Context of the thesis
Satellite telecommunications are moving towards applications requiring ever-increasing bit rates. Applications such as VoIP, television broadcasting in UHD or very high speed internet require the transmission of a lot of information through frequency bands, whose expansion is not indefinitely possible. The latest evolution of the Digital Video Brodcasting standard (DVB-S2X called "DVB-S2 revolution") enables for SATCOMS an increase in spectral efficiency of almost 15% . To achieve this, two major evolutions have been proposed: on the one hand, lower roll-offs optimizing the exploitation of the band allocated to the user (5%, 10%). On the other hand, modulations with a large number of states to communicate more bits with the same symbol (64APSK, 128APSK, 256APSK). These high spectral efficiency modulations are less robust to the Gaussian noise present in the telecommunications chains, and their use is only possible at high signal-to-noise ratio. However, very high speed SATCOM systems are evolving towards solutions where the amplified channels onboard the satellite are becoming wider, generally dealing with multiple carriers. In this context, nonlinear distortions introduced into the satellite channel can become problematic. . The innovations proposed in the DVB-S2x standard are therefore exploitable only in very specific favorable scenarios and do not cover all the communications faced by the operators.
In this thesis, new waveforms will be investigated, allowing an improvement in the spectral efficiency of the satellite channel with high constraints. Among the candidates, we will focus on the extensions of classical single-carrier modulations (SC) and their circular block version, the SC-OFDM, to packing techniques such as Time Frequency Packing (TFP) or Faster-then-Nyquist (FTN). For each case, we will evaluate their performance and their possible extensions to the multi-carrier case.
SC FTN and its TFP generalization allow the increase of the bit rate in an unchanged occupancy band by accelerating the emission rate of the symbols (or by overlapping carriers) without modifying the transmission filtering, inducing a rupture of the Nyquist criterion and therefore the introduction of strong inter-symbol interference . Without processing on the receiver side, this signal is not exploitable and therefore requires the use of equalization, linear or non-linear according to the regime of the amplifier in question. In a thesis started in late Autumn 2013 and completed in December 2017 in collaboration with the IRT Saint Exupéry, we were able to show the interest of the FTN waveform in a satellite channel in a single-carrier scenario. Several interesting physical properties in a nonlinear context of the waveform have been exhibited . From pre-distortion at the emitter to equalization at the receiver side, many algorithms have already been studied and make it possible to now which tools to use or not to use to enable efficient use of FTN based schemes. With these first results, this study was extended to equalization based on a Volterra model or SVA-type pre-distortion. The synchronization aspects in context of strong ISI have also been the subject of several investigations. In another thesis defended in July 2018, we also evaluated the gain for the single-carrier case for aeronautical communications and a DVB-RCS2 type return link. Among the solutions considered, we have extensively studied the SC-OFDM case extended to the FTN case. In the same way, the tools of joint optimization code-modulation FTN are mastered and allow a joint design .
All the elements are therefore relatively well mastered in the context of a single-carrier scheme. However, in the case of multi-carrier transponders, few studies are available to characterize and optimize the FTN schemes or its extension to the TFP case. In particular, the immunity to interference between carriers can be an important factor of discrimination between two powerful solutions, as well as the complexity of the receiver (synchronization, detection) in a frame where one considers at the same time the interference generated by the FTN and carrier interferences. This is essentially the objective of the proposed thesis.
Objectives of the thesis
In the works mentioned above, the highlighted gains are only possible by assuming optimal detection algorithms such as MAP type or quasi-optimal MAP algorithms, regardless of their actual cost of hardware implementation. The main problem is that there is an asymptotic loss that can be quite significant with high spectral efficiency for (turbo) MMSE solutions. It is therefore necessary to study further the performance-complexity trade-off. Thus, various methods of reducing the complexity of the algorithms can be studied. The equalizers used in the previous works are not the only existing ones and other detectors can be taken into consideration (algorithms based on graphs such as extended BP or graph based LMMSE or self-iterated receivers by propagation expectation, etc ...) to find the receiver offering the best performance / feasibility compromise. The aim is to achieve performances comparable to optimized sub-optimal approaches based on M * / M-BCJR type algorithms (with channel shortening methods or not) and whose performances vary enormously depending on the compression conditions, especially in the nonlinear case. The methods based on graphs make it possible to consider joint detection and decoding approaches by considering a single structure for the detector and the channel decoder. Moreover, it is in our view the only approaches that allow to scale complexity for the multi-carrier case.
Once these complexity aspects will be addressed, we will be interested in the case of the FTN in a "multi-carrier" framework [8-10]. Indeed, it is necessary to take into account the interference of adjacent carriers when considering the forward path of a broadband satellite access system. All the points discussed in the single-carrier framework can then be dealt with in this generalized framework: synchronization, joint or disjoint detection in the saturation regime, design and optimization for the FTN for the management of the ICAs, considering precoding or not. In addition, we can consider extensions currently foreseen such as the ability for FTN to be used in a MIMO framework and compatibility with "carrier-aggregation" approaches.
The objective of this thesis lies in the investigation of llow complexity processing at the receiver while taking into account the multi-carrier aspects in the FTN / TFP transmission framework.
Main objectives and work plan
The main scientific locks can be summarized as follows:
A / The complexity reduction of BCJR-based detection algorithms    is necessary in order to make FTN/ TFP compatible with current technologies, without penalizing latency for real time applications. With this in mind, trellis based algorithms will be implemented in conjunction with extended channel shortening to the nonlinear case and tested to ensure that they do not induce an excessive reduction of previously measured gains.
B / Other graph-based receivers (extended BP, graph-based LMMSE for Volterra models, or approximate message passing (AMP) or Expectation-Propagation with self-iterated receivers) will be considered. Possibly hybrid structures will be proposed.
C / Transmitter aspects in the multi-carrier framework: optimization of the FTN / TFP parameters for the case of a joint or distributed design. Impact of joint or distributed predistorsion.
D / Receiving aspects: synchronization problems and joint / separate demodulation in the multi-carrier frame. Management of the ACI. Scaling of previously studied algorithms for joint processing of intercarrier interferences.
Le candidat devra avoir un Master 2 ou diplôme d'ingénieur dans le domaine des télécommunications, traitement du signal pour les communications numériques.
Pour postuler à cette offre, nous vous invitons à vous rapprocher du directeur/rice de thèse et compléter avec son aide la partie cofinancement du formulaire en ligne (Répondre à l’offre) avant le 1er avril 2019