achieved through the sharing of the same
or more predefined regions within the time-frequency plane of time-frequency resources among SCMA layers of multiple
OFDMA or SC-FDMA is assigned to contention based active users. Upon the reception of signals, multiple SCMA
transmission. The assigned time-frequency resources are layers may collide with the same SCMA block. The number of
shared among a pool of users which are allowed to use a simultaneous layers/users within a time slot varies depending
specific region for contention transmission. Upon the arrival of on the traffic loading and possible packet retransmissions.
a packet, a user attempts to transmit the data through a portion Assume a time slot in which the signals of users arrive
of the resources exist within a contention region. As multiple through different channels. Within an SCMA block, user
users share the same time-frequency resources, it might lead to
transmits codeword selected from a
non-orthogonal collision of the users’ arriving signals at the
codebook. The received signal after the synchronous layer
receiver. As a result, the contention based OFDMA/SC-FDMA
multiplexing can be expressed as [3]
data transmissions incur significant performance degradations
once the load of the system increases with a large number of
active users. (1)
where is the received vector over an SCMA block, IV. PHY AND MAC LAYER DESIGN
is the vector of SCMA codeword of
user , is the received signal power of user , A. UL SCMA Resource Definition
is the channel vector of user over In order to support UL SCMA with contention based
OFDMA tones of an SCMA block, is a diagonal access, radio resources are defined for the proposed multiple
matrix where its -th diagonal element is , and is the access scheme. The basic resource for contention transmission
ambient AWGN noise plus out-of-cell interference. is called a contention transmission unit (CTU). A combination
of time, frequency, SCMA codebook, and pilot sequence
Due to the sparsity feature of SCMA codewords, the defines a CTU as shown in Fig. 2. There are J unique
number of non-zero elements colliding at a give tone of is codebooks defined over a time-frequency resource. For each
effectively less than the total number of active layers ( . To codebook, L pilot sequences are associated with it. A total of
take advantage of the sparsity and limit the complexity of unique pilot sequences are defined. There are a total of
detection, a MAP multi-user detection scheme is employed. In CTUs in the given time-frequency region.
a typical scenario of UL transmission with 2 receive antennas,
the complexity of MPA detection of 6 SCMA layers with Several terminals may share the same codebook, and due to
150% overloading is less than 4 times of a linear MMSE the random access nature, those terminals may actively
receiver. The details of the MPA detection can be found in [3] transmit at the same time. As codebooks go through different
and the references within. wireless channels, the MPA receiver is still able to detect the
data streams even if they are carried over identical codebook as
B. SCMA Scalability for Massive Connectivity long as different pilot sequences are used. The receiver can
The maximum number of codebooks, J, that can be estimate the channels of different terminals with the different
generated is a function of N (non-zero entries in a codeword) pilots. Therefore, in a practical scenario, the number of active
and K (codeword length). Selection of N non-zero positions users ( ) at a specific time slot can be potentially more than .
within K elements is simply a combination problem. The Codebook reuse can help to increase the effective overloading
maximum number of such combinations is given by the factor and the number of connections to realize the massive
binomial coefficient as follows connectivity.
Consequently, user collision happens only if two or more
. (2)
users pick the same pilot sequence within a contention region.
By multiplexing codewords from the codebooks over Pilot collisions need to be resolved via the random back-off
resources, the overloading factor (OF) is defined as: mechanism as described in the next sub-section.

. (3)
By adjusting the spreading factor, K, and the number of
non-zero entries, N, different levels of overloading can be
achieved with different number of codebooks. In order to
support massive connectivity, it is desired to have an
overloading factor much greater than 1.
As an example, for a 4-dimensional complex codebook
( ) with 2 non-zero entries ( ), a set of
codebooks can be generated from (2). An illustration is
provided in Fig. 1. Two bits (b1, b2) from a data stream are Fig. 2. Definition of a contention transmission unit (CTU).
mapped to a codeword. The data are then spread over 4
subcarriers. As expressed in (1), data streams of multiple users The time-frequency resources on which the codebooks are
are overlaid with codewords from different codebooks. overlaid form a contention region. This is shown in the
designated time-frequency resources in Fig. 3. The size and the
CodeBK1 CodeBK2 CodeBK3 CodeBK4 CodeBK5 CodeBK6
number of access regions depend on many factors such as the
expected number of terminals and/or applications that are
suitable for UL SCMA.

(b1,b2) (1,1) (1,0) (1,0) (0,0) (0,1) (1,1)

Fig. 1. SCMA codebooks, encoding, and multiplexing.

 (TTI)) from a back-off window and retransmit on the
predefined CTU as the original transmission.
With the pre-defined contention regions and CTU
assignments to UEs, and the blind detection techniques the
dynamic UL grant is no longer necessary for contention based
SCMA. Since UL grants occupy DL control channel resources,
elimination of grant signaling for traffic with UL contention
based SCMA means that DL signaling overhead can be
reduced.
V. NUMERICAL RESULTS AND ANALYSIS
In this section, we present some analysis on the scalability
of SCMA and system-level simulation of the potential gain of
SCMA over OFDMA contention-based access on the uplink.
A. Scalability Analysis
Fig. 3. Uplink contention-based regions within an OFDMA time-frequency
plane. A plot of the maximum number of codebooks that can be
generated as a function of the spreading factor and the
B. Transmission Mechanism number of non-zero entries is shown in Fig. 4.
The proposed UL contention-based SCMA solution
consists of the following two major components: i) a
contention based multiple access mechanism with the
possibility of multiple UEs contending for the same resource,
and ii) non-adaptive transmission with predefined set(s) of
coding and modulation levels.
Each UE transmits data in the predefined contention region
with a CTU. The determination of the access region and CTU
can be assigned explicitly by the network via semi-static
signaling or implicitly derived from the UE ID. An example of
a simple UE-to-CTU mapping rule can be CTU_index =
UE_ID mod , where the CTU index that a UE transmits its
data over is a function of the UE ID and the total number of
resources ( ). This mapping rule can be changed in a
predefined manner from time to time in order to provide Fig. 4. Number of codebooks, J, as a function of N and K.
collision diversity.
It can be seen that by varying the spreading factor and the
The multiple access mechanism allows contention to occur number of non-zero entries, the number of codebooks can be
as multiple UEs may be assigned the same resource (CTU) as increased dramatically. As an example, when
shown in Fig. 3. The network detects the uplink packets by 70 codebooks can be generated. When more codebooks are
attempting reception using all possible access codes assigned generated, the number of CTU’s in a contention region
to the predefined contention region. Some blind detection and increases.
compressive sensing techniques can be used to perform joint
data and activity detections [7]. In addition, cyclic redundancy B. Simulation Setup and Assumptions
check masked with UE ID or header identification may be used System-level simulation for a small packet application
as detection criteria and enablers for HARQ retransmissions. scenario is performed on a 19-cell 3-sector network. Evaluation
Uplink synchronization can be maintained with a timing is done on the 2GHz carrier frequency. It should be noted that
advance procedure similar to existing LTE systems [2]. In the existing cellular bands will be part of the 5G system that
case where uplink timing alignment is lost and only traffic includes licensed, unlicensed and high frequency spectrum.
carried over contention-based SCMA is available, the timing Different numbers of users are dropped randomly in the
advance procedure precedes actual data transmission. In a network and statistics are collected for a sector in the centre
mixed traffic scenario, a UE can maintain uplink timing cell. Uplink traffic for each user follows a Poisson distribution
alignment by relying on traffic exchange in regular scheduled with a mean packet inter-arrival time of 160 ms per user. A
transmissions. specific traffic load in each sector is obtained by configuring a
different number of active users in the sector. A contention
Since multiple UEs can be mapped to the same CTU for region with a size of 4 LTE resource block (RB) pairs are
data transmission, collision can occur when more than one UE simulated and flat Rayleigh fading channels are modeled. An
has data to transmit at the same time. Collision resolution can RB pair is equivalent to 12 subcarriers in frequency and 14
employ the random back-off procedure. Each UE picks a OFDM symbols in time [2]. For fair comparison of the SCMA
random back-off time (in units of transmission time interval and OFDMA schemes, same data size that fits into the
equivalent of one RB pair OFDMA resource is considered for more than 2%, and the system outage is defined by the
each transmission with a fixed spectral efficiency of 1 bit/s/Hz. percentage of the users in outage.
The data transmitted in SCMA is spread over the 4 RB pairs.
For OFDMA, it is transmitted in one of the 4 RB pairs. Open-
loop power control for LTE [4] is applied to determine the
transmission power at the user terminal with fixed MCS. As a
result, the interference from the neighboring sectors to the
center sector can be captured, which varies in time. The MPA
receiver is employed for SCMA [3] to handle non-orthogonal
signal processing. A linear MMSE receiver is employed for
OFDMA as a baseline LTE scenario.
We consider an application scenario for small packet
transmission with low latency. Given a latency constraint, a
packet is dropped when the waiting and transmission time is
beyond the latency requirement. Perfect channel estimation is
assumed in the simulation. The detailed simulation
assumptions are described in Table I.

TABLE I. SIMULATION ASSUMPTIONS

Parameters Settings
Network 19 cells with 3 sectors per cell Fig. 5. CDF of user packet drop rate as a function of traffic loading for
layout SCMA and OFDMA.
Inter-site 500 m
distance It is found that SCMA and OFDMA perform in different
Channel Flat Rayleigh fading operating points in terms of traffic load to achieve the same
Antenna SIMO 1x2, uncorrelated antennas QoS. Specifically, SCMA demonstrates a large advantage over
configuration OFDMA in terms of supported number of active users to
Modulation Fixed.
and coding QPSK ½ for OFDMA;
achieve the same system outage. For example, to achieve a
Codebook with K = 4, N = 2, J = 6, code rate ½ for system outage of 2%, SCMA can support 119 users while
SCMA OFDM support 42 users. Fig. 6 shows the system capacity per
Channel Perfect sector for SCMA and OFDMA when they achieve the system
estimation outage of 2% and 5%, respectively. As shown in Fig. 7, the
Traffic pattern Poisson packet arrival with a mean of 160 ms simulated SCMA gains over OFDMA are 2.83 and 2.74 in
configurable number of active users in each sector
terms of number of supported users for given system outages
UE max Tx 23 dBm
power of 2% and 5%, respectively.
Power control = 1.0 (full pathloss compensation), = -95 dBm
[4]
Receiver MPA receiver for SCMA; linear MMSE receiver for
model OFDMA

C. Performance and analysis

The UL contention based OFDMA and SCMA schemes are
evaluated for transmission of small packets with low latency
requirements such that the delay is within 5 ms interval. Due to
the tight latency requirement, no retransmission opportunity is
allowed.
Fig. 5 demonstrates the distribution of user packet drop rate
(or loss) as a function of traffic loading for contention based
SCMA and contention based OFDMA. By definition, a packet
is lost if its transmission fails at the first attempt. It is shown Fig. 6. Capacity supported based on different system outages.
clearly that the OFDMA packet drop rate performance
degrades much faster with load increasing than SCMA. SCMA
has a lower user packet drop rate distributions than OFDMA
for various traffic loadings.
In the following, the system performance is evaluated based
on outage criteria. The system operates in a traffic load to
satisfy the QoS conditions defined by both UE outage and
network outage. A UE is in outage if its packet drop rate is
 traffic connectivity requirement of 5G wireless networks. We
describe the system design aspects of the new multiple access
scheme. System-level evaluations of a small packet application
scenario are provided for contention-based UL SCMA. SCMA
is compared to OFDMA in terms of connectivity and drop rate
under a tight latency requirement. The simulation results
demonstrate the potential gain of contention-based SCMA over
contention-based OFDMA scheme for small packet
transmission. SCMA can provide around 2.8 times gain over
OFDMA in terms of supported active users in a contention
based system with low latency traffic.
Therefore, the uplink contention-based SCMA scheme can
be a promising technology for 5G wireless networks for data
Fig. 7. Outage capacity gain of SCMA over OFDMA. transmission with low signaling overhead, low delay, and
support of massive connectivity.
D. PAPR Reduction for SCMA
One issue related to OFDMA signal is its high range of REFERENCES
peak-to-average power ratio (PAPR). SC-FDMA is an [1] Huawei Whitepaper, “5G: A Technology Vision,” Nov 2013.
alternative solution adopted by UL LTE to reduce PAPR which http://www.huawei.com/ilink/en/download/HW_314849.
is more suitable for low cost terminals. As mentioned before, [2] Erik Dahlman, Stefan Parkvall, and Johan Skold, 4G: LTE/LTE-
SCMA codewords are carried over OFDMA tones. Hence, the Advanced for Mobile Broadband, 2nd ed. Waltham, MA, USA:
Elsevier, 2014.
problem of high PAPR can inherently pass to SCMA-OFDM
[3] Hosein Nikopour and Hadi Baligh, “Sparse Code Multiple Access,”
transmission mode as well. As opposed to original OFDM, in PIMRC, 2013, pp. 332 – 336.
SCMA-OFDM there is an extra degree of freedom in which the [4] 3GPP TS 36.213, 3GPP Technical Specification Group Radio Access
issue of PAPR can be addressed during the process of the Network; E-UTRA; Physical Layer Procedures.
SCMA codebook design. The general method for SCMA [5] 3GPP TS 36.211, 3GPP Technical Specification Group Radio Access
codebook design is introduced in [8]. Following this method, Network; E-UTRA; Physical Channels and Modulation.
low projection codebooks are designed such that the outcome [6] IEEE Standard for Information technology--Telecommunications and
guarantees low PAPR (<5 dB) for SCMA-OFDM transmission information exchange between systems Local and metropolitan area
in certain narrow band scenarios. This is comparable with networks--Specific requirements Part 11: Wireless LAN Medium
current low PAPR SC-FDMA waveform. Access Control (MAC) and Physical Layer (PHY) Specifications, 2012.
[7] Alireza Bayesteh, Eric Yi, Hosein Nikopour, and Hadi Baligh, “Blind
Detection of SCMA for Uplink Grant-Free Multiple-Access,” Accepted
for IEEE ISWCS 2014.
VI. CONCLUSION [8] Mahmoud Taherzadeh, Hosein Nikopour, Alireza Bayesteh, and Hadi
In this paper, an uplink contention-based SCMA scheme is Baligh, “SCMA Codebook Design,” Accepted for VTC fall 2014.
proposed. The SCMA parameters can be adjusted to provide
different levels of overloading, thus suitable to meet a diverse