Speaker
Description
As particle physics sensor arrays expand in size and complexity, single-point failures pose a significant obstacle to achieving the next generation detectors at scale. Failures can originate from manufacturing defects or be induced by radiation damage or cryogenic temperatures during operation.
Artificial Intelligence paradigms in dynamic mesh communication, both on-chip and in multi-chip systems, hold the potential to enable highly fault-tolerant sensors and systems, addressing both manufacturing yield and the extreme radiation or temperature environments encountered in high-energy particle physics detectors. The same intelligent network architecture aimed at fault tolerance can be leveraged for lower power and higher bandwidth readout.
Radiation-induced single-event effects (SEE) are a major concern for hadron collider experiments, such as those conducted at the Large Hadron Collider (LHC). Cryogenic system reliability is a significant concern for large liquid noble element detectors, such as DUNE. Wafer yield is a major concern for emerging wafer-scale monolithic sensors that will eventually be utilized at future electron-positron Higgs factory colliders.
Fault-tolerance, low latency, and low-power sensor readout are crucial for all of the above. In this project, we re-envision the nature of sensor readout and operation to address these challenges through on-chip intelligence. This presentation will describe the current branches of the NIFTY project: dynamic network routing for the LArPix chip employed in the DUNE LAr TPC, a network-based digital architecture for future 4D Pixel detectors, and spiking neural network-based network routers.