Optical computing chip based on multiplexing and design method thereof

Abstract

The invention relates to the technical field of semiconductors, in particular to a multiplexing-based optical computing chip and a design method thereof.The multiplexing-based optical computing chip comprises a computing area formed by crossing a first waveguide and a second waveguide, and the computing area comprises a branch waveguide connected with the first waveguide and the second waveguide, a beam splitting part, an intensity modulation part and a first signal conversion part are sequentially arranged on the branch waveguide, the beam splitting part is used for splitting optical signals in a specific proportion, the intensity modulation part is used for modulating the signal intensity of the optical signals into a specific value, and the first signal conversion part is used for setting or maintaining the optical signals in a specific mode; the specific mode means that the optical signal has a specific polarization direction and / or a specific order; the second waveguide is provided with a calculation unit for calculating the sum of the optical powers of the plurality of optical signals. According to the multiplexing scheme provided by the invention, the data processing amount of the optical computing chip can be improved.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and in particular to a multiplexing-based optical computing chip and a design method thereof. Background Art

[0002] With the application of 5G technology, the rise of live video broadcasting, and the surge in big data services, people are increasingly demanding the improvement of the computing power of optical communication systems. Currently, the commonly used method to improve data processing capabilities is wavelength division multiplexing.

[0003] For example, CN118428430A discloses a silicon-based integrated all-optical deep neural network chip and its training method and intelligent device. It discloses a silicon-based integrated all-optical deep neural network chip and its training method and intelligent device. The chip includes: an input layer, an output layer and multiple fully connected layers; each fully connected layer and the output layer contains a beam splitter, a microcavity array and a nonlinear gain unit. The beam splitter divides the wavelength signals output by the previous layer into M parts, and different rows of the microcavity array are used to load weights on different parts of the divided signals; the nonlinear gain unit is used to perform nonlinear optical amplification calculations on each column to generate a regenerated signal; each fully connected layer uses a wavelength division multiplexing unit to multiplex the regenerated signal onto a waveguide and transmit it to the next layer; the output layer uses a photodetector array to perform photoelectric conversion on the regenerated signal to obtain the inference result.

[0004] However, the applicant has noticed that when wavelength division multiplexing technology is applied to computing optical chips, the signal loss is relatively large. Especially with the gradual expansion of current optical computing application scenarios, such as the widespread application of artificial intelligence in fields such as machine vision, natural language processing, and autonomous driving, higher requirements are placed on the data transmission computing capabilities in optical computing chips. However, the data processing capabilities that wavelength division multiplexing can support are very limited, and the loss is difficult to control, making it difficult to meet the computing needs of higher-specification optical computing chips.

[0005] In other words, as the requirements for chip integration and computing performance gradually increase, traditional multiplexing solutions are difficult to meet the real-time computing needs of data. Summary of the invention

[0006] The purpose of the present invention is to provide a multiplexing-based optical computing chip and a design method thereof, which can partially solve or alleviate the above-mentioned deficiencies in the prior art, improve the computing power of the optical computing chip, and at the same time achieve a good balance and optimization between its architectural complexity and computing accuracy.

[0007] In order to solve the above-mentioned technical problems, the present invention specifically adopts the following technical solutions: A first aspect of the present invention is to provide an optical computing chip based on multiplexing, comprising: a plurality of first waveguides; A plurality of second waveguides arranged opposite to the first waveguide, wherein the first waveguides and the second waveguides intersect to form a plurality of calculation areas; Wherein, the calculation area includes: a branch waveguide derived from the first waveguide, the first end and the second end of the branch waveguide are respectively connected to the first waveguide and the second waveguide, and the branch waveguide is provided with a beam splitter, an intensity modulation part, and a first signal conversion part in sequence along the direction from the first end to the second end thereof, wherein the beam splitter is used to separate a specific proportion of the first optical signal from the input optical signal from the first waveguide according to a set beam splitting ratio; the intensity modulation part is used to modulate the signal intensity of the input first optical signal to a specific value, and correspondingly generate and output a second optical signal; the first signal conversion part is used to set or maintain the second optical signal to a specific mode, and correspondingly generate and output a third optical signal, and the specific mode refers to that the third optical signal has a specific polarization direction and / or a specific order, and the third optical signal is transmitted to the corresponding second waveguide; wherein the third optical signals received at different positions of the same column of the second waveguide have different specific modes; A calculation unit is provided corresponding to the second waveguide, and the calculation unit is used to calculate the sum of the optical powers of the plurality of third optical signals.

[0008] In some embodiments, the beam splitting portion adopts a symmetrical light splitting structure, and correspondingly, the first waveguide includes: at least one row bus waveguide; and a group of first-level branch waveguides extending from the row bus waveguide, wherein the first-level branch waveguides are connected to the bus waveguide using the symmetrical splitting structure to separate a first-level optical signal of half the intensity from an input optical signal at a current position of the row bus waveguide; the first-level branch waveguide comprises: a first branch waveguide, and a second branch waveguide extending from the first branch waveguide, wherein the second branch waveguide is connected to the first branch waveguide using the symmetrical splitting structure to separate a second-level optical signal of half the intensity from an optical signal at a current position, wherein the secondary optical signal is input to the intensity modulation unit correspondingly, and the intensity modulation unit outputs the corresponding second optical signal. In some embodiments, the first waveguide further includes: a primary branch waveguide extending from the bus waveguide, wherein the primary branch waveguide is connected using a symmetrical splitting structure to separate a primary optical signal of half the intensity from the input optical signal at the current position.

[0009] In some embodiments, the specific order of the third optical signal input in the second waveguide decreases successively along a first direction, and correspondingly, the width of the second waveguide decreases successively along the first direction, wherein the first direction is a direction opposite to the transmission direction of the optical signal in the second waveguide.

[0010] In some embodiments, the calculation unit is a photodetector; and / or the first signal conversion unit is an asymmetric directional coupler.

[0011] In some embodiments, the second waveguide is a multimode waveguide.

[0012] In some embodiments, the branch waveguide further includes: at least two first sub-waveguides, a combining waveguide connecting at least two of the first sub-waveguides, and a second sub-waveguide for transmitting a signal of the combining waveguide to the second waveguide; Correspondingly, the first waveguide includes: a first row of waveguides arranged in sequence, and at least one middle row of waveguides arranged in sequence; A plurality of first-row calculation units are arranged on the first-row waveguide, and the first-row calculation units include: a beam splitter arranged on the first-row waveguide, and a first sub-waveguide connected through the beam splitter, an intensity modulator is arranged on the first sub-waveguide, and an output end of the first sub-waveguide is connected to the corresponding wave combining waveguide; The intermediate row waveguide is provided with a plurality of intermediate row calculation units, and the intermediate row calculation units include: a beam splitter provided on the intermediate row waveguide, and a first sub-waveguide connected through the beam splitter, the first sub-waveguide is provided with an intensity modulator, and the first sub-waveguide is connected to the combiner waveguide through a second signal conversion unit, wherein different row waveguides use light of different wavelengths as the input optical signal, and the second signal conversion unit is used to combine at least two optical signals of different wavelengths into one optical signal, and input the optical signal into the combiner waveguide; The second sub-waveguide is connected to the second waveguide through the first signal conversion section.

[0013] In some embodiments, the first waveguide is disposed in a first layer, the second waveguide is disposed in a second layer, and the first layer and the second layer are connected via the second sub-waveguide.

[0014] The present invention also provides a design method corresponding to the composite optical computing chip, which comprises the steps of: S101, providing an optical computing chip, wherein the optical computing chip is provided with n first waveguides, and the first waveguides include: m running waveguides, the m running waveguides include: 1 first running waveguide and a middle running waveguides; and the m running waveguides are arranged at a first spacing; S102, setting at least one set of different test conditions, the test conditions comprising: a first set value of m, and a second set value of the first interval; S103, testing the optical computing chip according to at least one set of the test conditions, and correspondingly collecting at least one set of first test results, wherein the first test results include: calculation results output by m*n second waveguides; wherein the calculation result is the sum of the powers of the m*n optical signals output by the second waveguides, and the m traveling waveguides respectively use light of m different wavelengths as input optical signals; S104, selecting at least one second test result satisfying a first rule from the first test results, wherein the first rule requires that a difference ratio of the calculated results is less than a number of a set composite threshold and greater than a set reference number, wherein the difference ratio is |calculated result-standard result| / standard result; S105, calculating at least one evaluation index L of the second test result according to the corresponding test condition, wherein L=αX+βY; wherein X is a first relative index of a first set value, Y is a second relative index of a second set value, α is a first weight, and β is a second weight; wherein the numerical value of the first set value is divided into at least two levels from small to large, and the levels correspond to a relative value for describing the first relative index; the numerical value of the second set value is divided into at least two levels from small to large, and the levels correspond to a relative value for describing the second relative index; S106, recommending corresponding test conditions to the user according to the evaluation index L as a preferred design solution.

[0015] In some embodiments, the order difference between the front and rear third optical signals coupled into the second waveguide is greater than a set order threshold, and the test condition further includes: a third set value of the order difference; the value of the third set value is divided into at least two levels from small to large, and the level corresponds to a relative value for describing the third relative index; Correspondingly, the evaluation index L is: L=αX+βY+γZ1, γ is the third weight, and Z1 is the third relative index of the order difference.

[0016] In some embodiments, the test condition further includes: a fifth setting value of n; correspondingly, the value of the fifth setting value is divided into at least two levels from small to large, and the level corresponds to a relative value describing the fifth relative indicator; Correspondingly, the evaluation index L is: L = αX + βY + εZ3; Among them, ε is the fifth weight, and Z3 is the fifth relative index.

[0017] Beneficial technical effects: The present invention provides an optical computing chip that synchronously transmits and calculates large data optical signals through mode differentiation. The transmission of differentiated modes of multiple types of signals in the second waveguide can simplify the overall architecture of the optical computing chip. For example, it can directly sum the power of multiple optical signals through a photodetector at the end of the second waveguide, and there is no need to set up a power calculation device at each first waveguide to perform independent signal calculations. In other words, this mode-differentiated operating mechanism can improve data transmission efficiency through unified calculation processing, that is, optical signals from different first waveguides can be quickly converted and coupled to the second waveguide only through mode changes, and continue to be transmitted at high speed in the form of light, and only one overall calculation is performed at the end of the second waveguide. Therefore, the overall computing efficiency of the optical computing chip is relatively high.

[0018] Furthermore, in order to reduce the computational error of large-scale optical computing chips, an array structure based on tree-like cascade equal-proportional splitting can be preferably adopted. That is, the present invention also provides a restrictive splitting mechanism, which forcibly requires that each splitting position under the tree-like cascade structure adopts the same proportion of light beams for splitting adjustment, thereby avoiding the introduction of more errors in the splitting process during the mode conversion process.

[0019] On the other hand, in order to reduce the signal crosstalk and difficult mode management problems faced by large-scale optical computing chips when performing differentiated light mode transmission, a differentiated transmission mechanism based on mode and wavelength coordination is proposed (that is, a composite optical computing chip is provided, which has the characteristics of high computing speed and low loss) to reduce the difficulty of mode difference setting while reducing the risk of signal crosstalk.

[0020] This composite multi-stage cascade architecture solution is conducive to a certain degree of balanced design between the difficulty of signal differentiation setting and the amount of signal loss. Specifically, by introducing a limited number of input optical signals of different wavelengths in the first waveguide (that is, a group of traveling waveguides), the difficulty of order difference can be reduced, that is, the number of necessary orders required for mode differentiation is reduced to m, that is, the original 1 / n order. Therefore, for large-scale optical computing chips (such as 124*124 optical computing chips), the technical difficulty of mode differentiation arrangement can be greatly reduced. In addition, by setting the coupling difference of wavelength and order (that is, by restrictively introducing only a limited number of optical signals of different wavelengths through the cooperation of multiple groups of traveling waveguides), the technical difficulty and cost of multi-wavelength light sources can also be reduced. Therefore, through the coordinated setting of the number of wavelengths and the number of orders, a certain degree of balance and optimization can be achieved between the cost of the light source and the signal crosstalk.

[0021] Furthermore, in order to make the composite optical computing chip proposed by the present invention easier to implement, the present invention also proposes a design method for the composite optical computing chip. The reliability and size characteristics of the optical computing chip are comprehensively evaluated by the two key factors of the number of wavelengths (m) set differently and the first spacing between the traveling waveguides (which can alleviate the risk of crosstalk to a certain extent), so that users can quickly select suitable design specifications for this new type of composite optical computing chip.

[0022] In other words, the present invention performs a dual evaluation of the number of wavelength differences that can be accommodated in the traveling waveguide and the spacing of the traveling waveguides, so as to alleviate the possible signal crosstalk or signal quality problems on the basis of improving the computing capacity of the optical computing chip. It should be noted that the applicant has noticed that optical input signals of different wavelengths are prone to introduce signal loss problems on the basis of improving computing performance. Therefore, by dual evaluation of the number of wavelengths and the spacing setting, the possible increase in computing errors of large-scale optical computing chips can be limited as much as possible, so that the composite computing chip solution is easier to implement.

[0023] Furthermore, the present invention also proposes a design scheme that can comprehensively evaluate the setting interval of the combined waveguide, the number of wavelength differences and the interval of the traveling waveguide. Similarly, as the scale of optical computing chips continues to expand, the multi-level cascade scheme of the composite optical computing chip will become more complicated. In particular, in the actual chip architecture, it is difficult to achieve such Figure 5 or Figure 6 The ideal straight-line arrangement may have the problem of waveguide bending, which will further increase the strength or loss difference of signals of different orders and the risk of crosstalk. In this regard, the present invention focuses on selecting the spacing between the combined waveguide and the second waveguide as the key factor to comprehensively evaluate the signal loss and signal crosstalk factors, so as to further achieve a balance between wavelength differentiation and order differentiation.

[0024] Preferably, the present invention also proposes a scheme for layered arrangement of the first and second waveguides. On the one hand, the first and second waveguides are respectively arranged in different layers, and the interlayer spacing is controlled to reduce the mode crosstalk caused by scattering; on the other hand, it is also conducive to realizing a flexible layout of the second waveguide and reducing the loss and crosstalk introduced by waveguide bending.

[0025] Furthermore, the present invention can also comprehensively consider the degree of order differentiation (such as the order distribution corresponding to different first waveguides, that is, the difference between the orders corresponding to adjacent first waveguides) to balance the wavelength differentiation and order differentiation setting resources of the composite optical computing chip, so as to reduce the difficulty of setting wavelength differentiation and order differentiation (reducing the performance requirements for the first and second signal conversion units) on the basis of ensuring the overall operational feasibility, thereby collaboratively managing signal loss and signal crosstalk problems so that it can be suitable for the optical computing needs of large-scale optical computing chips. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale. Obviously, the drawings described below are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without paying creative labor.

[0027] Figure 1 is a schematic diagram of an array structure of an optical computing chip in an exemplary embodiment of the present invention; Figure 2a Schematic diagram of the distribution of the electric field amplitude of light with different mode orders along the transverse direction (perpendicular to the transmission direction); Figure 2b Schematic diagram of the distribution of each order mode field in the TE and TM directions in the waveguide; Figure 3 is a schematic structural diagram of a second waveguide in an exemplary embodiment of the present invention; Figure 4 It is a schematic diagram of the array structure of the tree-like cascade optical computing chips in an exemplary embodiment of the present invention; Figure 5 is a schematic diagram of a partial structure of a composite optical computing chip in an exemplary embodiment of the present invention; Figure 6 is a schematic diagram of the structure of a composite optical computing chip in an exemplary embodiment of the present invention; Figure 7 The figure is a schematic flow chart of the design method steps in an exemplary embodiment of the present invention.

[0028] Summary of reference numerals: First waveguide 1, second waveguide 2, branch waveguide 3 (also called: third waveguide), intensity modulation unit 6, first signal conversion unit 7, beam splitting unit 8, row bus waveguide 11, primary branch waveguide 12, first branch waveguide 121, second branch waveguide 122, first sub-waveguide 31, combining waveguide 32, second sub-waveguide 33, second signal conversion unit 34, first row waveguide 41, intermediate row waveguide 42; waveguide core layer 01, waveguide cladding 02, wafer substrate 03. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0030] Herein, suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present invention, and have no specific meanings by themselves. Therefore, "module", "component" or "unit" can be used mixedly.

[0031] In this document, the terms "upper", "lower", "inner", "outer", "front", "back", "one end", "the other end" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0032] In this document, unless otherwise clearly specified and limited, the terms "installed", "provided with", "connected", etc. should be understood in a broad sense. For example, "connected" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0033] Herein "and / or" includes any and all combinations of one or more of the associated listed items.

[0034] Herein, "plurality" means two or more than two, ie, it includes two, three, four, five, etc.

[0035] As used in this specification, the term "about" typically means + / - 5% of the stated value, more typically + / - 4% of the stated value, more typically + / - 3% of the stated value, more typically + / - 2% of the stated value, even more typically + / - 1% of the stated value, and even more typically + / - 0.5% of the stated value.

[0036] In this specification, some embodiments may be disclosed in a format of being in a range. It should be understood that this description of "being in a range" is only for convenience and brevity, and should not be interpreted as a rigid limitation on the disclosed range. Therefore, the description of the range should be considered to have specifically disclosed all possible sub-ranges and independent numerical values ​​within this range. For example, the description of the range 1-6 should be considered to have specifically disclosed sub-ranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6, etc., as well as individual numbers within this range, such as 1, 2, 3, 4, 5 and 6. Regardless of the breadth of the range, the above rules apply.

[0037] Computation area: In this article, the computation area refers to the area formed by one or more computation units. Generally, an N*N optical computing chip (hereinafter referred to as a chip) means that it has N*N computation units. Figure 1 For example, a computing unit generally refers to a functional unit including a section of the first waveguide, a beam splitting portion, an intensity modulation portion 6, a section of the branch waveguide, and a section of the second waveguide. In particular, when the computing unit is in the second row or a subsequent row (i.e., the computing unit is not in the first row) along the direction in which the optical signal is transmitted in the second waveguide, it refers to a functional unit including a section of the first waveguide, a beam splitting portion, an intensity modulation portion 6, a section of the branch waveguide, a section of the second waveguide, and a signal conversion portion (such as a first signal conversion portion or a second signal conversion portion).

[0038] For example, in Figure 1 In the exemplary embodiment shown, one computing region may refer to one computing unit.

[0039] For example, in Figure 6 In the exemplary embodiment shown, when the first waveguide further includes m row waveguides, a calculation area formed by the intersection of the first waveguide and the second waveguide can also refer to a functional area formed by m calculation units. Specifically, the calculation area in this exemplary embodiment includes: 1 first row calculation unit, and a middle row calculation units.

[0040] See also Figure 2bAs shown, the waveguide is sequentially provided with a waveguide core layer 01, a waveguide cladding layer 02 and a wafer substrate 03. The waveguide realizes light transmission based on the principle of total reflection, and its lateral spatial restriction enables the light field amplitude distribution that satisfies the lateral resonance condition (corresponding to a specific wave vector or effective refractive index) to be stably transmitted in the waveguide, i.e., the waveguide mode.

[0041] Light waves are electromagnetic waves, and the polarization direction of the optical waveguide mode refers to the relationship between the direction of the electric field vector and the propagation plane. For example, the TE and TM polarization directions refer to the main components of the electric field along the tangential and normal directions of the wafer substrate, respectively (such as Figure 2b (indicated by the arrow in the middle).

[0042] According to the number of nodes (nodes are points where the field strength is 0) in the stable mode field distribution formed by the lateral resonance of the waveguide, it can be divided into fundamental mode TE0, TM0 (number of nodes is 0), first-order mode TE10, TM10, TE01, TM01 (number of nodes is 1), and higher-order modes such as TE20, TM20, etc. (see Figure 2b As shown). Correspondingly, light of different modes refers to light signals of different orders (or, numbers of orders). Light of each order mode has a specific electric field distribution, effective refractive index and group refractive index. In a uniform waveguide, the modes are mutually orthogonal and are transmitted independently without conversion. The total optical power in the waveguide is the sum of the powers of each order mode.

[0043] For example, see Figure 2a As shown in the figure, taking a one-dimensional planar waveguide as an example, the distribution of TE0~TE3 modes is shown in the figure. The number of nodes in the lateral distribution is e, and the mode is the e-th order mode TEe. The light wave is a transverse wave. In an integrated optical waveguide, when the electric field (main direction) is along the normal direction of the wafer, it is a TM wave; when it is along the tangential direction of the wafer, it is a TE wave.

[0044] In the present invention, light (such as optical signals) is transmitted in the same waveguide using different modes, which can avoid the use of low-speed multi-path parallel electrical summing devices, thereby simplifying the structure of the chip system and improving the data processing (or calculation) speed.

[0045] With the rise of big data services, such as machine vision, natural language processing and autonomous driving, higher requirements are placed on the data processing capabilities of optical computing chips. In response to this, the present invention proposes an optical computing chip with high data processing speed and low loss characteristics.

[0046] Embodiment 1 See also Figure 1 As shown, the present invention provides an optical computing chip based on multiplexing, comprising: A plurality of first waveguides 1; A plurality of second waveguides 2 arranged opposite to the first waveguide, wherein the first waveguide 1 and the second waveguide 2 intersect to form a plurality of calculation areas; Wherein, the calculation area includes: a branch waveguide 3 (also referred to as: a third waveguide) extending from the first waveguide 1, the first end and the second end of the branch waveguide 3 are connected to the first waveguide and the second waveguide respectively, and the branch waveguide is provided with a beam splitter 8, an intensity modulation part 6, and a first signal conversion part 7 in sequence along the direction from the first end to the second end thereof, wherein the beam splitter 8 is used to separate a specific proportion of optical signals from the first waveguide 1 according to a set beam splitting ratio (from the input optical signal); the intensity modulation part 6 is used to modulate the signal intensity of the input optical signal to a specific value, and correspondingly generate and output a second optical signal, and the first signal conversion part 7 is used to set or maintain the second optical signal to a specific mode, and correspondingly generate and output a third optical signal, wherein the specific mode refers to that the optical signal has a specific polarization direction and / or a specific order, and the third optical signal is transmitted to the corresponding second waveguide; wherein the third optical signals received at different positions of the same column of the second waveguide have different specific modes; A calculation unit is provided corresponding to the second waveguide, and the calculation unit is used to calculate the sum of the optical powers of the plurality of third optical signals.

[0047] It should be noted that, in some embodiments, the branch waveguide 3 is mainly used to lead out a specific proportion of optical signals from the first waveguide 2 , so it can also be regarded as a part of the first waveguide 2 .

[0048] In some embodiments, the calculation part may be a photodetector disposed at the end of the second waveguide. The beam splitting part may be provided with a beam splitter, and the intensity modulation part may be provided with an intensity modulator.

[0049] See also Figure 1 As shown, a calculation area in this embodiment specifically refers to a calculation unit, and the first waveguides in the i-1th row, the i-th row, and the i+1th row are arranged in sequence in the lateral direction, and the second waveguides in the j-1th column, the jth column, and the j+1th column are arranged in sequence in the lateral direction. Correspondingly, in the calculation unit composed of the waveguides in the i-th row and the j-th column, the optical power output by the beam splitter is , the transmittance is After the intensity modulator, the optical power is The modulator output light beam is transmitted in TE00 mode, and is converted into a specific mode of the column bus multimode waveguide through the signal conversion unit; the output light of each row is coupled to N different modes of the column bus multimode waveguide, and transmitted to the photodetector of the column together. The detection light power is the sum of the power of each mode. , realizing the summation of the optical power of the calculation units in this column.

[0050] In this embodiment, the input TE 00 mode light beam can be converted to the TE / TM N0 mode of the bus waveguide without affecting the transmission of other modes in the bus waveguide. When the effective refractive index of the TE 00 mode of the input waveguide (equivalent to the first waveguide) is the same as that of the TE / TM K0 mode in the bus waveguide (equivalent to the second waveguide), during the parallel transmission process, the optical power will be completely converted from one waveguide to another waveguide. Correspondingly, the signal conversion unit can be provided with an asymmetric directional coupler.

[0051] It is worth noting that the transmission mechanism of multi-mode optical signals in the bus waveguide adopted in the present invention can set only one photodetector at the end of the bus waveguide to calculate the sum of the powers of all optical signals at one time, without having to calculate the power separately at the end of each computing unit. This transmission mechanism simplifies the device design of the computing array on the one hand, and improves the computing speed on the other.

[0052] Specifically, the optical transmission mechanism of different modes provided by the present invention avoids the need to arrange multi-wavelength lasers on the chip, such as optical frequency combs or lasers of multiple different wavelengths. The reduction in the number of lasers also reduces the difficulty of chip system packaging and reduces costs. On the other hand, the reduction in the number of required signal wavelengths avoids the need to arrange large-scale feedback circuits for stable control of wavelengths, thus simplifying system complexity.

[0053] Preferably, in this embodiment, each row of first waveguides uses an optical signal with the same wavelength as the input optical signal.

[0054] Of course, in other embodiments, when the computing array scale is further expanded, optical signals of different wavelengths may also be used as input optical signals.

[0055] Preferably, along the transmission direction of the column bus waveguide, the order of the mode is preferentially from low order to high order. For example, in some embodiments, the 1-4 rows of input light are respectively converted into the TE00, TE10, TE20, TE30 modes of the column bus waveguide; or, in other embodiments, the 1-4 rows of input light are respectively converted into the TE00, TE10, TM00, TM10 modes of the column bus waveguide. In this embodiment, when the polarization direction is converted, a polarization rotator needs to be added before the mode multiplexer to convert the TE00 mode into the TM00 mode. Preferably, the second waveguide uses a waveguide that only supports TE00 and / or TM00 modes to limit the mode type and avoid coupling of different optical signals during side wall scattering and bending radiation, which causes large deviations in the optical signal during transmission.

[0056] Ideally, the TE 00 mode is completely converted to the higher-order mode TE / TM N0, but due to the structural preparation deviation of the mode multiplexer, part of the input light beam may remain in the input waveguide, or part of it may be converted (or crosstalked) to other modes TE / TM K0; for example, when it is desired to convert from TE00 to TE20, but because the effective refractive index of TE10 is similar to that of TE20, a small part of the light may be converted to TE10, causing crosstalk. In addition, in the bus waveguide, structures such as waveguide crossing and sidewall roughness may cause part of the power in the transmission mode to be converted to other modes, thus forming crosstalk.

[0057] In this regard, in this embodiment, a mode multiplexer with an adiabatic gradient asymmetric coupling structure is preferably used to reduce the impact of process deviations on coupling efficiency and crosstalk. Specifically, the widths of the two mutually coupled waveguides in the adiabatic gradient asymmetric coupling structure are adiabatically gradient along the transmission direction, so that the mode of the transmitted light field remains at the original order, and only the distribution width changes; because mode coupling mainly occurs in the waveguide width range where the effective refractive index matches. When there is a process deviation, the change in waveguide width mainly affects the coupling position, but not the coupling efficiency.

[0058] Further, see Figure 3 As shown, the width of the column waveguide gradually widens with the increase of the number of rows, and the number of supported modes increases from small to large. The signals are coupled sequentially from low to high to reduce mode crosstalk. Preferably, the sidewall roughness of the silicon waveguide can be reduced by post-etching annealing, sidewall oxidation and other process methods; preferably, a material with a lower refractive index can be selected as the column bus waveguide, such as doped silicon oxide, silicon nitride, silicon oxynitride, etc., to reduce the sidewall scattering intensity.

[0059] In this embodiment, the risk of crosstalk generated by multi-mode optical signals can be reduced by optimizing the sidewall scattering intensity of the column bus waveguide.

[0060] In some embodiments, see Figure 4 As shown, the first waveguide comprises: at least one row bus waveguide 11; and a group of primary branch waveguides 12 extending from the row bus waveguide 11, wherein the primary branch waveguides 12 are connected to the row bus waveguide 11 by a symmetrical splitting structure (equivalent to the beam splitting section) to separate a primary optical signal of half the intensity from an input optical signal at a current position of the row bus waveguide 11; the primary branch waveguide 12 further comprises: a first branch waveguide 121, and a second branch waveguide 122 extending from the first branch waveguide 121, wherein the second branch waveguide 122 is connected to the first branch waveguide 121 by a symmetrical splitting structure to separate a secondary optical signal of half the intensity from an optical signal at a current position (in other words, the first branch waveguide and the second branch waveguide are respectively input with optical signals of the same intensity), and the secondary optical signal is correspondingly input to the intensity modulation section, and the intensity modulation section outputs the corresponding second optical signal.

[0061] It is understandable that in this embodiment, the branch waveguide 3 (for example, the primary branch waveguide 12) can also be regarded as a portion extending from the first waveguide. In some embodiments, the branch waveguide can be a waveguide structure extending directly from the first waveguide (especially, the row bus waveguide 11). Alternatively, in other embodiments, the branch waveguide can also be a waveguide structure connected to the first waveguide through a beam splitter or other connecting devices.

[0062] In some embodiments, the first waveguide further includes: a primary branch waveguide (equivalent to a section of branch waveguide 3) extending from the bus waveguide 11, and the primary branch waveguide is connected using a symmetrical splitting structure to separate a primary optical signal of half the intensity from the optical signal at the current position.

[0063] This embodiment provides an array structure based on tree-like cascade equal-proportional splitting. In other words, for the transmission scenarios of optical signals of different specific modes, the present invention proposes a restrictive splitting mechanism to reduce the deviation that may be caused by uneven splitting on the mode modulation of the optical signal, thereby reducing the transmission error that may be caused by large-scale computing chips in signal transmission.

[0064] It is worth noting that due to the manufacturing process, different beam splitters may have certain deviations in the splitting performance, especially when the beam splitter splits different proportions of optical signals, the splitting deviation that may occur during the actual operation is also difficult to predict. In this regard, a restrictive splitting mechanism proposed in the present invention requires that each splitting position under the tree-like cascade structure adopts the same proportion of light beams for splitting adjustment, thereby avoiding the introduction of more errors in the splitting process in mode division multiplexing.

[0065] Preferably, the present invention adopts a beam splitter with a symmetrical structure, which has higher reliability when applied to equal-proportional light splitting.

[0066] In some embodiments, the second waveguide is a multimode waveguide.

[0067] Correspondingly, in order to meet the requirements of high data computing capability and low loss, the present invention also provides a composite multiplexing computing system (or composite optical computing chip) based on mode division and wavelength division cross-connection. Specifically, in some embodiments, the branch waveguide (also referred to as the third waveguide) extending from the first waveguide further includes: at least two first sub-waveguides 31, and a combining waveguide 32 connecting at least two of the first sub-waveguides 31, and a second sub-waveguide 33 for transmitting the signal of the combining waveguide to the second waveguide 2; Correspondingly, the first waveguide includes: a first row waveguide 41 arranged in sequence, at least one middle row waveguide 42 arranged in sequence, such as a middle row waveguide; and at least two first sub-waveguides 31 are respectively connected to the first row waveguide 41 and the middle row waveguide 42, and the output ends of the two first sub-waveguides 31 are connected through a combined waveguide 32, and the output end of the combined waveguide 32 is connected to the second waveguide 2 through a second sub-waveguide 33; Correspondingly, a plurality of first-row calculation units are formed on the first-row waveguide 41 along the length direction, and the first-row calculation unit includes: a beam splitter provided on the first-row waveguide 41, and a first sub-waveguide 31 connected through the beam splitter, an intensity modulator is provided on the first sub-waveguide 31, and an output end of the first sub-waveguide 31 is connected to the corresponding combiner waveguide 32; The intermediate row waveguide 42 is provided with a plurality of intermediate row calculation units along the length direction, and the intermediate row calculation units include: a beam splitter provided on the intermediate row waveguide 42, and a first sub-waveguide 31 connected through the beam splitter, the first sub-waveguide 31 is provided with an intensity modulator, and the first sub-waveguide 31 is connected to the combining waveguide 32 through a second signal conversion unit 34, wherein the second signal conversion unit is used to combine at least two optical signals of different wavelengths into one optical signal, and input the optical signal into the combining waveguide 32; That is to say, a group of optical signals with a+1 different wavelengths are coupled in the combining waveguide 32, and the group of optical signals are combined into a beam of light through the second signal conversion unit to be transmitted in the second waveguide; since different optical signals have different wavelengths, they can be distinguished and identified by the detector when they are synchronously transmitted in the same second waveguide, so as to complete the calculation of the output results of different calculation areas.

[0068] Below, see Figure 5-Figure 6 The multiplexing mode in the present invention is explained as follows: Figure 6A 4*4 computing unit array is shown. Two groups of first waveguides are respectively provided, and the first row of waveguides (equivalent to the first row of waveguides 41) in the first waveguides inputs the first incident light (i.e., optical signal) of the first wavelength λ1, and the second row of waveguides (equivalent to the middle row of waveguides 42) inputs the second incident light of the second wavelength λ2; the optical signals output by the two rows of waveguides are intensity modulated and then combined into the combining waveguide 32 through a wavelength division multiplexer. The first wavelength and the second wavelength are not equal. Furthermore, the optical signal in the combining waveguide 32 is transferred to the second waveguide 2 in the TE0 mode for transmission.

[0069] The first row of waveguides in the second first waveguide can also input the first incident light of the first wavelength λ1, and the second row of waveguides can input the second incident light of the second wavelength λ2. The optical signals output by the two rows of waveguides are intensity modulated and then combined in the combining waveguide 33 through the wavelength division multiplexer. Furthermore, the optical signal in the combining waveguide will be adjusted to the TE1 mode through the signal conversion unit to be transferred to the second waveguide 2 for transmission. At this time, four optical signals are input into the second waveguide together, and each optical signal has a different wavelength or order, so they can be transmitted simultaneously in the second waveguide without easily causing mutual interference and other effects.

[0070] It is worth noting that, through the composite series connection setting in the present invention, the tolerance of the optical computing chip to the synchronous transmission of a large number of optical signals can be improved without increasing the number of traveling waveguides (or first waveguides).

[0071] Preferably, the second signal conversion unit in the present invention can adopt a wavelength division multiplexer with a micro-ring structure, which has a high wavelength correlation; the first signal conversion unit adopts an asymmetric directional coupler structure, whose working performance is insensitive to wavelength, and can convert optical signals in a certain wavelength range λ1~λ2 from TE0 to the same order mode, thereby realizing the combination of mode and wavelength.

[0072] In some embodiments, the first waveguide is arranged in the first layer, the second waveguide is arranged in the second layer, and the first layer and the second layer are connected by the second sub-waveguide 33. In this embodiment, in order to alleviate the crosstalk problem in large-scale optical computing chips, a double-layer waveguide arrangement with a long distance between them is preferably used.

[0073] In some embodiments, the first layer also includes: a first sublayer and a second sublayer, and the first sublayer and the second sublayer are respectively arranged on the upper and lower sides of the second layer, and correspondingly, the first waveguides adjacent to each other along the signal transmission direction of the second waveguide can be respectively arranged in different sublayers. In this embodiment, a cross-layered layout mode is also proposed for the computing chip with wavelength division and mode division multiplexing distributed in series, thereby combining the double reuse mode and the layered arrangement of the lines so that the computing chip can accommodate more optical signal types while reducing the possibility of crosstalk between different signals. In other words, through the superposition of multiplexing modes and the position layout in physical space, the optical computing chip's tolerance to optical signal types is improved, while the difference in optical signals between adjacent waveguides can be increased, and the interference problem between lines can be alleviated.

[0074] In this embodiment, m first waveguides are arranged in the large-scale optical computing chip, and n row waveguides are arranged in the first waveguide (such as 1 first row waveguide and a intermediate row waveguides, that is, n=a+1). Correspondingly, m*n different types of optical signals can be coupled in a second waveguide, and the different types here refer to optical signals having different wavelengths or different orders.

[0075] That is to say, for an N*N (N=m*n) computing array, if a single order adjustment is used, n*m optical signals of different orders need to be set to complete the computing task. Once the chip size becomes larger (such as the specification of the computing chip is 124*124), 124 orders of optical signals will be required. On the contrary, the invention can reduce the necessary order to m, which is 1 / n of the original.

[0076] On the one hand, this reduction in order can reduce the pressure of setting the order, thereby reducing the performance requirements for the first and second signal conversion parts. On the other hand, it is also beneficial to reduce the probability of crosstalk between adjacent lines. Specifically, the reduction in order is also beneficial to increase the order difference between adjacent waveguides to alleviate the risk of crosstalk between adjacent waveguides.

[0077] In some embodiments, the specific order of the third optical signal input in the second waveguide decreases successively along a first direction, and correspondingly, the width of the second waveguide decreases successively along the first direction, wherein the first direction is a direction opposite to the transmission direction of the optical signal in the second waveguide.

[0078] Embodiment 2 Furthermore, for such a large-scale composite optical computing chip with multiple waveguides (such as a first waveguide also introducing a combined waveguide), the present invention also provides a design method to coordinate the size and computing accuracy of the optical computing chip. Figure 7 As shown, the method comprises the steps of: S101, providing an optical computing chip, wherein the optical computing chip is provided with n first waveguides, and the first waveguides include: m traveling waveguides; and the m traveling waveguides are arranged at a first pitch; It can be understood that the optical computing chip can be the composite optical computing chip described in any of the above embodiments.

[0079] S102, setting at least one set of different test conditions, the test conditions comprising: a first set value of m, and a second set value of the first interval; For example, in some embodiments, for optical computing chips with different specification ranges (such as different design size requirements), they may have some setting reference values, such as the first setting value can be selected from H1 reference values, and the second setting value can also be selected from H2 reference values. Alternatively, the reference values ​​required for these tests can also be manually set by engineers based on historical experience. Correspondingly, in the initial stage of design, multiple groups of test conditions can be set according to the above method to select appropriate design specifications through evaluation of multiple groups of test conditions.

[0080] S103, testing the optical computing chip according to at least one set of the test conditions, and correspondingly collecting at least one set of first test results, the test results including: calculation results output by m*n second waveguides; wherein the calculation result is the sum of the powers of the m*n optical signals output by the second waveguides, and the m traveling waveguides respectively use light of m different wavelengths as input optical signals; For example, in some embodiments, the optical computing chip may be simulated and tested using simulation software, or a physical chip may be run and tested.

[0081] S104, selecting at least one second test result satisfying a first rule from the first test results, wherein the first rule requires that a difference ratio of the calculated results is less than a number of a set composite threshold and greater than a set reference number, wherein the difference ratio is |calculated result-standard result| / standard result; For example, in some embodiments, in the first rule, a calculation result is considered to be relatively reliable only when the deviation between the calculation result and the standard result is less than a set composite threshold, and when multiple calculation results are determined to be reliable, the current test result is considered to be reliable as a whole.

[0082] In this embodiment, the calculation result refers to the sum of the optical powers calculated by the photodetector. Correspondingly, at least one standard matrix multiplication can be selected as a test scheme during the test process, and the standard matrix multiplication has a predetermined accurate calculation result (that is, a standard result). By running the standard matrix multiplication on the optical computing chip and comparing the actual calculation result output by it with the standard result, the accuracy of the optical computing chip can be verified. Therefore, the test process can make an overall assessment of the size, optical loss, and optical crosstalk problems of the composite optical computing chip.

[0083] S105, calculating at least one evaluation index L of the second test result, where L=αX+βY (equivalent to the first evaluation model); wherein X is a first relative index of a first set value, Y is a second relative index of a second set value, α is a first weight, and β is a second weight.

[0084] For example, in some embodiments, the first setting value is divided into at least two levels from small to large according to its numerical value, and each level corresponds to a relative value for describing the relative index. For example, it can be divided into level 1 and level 2, and its relative index can be set to 1 and 2 accordingly.

[0085] For example, in some embodiments, the second setting value is divided into at least two levels in ascending order according to its numerical value, and each level corresponds to a relative value for describing the relative indicator.

[0086] Of course, it can be understood that the first relative index is mainly used to describe the number of optical signals of different wavelengths, while the second relative index is mainly used to evaluate the size of the first interval, and the specific relative values ​​thereof can be freely adjusted.

[0087] S106, recommending corresponding test conditions to the user according to the evaluation index L as a preferred design solution.

[0088] For example, when the evaluation index L is relatively small, it means that the performance requirements of the corresponding signal conversion unit and the overall size of the optical computing chip are relatively limited, so it can be used as a preferred design solution.

[0089] Alternatively, in some embodiments, feasible design solutions can be quickly screened out through the above design solutions, so that engineers can quickly obtain initial design solutions, which is beneficial for subsequent rapid specification.

[0090] Furthermore, in some embodiments, the test conditions further include: order difference setting conditions. Specifically, the m first waveguides respectively input m optical signals of different orders (that is, the third optical signal mentioned above) into the second waveguide, and the order difference between adjacent optical signals (that is, the third optical signal input from the adjacent first waveguide to the second waveguide) is greater than the set order threshold. Correspondingly, the test conditions may include two or more order differences.

[0091] For example, in this embodiment, the multiple test conditions may be: test condition 1 (where m=2, first spacing=D1, order difference=1), test condition 2 (where m=3, first spacing=D1, order difference=1), test condition 3 (where m=2, first spacing=D2, order difference=1), test condition 4 (where m=3, first spacing=D2, order difference=1), test condition 5 (where m=2, first spacing=D1, order difference=2), test condition 6 (where m=3, first spacing=D1, order difference=2), test condition 7 (where m=2, first spacing=D2, order difference=2), test condition 8 (where m=3, first spacing=D2, order difference=2). Correspondingly, the 8 test conditions may be simulated and tested respectively to obtain the calculation results after their operation.

[0092] Correspondingly, the evaluation index can also adopt the following model: L=αX+βY+γZ1 (equivalent to the second evaluation model); wherein γ is the third weight, and Z1 is the third relative index of the order difference. For example, according to the numerical value of the order difference, it can also be divided into at least two levels from small to large, and each level corresponds to a relative value to describe the relative index.

[0093] In this embodiment, the performance requirements of the first and second signal conversion parts are comprehensively evaluated through the first and third relative indicators (in other words, this embodiment comprehensively analyzes the signal loss and crosstalk problems through the first and third relative indicators), and the size and crosstalk problems of the optical computing chip are synchronously evaluated through the second relative indicator. Therefore, this evaluation indicator model can achieve a certain balance between the wavelength division and mode division multiplexing mechanisms on the basis of taking into account the size scale, so that the loss caused by the wavelength difference and the error caused by the signal crosstalk of this scheme are controlled as a whole.

[0094] In particular, in some embodiments, the relative index may also be directly represented by the order difference.

[0095] Furthermore, in some embodiments, the test condition further introduces a setting parameter of the combiner waveguide, that is, a second spacing between the combiner waveguide and the second waveguide. Correspondingly, the combiner waveguide and the second waveguide are arranged with the second spacing, and correspondingly, the evaluation index is: L=αX+βY+δZ2 (equivalent to the third evaluation model); δ is the fourth weight, and Z2 is the third relative index of the second interval. According to the numerical value of the second interval, it can also be divided into at least two levels from small to large, and each level corresponds to a relative value for describing the fourth relative index.

[0096] Of course, in other embodiments, when higher performance requirements are placed on the optical computing chip, the fourth evaluation model, ie, L=αX+βY+γZ1+δZ2, may also be used.

[0097] In some embodiments, along the transmission direction of the second waveguide, the width of the second waveguide gradually increases.

[0098] In some embodiments, the test conditions also include: a fifth set value of n; correspondingly, the numerical value of the fifth set value is divided into at least two levels from small to large, and the level corresponds to a relative value describing a fifth relative indicator; correspondingly, the evaluation index L is: L=αX+βY+εZ3; wherein ε is the fifth weight, and Z3 is the fifth relative indicator.

[0099] For example, in Figure 6 In the optical computing chip of the specification shown, n is 2. Of course, as the specifications of the optical computing chip gradually expand, n can be flexibly selected and set according to the number of traveling waveguides in the first wavelength and the overall specifications of the optical computing chip.

[0100] In this embodiment, the composite conditions of the composite optical computing chip are preferably selected based on the two main dimensions of the number of wavelengths and the number of orders, so as to quickly find a better balance between reducing the technical difficulty of multi-wavelength light sources and the technical difficulty of multi-order light source settings.

[0101] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0102] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a computer terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0103] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A multiplexing-based optical computing chip, characterized in that: include: A plurality of first waveguides (1); a plurality of second waveguides (2) arranged opposite to the first waveguide, wherein the first waveguide (1) and the second waveguide (2) intersect to form a plurality of calculation areas; The calculation area comprises: a branch waveguide (3) derived from the first waveguide (1), the first end and the second end of the branch waveguide (3) being connected to the first waveguide and the second waveguide respectively, the branch waveguide being provided with a beam splitter (8), an intensity modulator (6), and a first signal converter (7) in sequence along the direction from the first end to the second end thereof, wherein the beam splitter is used to separate a specific proportion of the first optical signal from the input optical signal from the first waveguide according to a set beam splitting ratio; the intensity modulator is used to modulate the signal intensity of the input first optical signal to a specific value, and correspondingly generate and output a second optical signal; the first signal converter is used to set or maintain the second optical signal to a specific mode, and correspondingly generate and output a third optical signal, wherein the specific mode refers to that the third optical signal has a specific polarization direction and / or a specific order, and the third optical signal is transmitted to the corresponding second waveguide; wherein the third optical signals received at different positions of the same column of the second waveguide have different specific modes; A calculation unit is provided corresponding to the second waveguide, and the calculation unit is used to calculate the sum of the optical powers of the plurality of third optical signals.

2. The optical computing chip according to claim 1, characterized in that: The beam splitting part (8) adopts a symmetrical beam splitting structure, and correspondingly, the first waveguide comprises: at least one row bus waveguide (11); and a group of primary branch waveguides (12) extending from the row bus waveguide (11), wherein the primary branch waveguide (12) and the row bus waveguide (11) are connected by the symmetrical splitting structure to separate a primary optical signal of half the intensity from an input optical signal at a current position of the row bus waveguide (11); the primary branch waveguide (12) comprises: a first branch waveguide (121), and a second branch waveguide (122) extending from the first branch waveguide (121), wherein the second branch waveguide (122) and the first branch waveguide (121) are connected by the symmetrical splitting structure to separate a secondary optical signal of half the intensity from an optical signal at a current position, wherein the secondary optical signal is input into the intensity modulation unit, and the intensity modulation unit outputs the corresponding second optical signal.

3. The optical computing chip according to claim 2, characterized in that: The first waveguide also includes: a primary branch waveguide extending from the row bus waveguide (11), the primary branch waveguide being connected using a symmetrical light splitting structure to separate a primary light signal of half the intensity from an input light signal at a current position.

4. The optical computing chip according to claim 1, characterized in that: The specific order of the third optical signal input into the second waveguide decreases sequentially along a first direction, and correspondingly, the width of the second waveguide decreases sequentially along the first direction, wherein the first direction is a direction opposite to a transmission direction of the optical signal in the second waveguide; And / or, the calculation unit is a photodetector; And / or, the first signal conversion unit is an asymmetric directional coupler.

5. The optical computing chip according to claim 1, characterized in that: The second waveguide is a multimode waveguide.

6. The optical computing chip according to any one of claims 4 or 5, characterized in that: The branch waveguide further comprises: at least two first sub-waveguides (31), a combining waveguide (32) connecting the at least two first sub-waveguides (31), and a second sub-waveguide (33) for transmitting the signal of the combining waveguide to the second waveguide; Correspondingly, the first waveguide comprises: a first row of waveguides (41) arranged in sequence, and at least one middle row of waveguides (42) arranged in sequence; A plurality of first-row calculation units are arranged on the first-row waveguide (41), and the first-row calculation units include: a beam splitter arranged on the first-row waveguide (41), and a first sub-waveguide (31) connected through the beam splitter, the first sub-waveguide (31) being provided with an intensity modulation unit, and the output end of the first sub-waveguide (31) being connected to the corresponding combiner waveguide (32); The intermediate row waveguide (42) is provided with a plurality of intermediate row calculation units, the intermediate row calculation units comprising: a beam splitter provided on the intermediate row waveguide (42), and a first sub-waveguide (31) connected via the beam splitter, the first sub-waveguide (31) being provided with an intensity modulation unit, the first sub-waveguide being connected to the combiner waveguide (32) via a second signal conversion unit, wherein different row waveguides use light of different wavelengths as the input optical signal, and the second signal conversion unit is used to combine at least two optical signals of different wavelengths into one optical signal, and input the optical signal into the combiner waveguide (32); The second sub-waveguide (33) is connected to the second waveguide via the first signal conversion portion.

7. The optical computing chip according to claim 6, characterized in that: The first waveguide is arranged in the first layer, the second waveguide is arranged in the second layer, and the first layer and the second layer are connected via the second sub-waveguide (33).

8. A design method for an optical computing chip based on claim 7, characterized in that: Includes steps: S101, providing an optical computing chip, wherein the optical computing chip is provided with n first waveguides, and the first waveguides include: m running waveguides, the m running waveguides include: 1 first running waveguide and a middle running waveguides; and the m running waveguides are arranged at a first spacing; S102, setting at least one set of different test conditions, the test conditions comprising: a first set value of m, and a second set value of the first interval; S103, testing the optical computing chip according to at least one set of the test conditions, and correspondingly collecting at least one set of first test results, wherein the first test results include: calculation results output by m*n second waveguides; wherein the calculation result is the sum of powers of m*n optical signals output by the second waveguides, and the m traveling waveguides respectively use light of m different wavelengths as input optical signals; S104, selecting at least one second test result satisfying a first rule from the first test results, wherein the first rule requires that a difference ratio satisfying the calculation result is less than a number of a set composite threshold and greater than a set reference number, wherein the difference ratio is |calculation result-standard result| / standard result; S105, calculating at least one evaluation index L of the second test result according to the corresponding test condition, wherein L=αX+βY; wherein X is a first relative index of a first set value, Y is a second relative index of a second set value, α is a first weight, and β is a second weight; wherein the numerical value of the first set value is divided into at least two levels from small to large, and the levels correspond to a relative value for describing the first relative index; the numerical value of the second set value is divided into at least two levels from small to large, and the levels correspond to a relative value for describing the second relative index; S106, recommending corresponding test conditions to the user according to the evaluation index L as a preferred design solution.

9. The method for designing an optical computing chip according to claim 8, characterized in that: The order difference between the front and rear third optical signals coupled into the second waveguide is greater than a set order threshold, and the test condition further includes: a third set value of the order difference; the value of the third set value is divided into at least two levels from small to large, and the level corresponds to a relative value for describing a third relative index of the third set value; Correspondingly, the evaluation index L is: L=αX+βY+γZ1, γ is the third weight, and Z1 is the third relative index.

10. The method for designing an optical computing chip according to claim 8, characterized in that: The test condition further includes: a fifth setting value of n; correspondingly, the value of the fifth setting value is divided into at least two levels from small to large, and the level corresponds to a relative value describing the fifth relative index; Correspondingly, the evaluation index L is: L = αX + βY + εZ3; Among them, ε is the fifth weight, and Z3 is the fifth relative index.

Images