The new chip does away with the need to use specific decoding hardware and could increase efficiency in gaming systems, 5G networks, and the Internet of Things.

All data sent over the internet can be affected by noise, including electromagnetic interference from a Bluetooth or microwave device. Data are encoded so that, when they reach their destination, a decoding algorithm can undo any adverse effects and retrieve the original data.

Most error-correcting codes have been created with decoding algorithms since the 1950s. Each code was designed to correspond with a specific, complex decoding algorithm. This often required dedicated hardware.

Researchers from MIT, Boston University, and Maynooth University in Ireland created the first silicon chip capable of decoding any code with maximum accuracy. They used a universal algorithm called Guessing random additive noise decoding (GRAND). GRAND eliminates the need to have multiple complex, computationally complicated decoders. This allows for greater efficiency, which could be used in augmented and virtual realities, gaming, 5G networks, and other connected devices that require high data volumes and minimal delay.

Muriel Medard is the Cecil H. and Ida Green professor in the Department of Electrical Engineering and Computer Science at MIT. The research was co-authored at MIT by Wei Ann and Amit Solomon, both graduate students at MIT; Rabia Tuga Yazicigil, assistant professor of electrical and computing engineering at Boston University; Arslan R. Bansal and Vaibhav Bansal, both graduate students at Boston University; Ken R. Duffy director of the Hamilton Institute at the National University of Ireland Maynooth; and Kevin Galligan graduate from Maynooth graduate. Next week, the research will be presented at European Solid-States Device Research and Circuits Conference.

Concentrate on the noise.

These codes can be considered redundant hashes, a series of 1s and 0s added to the end date. A specific codebook contains the rules that create this hash.

Noise or energy that disrupts a signal and is often generated by electronic devices can affect encoded data traveling over a network. The decoding algorithm uses the structure of the hash and the coded data to determine what they are.

GRAND uses noise patterns to determine the original information. Instead of guessing what noise caused the message to be sent, GRAND does this by deducing the noise pattern. GRAND generates a sequence of noise sequences in the most likely order, subtracts them from the received data, and then checks to ensure the codeword is in a suitable codebook.

Although the noise may appear random, the algorithm can use probabilistic structures to predict what it could be.

It is similar to troubleshooting. The mechanic won’t begin by mapping out the car. Instead, they ask, “What is the most likely to go wrong?” Maybe it’s just gas. What if that fails? Medard suggests that the battery might be dead.

Novel hardware

The GRAND chip has a three-tiered structure. It starts with the most straightforward solutions in the first stage and then moves on to more complicated noise patterns in the second and third stages. Each step can operate independently, increasing the system’s throughput and saving power.

It can also switch between two codebooks seamlessly. The device has two static random-access memory chips that can crack codewords and the other that loads a new book, then switch seamlessly to decoding.

Researchers tested the GRAND chip, and it could decode any moderate redundancy code up to 128 bits long with just a microsecond latency.

Medard and her colleagues had previously demonstrated the success of the algorithm. However, this new work demonstrates the effectiveness of GRAND in hardware.

Medard states that researchers had to abandon all preconceived notions to develop hardware for the new decoding algorithm.

“We could not go out and reuse items that were already done. It was like a whiteboard. It was a challenge to think through every component. It was a journey that required a lot of reflection. She says that when we make our next chip, we will discover things we didn’t do out of habit or assume that we could do better.

The chip of the future

GRAND uses codebooks only for verification. The chip can be used with legacy codes and codes that have yet to be introduced.

Communication companies and regulators need help to agree on which codes to use in 5G’s implementation. Finally, regulators used two traditional codes for 5G infrastructure in different circumstances. Medard believes that GRAND could replace the need for rigid standardization shortly.

GRAND chips could even lead to new opportunities in coding.

“People approach programming with awe for reasons I don’t know. It is like black magic. Because the process is mathematically difficult, people use codes that are already in existence. She says, “I hope this will change the discussion, so it isn’t standards-oriented. It will allow people to use existing and create new codes.”

Medard and her colleagues plan to address the soft detection problem with a retooled GRAND chip—smooth detection results in less precise data.

They also plan to test GRAND’s ability to crack more extended codes and adjust silicon chip structure to increase its energy efficiency.

The Battelle Memorial Institute, Science Foundation of Ireland, funded the research.