Optical calculation module and optical calculation array

Through cross-set optical waveguides and electrical buses, combined with independent photoelectric signal processing units, the optical signal is directly converted into current signals, solving the challenges of existing optical computing solutions in high-frequency switching, low power consumption and large-scale integration, and achieving an efficient and low-power optical computing module design.

CN120065431AActive Publication Date: 2025-05-30LIGHTSTANDARD CO LTD
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202510528072.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-04-25
Publication Date
2025-05-30
Estimated Expiration
2045-04-25

AI Technical Summary

Technical Problem

Existing optical computing solutions have significant challenges in device structure, power consumption and trainability, and are difficult to meet the needs of high-frequency switching, low-power consumption and large-scale integration.

Method used

The cross-set input optical waveguide and electrical bus are adopted, combined with an independent photoelectric signal processing unit, and encode and adjust the photoelectric characteristics of the light absorption layer to realize the direct conversion of the optical signal into a current signal and processed through the electrical bus.

Benefits of technology

It achieves the requirements of high-frequency switching, low power consumption and large-scale integration while ensuring computing accuracy, avoids the wiring difficulties and large-scale problems of traditional discrete structures, and improves integration density and energy efficiency performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120065431A_ABST
    Figure CN120065431A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field, and particularly relates to an optical calculation module and an optical calculation array, the optical calculation module comprises an input optical waveguide and an electric bus which are arranged in a crossed mode, the input end of the input optical waveguide is provided with an optical splitter, and the output end of the optical splitter is connected with the first input end of a photoelectric signal processing unit through a transmission optical waveguide; different from the scheme of a discrete optical calculation unit in the prior art, the independent photoelectric signal processing unit is adopted in the optical calculation unit, rapid conversion from an unmodulated optical signal to an electric signal can be achieved, and the optical calculation unit has the advantages that the optical calculation unit is simple in structure, convenient to use and high in practicability. The size of the detector basically remains unchanged compared with the size of a detector device in the prior art, so that the size of a single optical calculation unit is smaller, a larger-scale matrix is realized compared with an existing scheme, the problem of difficulty in wiring of an existing discrete device is avoided, and the service life of the detector is the same as that of a conventional silicon-based device. And frequent switching requirements of training levels can be met.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of semiconductor optoelectronic devices, and particularly relates to an optical computing module and an optical computing array. Background Art

[0002] Optical computing technology has become an important research direction for breaking through the limitations of traditional computing due to its high parallelism, low latency, and potential high energy efficiency characteristics. However, existing optical computing solutions still pose significant challenges in terms of device structure, power consumption, and trainability, restricting their application in large-scale computing and training tasks.

[0003] Currently, the core structure of optical computing units is typically based on a combination of a modulator and a detector. The modulator is responsible for encoding or performing computing operations on optical signals, while the detector completes optoelectronic conversion. This discrete design makes it difficult to further optimize the area and power consumption of a single computing unit, thereby limiting the integration scale and energy efficiency performance of optical computing chips. Specifically, the above-mentioned discrete prior art solutions have the following limitations: 1. Optical computing architectures based on phase change materials (such as the applicant's prior application document CN118394171B): The switching speed of phase change materials is limited, making it difficult to meet the requirements of high-frequency weight updates in training tasks; and during frequent switching, the dynamic power consumption will increase significantly, resulting in a decrease in overall energy efficiency.

[0004] 2. Optical computing architectures based on carrier absorption (such as the prior art CN109960310A): Its power consumption is relatively high, mainly due to the carrier injection loss when the modulator is forward-conducting, and the combination of discrete modulators and detectors further increases the unit area, restricting the integration density.

[0005] That is to say, the above-mentioned discrete design solutions are greatly limited in size and also have relatively high power consumption.

[0006] Therefore, there is an urgent need for an optical computing unit that can meet the requirements of high-frequency switching, low power consumption, and large-scale integration while ensuring computing accuracy. Summary of the Invention

[0007] The purpose of the present invention is to provide an optical computing module and an optical computing array to partially alleviate or solve the above deficiencies and meet the requirements of high-frequency switching, low power consumption, and large-scale integration.

[0008] To solve the above-mentioned technical problems, the present invention specifically adopts the following technical solutions: In a first aspect of the present invention, there is provided an optical computing module, comprising: An input optical waveguide and an electrical bus are arranged crosswise. A beam splitter is arranged at the input end of the input optical waveguide. The output end of the beam splitter is connected to a first input end of an optoelectronic signal processing unit through a transmission optical waveguide. The electrical output end of the optoelectronic signal processing unit is connected to the electrical bus; Wherein, the optoelectronic characteristics of the light absorption layer in the optoelectronic signal processing unit can be changed by writing a control signal to its second input end, and the optoelectronic signal processing unit can convert the optical signal input from the first input end into a corresponding current signal and send it to the electrical bus through the electrical output end; the optoelectronic characteristics include the light absorption coefficient of the light absorption layer; Wherein, the product of the multiplier value and the multiplicand value is encoded in the current signal.

[0009] As an improvement, the optoelectronic signal processing unit uses pure germanium as the light absorption layer; The optoelectronic signal processing unit includes: a silicon substrate and an insulating layer deposited on the silicon substrate; a waveguide layer is arranged on the insulating layer; a germanium absorption layer is embedded in the waveguide layer; the first depth at which the germanium absorption layer is embedded in the waveguide layer is less than the thickness of the waveguide layer, or the second depth at which the germanium absorption layer is embedded in the waveguide layer is equal to the thickness of the waveguide layer.

[0010] As an improvement, a heating unit is arranged at the top and / or bottom of the germanium absorption layer, and the heating unit extends along the length direction of the germanium absorption layer; or, Heating units are respectively arranged on both sides of the germanium absorption layer, and the heating units extend along the length direction of the germanium absorption layer; or, Heating unit groups are respectively arranged on both sides of the germanium absorption layer, and each heating unit group on each side includes a plurality of heating units evenly spaced along the length direction of the germanium absorption layer. Among them, the heating levels of two adjacent heating units on the same side are respectively primary heating and secondary heating.

[0011] As an improvement, the germanium absorption layer includes a light guiding region and a light absorption region arranged in sequence along the light propagation direction, Wherein, the cross-sectional area of the light guiding region gradually increases from the incident end along the direction close to the light absorption region; the cross-sectional area of the light absorption region is the same along the light propagation direction.

[0012] As an improvement, when viewed from above, the width of the light guiding region gradually increases from the incident end along the light propagation direction; and when viewed from the side, the width of the light guiding region gradually decreases from the top to the bottom; or, the interface between the light guiding region and the waveguide layer is fan-shaped; and / or, when viewed from the side, the light absorption region is rectangular.

[0013] The second aspect of the present invention lies in providing an optical computing array, including: A crossbar switch matrix architecture, which is formed by the intersection of multiple rows of parallel input optical waveguides and multiple columns of parallel electrical buses; the optical input signal is transmitted through the input optical waveguides; A plurality of optical splitters, which are arranged at the front ends of the intersection nodes of each row of optical input waveguides and each column of electrical buses; A plurality of optoelectronic signal processing units, each of the optoelectronic signal processing units corresponding to one of the cross nodes, and a first input end of the optoelectronic signal processing unit being connected to an output end of the optical splitter through a transmission optical waveguide, and an electrical output end of the optoelectronic signal processing unit being connected to the electrical bus in the corresponding cross node; Wherein, the optoelectronic characteristics of the optical absorption layer in the optoelectronic signal processing unit can be changed by writing a control signal to its second input end, and the optoelectronic signal processing unit can convert the optical signal input to the first input end into a corresponding current signal and send it to the electrical bus through the electrical output end; the optoelectronic characteristics include an optical absorption coefficient; wherein, the product of the multiplier value and the multiplicand value is encoded in the current signal.

[0014] As an improvement, the optoelectronic signal processing unit uses pure germanium as the optical absorption layer; the optoelectronic signal processing unit includes: a silicon substrate and an insulating layer deposited on the silicon substrate; a waveguide layer is arranged on the insulating layer; a germanium absorption layer is embedded in the waveguide layer; a first depth at which the germanium absorption layer is embedded in the waveguide layer is less than the thickness of the waveguide layer, or a second depth at which the germanium absorption layer is embedded in the waveguide layer is equal to the thickness of the waveguide layer.

[0015] As an improvement, a heating unit is arranged at the top and / or bottom of the germanium absorption layer, and the heating unit extends along the length direction of the germanium absorption layer; or, heating units are respectively arranged on both sides of the germanium absorption layer, and the heating units extend along the length direction of the germanium absorption layer; or, heating unit groups are respectively arranged on both sides of the germanium absorption layer, and each heating unit group on each side includes a plurality of heating units uniformly spaced along the length direction of the germanium absorption layer, wherein, heating levels of two adjacent heating units on the same side are respectively a first-level heating and a second-level heating.

[0016] As an improvement, the germanium absorption layer includes an optical guiding region and an optical absorption region arranged in sequence along the light propagation direction, wherein, a cross-sectional area of the optical guiding region gradually increases from the incident end along the direction close to the optical absorption region; a cross-sectional area of the optical absorption region is the same along the light propagation direction.

[0017] As an improvement, a Pad is connected to a second input end of each of the optoelectronic signal processing units, and a Pad is arranged on each column of the electrical buses; and / or, an EOM is arranged on each row of the input optical waveguides.

[0018] The principle and beneficial technical effects of the present invention are as follows: This solution uses an independent optoelectronic signal processing unit to simultaneously achieve modulation and detection. Specifically, the optical signal input to the optoelectronic signal processing unit is unmodulated and is a stable value. When it is processed by the optoelectronic signal processing unit, the absorbed part of the optical signal is directly converted into a current signal as output for operations such as summation, without the need to separately convert the unabsorbed optical signal (i.e., the remaining optical signal) into an electrical signal through a detector. Moreover, the volume of the optoelectronic signal processing unit remains basically unchanged compared to the size of the detector device in the prior art, enabling the size of a single optical computing module to be smaller, thus achieving a larger-scale matrix compared to the existing solution, and avoiding the problems of difficult wiring and large volume in the prior art (such as in the field of optical communication) caused by the discrete structure of a modulator + detector (this discrete structure first modulates the optical signal using a modulator and then inputs the modulated optical signal into a detector to convert it into an electrical signal). In addition, the lifespan performance of the optoelectronic signal processing unit in this solution is the same as that of conventional silicon-based devices, which can meet the frequent switching requirements at the training level. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. In all the drawings, similar elements or parts are generally denoted by similar reference numerals. In the drawings, the elements or parts do not necessarily draw to actual scale. Obviously, the following-described drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is a schematic side view of the prior art; Figure 2 It is a schematic side view of the fourth embodiment of the present invention; Figure 3 It is a schematic top view of the fourth embodiment of the present invention; Figure 4 It is a top view of the progressive section in the fourth embodiment of the present invention; Figure 5 It is an exemplary side view of the modulator in the third embodiment of the present invention; Figure 6 It is another exemplary side view of the modulator in the third embodiment of the present invention; Figure 7 It is another exemplary side view of the modulator in the third embodiment of the present invention; Figure 8 It is a top view of the modulator in the third embodiment of the present invention; Figure 9 This is the front view of the germanium absorption layer in Embodiment 3 and the germanium strip in Embodiment 4 of the present invention; Figure 10 This is the top view of the germanium absorption layer in Embodiment 3 and the germanium strip in Embodiment 4 of the present invention; Figure 11 This is the schematic diagram of the optical computing module in Embodiment 1 of the present invention; Figure 12 This is the schematic diagram of the optical computing array in Embodiment 2 of the present invention; Figure 13 This is the schematic diagram of another optical computing array in Embodiment 2 of the present invention; Figure 14 This is the schematic diagram of yet another optical computing array in Embodiment 2 of the present invention; Figure 15 This is the schematic diagram of the optical computing module in Embodiment 5 of the present invention; Figure 16 This is the schematic diagram of the signal comparison by the detection unit in Embodiments 5 - 7 of the present invention; Figure 17 This is the flowchart of the detection method in Embodiment 7 of the present invention.

[0021] Markings in the figure: 1. Silicon substrate; 2. Insulating layer; 3. Ridge optical waveguide / waveguide layer; 31. Silicon ridge; 32. Flat layer; 4. Electrode; 5. Germanium strip / germanium absorption layer; 51. Gradual section / light guiding region; 52. Straight section / light absorption region; 6. Silicon strip; 7. Heating unit; 101. Photoelectric signal processing unit; 102. Input optical waveguide; 103. Transmission optical waveguide; 104. Electrical bus; 105. Beam splitter. Detailed implementation manners

[0022] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] In this text, suffixes such as "module", "component", or "unit" used to represent elements are only for facilitating the description of the present invention and have no specific meaning in themselves. Therefore, "module", "component", or "unit" can be used interchangeably. In this text, terms such as "upper", "lower", "inner", "outer", "front", "rear", "one end", "the other end", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for facilitating the description of the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation on the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0024] In this text, unless otherwise clearly specified and defined, terms such as "installed", "provided with", "connected", etc. shall 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, and can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations. In this text, "a plurality of" means two or more, that is, it includes two, three, four, five, etc.

[0025] In this text, the "optical splitter" can utilize components such as optical waveguides, optical couplers, heaters, phase shifters, etc. with specific structures to achieve the distribution of optical signals inside the chip in an active or passive manner.

[0026] In this text, the Pad is used to transmit electrical signals or optical signals in the optical computing unit or the electrical bus; the EMO is an optical modulator based on the electro-optic effect, which controls the power, phase, and polarization of the laser beam through an electronic control signal and is used to preliminarily modulate the input optical signal, so as to provide an optical signal that meets the processing range of the optical signal processing unit.

[0027] Embodiment 1 This embodiment provides an optical computing unit, including: An input optical waveguide and an electrical bus arranged crosswise, a splitter is provided at the input end of the input optical waveguide, the output end of the splitter is connected to the first input end of an optoelectronic signal processing unit through a transmission optical waveguide, and the electrical output end of the optoelectronic signal processing unit is connected to the electrical bus.

[0028] Among them, the optoelectronic characteristics of the light absorption layer in the optoelectronic signal processing unit can be changed by writing a control signal to its second input end, and the optoelectronic signal processing unit can convert the optical signal input from the first input end into a corresponding current signal and send it to the electrical bus through the electrical output end; the optoelectronic characteristics include the light absorption coefficient of the light absorption layer. Among them, the product of the multiplier value and the multiplicand value is encoded in the current signal.

[0029] The optical computing unit in the prior art combines discrete modulators and detectors. Part of the optical signal input into the modulator is absorbed, and the unabsorbed part of the optical signal (the remaining optical signal) is output from the optical output end of the modulator and enters the detector to be converted into an electrical signal, and then enters the data terminal for processing; while the optical computing unit in this application is completely different from the above scheme. It uses an independent optoelectronic signal processing unit, which directly converts the absorbed part of the optical signal into an electrical signal as the output, performs operations such as summation, and then inputs it into the backend data terminal for data processing, without the need for additional processing of the remaining optical signal, so as to directly obtain the processed electrical signal.

[0030] The volume of the optoelectronic signal processing unit in this scheme is basically the same as the size of the detector device in the prior art. Furthermore, it can make the size of a single optical computing unit smaller, thus realizing a larger-scale matrix compared with the existing scheme, and avoiding the problem of difficult wiring of existing discrete devices. Its lifespan performance is the same as that of conventional silicon-based devices and can meet the frequent switching requirements at the training level.

[0031] Embodiment 2 This embodiment provides an optical computing array, including a plurality of optical computing modules in Embodiment 1, including: A crossbar switch matrix architecture, which is formed by the intersection of multiple rows of mutually parallel input optical waveguides and multiple columns of mutually parallel electrical buses; the optical input signal is transmitted through the input optical waveguides. A plurality of optical splitters, which are arranged at the front ends of the intersection nodes of each row of optical input waveguides and each column of electrical buses. A plurality of optoelectronic signal processing units, each optoelectronic signal processing unit corresponding to one of the cross nodes, and the first input end of the optoelectronic signal processing unit is connected to the output end of the optical splitter through a transmission optical waveguide, and the electrical output end of the optoelectronic signal processing unit is connected to the electrical bus in the corresponding cross node. Among them, the optoelectronic characteristics of the light absorption layer in the optoelectronic signal processing unit can be changed by writing a control signal to its second input end, and the optoelectronic signal processing unit can convert the optical signal input from the first input end into a corresponding current signal and send it to the electrical bus through the electrical output end; the optoelectronic characteristics include the light absorption coefficient. Wherein, the product of the multiplication value and the value to be multiplied is encoded in the current signal.

[0032] In some embodiments, a Pad is connected to the second input end of each of the optoelectronic signal processing units, and a Pad is provided on each column of the electrical buses; and / or, an EOM is provided on each row of the input optical waveguides.

[0033] Embodiment III This embodiment provides an optoelectronic signal processing unit, which uses pure germanium as the optical absorption layer, as an example of the optoelectronic signal processing unit in Embodiment I and / or Embodiment II: The optoelectronic signal processing unit includes: a silicon substrate and an insulating layer deposited on the silicon substrate; a waveguide layer is provided on the insulating layer; a germanium absorption layer is embedded in the waveguide layer; the first depth at which the germanium absorption layer is embedded in the waveguide layer is less than the thickness of the waveguide layer (this is "shallow etching" at this time), or, the second depth at which the germanium absorption layer is embedded in the waveguide layer is equal to the thickness of the waveguide layer (this is "full etching" at this time).

[0034] In the above embodiments, when the second depth is equal to the thickness of the waveguide layer 3, that is, taking the insulating layer 2 as the etching end point, on the one hand, it can make the germanium absorption layer 5 have a larger cross-section, so that the coincidence time between the modulation region and the optical field is the longest. At the same length, the responsivity is higher; at the same responsivity, the size is the smallest. On the other hand, the etching depth is easy to control, which can ensure the consistency of the etching groove depth between different modulators. In the traditional shallow etching process, due to the relatively shallow etching depth and difficult to accurately control, it is easily affected by factors such as uneven silicon layer thickness, resulting in inconsistent etching depth, which in turn affects the performance uniformity of the modulators. The etching groove process with a depth reaching the insulating layer 2 has better stability, which is beneficial to the large-scale production of high-quality modulators.

[0035] In some embodiments, the first depth of the waveguide layer may also be greater than the thickness of the waveguide layer, and this is "over-etching" at this time.

[0036] In some embodiments, a heating unit 7 is provided on at least one side of the germanium absorption layer 5.

[0037] In some specific embodiments, a heating unit 7 is provided at the top and / or bottom of the germanium absorption layer 5, and the heating unit 7 extends along the length direction of the germanium absorption layer 5; alternatively, heating units 7 are respectively provided on both sides of the germanium absorption layer 5, and the heating units 7 extend along the length direction of the germanium absorption layer 5; alternatively, heating unit groups 7 are respectively provided on both sides of the germanium absorption layer 5, and each heating unit group 7 on each side includes a plurality of heating units 7 evenly spaced along the length direction of the germanium absorption layer 5, wherein the heating levels of two adjacent heating units 7 on the same side are primary heating and secondary heating respectively; the heating unit 7 is used to change the optoelectronic characteristics of the light absorption layer.

[0038] Completely different from the traditional idea of relying on a high electric field strength to increase the absorption coefficient to ensure the uniformity of the photocurrent density, the present invention provides a thermal compensation mechanism only for the end effect. Specifically, by providing a heating unit 7 on at least one side (bottom and / or top, or front and back sides) of the germanium absorption layer 5, the heating unit 7 acts on the germanium absorption layer 5 along the length direction of the germanium absorption layer to ensure the consistency and uniformity of the temperature of the entire germanium absorption layer 5, thereby increasing the valence band electrons that can transition, and then changing the light absorption coefficient of the germanium absorption layer 5 while reducing the operating voltage of the electro-absorption modulator. In this process, only the final light absorption rate (i.e., the ratio of the input light to the output light / input light) needs to be ensured, without considering the uniformity of the light absorption coefficient of the germanium absorption layer 5.

[0039] That is to say, the present application provides a thermal compensation mechanism for double-regulating the light absorption coefficient only for the end effect. First, by heating the germanium absorption layer to adjust the light absorption coefficient of the germanium absorption layer from the germanium modulator itself, and then, by adjusting the input voltage to further adjust the light absorption coefficient of the germanium absorption layer. That is to say, the adjustment requirement for the input voltage in this process is not high, that is, the response speed of the computing array can be ensured while ensuring low energy consumption, and thus the requirements of a large-scale matrix can be met.

[0040] In some embodiments, the germanium modulator further includes: A temperature monitoring module for real-time monitoring of the actual temperature value of the germanium absorption layer 5.

[0041] A first judgment module for judging whether the actual temperature value monitored by the temperature monitoring module is greater than a first preset temperature threshold. If the actual temperature value is greater than the first preset temperature threshold, a first control signal indicating to stop heating is generated and sent to the heating unit 7 / the heating unit group; if the actual temperature value is less than the first preset temperature threshold, a second control signal indicating to continue monitoring is generated and sent to the temperature monitoring module; and / or, A temperature monitoring module for monitoring in real time the actual temperature value of the germanium absorption layer 5 after being heated by the heating unit 7 at the first-level heating level and in the heating state.

[0042] A second judgment module for judging whether the actual temperature value monitored by the temperature monitoring module reaches a preset target temperature threshold after a preset time. If not, a third control signal is generated and sent to activate the heating unit 7 at the second-level heating level and in the dormant state, so that the actual temperature of the germanium absorption layer 5 reaches the preset target temperature threshold.

[0043] Among them, the first temperature threshold can be obtained through a large number of experiments in advance or preset by the user according to experience. When the actual temperature is near the first preset temperature threshold, the number of valence band electrons that can transition tends to be maximized, but at the same time, it will not cause too many adverse effects on the material itself.

[0044] The division of the heating levels (first level and second level) is only for distinguishing the two, and cannot limit their heating power or other characteristics. When only the first-level heating is turned on, the heating power is small, and when both the first-level heating and the second-level heating are turned on at the same time, the heating power increases, and the temperature of the germanium absorption layer 5 can also be correspondingly increased.

[0045] The present invention regulates the temperature of the germanium absorption layer 5 through a real-time feedback mechanism. On the one hand, it can make the number of valence band electrons that can transition tend to be maximized as much as possible. On the other hand, it can improve the stability of the temperature of the germanium absorption layer 5 and prevent the problem that the optical signal emitted by the germanium absorption layer 5 becomes unstable due to frequent temperature changes.

[0046] In some embodiments, the germanium absorption layer includes a light guiding region and a light absorption region arranged in sequence along the light propagation direction. Among them, the cross-sectional area of the light guiding region gradually increases from the incident end along the direction close to the light absorption region; the cross-sectional area of the light absorption region is the same along the light propagation direction.

[0047] In some specific embodiments, the longitudinal section of the light guiding region 51 is a right trapezoid, and the hypotenuse of the right trapezoid corresponds to the interface between the light guiding region 51 and the waveguide layer 3, and the angle α between it and the bottom edge is 86°-89°; the longitudinal section of the light absorption region 52 is a rectangle; or, the interface between the light guiding region 51 and the waveguide layer 3 is a fan shape. Among them, the longitudinal section refers to a vertical plane parallel to the length direction of the germanium absorption layer.

[0048] When viewed from above, the width of the light guiding region gradually increases along the light propagation direction from the incident end; and when viewed from the side, the width of the light guiding region gradually decreases from the top to the bottom; alternatively, the interface between the light guiding region and the waveguide layer is fan-shaped; and / or, the cross-section of the light absorption region is rectangular.

[0049] The purpose of the above settings is to achieve a smooth transition of the refractive index. Since the refractive index of silicon is about 3.4 and the refractive index of germanium is about 4.4, through this gradual structural change, the refractive index can gradually transition from the waveguide layer 3 (silicon waveguide) to germanium, reducing the reflection of light at the incident end. When light propagates at the interface of media with different refractive indices, the greater the refractive index difference, the more serious the reflection. This gradual transition structure can allow light to enter the germanium absorption layer 5 more smoothly from the silicon waveguide, improving the light coupling efficiency and thus enhancing the light absorption ability of the modulator.

[0050] Ideally, the light guiding region is a conical shape that is smaller at the front and larger at the back. The conical structure can achieve a more perfect gradual change in the refractive index, minimizing the reflection of light when it enters the germanium absorption layer 5 and theoretically maximizing the light coupling efficiency.

[0051] Due to process limitations, a conical light guiding region incurs high costs. Therefore, in this embodiment, in order to reduce the process difficulty, it is only necessary to make the cross-sectional area of the light guiding region gradually increase along the direction close to the light absorption region 52 from the incident end. For example, the light guiding region is a triangle that is narrow at the front and wide at the back when viewed from above; and a trapezoid that is larger at the top and smaller at the bottom when viewed from the front. Although it is different from the ideal conical shape, the trapezoidal structure can still achieve a gradual change in the refractive index to a certain extent, reducing light reflection and is a feasible solution under the existing process conditions.

[0052] That is to say, by restricting the shape of the light guiding region in the present invention, the cross-sectional area of the light guiding region gradually increases both in height and length from the incident end, which can significantly reduce the reflection of light at the incident end when light propagates in the light absorption layer (germanium absorption layer) with a larger cross-sectional area, improving the light absorption rate of the modulator. That is to say, the present application provides a full-etching solution with a thermal compensation mechanism that can reduce light reflection.

[0053] In some embodiments, the width and length of the heating unit 7 located at the top and / or bottom of the germanium absorption layer 5 are both smaller than the width and length of the germanium absorption layer 5. In this way, while ensuring the heating effect, unnecessary heat waste is not caused.

[0054] In summary, for the millisecond-level frequent switching requirements in the deep learning training scenario, the present application provides a high-response optical computing module / computing array with a thermal compensation mechanism that can reduce light reflection, so as to achieve the design of a low-power large-scale integrated computing array under the condition of meeting frequent switching.

[0055] Contrary to the existing methods that use phase change materials for modulation or change the carrier concentration by injecting current or applying voltage in doping to change the light absorption coefficient, the present application provides an optical computing array based on a reverse-biased germanium modulator, and a solution for multi-control of the light absorption characteristics of the germanium absorption layer at the same time.

[0056] First of all, the dark current of the optoelectronic signal processing unit with a pure germanium absorption layer in the present application is extremely low under reverse bias, and the photocurrent during its operation is at least two orders of magnitude smaller than that of a forward-conducting carrier absorption modulator or a phase change material modulator, thereby effectively reducing the power consumption of a single device during operation.

[0057] Furthermore, a heating unit is arranged on at least one side of the germanium absorption layer and is combined with a real-time feedback mechanism to ensure the consistency and uniformity of the temperature of the entire germanium absorption layer, thereby increasing the available valence band electrons that can transition, and further changing the light absorption coefficient of the germanium absorption layer while reducing the operating voltage of the optoelectronic signal processing unit.

[0058] Even further, for the case where the light absorption cross-sectional area of the light absorption region (germanium absorption layer) is relatively large (such as "full etching"), the present application designs the germanium absorption layer in zones (light guiding zone and light absorption zone), and guides the optical signal transmitted from the optical waveguide through a gradually changing interface formed by the gradual change of the height and width of the light guiding zone, that is, the cross-sectional area of the light guiding zone becomes gradually larger both in height and length from the incident end, which can greatly reduce the reflection of light at the incident end when propagating in the light absorption layer (germanium absorption layer) with a larger cross-sectional area, and improve the light absorption rate of the optoelectronic signal processing unit.

[0059] Embodiment 4 This embodiment provides an optoelectronic signal processing unit, which uses pure germanium as the light absorption layer. Different from Embodiment 3: The optoelectronic signal processing unit includes: a silicon substrate 1 and an insulating layer 2 deposited on the silicon substrate 1; an optical waveguide is arranged on the insulating layer 2; an etching groove reaching the insulating layer 2 or an etching groove embedded in the insulating layer 2 is opened on the optical waveguide; a germanium strip 5 is filled in the etching groove as the light absorption region, and electrodes 4 are arranged on both sides of the etching groove.

[0060] The bottom of the germanium strip 5 is flush with the upper surface of the insulating layer 2, or the bottom of the germanium strip 5 is embedded in the insulating layer 2, and the top of the germanium strip 5 protrudes from the upper surface of the optical waveguide.

[0061] In this embodiment, compared with Figure 1In the prior art for the shallow etching of optical waveguides, the etched groove depth reaches the insulating layer 2, that is, the bottom of the etched groove is flush with the insulating layer 2 or embedded in the insulating layer. This enables the bottom of the germanium strip 5 in the etched groove to be flush with the upper surface of the insulating layer, or the bottom of the germanium strip 5 to be embedded in the insulating layer 2, while the top of the germanium strip 5 can protrude from the upper surface of the optical waveguide. This allows the germanium strip 5 to have a larger cross-section, thereby maximizing the coincidence time between the detection region and the light field. At the same length, the responsivity is higher; at the same responsivity, the size is minimized.

[0062] In addition, using the insulating layer 2 as the etching end point makes it easy to control the etching depth, ensuring the consistency of the etched groove depth between different detectors. In traditional shallow etching processes, due to the shallow etching depth and difficulty in precise control, it is easily affected by factors such as uneven silicon layer thickness, resulting in inconsistent etching depths and thus affecting the performance uniformity of the detectors. The etching groove process reaching the insulating layer has better stability, which is beneficial for large-scale production of high-quality detectors.

[0063] More specifically, in this embodiment, the optical waveguide is a ridge optical waveguide 3, including a flat layer 32 and a silicon ridge 31 on the flat layer 32; the electrode 4 is disposed on the flat layer 32; and the etched groove penetrates through the silicon ridge 31.

[0064] The ridge optical waveguide 3 utilizes the propagation characteristics of light in media with different refractive indices. The refractive index of the silicon ridge 31 is higher than that of the surrounding environment, forming a confinement of light. When light propagates in the ridge optical waveguide, it is restricted to the region of the silicon ridge 31, reducing light scattering and loss, improving the light transmission efficiency, and providing a stable light transmission channel for the light detection process of the detector.

[0065] Disposing the electrode 4 on the flat layer 32 facilitates connection to the external circuit. The flat layer 32 provides a large planar area, which is beneficial for the fabrication and layout of the electrode 4, ensuring good electrical contact between the electrode 4 and the external circuit. At the same time, this layout avoids direct interference of the electrode 4 with the light propagation path, reducing light absorption or scattering caused by the presence of the electrode during light propagation, ensuring stable light signal transmission in the silicon ridge, and improving the light detection performance of the detector.

[0066] The etched groove penetrates through the silicon ridge 31, providing sufficient space for the filling of the germanium strip 5. The germanium strip 5, as the light absorption region, is in direct close contact with the light propagation path. When light propagates in the ridge optical waveguide, it passes through the germanium strip 5 in the etched groove, and the germanium strip 5 can fully absorb the light signal and convert it into an electrical signal. The design of the etched groove penetrating through the silicon ridge 31 ensures sufficient interaction between light and the germanium strip 5, enhancing the light absorption effect, thereby improving the responsivity of the detector and enhancing the detection sensitivity of the detector to light signals.

[0067] Such as Figure 3As shown, in some embodiments, silicon strips 6 are connected to both ends of the germanium strip 5. As Figure 4 shown, the germanium strip 5 includes a tapered section 51 (also referred to as the light guiding region herein) and a straight section 52 (also referred to as the light absorption region herein); the tapered section 51 is arranged at the incident end of the germanium strip 5, and the cross-sectional area of the tapered section 51 increases along the light incident direction (i.e., the light propagation direction); the shape of the exit end of the silicon strip 6 is matingly connected to the tapered section 51.

[0068] The purpose of the above arrangement is to achieve a smooth transition of the refractive index. Since the refractive index of silicon is about 3.4 and that of germanium is about 4.4, through this gradual structural change, the refractive index can gradually transition from the silicon waveguide to germanium, reducing the reflection of light at the incident end. When light propagates at the interface of media with different refractive indices, the greater the refractive index difference, the more serious the reflection. This gradual transition structure allows light to enter the germanium strip more smoothly from the silicon waveguide, improving the light coupling efficiency and thus enhancing the light absorption ability of the detector.

[0069] Ideally, the tapered section 51 is a conical shape with a smaller front and a larger rear. The conical structure can achieve a more perfect refractive index gradient, minimizing the reflection of light when entering the germanium strip, and theoretically maximizing the light coupling efficiency and the responsivity of the detector.

[0070] Due to process limitations, a conical tapered section 51 incurs high costs. Therefore, in this embodiment, to reduce the process difficulty, the tapered section 51 is a triangle with a narrower front and a wider rear when viewed from above and a trapezoid with a larger upper part and a smaller lower part when viewed from the front, as Figure 4 、 Figure 5 shown. Although it is different from the ideal conical shape, the trapezoidal structure can still achieve a refractive index gradient to a certain extent, reducing light reflection, and is a feasible solution under the existing process conditions. More preferably, the length of the tapered section 51 along the light propagation direction is the same as the length of the straight section 52 along the light propagation direction.

[0071] In some other embodiments, a silicon-germanium alloy transition layer is also provided between the incident end of the germanium strip 5 and the silicon strip 6. The silicon-germanium alloy transition layer is gradient-doped, and the germanium concentration increases from the silicon strip side to the germanium strip side. The germanium concentration increases from 10% on the silicon strip side to 100% on the germanium strip side. By providing the silicon-germanium alloy transition layer and increasing the germanium concentration from 10% on the silicon strip 6 side to 100% on the germanium strip side, the refractive index can gradually transition from the refractive index of silicon to that of germanium. In the transition layer, as the germanium concentration increases, its refractive index also gradually increases, forming a region with a gradual change in refractive index between the silicon strip and the germanium strip. In this way, during the propagation of light, since the refractive index changes gradually, the reflection caused by the sudden change in refractive index is reduced, enabling light to enter the germanium strip more smoothly from the silicon strip and improving the light coupling efficiency.

[0072] In addition, in this embodiment, the electrode 4 is a metal electrode made of copper or aluminum. Copper and aluminum have good processing properties in semiconductor manufacturing processes. They can be deposited by various common process methods, such as sputtering in physical vapor deposition (PVD), electron beam evaporation, etc., to form the required electrode pattern. In photolithography and etching processes, the shape and size of the electrode can also be accurately defined to meet the design requirements of the detector. Moreover, these materials are widely used in industrial production, and the related process equipment and technologies are mature, which is convenient for large-scale production and reduces production costs.

[0073] In this application, by increasing the light absorption area, the responsivity of the optoelectronic signal processing unit is improved while miniaturizing the device, which is conducive to large-scale integration. Specifically, by using full etching or over-etching methods, etching grooves are formed on the waveguide, thereby increasing the light absorption area (i.e., the cross-sectional area) along the light propagation direction, and then increasing the light absorption rate, significantly improving the responsivity of the optoelectronic signal processing unit, enabling it to better meet the requirements of high-sensitivity detection of optical signals. At the same time, it is not necessary to increase the length of the germanium strip, which is more conducive to the miniaturization of the device.

[0074] In addition, the present invention improves the problem of poor process stability. In the prior art during shallow etching, on the one hand, the process itself cannot be completely consistent each time, and on the other hand, small differences in the thickness of the silicon layer may lead to different etching depths, affecting the consistency of performance indicators such as the dark current and responsivity of the optoelectronic signal processing unit. In the present invention, due to the design that the etching groove reaches the insulating layer and uses the insulating layer as a stable etching termination layer, the sensitivity of the process to changes in the silicon layer thickness is greatly reduced, the process stability is improved, and the performance of the produced optoelectronic signal processing unit is more stable and uniform.

[0075] The present invention avoids performance losses caused by light leakage. In the traditional shallow etching structure, when light propagates in the germanium / silicon processing region, it is easy to leak to the bottom of the shallow etching groove, reducing the overlapping time of the light field and germanium, and affecting the processing efficiency. In the present invention, the design that the etching groove reaches the insulating layer optimizes the light propagation path, reduces light leakage, ensures sufficient interaction between the light field and the germanium strip, and improves the comprehensive performance of the optoelectronic signal processing unit.

[0076] In addition, in the present invention, in terms of cross-section, since the light absorption area increases, when light is incident on the interface between the germanium strip and silicon, the amount of reflected light will also increase. Therefore, in this application, through the gradual structural change of the silicon strip and the germanium strip, the refractive index can be gradually transitioned from the silicon waveguide to germanium, reducing the reflection of light at the incident end. When light propagates at the interface of media with different refractive indices, the greater the refractive index difference, the more serious the reflection. This gradual transition structure enables light to enter the germanium strip more smoothly from the silicon waveguide, improves the light coupling efficiency, and further enhances the light absorption ability of the detector.

[0077] Embodiment 5 This embodiment provides an optical computing module with a self-detection function. Refer to Figure 15 , including: An input optical waveguide and an output electrical bus are cross-set. A splitter is provided at the input end of the input optical waveguide. The output end of the splitter is connected to the first input end of an optoelectronic signal processing unit through a transmission optical waveguide. A second input end for inputting an external electrical driving signal is also provided on the optoelectronic signal processing unit; The electrical output end of the optoelectronic signal processing unit is connected to the electrical bus; A detection unit is connected to the optical output end of the optoelectronic signal processing unit. The detection unit includes a detector and a data processing unit connected in sequence; Wherein, the optical input end of the detector is connected to the optical output end of the optoelectronic signal processor. A first input end for inputting the electrical driving signal is provided on the data processing unit. The second input end of the data processing unit is connected to the electrical output end of the detector.

[0078] Wherein, the data processing unit is used to analyze the electrical driving signal and the output electrical signal output by the detector to obtain an analysis result. The data processing unit can be a host computer.

[0079] Wherein, the first input end of the optoelectronic signal processing unit is an optical input end, and the second input end is an electrical input end; the optical signal entering the optoelectronic signal processing unit through the optical input end is modulated by the electrical driving signal input through the electrical input end, and at the same time, the absorbed part of the optical signal is converted into an electrical signal and output to the electrical bus. The other part of the unabsorbed optical signal (residual optical signal) is conducted to the optical output end and enters the detector in the detection unit.

[0080] In some embodiments, the detection unit further includes a comparator. The comparator is arranged between the detection unit and the detector. Wherein, a first input end for inputting the electrical driving signal is provided on the comparator. The second input end of the comparator is connected to the electrical output end of the detector. The output end of the comparator is connected to the data processing unit.

[0081] It can be seen that the present application provides an optical computing module with an independent detection unit, which directly generates an output electrical signal corresponding to the remaining optical signal in the optical computing module, and then synchronously outputs the output electrical signal and the electrical drive signal to an external correction unit for comparison to directly obtain the distortion situation between the two. Compared with the prior art of inserting error correction codes into the calculated data signal to improve the calculation accuracy, the present application improves from the hardware structure. On the one hand, the cost is low and the calculation amount is small. On the other hand, the remaining optical signal (instead of the output electrical signal) is used for error correction, and the entire error correction process is carried out independently, without affecting other components or signals other than the detection unit in the optical computing module.

[0082] In some embodiments, the optoelectronic signal processing unit may be the optoelectronic signal processing unit in Embodiment 3. Specifically, the optoelectronic signal processing unit includes: a silicon substrate and an insulating layer deposited on the silicon substrate; a waveguide layer is provided on the insulating layer; a germanium absorption layer is embedded in the waveguide layer; the first depth at which the germanium absorption layer is embedded in the waveguide layer is less than the thickness of the waveguide layer, or the second depth at which the germanium absorption layer is embedded in the waveguide layer is equal to the thickness of the waveguide layer.

[0083] A heating unit is provided on the top and / or bottom of the germanium absorption layer, and the heating unit extends along the length direction of the germanium absorption layer; or, heating units are respectively provided on both sides of the germanium absorption layer, and the heating units extend along the length direction of the germanium absorption layer; or, heating unit groups are respectively provided on both sides of the germanium absorption layer, and each heating unit group on each side includes a plurality of heating units uniformly spaced along the length direction of the germanium absorption layer, wherein the heating levels of two adjacent heating units on the same side are primary heating and secondary heating respectively.

[0084] In summary, the present application actually provides an error correction scheme for "amplifying" the remaining optical signal and using an independent detection unit for detection.

[0085] Among them, the present application improves the optoelectronic signal processing unit from the structure, uses pure germanium as the absorption layer of the optoelectronic signal processor, and provides a thermal compensation mechanism for dual regulation of the light absorption coefficient. Specifically, by heating the germanium absorption layer to adjust the light absorption coefficient of the germanium absorption layer from the germanium modulator itself, and then further adjusting the light absorption coefficient of the germanium absorption layer by adjusting the input voltage. The adjustment requirement for the input voltage in this process is not high, that is, the response speed of the computing module can be greatly improved while ensuring low energy consumption. In other words, both the remaining optical signal and the output electrical signal in the computing module can be "amplified" to a certain extent.

[0086] Furthermore, by performing error correction analysis on the "amplified" residual optical signal, the distortion condition after "amplification" can be obtained, and thus the situation where the distortion degree is too small to be detected can also be avoided.

[0087] In some other embodiments, the optoelectronic signal processing unit may also be the optoelectronic signal processing unit in Embodiment 4. Specifically, the optoelectronic signal processing unit includes: a silicon substrate and an insulating layer deposited on the silicon substrate; a waveguide layer (also referred to as an optical waveguide) is provided on the insulating layer; an etching groove reaching the insulating layer or an etching groove embedded in the insulating layer is formed on the waveguide layer; a germanium absorption layer (also referred to as a germanium strip) is filled in the etching groove as a light absorption region, and electrodes are provided on both sides of the etching groove; The bottom of the germanium absorption layer is flush with the upper surface of the insulating layer, or the bottom of the germanium absorption layer is embedded in the insulating layer; and the top of the germanium absorption layer protrudes above the upper surface of the waveguide layer.

[0088] In some embodiments, the germanium absorption layer includes a light guiding region and a light absorption region arranged in sequence along the light propagation direction. Among them, the cross-sectional area of the light guiding region gradually increases from the incident end along the direction close to the light absorption region; the cross-sectional area of the light absorption region is the same along the light propagation direction.

[0089] In some embodiments, when viewed from above, the width of the light guiding region gradually increases from the incident end along the light propagation direction; and when viewed from the side, the width of the light guiding region gradually decreases from the top to the bottom; or, the interface between the light guiding region and the waveguide layer is fan-shaped; and / or, when viewed from the side, the light absorption region is rectangular.

[0090] In summary, the present application also provides an error correction solution that "amplifies" the residual optical signal and uses an independent detection unit for detection.

[0091] The difference from the above solution is that the optoelectronic processing unit in this solution improves the responsivity of the optoelectronic signal processing unit by increasing the light absorption area, can realize the miniaturization of the device, and thus is conducive to realizing large-scale integration.

[0092] Specifically, in the present application, a full-etching or over-etching method is adopted to etch an etching groove on the waveguide, thereby increasing the light absorption area (i.e., the cross-sectional area) along the light propagation direction, and further improving the light absorption rate. At the same time, the shape of the light guiding region is doubly restricted in its length direction and height direction, that is, the cross-sectional area of the light guiding region gradually increases both in height and length from the incident end, which can greatly reduce the reflection of light at the incident end when propagating in the light absorption layer (germanium absorption layer) with a larger cross-sectional area, improving the light absorption rate of the optoelectronic signal processing unit. That is, the remaining optical signals and output electrical signals in the computing module can also be "amplified" to a certain extent.

[0093] Furthermore, by performing error correction analysis on the "amplified" remaining optical signals, the "amplified" distortion situation can be obtained, and further, the situation where the distortion degree is too small to be detected can be avoided.

[0094] Embodiment Six This embodiment provides an optical computing array with a self-detection function, including a plurality of optical computing modules described in Embodiment Five. Specifically, the optical computing array includes: A cross-switch matrix architecture formed by the intersection of multiple rows of mutually parallel input optical waveguides and multiple columns of mutually parallel electrical buses; the optical input signals are transmitted through the input optical waveguides; A plurality of optical splitters disposed at the front ends of the intersection nodes of each row of optical input waveguides and each column of electrical buses; A plurality of optoelectronic signal processing units, each optoelectronic signal processing unit corresponding to one of the cross nodes, and a first input end of the optoelectronic signal processing unit is connected to an output end of the optical splitter through a transmission optical waveguide. A second input end for inputting an external electrical drive signal is further provided on the optoelectronic signal processing unit; an electrical output end of the optoelectronic signal processing unit is connected to the electrical bus in the corresponding cross node; an optical output end of the optoelectronic signal processing unit is connected to a detection unit, and the detection unit includes a detector and a data processing unit connected in sequence; Wherein, a light input end of the detector is connected to an electrical output end of the optoelectronic signal processor, a first input end for inputting the electrical drive signal is provided on the data processing unit, and a second input end of the data processing unit is connected to an electrical output end of the detector.

[0095] In some embodiments, a third input end for inputting a reference voltage is further provided on the optoelectronic signal processing unit. The reference voltage is used to provide a reference voltage value for the electrical output end, so that the electrical output end can draw a current signal according to the reference voltage value and enable the current signal to flow in a specified direction to converge, thereby realizing operations such as summation.

[0096] Exemplarily, refer to Figure 11 and the reference voltage may be 2V (volts). When the voltage value at the electrical output terminal is less than 2V, the current signal flows in the first direction a; when the voltage value at the electrical output terminal is greater than 2V, the current signal flows in the second direction b; wherein, the first direction a and the second direction b are opposite, and the current signals flowing in the same direction converge at the end of the electrical bus.

[0097] In some embodiments, the optoelectronic signal processing unit includes: a silicon substrate and an insulating layer deposited on the silicon substrate; a waveguide layer is provided on the insulating layer; a germanium absorption layer is embedded in the waveguide layer; the first depth at which the germanium absorption layer is embedded in the waveguide layer is less than the thickness of the waveguide layer, or the second depth at which the germanium absorption layer is embedded in the waveguide layer is equal to the thickness of the waveguide layer; a heating unit is provided on the top and / or bottom of the germanium absorption layer, and the heating unit extends along the length direction of the germanium absorption layer; or, heating units are respectively provided on both sides of the germanium absorption layer, and the heating units extend along the length direction of the germanium absorption layer; or, heating unit groups are respectively provided on both sides of the germanium absorption layer, and each heating unit group on each side includes a plurality of heating units uniformly spaced along the length direction of the germanium absorption layer, wherein the heating levels of two adjacent heating units on the same side are respectively primary heating and secondary heating.

[0098] In some embodiments, the germanium absorption layer includes a light guiding region and a light absorption region arranged in sequence along the light propagation direction, wherein, the cross-sectional area of the light guiding region gradually increases from the incident end along the direction close to the light absorption region; the cross-sectional area of the light absorption region is the same along the light propagation direction.

[0099] Embodiment Seven This embodiment provides a detection method for an optical computing module. Refer to Figures 15 - 17 , based on the optical computing module in Embodiment Five, or based on the optical computing array in Embodiment Six, including the following steps: S1. Obtain the electrical drive signal input to the optoelectronic signal processing unit and the output electrical signal output by the detector; S2. Synchronously input the electrical drive signal and the output electrical signal into the data processing unit to obtain the electrical signal difference in the current state; S3. Repeat steps S1 and S2 to obtain multiple electrical signal differences; S4. Calculate multiple signal means according to the multiple electrical signal differences; S5. Determine whether the signal mean value conforms to a preset signal range. If so, mark the optical computing module as normal; otherwise, mark the optical computing module as abnormal.

[0100] Among them, the preset signal range is pre-input by the user or obtained in advance based on a large number of experiments and stored in the data processing unit. When the signal mean value does not conform to the preset signal range, it indicates that the signal is distorted. At this time, it is marked as abnormal, and manual judgment is introduced to determine whether the abnormality is due to signal distortion of the signal source, problems with the device, or other circumstances.

[0101] In the above detection method, multiple electrical signal differences are obtained through multiple detections during detection, and the multiple electrical signal differences are analyzed, which can avoid the "false judgment" problem caused by signal delay.

[0102] In some embodiments, the detection method, based on the optical computing module in Embodiment 5 or the optical computing array in Embodiment 6, includes the following steps: S1. Obtain the electrical modulation signal S1 input to the optoelectronic signal processing unit and the output electrical signal I1 output by the detector.

[0103] S2. Synchronously input the electrical modulation signal S1 and the output electrical signal I1 into the data processing unit, and calculate the electrical signal difference between the electrical modulation signal S1 and the output electrical signal I1; among them, this difference has a constant relationship with the current signal O1 output by the optoelectronic signal processing unit, and deviation from this constant can be determined as the loss of the signal packet in the optoelectronic signal processing unit.

[0104] S3. Input the electrical signal difference into the data processing unit, and determine whether the electrical signal difference conforms to a preset difference range. If so, mark the optical computing module as normal; otherwise, mark the optical computing module as abnormal.

[0105] See Figure 16 , where a signal comparison diagram of normal signals and abnormal signals is provided in the figure. Among them, the abscissa t is time, and the ordinate P is power.

[0106] It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0107] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative rather than restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the spirit of the present invention and the scope protected by the claims. All of these fall within the protection scope of the present invention.

Claims

1. An optical computing module, characterized in that: include: An input optical waveguide and an electrical bus are arranged crosswise, wherein an optical splitter is arranged at the input end of the input optical waveguide, an output end of the optical splitter is connected to a first input end of an optoelectronic signal processing unit through a transmission optical waveguide, and an electrical output end of the optoelectronic signal processing unit is connected to the electrical bus; The photoelectric characteristics of the light absorption layer in the photoelectric signal processing unit can be changed by writing a control signal into the second input terminal thereof, and the photoelectric signal processing unit can convert the light signal inputted from the first input terminal into a corresponding current signal, and send the current signal to the electrical bus through the electrical output terminal; the photoelectric characteristics include the light absorption coefficient of the light absorption layer; Therein, the product of the multiplier value and the multiplicand value is encoded in the current signal.

2. The optical computing module according to claim 1, characterized in that: The photoelectric signal processing unit uses pure germanium as a light absorption layer; The optoelectronic signal processing unit includes: a silicon substrate and an insulating layer deposited on the silicon substrate; a waveguide layer is arranged on the insulating layer; a germanium absorption layer is embedded in the waveguide layer; a first depth of the germanium absorption layer embedded in the waveguide layer is less than the thickness of the waveguide layer, or a second depth of the germanium absorption layer embedded in the waveguide layer is equal to the thickness of the waveguide layer.

3. The optical computing module according to claim 2, characterized in that: A heating unit is disposed on the top and / or bottom of the germanium absorption layer, and the heating unit extends along the length direction of the germanium absorption layer; or, Heating units are respectively arranged on both sides of the germanium absorption layer, and the heating units extend along the length direction of the germanium absorption layer; or, Heating unit groups are respectively arranged on both sides of the germanium absorption layer, and the heating unit group on each side includes a plurality of heating units evenly spaced along the length direction of the germanium absorption layer, wherein the heating levels of two adjacent heating units on the same side are primary heating and secondary heating, respectively.

4. The optical computing module according to claim 2, characterized in that: The germanium absorption layer includes a light guiding region and a light absorption region sequentially arranged along the light propagation direction. The cross-sectional area of ​​the light guiding region gradually increases from the incident end in a direction close to the light absorbing region; and the cross-sectional area of ​​the light absorbing region is the same along the light propagation direction.

5. The optical computing module according to claim 4, characterized in that: When the light guiding area is viewed from above, its width gradually increases from the incident end along the direction of light propagation; and when the light guiding area is viewed from the side, its width gradually decreases from the top to the bottom; or, the interface between the light guiding area and the waveguide layer is fan-shaped; and / or, the light absorption area is rectangular when viewed from the side.

6. An optical computing array, characterized in that: include: A crossbar switch matrix architecture, wherein the crossbar switch matrix architecture is formed by crossing multiple rows of mutually parallel input optical waveguides and multiple columns of mutually parallel electrical buses; An optical input signal is transmitted through the input optical waveguide; A plurality of optical splitters, the optical splitters being arranged at the front end of the intersection node between each row of optical input waveguides and each column of electrical bus; A plurality of optoelectronic signal processing units, each of which corresponds to one of the cross nodes, and a first input end of the optoelectronic signal processing unit is connected to an output end of the optical splitter through a transmission optical waveguide, and an electrical output end of the optoelectronic signal processing unit is connected to the electrical bus in the corresponding cross node; The photoelectric characteristics of the light absorption layer in the photoelectric signal processing unit can be changed by writing a control signal at its second input terminal, and the photoelectric signal processing unit can convert the light signal input at the first input terminal into a corresponding current signal and send it to the electrical bus through the electrical output terminal; the photoelectric characteristics include a light absorption coefficient; Therein, the product of the multiplier value and the multiplicand value is encoded in the current signal.

7. The optical computing array according to claim 6, characterized in that: The photoelectric signal processing unit uses pure germanium as a light absorption layer; The optoelectronic signal processing unit includes: a silicon substrate and an insulating layer deposited on the silicon substrate; a waveguide layer is arranged on the insulating layer; a germanium absorption layer is embedded in the waveguide layer; a first depth of the germanium absorption layer embedded in the waveguide layer is less than the thickness of the waveguide layer, or a second depth of the germanium absorption layer embedded in the waveguide layer is equal to the thickness of the waveguide layer.

8. The optical computing array according to claim 7, characterized in that: A heating unit is disposed on the top and / or bottom of the germanium absorption layer, and the heating unit extends along the length direction of the germanium absorption layer; or, Heating units are respectively arranged on both sides of the germanium absorption layer, and the heating units extend along the length direction of the germanium absorption layer; or, Heating unit groups are respectively arranged on both sides of the germanium absorption layer, and the heating unit group on each side includes a plurality of heating units evenly spaced along the length direction of the germanium absorption layer, wherein the heating levels of two adjacent heating units on the same side are primary heating and secondary heating, respectively.

9. The optical computing array according to claim 7 or 8, characterized in that: The germanium absorption layer includes a light guiding region and a light absorption region sequentially arranged along the light propagation direction. The cross-sectional area of ​​the light guiding region gradually increases from the incident end in a direction close to the light absorbing region; and the cross-sectional area of ​​the light absorbing region is the same along the light propagation direction.

10. The optical computing array according to claim 7 or 8, characterized in that: The second input end of each of the optoelectronic signal processing units is connected to a Pad, and a Pad is arranged on each column of the electrical bus; and / or an EOM is arranged on each row of the input optical waveguide.

Citation Information

Patent Citations

  • Photoelectric computing unit

    CN109960310A

  • A photonic computing array and method for on-chip all-optical reasoning

    CN118394171B

  • Data processing method and system

    CN116795171A

  • Method for operating in optical domain by means of photon computing unit

    CN117896011A

  • Novel large-scale photon calculation array and method for on-chip all-optical reasoning

    CN118394171A