CN112506265B - Optical computing device and computing method - Google Patents

Optical computing device and computing method Download PDF

Info

Publication number
CN112506265B
CN112506265B CN202011270075.3A CN202011270075A CN112506265B CN 112506265 B CN112506265 B CN 112506265B CN 202011270075 A CN202011270075 A CN 202011270075A CN 112506265 B CN112506265 B CN 112506265B
Authority
CN
China
Prior art keywords
optical signals
waveguide array
waveguides
array
layer
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011270075.3A
Other languages
Chinese (zh)
Other versions
CN112506265A (en
Inventor
穆进峰
董晓文
吴彤宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Original Assignee
Huawei Technologies Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co Ltd filed Critical Huawei Technologies Co Ltd
Priority to CN202011270075.3A priority Critical patent/CN112506265B/en
Publication of CN112506265A publication Critical patent/CN112506265A/en
Application granted granted Critical
Publication of CN112506265B publication Critical patent/CN112506265B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06E—OPTICAL COMPUTING DEVICES; COMPUTING DEVICES USING OTHER RADIATIONS WITH SIMILAR PROPERTIES
    • G06E3/00—Devices not provided for in group G06E1/00, e.g. for processing analogue or hybrid data
    • G06E3/001—Analogue devices in which mathematical operations are carried out with the aid of optical or electro-optical elements
    • G06E3/005—Analogue devices in which mathematical operations are carried out with the aid of optical or electro-optical elements using electro-optical or opto-electronic means
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06E—OPTICAL COMPUTING DEVICES; COMPUTING DEVICES USING OTHER RADIATIONS WITH SIMILAR PROPERTIES
    • G06E3/00—Devices not provided for in group G06E1/00, e.g. for processing analogue or hybrid data
    • G06E3/008—Matrix or vector computation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
    • G06F7/38—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
    • G06F7/48—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state device; using unspecified devices
    • G06F7/544—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state device; using unspecified devices for evaluating functions by calculation
    • G06F7/5443—Sum of products
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
    • G06F7/38—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation
    • G06F7/48—Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state device; using unspecified devices
    • G06F7/57—Arithmetic logic units [ALU], i.e. arrangements or devices for performing two or more of the operations covered by groups G06F7/483 – G06F7/556 or for performing logical operations
    • G06F7/575—Basic arithmetic logic units, i.e. devices selectable to perform either addition, subtraction or one of several logical operations, using, at least partially, the same circuitry

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Physics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Optics & Photonics (AREA)
  • Algebra (AREA)
  • Nonlinear Science (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Optical Integrated Circuits (AREA)
  • Optical Modulation, Optical Deflection, Nonlinear Optics, Optical Demodulation, Optical Logic Elements (AREA)

Abstract

An optical computing device and a computing method, wherein a first layer waveguide array can receive a first group of optical signals. The second layer waveguide array is positioned on a different plane from the first layer waveguide array, receives the first group of optical signals from the first layer waveguide array based on the waveguide coupling principle, and the modulator array can modulate the first group of optical signals transmitted by the second layer waveguide array based on the second group of data and output a plurality of second optical signals. The modulator array modulates the first set of optical signals based on a second set of data, and then the combined waveguide array converges the plurality of second optical signals. Compared with a light splitting mode of a light splitter, the second-layer waveguide array can receive the first group of optical signals from the first-layer waveguide array in a lossless manner based on the waveguide coupling principle between waveguides, excessive noise can not be introduced into the first group of optical signals, and accuracy of multiply-accumulate calculation realized by utilizing the optical signals is improved.

Description

Optical computing device and computing method
Technical Field
The present application relates to the field of communications technologies, and in particular, to an optical computing device and a computing method.
Background
With rapid development and rapid increase of data volume in the field of artificial intelligence (ARTIFICIAL INTELLIGENCE, AI), various deep learning algorithms are widely used in the fields of image processing, speech recognition, radar signal processing, coherent optical communication and the like. Multiply-accumulate computation is the basic operation of various deep learning algorithms. At present, the optical signal can be used for realizing multiply-accumulate operation to obtain a calculation result, and the mode for realizing multiply-accumulate calculation through the optical principle has the advantages of ultra-wide calculation bandwidth and low energy consumption, and the calculation rate of the multiply-accumulate calculation is accelerated to a certain extent.
Among the numerous optical computing devices that utilize optical signals to implement multiply-accumulate, on-chip integrated optical computing devices are attracting attention, and the on-chip integrated optical computing devices can implement chip-level integration of each functional module, thereby improving system stability, reducing device size, and also reducing energy consumption ratio of the device.
However, the existing on-chip integrated optical computing device has a certain limitation, taking a cross array (crossbar) optical multiplication and addition device as an example, inputting an optical signal into the crossbar array, wherein the crossbar array comprises a waveguide which is transversely arranged and a waveguide which is longitudinally arranged, and a beam splitter is arranged at the cross part of the waveguide which is transversely arranged and the waveguide which is longitudinally arranged. The optical signals of the crossbar can be transmitted through the longitudinally arranged waveguides, the beam splitter splits the transmitted optical signals in the longitudinally arranged waveguides, and the optical signals generated by the splitting enter the modulator to be modulated, so that multiplication operation is realized. The modulator couples the optical signal after multiplication into the transversely arranged waveguides. Each waveguide which is transversely arranged can collect the optical signals after multiplication operation, and addition operation is completed.
In the crossbar optical multiplication and addition device, the transversely arranged waveguides and the longitudinally arranged waveguides are crossed, crosstalk exists among signals among the waveguides, noise is easily introduced into the optical signals after light splitting, and the accuracy of multiplication and addition operation is affected.
Disclosure of Invention
The application provides an optical computing device and a computing method, which are used for reducing noise introduced in the process of realizing multiply-accumulate computation by utilizing optical signals and improving the accuracy of the multiply-accumulate computation.
In a first aspect, the present application provides an optical computing device comprising a first layer waveguide array, a second layer waveguide array, a modulator array, and a beam combining waveguide array.
In the optical computing device, a first layer waveguide array may receive the first set of optical signals, the first set of optical signals including a plurality of optical signals indicative of a first set of data, the first layer waveguide array including a plurality of first waveguides, each first waveguide being operable to transmit one optical signal of the first set of optical signals. The second layer waveguide array is located in a different plane from the first layer waveguide array, and the second layer waveguide array may receive the first set of optical signals from the first layer waveguide array based on a waveguide coupling principle, which is not limited herein, and may receive one first set of optical signals or may receive a plurality of first sets of optical signals based on the waveguide coupling principle.
The modulator array may modulate the first set of optical signals transmitted by the second layer waveguide array based on the second set of data to output a plurality of second optical signals. If a plurality of first sets of optical signals are transmitted in the second layer waveguide array, the modulator array may modulate the plurality of first sets of optical signals based on a plurality of second sets of data, and in particular, the modulator array modulates a first set of optical signals based on a second set of data.
The beam combining waveguide array can converge a plurality of second optical signals output by the modulator array, and the second optical signals are converged into optical signals with a shorter distance.
In the device provided by the application, the second-layer waveguide array can receive the first group of optical signals from the first-layer waveguide array based on the waveguide coupling principle, so that the light splitting effect is realized. Compared with a beam splitter which is used for splitting, the second-layer waveguide array can receive the first group of optical signals from the first-layer waveguide array in a lossless manner based on the waveguide coupling principle between waveguides, excessive noise can not be introduced into the first group of optical signals, and accuracy of multiply-accumulate calculation through the optical signals is improved.
In one possible design, the thickness of the plurality of first waveguides in the first region is smaller than the thickness in other regions than the first region, and the plurality of first waveguides overlap the plurality of second waveguides in the first region, that is, there is an overlap region between the first and second waveguides, that is, the first region.
In the device provided by the application, the thickness of the plurality of first waveguides in the overlapping area is thinned, so that the first group of optical signals transmitted on the plurality of first waveguides can be coupled to the plurality of second waveguides, and the light splitting effect is achieved.
In one possible design, the combined beam waveguide array includes multiple layers of beam waveguides, each layer of beam waveguides configured to converge a portion of the plurality of second optical signals.
In the device provided by the application, the beam combining waveguide array adopts a layered design mode, and the plurality of second optical signals can be converged by utilizing the multi-layered beam combining waveguide, so that the plurality of second optical signals are converged into a plurality of second optical signals which are closer, and the number of the converged second optical signals can be increased by adopting the layered design mode.
In one possible design, each layer of beam waveguides includes a plurality of superlattice waveguides. In view of the fact that the distance between the superlattice waveguides can reach the sub-wavelength size, the structure is compact, namely each layer of the superlattice waveguides can comprise a large number of superlattice waveguides, the second optical signals with a large number of numbers can be collected, and the number of the second optical signals which can be collected is increased.
In one possible design, the optical computing device further includes a detector array, where the detector array may detect a plurality of second optical signals converged by the beam combining waveguide array, and the detector array may convert the plurality of second optical signals into electrical signals, implement summation of the signals, and detect an intensity of the optical signals.
In one possible design, the modulator array includes a plurality of modulators including at least one of the following modulators: lithium niobate thin film modulator, semiconductor optical amplifier, thermal modulator, MZI. The modulator array can comprise a plurality of modulators of the same type, and also can comprise a plurality of modulators of different types, and the mode of constructing the modulator array is more flexible.
In one possible design, the optical computing device may further include a light source module, where the light source module may generate the first set of optical signals, and the light source module may generate the first set of optical signals more conveniently and quickly so as to implement the multiplication and addition operation based on the first set of optical signals.
In one possible design, the light source module includes a plurality of lasers, each for generating a first optical signal. The generation of each first optical signal may be independently controlled, generated by one laser.
In one possible design, the light source module includes a laser and a beam splitter, the laser can generate an original light signal, the beam splitter can split the original light signal to generate a plurality of first light signals, the laser and the beam splitter cooperate to divide the original light signal into the plurality of first light signals, and the light source module has a relatively simple structure and reduces the difficulty of construction.
In a second aspect, the present application provides a computing method, and the beneficial effects can be seen from the description related to the first aspect, which is not repeated here. The method is performed by an optical computing device that includes a first layer waveguide array, a second layer waveguide array, a modulator array, and a combined beam waveguide array.
The first layer waveguide array may receive a first set of optical signals including a plurality of optical signals indicative of a first set of data, wherein the first layer waveguide array includes a plurality of first waveguides therein; the second layer waveguide array and the first layer waveguide array are positioned on different planes, the second layer waveguide array comprises a plurality of second waveguides, and can receive a first group of optical signals from the first layer waveguide array based on a waveguide coupling principle, wherein the second layer waveguide array and the first layer waveguide array are positioned on different planes; then, the modulator array modulates the first group of optical signals of the plurality of second waveguides based on the second group of data, and outputs a plurality of second optical signals; the beam combining waveguide array may converge the plurality of second optical signals.
In one possible design, the thickness of the plurality of first waveguides in the first region is less than the thickness in other regions than the first region, and the plurality of first waveguides overlap the plurality of second waveguides in the first region.
In one possible design, the combined beam waveguide array includes multiple layers of beam waveguides, each layer of beam waveguides configured to converge a portion of the plurality of second optical signals.
In one possible design, each layer of beam waveguides includes a plurality of superlattice waveguides.
In one possible design, after the converging waveguide array converges the plurality of second optical signals, the detector array may detect the plurality of second optical signals converged by the converging waveguide array.
In one possible design, the first set of optical signals may be generated by the light source module.
In a third aspect, the application provides a computing system, which may comprise a processor and an optical computing device as in the first aspect or any one of the possible implementations of the first aspect.
The processor may send data to be multiplied by the optical computing device. Specifically, the data to be subjected to multiply-add operation includes a first set of data and a second set of data. The optical computing device receives the data to be multiplied and added, generates a first group of optical signals for indicating the first group of data, and modulates the first group of optical signals based on the second group of data.
Drawings
FIG. 1 is a schematic diagram of an optical computing device according to the present application;
Fig. 2A to fig. 2B are schematic structural diagrams of a light source module according to the present application;
FIG. 3A is a schematic diagram of a dual-layer spectroscopic waveguide array according to the present application;
FIG. 3B is a schematic diagram of a first layer waveguide array according to the present application;
FIG. 3C is a schematic diagram of a second-layer waveguide array according to the present application;
FIG. 3D is a top view of a first waveguide and a second waveguide provided by the present application;
FIG. 3E is a side view of a first waveguide and a second waveguide provided by the present application;
FIG. 4 is a schematic diagram of a modulator array according to the present application;
FIG. 5 is a schematic diagram of a beam combining waveguide array according to the present application;
fig. 6A to 6B are schematic scanning diagrams of an optical computing device according to the present application;
FIG. 7 is a schematic diagram of a calculation method according to the present application;
fig. 8 is a schematic structural diagram of a computing system according to the present application.
Detailed Description
Before explaining the optical calculation device and the calculation method provided in the embodiment of the present application, a multiply-add operation will be described.
The multiply-add operation may be integrated as a matrix multiplication operation, the matrix multiplication operation being as follows:
B＝A*C
Namely:
Wherein B, A, C are matrices, C may be referred to as an input matrix, each element in C is an input vector, a may be referred to as an action matrix, B may be referred to as an output matrix, and each element in B may be an output vector. According to the multiplication operation of the matrix, each output vector is the calculation result of the multiplication and addition operation of one row of elements in the action matrix and each input vector.
In the optical multiply-add operation, one input vector may correspond to one optical signal (in the embodiment of the present application, the optical signal may be one optical signal in the first group of optical signals), the multiplication operation of the input vector and one element in the active matrix may be implemented by using modulation of the optical signal (in the embodiment of the present application, it may be understood that modulation of one optical signal in the first group of optical signals is performed by one element in the active matrix may be understood as modulation degree of modulation intensity or intensity of the optical signal), the multiplication operation result of the input vector and one element in the active matrix may be represented by one modulated optical signal (in the embodiment of the present application, the multiplication operation result of the input vector and one element in the active matrix may be represented by a plurality of modulated optical signals (in the embodiment of the present application, it may be represented by a plurality of second optical signals outputted by the modulator array), the multiplication operation result of one element in the active matrix and each input vector may be calculated by multiplying and adding modulation result of a plurality of optical signals, that is obtained by multiplying and adding a plurality of optical signals may be calculated by multiplying and adding optical signals.
As shown in fig. 1, an optical computing device 10 according to an embodiment of the present application includes a dual-layer splitting waveguide array 200, a modulator array 300, a beam combining waveguide array 400, and optionally, a light source module 100 and a detector array 500.
The light source module 100 is capable of generating an optical signal, and for convenience of description, the optical signal output from the light source module 100 will be referred to as a first optical signal. The light source module 100 may generate a plurality of first light signals, and the number of the first light signals is not limited herein. The number of first optical signals is related to the scale of the multiply-add operation to be performed, and the number of the plurality of first optical signals may be equal to the number of input vectors in the multiply-add operation, and the plurality of first optical signals constitute the first group of optical signals.
The dual-layer optical waveguide array 200 is capable of receiving a plurality of first optical signals output from the optical source module 100 and reproducing each of the first optical signals as a plurality of first optical signals. The dual-layer optical splitting waveguide array 200 may replicate the received first set of optical signals into a plurality of first sets of optical signals, and the first optical signals of the received first set of optical signals may be different in intensity from the first optical signals of the replicated plurality of first sets of optical signals. For example, the intensity of copying an optical signal into a plurality of optical signals C i,ci, where i is a positive integer no greater than n, may be less than C i.
Specifically in the dual-layer spectroscopic waveguide array 200, the dual-layer spectroscopic waveguide array 200 includes a first layer waveguide array 210 and a second layer waveguide array 220. The first layer waveguide array 210 is connected to the light source module 100 and is capable of receiving a plurality of first optical signals output from the light source module 100. The second layer waveguide array 220 is positioned in a different plane than the first layer waveguide array 210, the second layer waveguide array 220 being capable of receiving a first set of optical signals from the first waveguide array. For each first optical signal in the first set of optical signals, the second layer waveguide array 220 may split the first optical signal into a plurality of first optical signals, that is, the second layer waveguide array 220 may receive the first set of optical signals and output a plurality of first sets of optical signals.
The modulator array 300 is positioned behind the second layer waveguide array 220 and is capable of modulating the first set of optical signals output by the second layer waveguide array 220 to vary the intensity of the first set of optical signals. Taking a first set of optical signals output by the second layer waveguide array 220 as an example, the modulator array may modulate (implement multiplication) the first set of optical signals based on a second set of data, and output a second set of optical signals, where the second set of optical signals includes a plurality of second optical signals. The degree of modulation of the first set of optical signals by the modulator array 300 is determined based on a second set of data indicative of the data to be calculated, which may be a row of elements in the action matrix in the multiply-add operation, since the optical calculation means are used to implement the multiply-add operation.
The beam combining waveguide array 400 can collect a second group of optical signals output by the modulator array, and collect the second group of optical signals into a plurality of second optical signals that are closer to each other, so that the second group of optical signals can be detected or scanned by using the second group of optical signals.
The detector array 500 is configured to detect the intensity of the optical signals after the second group of optical signals output by the beam combining waveguide array are combined, in the detector array 500, when the second group of optical signals reach the detector array 500, the second group of optical signals are converted into an electrical signal, that is, a plurality of second optical signals are converted into an electrical signal, so as to realize the combination (addition operation) of the plurality of second optical signals, and the photoelectric conversion, and the detector array 500 can detect the intensity of the electrical signal to determine the result of the multiplication and addition operation.
In the embodiment of the application, the optical computing device does not introduce a beam splitter, but utilizes the coupling principle between waveguides to realize the beam splitting of the optical signals and divide the first optical signals into a plurality of first optical signals. By adopting the light splitting mode, noise carried in the first light signal after light splitting can be reduced to a large extent, and the accuracy of multiply-add operation can be ensured.
The following describes each component in the optical computing device according to the embodiment of the present application with reference to the accompanying drawings.
(1) A light source module 100.
The light source module 100 in the embodiment of the present application can output a plurality of first optical signals so as to input the plurality of first optical signals to the first layer waveguide array 210 of the dual-layer spectroscopic waveguide array 200.
The light source module 100 generates a plurality of first light signals in a plurality of ways, and the light source module 100 with different structures also generates a plurality of first light signals in a different way, and the following description describes two light source modules 100 according to the embodiment of the present application:
first, as shown in fig. 2A, the light source module 100 includes a laser array 110 formed by a plurality of lasers 111 and a first modulator array 120. The first modulator array 120 includes a plurality of modulators 121, and embodiments of the present application are not limited to the type of the modulators 121, and for example, the modulators 121 may be semiconductor laser amplifiers, lithium niobate modulators, metal heating modulators, and phase change material modulators. Any device capable of modulating an optical signal may be used as modulator 121 to form first modulator array 120.
In the light source module 100 shown in fig. 2A, each laser 111 can generate an original light signal, and one modulator 121 in the first modulator array 120 can modulate the original light signal generated by one laser 111 to output a first light signal. The plurality of modulators 121 in the first modulator array 120 can modulate the plurality of original optical signals generated by the plurality of lasers 111 to output a plurality of first optical signals.
The modulation process of the original optical signal by the first modulator array 120 is a process of loading information in the original optical signal, so that the generated first optical signal can indicate an input vector participating in the multiply-add operation.
Second, as shown in fig. 2B, another structure of a light source module 100 according to an embodiment of the application is shown, where the light source module 100 includes a laser 130, a beam splitter 140, and a second modulator array 150. The second modulator array 150 includes a plurality of modulators 151, and the embodiment of the present application is not limited to the type of the modulators 151, and the type of the modulators 151 may be referred to in the related description of the modulators 121, which is not repeated here.
The laser 130 can generate an original optical signal, the optical splitter 140 uniformly splits the original optical signal, and outputs a plurality of optical signals with the same light intensity, and one modulator 151 in the second modulator array 150 can modulate one optical signal output by the optical splitter 140 and output a first optical signal. The plurality of modulators 151 in the second modulator array 150 modulate the plurality of identical optical signals output from the optical splitter 140, and output a plurality of first optical signals.
The modulation process of the plurality of identical optical signals by the second modulator array 150 is a process of loading information in the plurality of identical optical signals such that the generated first optical signal is indicative of an input vector that participates in the multiply-add operation.
It should be noted that the above two structures of the light source module 100 are only examples, and the embodiments of the present application are not limited to the components included in the light source module 100, and any module capable of generating a plurality of first optical signals is suitable for the embodiments of the present application.
(2) Double-layer spectroscopic waveguide array 200
The dual-layer optical splitter array 200 is used for splitting a plurality of first optical signals, and unlike an optical splitter, the dual-layer optical splitter array includes two waveguide arrays, which are a first waveguide array 210 and a second waveguide array 220, respectively, for convenience of description. When the dual-layer optical splitting waveguide array 200 transmits the plurality of first optical signals, there is a coupling effect between waveguides of the waveguide arrays of different layers, and based on a coupling principle between waveguides (may also be simply referred to as a waveguide coupling principle), the plurality of first optical signals are split, and each first optical signal may be split into a plurality of second optical signals.
The coupling principle between waveguides means that when the effective refractive indexes of the fundamental modes in two waveguides close to each other are equal or approximately matched, the optical energy in the two waveguides generates a mutual coupling effect, and the optical energy in one waveguide can be transferred to the other waveguide, that is, the optical signal in one waveguide can be transferred to the other waveguide, so as to achieve the effect of light splitting. By controlling the interaction interval between the two waveguides (i.e. the part of the two waveguides that overlap) the relative phase of the two waveguides can be varied (such that the relative phase varies between 0-2 pi), a different degree of coupling between the two waveguides can be achieved, i.e. the light energy transferred from one waveguide to the other (also understood as the degree of intensity transferred to the other waveguide).
The structure of the dual-layer spectroscopic waveguide array 200 is described below:
as shown in fig. 3A, a schematic structural diagram of a dual-layer optical waveguide array 200 according to an embodiment of the present application is provided, where the dual-layer optical waveguide array 200 includes a first layer waveguide array 210 and a second layer waveguide array 220.
In the embodiment of the present application, the first layer waveguide array 210 is used for transmitting a plurality of first optical signals output by the light source module 100, and the second layer waveguide array 220 is close to the first layer waveguide array 210. For any first optical signal, the second layer waveguide array 220 can change the transmission direction of the first optical signal transmitted in the first layer waveguide array 210 by using the coupling principle between waveguides, so that a part of the first optical signal is coupled into the second layer waveguide array 220, and a part of the second layer waveguide array 220 coupled out of the first optical signal is only different from the first optical signal in light intensity, which can also be understood as that the second layer waveguide array 220 couples out the first optical signal from the first optical signal. The second-layer waveguide array 220 may couple one first optical signal a plurality of times to obtain a plurality of first optical signals.
Referring to fig. 3A, the first layer waveguide array 210 and the second layer waveguide array 220 are not in the same plane. For example, the first layer waveguide array 210 and the second layer waveguide array 220 may be located in two parallel planes, respectively, and there is an overlapping region between the first layer waveguide array 210 and the second layer waveguide array 220, that is, there is an overlapping portion between projections of the first layer waveguide array 210 and the second layer waveguide array 220 parallel to the two planes, and the overlapping portion corresponds to an overlapping region between the first layer waveguide array 210 and the second layer waveguide array 220. The overlapping region may also be understood as the region of overlap between the first layer waveguide array 210 and the second layer waveguide array.
The overlapping region of the first layer waveguide array 210 and the second layer waveguide array 220 enables a portion of the first optical signal transmitted in the first layer waveguide array 210 to pass from the first layer waveguide array 210 to the second layer waveguide array 220 (i.e., the principle of coupling between waveguides) as it passes through the overlapping region of the first layer waveguide array 210 and the second layer waveguide array 220.
In order to ensure that the first optical signal transmitted in the first layer waveguide array 210 can reduce the constraint of the first layer waveguide array 210 on the first optical signal when passing through the overlapping area, so that a part of the first optical signal can be smoothly transferred from the first layer waveguide array 210 to the second layer waveguide array 220, the thickness of the first layer waveguide array 210 in the overlapping area can be thinned, that is, the thickness of the first layer waveguide array 210 in different areas is different, and the thickness D in the overlapping area is smaller than the thickness D in other areas.
The first optical signal split from the first optical signal propagating in the first layer waveguide array 210 substantially reduces the intensity of the first optical signal in the first layer waveguide array 210, and the following relationship exists between the intensity of the optical signal C i in the first layer waveguide array 210 and the intensity of the optical signal C i in the second layer waveguide array 220:
Ci＝Sci
wherein S is the light splitting intensity, S is related to the coupling coefficient of the overlapping area, and the coupling coefficient can be configured by the structural design and the angle of the upper layer waveguide and the lower layer waveguide.
One first optical signal transmitted in the first layer waveguide array 210 may be divided into m first optical signals. For the plurality of first optical signals transmitted in the first layer waveguide array 210, each first optical signal may be divided into m first optical signals, and m second optical signals generated after each first optical signal is split may constitute m first groups of optical signals, each group of optical signals including the plurality of first optical signals, and each first group of optical signals is { c 1、c2、c3、…、cn }.
The following describes the structure of each layer of waveguide array, and as shown in fig. 3B, a schematic structural diagram of a first layer of waveguide array 210 according to an embodiment of the present application is provided, where the first layer of waveguide array 210 includes a plurality of first waveguides 211, and each first waveguide 211 is used for transmitting a first optical signal.
The thickness distribution of the first waveguide 211 is uneven, and the thickness D of the first waveguide 211 in the overlapping region of the first layer waveguide array 210 and the second layer waveguide array 220 is small, and the thickness D in the other region is large, that is, the thickness D of the first waveguide 211 in the overlapping region of the first layer waveguide array 210 and the second layer waveguide array 220 is smaller than the thickness D in the other region.
As shown in fig. 3C, in the schematic structural diagram of the second-layer waveguide array 220 according to the embodiment of the present application, the second-layer waveguide array 220 includes a plurality of groups of second waveguides 221, each group of second waveguides 221 includes a plurality of second waveguides 221, an overlapping area exists between one group of second waveguides 221 and each first waveguide 211 in the first-layer waveguide array 210, and the group of second waveguides 221 can split the first optical signals transmitted in the plurality of first waveguides 211 and output the first group of optical signals.
The positional relationship between the first layer waveguide array 210 and the second layer waveguide will be described below by taking one first waveguide 211 and one second waveguide 221 as an example.
As shown in fig. 3D, a top view of the first waveguide 211 and the second waveguide 221, that is, a schematic view of the first waveguide 211 and the second waveguide 221 in a direction perpendicular to the plane of the first layer waveguide array 210 and the second layer waveguide array 220 is shown.
As can be seen from fig. 3D, there is an included angle between the first waveguide 211 and the second waveguide 221, and there is an overlap region. The first waveguide 211 and the second waveguide 221 are abstracted as spatial straight lines, and the straight line corresponding to the first waveguide 211 and the straight line corresponding to the second waveguide 221 are different-plane straight lines.
As shown in fig. 3E, a side view of the first waveguide 211 and the second waveguide 221 is a side view of the first waveguide 211 and the second waveguide 221 in the thickness direction. As can be seen from fig. 3E, the first waveguide 211 and the second waveguide 221 are located in two parallel planes, and an insulating medium may be filled between the first waveguide 211 and the second waveguide 221, where the insulating medium may be silicon dioxide (SiO 2), and the embodiment of the present application does not limit the growth manner of the silicon dioxide, where the thickness of the first waveguide 211 is unevenly distributed, and the thickness of the overlapping area is smaller, and the thickness of the other area is larger, so that the structure is beneficial for coupling part of the first optical signal transmitted in the first waveguide 211 into the second waveguide 221 to form the second optical signal.
In the embodiment of the application, the coupling principle between the waveguides is utilized to realize the light splitting of the first optical signal, so that part of the first optical signal can be coupled into the second waveguide 221 in a lossless manner, the loss of the optical signal caused by light splitting can be greatly reduced, and the accuracy of multiply-add operation is ensured.
(3) A modulator array 300.
In the embodiment of the present application, the modulator array 300 can modulate the intensities of the plurality of first optical signals to generate a plurality of second optical signals, so as to implement multiplication. Illustratively, the modulator array 300 is capable of modulating a plurality of first sets of optical signals output in the second layer waveguide array 220, outputting a plurality of second sets of optical signals, each second set of optical signals including a plurality of second optical signals.
As shown in fig. 4, a plurality of modulators 310 may be included in the modulator array 300, each modulator 310 being capable of modulating one first optical signal, and the number of modulators 310 may be identical to the number of first optical signals.
For example, the modulator 310 may be disposed on each second waveguide 221 in the second layer waveguide array 220 to ensure that the second optical signal transmitted on the second waveguide 221 can be modulated.
Embodiments of the present application are not limited to the type of modulator 310, and the modulator 310 may be some or all of the following: lithium niobate thin film modulators, semiconductor optical amplifiers, thermal modulators, mach-zehnder interferometers (mach-zehnder interferometer, MZI). The modulator array 300 may comprise the same type of modulator 310 or may comprise different types of modulators 310.
The plurality of modulators 310 included in the modulator array 300 may be divided into a plurality of groups of modulators, each group of modulators may modulate a first group of optical signals { c1, c2, c3, …, cn } transmitted in the second layer waveguide array 220, the group of modulators modulating the first group of optical signals by a factor of [ t i1,ti2,ti3,…,tin ]. Assuming that the optical coupling efficiency between the set of modulators and the second layer waveguide array 220 is k, then the set of second optical signals passes through the set of modulators and then the set of third optical signals output through the set of modulators is k [ c 1ti1,c2ti2,c3ti3,…,cntin ]. The product of the modulation factor t i1,ti2,ti3,…,tin and the optical coupling efficiency k is determined from one row of elements in the action matrix. That is, the product of a matrix of modulation coefficients of the individual modulations of the modulator array and the optical coupling efficiency k is determined based on the action matrix.
(4) Beam combining waveguide array 400
The beam combining waveguide array 400 can converge a plurality of second optical signals, for example, one second group of optical signals output by each group of modulators, and the beam combining waveguide array 400 can converge the second optical signals in the second group of optical signals, so that the plurality of second optical signals are closer.
In practical applications, the total number of optical signals of the plurality of second groups of optical signals output by the modulator array is larger, so that the plurality of second groups of optical signals can be converged, and the beam combining waveguide array 400 can also adopt a layered structure.
Fig. 5 is a schematic structural diagram of a beam combining waveguide array 400 according to an embodiment of the present application, where the beam combining waveguide array 400 includes a plurality of beam combining waveguides. Each layer of the beam waveguides is capable of converging optical signals in a set of one or more second optical signals output by the modulator array.
Each layer of beam waveguide can be a superlattice waveguide, the superlattice waveguide can be realized through large-scale photoetching, the preparation method is the same as that of a common waveguide, the spacing distance between the superlattice waveguides is of a sub-wavelength scale, the structure is compact, and the number density of the integrated superlattice waveguides can be extremely high.
(5) Detector array 500
The detector array 500 is configured to detect the optical signals of the plurality of second sets of optical signals output by the beam-combining waveguide array 400, where the detector array 500 may include a plurality of detectors, and the embodiments of the present application are not limited to the types of detectors, and the detectors may be germanium detectors (Ge), silicon detectors (Si), and indium gallium arsenide detectors (InGaAs). The embodiment of the application is not limited to the deployment mode of the plurality of detectors, and only needs to ensure that the detection surfaces of the plurality of detectors can be coupled with the output end surfaces in the beam combining waveguide array 400 and can detect all optical signals output by the beam combining waveguide array 400.
It should be noted that, the embodiment of the present application is not limited to the specific form of the optical computing device, and for example, each component in the optical computing device may be integrated on a chip to form an on-chip optical computing device.
In the above description, the optical computing device is provided with the detector array 500 capable of detecting the total intensity of the optical information of the plurality of second groups of optical signals, and the optical computing device can be applied to the light beam scanning scene by changing the action matrix to realize different multiplication and addition operations to generate the second optical signals with different transmission directions.
In the light beam scanning scenario, the plurality of second groups of optical signals that need to be finally generated can cover a larger angle, and the first optical signals generated in the light source module 100 may be a plurality of optical signals with the same wavelength (i.e. optical signals with a single wavelength) or optical signals with different wavelengths (i.e. optical signals with multiple wavelengths).
The first optical signals output by the light source module 100 output a plurality of first groups of optical signals through the double-layer light splitting waveguide array 200, intensity modulation is achieved through the modulator array 300, the beam combining waveguide array 400 converges, and the second groups of optical signals output can interfere in far field space after exiting. By adjusting the intensity of the modulator array 300 for intensity modulating the first plurality of optical signals, the second plurality of optical signals can interfere in the far-field space to form a phased array, and scanning can be realized in space by using the main lobe (the maximum beam is the main lobe) of the phased array.
As shown in fig. 6A, when the light source module 100 in the computing device outputs a single wavelength light signal, a main lobe may be formed in the phased array to realize scanning in space.
As shown in fig. 6B, when the light source module 100 in the computing device outputs the optical signals of the plurality of wavelengths, the optical signals after the beam combination also have the optical signals of the plurality of wavelengths. For any wavelength optical signal, the optical signals after the optical signals with the wavelength are combined are interfered in a remote space to form a corresponding phased array, and a plurality of phased arrays can be formed due to the existence of signals with various wavelengths, and scanning can be realized in space by utilizing main lobes in each phased array.
In the light beam scanning scene, the optical computing device comprises the double-layer beam splitting waveguide and the beam combining waveguide array 400 formed by the multi-layer beam combining waveguides, so that a sufficient number of second optical signals can be generated, and a phased array can be formed to realize space scanning. In addition, the modulator array is arranged in the optical computing device, so that the modulation of the optical signals can be independently completed, and the on-chip control of the phase of the optical signals output by each unit of the integrated large-scale phased array antenna is facilitated.
In order to make the description of the solution clearer, the working procedure of the optical computing device provided in the embodiment of the present application will be generally described with reference to the previous embodiment, taking the optical computing device shown in fig. 1 and the computing method shown in fig. 7 as examples. In an embodiment of the present application, the light source module 100 generates a first set of light signals (1), wherein the first set of light signals comprises a plurality of light signals for indicating a first set of data; the light source module 100 transmits the first set of optical signals (2), and the first layer waveguide array 210 may receive the first set of optical signals (3) generated by the light source module 100, where the first layer waveguide array 210 includes a plurality of first waveguides; the second layer waveguide array 220 is located in a different plane than the first layer waveguide array 210; the first layer waveguide array 210 transmits the first set of optical signals (4), the second layer waveguide array 220 comprises a plurality of second waveguides and can receive the first set of optical signals (5) from the first layer waveguide array 210 based on waveguide coupling principle, wherein the second layer waveguide array 220 and the first layer waveguide array 210 are located on different planes, and the second layer waveguide array transmits the first set of optical signals (6); thereafter, the modulator array 300 modulates the first set of optical signals transmitted by the second-layer waveguide array based on the second set of data, outputting a plurality of second optical signals (7); the modulator array 300 transmits the plurality of second optical signals to the beam combining waveguide array 400 (8). The combined waveguide array 400 may converge (9) the plurality of second optical signals.
As a possible embodiment, the thickness of the plurality of first waveguides in the first region is smaller than that in the other region except the first region, and the plurality of first waveguides overlap with the plurality of second waveguides in the first region.
As one possible implementation, the combined beam waveguide array includes a plurality of layers of beam waveguides, each layer of beam waveguides being capable of converging a portion of the plurality of second optical signals.
As one possible implementation, each layer of beam waveguides may include a plurality of superlattice waveguides.
As a possible implementation, after the beam combining waveguide array 400 gathers the plurality of second optical signals, the detector array 500 may detect the plurality of second optical signals gathered by the beam combining waveguide array 400.
As shown in fig. 8, a computing system 30 is provided according to an embodiment of the present application, where the computing system 30 includes an optical computing device 10 and a processor 20. The structure of the light computing device 10 can be seen in the light computing device shown in fig. 1.
The processor 20 is connected to the optical computing device 10, and the processor 20 may send data to be subjected to multiply-add operation to the optical computing device 10 to instruct the optical computing device 10 to perform computation on the received data. The data to be subjected to multiply-add operation includes a first set of data and a second set of data.
The optical computing device may generate a first set of optical signals for indicating the first set of data after receiving the data to be subjected to the multiply-add operation, and may modulate the first set of optical signals based on the second set of data.
It should be noted that the examples provided by the present application are illustrative only. It will be apparent to those skilled in the art that, for convenience and brevity of description, in the above embodiments, descriptions of various embodiments are focused on, and for parts of one embodiment that are not described in detail, reference may be made to related descriptions of other embodiments. The features disclosed in the embodiments of the application, in the claims and in the drawings may be present independently or in combination. Features described in the form of hardware in embodiments of the application may be implemented in software and vice versa. And are not limited herein.

Claims (15)

1. An optical computing device, comprising:
A first layer waveguide array comprising a plurality of first waveguides for receiving a first set of optical signals, wherein the first set of optical signals comprises a plurality of first optical signals for indicating a first set of data;
A second layer waveguide array, located in a different plane than the first layer waveguide array, comprising a plurality of second waveguides for receiving the first set of optical signals from the first layer waveguide array based on waveguide coupling principles;
A modulator array for modulating the first set of optical signals transmitted by the second-layer waveguide array based on a second set of data, and outputting a plurality of second optical signals;
And the beam combining waveguide array is used for converging the plurality of second optical signals.
2. The apparatus of claim 1, wherein a thickness of the plurality of first waveguides in a first region is less than a thickness in other regions than the first region, the plurality of first waveguides overlapping the plurality of second waveguides in the first region.
3. The apparatus of claim 1 or 2, wherein the array of beam combining waveguides comprises a plurality of layers of beam combining waveguides, each layer of beam combining waveguides configured to converge a portion of the plurality of second optical signals.
4. The apparatus of claim 3, wherein each layer of the beam waveguides comprises a plurality of superlattice waveguides.
5. The apparatus of any one of claims 1,2, 4, wherein the optical computing device further comprises a detector array for detecting the plurality of second optical signals converged by the combined beam waveguide array.
6. The apparatus of any of claims 1,2, 4, wherein the modulator array comprises a plurality of modulators, the plurality of modulators comprising at least one of:
Lithium niobate thin film modulators, semiconductor optical amplifiers, thermal modulators, mach-zehnder interferometers MZI.
7. The apparatus of any one of claims 1, 2, 4, wherein the apparatus further comprises a light source module;
The light source module is used for generating the first group of light signals.
8. The apparatus of claim 7, wherein the light source module comprises a plurality of lasers, each laser for generating one of the first optical signals.
9. The apparatus of claim 7, wherein the light source module comprises a laser for generating an original light signal and a beam splitter for splitting the original light signal to generate a plurality of the first light signals.
10. A method of computing, the method comprising:
A first layer waveguide array receiving a first set of optical signals, the first set of optical signals including a plurality of optical signals for indicating a first set of data, the first layer waveguide array including a plurality of first waveguides;
A second layer waveguide array receives the first set of optical signals from the first layer waveguide array based on a waveguide coupling principle, wherein the second layer waveguide array comprises a plurality of second waveguides, and the second layer waveguide array and the first layer waveguide array are positioned on different planes;
The modulator array modulates the first group of optical signals transmitted by the second-layer waveguide array based on a second group of data and outputs a plurality of second optical signals;
the beam combining waveguide array converges the plurality of second optical signals.
11. The method of claim 10, wherein the plurality of first waveguides have a thickness in a first region that is less than a thickness in other regions than the first region, the plurality of first waveguides overlapping the plurality of second waveguides in the first region.
12. The method of claim 10 or 11, wherein the combined beam waveguide array comprises a plurality of layers of beam waveguides, each layer of beam waveguides configured to converge a portion of the plurality of second optical signals.
13. The method of claim 12, wherein each layer of the beam waveguides comprises a plurality of superlattice waveguides.
14. The method of any of claims 10, 11, 13, wherein after the converging the plurality of second optical signals, the beam-combining waveguide array further comprises:
The detector array detects a plurality of second optical signals converged by the beam combining waveguide array.
15. The method of any one of claims 10, 11, 13, wherein prior to the first layer waveguide array receiving the first set of optical signals, further comprising:
the light source module generates the first set of light signals.
CN202011270075.3A 2020-11-13 2020-11-13 Optical computing device and computing method Active CN112506265B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011270075.3A CN112506265B (en) 2020-11-13 2020-11-13 Optical computing device and computing method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011270075.3A CN112506265B (en) 2020-11-13 2020-11-13 Optical computing device and computing method

Publications (2)

Publication Number Publication Date
CN112506265A CN112506265A (en) 2021-03-16
CN112506265B true CN112506265B (en) 2024-08-09

Family

ID=74957564

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011270075.3A Active CN112506265B (en) 2020-11-13 2020-11-13 Optical computing device and computing method

Country Status (1)

Country Link
CN (1) CN112506265B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113161450B (en) * 2021-04-25 2022-11-25 中国科学院长春光学精密机械与物理研究所 Preparation method of integrated optical signal computing device
CN112988113B (en) * 2021-04-29 2021-09-14 中国科学院西安光学精密机械研究所 Photon matrix vector multiplier
CN113220268B (en) * 2021-06-11 2022-08-02 上海交通大学 Photoelectric mixed multiply-accumulate calculating structure
WO2023015438A1 (en) * 2021-08-10 2023-02-16 深圳市速腾聚创科技有限公司 Optical phased array chip and laser radar
CN116339444A (en) * 2021-12-23 2023-06-27 华为技术有限公司 Optical computing device, method and system
CN116502689A (en) * 2022-01-18 2023-07-28 华为技术有限公司 Optical calculation device and optical calculation method
CN117148496B (en) * 2023-08-28 2024-09-17 光本位科技(苏州)有限公司 Photon calculating chip for large-scale different-layer photon calculating array interconnection

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110568560A (en) * 2019-07-24 2019-12-13 浙江大学 Method for realizing large-scale optical switch topology array chip based on Benes structure and having balanced loss
TW202001694A (en) * 2018-06-05 2020-01-01 美商光子智能股份有限公司 Computing system

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110908428B (en) * 2019-10-25 2024-02-06 东南大学 Parallel light computing system for efficiently realizing large-scale matrix operation

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TW202001694A (en) * 2018-06-05 2020-01-01 美商光子智能股份有限公司 Computing system
CN110568560A (en) * 2019-07-24 2019-12-13 浙江大学 Method for realizing large-scale optical switch topology array chip based on Benes structure and having balanced loss

Also Published As

Publication number Publication date
CN112506265A (en) 2021-03-16

Similar Documents

Publication Publication Date Title
CN112506265B (en) Optical computing device and computing method
US11907832B2 (en) Optoelectronic computing systems
US11734556B2 (en) Optoelectronic computing systems
CN112001487B (en) Photonic neural network
US20190356394A1 (en) Photonic processing systems and methods
US9124373B2 (en) Integrated optical phased array
WO2023201970A1 (en) Computing chip, system, and data processing method
WO2021248810A1 (en) Calculation device, calculation method and calculation system
WO2022160784A1 (en) Optical computing device and system and convolution computing method
US20240063936A1 (en) Systems and methods for utilizing photonic degrees of freedom in a photonic processor
CN115222035B (en) Photon neural network convolution acceleration chip
CN117077750A (en) Parallel photon convolution operation chip and system
US20220147095A1 (en) Optical Computing Chip and System, and Data Processing Technology
US20230236621A1 (en) Photonic computing system
CN111949067B (en) Dammann convolution optical computer
Peserico et al. PhotoFourier: silicon photonics joint transfer correlator for convolution neural network
CN112232487A (en) Optical neural network chip and calculation method thereof
US20230142781A1 (en) Photonic Ising Compute Engine with An Optical Phased Array
CN112306145A (en) Optical computing chip, system and data processing technology
CN117709423B (en) Deep neural network photon acceleration chip and operation system thereof
US20230352465A1 (en) Led-based photonic communication and processing unit
Miller Self‐configuring algorithms, topologies and fundamental limits for photonic circuits and structures
Gottlieb et al. Photonic neural network and in-situ training in a synthetic frequency dimension
CN116502689A (en) Optical calculation device and optical calculation method
Forrest Optical Signal Processing

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant