A large-scale optical computing matrix with adjustable optical delay

Abstract

This invention discloses an ultra-large-scale optical computing matrix with adjustable optical delay, comprising a light-transmitting substrate on which the optical computing matrix is ​​integrated. Each optical computing unit in the optical computing matrix is ​​connected to an optical delay module. Each optical delay module includes a coarse adjustment module with a first adjustment accuracy level; it also includes a photodetector connected to the optical computing unit and connected to a main control module. The coarse adjustment module includes an optical waveguide delay structure. The optical delay module further includes a mid-adjustment module acting on the optical waveguide delay structure and a fine adjustment module acting on the optical waveguide of the optical computing unit. The mid-adjustment module has a second adjustment accuracy level, and the fine adjustment module has a third accuracy adjustment level; the first adjustment accuracy level > the second adjustment accuracy level > the third adjustment level. The optical delay generated by the optical delay module is adjustable, facilitating the uniform adjustment of the optical delay of all optical computing units in the computing matrix.

Description

Technical Field

[0001] This invention belongs to the field of optical chip technology, and in particular relates to an ultra-large-scale optical computing matrix with adjustable optical delay. Background Technology

[0002] With the exponential growth in computing power demands in fields such as artificial intelligence, big data processing, and high-speed communication, traditional electrical computing architectures are increasingly unable to meet the needs of ultra-large-scale parallel data processing due to limitations in electron mobility, excessively high power density, and signal crosstalk. Against this backdrop, optical computing technology, with its core advantages of high parallelism, low power consumption, and wide bandwidth, has become one of the key directions for breaking through computing power bottlenecks. Ultra-large-scale optical computing matrices, as the core computing carrier of optical computing systems, directly determine the accuracy and efficiency of optical computing.

[0003] The core operational logic of an optical computing matrix relies on the precise transmission and coordination of optical signals between various optical computing units. Consistent optical delay is a prerequisite for this logic to function—that is, the optical signals output from different optical computing units must arrive at the target unit synchronously within a preset time window. Otherwise, phase deviations and operational logic errors (such as interferometer misjudgments and inaccurate multiplication-accumulation operations) will occur, ultimately affecting the overall computational accuracy. Therefore, how to achieve precise and consistent calibration of the optical delay across all units in an ultra-large-scale optical computing matrix has become a core technological bottleneck restricting the transition of optical computing technology from the laboratory to industrial applications.

[0004] The prior art Chinese patent application CN201210529198.3 discloses a reflective adjustable optical delay line, which is composed of an optical delay structure and an optical reflection structure connected in series. The input optical signal is input from the input end of the optical delay structure, is optically delayed, is input to the optical reflection structure, is reflected, and then returns to the optical delay structure. After being optically delayed again by the optical delay structure, the delayed optical signal is output from the input end of the optical delay structure.

[0005] The aforementioned prior art proposes optical delay structures, which extend the optical path by lengthening the optical waveguide to change the optical delay. However, in these prior art, the optical delay structure is pre-set and cannot be adjusted. For ultra-large-scale optical computing matrices, which contain a large number of optical computing units, each unit may require an optical delay structure to change its optical delay. Since the optical delay structure is not adjustable, the optical delay structure of each photon unit needs to be extremely precise during fabrication to ensure the consistency of the delay of the entire optical computing matrix, which is extremely difficult. Summary of the Invention

[0006] The purpose of this invention is to provide an ultra-large-scale optical computing matrix with adjustable optical delay, which partially solves or alleviates the above-mentioned shortcomings in the prior art. The optical delay generated by its optical delay module can be adjusted as needed.

[0007] To solve the aforementioned technical problems, the present invention specifically adopts the following technical solution:

[0008] A first aspect of the present invention is to provide an ultra-large-scale optical computing matrix with adjustable optical delay, comprising a light-transmitting substrate, wherein an optical computing matrix is ​​integrated on the light-transmitting substrate; each optical computing unit in the optical computing matrix is ​​connected to an optical delay module; the optical delay module includes a coarse adjustment module having a first adjustment accuracy level; and further includes a photodetector connected to the optical computing unit, wherein the photodetector is connected to a main control module.

[0009] The coarse adjustment module includes an optical waveguide delay structure; the length L and width W of the optical waveguide delay structure satisfy the following condition:

[0010] Ti=T0;

[0011] Ti = tf + t0;

[0012] tf=(k1 *L)*(k2* W) / C;

[0013] Where k1 and k2 are adjustment coefficients, k2*W≤W0, W0 is the preset maximum width threshold; C is the propagation speed of light in the optical waveguide; T0 is the optical delay of the reference optical computing unit, Ti is the optical delay of the current optical computing unit, t0 is the initial optical delay of the current optical computing unit, and tf is the optical delay provided by the optical waveguide delay structure of the current optical computing unit.

[0014] The optical delay module further includes: a mid-tuning module acting on the optical waveguide delay structure, and a fine-tuning module acting on the optical waveguide corresponding to the optical computing unit; the mid-tuning module has a second adjustment accuracy level, and the fine-tuning module has a third accuracy adjustment level; the first adjustment accuracy level > the second adjustment accuracy level > the third accuracy adjustment level.

[0015] Furthermore, the fine-tuning module is configured to apply a preset third adjustment condition to the optical waveguide corresponding to the optical computing unit; the main control module is configured to perform delay calculation based on the optical signal detected by the photodetector, and compare the calculated optical delay Ti with T0 to obtain ΔT, ΔT=|Ti-T0|; and control the fine-tuning module to apply the third adjustment condition to the optical waveguide corresponding to the optical computing unit to change the refractive index of the optical waveguide corresponding to the optical computing unit, thereby reducing ΔT.

[0016] Furthermore, the fine-tuning module includes a fine-tuning element and a third adjustment unit. The fine-tuning element is a modulation layer laid on the optical waveguide corresponding to the optical computing unit. The third adjustment unit is configured to change the refractive index of the optical waveguide corresponding to the optical computing unit by switching the state of the fine-tuning element.

[0017] Furthermore, the intermediate adjustment module is configured to apply a preset second adjustment condition to the optical waveguide delay structure; the main control module is configured to perform delay calculation based on the optical signal detected by the photodetector, and compare the calculated optical delay Ti with T0 to obtain ΔT, ΔT=|Ti-T0|; and control the intermediate adjustment module to apply the second adjustment condition to the optical waveguide delay structure to change the refractive index of the optical waveguide delay structure, thereby reducing ΔT.

[0018] Furthermore, the intermediate modulation module includes an intermediate modulation element and a second adjustment unit. The intermediate modulation element is one of a thermal modulation circuit, an electrical modulation circuit, and an optical modulation circuit. The second adjustment unit is configured to apply heat to the optical waveguide delay structure using the intermediate modulation element to change its refractive index, or to apply an electric field to the optical waveguide delay structure using the intermediate modulation element to change its refractive index, or to apply light energy to the optical waveguide delay structure using the intermediate modulation element to change its refractive index.

[0019] Furthermore, the steps for adjusting the optical delay of the optical computing unit using the optical delay module are as follows:

[0020] Set the coarse adjustment termination threshold S1, the medium adjustment termination threshold S2, and the fine adjustment termination threshold S3;

[0021] The length and width of the optical waveguide delay structure are set according to the optical delay between the current optical computing unit and the reference optical computing unit, such that ΔT≤S1;

[0022] Based on the coarsely adjusted optical delay, a preset second adjustment condition is applied to the optical waveguide delay structure through the intermediate adjustment module, such that ΔT≤S2;

[0023] Based on the adjusted optical delay, a preset third adjustment condition is applied to the optical waveguide of the optical computing unit through the fine-tuning module, so that ΔT≤S3.

[0024] Furthermore, the optical waveguide delay structure is a spiral or serpentine optical waveguide.

[0025] Furthermore, when determining the length and width of the optical waveguide delay structure of a certain optical computing unit, the adjustment coefficient k2 is first modified to make the optical delay Ti of the optical computing unit approach the optical delay T0 of the reference optical computing unit under the premise that k2*W≤W0; then the adjustment coefficient k1 is modified to make the optical delay Ti of the optical computing unit further approach the optical delay T0 of the reference optical computing unit.

[0026] Furthermore, the optical computing units in the optical computing matrix are partitioned according to a preset partitioning rule to obtain several sub-matrices, and a sub-matrix optical waveguide delay structure is connected in series on each sub-matrix;

[0027] Each submatrix contains an internal reference light calculation unit, and all light calculation units in each submatrix use their light delay modules to adjust the light delay to be consistent with that of the internal reference light calculation unit.

[0028] A reference submatrix is ​​set in all submatrices, and each submatrix adjusts its optical delay to match that of the reference submatrix using its cascaded submatrix waveguide delay structure.

[0029] Furthermore, the partitioning rule is as follows: determine the size of the submatrix and divide the submatrix according to the physical location of the optical computing unit; or, cluster all optical computing units according to the optical delay between them and the reference optical computing unit, and divide the submatrix according to the clustering results.

[0030] Beneficial Effects: This invention addresses the core contradiction of balancing delay adjustment range and precision in ultra-large-scale scenarios by constructing a three-level adjustment architecture of coarse, medium, and fine adjustments, progressively increasing the precision levels. Specifically, the coarse adjustment module, an optical waveguide delay structure, utilizes the formula tf=(k1*L)*(k2*W) / C to achieve a wide range of delay compensation from microseconds to nanoseconds, quickly eliminating initial large deviations and laying the foundation for subsequent adjustments. The medium adjustment module, employing thermal / electrical / optical refractive index modulation, bridges the precision gap between coarse and fine adjustments, reducing deviations from hundreds of picoseconds to tens of picoseconds, adapting to medium-precision optimization needs. The fine adjustment module, involving modulation layer state switching, achieves ultimate calibration at the picosecond or even femtosecond level through precise techniques such as local crystallization of phase change materials and electro-optic micro-area modulation, meeting the stringent requirements for delay consistency in core operations such as optical interference and phase-sensitive multiplication.

[0031] The synergy of these three elements can cover a microsecond-level adjustment range and a femtosecond-level adjustment precision, which is the core guarantee for achieving high-fidelity computation of ultra-large-scale matrices.

[0032] To address the pain points of low efficiency and easy error accumulation in the global unified adjustment of ultra-large-scale matrices, this invention adopts a two-level architecture of local calibration within submatrices and global calibration between submatrices. Intra-submatric calibration uses an internal reference unit as an anchor point. Because the process / environmental errors of units within the same submatric are similar, calibration parameters can be reused in batches, significantly improving calibration efficiency compared to adjusting a single unit. Inter-submatric calibration uses a reference submatric as a global anchor point, achieving overall compensation through a series of optical waveguide delay structures within the submatrices. The adjustment target is reduced from the optical computing unit level to the submatric level, significantly reducing global calibration time. Simultaneously, it avoids error propagation at the unit-level adjustment level, ensuring that the final delay deviation of the entire matrix can be controlled within an ideal range. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0034] Figure 1 This is a schematic diagram of the structure of a single optical computing unit in this invention;

[0035] Figure 2 This is a schematic diagram of the partition adjustment in this invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0037] In this document, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" may be used interchangeably.

[0038] In this document, the terms "upper," "lower," "inner," "outer," "front," "rear," "one end," and "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the present invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0039] In this document, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," "connected," etc., should be interpreted broadly. For example, "connection" 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; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0040] In this document, "and / or" includes any and all combinations of one or more of the listed related items.

[0041] In this article, "multiple" means two or more, that is, it includes two, three, four, five, etc.

[0042] Optical AI chip: A photonic computing chip, also known as a photonic chip or optical computing chip, is a chip specifically designed for processing optical data and performing artificial intelligence (AI) inference. Based on optical principles, it takes optical signals as input and performs computational operations through a series of optical elements and devices. These optical elements can be components such as lasers, optical fibers, and optical modulators, used to process and transmit optical signals. For example, a photonic computing chip converts the optical signal into a weak photocurrent signal using a light-emitting diode (LED). This photocurrent signal is then converted into a voltage signal U by a transimpedance amplifier (TIA), and finally converted into a digital signal by an analog-to-digital converter (ADC). The digital signal is then output to an FPGA for processing. Simultaneously, the FPGA outputs the processed digital signal to a digital-to-analog converter (DAC) to convert the FPGA's digital signal back into an analog signal and input it back to the photonic computing chip.

[0043] Computational Unit: A single computational unit used to perform computation. For example, a single computational unit that utilizes a Mach-Zehnder interferometer (MZI) or a microring structure (MMR) to perform photonic computation. Other examples include photonic computational units utilizing carrier light absorption effects and photonic computational units based on the absorption effects of phase change materials. Specifically, computational units can be divided into photonic computational units and hybrid computational units.

[0044] Computational Array: An N*N computational array is constructed based on the above computational units. Each intersection of the input row waveguide and the output column waveguide in this array contains a computational unit. If the computational units are photonic computational units, the computational array can also be called a photonic computational array; if the computational units are optoelectronic hybrid computational units, the computational array can also be called an optoelectronic hybrid computational array.

[0045] Transparent substrate: In this document, a transparent substrate refers to a base substrate with a flat, smooth surface and specific geometric dimensions (such as a disc, square, or panel), and exhibits extremely high transmittance, for example, >99.999%, within a target wavelength range (such as the visible light region 380nm–2µm). It is commonly used as a "foundation" or "canvas" for various functional devices (such as circuits, sensors, and waveguides), providing a support platform for subsequent micro / nano fabrication and functional layer deposition. Examples include interposers or redistribution substrates for system-in-package (SIIP) applications, and the fabrication of electrical interconnect vias and optical waveguide structures (such as gratings) within the glass. Preferably, the transparent substrate can be a transparent substrate, etc.

[0046] Optical waveguide delay structure: In this article, an optical waveguide delay structure refers to a structure placed in the optical propagation path to extend the optical propagation path. Examples include optical waveguides of a certain length and width, or fiber delay lines.

[0047] Example 1: To address the issue of optical delay consistency in ultra-large-scale optical computing matrices, and to ensure the synchronization of output optical signals from all optical computing units within the matrix, thereby avoiding computational errors caused by delay deviations, such as... Figure 1 As shown, this embodiment provides an ultra-large-scale optical computing matrix with adjustable optical delay, which includes a light-transmitting substrate (such as a glass substrate), on which the optical computing matrix is ​​integrated; each optical computing unit in the optical computing matrix is ​​connected to an optical delay module; the optical delay module includes a coarse adjustment module with a first adjustment accuracy level; and also includes a photodetector connected to the optical computing unit, the photodetector being connected to a main control module;

[0048] The coarse adjustment module includes an optical waveguide delay structure; the length L and width W of the optical waveguide delay structure satisfy the following condition:

[0049] Ti=T0; Ti=tf+t0; tf=(k1*L)*(k2*W) / C; where k1 and k2 are adjustment coefficients, k2*W≤W0, W0 is the preset maximum width threshold; C is the propagation speed of light in the optical waveguide; T0 is the optical delay of the reference optical computing unit, Ti is the optical delay of the current optical computing unit, t0 is the initial optical delay of the current optical computing unit, and tf is the optical delay provided by the optical waveguide delay structure of the current optical computing unit.

[0050] The coarse adjustment module is the first stage of delay adjustment, responsible for quickly bringing the initial delay t0 of the optical computing unit from a large deviation to the range that the intermediate adjustment module can handle. Its core technology is to achieve quantifiable delay adjustment through the physical parameters of the optical waveguide delay structure, namely the length L and the width W.

[0051] t0 is the initial delay of the optical computing unit, caused by manufacturing process errors, material uniformity differences, etc., and is the reference deviation to be adjusted. The optical delay tf is the compensation delay provided by the optical waveguide delay structure. It is actively adjusted to offset the deviation of t0, gradually bringing the total delay Ti closer to, or even equal to, the target T0. L and W are the length and width of the optical waveguide delay structure, which are physical adjustment variables for coarse tuning. The length L directly determines the physical length of the light propagation path; the larger L is, the longer the light propagation time and the larger tf is. However, the space occupied by the substrate needs to be balanced. Therefore, in this embodiment, the optical waveguide delay structure is usually arranged in a spiral or serpentine shape to extend L within a limited space. The width W affects the propagation mode and speed of light in the waveguide. When the waveguide width increases, the equivalent refractive index decreases, and the light propagation speed increases. However, this is limited by k2*W≤W0, where W0 is the critical width, to avoid excessive width triggering multimode transmission (multimode transmission leads to mode dispersion, destroying delay stability).

[0052] k1 and k2 are adjustment coefficients that are related to the characteristics of the optical waveguide material and the shape of its cross-section. For example, k1 is related to the sensitivity of the material's refractive index to the length, and k2 is related to the correction coefficient of the waveguide width to the propagation speed. By dynamically adjusting k1 and k2, the process deviations of different batches of waveguides can be adapted to improve the adjustment flexibility.

[0053] C is the speed of light in the optical waveguide, which is determined by the refractive index n of the waveguide material. C = c0 / n, where c0 is the speed of light in a vacuum and is the reference physical quantity for time delay calculation.

[0054] Specifically, when determining the length and width of the optical waveguide delay structure of a certain optical computing unit, the adjustment coefficient k2 is first determined. Under the premise that k2*W≤W0, the optical delay Ti of the optical computing unit is made to approach the optical delay T0 of the reference optical computing unit.

[0055] Since the optical computing matrix is a highly integrated device, the utilization of space needs to be very precise. For the optical waveguide delay structure, increasing the width can solve the space problem more effectively than increasing the length. Therefore, in this embodiment, the width W of the optical waveguide delay structure is corrected first. The width W of the optical waveguide affects the light propagation mode and the equivalent refractive index. Therefore, the width W cannot be increased without limit and needs to follow the constraint of k2 * W ≤ W0. Therefore, by adjusting k2 first, the delay error related to the width can be corrected to the maximum extent on the premise of ensuring the safe operation state of the waveguide (single-mode transmission). For example, if the tf of an optical computing unit is small due to the actual small width, and Ti < T0, k2 can be appropriately increased, but it is necessary to ensure that k2 * W ≤ W0. By increasing the contribution weight of the width to the delay, tf is increased, and Ti approaches T0 initially.

[0056] The length L of the optical waveguide delay structure directly determines the physical length of the light propagation path. The longer L is, the longer the propagation time is theoretically. After correcting the initial width W, the gap between Ti and T0 needs to be compensated by correcting L, so that Ti approaches T0 more and more, that is, by modifying the adjustment coefficient k1, the optical delay Ti of the optical computing unit further approaches the optical delay T0 of the reference benchmark optical computing unit.

[0057] After adjusting k2 to make Ti initially close to T0, the purpose of modifying k1 is to compensate for the remaining length-related errors. The adjustment of k1 does not require a hard constraint like k2 * W ≤ W0 and can be corrected more flexibly. For example, if Ti is still slightly less than T0 after adjusting k2, k1 can be slightly increased to increase the contribution weight of the length to the delay, so that tf is further increased, and finally Ti approaches T0.

[0058] The adjustment of k2 mainly solves the step error, while the adjustment of k1 can be used for continuous errors. The combination of the two forms a complementary mechanism to cover different types of process errors and ensure that the deviation between Ti and T0 is finally controlled within the target accuracy.

[0059] In some scenarios, relying solely on the coarse adjustment module is likely to not achieve the final target accuracy. Therefore, in this embodiment, the optical delay module further includes: a medium adjustment module acting on the optical waveguide delay structure, and a fine adjustment module acting on the optical waveguide delay structure corresponding to the optical computing unit; the medium adjustment module has a second adjustment accuracy level, and the fine adjustment module has a third accuracy adjustment level; the first adjustment accuracy level > the second adjustment accuracy level > the third adjustment accuracy level.

[0060] Among them, the medium adjustment module acts on the optical waveguide delay structure. The core principle is to change its refractive index through external excitation, and then fine-tune the light propagation speed. When the refractive index n increases, the propagation speed C decreases and the delay increases; otherwise, the delay decreases.

[0061] After coarse adjustment, the difference between the optical delay of the current optical computing unit and the optical delay of the reference optical computing unit has been reduced to the upper limit of coarse adjustment accuracy, while the intermediate adjustment module can further reduce the deviation. Compared with coarse adjustment, intermediate adjustment has higher adjustment accuracy, the second adjustment accuracy level, which is usually at the picosecond level; in addition, the intermediate adjustment module acts on the optical waveguide delay structure without changing the physical structure of the optical waveguide.

[0062] Specifically, the intermediate adjustment module is configured to apply a preset second adjustment condition to the optical waveguide delay structure; the main control module is configured to perform delay calculation based on the optical signal detected by the photodetector, and compare the calculated optical delay Ti with T0 to obtain ΔT, ΔT=|Ti-T0|; and control the intermediate adjustment module to apply the second adjustment condition to the optical waveguide delay structure to change the refractive index of the optical waveguide delay structure, thereby reducing ΔT.

[0063] In this embodiment, the adjustment of the mid-tuning module essentially involves changing the refractive index *n* of the optical waveguide delay structure by applying a second adjustment condition, thereby controlling the propagation speed of light and ultimately achieving delay correction. The propagation speed of light in a medium follows the formula v=nc, where *c* is the speed of light in a vacuum and *n* is the refractive index of the medium. When the effective length of the optical waveguide is fixed, the delay *t* is directly proportional to the refractive index *n*. By changing *n*, the delay *t* can be linearly controlled, thereby reducing *ΔT*.

[0064] More specifically, the mid-tuning module in this embodiment includes a mid-tuning element and a second adjustment unit. The mid-tuning element is one of a thermal tuning circuit, an electrical tuning circuit, and an optical tuning circuit. The second adjustment unit is configured to apply heat to the optical waveguide delay structure using the mid-tuning element to change its refractive index, or to apply an electric field to the optical waveguide delay structure using the mid-tuning element to change its refractive index, or to apply light energy to the optical waveguide delay structure using the mid-tuning element to change its refractive index. Preferably, the mid-tuning element includes one or more of a thermal tuning circuit, an electrical tuning circuit, and an optical tuning circuit, and is the execution carrier for realizing refractive index control, changing the refractive index of the optical waveguide delay structure through different physical effects.

[0065] Preferably, the thermal adjustment circuit includes a microheater, such as a metal thin-film resistor, a doped semiconductor heating unit, and a temperature control circuit for precisely controlling the heating power, integrated in the vicinity of the optical waveguide delay structure. When the thermal adjustment circuit applies current through the second adjustment unit, the microheater generates Joule heating, causing the local temperature of the optical waveguide delay structure to rise, leading to intensified thermal motion of material molecules, changes in electronic polarization, and ultimately an increase in the refractive index n.

[0066] Preferably, the electrically adjustable circuit includes an electrode system such as parallel plate electrodes, coplanar waveguide electrodes, and a voltage driving circuit. The electrodes are typically symmetrically distributed on both sides of the optical waveguide delay structure to form a transverse electric field or on the upper and lower layers to form a longitudinal electric field. The electrode material is a highly conductive metal, and it is isolated from the waveguide by an insulating layer to avoid short circuits. When the second adjustment unit applies a voltage through the electrically adjustable circuit, an electric field E is formed between the electrodes, causing the molecular polarization directions of the optical waveguide material to align in an orderly manner, resulting in a change in the dielectric constant ε, and ultimately achieving precise control of the refractive index.

[0067] Preferably, the light modulation circuit includes a light source such as a laser diode (LD), an optical coupler, and an optical power controller. When the second modulation unit controls the light stripe circuit to apply light energy of a specific intensity to the optical waveguide delay structure, the light wave interacts with the material molecules, causing a change in the electronic distribution state of the material, which in turn causes a change in the refractive index n.

[0068] In some embodiments, the second adjustment unit is the control center of the intermediate adjustment module. According to the instructions of the main control module, it dynamically adjusts the adjustment conditions applied by the intermediate adjustment element, such as the intensity and duration of heat, electric field, light energy, etc., to ensure that the amount of refractive index change is accurately matched with the target delay correction amount.

[0069] In this embodiment, the fine-tuning module operates on the optical waveguide of the optical computing unit (specifically, the optical waveguide includes a waveguide layer and a modulation layer disposed on the waveguide layer). The core principle is to change the local refractive index of the optical waveguide by switching the state of the modulation layer. The modulation layer is usually a phase change material such as Ge2Sb2Te or an electro-optic material, which has significant differences in refractive index in different states such as amorphous / crystalline, enabling ultra-fine delay adjustment at the picosecond or even femtosecond level.

[0070] After intermediate tuning, fine-tuning is performed to ultimately control the deviation within the target threshold, ensuring the global latency consistency of the ultra-large-scale matrix. Compared to intermediate tuning, fine-tuning offers the highest adjustment precision, allowing the optical latency Ti of the current optical computing unit to approach T0 infinitely.

[0071] Specifically, the fine-tuning module is configured to apply a preset third adjustment condition to the optical waveguide (modulation layer) corresponding to the optical computing unit; the main control module is configured to perform delay calculation based on the optical signal detected by the photodetector, and compare the calculated optical delay Ti with T0 to obtain ΔT, ΔT=|Ti-T0|; and control the fine-tuning module to apply the third adjustment condition to the optical waveguide (modulation layer) corresponding to the optical computing unit to change the refractive index of the optical waveguide of the optical computing unit, thereby reducing ΔT.

[0072] In some embodiments, the fine-tuning module includes a fine-tuning element and a third adjustment unit. The fine-tuning element is a modulation layer deposited on the optical waveguide of the optical computing unit. The third adjustment unit is configured to change the refractive index of the optical waveguide of the optical computing unit by switching the state of the fine-tuning element. The modulation layer is the execution carrier of the fine-tuning module. The modulation layer is typically deposited directly on the surface of the optical waveguide of the optical computing unit using a thin-film deposition process. It is extremely thin, and the distance between the modulation layer and the waveguide is controlled at the nanometer level. The state change of the modulation layer can directly affect the equivalent refractive index of the optical waveguide. The modulation layer needs to meet the characteristics of switchable state, significant refractive index change, and good compatibility with the optical waveguide. Common materials include phase change materials such as Ge2Sb2Te.

[0073] In this embodiment, the state switching of the modulation layer refers to the tunable physical / chemical state change, such as from crystalline to amorphous, that occurs under external stimulation (e.g., electrical, thermal, or optical energy). This state change is directly related to the change in refractive index. The state switching of the modulation layer can be controlled at the nanometer level, and the corresponding delay correction accuracy can reach the femtosecond level, meeting the extreme requirements of delay consistency for ultra-large-scale matrices. In addition, the modulation layer is co-integrated with the optical waveguide using a thin-film process, without occupying additional substrate space, and has high optical field coupling efficiency with the optical waveguide, making it suitable for dense arrangement of ultra-large-scale cells.

[0074] In this embodiment, the third adjustment unit is the control center for the state switching of the modulation layer. According to the instructions of the main control module, it precisely adjusts the parameters of the excitation signal so that the degree of state switching of the modulation layer is strictly matched with the change in the target refractive index.

[0075] In some embodiments, combining the three modules of coarse adjustment, intermediate adjustment, and fine adjustment, the steps for adjusting the optical delay of the optical computing unit using the optical delay module in this embodiment are as follows:

[0076] 1. Set the coarse adjustment termination threshold S1, the medium adjustment termination threshold S2, and the fine adjustment termination threshold S3.

[0077] Among them, the coarse adjustment termination threshold S1 corresponds to the highest accuracy (first adjustment accuracy level) of the coarse adjustment module and is the minimum target that coarse adjustment needs to achieve. Its value must cover the lower limit of the adjustment range of the coarse adjustment module, while reserving sufficient adjustment space for the intermediate adjustment module. For example, if the adjustment range of the coarse adjustment module is 100ps~1000ps (accuracy ±50ps), then S1 can be set to 100ps, that is, after coarse adjustment, ΔT must be ≤100ps to ensure that the intermediate adjustment module can handle the remaining deviation.

[0078] Among them, the intermediate adjustment termination threshold S2 corresponds to the highest accuracy (second adjustment accuracy level) of the intermediate adjustment module, and should be between S1 and S3, serving as a transition target between coarse and fine adjustment. For example, if the adjustment range of the intermediate adjustment module is 10ps~100ps (accuracy ±5ps), then S2 can be set to 10ps, meaning that ΔT after intermediate adjustment should be ≤10ps, ensuring that the fine adjustment module can handle the remaining small deviations.

[0079] Among them, the fine-tuning termination threshold S3 corresponds to the highest precision (third adjustment precision level) of the fine-tuning module, which is the final goal and must meet the core computing requirements of the optical computing unit. For example, if the phase error of optical computing needs to be ≤0.1 radians, corresponding to a delay deviation of ≤0.5ps for 1550nm light, then S3 can be set to 0.5ps, that is, after fine-tuning, ΔT needs to be ≤0.5ps to ensure computing accuracy.

[0080] 2. Set the length and width of the optical waveguide delay structure according to the optical delay between the current optical computing unit and the reference optical computing unit, so that ΔT≤S1.

[0081] In this embodiment, the core advantage of the coarse adjustment module is its large adjustment range, making it suitable for handling large initial deviations. Its adjustment is based on the core formula Ti=t0+tf=T0 (tf=(k1*L)(k2*W) / C). That is, by designing L and W, the compensation delay tf provided by the optical waveguide delay structure is made to precisely offset the deviation of the initial delay t0, ultimately ensuring that the coarsely adjusted ΔT=|Ti-T0|≤S1.

[0082] In practice, the initial delay t0 of the current unit is first measured by a photodetector, and the required compensation delay tf = T0 - t0 is calculated. Then, based on tf = (k1*L)(k2*W) / C, combined with material parameters (k1, k2, C) and width constraints (k2W≤W0, to avoid multimode transmission), the target values ​​of L and W are calculated, thereby designing an optical waveguide with an initial structure of length L and width W. Then, the optical waveguide delay structure adjusted by K1 and K2 is fabricated through photolithography and etching processes to make ΔT≤S1.

[0083] 3. Based on the coarsely adjusted optical delay, a preset second adjustment condition is applied to the optical waveguide delay structure through the intermediate adjustment module, such that ΔT≤S2.

[0084] Although ΔT ≤ S1 after coarse adjustment, a moderate deviation still exists, exceeding the adjustment range of the fine adjustment module. The intermediate adjustment module achieves moderate precision adjustment by controlling the refractive index of the optical waveguide delay structure, compressing the deviation to a range that the fine adjustment can handle.

[0085] In practice, a photodetector is used to measure the coarsely adjusted Ti, and ΔT = |Ti - T0| is calculated. The main control module calculates the required second adjustment condition based on the mapping relationship between ΔT and the intermediate refractive index-delay of the intermediate adjustment module. The second adjustment unit of the intermediate adjustment module applies the target condition, monitors the change of ΔT in real time, and adjusts according to ΔT until ΔT ≤ S2.

[0086] 4. Based on the adjusted optical delay, a preset third adjustment condition is applied to the optical waveguide of the optical computing unit through the fine-tuning module, so that ΔT≤S3.

[0087] Although the ΔT is smaller after the mid-tuning, it may still affect the core operations of optical computing. The fine-tuning module achieves ultra-high precision adjustment by regulating the local refractive index of the optical waveguide of the optical computing unit, ensuring that the final deviation meets the system requirements.

[0088] The adjusted Ti is measured using a photodetector, and ΔT = |Ti - T0| is calculated. The main control module calculates the required third adjustment condition based on the mapping relationship between ΔT and the modulation layer state-delay of the fine-tuning module. The third adjustment unit of the fine-tuning module applies the target condition, monitors changes in ΔT, and adjusts according to ΔT until ΔT ≤ S3.

[0089] In other embodiments, due to the enormous number of optical computing units in a large-scale optical computing matrix, adjusting the optical delay one by one is a huge undertaking. Therefore, this embodiment also provides a method for partitioned adjustment, such as... Figure 2 As shown, the specific steps include:

[0090] 1. The optical computing units in the optical computing matrix are partitioned according to a preset partitioning rule to obtain several sub-matrices, and a sub-matrix optical waveguide delay structure is connected in series on each sub-matrix.

[0091] This embodiment provides two types of partitioning rules:

[0092] (1) First, determine the size of the submatrix and divide the submatrix according to the physical location of the optical computing unit.

[0093] The process errors of ultra-large-scale optical computing matrices, such as lithography precision, waveguide etching depth, and environmental interference, exhibit significant spatial regionality. Physically adjacent cells show highly similar patterns in their initial delay deviation. Therefore, subdividing the matrix according to physical location is a simple and efficient method, but it is less effective at handling individual abnormal cells. For example, specifying 16x16 as the cell submatrix results in multiple 16x16 submatrices for an ultra-large-scale optical computing matrix.

[0094] Preferably, the step of determining the size of the submatrix includes: specifying multiple candidate submatrixes, then performing region division through simulation to obtain N1 partitions based on the first candidate submatrix, N2 partitions based on the second candidate submatrix, and so on; then calculating the maximum internal delay (i.e., the maximum delay between computing units within each submatrix) and the maximum external delay difference between adjacent submatrixes for each candidate submatrix (such as the first candidate submatrix, or the second candidate submatrix, etc.); then calculating the cost index Cos=w1*maximum internal delay+w2*maximum external delay difference for each candidate submatrix, and taking the candidate submatrix corresponding to the minimum value as the target submatrix, where w1 and w2 are weighting coefficients, taken as empirical values. The division of the submatrix should not be as small as possible, because smaller divisions require more submatrix optical delay structures, which increases both integration and adjustment difficulty; larger divisions may not achieve the desired adjustment accuracy. Therefore, this embodiment provides a method to select a more suitable submatrix from the candidate submatrixes for region division.

[0095] (2) Second, all optical computing units are clustered according to the optical delay between them and the reference optical computing unit, and sub-matrixes are divided according to the clustering results.

[0096] In practice, the actual delay Ti of each optical computing unit is first measured using a photodetector, and the delay deviation ΔTi = |Ti - T0| between each optical computing unit and the reference unit is calculated to form a dataset. Then, a clustering algorithm suitable for a preset number of categories, such as the K-means algorithm, is used to divide the dataset into K clusters. The algorithm iteratively calculates to minimize the variance of ΔTi within the same cluster and maximize the variance of ΔTi between different clusters. For example, units with ΔTi of 5-15 ps are assigned to cluster 1, 16-25 ps to cluster 2, 26-35 ps to cluster 3, and so on.

[0097] In addition, each cluster corresponds to a submatrix. If the physical locations of the units within a cluster are scattered, they can be adjusted through proximity optimization. For example, the scattered units can be assigned to the clusters with the closest physical locations to avoid the submatrix span being too large, which would lead to complex wiring. In the end, each submatrix contains a set of units with similar delay deviations and relatively concentrated physical locations.

[0098] 2. Each submatrix contains an internal reference light calculation unit, and all light calculation units in each submatrix use their light delay modules to adjust the light delay to be consistent with that of the internal reference light calculation unit.

[0099] The purpose of this step is to eliminate local errors within the submatrix, laying the foundation for subsequent global calibration.

[0100] The internal reference light calculation unit serves as a local anchor point within the sub-matrix, simplifying the adjustment target of the unit within the sub-matrix from the global reference T0 to the internal reference Ti_ref, thus avoiding detection errors caused by excessive distance when the unit within the sub-matrix is ​​directly aligned with the global reference.

[0101] All optical computing units within the submatrix complete the calibration through their own optical delay modules. The specific process is the same as that for single-unit adjustment, and will not be described in detail here.

[0102] 4. Set a reference submatrix in all submatrices, and adjust the optical delay of each submatrix to be consistent with the reference submatrix using its cascaded submatrix waveguide delay structure.

[0103] The reference reference submatrix is ​​usually selected as the submatrix with the largest optical delay, and the delay of its internal reference reference cell is directly traced back to the global reference T0.

[0104] The sub-matrix optical waveguide delay structure is the carrier for calibration between sub-matrices. It is connected in series at the signal input end of each sub-matrix, meaning that the optical signal of the entire sub-matrix must first pass through this delay structure before entering the unit. Its function is to provide overall delay compensation for the entire sub-matrix.

[0105] Compared to the mode where all units are directly calibrated to the global benchmark, the submatrix calibration significantly reduces the complexity of global adjustment through overall submatrix compensation.

[0106] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0107] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a computer terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0108] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A large-scale optical computing matrix with adjustable optical delay, characterized in that: The system includes a light-transmitting substrate on which a light-transmitting substrate is integrated; each light-computing unit in the light-computing matrix is ​​connected to a light delay module; the light delay module includes a coarse adjustment module with a first adjustment accuracy level; and also includes a photodetector connected to the light-computing unit, the photodetector being connected to a main control module. The coarse adjustment module includes an optical waveguide delay structure; the length L and width W of the optical waveguide delay structure satisfy the following condition: Ti=T0; Ti = tf + t0; tf=(k1*L)*(k2*W) / C; Where k1 and k2 are adjustment coefficients, k2*W≤W0, W0 is the preset maximum width threshold; C is the propagation speed of light in the optical waveguide; T0 is the optical delay of the reference optical computing unit, Ti is the optical delay of the current optical computing unit, t0 is the initial optical delay of the current optical computing unit, and tf is the optical delay provided by the optical waveguide delay structure of the current optical computing unit. The optical delay module further includes: a mid-tuning module acting on the optical waveguide delay structure, and a fine-tuning module acting on the optical waveguide of the optical computing unit; the mid-tuning module has a second adjustment accuracy level, and the fine-tuning module has a third accuracy adjustment level; the first adjustment accuracy level > the second adjustment accuracy level > the third accuracy adjustment level.

2. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 1, characterized in that: The fine-tuning module is configured to apply a preset third adjustment condition to the optical waveguide of the optical computing unit; The main control module is configured to perform delay calculation based on the optical signal detected by the photodetector, compare the calculated optical delay Ti with T0 to obtain ΔT, ΔT=|Ti-T0|; and control the fine-tuning module to apply a third adjustment condition to the optical waveguide of the optical computing unit to change the refractive index of the optical waveguide of the optical computing unit, thereby reducing ΔT.

3. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 2, characterized in that: The fine-tuning module includes a fine-tuning element and a third adjustment unit. The fine-tuning element is a modulation layer laid on the optical waveguide of the optical computing unit. The third adjustment unit is configured to change the refractive index of the optical waveguide of the optical computing unit by switching the state of the fine-tuning element.

4. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 1, characterized in that: The mid-tuning module is configured to apply a preset second adjustment condition to the optical waveguide delay structure; The main control module is configured to perform delay calculation based on the optical signal detected by the photodetector, and compare the calculated optical delay Ti with T0 to obtain ΔT, where ΔT = |Ti - T0|; and control the intermediate adjustment module to apply a second adjustment condition to the optical waveguide delay structure to change the refractive index of the optical waveguide delay structure, thereby reducing ΔT.

5. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 4, characterized in that: The intermediate modulation module includes an intermediate modulation element and a second adjustment unit. The intermediate modulation element is one of a thermal modulation circuit, an electrical modulation circuit, and an optical modulation circuit. The second adjustment unit is configured to apply heat to the optical waveguide delay structure using the intermediate modulation element to change its refractive index, or to apply an electric field to the optical waveguide delay structure using the intermediate modulation element to change its refractive index, or to apply light energy to the optical waveguide delay structure using the intermediate modulation element to change its refractive index.

6. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 5, characterized in that... The steps for adjusting the optical delay of the optical computing unit using the optical delay module are as follows: Set the coarse adjustment termination threshold S1, the medium adjustment termination threshold S2, and the fine adjustment termination threshold S3; The length and width of the corresponding optical waveguide delay structure are set according to the optical delay between the current optical computing unit and the reference optical computing unit, such that ΔT≤S1; Based on the coarsely adjusted optical delay, a preset second adjustment condition is applied to the optical waveguide delay structure through the intermediate adjustment module, such that ΔT≤S2; Based on the adjusted optical delay, a preset third adjustment condition is applied to the optical waveguide of the optical computing unit through the fine-tuning module, so that ΔT≤S3.

7. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 1, characterized in that: The optical waveguide delay structure is a spiral or serpentine arrangement of optical waveguides.

8. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 1, characterized in that: When determining the width and length of the optical waveguide delay structure of a certain optical computing unit, the adjustment coefficient k2 is modified so that the optical delay Ti of the optical computing unit approaches the optical delay T0 of the reference optical computing unit, provided that k2*W≤W0. Modify the adjustment coefficient k1 so that the optical delay Ti of this optical computing unit is closer to the optical delay T0 of the reference optical computing unit.

9. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 1, characterized in that: The optical computing units in the optical computing matrix are partitioned according to a preset partitioning rule to obtain several sub-matrices, and a sub-matrix optical waveguide delay structure is connected in series on each sub-matrix. Each submatrix contains an internal reference light calculation unit, and all light calculation units in each submatrix use their light delay modules to adjust the light delay to be consistent with that of the internal reference light calculation unit. A reference submatrix is ​​set in all submatrices, and each submatrix adjusts its optical delay to match that of the reference submatrix using its cascaded submatrix waveguide delay structure.

10. The ultra-large-scale optical computing matrix with adjustable optical delay according to claim 9, characterized in that... The partitioning rules are as follows: determine the size of the submatrix and divide the submatrix according to the physical location of the optical computing units; or, cluster all optical computing units according to the optical delay between them and the reference optical computing unit, and divide the submatrix according to the clustering results.

Images