“Signal Processing in the AI era” was the tagline of this year’s IEEE International Conference on Acoustics, Speech and Signal Processing, taking place in Rhodes, Greece.
In this context, Brent de Weerdt, Xiangyu Yang, Boris Joukovsky, Alex Stergiou and Nikos Deligiannis presented ETRO’s research during poster sessions and oral presentations, with novel ways to process and understand graph, video, and audio data. Nikos Deligiannis chaired a session on Graph Deep Learning, attended the IEEE T-IP Editorial Board Meeting, and had the opportunity to meet with collaborators from the VUB-Duke-Ugent-UCL joint lab.
Featured articles:
- De Weerdt, B. et al., “Designing Transformer Networks for Sparse Recovery of Sequential Data Using Deep Unfolding” (https://ieeexplore.ieee.org/abstract/document/10094712)
- Yang X. et al., “Relevance Propagation through Deep Conditional Random Fields” (https://ieeexplore.ieee.org/abstract/document/10095075)
- Stergiou A. et al., “Play it back: Iterative attention for audio recognition” (https://ieeexplore.ieee.org/abstract/document/10096532, developed at the University of Bristol)
On May 9 2022 at 14.00 Zhiwei Zong will defend his PhD entitled “DESIGN OF VCOs AND PAs IN 22 NM FD-SOI FOR 5G MM-WAVE COMMUNICATIONS”.
Everybody is invited to attend the presentation online via this TEAMS link.
The use of spectrum in the millimeter-wave (mm-Wave) frequency range is considered as a key enabler to continue the insatiable demand for increased wireless data capacity. This spectrum will be adopted in the 5G wireless communication standard. To obtain a high integration degree for the implementation of 5G mm-Wave transceivers, advanced CMOS is the preferred technology. The higher operating frequency, compared to 4G, poses more design challenges on the key building blocks of a transceiver. This PhD thesis focuses on the design of the two key building blocks in a 5G mm-Wave transceiver, namely a voltage-controlled oscillator (VCO) and a power amplifier (PA). All building blocks designed in this PhD work are operating in the 20-30 GHz frequency region. All building blocks have been designed in a 22nm fully-depleted silicon-on-insulator (FD-SOI) CMOS technology.
First, a modified transformer-feedback VCO (TF-VCO) with a sourcebridging capacitor (Cs) is introduced. Thanks to the use of Cs, the phase noise (PN) in the 1/f2 and 1/f3 regions are both improved compared to earlier published TF-VCOs. The origin of the PN improvement by the use Cs is explained in this thesis. It is seen that with Cs we can improve the symmetry of the waveform of the voltage over the tank of the VCO. Also, with Cs the effective quality factor of the transformer can be increased, which also reduces phase noise. These theoretical investigations are proven with measurement results. With a second design, an LC-VCO design, another key design challenge is tackled, namely the suppression of flicker noise upconversion. A 22-29GHz voltage-biased LC-VCO is designed and implemented to suppress this flicker noise upconversion by using a flicker noise filtering technique. A self-coupled inductor and a common-centroid capacitor bank layout are proposed in this design to guarantee a good flicker noise suppression over the frequency tuning range.
Next, two 28GHz PAs are designed and implemented for 5G mm-Wave communications. The first PA focuses on generation of a high output power (Pout) with a high linearity. This is achieved in a first design that uses a two-way current combiner and an output stage that uses stacking of transistors. The stack of three transistors used in this design enables the generation of a high output power without overstressing the core devices. The second PA focuses on the power back-off (PBO) efficiency enhancement. This is important for communication with a high spectral efficiency: high-order modulation requires to operate at a relatively large back-off from the saturation level. The design is based on the Doherty architecture. By merging lumped passive components into a transformer, a transformer-based Doherty PA with a compact power combiner is obtained, achieving Doherty load modulation with a compact footprint. This design has the highest power density and ITRS PA figure-of-merit (FOM) among the published mm-wave Doherty PAs.
- Jan Cornelis becomes HoD.
- First Patent on an optical switch (IMOS) is filed.
- ETRO becomes co-founder of the International Postgraduate Course on Biomedical Engineering and Medical Physics (University of Patras) – Erasmus/Tempus.
- ETRO’s first EU framework project, BIOLAB: An Integrated Biomedical Laboratory – AIM- Advanced Informatics in Medicine.
- Data collection, drafting and validation of the first ever table of sizes for the Belgian apparel industry.
Pengpeng Hu has successfully defended his PhD in public on 22/09. He has obtained the “Grootste Onderscheiding”. In addition, the jury has unanimously decided to felicitate Pengpeng for his achievements in terms of scientific publications and industrial valorisation.
With this, he continues the tradition in his team of his promotor Adrian Munteanu and colleague Nikolaos Delligiannis, whom both also recieved this exceptional result for their respective PhD theses.
The tuition fees are described here: https://student.vub.be/en/student-administration#tuition-fee
PROJECT: INTOWALL: detection of leaks and isolation in walls
Our buildings generate 35% of the CO2-emissions and 40% of the energy use. 75% of our buildings need to be energetically renovated. In order to do so, one needs to know what the status is of the current isolation, and if there are any water leaks. The Vrije Universiteit Brussel researches with a consortium of partners towards a unique, patented technique to look into the walls without breaking them. With a compact and mobile radar system you can “read” your walls. This project is part of the Smart Hub Cleantech, climate ambitions of the province and the communes.
Partners: Vrije Universiteit Brussel i.s.m. WTCB, Green Energy Park, BAM, ING, Flux50 en Alter Reim.
After the conservation and restoration project for Jan Van Eyck’s masterpiece, the Royal Institute for Cultural Heritage documented both sides of the painting with hundreds of macro photography photos. Universum Digitalis then algorithmically assembled those images to produce gigapixel images of the artwork. The painting was previously digitized in 2015 using the same scientific protocol. Universum Digitalis seamlessly aligned both acquisitions, enabling a unique pixel-level comparison before and after restoration.
Comparison of the front and backside before and after restoration.
The restored painting will be exhibited at The Louvre Museum during the exhibition “Revoir Van Eyck – La Vierge du chancelier Rolin” from March 20th to June 17th, 2024. In parallel with the exhibition’s opening, the gigapixel images produced by Universum Digitalis will be made publicly accessible on the Closer to Van Eyck website.
https://www.louvre.fr/en/what-s-on/exhibitions/a-new-look-at-jan-van-eyck
