DVB-S2x Enabled Precoding for High
Throughput Satellite Systems
Pantelis-Daniel Arapoglou, Alberto Ginesi, Stefano Cioni, ESA/ESTEC
Stefan Erl, Federico Clazzer, DLR
Stefano Andrenacci, Alessandro Vanelli-Coralli, University of Bologna
Abstract
Multi-user Multiple-Input Multiple-Output (MU-MIMO) has allowed recent releases of terrestrial LTE
standards to achieve significant improvements in terms of offered system capacity. The publications of
the DVB-S2x standard and particularly of its novel superframe structure is a key enabler for applying
similar interference management techniques -such as precoding- to multibeam High Throughput Satellite
(HTS) systems. This paper presents results resulting from European Space Agency (ESA) funded R&D
activities concerning the practical issues that arise when precoding is applied over an aggressive frequency
re-use HTS network. In addressing these issues, the paper also proposes pragmatic solutions that have
been developed in order to overcome these limitations. Through the application of a comprehensive
system simulator, it is demonstrated that important capacity gains (beyond 40%) are to be expected from
applying precoding even after introducing a number of significant practical impairments.
I. Introduction – The Need for Precoding
A variety of single-user and multi-user Multiple-Input Multiple-Output (SU/MU-MIMO) transmission
modes (TMs) has enabled consecutive releases of Long Term Evolution (LTE) and LTE-Advanced to achieve
quantum leaps in terms of offered spectral efficiencies [1], [2]. Although in early releases of LTE only
elementary MU-MIMO TMs are present, Release 10 onwards fully developed (downlink) MU-MIMO for
up to 8 antennas because:
Most of the user terminals (UTs) are not able to support a large number of receive antennas due
to terminal size/complexity.
MU channels are not susceptible to high spatial antenna correlation as the corresponding SU ones.
The satellite community has witnessed a similar capacity push for its interactive satellite networks leading
to a proliferation of High Throughput Satellite (HTS) systems. The total system throughput achieved by
such systems is in the order of hundreds Gigabits/s with future (2025) generations of large satellites
targeting up to one Terabit/s [3]. HTS systems typically employ the Ka frequency band (20/30 GHz) on the
link from the satellite towards the UTs forming multiple spot beams on ground and re-using the available
system bandwidth based on a certain pattern in frequency and polarization (colouring scheme,see Figure
1). Current large HTS systems employing cutting edge technology –like Viasat-1 with a throughput of about
140 Gbit/s– typically split the available bandwidth in two frequency bands and two orthogonal
polarizations generating the so-called four colour beam pattern across the coverage area. The
corresponding spectrum necessary to support the user services needs to be available to the feeder link,
that is the link between the gateway stations (GWs) and the satellite.
Figure 1 Multibeam HTS system architecture.
In terms of waveform, due to its high number of modulation and coding schemes (ModCods), its available
mechanisms to adapt the ModCods to the channel conditions (ACM) and its high performance, Digital
Video Broadcasting via Satellite 2nd Generation (DVB-S2) [4] has been by far the most popular choice for
providing fixed satellite services (FSS) through HTS systems. DVB-S2 is based on amplitude and phase shift
keying (APSK) modulation and BCH/LDPC (Bose Chaudhuri Hocquenghem / Low Density Parity Check)
concatenation of channel codes.
Looking at the limited spectrum available in Ka-band for FSS UTs, further leaps in terms of HTS system
capacity seem only possible by diverging from the paradigm of the four colour frequency re-use and
moving instead to a new paradigm of higher frequency re-use (of 2 or even full frequency re-use of 1).
Under such an arrangement of the frequency assignments, the interference environment between the
co-channel beams becomes harsh, as there is no longer spatial isolation between co-channel beams and
their side lobes. This negates any potential capacity increase stemming from the use of the additional
spectrum as the high co-channel interference leads to very low signal-to-noise plus interference ratio
(SNIR) and, in turn, to low spectral efficiency.
Following in the footsteps of terrestrial LTE, MU-MIMO seems like an excellent solution to manage this
high intra-system interference originating from the co-channel beams in the system. Indeed, MU-MIMO
techniques have been proposed for the forward link of HTS broadband interactive systems and are
collectively referred to as precoding [5]. That is, the GW on ground precodes across the signals intended
to the UTs that are distributed over a high number of spot beams via a multi-feed satellite antenna.
Following the early works of [6], [7], [8], the literature on satellite based precoding has intensified recently
[9], [10] demonstrating very high theoretical precoding gains in terms of system capacity even when sub-
optimal linear precoding techniques are applied over the multibeam fixed satellite channel. Despite these
promising results, only very recently researchers have started looking into the challenges of implementing
precoding in a DVB-S2 based HTS practical system with an aggressive frequency re-use [11]. These
implementation challenges, which are unfolded in Section II, are the main focus of this paper which
reports the work carried out in the last phase of the ESA Artes 1 activity “Next Generation Waveform for
Improved Spectral Efficiency (NGW)” [12]. Many of the difficulties have been accommodated by the novel
superframe format that was introduced in the extension of DVB-S2 (DVB-S2x) [13]. How the superframe
has helped in overcoming issues related to precoding is explained in Section III. The rest of the
fundamental implementation challenges for applying precoding over HTS systems are related to multicast
(or frame-based) precoding and to the UTs synchronization strategy, topics which are addressed in Section
IV and Section V, respectively. After modeling these features, the performance of a practical system
making use of the multicast precoding technique is compared against a reference four colour system in
Section VI. Useful conclusions are drawn in Section VII.
II. Implementation Challenges in Practical Systems
The support of classical (LTE-like) precoding algorithms in DVB-S2-based networks presents a number of
challenges ultimately due to the fact that the DVB-S2 standard had been conceived with broadcast Direct-
to-Home (DTH) services as the primary application. In the following paragraphs, a number of such
implementation challenges are discussed individually.
Multicast PLFRAME
In the structure of Figure 2 corresponding to the DVB-S2 physical layer frame (PLFRAME), the Layer 2
packets of different users are multiplexed together within a single codeword. In addition, their bits are
interleaved together by the DVB-S2 bit interleaver so that each resulting channel symbol might stem from
the mapping of bits belonging to different users. As precoding works at symbol level (and thus needs to
match a given channel symbol to a given user) it would be virtually impossible to generate a classical
precoding matrix this way. In addition, within the PLFRAME there are bits which do not belong to any
specific user, as they are meant to be multicasted to all the users of the PLFRAME: these are the PLH
(physical layer header) bits as well as the LDPC and BCH parity redundancy bits.
Figure 2 Structure of a DVB-S2 PLFRAME.
From a precoding point of view, this approach to DVB-S2 PL framing is adding an important practical
constraint since it implies that the precoder cannot be designed on a user-by-user basis (conventional
unicast precoding). Rather, some type of `equivalent' frame based precoding should operate on the
channels of multiple UTs encapsulated in a frame. This has given rise to a new research area within the
satellite community called multicast precoding [14], which involves a user selection and a non-
conventional precoding method. Multicast or frame based precoding approaches are briefly discussed in
Section IV.
PLFRAME of Variable Length
The DVB-S2 PLFRAME size at the output of the encoder is constant and equal to either 16k or 64k bits [4].
Nevertheless, depending on the selected modulation, the size of the resulting PLFRAME in symbols is
variable. In a multibeam HTS system where ACM is employed, it turns out that every beam transmits a
different frame size in symbols. To explain how this comes about, let us consider Figure 3.
Figure 3 Mis-alignment of the DVB-S2 PLFRAMES across co-channel carriers when using ACM mode.
In this figure, the PLFRAMEs related to different DVB-S2 carriers serving different co-frequency beams are
depicted. In this configuration, due to the variable PLFRAME length when using ACM, even if at system
initialization the PLFRAMEs across the 4 carriers were aligned in time, the alignment would soon be
violated. As a consequence, the rate of the precoding matrix computation could approach the symbol rate
(as illustrated in the same figure). Also the pilot symbols necessary to estimate the channel would be
misaligned. This issue calls for a more regular physical layer framing structure. A solution along this line
Precod.
matrix A
Precod.
matrix B
Precod.
matrix C
has been implemented in the Annex E of DVB-S2x [13], through the exploitation of the Bundled PLFRAMEs.
This is further elaborated in Section III of this paper.
Imperfect Channel Estimation
Precoding assumes the knowledge of the UTs channel state information (CSI) at the GW transmitter. CSI
should be available at the GW so that multiuser precoding can be performed. However, in the case of
precoding, each UT needs to estimate a whole vector of channels –the so called channel direction
information (CDI)- instead of a single element in the channel matrix. In addition, as already the case in
existing DVB-S2, each UT should provide to the GW also the channel quality indicator (CQI), i.e. its SNIR.
These channel estimation operations take place under a higher interference environment (compared to
the usual four colour systems) due to the more aggressive interference reuse, with the strongest
interference approaching the same power as the main carrier and a number of additional non-negligible
interferers. A possible distribution of C/I and C/N for a given geographical point over a full-frequency re-
use network, is represented in Figure 4.
Figure 4 A possible distribution of C/I and C/N for a geographical location in a full frequency re-use network.
In this case, the receiver needs to collect CSI on every single carrier with not negligible power. This implies
locking onto and performing channel estimation on the main signal, the first interference at 0dB, the
second interference at -4dB, the third interference at -8dB, and finally the fourth interference at -12 dB.
In other words, the receiver would need to perform frame, carrier, and timing synchronization as well as
channel amplitude and phase estimation on a signal which is 12 dB below the carrier and possibly a few
dBs below noise, which is a non-trivial task.
This calls for a much more sophisticated UT synchronization and channel estimation approach that has
been developed during the ESA NGW R&D activity [12] and is overviewed in Section V. Despite putting in
place these mechanisms, the CSI estimate still contains residuals errors which reduce the expected gains
of the technique. Along similar lines, also the number of subchannel coefficients that is included in the
CDI is limited, due to the limited capability of estimating below a certain carrier-to-interference level. The
effect of both imperfect CSI estimation as well as the threshold effect below which coefficients cannot be
accessed on the system performance is reported in the system simulation results presented in Section VI.
4
th
Interf: -12 dB
3
rd
Interf: -8 dB
2
nd
Interf: -4 dB
1
st
interf: 0 dB
Main Signal
Noise Level
![Figure 5 Super-frames of format with bundled PLFRAMEs (64 800 payload size) [13].](/figures/figure-5-super-frames-of-format-with-bundled-plframes-64-800-xzocz6s3.webp)
