Optical Routing Layer for Full-Mesh Interconnects

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Implementing a full-mesh, point-to-point interconnect topology in multi-chip modules (MCMs) using silicon optical waveguides results in a large number of optical-waveguide crossings, leading to channel loss and crosstalk, which significantly impacts performance and energy efficiency.

Innovation Solution

A multi-chip module with a single optical routing layer that includes N optical waveguides optically coupled to N chips and a cyclic de-multiplexer, which routes optical signals without optical waveguide crossings, using an array-waveguide-grating or echelle-grating wavelength router to provide all-to-all connectivity and reduce the need for non-blocking dedicated optical channels.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a full-mesh, point-to-point interconnect topology is implemented using silicon optical waveguides, then all-to-all connectivity among chips is achieved, but the number of optical-waveguide crossings increases significantly, leading to channel loss and crosstalk

Engineering Contradiction:
ImproveconnectivityVSAvoidchannel loss
Core Design Contradiction:
Adaptability or versatilityVSLoss of energy

Solution Approach 1:

The patent transitions from planar 2D waveguide routing to 3D vertical routing using multiple layers. Optical waveguides are routed in different vertical layers (first layer and second layer) to avoid crossings, utilizing the third dimension (height) to resolve conflicts that cannot be solved in a single plane.

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

Solution Approach 2:

The patent implements nested routing where waveguides in one layer pass through or alongside structures in another layer without interaction. The first and second optical waveguide layers are vertically stacked, with each layer containing complete routing paths that would otherwise require crossings in a single layer.

Inventive Principle:
Principle #7Nested doll (Nesting)

2Object-affected harmful factors

If surface normal optical proximity couplers (OPxCs) and dual-layer routing are used to eliminate optical-waveguide crossings, then connectivity is maintained, but the number of OPxC hops increases to four per WDM link, adding 12 dB to link loss

Engineering Contradiction:
ImprovecrosstalkVSAvoidlink loss
Core Design Contradiction:
Object-affected harmful factorsVSLoss of energy

Solution Approach 1:

The patent extracts the wavelength routing function from the proximity coupler and relocates it to dedicated wavelength router components (AWG or echelle grating) positioned at the edges of the MCM. This removes the need for multiple OPxC hops, reducing link loss while maintaining the ability to route different wavelengths to different destinations.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces wavelength routers as intermediary devices that perform wavelength-based routing externally to the main MCM structure. These routers act as mediators that receive optical signals, identify their destination wavelength, and route them through the appropriate vertical layer without requiring multiple OPxC coupling stages.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Reliability

If optical-waveguide crossings are eliminated using special crossing designs, then channel loss and crosstalk are reduced, but the complexity and number of components increase

Engineering Contradiction:
Improvesignal qualityVSAvoidrouting structure
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

Instead of using complex planar crossing structures, the patent resolves crossings by moving to a third dimension. Simple vertical vias connect waveguides between layers, replacing complex 2D crossing patterns with straightforward 3D routing that is both simpler and more reliable.

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

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

This solution reduces the number of optical-waveguide crossings, minimizing channel loss and crosstalk, thereby enhancing the energy efficiency and scalability of the interconnect network while maintaining full-mesh, point-to-point connectivity among chips.

Implementation Method 1

a cyclic de-multiplexer, optically coupled to the N optical waveguides, that routes optical signals among the N optical waveguides without optical waveguide crossing in the optical routing layer. Note that the cyclic de-multiplexer may include: an array-waveguide-grating (AWG) wavelength router and/or an echelle-grating wavelength router.

Methodology Applied
Scientific EffectDiffraction: Diffraction

Data Source

PatentUS9316784B2Single- layer full-mesh, point-to-point network
Publication Date: 2016.04.19 ORACLE INT CORP
  • US9316784B2 patent drawing
  • US9316784B2 patent drawing
  • US9316784B2 patent drawing

AI summary

An MCM may include a single optical routing layer that provides point-to-point connectivity among N chips in the MCM, such as all-to-all connectivity or full-mesh point-to-point connectivity. Moreover, the optical routing layer may include: N optical waveguides optically coupled to the N chips and a cyclic de-multiplexer, optically coupled to the N optical waveguides, that routes optical signals among the N optical waveguides without optical-waveguide crossing in the optical routing layer. For example, the cyclic de-multiplexer may include: an array-waveguide-grating (AWG) wavelength router and/or an echelle-grating wavelength router.