The significance of fibre optic networks in the digital age cannot be overstated. With the increasing demand for higher bandwidths and lower latencies across various industries, the need for improved data transmission capabilities is paramount. This is especially crucial as we look towards the future of 6G networks, which will require exceptionally high data transmission speeds.
To address the challenge of deteriorating signal quality as data rates escalate, Fraunhofer IIS is exploring innovative solutions beyond traditional digital signal processors. The SpikeHERO project aims to revolutionize fibre optic systems by introducing a unique AI processor architecture that merges optical and electrical spiking neural network chips.
These neural networks will play a pivotal role in continuously monitoring communication channels, analyzing signals, and rectifying any interference at the receiver end using control parameters. By maintaining optimal signal quality, the project aims to unlock new possibilities for enhancing data rates in fibre optic networks.
The project’s ultimate goal is to triple the bandwidth capacity from 10 GHz to 30 GHz, while simultaneously reducing latency from 10 microseconds to under 6 nanoseconds. Additionally, the energy consumption is expected to decrease significantly from 7-10 watts to a mere 1-2 watts, making the system more efficient and sustainable in the long run.
Spiking neural networks (SNNs) serve as the foundation for this groundbreaking project, offering a promising avenue for advancements in artificial intelligence. Modeled after the human brain’s functioning, SNNs process information through pulses only when a certain threshold of relevance is surpassed. This makes them an ideal candidate for AI applications requiring real-time responsiveness and energy efficiency.
By combining both optical and electrical SNN chip technologies, SpikeHERO aims to leverage the unique advantages of each approach. The project will utilize the SENNA chip, developed by Fraunhofer IIS and Fraunhofer EMFT, for the electrical SNN chip component, with ongoing work on a second-generation chip promising even higher spike rates and lower energy consumption.
