Peak-Efficiency Non-Coherent ICs for Low-Power THz Communications

• Applicant: Professor Dr. Corrado Carta
• Subject Area: Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 566994190

The infrastructure enabling and supporting the modern digital society requires data centers and computing capabilities of unprecedented scales and subject to constantly increasing demands. The often-exponential nature of the growth in several key application areas – such as artificial intelligence and machine learning, cloud computing or more generally the global Internet traffic – poses challenges to the scalability, energetic sustainability and resilience of the computer architectures at play. At a basic level, the computing hardware at the core of a data center can be represented as a cluster of interconnected units, such as CPUs, GPUs, memory units of diverse access speeds and storage sizes. The maximal datarate of the transmission links between two units is often a performance bottleneck of the overall system, which results in the requirement of supporting the highest possible speeds. The data transmission links can be realized with different technologies, typically on-chip or on-board connections, with an increasing interest towards the realization of on-board optical links for the high datarate connection of units over physical distances compatible with the maximal size of the board. An attractive approach to enable additional degrees of freedom is the realization of chip-to-chip or board-to-board wireless links capable of performance comparable to short optical links in terms of datarates and energy per transferred bit. This proposal aims at developing a BiCMOS transceiver chipset to support data transmission over short wireless links required between two-chips in physically close proximities. To support both high datarates – in the order of 30-50Gbps – and high efficiency – in the order of about 12-16pJ per transferred bit – simple, non-coherent modulation schemes such as binary phase-shift keying (BPSK), differential phase-shift keying (DPSK) and on-off keying (OOK), will be studied against their impact on the efficiency of both transmitter and receiver. Complex modulated schemes such as 16/64/256-QAM or higher is more spectrum efficient, but the transmitter suffers from a low energy efficiency since it must operate in its linear region. In its receiver, a coherent phase-locked source is required, significantly impacting the receiver energy efficiency. An expected outcome of this effort is that efficiency will result from a drastic reduction of spectral efficiency, which can be enabled by the non-coherent operation over a THz carrier in the order of 500 GHz, as this results in a natural RF bandwidth of approximately 50 GHz and can support data rates of 30 Gbps or more. Such high frequency of operation is within the reach of the SiGe HBT transistors available in the SG13G3 technology of IHP, which provide a nominal fmax of about 700GHz: this would not be possible with any of the plain Si CMOS technologies, as their fmax saturates around 400GHz and does not benefit from scaling further than already achieved by the most advanced nodes.

• DFG Programme: Research Grants
• International Connection: South Korea
• Partner Organisation: National Research Foundation of Korea, NRF
• Co-Investigator: Dr.-Ing. Andrea Malignaggi
• Cooperation Partner: Professor Dr. Munkyo Seo