Lawrence Berkeley National Laboratory unveils ASIC chip failure bypass system
BERKELEY, Calif.–(BUSINESS WIRE)–Scientists at Lawrence Berkeley National Laboratory (Berkeley Lab) have developed a technique to automatically bypass failed application-specific integrated circuits (ASICs) deployed in large networks by dynamically rerouting them around them. In the new system, called “Hydra I/O”, each ASIC chip is connected to its nearest neighbors on all four sides, unlike traditional networked systems where chip-to-chip data paths are typically wired. This configuration allows data to be redirected to adjacent neighbors in the event of an individual chip or multiple chip failure. The system includes software that dynamically redirects readback commands or data in such chip failure cases in a manner that is transparent to the user.
“To cope with the possibility of chip failure, current systems require the construction of significant redundancy in the event of chip failure,” said Carl Grace, group leader of integrated circuits and chip detectors. semiconductors from Berkeley Lab. “The Hydra I/O system eliminates the need for expensive redundant systems, and it’s highly resilient by design.”
The system was originally developed by Berkeley Lab for the Deep Underground Neutrino Experiment (DUNE) in which the custom chips that instrument the particle detectors will be placed in a sealed cryostat filled with cryogenic liquid located underground. Since the chips are physically inaccessible, the scientists developed a way to deal with inevitable chip failures during the experiment’s expected 30-year lifespan. The system was named “Hydra I/O” because when a single point fails, the rest of the system continues to operate. This feature is similar to how the Internet automatically redirects traffic around failing servers.
Lab scientists Dan Dwyer and Grace, Principal Investigators for Hydra I/O, quickly recognized its potential applications in other environments where repairing or replacing ASIC arrays is equally difficult or expensive, such as in space, underwater, in remote locations or in 5G. systems that include multiple phased arrays.
“Repairing or replacing failed ASIC chips is especially difficult when the chips are inaccessible, such as in the DUNE project,” said Dwyer, leader of Berkeley Lab’s Neutrino Physics group. “It was essential for the scientific mission that the tiles of the neutrino detectors be highly resistant. In our current prototype detector, we’ve encountered about a dozen chip failures, and Hydra I/O has made those failures non-events.
The system is highly scalable. To date, the Berkeley Lab team has successfully operated a prototype detection system containing 1,600 ASICs paired with nearly 100,000 sensors. The Berkeley Lab team plans to build hundreds of thousands of ASICs for the DUNE project; further scaling up of the technology is entirely feasible.
For more information on licensing this technology, contact [email protected]
Founded in 1931 on the belief that the greatest scientific challenges are best met by teams, Lawrence Berkeley National Laboratory and its scientists have been awarded 13 Nobel Prizes. Today, Berkeley Lab researchers are developing sustainable energy and environmental solutions, creating useful new materials, pushing the boundaries of computing, and probing the mysteries of life, matter, and the universe. Scientists around the world rely on the laboratory’s facilities for their own scientific discovery. Berkeley Lab is a multi-program national laboratory, operated by the University of California for the US Department of Energy’s Office of Science.
The DOE’s Office of Science is the largest supporter of basic physical science research in the United States and works to address some of the most pressing challenges of our time. For more information, please visit science.energy.gov.