Scientific direction Development of key enabling technologies
Transfer of knowledge to industry

PhD : selection by topics

Technological challenges >> Communication networks, IOT, radiofrequencies and antennas
8 proposition(s).

See all positions [+]

5G mmW integrated BiDirectional TRX for hybrid and digital beamforming system

Département Architectures Conception et Logiciels Embarqués (LIST-LETI)

Laboratoire Architectures Intégrées Radiofréquences

01-10-2020

SL-DRT-20-0478

baudouin.martineau@cea.fr

Communication networks, IOT, radiofrequencies and antennas (.pdf)

This thesis addresses the topic of compact, low-cost millimeter wave transceiver in the context of the new coming 5G FR2. Indeed, a considerable number of chips and an area-efficient design will be necessary for hybrid and digital MIMO beamforming. However, conventional transceiver designs use switch-based bidirectional approach with one Tx and one Rx working alternatively in time duplex. For this reason, bi-directional transceiver completely sharing amplifiers and matching networks between the transmitter and the receiver is proposed. Additionally, bidirectional phase shifter, quadrature mixer and baseband amplifier will be studied and design offering a complete solution for hybrid and digital beamforming architecture. The thesis study will cover the architecture, the design and the measurement of such blocs in standalone and the full transceiver. The awaited innovation will encompass several aspect: bidirectional front-end compatible with hybrid configuration, mmW digital beamforming compatible, LO multiplication and local quadrature generation, CMOS SOI process. This phd research will give the opportunity to work in cross-scientific disciplinary from millimeter wave to baseband design and transceiver system architecture offering a very large panel of experiences and competencies. The thesis will take place in the CEA Leti institute under the supervision of Mr Martineau Dr and Mr Belot Hab. The publication in journals and international conferences will be encouraged and facilitated.

Download the offer (.zip)

Optomechanical reference oscillators

Département Composants Silicium (LETI)

Laboratoire Composants Micro-Capteurs

01-09-2020

SL-DRT-20-0592

marc.sansaperna@cea.fr

Communication networks, IOT, radiofrequencies and antennas (.pdf)

Clocks (reference oscillators) are ubiquitous elements in electronic circuits. The arrival of new technologies such as 5G or autonomous vehicles requires a level of performance that is not attainable by commercial clock technologies. One of the most promising routes to improve performance is the development of clocks based on micro-electromechanical (MEMS) resonators at high frequency (1-5 GHz, tens of GHz in the future). However, it is challenging to build high-performance MEMS resonators in the GHz range, mainly due to the difficulty of detecting their minuscule vibration amplitudes. Recently several groups have demonstrated the possibility of building optomechanical devices in piezoelectric materials. This technology, which was confined to fundamental studies, is now mature enough to evolve towards applications, and solves many of the difficulties involved in the implementation of MEMS clocks in the GHz range. The objective of the thesis is to develop a MEMS clock based on this novel optomechanical technology. The thesis will take place in the Microsensors Laboratory of the CEA-Leti, in collaboration with the RF Components Laboratory. The Leti is a pioneer in the implementation of on-chip optomechanical and piezoelectric resonators. The PhD student will work in collaboration with Leti Engineers to design the MEMS resonators and their fabrication process, based on an analytical study and finite-element simulations. Then, the student will have the opportunity to contribute to the fabrication of the devices in clean room. Finally, the student will characterize them in the Leti's laboratories, to extract their performance and implement a first demonstrator of MEMS clock.

Download the offer (.zip)

Resonators and devices based on elastic waves obtained through the hybridization of surface and bulk waves

Département Composants Silicium (LETI)

Laboratoire Composants Radiofréquences

01-09-2020

SL-DRT-20-0668

alexandre.reinhardt@cea.fr

Communication networks, IOT, radiofrequencies and antennas (.pdf)

Bulk or surface elastic wave devices are currently an enabling technology for radiofrequency emission/reception circuits used in mobile phones. Since, at constant frequency, the wavelength of elastic wave is close to 100,000 times smaller than electromagnetic wavelengths, the treatment of a signal carried by elastic waves instead of an electrical signal offers a tremendous miniaturization. With the increase in frequency bands operated simultaneously by each single mobile phone, requirements brought onto radiofrequency filters become more and more stringent. This motivates the research on new types of components exploiting new elastic waves. Conventional technologies rely on bulk acoustic waves (BAW) or surface acoustic waves (SAW) propagating respectively along the thickness or the surface of a piezoelectric material. Such kind of materials offer the possibility to couple electric signals into elastic waves, and conversely. In the last few years, a new kind of propagation mode, called "hybrid SAW/BAW" has been proposed, based on the excitation of waves by a periodic array of piezoelectric stubs. First realizations have been proposed, but their properties are not yet fully determined. This PhD subject focuses therefore on the study of the potentialities offered by these new kinds of modes. On one hand, the properties of such waves are strongly related to the combination of piezolectric material, of the nature of the substrate, on their respective crystal orientations as well as on the geometric dimensions of the piezoelectric stubs. The candidate will therefore investigate the design space in order to reveal what performances can be expected from such structures and optimise their design towards applications such as RF filters or time references, ideally for applications above 3 GHz. This work will leverage the simulation models available at CEA-LETI and those developped by the FrecNSys company. A second part of the PhD is expected also to explore more fundamental possibilites opened by these modes arising from the coupling between elastic surface waves and a periodic array of electrically active structures. Such periodic structures belong to the broader range of so-called elastic metamaterials, which offer unusual propagation properties such as frequency ranges in which wave propagation is forbidden, artificial slowing of waves, strong confinement or nonreciprocal propagation. Since active structures are involved, additional interesting effects may be explored. The candidate will leverage the expertise on elastic metamaterials brought by the acoustic department of ISEN. Eventually, an experimental part will be devoted to the proposition of designs to be implemented in the clean rooms of CEA-LETI and participation to the technological developments. The goal is here to assess the exprimental feasibility of such structures.

Download the offer (.zip)

Optimized coding techniques for the design of deep neural network hardware accelerators

Département Architectures Conception et Logiciels Embarqués (LIST-LETI)

Laboratoire Architectures Intégrées Radiofréquences

01-10-2020

SL-DRT-20-0689

joseluis.gonzalezjimenez@cea.fr

Communication networks, IOT, radiofrequencies and antennas (.pdf)

Artificial neural network based approaches have significantly improved performance in many areas of computer vision, such as classification, segmentation, and object detection. While the deep learning approach is now well established and the number of applications is still strongly increasing, the computational complexity and memory needs of these networks make them difficult to embed in low power hardware platforms. To overcome these difficulties, several research branches have produced different techniques that allow to reduce memory and computation footprint of artificial neural networks: reduction of the number of parameters, low precision quantization, etc. This thesis aims to go a step further in this optimization by working on the encoding of the data that is propagated through the network. In this thesis, we proposed to explore novel coding methods by working directly on the information encoding of the neural network. In particular, this thesis will aim to generalize several ideas from the domain of neuromorphic event-based computing, and adapt them to the constraints of current digital hardware. Depending on the background of the candidate, the thesis can have a focus on theoretical analysis and algorithm design, or on optimized hardware implementations. Main tasks and responsibilities: ? Analysis of activation and weight distributions of state-of-the-art DNN topologies ? Development of a framework for data driven optimization of DNNs for hardware efficient implementations ? Development of novel coding schemes for information flow in DNNs, and implementation of specific optimized hardware operators ? Publication of scientific articles in top tier international conferences and journals on neuromorphic computing and machine learning Skills and qualifications: ? Good Master (or equivalent) degree in computer science, machine learning, electrical engineering, Physics, Math or a related discipline ? Theoretical understanding and first practical experiences with machine learning and deep neural networks, in particular in computer vision ? Good programming skills (at least in Python, preferably also in C++) ? Basic understanding of embedded systems and their constraints is a plus. For a hardware heavy thesis, practical experience with HDLs is required ? Good written and spoken English skills (French knowledge is not required) Our lab possesses a strong experience in design and optimization of neural networks for embedded applications, DNN hardware accelerators and neuromorphic computing, with publications in major AI and hardware conferences and research collaborations with leading international industrial and academic partners. Our group is the main developer of the French DNN optimization and deployment framework N2D2, and the position offers the possibility to contribute to the open source version of the framework (C++ programming skills required). If you are interested, you can apply to directly with your CV, recent university transcripts, potential references for recommendations, and a brief description why you are interested in working on this specific topic.

Download the offer (.zip)

Advanced network management for controlling the real-time redeployment of a mobile network infrastructure under traffic performance constraints

Département Intelligence Ambiante et Systèmes Interactifs (LIST)

Laboratoire Systèmes Communiquants

01-09-2020

SL-DRT-20-0865

Michael.Boc@cea.fr

Communication networks, IOT, radiofrequencies and antennas (.pdf)

The digitization of industries introduces the need for providing high-speed wireless connectivity on industrial sites, which is extremely difficult due to the constraints imposed by these environments. To address them, this PhD thesis will investigate opportunities to increase real-time reconfiguration capabilities of the wireless infrastructure by means of an SDN-oriented management of the network. This network management will control the mobility of the infrastructure equipment as an additional degree of freedom in order to improve the performance of the data flows. This capability should provide two key benefits: 1) not having to rely on a lengthy and costly planning phase for network deployment, and 2) being able to implement new and more sophisticated network reconfiguration strategies to increase its overall performance level at any time. The mobility of the infrastructure could be provided by mobile robots that can be controlled through an SDN protocol and carrying some of the network equipment. In the case of a nuclear dismantling operation, for example, we could consider the wireless communication infrastructure as being composed of a fleet of mobile robots (terrestrial or aerial) whose mobility is managed by a network management system (SDN) in charge of ensuring the proper performance of the connectivity for dismantling robots remotely operated. The objective of the proposed thesis work is to define an advanced and centralized network management system for the control of the real-time redeployment of a mobile network infrastructure under performance constraints of data flows. This system should be able to 1) identify when a topological change becomes relevant considering the types of data flow performance problems and the limitations of existing network optimization solutions, 2) to define and pilot the redeployment of the network infrastructure in order to improve the performance of these data flows.

Download the offer (.zip)

Département Systèmes (LETI)

Laboratoire Sans fils Haut Débit

01-10-2020

SL-DRT-20-1008

valentin.savin@cea.fr

Communication networks, IOT, radiofrequencies and antennas (.pdf)

Download the offer (.zip)

Digital-to-Light Communications on micro-(O)LED matrices

Département Systèmes (LETI)

Laboratoire Sans fils Haut Débit

01-10-2020

SL-DRT-20-1017

luc.maret@cea.fr

Communication networks, IOT, radiofrequencies and antennas (.pdf)

In Optical Wireless Communications (OWC) with LED or OLED, increasing the data rate will probably imply the use of microLEDs that show bandwidths approaching the GHz (Organic or GaN technologies). However these micro(O)LED will be limited in terms of optical power and thus communication range. The development of microLED matrices will allow recovering enough optical power to open access to mid-range applications. An hybridization of these matrices onto another CMOS matrices of pixelated current drives will preserve the bandwidth. But it will also offer an independent access to a huge number of optical micro-sources. This type of design has recently yielded to the implementation of the so-called « Digital-to-Light » (D2L) modulations where the quantized digital signal is directly used to pilot the matrix by thermometric codes, removing the digital-to-analogue converters used in classical OWC. But this new ?Digital-to-Optical Converter? (DOC) will face new challenges due to disparities in the behavior of microLEDs when designed within a same matrix, leading to e.g. non-linearities in the emitted optical signal. This thesis must tackle these new challenges, understand the new sources of impairments and try to mitigate them by specific system architectures designs and signal processing techniques such as the implementation of efficient D2L modulations, non-uniform quantization, compensation of time constants,? The final objective is to provide an optical wireless transmission of more than 10Gb/s at ranges around one meter. Eventually, hardware implementation will be possible using micro(O)LED matrices fabricated at Leti.

Download the offer (.zip)

Reliability of power amplifiers: modeling, design and test strategies

Département Composants Silicium (LETI)

Laboratoire de Caractérisation et Test Electrique

01-10-2020

SL-DRT-20-1251

alexis.divay@cea.fr

Communication networks, IOT, radiofrequencies and antennas (.pdf)

With the rise of 5G, RF and millimeter circuits are increasing in the telecommunications sector. In the reception / transmission chain, the power amplifier (PA) is a key element and it is critical to the proper functioning of transmissions. The reliability of the latter is an important element because the RF excursions at the terminals of the device are strong and the signal shapes are complex, bringing dynamic aging mechanisms into play. In this context, the designer must rely on aging simulation tools to be able to correctly size his circuit. One of the first objectives of the thesis is to better understand the aging mechanisms put into play at the CMOS level through stress campaigns in DC and in broad signal RF on a Load-Pull bench. The transistors studied will be MOS with fine oxide as well as devices optimized for high power (DMOS). Then, different topologies of the power stage will be studied in order to correlate the design with a possible gain in reliability. Finally, an in-situ monitoring system for the degradation of the device will be proposed to compensate for performance losses or to alert in the event of an imminent failure.

Download the offer (.zip)

See all positions