Master Thesis: Design of a Low Noise Amplifier for Neuromorphic Sensor Frontends

Das Fraunhofer EMFT betreibt angewandte Forschung an Sensoren und Aktoren für Mensch und Umwelt. Die rund 180 Mitarbeitenden an den drei Standorten München, Oberpfaffenhofen und Regensburg verfügen über langjährige Erfahrung und umfangreiches Know-how in Mikroelektronik und Mikrosystemtechnik. Das Technologieangebot des Instituts reicht von Halbleiterprozessen und MEMS-Technologien über 3D-Integration bis hin zur Folienelektronik. Diese Nano- und Mikrotechnologien sind die Basis für die anderen Kompetenzfelder des Fraunhofer EMFT: Sensorlösungen, sichere Elektronik und Mikropumpen. Gerade das interdisziplinäre Zusammenspiel dieser Kompetenzen bringt zukunftsweisende Lösungen hervor. Dadurch ist das Fraunhofer EMFT bestens aufgestellt, um zur Bewältigung der aktuellen Herausforderungen unserer Gesellschaft beizutragen.

The rapid development of artificial intelligence (AI) motivated moving AI closer to the edge or even in sensors to make daily human life more convenient. Conventional time-discrete signal processing flow, where the signal is sensed and quantized by a frontend and processed in a digital processing unit (DPU), is less practical due to significant power consumption required for the processing of the massive amount of raw signal data. In contrast to this, the event-based signal processing flow shows great potential to solve this problem because it only consumes power if a particular event occurs. It can be realized by combining an event-based analog frontend (AFE) and a spiking neural network (SNN) core.

In order to use the AFE for audio applications, a Low Noise Amplifier (LNA) is necessary. Its purpose is to act as a direct interface to a MEMS microphone and to amplify the signals provided by the microphone while introducing as little noise as possible. For increased dynamic range, the LNA shall be implemented with programmable gain settings. The main challenge is to implement the design within strict constraints on power consumption in the sub-µW range, since low-power edge AI is the intended use case.

Our group aims to develop analog/mixed-signal integrated systems for applications such as sensors, high-performance data converters, neuromorphic computing systems, and RF components. As a part of our group, you will perform a literature study on LNAs and MEMS microphones, design your own LNA with programmable gain settings and power consumption in the nW-range, and verify your design considering process variation and chip yield. Under certain conditions, publication of the results is possible and encouraged.

• Study state of the art of LNAs and MEMS microphones
• Select a suitable topology for LNA based on literature study
• Design and implement your LNA including programmable gain settings
• Documentation and presentation of the results

• Course of study in the field of Electrical Engineering or similar 
• Good knowledge of integrated analog circuits
• First experience with Cadence Virtuoso ADE is required
• Ability to work independently
• Good analytical and structured working style
• Ability to coordinate with an international team
• Good German or English language skills

We offer you a challenging task with responsibilities, room for creativity and many opportunities to gain new experience. You can expect an international, open-minded, and dynamic team, where you will be trusted from day one and encouraged to work independently. Your contributions will play a vital role in advancing our research in the field of neuromorphic computing.

Mr. Damian Panter
E-Mail: damian.panter@emft.fraunhofer.de
Telephone: +49 89 54759-233