Integrated silicon photonics and memristor dot product engine systems and methods

Patent metadata

Field	Value
Publication number	US-10845535-B2
Application number	US-201916708549-A
Country	US
Kind code	B2
Filing date	Dec 10, 2019
Priority date	Apr 23, 2018
Publication date	Nov 24, 2020
Grant date	Nov 24, 2020

Abstract

Systems and methods are provided for processing an optical signal. An example system may include a source disposed on a substrate and capable of emitting the optical signal. A first waveguide is formed in the substrate to receive the optical signal. A first coupler is disposed on the substrate to receive a reflected portion of the optical signal. A second waveguide is formed in the substrate to receive the reflected portion from the first coupler. A second coupler is formed in the substrate to mix the optical signal and the reflected portion to form a mixed signal. Photodetectors are formed in the substrate to convert the mixed signal to an electrical signal. A processor is electrically coupled to the substrate and programmed to convert the electrical signal from a time domain to a frequency domain to determine a phase difference between the optical signal and the reflected portion.

First claim

What is claimed is:

1. A method for processing one or more optical signals, comprising:
emitting an optical signal into free space from an optical source disposed on a substrate;
receiving the optical signal from the optical source via a first waveguide formed in the substrate;
receiving a reflected portion of the optical signal via a first coupler disposed on or formed in the substrate;
receiving the reflected portion of the optical signal from the first coupler via a second waveguide formed in the substrate;
mixing, in a second coupler formed in the substrate, the optical signal from the first waveguide and the reflected portion of the optical signal from the second waveguide to form a linearly mixed signal;
converting the linearly mixed signal to an electrical signal via one or more photodetectors formed in the substrate;
converting the electrical signal from a time domain to a frequency domain to determine a phase difference between the optical signal and the reflected portion of the optical signal;
wherein converting the electrical signal from the time domain to the frequency domain further comprises manipulating a first plurality of memristors of a first set of dot product engines to perform a digital Fourier transform to convert the electrical signal from the time domain to the frequency domain;
calculating a distance of an object in the free space from the substrate based on the phase difference between the optical signal and the reflected portion of the optical signal; and
processing, via a neural network deployed by a second set of dot product engines, a plurality of inputs including the distance of the object to determine a real time action,
wherein processing, via the neural network deployed by the second set of dot product engines, the plurality of inputs including the distance of the object to determine the real time action, further comprises manipulating a second plurality of memristors of the second set of dot product engines, the first set of dot product engines and the second set of dot product engines included in a processor flip chipped to the substrate.

2. The method of claim 1 , further comprising:
matching a mode of the optical signal traveling through the first waveguide with a mode of the reflected portion of the optical signal traveling through the second waveguide.

3. The method of claim 1 , further comprising:
forming, in part, a 3D point cloud based on the determined phase difference between the optical signal and the reflected portion of the optical signal.

4. The method of claim 1 , wherein the optical source is a quantum dot laser that, in operation, emits the optical signal substantially perpendicularly from a surface of the substrate, and the optical signal is an optical frequency comb.

5. A system for processing one or more LiDAR optical signals, comprising:
a silicon interposer;
a quantum dot laser disposed on the silicon interposer that, in operation, emits a LiDAR optical signal into free space in a direction perpendicular to a surface of the silicon interposer;
a first waveguide formed in the silicon interposer to receive the LiDAR optical signal from the quantum dot laser;
a first coupler disposed on or formed in the silicon interposer to receive a reflected portion of the LiDAR optical signal from an object in the free space at a distance from the system;
a second waveguide formed in the silicon interposer to receive the reflected portion of the LiDAR optical signal from the first coupler;
a second coupler formed in the silicon interposer to mix the LiDAR optical signal from the first waveguide and the reflected portion of the LiDAR optical signal from the second waveguide to form a linearly mixed signal;
one or more photodetectors formed in the silicon interposer to convert the linearly mixed signal to an electrical signal; and
a processor electrically coupled to the silicon interposer and programmed to
convert the electrical signal from a time domain to a frequency domain to determine a phase difference between the LiDAR optical signal and the reflected portion of the LiDAR optical signal; and
calculate the distance of the object from the system based on the phase difference between the LiDAR optical signal and the reflected portion of the LiDAR optical signal.

6. The system of claim 5 , wherein the processor comprises:
a sampling circuit communicatively coupled to the one or more photodetectors to sample the electrical signal to form a sampled electrical signal;
a first set of dot product engines communicatively coupled to the sampled circuit and including a first plurality of memristors arranged to receive the sampled electrical signal and to
perform a digital Fourier transform to convert the sampled electrical signal from the time domain to the frequency domain to determine the phase difference between the optical signal and the reflected portion of the optical signal, the first plurality of memristors being capable of non-volatile storage of values of the digital Fourier transform; and
calculate the distance of the object in the free space from the system based on the phase difference between the optical signal and the reflected portion of the optical signal; and

a second set of dot product engines communicatively coupled to the first set of dot product engines and an external interface to receive a plurality of inputs including the distance of the object in the free space, each dot product engine of the second set of dot product engines including a second plurality of memristors arranged to perform neural network processing to determine a real time action, the second plurality of memristors being capable of non-volatile storage of synaptic weights of a neural network.

7. The system of claim 6 , wherein the first waveguide and the second waveguide are configured such that a mode of the LiDAR optical signal traveling through the first waveguide and a mode of the reflected portion of the LiDAR optical signal traveling through the second waveguide are matched.

Assignees

• Hewlett Packard Entpr Dev Lp

Inventors

• Sharma Amit S
• Strachan John Paul
• Fiorentino Marco

Classifications

• G06N3/065

  Analogue means · CPC title

• G02B6/12004 Primary

  Combinations of two or more optical elements · CPC title

• G02B6/12002

  Three-dimensional structures · CPC title

• G01S17/36

  with phase comparison between the received signal and the contemporaneously transmitted signal · CPC title

• G01S7/4917

  superposing optical signals in a photodetector, e.g. optical heterodyne detection · CPC title