Optical Shuffle Network Lenslet Arrays for Low-Latency Interconnects

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Computational systems face limitations due to long wires that cause signal delay and high power consumption, and existing optical interconnects fail to achieve high density and scalable interconnects with minimal delay and low power consumption.

Innovation Solution

An optical network with densely spaced transmitters and receivers, utilizing lenslet arrays for collimating and focusing beams to enable efficient signal routing between collected and dispersed arrays of tiles, allowing for high-density, low-power, and scalable interconnects through free-space optical trains and optoelectronic transducers.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If long wires are used to interconnect processing elements, then the system can accommodate larger computational systems, but the signal delay and power consumption increase

Engineering Contradiction:
Improvesystem sizeVSAvoidsignal delay
Core Design Contradiction:
Area of stationary objectVSLoss of time

Solution Approach 1:

The patent replaces electrical wire-based interconnects with optical interconnects using light propagation. This substitution eliminates the quadratic delay growth characteristic of electrical wires by using optical signals that can traverse longer distances with minimal delay, directly addressing the signal delay problem while maintaining system scalability

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces a third dimension (vertical stacking with lenslet arrays) to achieve long-range connections without requiring long horizontal wires. By stacking processing elements vertically and using optical lenses to establish long-range connections between layers, the system achieves scalable interconnectivity without the delay penalties of long planar wires

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Area of stationary object

If long wires are used to interconnect processing elements, then the system can accommodate larger computational systems, but the power consumption increases

Engineering Contradiction:
Improvesystem sizeVSAvoidpower consumption
Core Design Contradiction:
Area of stationary objectVSUse of energy by stationary object

Solution Approach 1:

The patent replaces electrical signal transmission through long wires with optical signal transmission. This substitution dramatically reduces power consumption because optical signals experience minimal attenuation over distance and do not require the continuous buffering and signal regeneration that electrical wires require, directly addressing the linear power growth problem

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Speed

If optical waveguides are used for interconnections, then high speed optical interconnects are achieved, but the number of channels cannot scale with chip area

Engineering Contradiction:
Improveinterconnect speedVSAvoidchannel scalability
Core Design Contradiction:
SpeedVSAdaptability or versatility

Solution Approach 1:

The patent moves from planar waveguide interconnects to three-dimensional free-space optical interconnects using stacked lenslet arrays. This dimensional transition allows channels to scale with chip area by adding vertical layers, with each layer providing additional interconnect channels while maintaining high optical speeds

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent divides the optical interconnect system into discrete lenslet array layers, with each layer providing a set of independent optical channels. This segmentation allows the system to scale by adding more layers, with each layer contributing additional channels proportional to the chip area

Inventive Principle:
Principle #1Segmentation

4Adaptability or versatility

If space-variant interconnection patterns are implemented, then routing flexibility is improved, but the interconnect density decreases

Engineering Contradiction:
Improverouting flexibilityVSAvoidinterconnect density
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent implements dynamically reconfigurable optical interconnects using programmable optical elements in the lenslet arrays. This allows the system to adapt connection patterns in real-time based on computational requirements while maintaining high density, resolving the trade-off between routing flexibility and interconnect density

Inventive Principle:
Principle #15Dynamics

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

The solution enables effective routing of signals with minimal delay and low power consumption, maintaining high density interconnects while scaling with complex computational systems, suitable for applications like artificial neural networks and systolic arrays.

Implementation Method 1

An originating lenslet array having two parallel layers of lenslets is parallel and adjacent the array of originating tiles. Each lenslet pair within the lenslet array pair corresponds to a tile. The lenslet array pair collimates beams from transmitters such that the location of a transmitter within its tile determines the direction of its resulting collimated beam.

Methodology Applied
Scientific EffectCollimation: Lens

Implementation Method 2

Similarly, a terminating lenslet array focuses collimated beams to receivers such that the direction of a collimated beam determines which receiver it is focused on.

Methodology Applied
Scientific EffectFocusing: Lens

Data Source

PatentUS11444695B2Optical shuffle computation network
Publication Date: 2022.09.13 FATHOM RADIANT PBC
  • US11444695B2 patent drawing
  • US11444695B2 patent drawing
  • US11444695B2 patent drawing

AI summary

Optical communication system communicates between an array of originating tiles and an array of terminating tiles. Each array is associated with a lenslet array, such as a two-layer array which has two layers of lenslets. Each originating tile has an array of transmitters and each terminating tile has an array of receivers. Each tile is associated with a common lenslet or lenslet pair. A beamlet from a representative transmitter passes through the lenslet pair adjacent to its tile to become a collimated beam whose angle is related to the location of the transmitter. The collimated beam passes through the receiver lenslet pair adjacent to the tile containing the receiver associated with the representative transmitter, and is focused onto that receiver by that lenslet pair. The system may operate in the reverse direction, wherein the transmitters are transmitter-receivers, the receivers are receiver-transmitters, and a beam from a receiver-transmitter is directed to its corresponding transmitter-receiver.