Wireless Communications

Algorithms and Building Blocks for Communications

Very high throughput polar decoder on APU for

Software-defined Radio Applications

 

Description: Polar codes are a new class of capacity-achieving error-correcting codes with low encoding and decoding complexity. Their low-complexity decoding algorithms render them attractive for use in software-defined radio applications where computational resources are limited.

High performance software implementation of polar decoders has received little attention so far. Most implementations focussed on desktop and embedded processors (x86-64 and ARM) with only one implementation for general-purpose graphical processing units (GPGPUs). While good throughput was achievable on GPGPU, it was observed that the latency of transfers between the host and the device memories significantely limits the usefullness of a GPGPU implementation. APUs however, are graphical cores that also have potential for highly parallel processing that share memory with the main processor.

Hence, for this project, the candidate should first go over the required background on error-correction and polar codes. The candidate should then implement a reference decoder in a language of his/her choice (subject to approval by the supervisor) that will be used later for verification. A parametrisable implementation in OpenCL specifically targeted at APUs should then be carried out in order to exploit the highly parallel architecture of these types of ASICs. Finally, the implementation should be testbenched in order to measure its effective latency and throughput for comparison against the state-of-the-art software implementations targeting other types of processors.

Prior knowledge: Interest in real-time signal processing. Knowledge of a programming language that ressembles C is a clear asset.

Kind of work: 30% background and reference implementation, 40% software architecture and 30% verification and test.

Supervisor: Pascal Giard

Date added: 30-11-2016

Implemention of a packet encoder/decoder pair for

the GNU Radio framework 

Description: The GNU Radio framework is the most popular toolkit that provides signal processing blocks to implement software-defined radios. It has been used during the DARPA Spectrum Challenge, to communicate with the ISEE-3 space probe, to teach digital communication in universities, etc. We will use it to communicate with IoT objects implementing the LoRa PHY.

Due GNU Radio’s organic nature and to contributors often scratching their own itch, the quality of the blocks within the framework varies greatly. Notably, the packet encoder/decoder pair lacks in functionality and flexibility. Reliable wireless communications often resort to soft error-correction coding, something that the current packet encoder/decoder pair does not support. Furthermore, while high-order modulations are beneficial to improve the data throughput when the channel conditions are good, robust synchronisation is better achieved with binary phase-shift keying. Using ready-made blocks, it is currently not trivial to build a packet (or frame) mixing two modulation types.

For this project, the candidate shall first familiarize himself (herself) with the GNU Radio framework. The candidate should then briefly review the existing packet (or frame) encoder/decoder blocks available. Following this survey, the candidate shall either adapt existing blocks or implement custom blocks to be integrated in GNU Radio. The blocks should be documented and of sufficient quality to be contributed back into the free software framework. Ideally, the core of the blocks should be sufficiently decoupled from GNU Radio so that they could easily be integrated into another framework (e.g. NI Labview) at the cost of writing another wrapper.

Finally, the candidate shall create a simple transceiver/receiver pair running on USRP hardware to make a live demonstration making use of his/her blocks.

Prior knowledge: Interest in real-time signal processing and software-defined radios. Good knowledge of C++ and Python is a clear asset.

Kind of work: 20% background, 50% software implementation and 30% demonstration.

Supervisor: Pascal Giard

Date added: 30-11-2016

Communications System Design and Prototyping

60 GHz Data Kiosk Demonstrator

 

Semester Project

Description: One of the promising applications of 60 GHz communication are wireless data kiosks. Terminals where you can pass by with your mobile device and download data. 60 GHz communication offers the opportunity to transfer gigabytes of data within several seconds, this enables wireless movie rentals at the airport or to sync our phone media library with the cloud within seconds.

Goal of this project is to build a 60 GHz radio link for LOS conditions. To reduce the complexity of the project we will try to rely on as many preexisting components as possible. Your task will be the assembly of the different components, engineering of the control path and basic signal processing of the received data. Although the project will be based on existing components hands-on electronics experience is strongly recommended.

Areas: RF, Electronics, System Level Programming, Signal Processing

Work Distribution: 40% Labview, 40% Hardware, 20% Measurements

Supervisor: Nicholas Preyss

Semester Project

Description:

  • All modern communication systems operate in half-duplex mode
  • Bi-directional communications achieved by dividing time and/or frequency
  • True full-duplex will off er up to 2x throughput gain! (and simplify the MAC!)
  • … but it su ffers from strong self-interference
  • The objective is to build the first demonstration of a full-duplex radio at EPFL!
  • Use a powerful new multi-antenna software-radio platform
  • Suppress self-interference by a combination of diff erent methods
  • Implement a complete OFDM transmitter and receiver
  • Carry out extensive measurements to evaluate the performance of the link

Prerequisites: Basic background in signal processing and communications systems. Some knowledge of Labview and MATLAB is an advantage.

Project Type: 60% Labview, 30% Measurements, 10% MATLAB

Supervisor: Pavle Belanovic

 

60 GHz Wideband Channel Estimation and Tracking

Master Thesis or Semester Project

 

 

Description: Intersymbol interference is one of the biggest issues one has to deal with when operating in the 60 GHz band. But good channel equalization can only be achieved when precise information about the channel conditions are available. The intial estimation and the tracking of the channel impulse response is critical for the performance of the receiver. The IEEE802.11ad standard specifies Golay sequences at the beginning of each frame and and in regular intervals during the transmission to allow channel estimations.
The existing MATLAB simulation environment should be extended to include channel estimation and the simulation with imperfect channel state information. A high-speed low-complexity hardware architecture should be designed and implemented which can cope with the gigasymbol per second rate.

Areas: Digital Signal Processing, Simulation, VLSI

Work Distribution: 40% Algorithm / Theory, 40% MATLAB, 20% VHDL

Supervisor: Nicholas Preyss

Master Thesis or Semester Project

Description:

  • Cooperative communication is an important new paradigm
  • Organized sharing of resources and cooperation among users brings big benefi ts
  • Relaying is the fundamental form of such cooperation
  • The objective is to build a fully operational relay link!
  • Use a powerful new multi-antenna software-radio platform
  • One of the key aspects will be the synchronization among the nodes
  • Investigate decode-and-forward (DF) and amplify-and-forward (AF) schemes
  • Carry out extensive measurements to evaluate the performance of the link

Prerequisites: Basic background in signal processing and communications systems. Some knowledge of Labview and MATLAB is an advantage.

Project type: 70% Labview, 20% Measurements, 10% MATLAB

Supervisor: Pavle Belanovic

Low-Power Linear Time-Domain Prefilter at 1.76 Gsym/s

Semester Project

 

 

Description: Communication systems in the 60 GHz band has lead to a shy revival of the well known linear equalizer. This is mainly due to its ability to scale the power consumption with the complexity of the equalization problem. Such a behaviour is especially critical for mobile devices which often face varying channel conditions and have strict power constraints. Although the idea of linear equalizers is known for a long time the throughput required for 60 GHz operation and the power constraints are by orders different than for other communication systems. Goal of this project is to design and implement a high-speed low power FIR filter and characterize its area and power consumption of the implementation over the design space.

Areas: Digital Signal Processing, VLSI, EDA Design Flow, Power Simulation

Work Distribution: 10% Algorithm, 20% MATLAB, 40% VHDL, 30% Backend Design

Supervisor: Nicholas Preyss

IEEE802.11ad FFT-based Frequency Domain Equalizer for SC and

 

Master Thesis or Semester Project

Description: Communication at 60 GHz offers new unmatched opportunities for short-range communication. Channel bandwidths of multiple GHz and the possibility to operate multiple nodes on the same frequency allow practical transmission throughputs in the range of multiple Gb/s. The significant delay-spread and sampling rates of multiple GHz require new low-complexity approaches for channel equalization and circuit design. OFDM transmission offers excellent channel equalization properties in such environments. Unfortunately it suffers from several problems on the transmission side. By inserting pilot words the IEEE802.11ad standard enables the possibility to use the principle of frequency domain equalization also with single-carrier transmissions.

Goal of this project is to design and implement a high-speed reconfigurable FFT/IFFT-based filter block that can be used in OFDM and SC operation mode of IEEE802.11ad.

Areas: Digital Signal Processing, Simulation, VLSI

Work Distribution: 20% Algorithm, 30% MATLAB, 50% VHDL

Supervisor: Nicholas Preyss