Optical multiplexing arithmetic circuit

Abstract

This optical multiplexing arithmetic circuit comprises a spreading circuit, a product-sum arithmetic circuit, a modulation circuit, and despreading circuit. The spreading circuit is capable of generating a plurality of spread signal strings by using a plurality of mutually different code strings to multiply each of a plurality of input signal strings by a different code string for the respective input signal strings. The product-sum arithmetic circuit is capable of generating a plurality of first modulation signals by using the plurality of spread signal strings to modulate each of a plurality of optical signals by a different spread signal string for the respective optical signals. The product-sum arithmetic circuit is further capable of generating a combined optical signal by combining the plurality of first modulation signals that have been generated. The modulation circuit is capable of generating a specific modulated optical signal by performing specific modulation on the combined optical signal. The despreading circuit is capable of generating a plurality of output signal strings by calculating the correlation between the specific modulated optical signal and the plurality of code strings for each code string.

Description

Optical multiplex calculation circuit

[0001] This disclosure relates to an optical multiplexing circuit.

[0002] As an example of an optical multiplexing circuit, a sum-of-accumulate circuit using wavelength multiplexing is known (see, for example, Patent Document 1 and Non-Patent Document 1). It is also known to multiply one or more optical signals by one or more matrix element values ​​using optical amplitude modulation (see, for example, Patent Document 2).

[0003] Special table 2023-536703 publication Special table 2021-527287 publication

[0004] Feldmann, et. al., "Parallel convolutional processing using an integrated photonic tensor core," Vol 589, 7 January 2021.

[0005] Incidentally, optical multiplexing circuits have traditionally faced the problem of being large in scale, making it difficult to reduce the chip area. It is desirable to provide an optical multiplexing circuit that can reduce the size of the optical processing circuit.

[0006] An optical multiplexing circuit relating to one aspect of this disclosure comprises a diffusion circuit, a multiply-accumulate circuit, a modulation circuit, and a despreading circuit. The diffusion circuit can generate multiple diffusion signal sequences by multiplying each of a plurality of input signal sequences by a different code sequence for each input signal sequence using a plurality of different code sequences. The multiply-accumulate circuit can generate multiple first modulated signals by modulating each of a plurality of optical signals with a different diffusion signal sequence for each optical signal using a plurality of diffusion signal sequences. The multiply-accumulate circuit can further generate a composite optical signal by combining the generated plurality of first modulated signals. The modulation circuit can generate a specific modulated optical signal by performing specific modulation on the composite optical signal. The despreading circuit can generate multiple output signal sequences by calculating the correlation between the specific modulated optical signal and a plurality of code sequences for each code sequence.

[0007] An optical multiplexing circuit according to one aspect of this disclosure comprises a spread circuit, a plurality of multiply-accumulate circuits, a plurality of modulation circuits, and a despread circuit. The spread circuit is capable of generating a plurality of spread signal sequences by multiplying each of a plurality of sets of input signal sequences by a plurality of different code sequences. Each input signal sequence group includes a plurality of input signal sequences. Each spread signal sequence group includes a plurality of spread signal sequences. A plurality of multiply-accumulate circuits are provided for each spread signal sequence group. Each multiply-accumulate circuit is capable of generating a plurality of first modulated signals by modulating each of the plurality of optical signals with a different spread signal sequence for each optical signal using the corresponding spread signal sequence group. Each multiply-accumulate circuit is further capable of generating a composite optical signal by combining the plurality of first modulated signals that have been generated. A plurality of modulation circuits are provided for each multiply-accumulate circuit. Each modulation circuit is capable of generating a specific modulated optical signal by performing specific modulation on the composite optical signal obtained from the corresponding multiply-accumulate circuit. A despreader circuit can generate multiple output signal sequences by calculating the correlation between a composite signal of multiple electrical signals corresponding to multiple specific modulated optical signals and multiple code sequences for each code sequence.

[0008] Figure 1 is a diagram showing an example of the functional block of an optical multiplexing circuit according to the first embodiment of this disclosure. Figure 2 is a diagram showing an example of the cross-sectional configuration of the optical multiplexing circuit of Figure 1. Figure 3 is a diagram showing an example of one output of the multiplication circuit of Figure 1. Figure 4 is a diagram showing an example of the circuit configuration of the sum-of-accumulate circuit of Figure 1. Figure 5 is a diagram showing an example of the cross-sectional configuration near the optical coupler of Figure 4. Figure 6 is a diagram showing an example of the circuit configuration of the modulation circuit of Figure 1. Figure 7 is a diagram showing an example of the circuit configuration of the detection circuit of Figure 1. Figure 8 is a diagram showing an example of each output of the multiplication circuit of Figure 1. Figure 9 is a diagram showing a modified example of each output of the multiplication circuit of Figure 1. Figure 10 is a diagram showing an example of the circuit configuration of the correlation circuit of Figure 1. Figure 11 is a diagram showing an example of correlation processing in the correlation circuit of Figure 10. Figure 12 is a diagram showing an example of the output of the correlation circuit of Figure 10. Figure 13 is a diagram showing an example of the functional block of an optical multiplexing circuit according to the second embodiment of this disclosure. Figure 14 is a diagram showing an example of the cross-sectional configuration of the optical multiplexing circuit of Figure 13. Figure 15 is a diagram showing an example of the functional blocks of the sum-of-accumulate circuit in Figure 13. Figure 16 is a diagram showing an example of the circuit configuration of the sum-of-accumulate circuit in Figure 15. Figure 17 is a diagram showing an example of the cross-sectional configuration near the optical coupler in Figure 16. Figure 18 is a diagram showing an example of the circuit configuration of the diffusion circuit in Figure 13.

[0009] <Background> Artificial intelligence (AI) is primarily composed of neural networks that utilize multiply-accumulate operations. The implementation of neural networks mainly requires integrated circuits such as CPUs (Central Processing Units) and GPUs (Graphics Processing Units). Multiply-accumulate operations are typically performed in CPUs and GPUs using digital multipliers and accumulators. These consist of a large number of registers and electronic logic circuits. Registers and electronic logic circuits are switched on and off simultaneously and in parallel at the frequency of the system clock. Therefore, CPUs and GPUs that implement AI consume enormous amounts of power.

[0010] Therefore, in recent years, optical computing circuits have attracted attention for the purpose of reducing power consumption. Optical computing circuits are composed of optical waveguides and the like. In optical computing circuits, the amplitude and phase of the optical signal are modulated as the optical signal propagates through the optical waveguide, thereby weighting the optical signal. Consequently, optical computing circuits essentially consume no energy. Furthermore, since calculations are performed while the optical signal propagates through the optical waveguide at the speed of light, high-speed calculations are possible.

[0011] For example, Non-Patent Document 1 introduces a neural network using optical computing circuits (optical neural network). Non-Patent Document 1 also describes a comparison between optical neural networks and conventional GPU products. Non-Patent Document 1 also describes the processing power (TOPS / mm²) of optical neural networks per unit area. ２ It is stated that the processing power per unit power (TOPS / W) of the optical neural network, and the processing power of conventional GPU products, far exceeds that of conventional GPU products. Furthermore, Non-Patent Document 1 proposes inputting multiple input signals in parallel. The wavelengths of each of the multiple input signals are different from each other. The multiple input signals are weighted at each node of the crossbar circuit.

[0012] Incidentally, the optical neural network described in Non-Patent Document 1 had the problem that the optical computing circuit was large, making it difficult to reduce the chip area. It is desirable to provide an optical multiplexing circuit that can reduce the size of the optical computing circuit.

[0013] The embodiments for implementing this disclosure will be described in detail below with reference to the drawings. The following description is one specific example of this disclosure and is not limited to the embodiments described below. The description will be given in the following order: 1. First embodiment (Figures 1 to 12) 2. Second embodiment (Figures 13 to 18)

[0014] <1. First Embodiment> [Configuration Example] First, the configuration of the optical multiplexing circuit 1 according to the first embodiment of the present disclosure will be described. Figure 1 shows an example of the functional block of the optical multiplexing circuit 1. Figure 2 shows an example of the cross-sectional configuration of the optical multiplexing circuit 1. The optical multiplexing circuit 1, for example, as shown in equation (1), has multiple (K) input signal sequences x （ｋ） This is a circuit that multiplies each of (1 ≤ k ≤ K) by a scalar quantity α. The optical multiplexing circuit 1 comprises, for example, an optical circuit 10 and an electrical circuit 20, as shown in Figure 1. In this specification, a plurality of (K) input signal sequences x （ｋ） This refers to the input signal sequence x （１） ...x （ｋ） ...x （Ｋ） It refers to.

[0015]

[0016] The optical circuit 10 is provided within the substrate 10A, for example, as shown in Figure 2. The electrical circuit 20 is provided within the substrate 20A, for example, as shown in Figure 2. The substrates 10A and 20A are laminated together by so-called hybrid bonding, for example, as shown in Figure 2. Hybrid bonding refers to the bonding of the substrates 10A and 20A by strong intermolecular bonding between the insulating film (e.g., silicon oxide) on the surface of the substrate 10A and the insulating film (e.g., silicon oxide) on the surface of the substrate 20A, and by metallic bonding between multiple copper (Cu) pads on the surface of the substrate 10A and multiple Cu pads on the surface of the substrate 20A. The multiple Cu pads used in hybrid bonding constitute part of the wiring that electrically connects the optical circuit 10 and the electrical circuit 20.

[0017] The optical circuit 10 includes, for example, a laser light source 11 and a splitter 12 as shown in FIG. 1 and FIG. 2. The laser light source 11 corresponds to a specific example of "light source" according to an embodiment of the present disclosure. The laser light source 11 includes a semiconductor laser capable of emitting a single laser beam L1 having a center wavelength λ. The center wavelength λ is, for example, a wavelength in the near-infrared region. The laser light source 11 may be provided separately from the substrate 10A, for example. The laser light source 11 may be, for example, mounted on a surface of the substrate 20A, and may be capable of irradiating the laser beam L1 onto a side surface of the substrate 10A. The splitter 12 is an optical component that can split the laser beam L1 into a plurality of (K) laser beams L2 without wavelength modulation. The splitter 12 can split the laser beam L1 to generate a plurality of (K) laser beams L2. The laser beam L1 corresponds to a specific example of "initial optical signal" according to an embodiment of the present disclosure. The laser beam L2 corresponds to a specific example of "optical signal" and "split optical signal" according to an embodiment of the present disclosure. The splitter 12 is configured by, for example, an optical coupler. The splitter 12 is configured by, for example, an SOI waveguide on a silicon substrate.

[0018] The optical circuit 10 further includes a product-sum operation circuit 110, for example, as shown in FIG. 1 and FIG. 2. The product-sum operation circuit 110 includes, for example, a modulation circuit 13 and an addition circuit 14 as shown in FIG. 1 and FIG. 2. The plurality of (K) laser beams L2 obtained by the splitter 12 are incident on the modulation circuit 13. The modulation circuit 13 modulates each of the incident plurality of (K) laser beams L2 with a plurality of (K) different spreading sequence codes x^ （ｋ） such that each laser beam L2 is modulated by a different spreading sequence code x^ （ｋ） to obtain a plurality of (K) laser beams L3 （ｋ） that can be generated. In this specification, the plurality of (K) spreading sequence codes x^ （ｋ） refers to spreading sequence codes x^ （１） ...x^ （ｋ） ...x^ （Ｋ） . The modulation circuit 13 can output the generated plurality of (K) laser beams L3 （ｋ） to the addition circuit 14. In this specification, the plurality of (K) laser beams L3 （ｋ） refers to laser beams L3（１） ...L3 （ｋ） ...L3 （Ｋ） This refers to laser light L3. （ｋ） This corresponds to one specific example of the "first modulated signal" according to one embodiment of the present disclosure.

[0019] The modulation circuit 13 is, for example, a plurality (K) of modulation circuits 13 as shown in Figures 1 and 2. （ｋ） It is composed of including the following. In this specification, a plurality (K) modulation circuits 13 （ｋ） This refers to the modulation circuit 13 （１） ...13 （ｋ） ...13 （Ｋ） This refers to each modulation circuit 13 （ｋ） One of the multiple (K) laser beams L2 obtained by the splitter 12 is incident on it. In other words, the laser beam L1 is incident on the multiple (K) modulation circuits 13 by the splitter 12. （ｋ） It will be distributed to them.

[0020] Spreading signal train x^ （ｋ） This is expressed by equation (2). In equation (2), x （ｋ） This is the input signal sequence given to the optical multiplexing circuit 1. In equation (2), x （ｋ） and c （ｋ） The symbol between and indicates the Kronecker product. Input signal sequence x （ｋ） This is represented by the vector shown in equation (3). In equation (3), M is the input signal sequence x （ｋ） This is the number of symbols contained in the vector. T is the transpose of the vector. Below, the input signal sequence x （ｋ） The symbol length of each symbol included is T ｘ This is expressed as follows. In equation (3), m is a positive integer between 1 and M, inclusive. In equation (3), c （ｋ） This is the code sequence required for optical multiplexing. Code sequence c （ｋ） This is represented by the vector shown in equation (4). In equation (4), N is the code sequence c （ｋ） This is the number of symbols contained in the code sequence c. （ｋ） The symbol length of each symbol included is T ｃ This is expressed as follows. In equation (4), n is a positive integer between 1 and N (inclusive). Symbol length T ｃ The symbol length is T.ｘ It is shorter than the symbol length T. ｃ For example, as shown in Figure 3, T ｘ It is equal to / N. Diffusion signal train x^ （ｋ） This is expressed by equation (5). Equation (5) can be represented by a concept such as that shown in Figure 3. In Figure 3, the horizontal axis represents time.

[0021]

[0022] Each modulation circuit 13 （ｋ） The laser light L2 is diffused into a signal train x^ （ｋ） By modulating with, the laser light L3 （ｋ） It is possible to generate each modulation circuit 13 （ｋ） The generated laser light L3 （ｋ） The output can be sent to the summing circuit 14. Each modulation circuit 13 （ｋ） This is, for example, composed of SOI waveguides on a silicon substrate.

[0023] The adding circuit 14 uses multiple (K) laser beams L3 （ｋ） The following can be combined to generate laser light L4. The summing circuit 14 can output the generated laser light L4 to the modulation circuit 15 described later. The laser light L4 corresponds to one specific example of the "synthesized optical signal" according to one embodiment of the present disclosure. The summing circuit 14 is composed of, for example, an optical coupler. The summing circuit 14 is composed of, for example, an SOI waveguide on a silicon substrate.

[0024] Figure 4 shows an example of a planar configuration of the sum-of-products circuit 110. Figure 5 shows an example of a cross-sectional configuration of the optical coupler Cpk and its surroundings, which will be described later, within the sum-of-products circuit 110. The sum-of-products circuit 110 realizes the functions of the modulation circuit 13 and the adder circuit 14 using waveguides. The sum-of-products circuit 110 has, for example, a plurality (K) of first waveguides Pak (1 ≤ k ≤ K), a second waveguide Pb, and a plurality (K) of optical couplers Cpk (1 ≤ k ≤ K), as shown in Figure 4. In this specification, the plurality (K) of first waveguides Pak refers to the first waveguides Pa1...Pak...PaK. In this specification, the plurality (K) of optical couplers Cpk refers to the optical couplers Cp1...Cpk...CpK.

[0025] Multiple (K) first waveguides Pak are provided, one for each laser beam L2. The multiple (K) first waveguides Pak are formed, for example, in the SOI waveguide of the silicon substrate 111. The multiple (K) first waveguides Pak are capable of propagating multiple (K) laser beams L2. The second waveguide Pb is provided at a position adjacent to the multiple (K) first waveguides Pak and at a position intersecting each of the first waveguides Pak. The second waveguide Pb is formed, for example, in the SOI waveguide of the silicon substrate 111.

[0026] Multiple (K) optical couplers Cpk are provided, for example, at each intersection where multiple (K) first waveguides Pak and second waveguides Pb intersect, as shown in Figure 4. Each optical coupler Cpk is formed, for example, in the SOI waveguide of the silicon substrate 111. Multiple (K) optical couplers Cpk have different spreading signal sequences x^ at each of the aforementioned intersections (optical coupler Cpk). （ｋ） A voltage corresponding to the voltage is applied. In each optical coupler Cpk, the spread signal train x^ （ｋ） The corresponding voltage is input from the diffusion circuit 23 described later. Each optical coupler Cpk receives the diffusion signal sequence x^ （ｋ） An electrode is provided to which a voltage corresponding to the voltage is applied. A Cu pad is connected to this electrode. A Cu pad connected to wiring extending from the diffusion circuit 23, described later, is bonded to this Cu pad. In other words, these Cu pads are stacked on top of each other by so-called hybrid bonding.

[0027] The multiply-accumulate circuit 110 may, if necessary, be provided with one electronic integrated circuit 113 for each optical coupler Cpk. Each electronic integrated circuit 113 may, for example, provide a spread signal train x^ （ｋ） It is possible to adjust the timing of applying the appropriate voltage to the optical coupler Cpk. Each electronic integrated circuit 113 may be provided on the silicon substrate 111 via, for example, an interposer 112. In this case, the optical coupler Cpk in the silicon substrate 111 and the wiring extending from the diffusion circuit 23 provided on the interposer 112 are electrically connected, for example, by the hybrid bonding described above. Multiple electronic integrated circuits 113 functionally constitute a part of the diffusion circuit 23. Therefore, multiple electronic integrated circuits 113 may be provided within the substrate 20A.

[0028] Each optical coupler Cpk controls the spread signal train x^ （ｋ） The degree of coupling, determined by the magnitude of the corresponding voltage, allows the laser beam L2 propagating through the first waveguide Pak to be transferred to the second waveguide Pb. The second waveguide Pb can propagate laser beam L4, which is the composite light of multiple laser beams L2 transferred from each first waveguide Pak by each optical coupler Cpk. The laser beam L4 propagated through the second waveguide Pb is output to the modulation circuit 15, which will be described later.

[0029] The optical circuit 10 further includes a modulation circuit 15 and a detection circuit 16, as shown in Figures 1 and 2, for example. The modulation circuit 15 can generate laser light L5 by performing specific modulation on the laser light L4. The modulation circuit 15 can generate laser light L5 by performing amplification processing on the laser light L4 as specific modulation. The laser light L5 corresponds to one specific example of the "specifically modulated optical signal" according to one embodiment of the present disclosure.

[0030] The modulation circuit 15 is configured, for example, by a Mach-Zehnder interferometer. The modulation circuit 15 is configured, for example, as shown in Figure 6, by including waveguides P1 and P2 through which laser light L4 can propagate, and a pair of electrodes E1 and E2 that sandwich waveguide P2. One end of waveguides P1 and P2 is connected to each other, and the other ends of waveguides P1 and P2 are connected to each other. Laser light L2 is incident from one end of waveguides P1 and P2, and the laser light that has propagated through waveguides P1 and P2 interferes with each other at the other ends of waveguides P1 and P2. Laser light L5 is generated by the interference at the other ends of waveguides P1 and P2. Waveguides P1 and P2 are configured, for example, by SOI waveguides on a silicon substrate.

[0031] A fixed voltage (e.g., several volts) higher than that applied to electrode E1 is applied to electrode E2. A fixed voltage (e.g., 0 volts) is applied to electrode E2. In waveguide P1, the voltage applied by electrodes E1 and E2 adjusts the phase of the laser light L4 propagating through waveguide P1 to a predetermined value. Therefore, the laser light L5 is obtained by the interference of laser light with a predetermined phase modulation relative to the laser light L4 and the unphase-modulated laser light L4 propagating through waveguide P2.

[0032] The modulation circuit 15 may be configured by a modulation circuit based on a different principle than phase modulation, for example. The modulation circuit 15 may be configured by on-off keying, for example.

[0033] The detection circuit 16 can convert laser light L5 into an analog electrical signal Va. The detection circuit 16 is configured, for example, to include a photodiode, and when laser light L5 is input to the light-receiving surface of the photodiode, the input laser light L5 can be converted into an electrical signal Va by the photodiode. The detection circuit 16 can output the electrical signal Va obtained by the conversion to the AD converter 24 described later. The detection circuit 16 is configured, for example, as shown in Figure 7, to include a waveguide P3 through which laser light L5 can propagate and a photodiode 16A coupled to the waveguide P3. The waveguide P3 is configured, for example, by an SOI waveguide on a silicon substrate. The photodiode 16A can detect the laser light L5 that has propagated through the waveguide P3. The detection circuit 16 can output the electrical signal Va obtained by detecting the laser light L5 in the photodiode 16A to the AD converter 24. The detection circuit 16 can output the signal shown in equation (6) as the electrical signal Va.

[0034]

[0035] Next, the electrical circuit 20 will be described. The electrical circuit 20 includes, for example, a memory circuit 21, a generation circuit 22, a diffusion circuit 23, an AD converter 24, and a correlation circuit 25, as shown in Figures 1 and 2. The circuit including the detection circuit 16, the AD converter 24, and the correlation circuit 25 corresponds to one specific example of a "reverse diffusion circuit" according to one embodiment of the present disclosure. The generation circuit 22, the diffusion circuit 23, and the correlation circuit 25 are each configured to include, for example, a processor. The diffusion circuit 23 is, for example, a plurality (K) of diffusion circuits 23, as shown in Figure 1. （ｋ） The correlation circuit 25 has, for example, a plurality (K) of correlation circuits 25 as shown in Figure 1. （ｋ） The electrical circuit 20 may optionally have a driver 26 capable of driving the laser light source 11, as shown in Figure 2, for example. In this specification, a plurality (K) of diffusion circuits 23 （ｋ） This refers to the diffusion circuit 23 （１） ...23 （ｋ） ...23 （Ｋ） This refers to a plurality (K) of correlation circuits 25 （ｋ） Correlation circuit 25（１） ...25 （ｋ） ...25 （Ｋ） It refers to.

[0036] The memory circuit 21 stores multiple (K) code sequences c （ｋ） To store. In this specification, multiple (K) code sequences c （ｋ） This refers to the code sequence c （１） ...c （ｋ） ...c （Ｋ） This refers to the code sequence c. （ｋ） These are, for example, pseudorandom sequences, Hadamard codes, or the Walsh function. Pseudorandom sequences include, for example, M sequences and Gold code sequences. Each code sequence c （ｋ） Symbol length T ｃ These are equal to each other. The memory circuit 21 is composed of, for example, non-volatile memory. The memory circuit 21 is composed of, for example, EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, resistive random-access memory, etc.

[0037] The generation circuit 22 receives multiple (K) code sequences c from the memory circuit 21. （ｋ） It is possible to read out the following. The generation circuit 22 reads out multiple (K) code sequences c from the memory circuit 21. （ｋ） The output can be sent to the diffusion circuit 23. The generation circuit 22 has multiple (K) diffusion circuits 23 （ｋ） For each of (described later), multiple (K) code sequences c （ｋ） Using the diffusion circuit 23 （ｋ） Each has a different code sequence c （ｋ） The generation circuit 22 can output the code sequence c. （ｋ） Diffusion circuit 23 （ｋ） It can be output to [this format].

[0038] The diffusion circuit 23 receives multiple (K) input signal sequences x from the optical multiplexing circuit 1. （ｋ） It is possible to accept the following. The diffusion circuit 23 receives multiple (K) input signal sequences x （ｋ） By performing a predetermined calculation on it, multiple (K) spreading signal sequences x^ （ｋ）can be generated. Specifically, the spreading circuit 23 comprises a plurality (K pieces) of input signal sequences x （ｋ） for each of which, a plurality (K pieces) of different code sequences c are used （ｋ） to multiply the input signal sequence x （ｋ） by a different code sequence c for each （ｋ） , whereby a plurality (K pieces) of spread signal sequences x^ （ｋ） can be generated. The spreading circuit 23 can output the generated plurality (K pieces) of spread signal sequences x^ （ｋ） to the product-sum operation circuit 110. The spreading circuit 23 can output the spread signal sequence x^ （ｋ） to the modulation circuit 13 （ｋ） .

[0039] The spreading circuit 23, for example, can adjust the timing and the like for applying voltages corresponding to the plurality (K pieces) of spread signal sequences x^ （ｋ） to a plurality (K pieces) of optical couplers Cpk by using a plurality of electronic integrated circuits 113. The spreading circuit 23, for example, as shown in Fig. 8, can simultaneously output a plurality (K pieces) of spread signal sequences x^ （ｋ） to a plurality (K pieces) of optical couplers Cpk. The spreading circuit 23, for example, as shown in Fig. 9, may be configured to be capable of outputting a plurality (K pieces) of spread signal sequences x^ （ｋ） to a plurality (K pieces) of optical couplers Cpk asynchronously. The spreading circuit 23, for example, as shown in Fig. 9, may delay the output timing of the spread signal sequence x^ （２） by Δt compared to the output timing of another spread signal sequence x^ （１） （１） . This is because the plurality (K pieces) of laser beams L3 modulated by the plurality (K pieces) of spread signal sequences x^ （ｋ） do not interfere with each other, and the plurality (K pieces) of laser beams L3 （ｋ） do not need to be generated at the same time （ｋ） .

[0040] The AD converter 24 can convert an analog electric signal Va input from the detection circuit 16 into a digital electric signal Vd. The AD converter 24 can sample the electric signal Va at a predetermined sampling period. The sampling period of the AD converter 24 is the symbol length T of the code sequence c （ｋ） ​​ｃ Shorter than, for example, T ｃ The formula is / X (where X is a positive integer). The AD converter 24 converts the electrical signal Vd obtained by the conversion to each correlation circuit 25 （ｋ） It can be output to [this format].

[0041] The correlation circuit 25 connects the laser beam L5 with a sequence of multiple (K) code elements c. （ｋ） Correlation with the code sequence c （ｋ） Calculate each one, and thereafter obtain multiple (K) correlation signals αx （ｋ） | c （ｋ） | ２ It is possible to generate multiple (K) correlation signals αx. （ｋ） | c （ｋ） | ２ This refers to the correlation signal αx （１） | c （１） | ２ …αx （ｋ） | c （ｋ） | ２ …αx （Ｋ） | c （Ｋ） | ２ This refers to the correlation circuit 25, which receives multiple (K) correlation signals αx. （ｋ） | c （ｋ） | ２ Multiple (K) output signal trains (electrical signal V ｏｕｔ （ｋ） ) can be output as. In this specification, multiple (K) electrical signals V ｏｕｔ （ｋ） This refers to the electrical signal V ｏｕｔ （１） ...V ｏｕｔ （ｋ） ...V ｏｕｔ （Ｋ） This refers to each correlation circuit 25 （ｋ） This is the laser light L5 and the code sequence c （ｋ） The correlation is calculated, and the correlation signal αx is obtained from this. （ｋ） | c （ｋ） | ２ It is possible to generate the following. The correlation circuit 25 connects the electrical signal Vd obtained from the laser light L5 with a sequence of multiple (K) code elements c （ｋ） Correlation with the code sequence c （ｋ） Calculate each one, and thereafter obtain multiple (K) correlation signals αx （ｋ） | c （ｋ） |２ It is possible to generate each correlation circuit 25 （ｋ） This is the electrical signal Vd and code sequence c obtained from the laser beam L5. （ｋ） The correlation is calculated, and the correlation signal αx is obtained from this. （ｋ） | c （ｋ） | ２ It is possible to generate this.

[0042] Figure 10 shows the correlation circuit 25 （ｋ） This shows an example of the circuit configuration. In Figure 10, the analog electrical signal Va is represented by the code sequence c （ｋ） (c １ （ｋ） ...c ｎ （ｋ） ...c Ｎ （ｋ） It is shown that the signal is obtained from the laser light L5 modulated by ). The analog electrical signal Va is, for example, as shown in Figure 10, Va １ （ｋ） ...Va ｎ （ｋ） ...Va Ｎ （ｋ） Each correlation circuit 25 （ｋ） For example, as shown in Figure 10, register 25A （ｋ） And, register 25B （ｋ） And, multiplication circuit 25C （ｋ） And, the adding circuit 25D （ｋ） It has the following characteristics.

[0043] Register 25A （ｋ） Register 25A can store the electrical signal Vd output from the AD converter 24. （ｋ） Register 25A can sequentially store the b-bit word-length quantized electrical signal Vd output from the AD converter 24, one word at a time. （ｋ） It can store N x X words of data, where N is the code sequence c. （ｋ） The code c included ｎ （ｋ） This is the number. Here, let's assume X is 2. At this time, register 25A （ｋ） It can store 2N words of data (see Figure 10). Register 25B （ｋ） is the code sequence c （ｋ） (c １（ｋ） ...c ｎ （ｋ） ...c Ｎ （ｋ） ) Remember this.

[0044] Multiplication circuit 25C （ｋ） This is register 25A （ｋ） From there, for each sample period of the AD converter 24, N words of data (D １ （ｋ） ...D ｎ （ｋ） ...D Ｎ （ｋ） It is possible to read ) here. １ （ｋ） In the analog electrical signal Va, the code c １ （ｋ） Modulated data Va １ （ｋ） This is digital data. ｎ （ｋ） In the analog electrical signal Va, the code c ｎ （ｋ） Modulated data Va ｎ （ｋ） This is digital data. Ｎ （ｋ） In the analog electrical signal Va, the code c Ｎ （ｋ） Modulated data Va Ｎ （ｋ） It is digital data.

[0045] Multiplication circuit 25C （ｋ） This is register 25A （ｋ） From the N x X word data stored, data is extracted at X word intervals, thereby generating N word data (D １ （ｋ） ...D ｎ （ｋ） ...D Ｎ （ｋ） ) can be generated. Register 25A （ｋ） The order of the N-word data obtained changes with each sample period of the AD converter 24. Register 25A （ｋ）The N-word data obtained changes in order, for example, as time progresses with each sample period of the AD converter 24, as shown in Figure 11.

[0046] For example, at time ta, register 25A （ｋ） The N-word data obtained from D Ｎ－１ （ｋ） D Ｎ （ｋ） ...D ｎ－２ （ｋ） ...D Ｎ－２ （ｋ） For example, at time tb, register 25A （ｋ） The N-word data obtained from D Ｎ （ｋ） D １ （ｋ） ...D ｎ－１ （ｋ） ...D Ｎ－１ （ｋ） For example, at time tc, register 25A （ｋ） The N-word data obtained from D １ （ｋ） D ２ （ｋ） ...D ｎ （ｋ） ...D Ｎ （ｋ） For example, at time td, register 25A （ｋ） The N-word data obtained from D ２ （ｋ） D ３ （ｋ） ...D ｎ＋１ （ｋ） ...D １ （ｋ） That's how it is.

[0047] Multiplication circuit 25C （ｋ） This is register 25B （ｋ） From there, for each sample period of the AD converter 24, N words of data c １ （ｋ） ...c ｎ （ｋ） ...c Ｎ （ｋ） It is possible to read this. Multiplication circuit 25C （ｋ） This is register 25A （ｋ）The N-word data obtained from and register 25B （ｋ） The N words of data obtained from are multiplied by common digits, and the resulting N data are added by the adder 25D （ｋ） Output is possible. Adding circuit 25D （ｋ） This is the multiplication circuit 25C （ｋ） The N data points obtained are added together to form an electrical signal V. ｏｕｔ （ｋ） It is possible to output the following. Adding circuit 25D （ｋ） This is an electrical signal V ｏｕｔ （ｋ） Therefore, the signal shown in equation (7) can be output. Accordingly, the correlation circuit 25 can output multiple (K) input signal sequences x （ｋ） The signal obtained by multiplying by a scalar quantity α is a sequence of multiple (K) output signals (multiple (K) correlated signals αx （ｋ） | c （ｋ） | ２ It is possible to output ).

[0048]

[0049] Register 25A （ｋ） The N-word data obtained from this is called data D. Register 25B （ｋ） The N-word data obtained from the source is called data C. When the correlation between data D and data C is large, the correlation circuit 25 （ｋ） The output will be maximized. A large correlation between data D and data C means, specifically, that the modulation factor (code c) in data D is large. ｎ （ｋ） The order of the elements is the code c of data C. ｎ （ｋ） This refers to matching the order of the data. When the correlation between data D and data C is small, the correlation circuit 25 （ｋ） The output will be minimized. A small correlation between data D and data C means that, specifically, the modulation factor (code c) in data D is small. ｎ （ｋ） The order of the elements is the code c of data C. ｎ （ｋ） This refers to a discrepancy in the order of the elements. (Adding circuit 25D) （ｋ）The output is maximized at time tc, for example, when the correlation between data D and data C is high, as shown in Figure 12.

[0050] [Effects] Next, the effects of the optical multiplexing circuit 1 will be explained.

[0051] In this embodiment, multiple (K) input signal sequences x （ｋ） For each of these, a number of distinct code sequences c (K) （ｋ） Using the input signal sequence x （ｋ） Each has a different code sequence c （ｋ） By multiplying by x^, multiple (K) spreading signal sequences x^ （ｋ） This is generated. Each of the multiple (K) laser beams L2 generates multiple (K) diffuse signal trains x^ （ｋ） Using this method, a different diffusion signal train x^ for each laser beam L2 （ｋ） Multiple (K) laser beams L3 are modulated by this. （ｋ） Multiple (K) laser beams L3 are generated. （ｋ） By combining these, laser light L4 is generated. By applying specific modulation to laser light L4, laser light L5 is generated. Laser light L5 and multiple (K) code sequences c （ｋ） The correlation with the code sequence c （ｋ） Each is calculated, thereby generating multiple (K) electrical signals V ｏｕｔ （ｋ） This is generated.

[0052] Thus, in this embodiment, multiple (K) code sequences c （ｋ） Using this, multiple (K) input signal sequences x （ｋ） The signals are optically multiplexed, and the resulting optical signal (laser light L4) is multiplied by a scalar quantity α. Then, from the optical signal (laser light L5) obtained by multiplying the optical signal (laser light L4) by the scalar quantity α, multiple (K) code sequences c are generated. （ｋ） Using this, multiple (K) output signal trains (multiple (K) electrical signals V ｏｕｔ （ｋ） ) is obtained. Multiple (K) output signal sequences are obtained from multiple (K) input signal sequences x （ｋ） This corresponds to the signal obtained by multiplying the input signal sequence x by a scalar quantity α.（ｋ） By performing optical calculations separately for each signal, the size of the optical calculation circuit can be reduced compared to obtaining multiple (K) output signal trains, and the amount of calculation performed by the optical calculation circuit can also be reduced. Therefore, in this embodiment, the size of the optical calculation circuit can be reduced, and the amount of calculation performed by the optical calculation circuit can also be reduced.

[0053] In this embodiment, the sum-of-accumulate circuit 110 is configured to include a plurality of first waveguides Pak, a second waveguide Pab, and a plurality of optical couplers Cpk. As a result, a plurality (K) of code sequences c are connected to the plurality of optical couplers Cpk. （ｋ） Multiple (K) spreading signal sequences x^ obtained using （ｋ） The coupling ratio between multiple first waveguides Pak and second waveguides Pab can be controlled by the corresponding voltage. As a result, multiple (K) input signal trains x can be connected in a simple configuration. （ｋ） Optical multiplexing can be achieved. Therefore, the size of the optical computing circuit can be reduced, and furthermore, the amount of computation performed by the optical computing circuit can be reduced.

[0054] In this embodiment, multiple laser beams L2 input to the sum-of-accumulate circuit 110 are generated by a laser light source 11 capable of generating a single laser beam L1 and a splitter 12. This allows for a smaller light source compared to using a sum-of-accumulate circuit with wavelength division multiplexing, and further reduces the power consumption required for the light source.

[0055] In this embodiment, a circuit including a detection circuit 16, an AD converter 24, and a correlation circuit 25 generates multiple (K) output signal trains (multiple (K) correlation signals αx （ｋ） | c （ｋ） | ２ This yields the result. This makes it possible to perform multiplication of a scalar quantity α via an optical processing circuit (multiply-accumulate circuit 110).

[0056] In this embodiment, the modulation circuit 15 included in the optical circuit 10 performs multiplication by a scalar quantity α. As a result, the input signal train x （ｋ）Compared to the case where each scalar quantity α is multiplied separately, the size of the optical computing circuit can be reduced, and furthermore, the computation amount performed by the optical computing circuit can be reduced. Therefore, in this embodiment, the size of the optical computing circuit can be reduced, and furthermore, the computation amount performed by the optical computing circuit can be reduced.

[0057] In this invention, the code sequence c （ｋ） Pseudorandom sequences, Hadamard codes, or Walsh functions are used as the multiplication method. This enables multiplication of scalar quantities α via an optical processing circuit (multiply-accumulate circuit 110).

[0058] In this invention, each code sequence c （ｋ） Symbol length T ｃ This is each input signal sequence x （ｋ） Symbol length T ｘ It is shorter than the code sequence c. （ｋ） Symbol length T ｃ For example, as shown in Figure 3, T ｘ It is equal to / N. This results in multiple (K) spreading signal sequences x^ （ｋ） It can be easily generated.

[0059] In this embodiment, a circuit including a multiply-accumulate circuit 110, a modulation circuit 15, and a detection circuit 16 is provided in the optical circuit 10. A circuit including a diffusion circuit 23, an AD converter 24, and a correlation circuit 25 is provided in the electrical circuit 20. Thus, in this embodiment, the optical multiplexing circuit 1 is divided into an optical circuit 10 and an electrical circuit 20. As a result, for example, the optical multiplexing circuit 1 can be realized by stacking a substrate 10A including the optical circuit 10 and a substrate 20A including the electrical circuit 20 on each other by hybrid bonding (metal bonding). As a result, the optical multiplexing circuit 1 can be formed in a small and simple configuration. Furthermore, the reliability of the wiring between the optical circuit 10 and the electrical circuit 20 can be improved.

[0060] <2. Second Embodiment> Next, the configuration of the optical multiplexing circuit 2 according to the second embodiment of the present disclosure will be described. In the following, explanations of configurations and effects common to the first embodiment will be omitted as appropriate. Figure 13 shows an example of the functional block of the optical multiplexing circuit 2. Figure 14 shows an example of the cross-sectional configuration of the optical multiplexing circuit 2. The optical multiplexing circuit 2, for example, as shown in equation (8), assigns a vector quantity α to each of the multiple (K) input signal matrices Mtx(k). １ …α ｊ …α Ｊ This is a circuit that multiplies by . In this specification, multiple (K) input signal matrices Mtx(k) refer to input signal matrices Mtx(1)...Mtx(k)...Mtx(K). In other words, the optical multiplexing circuit 2 is a circuit capable of performing a sum-of-products operation between a K×J matrix A and a J×1 vector B, for example, as shown in equation (9). Matrix A and vector B will be described in detail later. The optical multiplexing circuit 2 comprises an optical circuit 30 and an electrical circuit 40, for example, as shown in Figure 13.

[0061]

[0062] The optical circuit 30 is provided within the substrate 30A, for example, as shown in Figure 14. The electrical circuit 40 is provided within the substrate 40A, for example, as shown in Figure 14. The substrates 30A and 40A are laminated together by so-called hybrid bonding, for example, as shown in Figure 14. Hybrid bonding refers to the bonding of the substrates 30A and 40A by strong intermolecular bonding between the insulating film (e.g., silicon oxide) on the surface of the substrate 30A and the insulating film (e.g., silicon oxide) on the surface of the substrate 40A, and by metallic bonding between multiple copper (Cu) pads on the surface of the substrate 30A and multiple Cu pads on the surface of the substrate 40A. The multiple Cu pads used in hybrid bonding constitute part of the wiring that electrically connects the optical circuit 30 and the electrical circuit 40.

[0063] The optical circuit 30 includes, for example, a laser light source 11, splitters 12 and 17, a multiply-accumulate circuit 110X, a modulation circuit 15, and a detection circuit 16, as shown in Figures 13 and 14. Splitter 12 corresponds to a specific example of the "first splitter" according to one embodiment of the present disclosure.

[0064] The splitter 17 has a plurality (J) of splitters 17-j (1 ≤ j ≤ J). In this specification, the plurality (J) of splitters 17-j refers to splitters 17-1...17-k...17-K. The splitter 17-j corresponds to a specific example of the "second splitter" according to one embodiment of the present disclosure. Each splitter 17-j is an optical component capable of splitting laser light L2 into a plurality (K) of laser light L17 without wavelength modulation, as shown in Figure 15, for example. The laser light L17 corresponds to a specific example of the "optical signal" according to one embodiment of the present disclosure. Each splitter 17-j is capable of splitting laser light L2 to generate a plurality (K) of laser light L17. The splitter 17 is configured, for example, by an optical coupler. The splitter 17 is configured, for example, by an SOI waveguide on a silicon substrate.

[0065] The sum-of-accumulate circuit 110X has, for example, a plurality (J) of sum-of-accumulate circuits 110-j, as shown in Figure 13. In this specification, the plurality (J) of sum-of-accumulate circuits 110-j refers to sum-of-accumulate circuits 110-1...110-k...110-K. Each sum-of-accumulate circuit 110-j is configured to include a multiple modulation circuit 13-j and an adder circuit 14-j, as shown in Figure 13, for example.

[0066] Multiple (K) laser beams L17 obtained by the splitter 17 are incident on the multiple modulation circuit 13-j. The multiple modulation circuit 13-j processes each of the incident (K) laser beams L17 into multiple (K) distinct diffuse signal trains x ｊ ^ （ｋ） Using a laser beam L17, a different diffusion signal train x ｊ ^ （ｋ） By modulating with this, multiple (K) laser beams L13(j) （ｋ） It is possible to generate a plurality (K) spreading signal train x ｊ ^（ｋ） This refers to the spreading signal train x ｊ ^ （１） ...x ｊ ^ （ｋ） ...x ｊ ^ （Ｋ） This refers to the spreading signal train x. ｊ ^ （１） ...x ｊ ^ （ｋ） ...x ｊ ^ （Ｋ） These are appropriately referred to as the spread signal train group SG^(j) (see Figure 18 below). Multiple (J) sum-of-accumulate circuits 110-j are provided for each spread signal train group SG^(j). In this specification, multiple (K) laser beams L13(j) （ｋ） This refers to the laser light L13(j) （１） ...L13(j) （ｋ） ...L13(j) （Ｋ） This refers to the multiple modulation circuit 13-j which generates multiple (K) laser beams L13(j) （ｋ） This can be output to the adder circuit 14-j. Laser light L13(j) （ｋ） This corresponds to one specific example of the "first modulated signal" according to one embodiment of the present disclosure.

[0067] The multiple modulation circuit 13-j is, for example, a plurality (K) of modulation circuits 13-j as shown in Figure 15. （ｋ） It is composed of including the following. In this specification, a plurality (K) modulation circuits 13-j （ｋ） This refers to the modulation circuit 13-j （１） ...13-j （ｋ） ...13-j （Ｋ） This refers to each modulation circuit 13-j （ｋ） One of the multiple (K) laser beams L17 obtained by the splitter 17 is incident on it. In other words, the laser beam L2 is incident on the multiple (K) modulation circuits 13-j （ｋ） It will be distributed to them.

[0068] Spreading signal train x ｊ ^ （ｋ） This is expressed by equation (10). In equation (10), x ｊ （ｋ） This is the input signal sequence provided to the optical multiplexing circuit 2. In equation (10), x ｊ （ｋ）and c （ｋ） The symbol between and indicates the Kronecker product. Input signal sequence x ｊ （ｋ） This is represented by the vector shown in equation (11). In equation (11), M is the input signal sequence x ｊ （ｋ） This is the number of symbols included in the input signal sequence x. ｊ （ｋ） The symbol length of each symbol included is T ｘ This is expressed as follows. In equation (11), m is a positive integer between 1 and M, inclusive. In equation (10), c （ｋ） This is the code sequence required for optical multiplexing. Code sequence c （ｋ） Symbol length T ｃ The input signal sequence x ｊ （ｋ） Symbol length T ｘ It is shorter than the symbol length T. ｃ For example, T ｘ It is equal to / N. Diffusion signal train x ｊ ^ （ｋ） This is expressed by equation (12).

[0069]

[0070] Each modulation circuit 13-j （ｋ） The laser light L17 is diffused into a signal train x ｊ ^ （ｋ） By modulating with, the laser light L13(j) （ｋ） It is possible to generate each modulation circuit 13-j （ｋ） The generated laser light L13(j) （ｋ） This can be output to the summing circuit 14-j. Each modulation circuit 13-j （ｋ） This is, for example, composed of SOI waveguides on a silicon substrate.

[0071] The adding circuit 14-j uses multiple (K) laser beams L13(j) （ｋ）The summing circuit 14-j can synthesize these signals, thereby generating laser light L14(j). The summing circuit 14-j can output the generated laser light L14(j) to the modulation circuit 15-j described later. The laser light L14(j) corresponds to a specific example of the "synthesized optical signal" according to one embodiment of the present disclosure. The summing circuit 14-j is configured, for example, by an optical coupler. The summing circuit 14-j is configured, for example, by an SOI waveguide on a silicon substrate.

[0072] Figure 16 shows an example of a planar configuration of the sum-of-accumulate circuit 110-j. Figure 17 shows an example of a cross-sectional configuration of the optical coupler Cpk and its surroundings within the sum-of-accumulate circuit 110-j. The sum-of-accumulate circuit 110-j is a modulation circuit 13-j （ｋ） The functions of the summing circuit 14-j are realized using waveguides. The sum-of-products operation circuit 110-j has, for example, a plurality (K) of first waveguides Pak, a second waveguide Pb, and a plurality (K) of optical couplers Cpk, as shown in Figure 16.

[0073] Multiple (K) first waveguides Pak are provided, one for each laser beam L17. The multiple (K) first waveguides Pak are formed, for example, in the SOI waveguide of the silicon substrate 111. The multiple (K) first waveguides Pak are capable of propagating multiple (K) laser beams L17. The second waveguide Pb is provided at a position adjacent to the multiple (K) first waveguides Pak and at a position intersecting each of the first waveguides Pak. The second waveguide Pb is formed, for example, in the SOI waveguide of the silicon substrate 111.

[0074] Multiple (K) optical couplers Cpk are provided, for example, at each intersection where (K) first waveguides Pak and second waveguides Pb intersect, as shown in Figure 16. Each optical coupler Cpk is formed, for example, in the SOI waveguide of the silicon substrate 111. Multiple (K) optical couplers Cpk have different spreading signal sequences x at each of the aforementioned intersections (optical coupler Cpk). ｊ ^ （ｋ） A voltage corresponding to the value is applied. In each optical coupler Cpk, the spread signal train x ｊ ^ （ｋ） The corresponding voltage is input from the diffusion circuit 28 described later. Each optical coupler Cpk receives the diffusion signal train xｊ ^ （ｋ） An electrode is provided to which a voltage corresponding to the voltage is applied. A Cu pad is connected to this electrode. A Cu pad connected to wiring extending from the diffusion circuit 28, described later, is bonded to this Cu pad. In other words, these Cu pads are stacked on top of each other by so-called hybrid bonding.

[0075] The multiply-accumulate circuit 110-j may, if necessary, be provided with one electronic integrated circuit 113 for each optical coupler Cpk. Each electronic integrated circuit 113 may, for example, provide a spread signal train x ｊ ^ （ｋ） It is possible to adjust the timing of applying the appropriate voltage to the optical coupler Cpk. Each electronic integrated circuit 113 may be provided on the silicon substrate 111 via, for example, an interposer 112. In this case, the optical coupler Cpk in the silicon substrate 111 and the wiring extending from the diffusion circuit 28 provided on the interposer 112 are electrically connected, for example, by the hybrid bonding described above. Multiple electronic integrated circuits 113 functionally constitute a part of the diffusion circuit 28. Therefore, multiple electronic integrated circuits 113 may be provided within the substrate 40A.

[0076] Each optical coupler Cpk controls the spread signal train x ｊ ^ （ｋ） The degree of coupling, determined by the voltage magnitude, allows the laser light L2 propagating through the first waveguide Pak to be transferred to the second waveguide Pb. The second waveguide Pb is capable of propagating the laser light L14(j), which is the composite light of multiple laser beams L17 transferred from each first waveguide Pak by each optical coupler Cpk. The laser light L14(j) propagating through the second waveguide Pb is output to the modulation circuit 15-j.

[0077] The modulation circuit 15 has a plurality (J) of modulation circuits 15-j, for example, as shown in Figure 13. In this specification, the plurality (J) of modulation circuits 15-j refers to modulation circuits 15-1...15-k...15-K. One of the plurality (J) of modulation circuits 15-j is provided for each multiply-accumulate operation circuit 110-j. The modulation circuits 15-j are provided in correspondence with the multiply-accumulate operation circuit 110-j. Each modulation circuit 15-j is capable of generating laser light L15(j) by performing a specific modulation on the laser light L14(j). Each modulation circuit 15-j is capable of generating laser light L15(j) by performing an amplification process on the laser light L14(j) as a specific modulation. The gain of each modulation circuit 15-j is represented by αj. The gains αj of each modulation circuit 15-j may be set to a common value or to be set independently of each other. The laser beam L15(j) corresponds to a specific example of the "specifically modulated optical signal" according to one embodiment of the present disclosure. Each modulation circuit 15-j is configured, for example, by a Mach-Zehnder interferometer. Each modulation circuit 15-j may be configured, for example, by a modulation circuit based on a different principle than phase modulation. Each modulation circuit 15-j may be configured, for example, by on-off keying.

[0078] The detection circuit 16 has, for example, a plurality (J) of detection circuits 16-j, as shown in Figure 13. In this specification, the plurality (J) of detection circuits 16-j refers to detection circuits 16-1...16-k...16-K. One detection circuit 16-j is provided for each laser beam L15(j). Each detection circuit 16-j is capable of converting the laser beam L15(j) into an analog electrical signal Va(j). Each detection circuit 16-j is configured, for example, to include a photodiode, and when the laser beam L15(j) is input to the light-receiving surface of the photodiode, the input laser beam L15(j) can be converted into an electrical signal Va(j) by the photodiode. Each detection circuit 16-j can output the electrical signal Va(j) obtained by the conversion to the AD converter 24-j described later. Each detection circuit 16-j has a configuration similar to, for example, the detection circuit 16 shown in Figure 7. Each detection circuit 16-j can output the signal shown in equation (13) as an electrical signal Va(j).

[0079]

[0080] Next, the electrical circuit 40 will be described. The electrical circuit 20 includes, for example, a memory circuit 21, a generation circuit 27, a diffusion circuit 28, an AD converter 24, a correlation circuit 25, and an adder circuit 29, as shown in Figures 13 and 14. The circuit including the detection circuit 16, the AD converter 24, and the correlation circuit 25 corresponds to one specific example of a "de-spreading circuit" according to one embodiment of the present disclosure. The generation circuit 27, the diffusion circuit 28, and the correlation circuit 25 are each configured to include, for example, a processor. The AD converter 24 has, for example, a plurality (J) of AD converters 24-j. The correlation circuit 25 has, for example, a plurality (K) of correlation circuits 25 （ｋ） The diffusion circuit 28 has, for example, a plurality (J) of diffusion circuits 28 （ｊ） It has. In this specification, multiple (J) AD converters 24-j refer to AD converters 24-1...24-k...24-K. In this specification, multiple (K) correlation circuits 25 （ｋ） This refers to the correlation circuit 25 （１） ...25 （ｋ） ...25 （Ｋ） This refers to a plurality (J) of diffusion circuits 28. （ｊ） This refers to the diffusion circuit 28 （１） ...28 （ｋ） ...28 （Ｋ） This refers to the electrical circuit 20, which may optionally have a driver 26 capable of driving the laser light source 11, as shown in Figure 14, for example.

[0081] The generation circuit 27 receives multiple (K) code sequences c from the memory circuit 21. （ｋ） It is possible to read out the following. The generation circuit 27 reads out multiple (K) code sequences c from the memory circuit 21. （ｋ） The output can be sent to the diffusion circuit 28. The generation circuit 27 can send multiple (J) diffusion circuits 28 （ｊ） For each of these, multiple (K) code sequences c （ｋ） It is possible to output this.

[0082] The diffusion circuit 28 can accept multiple (K) input signal matrices Mtx(k) that are given to the optical multiplexing circuit 2, for example, as shown in Figure 18. The input signal matrix Mtx(k) is expressed by equation (14). The multiple (K) input signal matrices Mtx(k) are given to the diffusion circuit 28 in a time-sequential manner. The diffusion circuit 28 as a whole can accept matrix A, for example, shown in equation (15).

[0083]

[0084] Each diffusion circuit 28 （ｊ） The input signal sequence group SG(j), which is one of the multiple (J) row components of matrix A, is input to this. The input signal sequence group SG(j) consists of multiple (K) input signal sequences x ｊ （ｋ） It is composed of and specifically represented by formula (16). In this specification, a plurality of (K) input signal sequences x ｊ （ｋ） This refers to the input signal sequence x ｊ （１） ...x ｊ （ｋ） ...x ｊ （Ｋ） It refers to.

[0085]

[0086] The diffusion circuit 23 can generate multiple (J) diffusion signal sequence groups SG^(j) by performing predetermined arithmetic processing on the multiple (J) input signal sequence groups SG(j) that it has received. In this specification, multiple (J) input signal sequence groups SG(j) refer to input signal sequence groups SG(1)...SG(k)...SG(K). Each diffusion signal sequence group SG^(j) is a plurality of (K) diffusion signal sequences x ｊ ^ （ｋ） It is composed of the above and is specifically represented by formula (17).

[0087]

[0088] Each diffusion circuit 28 （ｊ） x is a sequence of multiple (K) input signals. ｊ （ｋ）For each of these, a number of distinct code sequences c (K) （ｋ） Using the input signal sequence x ｊ （ｋ） Each has a different code sequence c （ｋ） By multiplying by this, multiple (K) spreading signal sequences x ｊ ^ （ｋ） It is possible to generate each diffusion circuit 28 （ｊ） This is the generated spread spectrum x of multiple (K) signals. ｊ ^ （ｋ） The output can be sent to the sum-of-accumulate circuit 110-j. Each diffusion circuit 28 （ｊ） This is the spreading signal train x ｊ ^ （ｋ） modulation circuit 13-j of sum-of-accumulate circuit 110-j （ｋ） It can be output to [this format].

[0089] The AD converter 24 can convert multiple (J) analog electrical signals Va(j) input from the detection circuit 16 into multiple (J) digital electrical signals Vd(j) (1 ≤ j ≤ J). In this specification, multiple (J) electrical signals Vd(j) refer to electrical signals Vd(1)...Vd(k)...Vd(k). The AD converter 24 can sample each electrical signal Va(j) at a predetermined sampling period. The sampling period of the AD converter 24 is given by the code sequence c （ｋ） Symbol length T ｃ Shorter than, for example, T ｃ The formula is / X (where X is a positive integer). The AD converter 24 can output multiple (J) digital electrical signals Vd(j) obtained by the conversion to the adder circuit 29. Each AD converter 24-j can convert the analog electrical signal Va(j) input from the detection circuit 16-j into a digital electrical signal Vd(j). Each AD converter 24-j can output the electrical signal Vd(j) obtained by the conversion to the adder circuit 29.

[0090] The summing circuit 29 adds up multiple (J) electrical signals Vd(j) obtained from the AD converter 24, and the resulting electrical signal Vd is then connected to each correlation circuit 25 （ｋ） It can be output to the following. The electrical signal Vd corresponds to one specific example of a "composite signal" according to one embodiment of the present disclosure.

[0091] The correlation circuit 25 connects an electrical signal Vd with a sequence of multiple (K) code elements c. （ｋ） Correlation with the code sequence c （ｋ） Calculate each one, and thereafter obtain multiple (K) correlation signals (α １ x １ （ｋ） …α ｊ x ｊ （ｋ） …α Ｊ x Ｊ （ｋ） )|c （ｋ） | ２ It is possible to generate multiple (K) correlation signals (α １ x １ （ｋ） …α ｊ x ｊ （ｋ） …α Ｊ x Ｊ （ｋ） )|c （ｋ） | ２ This refers to the correlation signal (α １ x １ （１） …α ｊ x ｊ （１） …α Ｊ x Ｊ （１） )|c （１） | ２ …(α １ x １ （ｋ） …α ｊ x ｊ （ｋ） …α Ｊ x Ｊ （ｋ） )|c （ｋ） | ２ …(α １ x １ （Ｋ） …α ｊ x ｊ （Ｋ） …α Ｊ x Ｊ （Ｋ） )|c （Ｋ） | ２ This refers to the correlation circuit 25, which generates multiple (K) correlation signals (α １ x １ （ｋ） …α ｊ xｊ （ｋ） …α Ｊ x Ｊ （ｋ） )|c （ｋ） | ２ Multiple (K) electrical signals V ｏｕｔ （ｋ） It can be output as (Equation (18)). Each correlation circuit 25 （ｋ） This consists of an electrical signal Vd and a code sequence c. （ｋ） The correlation is calculated and the correlation signal (α) is obtained from this. １ x １ （ｋ） …α ｊ x ｊ （ｋ） …α Ｊ x Ｊ （ｋ） )|c （ｋ） | ２ It is possible to generate each correlation circuit 25 （ｋ） For example, the circuit configuration is as shown in Figure 10.

[0092]

[0093] [Effects] Next, we will explain the effects of the optical multiplexing circuit 2.

[0094] In this embodiment, for each of the multiple (J) input signal sequence groups SG(j), multiple (K) distinct code sequences c are provided. （ｋ） By multiplying by this, a set of multiple (J) diffuse signal trains SG^(j) is generated. Each of the multiple (K) laser beams L17 uses the diffuse signal train train set SG^(j) to generate a different diffuse signal train x for each laser beam L17. ｊ ^ （ｋ） By being modulated, multiple (K) laser beams L3(j) （ｋ） Multiple (K) laser beams L3(j) are generated. （ｋ） By combining these, laser light L14(j) is generated. By applying specific modulation to laser light L14(j), laser light L15(j) is generated. The electrical signal Vd is a composite signal of multiple electrical signals Vd(j) corresponding to laser light L15(j), and the code sequence c consists of multiple (K) code sequences. （ｋ） The correlation with the code sequence c （ｋ）Each is calculated, thereby generating multiple (K) electrical signals V ｏｕｔ （ｋ） This is generated.

[0095] Thus, in this embodiment, multiple (K) code sequences c （ｋ） Using this, multiple (K) input signal matrices Mtx(k) are optically multiplexed, and the optical signal obtained by optical multiplexing (laser light L14(j)) is subjected to a vector quantity α １ …α ｊ …α Ｊ This is multiplied. Then, the vector quantity α is applied to the optical signal (laser light L14(j)). １ …α ｊ …α Ｊ From the optical signal (laser light L15(j)) obtained by multiplication, a sequence of multiple (K) code elements c （ｋ） Using this, multiple (K) output signal trains (multiple (K) electrical signals V ｏｕｔ （ｋ） ) is obtained. The multiple (K) output signal sequences are a vector quantity α for the multiple (K) input signal matrices Mtx(k). １ …α ｊ …α Ｊ This corresponds to the signal obtained by multiplication. As a result, the size of the optical processing circuit can be reduced, and the amount of computation performed by the optical processing circuit can be reduced, compared to the case where multiple (K) output signal sequences are obtained by performing optical calculations separately for each input signal matrix Mtx(k). Therefore, in this embodiment, the size of the optical processing circuit can be reduced, and the amount of computation performed by the optical processing circuit can be reduced.

[0096] In this embodiment, each sum-of-accumulate circuit 110-j is configured to include a plurality of first waveguides Pak, a second waveguide Pab, and a plurality of optical couplers Cpk. As a result, a plurality (K) of code sequences c are connected to the plurality of optical couplers Cpk. （ｋ） Multiple (K) spreading signal sequences x obtained using ｊ ^ （ｋ）The coupling ratio between multiple first waveguides Pak and second waveguides Pab can be controlled by the corresponding voltage. As a result, optical multiplexing of multiple (K) input signal matrices Mtx(k) can be realized with a simple configuration. Therefore, the size of the optical computing circuit can be reduced, and furthermore, the amount of computation performed by the optical computing circuit can be reduced.

[0097] In this embodiment, the multiple laser beams L17 input to each sum-of-accumulate circuit 110-j are generated by a laser light source 11 capable of generating a single laser beam L1 and two splitters 12 and 17. This allows for a smaller light source compared to using a sum-of-accumulate circuit with wavelength division multiplexing, and further reduces the power consumption required for the light source.

[0098] In this embodiment, a circuit including a detection circuit 16, an AD converter 24, and a correlation circuit 25 generates multiple (K) output signal trains (multiple (K) correlated signals (α １ x １ （ｋ） …α ｊ x ｊ （ｋ） …α Ｊ x Ｊ （ｋ） )|c （ｋ） | ２ This yields the vector quantity α via the optical computing circuit (each multiply-accumulate circuit 110-j). １ …α ｊ …α Ｊ Multiplication can be achieved.

[0099] In this embodiment, the modulation circuit 15 included in the optical circuit 30 modulates the vector quantity α １ …α ｊ …α Ｊ Multiplication is performed. This results in a separate vector quantity α for each input signal matrix Mtx(k). １ …α ｊ …α ＪCompared to the case where multiplication is performed, the size of the optical computing circuit can be reduced, and furthermore, the amount of computation performed by the optical computing circuit can be reduced. Therefore, in this embodiment, the size of the optical computing circuit can be reduced, and furthermore, the amount of computation performed by the optical computing circuit can be reduced.

[0100] In this embodiment, a circuit including a multiply-accumulate circuit 110X, a modulation circuit 15, and a detection circuit 16 is provided in the optical circuit 30. A circuit including a diffusion circuit 28, an AD converter 24, and a correlation circuit 25 is provided in the electrical circuit 40. Thus, in this embodiment, the optical multiplexing circuit 2 is divided into an optical circuit 30 and an electrical circuit 40. As a result, for example, the optical multiplexing circuit 2 can be realized by stacking a substrate 30A including the optical circuit 30 and a substrate 40A including the electrical circuit 40 on each other by hybrid bonding (metal bonding). Consequently, the optical multiplexing circuit 2 can be formed with a simple configuration.

[0101] Although this technology has been described above with reference to multiple embodiments and application examples, this technology is not limited to these embodiments, and various modifications are possible. Furthermore, the effects described herein are merely illustrative and not limiting, and other effects may also exist.

[0102] Furthermore, this technology can be configured as follows: (1) Multiple (K) input signal trains (x （ｋ） For each of the above, a set of (K) distinct code sequences (c （ｋ） ) using the input signal sequence (x （ｋ） ) each different code sequence (c （ｋ） By multiplying by ), multiple (K) spreading signal trains (x^ （ｋ） A diffusion circuit (23) capable of generating ) and each of the multiple (K) optical signals (L2) is connected to the multiple (K) diffusion signal train (x^ （ｋ） ) is used to create a different spreading signal sequence (x^) for each of the optical signals (L2). （ｋ）A sum-of-accumulate circuit (110) capable of generating multiple (K) first modulated signals (L3) by modulating with (c), and generating a combined optical signal (L4) by combining the generated multiple (K) first modulated signals (L3), a modulation circuit (15) capable of generating a specific modulated optical signal (L5) by performing specific modulation on the combined optical signal (L4), the specific modulated optical signal (L5), and the multiple (K) code sequences (c （ｋ） The correlation with the above code sequence (c （ｋ） By calculating each of these, multiple (K) output signal trains (V) can be generated. ｏｕｔ （ｋ） (2) An optical multiplexing circuit comprising a despreading circuit (16, 24, 25) capable of generating a multiply （ｋ） (1) The optical multiplexing circuit described in (1), comprising a plurality of optical couplers (Cpk) to which a voltage corresponding to ) is applied. (3) The optical multiplexing circuit described in (1) or (2), further comprising a light source (11) capable of generating an initial optical signal (L1), and a splitter (12) capable of branching the initial optical signal (L1) to generate the plurality of optical signals (L2). (4) The despreading circuit (16, 24, 25) comprises a detection circuit (16) capable of converting the specific modulated optical signal (L5) into an electrical signal (Va), the electrical signal (Va), and the plurality of code sequences (c （ｋ） The correlation with the above code sequence (c （ｋ） The plurality of output signal trains (V) are calculated for each of them. ｏｕｔ （ｋ）An optical multiplexing circuit according to any one of (1) to (3), comprising a correlation circuit (24, 25) capable of generating a specific modulated optical signal (L5). (5) An optical multiplexing circuit according to any one of (1) to (4), wherein the modulation circuit (15) is capable of generating the specific modulated optical signal (L5) by performing amplification processing on the composite optical signal (L4) as the specific modulation. (6) The diffusion circuit (23) comprises the code sequence (c （ｋ） (1) to (5) an optical multiplexing circuit can use a pseudorandom sequence, an Hadamard code, or a Walsh function as (7) each of the code sequences (c （ｋ） The symbol length (T) of each symbol included in ) ｃ ) is the input signal sequence (x （ｋ） The symbol length (T) of each symbol included in ) ｘ (8) Each of the code sequences (c （ｋ） The symbol length (T) of each symbol included in ) ｃ ) is the input signal sequence (x （ｋ） The symbol length (T) of each symbol included in ) ｘ ) 1 / (the aforementioned code sequence (c （ｋ） (7) The optical multiplexing circuit described in (9) The circuit including the sum-of-accumulate circuit (110), the modulation circuit (15), and the detection circuit (16) is composed of an optical circuit (10A), and the circuit including the diffusion circuit (23) and the correlation circuits (24, 25) is composed of an electrical circuit (20A), as described in (4). (10) The optical circuit (10A) and the electrical circuit (20A) are stacked on top of each other by hybrid bonding, as described in (9). (11) Each of the input signal trains (x) is equal to a plurality of (K) input signal trains (x ｊ （ｋ） For each of the multiple (J) sets of input signal sequences (SG(j)) including ), a multiple (K) set of distinct code sequences (c （ｋ） ) is multiplied, and thereafter each is a spread signal train of multiple (K) (x ｊ ^ （ｋ）A diffusion circuit (28) capable of generating multiple (J) sets of diffusion signal trains (SG^(j)) including ), and a different diffusion signal train (x) for each of the multiple (K) optical signals (L17) provided for each diffusion signal train (SG^(j)), using the corresponding diffusion signal train (SG^(j)). ｊ ^ （ｋ） By modulating with ), multiple (K) first modulated signals (L13(j) （ｋ） ) generates the multiple (K) first modulated signals (L13(j) （ｋ） A plurality of (J) multiply-accumulate circuits (110-j) capable of generating a composite optical signal (L14(j)) by combining the following: A plurality of (J) modulation circuits (15-j) provided for each of the multiply-accumulate circuits (110-j), capable of generating a specific modulated optical signal (L15(j)) by performing specific modulation on the composite optical signal (L14(j)) obtained from the corresponding multiply-accumulate circuit (110-j): A composite signal (Vd) of a plurality of (J) electrical signals (Vd(j)) corresponding to the plurality of (J) specific modulated optical signals (L15(j)), and a plurality of (K) code sequences (c （ｋ） The correlation with the above code sequence (c （ｋ） By calculating each of these, multiple (K) output signal trains (V) can be generated. ｏｕｔ （ｋ） An optical multiplexing circuit comprising a despreading circuit (16, 24, 25) capable of generating a multiply-accumulate circuit (110-j), each of the optical signals (L17) is provided with a plurality of first waveguides (Pak) capable of propagating the plurality of (K) optical signals (L17), a second waveguide (Pb) provided at a position adjacent to the plurality of (K) first waveguides (Pak) and at a position intersecting each of the first waveguides (Pak) capable of propagating the combined optical signal (L14(j)), and a different spread signal sequence (x^) provided at each intersection point where the plurality of (K) first waveguides (Pak) and the second waveguides (Pb) intersect each other. （ｋ）The optical multiplexing circuit according to (11), comprising a plurality of optical couplers (Cpk) to which a voltage corresponding to ) is applied. (13) The optical multiplexing circuit according to (11) or (12), further comprising: a light source (11) capable of generating an initial optical signal (L1); a first splitter (12) capable of splitting the initial optical signal (L1) to generate a plurality of branched optical signals (L2); and a second splitter (17) provided for each of the sum-of-accumulate circuits (110-j) capable of splitting the branched optical signals (L2) to generate the plurality of optical signals (L17). (14) The despreading circuit (16, 24, 25) is provided for each specific modulated optical signal (L15(j)) and comprises a plurality of detection circuits (16-j) capable of converting the corresponding specific modulated optical signal (L15(j)) into an electrical signal (Vd(j)), a composite signal (Vd) which is a composite signal of the plurality of electrical signals (Vd(j)) obtained by the plurality of detection circuits (16-j), and the plurality of code sequences (c （ｋ） The correlation with the above code sequence (c （ｋ） The plurality of output signal trains (V) are calculated for each of them. ｏｕｔ （ｋ） An optical multiplexing circuit according to any one of (11) to (13), further comprising correlation circuits (24, 25) capable of generating (15) a specific modulated optical signal (L15(j)) as the specific modulation, wherein each of the modulation circuits (15-j) is capable of generating the specific modulated optical signal (L15(j)) by performing amplification processing on the composite optical signal (Vd), as the specific modulation, as an optical multiplexing circuit according to any one of (11) to (14). (16) An optical multiplexing circuit according to (14), wherein the circuit comprising the plurality of (J) multiply-accumulate circuits (110-j), the plurality of (J) modulation circuits (15-j), and the plurality of detection circuits (16-j) is composed of an optical circuit (30A), and the circuit comprising the diffusion circuit (28) and the correlation circuits (24, 25) is composed of an electrical circuit (40A). (17) The optical circuit (30A) and the electrical circuit (40A) are stacked on top of each other by hybrid bonding, as described in (16).

[0103] In this disclosure, multiple spread signal sequences are generated by multiplying each of multiple input signal sequences by a different code sequence for each input signal sequence using multiple different code sequences. Multiple first modulated signals are generated by modulating each of the multiple optical signals with a different spread signal sequence for each optical signal using the multiple spread signal sequences. A composite optical signal is generated by combining the generated multiple first modulated signals. A specific modulated optical signal is generated by performing specific modulation on the composite optical signal. The correlation between the specific modulated optical signal and the multiple code sequences is calculated for each code sequence, thereby generating multiple output signal sequences.

[0104] Thus, in this disclosure, multiple input signal sequences are optically multiplexed using multiple code sequences, and specific modulation is performed on the optical signals obtained by optical multiplexing. Then, from the specifically modulated optical signals obtained by performing specific modulation on the optical signals, multiple output signal sequences (multiple correlated signals) corresponding to the signals obtained by performing specific modulation on multiple input signal sequences using multiple code sequences are obtained. As a result, the size of the optical computing circuit can be reduced, and the amount of computation performed by the optical computing circuit can be reduced, compared to the case where multiple output signal sequences (multiple correlated signals) are obtained by performing optical calculations separately for each input signal sequence.

[0105] In this disclosure, multiple sets of spread signal sequences are generated by multiplying each of multiple sets of input signal sequences by multiple different code sequences. Multiple first modulated signals are generated by modulating each of the multiple optical signals with a different spread signal sequence using the spread signal sequence. A composite optical signal is generated by combining the multiple first modulated signals generated. A specific modulated optical signal is generated by performing specific modulation on the composite optical signal. The correlation between the composite signal of multiple electrical signals corresponding to the specific modulated optical signal and multiple code sequences is calculated for each code sequence, thereby generating multiple electrical signals.

[0106] Thus, in this disclosure, multiple sets of input signal sequences are optically multiplexed using multiple code sequences, and specific modulation is performed on the optical signals obtained by optical multiplexing. Then, multiple output signal sequences (multiple correlated signals) are obtained from the specifically modulated optical signals obtained by performing specific modulation on the optical signals, using multiple code sequences. The multiple output signal sequences correspond to signals obtained by performing specific modulation on multiple sets of input signal sequences. As a result, the size of the optical computing circuit can be reduced, and the computation amount performed by the optical computing circuit can be reduced, compared to the case where multiple output signal sequences are obtained by performing optical calculations separately for each set of input signal sequences.

[0107] This application claims priority based on U.S. Provisional Application No. 63 / 779923, filed with the U.S. Patent and Trademark Office on 28 March 2025, and all the contents of that application are incorporated herein by reference.

[0108] Those skilled in the art will understand that various modifications, combinations, subcombinations, and changes can be conceived depending on design requirements and other factors, and that these fall within the scope of the attached claims and their equivalents.

Claims

1. An optical multiplexing circuit comprising: a diffusion circuit capable of generating multiple diffusion signal sequences by multiplying each of multiple input signal sequences by a different set of code sequences for each of the input signal sequences; a multiply-accumulate circuit capable of generating multiple first modulated signals by modulating each of multiple optical signals with a different diffusion signal sequence for each of the optical signals using the multiple diffusion signal sequences, and generating a composite optical signal by combining the generated multiple first modulated signals; a modulation circuit capable of generating a specific modulated optical signal by performing specific modulation on the composite optical signal; and a despreader circuit capable of generating multiple output signal sequences by calculating the correlation between the specific modulated optical signal and the multiple code sequences for each of the code sequences.

2. The optical multiplexing circuit according to claim 1, comprising: a plurality of first waveguides provided for each optical signal and capable of propagating the plurality of optical signals; a second waveguide provided at a position adjacent to the plurality of first waveguides and at a position intersecting each of the first waveguides, and capable of propagating the combined optical signal; and a plurality of optical couplers provided at each intersection point where the plurality of first waveguides and the second waveguides intersect each other, to which a voltage corresponding to the different diffusion signal sequences is applied at each intersection point.

3. The optical multiplexing circuit according to claim 1, further comprising a light source capable of generating an initial optical signal, and a splitter capable of branching the initial optical signal to generate the plurality of optical signals.

4. The optical multiplexing circuit according to claim 1, wherein the despreading circuit includes a detection circuit capable of converting the specific modulated optical signal into an electrical signal, and a correlation circuit capable of generating the plurality of output signal sequences by calculating the correlation between the electrical signal and the plurality of code sequences for each code sequence.

5. The optical multiplexing circuit according to claim 1, wherein the modulation circuit is capable of generating the specific modulated optical signal by performing amplification processing on the composite optical signal as the specific modulation.

6. The optical multiplexing circuit according to claim 1, wherein the diffusion circuit is capable of using a pseudo-random sequence, an Hadamard code, or a Walsh function as the code sequence.

7. The optical multiplexing circuit according to claim 1, wherein the symbol length of each symbol included in each of the code sequences is shorter than the symbol length of each symbol included in the input signal sequence.

8. The optical multiplexing circuit according to claim 7, wherein the symbol length of each symbol included in each of the code sequences is equal to 1 / (number of code sequences) of the symbol length of each symbol included in the input signal sequence.

9. The optical multiplexing circuit according to claim 4, wherein the circuit including the sum-of-products circuit, the modulation circuit and the detection circuit is an optical circuit, and the circuit including the diffusion circuit and the correlation circuit is an electrical circuit.

10. The optical multiplexing circuit according to claim 9, wherein the optical circuit and the electrical circuit are stacked on top of each other by hybrid bonding.

11. An optical multiplexing circuit comprising: a diffusion circuit capable of multiplying each of a plurality of sets of input signal sequences, each containing a plurality of input signal sequences, by a plurality of different code sequences, thereby generating a plurality of sets of diffusion signal sequences, each containing a plurality of diffusion signal sequences; a plurality of multiply-accumulate circuits provided for each set of diffusion signal sequences, capable of generating a plurality of first modulated signals by modulating each of a plurality of optical signals with a diffusion signal sequence different for each optical signal using the corresponding set of diffusion signal sequences, and generating a composite optical signal by combining the generated plurality of first modulated signals; a plurality of modulation circuits provided for each multiply-accumulate circuit, capable of generating a specific modulated optical signal by performing specific modulation on the composite optical signal obtained from the corresponding multiply-accumulate circuit; and a despreader circuit capable of generating a plurality of output signal sequences by calculating the correlation between a composite signal of a plurality of electrical signals corresponding to the plurality of specific modulated optical signals and the plurality of code sequences for each code sequence.

12. The optical multiplexing circuit according to claim 11, wherein each of the sum-of-accumulate circuits comprises: a plurality of first waveguides provided for each optical signal and capable of propagating the plurality of optical signals; a second waveguide provided at a position adjacent to the plurality of first waveguides and at a position intersecting each of the first waveguides, and capable of propagating the combined optical signal; and a plurality of optical couplers provided at each intersection point where the plurality of first waveguides and the second waveguides intersect each other, to which a voltage corresponding to the different diffusion signal sequences is applied at each intersection point.

13. The optical multiplexing circuit according to claim 11, further comprising: a light source capable of generating an initial optical signal; a first splitter capable of branching the initial optical signal to generate a plurality of branched optical signals; and a second splitter provided for each of the sum-of-accumulate circuits, capable of branching the branched optical signals to generate the plurality of optical signals.

14. The optical multiplexing circuit according to claim 11, wherein the despreading circuit includes a plurality of detection circuits provided for each specific modulated optical signal and capable of converting the corresponding specific modulated optical signal into an electrical signal, and a correlation circuit capable of generating a plurality of output signal sequences by calculating the correlation between the combined signal, which is a combined signal of the plurality of electrical signals obtained by the plurality of detection circuits, and the plurality of code sequences for each of the code sequences.

15. The optical multiplexing circuit according to claim 11, wherein each of the modulation circuits is capable of generating the specific modulated optical signal by performing amplification processing on the composite optical signal as the specific modulation.

16. The optical multiplexing circuit according to claim 14, wherein the circuit including the plurality of multiply-accumulate circuits, the plurality of modulation circuits, and the plurality of detection circuits is an optical circuit, and the circuit including the diffusion circuit and the correlation circuit is an electrical circuit.

17. The optical multiplexing circuit according to claim 16, wherein the optical circuit and the electrical circuit are stacked on top of each other by hybrid bonding.

Images