All-optical signal processor

Abstract

The application provides a kind of all-optical signal processor, comprising: signal expansion layer, optical reservoir and optical path splitting layer;Wherein, the signal expansion layer is used to sample continuous optical signal, and determines time division spread signal based on mask matrix and sampled optical signal;The time division spread signal represents different signal state of each time;The optical reservoir is connected with the signal expansion layer and the optical path splitting layer respectively, for spatial mapping to the time division spread signal, to obtain the mapped signal;The optical path splitting layer is used to split the mapped signal for optical path processing, to obtain the processed signal.Through signal expansion layer, optical reservoir and optical path splitting layer, the processing of continuous optical signal is realized, which can process continuous optical signal of different wavelengths in high-speed, multi-wavelength complex environment, avoid signal damage caused by wavelength difference, and improve the processing performance of optical signal.

Description

Technical Field

[0001] This invention relates to the field of chip technology, and more particularly to an all-optical signal processor. Background Technology

[0002] In the fields of data centers and high-performance computing, with the exponential expansion of large-scale artificial intelligence (AI) models, large-scale training tasks have gradually evolved into a distributed computing model that relies on multiple data centers to collaborate through optical interconnect (OI) networks. In order for numerous graphics processing units (GPUs) to work collaboratively as a single large GPU, the optical interconnect system must have ultra-low latency and highly consistent latency characteristics.

[0003] Traditional optical interconnect solutions typically employ an OI (Optical Integrator) approach combined with Digital Signal Processing (DSP). However, the number of DSPs increases significantly with data volume and computational complexity, leading to a substantial increase in overall system latency. Furthermore, while some existing all-optical interconnect solutions can reduce latency to some extent, they still suffer from wavelength sensitivity issues when dealing with complex scenarios involving high speeds and multiple wavelengths. This means that signals of different wavelengths may experience performance degradation during processing due to factors such as dispersion.

[0004] Therefore, there is an urgent need for an all-optical signal processor that can maintain low latency and high consistency in multi-wavelength, high-speed environments to achieve optical signal processing. Summary of the Invention

[0005] This invention provides an all-optical signal processor to solve the problem of achieving optical signal processing by providing an all-optical signal processor that can maintain low latency and high consistency in multi-wavelength, high-speed environments.

[0006] This invention provides an all-optical signal processor, comprising: a signal extension layer, an optical reservoir, and an optical path splitting layer; wherein, The signal expansion layer is used to sample continuous optical signals and determine time-division expanded signals based on the mask matrix and the sampled optical signals; the time-division expanded signals represent different signal states at each time. The optical reservoir is connected to the signal extension layer and the optical path splitting layer respectively, and is used to perform spatial mapping on the time-division unfolded signal to obtain the mapped signal. The optical path splitting layer is used to split the mapped signal into optical paths to obtain a processed signal.

[0007] According to an all-optical signal processor provided by the present invention, the optical storage pool includes multiple series-connected optical storage units, wherein the optical storage unit includes an adjustable feedback delay loop, and the adjustable feedback delay loop is composed of multiple phase shifters and multiple nonlinear activation nodes.

[0008] According to an all-optical signal processor provided by the present invention, the delay of the adjustable feedback delay loop is a non-integer multiple of the symbol period Ts.

[0009] According to an all-optical signal processor provided by the present invention, the phase shifter is used to control the feedback strength of the adjustable feedback delay loop.

[0010] According to an all-optical signal processor provided by the present invention, the nonlinear activation node is a multiply-accumulate activator constructed based on an interferometer, or the nonlinear activation node is a semiconductor optical amplifier.

[0011] According to an all-optical signal processor provided by the present invention, the optical reservoir is specifically used for: The time-division unfolded signal is sequentially input into the multi-stage cascaded optical storage unit to obtain the target signal output by each stage of the optical storage unit; Based on each of the target signals, the mapped signal is determined.

[0012] According to an all-optical signal processor provided by the present invention, the optical reservoir is further used for: The time-division unfolded signal is input to the first-stage optical storage unit in the multi-stage cascaded optical storage unit to obtain the first signal corresponding to the adjustable feedback delay loop in the first-stage optical storage unit; Based on the first signal, a first target signal corresponding to the first signal and a target signal of the first-stage optical storage unit are determined; the first target signal is used to drive the second-stage optical storage unit. The first target signal is input to the second-stage optical storage unit to obtain the second signal corresponding to the adjustable feedback delay loop in the second-stage optical storage unit; Based on the second signal, a second target signal corresponding to the second signal and a target signal for the second-stage optical storage unit are determined; the second target signal is used to drive the third-stage optical storage unit. Based on the second target signal, the target signal for each remaining optical storage unit is determined in the same manner as described above.

[0013] According to an all-optical signal processor provided by the present invention, the optical path splitting layer is specifically used for: The mapped signal is split into optical paths to obtain the first optical path signal corresponding to each optical path in the multiple optical paths; For each optical path, the optical path signal is passed through the optical path to obtain the second optical path signal; Multiply the second optical path signal by the complex weighting coefficients corresponding to the optical path to obtain the weighted signal; The processed signal is determined based on each of the weighted signals.

[0014] According to an all-optical signal processor provided by the present invention, the optical path splitting layer is further specifically used for: The weighted signals are summed to obtain the summed signal; The summed signals are fed into a photodetector to obtain the processed signal.

[0015] According to the all-optical signal processor provided by the present invention, the delay of multiple optical paths increases with the target delay.

[0016] According to the all-optical signal processor provided by the present invention, each optical path is connected to an interferometer, and the complex weighting coefficients are obtained based on the interferometer.

[0017] According to an all-optical signal processor provided by the present invention, at least one of the etching depth, width and sidewall tilt angle of the waveguide cross section of the adjustable feedback delay loop is set so that the waveguide exhibits flat anomalous dispersion characteristics.

[0018] According to an all-optical signal processor provided by the present invention, at least one of the following is provided in the coupling structure between multi-level series-connected optical storage units: coupling spacing, sidewall tilt angle, width, and thickness, so that the coupling strength of optical signals of different wavelengths is the same.

[0019] The present invention also provides a chip, characterized in that it includes: any of the above-mentioned all-optical signal processors.

[0020] The all-optical signal processor provided by this invention includes: a signal expansion layer, an optical reservoir, and an optical path splitting layer. The signal expansion layer samples continuous optical signals and determines a time-division expanded signal based on a mask matrix and the sampled optical signals. The time-division expanded signal represents different signal states at each time point. The optical reservoir is connected to both the signal expansion layer and the optical path splitting layer, and is used to spatially map the time-division expanded signal to obtain a mapped signal. The optical path splitting layer splits the mapped signal into optical paths to obtain a processed signal. Through the signal expansion layer, optical reservoir, and optical path splitting layer, continuous optical signal processing is achieved, enabling the processing of continuous optical signals of different wavelengths in complex high-speed, multi-wavelength environments, avoiding signal damage caused by wavelength differences, and improving the processing performance of optical signals. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is one of the structural schematic diagrams of the all-optical signal processor provided by the present invention.

[0023] Figure 2 This is a flowchart illustrating the determination of the time-division unfolded signal provided by the present invention.

[0024] Figure 3 This is a schematic diagram showing the group velocity dispersion comparison of optical signals of different wavelengths provided by the present invention.

[0025] Figure 4 This is a schematic diagram of the processing procedure for the optical reservoir and optical path splitting layer provided by the present invention.

[0026] Figure 5 This is the second schematic diagram of the all-optical signal processor provided by the present invention.

[0027] Figure 6 This is a schematic diagram of the chip structure provided by the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this 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 this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0029] The following is combined Figures 1-6 The all-optical signal processor of the present invention is described.

[0030] Figure 1 This is one of the structural schematic diagrams of the all-optical signal processor provided by the present invention, such as... Figure 1 As shown, the all-optical signal processor includes: a signal expansion layer 101, an optical reservoir 102, and an optical path splitting layer 103; wherein, the signal expansion layer 101 is used to sample continuous optical signals and determine a time-division unfolded signal based on a mask matrix and the sampled optical signals; the time-division unfolded signal represents different signal states at each time; the optical reservoir 102 is connected to the signal expansion layer 101 and the optical path splitting layer 103 respectively, and is used to spatially map the time-division unfolded signal to obtain a mapped signal; the optical path splitting layer 103 is used to split the mapped signal into optical paths to obtain a processed signal.

[0031] Specifically, Figure 2 This is a flowchart illustrating the determination of the time-division unfolded signal provided by the present invention, as shown below. Figure 2 As shown, to enable the continuous optical signal to be dynamically modeled by the optical reservoir 102, the signal extension layer 101 first samples the continuous optical signal u(t) to obtain the sampled optical signal u(k), which represents the signal value at each time point. Based on the mask matrix (M∈R^{N×1}) and the sampled optical signal u(k), the following steps are taken: Determine the time-division expanded signal J(t); the time-division expanded signal represents the different signal states at each time point. Inject the timing node corresponding to the time-division expanded signal J(t) into the optical storage pool 102 to complete the information initialization. The timing node represents the signal value of the time-division expanded signal J(t) at each time point t.

[0032] It should be noted that the time-division expanded signal J(t) can adapt to various modulation methods, data rates, and channel wavelengths. For example, the different modulation methods are On-Off Keying (OOK), Amplitude Shift Keying (ASK), and Frequency Shift Keying (FSK), with a data rate of 100Gbps and channel wavelengths of B-band and C-band.

[0033] The optical reservoir 102 is the core of the entire computation. The optical reservoir 102 is responsible for mapping the input time-division expanded signal J(t) to a high-dimensional dynamic space to obtain the mapped signal X(t). The mapped signal represents the multidimensional signal state at each time point.

[0034] To extract useful information from the high-dimensional signal state, the optical path splitting layer 103 splits the mapped signal into optical paths to obtain the processed signal y(t).

[0035] The all-optical signal processor provided by this invention includes: a signal expansion layer, an optical reservoir, and an optical path splitting layer. The signal expansion layer samples continuous optical signals and determines a time-division expanded signal based on a mask matrix and the sampled optical signals. The time-division expanded signal represents different signal states at each time point. The optical reservoir is connected to both the signal expansion layer and the optical path splitting layer, and is used to spatially map the time-division expanded signal to obtain a mapped signal. The optical path splitting layer splits the mapped signal into optical paths to obtain a processed signal. Through the signal expansion layer, optical reservoir, and optical path splitting layer, continuous optical signal processing is achieved, enabling the processing of continuous optical signals of different wavelengths in complex high-speed, multi-wavelength environments, avoiding signal damage caused by wavelength differences, and improving the processing performance of optical signals.

[0036] Optionally, the optical reservoir includes multiple series-connected optical storage units, wherein each optical storage unit includes an adjustable feedback delay loop, which consists of multiple phase shifters and multiple nonlinear activation nodes.

[0037] Optionally, the delay of the adjustable feedback delay loop is a non-integer multiple of the symbol period Ts.

[0038] Optionally, the phase shifter is used to control the feedback strength of the adjustable feedback delay loop.

[0039] Optionally, the nonlinear activation node is a multiply-add activator built based on an interferometer, or the nonlinear activation node is a semiconductor optical amplifier.

[0040] Specifically, the optical storage pool comprises multiple levels of cascaded optical storage units. The number of optical storage units can be set according to actual needs; for example, setting three optical storage units means the optical storage pool comprises three levels of cascaded optical storage units. Each optical storage unit includes an adjustable feedback delay loop. The delay of each level of the adjustable feedback delay loop can be the same or different. The delay of the adjustable feedback delay loop in the first level of the optical storage unit can be expressed as... The time delay of the adjustable feedback delay loop included in the first-stage optical storage unit can be expressed as: The time delay of the adjustable feedback delay loop included in the Nth-level optical storage unit can be expressed as: Delay of the adjustable feedback delay loop For example, a non-integer multiple of the symbol period Ts, With durations of 0.8Ts, 1.3Ts, or 1.7Ts, the design breaks temporal symmetry by using non-integer multiples, thereby enhancing short-term memory capacity and dynamic characteristics.

[0041] The phase shifter is used to control the feedback strength K of the adjustable feedback delay loop. The feedback strength of the adjustable feedback delay loop included in the first-stage optical storage unit can be expressed as K1, the feedback strength of the adjustable feedback delay loop included in the first-stage optical storage unit can be expressed as K2, and the feedback strength of the adjustable feedback delay loop included in the Nth-stage optical storage unit can be expressed as K3.

[0042] Nonlinear activation nodes can be multiply-accumulate activators built based on interferometers, such as Mach-Zehnder interferometers (MZI); alternatively, nonlinear activation nodes can be semiconductor optical amplifiers (SOA). Saturated nonlinear activation can be achieved through multiply-accumulate activators or semiconductor optical amplifiers.

[0043] Optionally, at least one of the etching depth, width, and sidewall tilt angle of the waveguide cross section of the adjustable feedback delay loop is set to make the waveguide exhibit flat anomalous dispersion characteristics.

[0044] Specifically, the waveguide cross-section of the adjustable feedback delay loop can be configured by setting at least one of the etching depth, width, and sidewall tilt angle. By adjusting the geometry of the waveguide cross-section, the geometric dispersion of the optical field is designed, reversing the inherent dispersion of the material. This results in the waveguide exhibiting flat anomalous dispersion characteristics, allowing optical signals of different wavelengths to have similar group velocity dispersion (GVD) characteristics during transmission, i.e., wavelength insensitivity. In other words, designing a waveguide structure with a flat dispersion response ensures that optical signals of different wavelengths experience consistent dispersion effects in the optical reservoir, thereby avoiding signal damage caused by wavelength differences.

[0045] Optionally, at least one of the following is provided in the coupling structure between the multi-level series optical storage units: coupling spacing, sidewall tilt angle, width, and thickness, so that the coupling strength of optical signals of different wavelengths is the same.

[0046] Specifically, there is a coupling structure between the multi-stage series optical storage units. This coupling structure is a waveguide structure. At least one of the coupling spacing, sidewall tilt angle, width, and thickness of the coupling structure between the multi-stage series optical storage units can be set to achieve a controllable distribution of evanescent field intensity. For optical signals of different wavelengths, corresponding coupling structures are designed to ensure that the coupling intensity of optical signals at different wavelengths is basically consistent, thereby ensuring the consistency and low-loss transmission of multi-wavelength signals during the coupling process.

[0047] The coupling structure can be symmetrical, which reduces the directional deviation of coupling between wavelengths and improves the consistency of processing between wavelengths.

[0048] For the waveguide and coupling structure of the tunable feedback delay loop, the waveguide material can be either silicon nitride (SiN) or lithium niobate (LN). Silicon nitride has low loss (<1 dB / cm) and tunable dispersion, making it suitable for short-to-medium distance high-speed interconnects. Lithium niobate has strong electro-optic modulation capabilities, making it suitable for high-speed MZI weighted modulation. Both silicon nitride and lithium niobate support silicon photonics platforms compatible with CMOS processes, facilitating large-scale manufacturing.

[0049] Figure 3 This is a schematic diagram of the group velocity dispersion comparison of optical signals of different wavelengths provided by the present invention, as shown below. Figure 3 As shown, the left figure is a schematic diagram of setting at least one of the etching depth, width, and sidewall tilt angle of the waveguide cross-section of the adjustable feedback delay loop provided by the present invention, so that the waveguide exhibits flat anomalous dispersion characteristics. The right figure is a schematic diagram of the dispersion characteristics of an existing waveguide. For waveguide dispersion modulation, LN material is selected for all cases. The waveguide width in the left figure is 2 micrometers and the depth is 1 micrometer, while the waveguide width in the right figure is 1.2 micrometers and the depth is 0.6 micrometers. The bending radii are 60, 80, 100, and 150 micrometers, respectively, and the transverse electric (TE) fundamental mode is selected.

[0050] Optionally, the optical reservoir is specifically used for: The time-division unfolded signal is sequentially input into the multi-stage cascaded optical storage unit to obtain the target signal output by each stage of the optical storage unit; based on each target signal, the mapped signal is determined.

[0051] Specifically, by sequentially inputting the time-division expanded signal into a multi-stage cascaded optical storage unit, the target signal x output by each stage of the optical storage unit can be obtained.i (t). The target signals x i Adding (t) together, we can obtain the mapped signal X(t).

[0052] Optionally, the optical reservoir is further used for: The time-division unfolded signal is input to the first-stage optical storage unit in the multi-stage cascaded optical storage unit to obtain a first signal corresponding to the adjustable feedback delay loop in the first-stage optical storage unit; based on the first signal, a first target signal corresponding to the first signal and a target signal of the first-stage optical storage unit are determined; the first target signal is used to drive the second-stage optical storage unit; the first target signal is input to the second-stage optical storage unit to obtain a second signal corresponding to the adjustable feedback delay loop in the second-stage optical storage unit; based on the second signal, a second target signal corresponding to the second signal and a target signal of the second-stage optical storage unit are determined; the second target signal is used to drive the third-stage optical storage unit; based on the second target signal, the target signal of each remaining stage of optical storage unit is determined in the same manner as described above.

[0053] Specifically, the time-division expanded signal J(t) is input to the first-stage optical storage unit in a multi-stage cascaded optical storage unit, using... The first signal corresponding to the adjustable feedback delay loop in the first-stage optical storage unit can be obtained. ,in, This represents a nonlinear transformation (which can be implemented using an MZI or PD+SOA structure). Indicates the number of stages of optical storage units. Let be the time delay of the adjustable feedback delay loop in the i-th stage optical storage unit. This represents the signal corresponding to the adjustable feedback delay loop of stage i.

[0054] The first signal is multiplied by the coupling strength S of the coupling structure between the multi-stage cascaded optical storage units to obtain the first target signal corresponding to the first signal. The first signal is multiplied by the output strength T of the first-stage optical storage unit, and the target signal x1(t) of the first-stage optical storage unit is obtained through a nonlinear device. The first target signal is used to drive the second-stage optical storage unit to form inter-stage residual connections and improve the expression depth.

[0055] The first target signal is input to the second-stage optical storage unit, using... The second signal corresponding to the adjustable feedback delay loop in the second-stage optical storage unit can be obtained. The second signal is multiplied by the coupling strength S of the coupling structure between the multi-stage cascaded optical storage units to obtain the second target signal corresponding to the second signal. The second signal is then multiplied by the output strength T of the second-stage optical storage unit, and the target signal x2(t) of the second-stage optical storage unit is obtained through a nonlinear device. The second target signal is used to drive the third-stage optical storage unit, forming inter-stage residual connections and improving the expression depth. Based on the second target signal, the target signal x of each remaining stage of optical storage unit is determined in the same manner as described above. i (t).

[0056] It should be noted that the signal corresponding to the adjustable feedback delay loop in the final stage optical storage unit... The target signal x is directly identified as the final stage optical storage unit. N (t).

[0057] Optionally, the optical path splitting layer is specifically used for: The mapped signal is split into optical paths to obtain a first optical path signal corresponding to each optical path in the multiple optical paths; for each optical path, the optical path signal is passed through the optical path to obtain a second optical path signal; the second optical path signal is multiplied by the complex weighting coefficients corresponding to the optical path to obtain a weighted signal; based on each weighted signal, the processed signal is determined.

[0058] Optionally, the delays of multiple optical paths increase in increments of the target delay. The target delay is represented as... ,For example, =4ps.

[0059] Optionally, each optical path is connected to an interferometer, and the complex weighting coefficients are obtained based on the interferometer.

[0060] Specifically, the number of optical paths can be set according to actual conditions. For example, the number of optical paths M can be set to 16. Each optical path is connected to an interferometer, which can be a Mach-Zehnder interferometer (MZI). Based on the interferometer, the current and voltage are adjusted by electrically adjusting the phase to achieve complex weighting, thereby continuously adjusting the complex weighting coefficients to obtain complex weighting coefficients that meet preset conditions.

[0061] By splitting the mapped signal X(t) into optical paths, we can obtain the first optical path signal corresponding to each of the multiple optical paths. x j For each optical path, the optical path signal x j After passing through the optical path, the second optical path signal is obtained. Multiplying the second optical path signal and the complex weighting coefficients corresponding to the optical path yields the weighted signal; based on each weighted signal, the processed signal can be determined.

[0062] Optionally, the optical path splitting layer is further specifically used for: The weighted signals are summed to obtain a summed signal; the summed signal is then fed into a photodetector to obtain the processed signal.

[0063] Specifically, the weighted signals are summed to obtain a summed signal; this summed signal is then fed into a photodetector (PD) to obtain a processed signal. Using formula express, Indicates the first j Complex weighting coefficients corresponding to each optical path Indicates the first j The second optical path signal corresponding to the optical path, Indicates the first j The time delay of each optical path, M represents the total number of optical paths.

[0064] It should be noted that the optical path splitting layer integrates PD and thermally tuned MZI, supporting online adjustment of complex weighting coefficients and thermal drift compensation to ensure the stability of waveguide parameters.

[0065] Figure 4 This is a schematic diagram of the processing procedure for the optical reservoir and optical path splitting layer provided by the present invention, as shown below. Figure 4 As shown, the time-division expanded signal J(t) is input to the first-stage optical storage unit in a multi-stage cascaded optical storage unit to obtain the first signal corresponding to the adjustable feedback delay loop in the first-stage optical storage unit. ; the first signal The coupling strength S between the multi-stage cascaded optical storage units is multiplied to obtain the first target signal corresponding to the first signal. The first signal is then multiplied by the output strength T of the first-stage optical storage unit, and the target signal x1(t) of the first-stage optical storage unit is obtained through a nonlinear device. The first target signal is used to drive the second-stage optical storage unit, forming inter-stage residual connections and improving the representation depth. The first target signal is input to the second-stage optical storage unit to obtain the second signal corresponding to the adjustable feedback delay loop in the second-stage optical storage unit. Based on the second signal The coupling strength S between the multi-stage cascaded optical storage units is multiplied to obtain the second target signal corresponding to the second signal. The second signal is then multiplied by the output strength T of the second-stage optical storage unit, and the target signal x2(t) of the second-stage optical storage unit is obtained through a nonlinear device. This second target signal drives the third-stage optical storage unit, forming inter-stage residual connections and enhancing the representation depth. Based on the second target signal, the target signals of each remaining stage of optical storage units are determined in the same manner, with the target signal of the last stage being... The target signal x at each level i The signals X(t) and X(t) are added together to obtain the mapped signal X(t). The mapped signal X(t) is then processed by optical path splitting to obtain the first optical path signal corresponding to each of the multiple optical paths. x j For each optical path, the optical path signal x j After passing through the optical path, the second optical path signal is obtained. The complex weighting coefficients corresponding to the second optical path signal and the optical path are calculated. W j Multiply the signals to obtain a weighted signal; sum the weighted signals to obtain a summed signal; input the summed signal into a photodetector to obtain a processed signal. .

[0066] Figure 5 This is the second schematic diagram of the all-optical signal processor provided by the present invention, as shown below. Figure 5 As shown, the signal spreading layer uses a mask matrix (M) and the sampled optical signal u(k). The time-division expanded signal J(t) is determined; the time-division expanded signal J(t) is input to the optical reservoir, which consists of three cascaded optical reservoir units. Each optical reservoir unit includes an adjustable feedback delay loop, which is composed of multiple phase shifters and multiple nonlinear activation nodes. It should be noted that, for illustration purposes, only one adjustable feedback delay loop is used. This represents the delay of the adjustable feedback delay loop. This represents the time delay interval between nonlinear active nodes, which is based on... Determine, for example (The diagram illustrates 7 non-linear activation nodes.)

[0067] The time-division expanded signal J(t) is sequentially input into a multi-stage cascaded optical storage unit to obtain the target signal output by each stage of the optical storage unit. The target signals are then summed to obtain the mapped signal. The mapped signal is then split into optical paths to obtain the first optical path signal for each of the multiple optical paths. For each optical path, the first optical path signal is passed through the optical path to obtain the second optical path signal. The second optical path signal is multiplied by the complex weighting coefficients corresponding to the optical path to obtain the weighted signal. The weighted signals are then summed to obtain the summed signal. Finally, the summed signal is input into a photodetector to obtain the processed signal.

[0068] The all-optical signal processor provided by this invention enhances the system's nonlinear characteristics and processing power through a multi-stage cascaded optical storage unit design combined with MZI complex weighting. By optimizing the waveguide cross-section and coupling structure of the optical storage units, wavelength insensitivity is achieved, maintaining consistent processing performance across multiple wavelength environments. The all-optical interconnect scheme ensures consistently ultra-low latency regardless of data rate expansion. Adjusting the complex weighting coefficients during training requires no additional energy consumption or chip area, offering significant advantages in energy efficiency and hardware resource utilization.

[0069] The all-optical signal processor provided by this invention overcomes the performance bottleneck of traditional solutions in multi-wavelength scenarios due to its wavelength insensitivity, and can adapt to the high-density wavelength multiplexing requirements of data centers. Its all-optical interconnect architecture achieves ultra-low latency and consistent transmission, significantly improving the real-time collaborative efficiency of computing clusters compared to traditional OI+DSP solutions. Furthermore, its compatibility with multiple modulation formats, rates, and wavelengths makes it widely applicable in diverse scenarios such as 5G and AI supercomputing.

[0070] Based on any of the above embodiments, the present invention also provides a chip, which will be described below in conjunction with... Figure 6 The structure of the chip is explained.

[0071] Figure 6 This is a schematic diagram of the chip structure provided by the present invention, as shown below. Figure 6As shown, chip 600 includes an all-optical signal processor 6001. Since chip 600 includes the all-optical signal processor 6001, the all-optical signal processor 6001 includes a signal expansion layer, an optical reservoir, and an optical path splitting layer. The signal expansion layer is used to sample continuous optical signals and determine a time-division unfolded signal based on a mask matrix and the sampled optical signals. The time-division unfolded signal represents different signal states at each time point. The optical reservoir is connected to both the signal expansion layer and the optical path splitting layer, and is used to spatially map the time-division unfolded signal to obtain a mapped signal. The optical path splitting layer is used to split the mapped signal into optical paths to obtain a processed signal. Therefore, through the signal expansion layer, the optical reservoir, and the optical path splitting layer, continuous optical signals of different wavelengths can be processed in complex environments with high speeds and multiple wavelengths, avoiding signal damage caused by wavelength differences and improving the processing performance of optical signals.

[0072] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An all-optical signal processor, characterized in that, include: Signal extension layer, optical reservoir, and optical path splitting layer; among which, The signal expansion layer is used to sample continuous optical signals and determine time-division expanded signals based on the mask matrix and the sampled optical signals; the time-division expanded signals represent different signal states at each time. The optical reservoir is connected to the signal extension layer and the optical path splitting layer respectively, and is used to perform spatial mapping on the time-division unfolded signal to obtain the mapped signal. The optical path splitting layer is used to split the mapped signal into optical paths to obtain a processed signal.

2. The all-optical signal processor according to claim 1, characterized in that, The optical reservoir includes multiple series-connected optical storage units, wherein each optical storage unit includes an adjustable feedback delay loop, which is composed of multiple phase shifters and multiple nonlinear activation nodes.

3. The all-optical signal processor according to claim 2, characterized in that, The delay of the adjustable feedback delay loop is a non-integer multiple of the symbol period Ts.

4. The all-optical signal processor according to claim 2, characterized in that, The phase shifter is used to control the feedback strength of the adjustable feedback delay loop.

5. The all-optical signal processor according to claim 2, characterized in that, The nonlinear activation node is either a multiply-add activator built based on an interferometer, or a semiconductor optical amplifier.

6. The all-optical signal processor according to claim 2, characterized in that, The optical reservoir is specifically used for: The time-division unfolded signal is sequentially input into the multi-stage cascaded optical storage unit to obtain the target signal output by each stage of the optical storage unit; Based on each of the target signals, the mapped signal is determined.

7. The all-optical signal processor according to claim 6, characterized in that, The optical reservoir is also used for: The time-division unfolded signal is input to the first-stage optical storage unit in the multi-stage cascaded optical storage unit to obtain the first signal corresponding to the adjustable feedback delay loop in the first-stage optical storage unit; Based on the first signal, a first target signal corresponding to the first signal and a target signal of the first-stage optical storage unit are determined; the first target signal is used to drive the second-stage optical storage unit. The first target signal is input to the second-stage optical storage unit to obtain the second signal corresponding to the adjustable feedback delay loop in the second-stage optical storage unit; Based on the second signal, a second target signal corresponding to the second signal and a target signal for the second-stage optical storage unit are determined; the second target signal is used to drive the third-stage optical storage unit. Based on the second target signal, the target signal for each remaining optical storage unit is determined in the same manner as described above.

8. The all-optical signal processor according to claim 1, characterized in that, The optical path splitting layer is specifically used for: The mapped signal is split into optical paths to obtain the first optical path signal corresponding to each optical path in the multiple optical paths; For each optical path, the optical path signal is passed through the optical path to obtain the second optical path signal; Multiply the second optical path signal by the complex weighting coefficients corresponding to the optical path to obtain the weighted signal; The processed signal is determined based on each of the weighted signals.

9. The all-optical signal processor according to claim 8, characterized in that, The optical path splitting layer is also specifically used for: The weighted signals are summed to obtain the summed signal; The summed signals are fed into a photodetector to obtain the processed signal.

10. The all-optical signal processor according to claim 8, characterized in that, The delay of multiple optical paths increases with the target delay.

11. The all-optical signal processor according to claim 8, characterized in that, Each optical path is connected to an interferometer, and the complex weighting coefficients are obtained based on the interferometer.

12. The all-optical signal processor according to claim 2, characterized in that, The waveguide cross-section of the adjustable feedback delay loop is configured with at least one of the etching depth, width, and sidewall tilt angle to make the waveguide exhibit flat anomalous dispersion characteristics.

13. The all-optical signal processor according to claim 2, characterized in that, The coupling structure between multi-stage cascaded optical storage units is configured with at least one of the following: coupling spacing, sidewall tilt angle, width, and thickness, so that the coupling strength of optical signals of different wavelengths is the same.

14. A chip, characterized in that, include: The all-optical signal processor according to any one of claims 1 to 13.

Images