WO2023168629A1

WO2023168629A1 - Optical computing system and optical signal processing method

Info

Publication number: WO2023168629A1
Application number: PCT/CN2022/079960
Authority: WO
Inventors: 陶明慧; 周雷; 欧阳伟龙; 李芮
Original assignee: 华为技术有限公司
Priority date: 2022-03-09
Filing date: 2022-03-09
Publication date: 2023-09-14

Abstract

The present application provides an optical computing system and an optical signal processing method. The digital-to-analog conversion of a high-bit digital signal is completed by means of the cooperation of a plurality of digital-to-analog converters (DACs), so that the whole optical computing system can realize optical operation based on an 8-bit digital signal and a digital signal of 8 bits or more. The optical computing system of the present application comprises: a first DAC, a second DAC and an optical processing unit (OPU), wherein after the first DAC and the second DAC cooperatively receive a digital signal from an electric processor, the first DAC can convert a first effective bit of the digital signal into a first analog signal, and the second DAC can convert a second effective bit of the digital signal into a second analog signal. Then, the OPU can convert the first analog signal into a first optical signal, convert the second analog signal into a second optical signal, and perform operation on the first optical signal and the second optical signal respectively to obtain a third optical signal and a fourth optical signal. Finally, the OPU can amplify and superimpose the third optical signal and the fourth optical signal to obtain a computing result of the digital signal.

Description

An optical computing system and optical signal processing method

Technical field

The present application relates to the field of optical communications, and in particular, to an optical computing system and an optical signal processing method.

Background technique

Photon computing chip can also be called optical processing unit (OPU). OPU uses optical signals as the carrier of data, so OPU can process data by performing operations on optical signals. The optical computing system based on OPU is considered to be the most promising solution in various fields such as artificial intelligence and mobile communications that require big data processing due to its advantages such as high-speed parallelism and low power consumption.

Currently, optical computing systems can include analog-to-analog converter (DAC), electro-optical conversion (E/O), OPU, optical-electro conversion (O/E) and Analog to digital converter (ADC). After the DAC receives the digital signal to be processed, it can be converted into an analog signal and sent to the E/O. The E/O can then convert the analog signal into an optical signal and send it to the OPU. Then, the OPU can perform operations on the optical signal to obtain the calculation result of the digital signal in the optical field. Finally, the calculation result of the digital signal in the optical field can be converted into the calculation result of the digital signal in the electrical field through O/E and ADC. At this point, the optical calculation of the digital signal is completed.

However, the DAC is limited by its own design process and materials, and its digital signal processing accuracy is no higher than 4 bits (4 bits). As a result, the current optical computing system cannot realize optical operations of digital signals of 8 bits (8 bits) and above. .

Contents of the invention

Embodiments of the present application provide an optical computing system and an optical signal processing method that use multiple DACs to collaboratively complete digital-to-analog conversion of high-bit digital signals, so that the entire optical computing system can achieve digital signal conversion of 8-bit or more. Light operations.

The first aspect of the embodiment of the present application provides an optical computing system. The optical computing system includes: multiple DACs and OPUs. The input end of each DAC is connected to the output end of an electric processor outside the system. The input end of each DAC is The output terminal is connected to the input terminal of the OPU.

When the electronic processor outputs a digital signal to be operated, since the digital signal can be split by bits, the digital signal can be split into multiple sub-digital signals. Among the multiple sub-digital signals, each sub-digital signal corresponds to at least one bit of the digital signal. Therefore, after the multiple sub-digital signals are spliced in order, the digital signal can be composed. For example, suppose there is a digital signal representing 409 X, the digital signal X is 110011001. Since the digital signal .

In this way, the electronic processor can send the digital signal to multiple DACs in the form of multiple sub-digital signals. It should be noted that the first DAC and the second DAC are two of the plurality of DACs, then the first DAC can receive the first sub-digital signal (the first sub-digital signal) corresponding to the first valid bit of the digital signal. The sub-digital signal is one of the aforementioned plurality of sub-digital signals), and the second DAC can receive the second sub-digital signal corresponding to the second effective bit of the digital signal (the second sub-digital signal is one of the aforementioned plurality of sub-digital signals). of another).

Then, the first DAC can perform digital-to-analog conversion on the first sub-digital signal it receives to obtain the first analog signal, which is equivalent to the first DAC converting the first effective bit of the digital signal into the first analog signal and sending to OPU. Similarly, the second DAC can perform digital-to-analog conversion on the second sub-digital signal it receives to obtain a second analog signal, which is equivalent to the second DAC converting the second effective bit of the digital signal into a second analog signal, and Send to OPU.

After receiving the first analog signal and the second analog signal, the OPU can perform electro-optical conversion on the first analog signal to obtain the first optical signal, and perform electro-optical conversion on the second analog signal to obtain the second optical signal. Then, the OPU performs calculations based on the first optical signal to obtain a third optical signal, and performs calculations based on the second optical signal to obtain a fourth optical signal. Finally, the OPU can amplify and superimpose the third optical signal and the fourth optical signal to obtain the calculation result of the digital signal.

The above optical computing system includes a first DAC, a second DAC and an OPU. After the first DAC and the second DAC cooperate to receive the digital signal from the electrical processor, the first DAC can convert the first effective bit of the digital signal into The first analog signal, the second DAC can convert the second significant bit of the digital signal into a second analog signal. Then, the OPU can convert the first analog signal into a first optical signal, convert the second analog signal into a second optical signal, and perform operations on the first optical signal and the second optical signal respectively to obtain the third optical signal and the third optical signal. Four light signals. Finally, the OPU can amplify and superimpose the third optical signal and the fourth optical signal to obtain the calculation result of the digital signal. In the aforementioned process, since the accuracy of the DAC is usually 4 bits, if the number of digital signals is greater than 4 bits, the digital signal can be split into multiple parts and received by multiple DACs to achieve digital-to-analog conversion. In this way, even if the accuracy of each DAC itself is limited, multiple DACs can cooperate to complete the digital-to-analog conversion of high-bit digital signals, so that the entire optical computing system can realize optical operations of 8-bit and above digital signals.

In a possible implementation, the accuracy of the first DAC and the accuracy of the second DAC are d bits, the digital signal contains r bits, and the first effective bit is the (i-1)*d+1th bit to The i*d bit, the second significant bit is the (j-1)*d+1 to j*d bit of the digital signal, where r=h*d, r, h and d are all greater than zero Integer, i=1,…,h,j=1,…,h,i≠j. In the aforementioned implementation, when the number of bits of the digital signal is divisible by the accuracy of the DAC (d bits), the first significant bit and the second significant bit are both d bits, and the first significant bit is the (i-th) of the digital signal. -1)*d+1 to the i*d bit, the second significant bit is the (j-1)*d+1 to j*d bit of the digital signal, for example, the first significant bit is The 1st to dth bits of the digital signal, the second significant bit is the d+1th to 2dth bit of the digital signal, and for example, the first significant bit is the (h-2)*dth bit of the digital signal +1 bit to the (h-1)*d bit, the second significant bit is the (h-1)*d+1 to h*d bit of the digital signal, and so on.

In a possible implementation, the accuracy of the first DAC and the accuracy of the second DAC are d bits, the digital signal contains r bits, and the first effective bit is the (i-1)*d+1th bit to The i*d bit, the second significant bit is the h*d+1 to r-th bit of the digital signal, where r=h*d+v, r, h, d and v are all integers greater than zero, i=1,…,h. In the aforementioned implementation, when the number of bits of the digital signal is not divisible by the accuracy of the DAC (d bits), the first effective bit is the d bit, and the second effective bit is the v bit, and the first effective bit is the digital signal. The (i-1)*d+1th bit to the i*dth bit, the second significant bit is the h*d+1th to r-th bit of the digital signal, for example, the first significant bit is the digital signal's The 1st to dth bits, the second significant bits are the h*d+1 to rth bits of the digital signal. Another example, the first significant bit is the (h-2)*d+1th bit of the digital signal. bit to the (h-1)*d bit, the second significant bit is the h*d+1 bit to the r-th bit of the digital signal, and so on.

In a possible implementation, the third optical signal is amplified by a factor greater than or equal to 2 to the power of (i-1)*d, and the fourth optical signal is amplified by a factor greater than or equal to 2(j-1) *d power. In the aforementioned implementation, since the third optical signal is generated based on the first sub-digital signal, and the first sub-digital signal corresponds to the first effective bit of the digital signal, that is, the (i-1)*d+1th bit of the digital signal bit to the i*dth bit, so the third optical signal needs to be amplified by a factor greater than or equal to 2 to the (i-1)*d power. Similarly, since the fourth optical signal is generated based on the second sub-digital signal, and the second sub-digital signal corresponds to the second effective bit of the digital signal, that is, the (j-1)*d+1th bit to The j*dth bit, so the amplification factor of the fourth optical signal is greater than or equal to 2 to the (j-1)*d power.

In a possible implementation, the third optical signal is amplified by a factor greater than or equal to 2 to the (i-1)*d power, and the fourth optical signal is amplified by a factor greater than or equal to 2 h*d power. . In the aforementioned implementation, since the third optical signal is generated based on the first sub-digital signal, and the first sub-digital signal corresponds to the first effective bit of the digital signal, that is, the (i-1)*d+1th bit of the digital signal bit to the i*dth bit, so the third optical signal needs to be amplified by a factor greater than or equal to 2 to the (i-1)*d power. Similarly, since the fourth optical signal is generated based on the second sub-digital signal, and the second sub-digital signal corresponds to the second effective bit of the digital signal, that is, the h*d+1th to rth bits of the digital signal, Therefore, the amplification factor of the fourth optical signal is greater than or equal to 2 h*d power.

In a possible implementation, the OPU includes a first modulator, a second modulator, an amplifier and an adder; the first modulator is used to convert the first analog signal into a first optical signal, and The signal is operated to obtain a third optical signal; the second modulator is used to convert the second analog signal into a second optical signal, and performs operations based on the second optical signal to obtain a fourth optical signal; the amplifier is used to The third optical signal is amplified to obtain the fifth optical signal; the fourth optical signal is amplified to obtain the sixth optical signal; the adder is used to superimpose the fifth optical signal and the sixth optical signal to obtain the calculation of the digital signal result. In the aforementioned implementation, the accuracy of the modulator in the OPU is usually 4 bits. By arranging multiple modulators in the OPU, even if the accuracy of each modulator itself is limited, high-bit digital signal operations can also be implemented. In this way, the optical computing system provided by the embodiments of the present application can implement high-precision optical computing of digital signals by deploying low-precision devices, which is beneficial to improving the computing performance of optical computing without any performance loss compared to electrical computing.

In a possible implementation, the first modulator and the second modulator are microring modulators, electroabsorption external modulators or Mach-Zehnder interferometers.

In a possible implementation, the aforementioned operation is a matrix multiplication operation.

A second aspect of the embodiment of the present application provides an optical signal processing method. The method is implemented through an optical computing system. The system includes a first DAC, a second DAC, and an OPU. The method includes: processing the third portion of the digital signal through the first DAC. One valid bit is converted into a first analog signal; the second valid bit of the digital signal is converted into a second analog signal through the second DAC; the OPU performs an operation based on the first optical signal derived from the first analog signal to obtain a third optical signal signal; the OPU performs calculations based on the second optical signal derived from the second analog signal to obtain a fourth optical signal; the OPU amplifies and superimposes the third optical signal and the fourth optical signal to obtain a calculation result of the digital signal.

The optical computing system used to implement the above method includes a first DAC, a second DAC and an OPU. After the first DAC and the second DAC cooperate to receive the digital signal from the electrical processor, the first DAC can convert the third part of the digital signal. A valid bit is converted into a first analog signal, and the second DAC can convert a second valid bit of the digital signal into a second analog signal. Then, the OPU can convert the first analog signal into a first optical signal, convert the second analog signal into a second optical signal, and perform operations on the first optical signal and the second optical signal respectively to obtain the third optical signal and the third optical signal. Four light signals. Finally, the OPU can amplify and superimpose the third optical signal and the fourth optical signal to obtain the calculation result of the digital signal. In the aforementioned process, since the accuracy of the DAC is usually 4 bits, if the number of digital signals is greater than 4 bits, the digital signal can be split into multiple parts and received by multiple DACs to achieve digital-to-analog conversion. In this way, even if the accuracy of each DAC itself is limited, multiple DACs can cooperate to complete the digital-to-analog conversion of high-bit digital signals, so that the entire optical computing system can realize optical operations of 8-bit and above digital signals.

In a possible implementation, the accuracy of the first DAC and the accuracy of the second DAC are d bits, the digital signal contains r bits, and the first effective bit is the (i-1)*d+1th bit to The i*d bit, the second significant bit is the (j-1)*d+1 to j*d bit of the digital signal, where r=h*d, r, h and d are all greater than zero Integer, i=1,…,h,j=1,…,h,i≠j.

In a possible implementation, the accuracy of the first DAC and the accuracy of the second DAC are d bits, the digital signal contains r bits, and the first effective bit is the (i-1)*d+1th bit to The i*d bit, the second significant bit is the h*d+1 to r-th bit of the digital signal, where r=h*d+v, r, h, d and v are all integers greater than zero, i=1,…,h.

In a possible implementation, the third optical signal is amplified by a factor greater than or equal to 2 to the power of (i-1)*d, and the fourth optical signal is amplified by a factor greater than or equal to 2(j-1) *d power.

In a possible implementation, the third optical signal is amplified by a factor greater than or equal to 2 to the (i-1)*d power, and the fourth optical signal is amplified by a factor greater than or equal to 2 h*d power. .

In a possible implementation, the OPU includes a first modulator, a second modulator, an amplifier and an adder, and the steps performed by the OPU include: converting the first analog signal into a first optical signal through the first modulator, and perform operations based on the first optical signal to obtain a third optical signal; convert the second analog signal into a second optical signal through the second modulator, and perform operations based on the second optical signal to obtain a fourth optical signal; through the amplifier The third optical signal is amplified to obtain the fifth optical signal; the fourth optical signal is amplified to obtain the sixth optical signal; the fifth optical signal and the sixth optical signal are superimposed by an adder to obtain the calculation result of the digital signal .

The optical computing system provided by the embodiment of the present application includes a first DAC, a second DAC and an OPU. After the first DAC and the second DAC cooperate to receive the digital signal from the electrical processor, the first DAC can convert the third part of the digital signal. A valid bit is converted into a first analog signal, and the second DAC can convert a second valid bit of the digital signal into a second analog signal. Then, the OPU can convert the first analog signal into a first optical signal, convert the second analog signal into a second optical signal, and perform operations on the first optical signal and the second optical signal respectively to obtain the third optical signal and the third optical signal. Four light signals. Finally, the OPU can amplify and superimpose the third optical signal and the fourth optical signal to obtain the calculation result of the digital signal. In the aforementioned process, since the accuracy of the DAC is usually 4 bits, if the number of digital signals is greater than 4 bits, the digital signal can be split into multiple parts and received by multiple DACs to achieve digital-to-analog conversion. In this way, even if the accuracy of each DAC itself is limited, multiple DACs can cooperate to complete the digital-to-analog conversion of high-bit digital signals, so that the entire optical computing system can realize optical operations of 8-bit and above digital signals.

Description of the drawings

Figure 1 is a schematic diagram of a modulator in the related art;

Figure 2 is a schematic structural diagram of an optical computing system provided by an embodiment of the present application;

Figure 3 is another structural schematic diagram of an optical computing system provided by an embodiment of the present application;

Figure 4 is a schematic diagram of an application example of the optical computing system provided by the embodiment of the present application;

Figure 5 is a schematic diagram of another application example of the optical computing system provided by the embodiment of the present application;

FIG. 6 is a schematic flowchart of an optical signal processing method provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application.

The terms "first", "second", etc. in the description and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the terms so used are interchangeable under appropriate circumstances, and are merely a way of distinguishing objects with the same attributes in describing the embodiments of the present application. Furthermore, the terms "include" and "having" and any variations thereof, are intended to cover non-exclusive inclusion, such that a process, method, system, product or apparatus comprising a series of elements need not be limited to those elements, but may include any other elements specifically listed or inherent to such processes, methods, products or equipment.

Photon computing chip can also be called OPU. OPU uses optical signals as the carrier of data, so OPU can process data by performing operations on optical signals. The optical computing system based on OPU is considered to be the most promising solution in various fields such as artificial intelligence and mobile communications that require big data processing due to its advantages such as high-speed parallelism and low power consumption.

Currently, the optical computing system can include devices such as DAC, E/O, OPU, O/E and ADC. It should be noted that the current E/O and O/E can be integrated in the OPU. For example, the OPU can be presented as a Modulator (for example, the modulator can include a microring modulator and a photodetector, and the output end of the microring modulator is connected to the input end of the photodetector). The modulator can realize both the electro-optical conversion function and the photoelectric conversion function. Can realize light computing function. Specifically, after the DAC receives the digital signal to be operated, it can be converted into an analog signal and sent to the modulator. Then, the modulator can convert the analog signal into an optical signal, and perform calculations on the optical signal to obtain the calculation result of the digital signal in the optical field. Then, the modulator can convert the calculation result of the digital signal in the optical domain into the calculation result of the digital signal in the electrical domain, and send it to the ADC (the ADC can perform analog-to-digital conversion on the result and output it to subsequent digital operations. unit for further signal processing). At this point, the optical calculation of the digital signal is completed.

The following is a brief introduction to the modulator workflow with an example. For example, suppose you need to implement the multiplication operation of two 4×4 matrices, that is, matrix A multiplied by matrix B. The precision of the elements in matrix A and matrix B is 4 bits (4bit), that is, each element in matrix A and matrix B is a 4-bit binary number, so it can be presented as a 4-bit digital signal. Matrix A and matrix B can be expressed as formula (1):

Then, it can be realized by a micro-ring modulator as shown in Figure 1 (Figure 1 is a schematic diagram of a modulator in the related art. It should be noted that the superscript "4" of an element indicates that the precision of the element is 4 bits). Multiplication of matrix A and matrix B. Specifically, at the current moment, the digital signal a ₁₁ representing the value 1, the digital signal a ₁₂ representing the value 2, the digital signal a ₁₃ representing the value 3 and the digital signal representing the value 4 can be input to the four DACs on the left respectively. a ₁₄ , and these four digital signals are all 4-bit (4bit) digital signals. Then, the four DACs can respectively convert the four digital signals into analog signals corresponding to a ₁₁ , an analog signal corresponding to a ₁₂ , an analog signal corresponding to a ₁₃ , and an analog signal corresponding to a ₁₄ . Then, these four analog signals can be respectively modulated on optical signals with corresponding wavelengths to obtain an optical signal corresponding to a ₁₁ (the wavelength of the optical signal is λ1), and an optical signal corresponding to a ₁₂ (the wavelength of the optical signal is λ1). is λ2), the optical signal corresponding to a ₁₃ (the wavelength of the optical signal is λ3) and the optical signal corresponding to a ₁₄ (the wavelength of the optical signal is λ4), and these four optical signals can be input to the first The four microrings of the column are respectively input to the four microrings of the second column, the four microrings of the third column are respectively input, and the four microrings of the fourth column are respectively input. Similarly, the digital signal b ₁₁ representing the value 5, the digital signal b ₂₁ representing the value 6, the digital signal b ₃₁ representing the value 7 and the digital signal b representing the value 8 can be respectively input to the four DACs in the first column on the right. ₄₁ , and these four digital signals are all 4-bit (4bit) digital signals. Then, the four DACs can respectively convert the four digital signals into an analog signal corresponding to b ₁₁ , an analog signal corresponding to b ₂₁ , an analog signal corresponding to b ₃₁ , and an analog signal corresponding to b ₄₁ . Then, these four analog signals are input to the four microrings in the first column respectively, so the analog signal corresponding to b ₁₁ can be modulated on the optical signal corresponding to a ₁₁ to obtain the optical signal corresponding to a ₁₁ × b ₁₁ , The analog signal corresponding to b ₂₁ can be modulated on the optical signal corresponding to a ₁₂ to obtain an optical signal corresponding to a ₁₂ × b _21. The analog signal corresponding to b ₃₁ can be modulated on the optical signal corresponding to a ₁₃ . An optical signal corresponding to a ₁₃ × b ₃₁ is obtained. The analog signal corresponding to b ₄₁ can be modulated on the optical signal corresponding to a ₁₄ to obtain an optical signal corresponding to a ₁₄ × b ₄₁ . By accumulating these four optical signals, the optical signal corresponding to c ₁₁ can be obtained, c ₁₁ =a ₁₁ ×b ₁₁ +a ₁₂ ×b ₂₁ +a ₁₃ ×b ₃₁ +a ₁₄ ×b ₄₁ . In the same way, the optical signal corresponding to c ₁₂ can also be obtained, c ₁₂ = a ₁₁ × b ₁₂ + a ₁₂ × b ₂₂ + a ₁₃ × b ₃₂ + a ₁₄ × b ₄₂ , and the optical signal corresponding to c ₁₂ , c ₁₃ =a ₁₁ ×b ₁₃ +a ₁₂ ×b ₂₃ +a ₁₃ × _{b 33} ₊ a 14 ×b ₄₃ , and the optical signal corresponding to c ₁₄ , c ₁₁ =a ₁₁ ×b ₁₄ +a ₁₂ ×b ₂₄ + a ₁₃ ×b ₃₄ +a ₁₄ ×b ₄₄ .

At the next moment, the inputs of the four DACs on the left are modified to digital signal a ₂₁ , digital signal a ₂₂ , digital signal a ₂₃ and digital signal a ₂₄ , and these four digital signals are all 4-bit (4bit) digital signals. , the inputs of the 16 DACs on the right remain unchanged, so the optical signal corresponding to c ₂₁ , the optical signal corresponding to c ₂₂ , the optical signal corresponding to c ₂₃ and the optical signal corresponding to c ₂₄ can be obtained. By analogy, until an optical signal corresponding to the entire matrix C is obtained, the matrix C is shown in formula (2):

Of course, after obtaining the optical signal corresponding to the entire matrix C, the photodetector (not shown in Figure 1) connected to the microring modulator can also convert the optical signal into an analog signal corresponding to the entire matrix C. And output to ADC for subsequent signal processing. At this point, the microring modulator has completed the matrix multiplication operation between matrix A and matrix B in the optical field, that is, it has realized the optical operation between the two matrices.

However, DACs and modulators are limited by their own design processes and materials, and their signal processing accuracy is no higher than 4 bits (for example, DACs can only accurately achieve digital-to-analog conversion of 4-bit digital signals, and modulators can only accurately achieve digital-to-analog conversion of 4-bit digital signals. Can accurately realize electro-optical conversion of 4-bit analog signals, etc.), resulting in the current optical computing system being unable to realize optical operations of 8-bit (8-bit) and above digital signals.

In order to solve the above problems, embodiments of the present application provide a new optical computing system, as shown in Figure 2 (Fig. 2 is a structural schematic diagram of the optical computing system provided by the embodiment of the present application). The optical computing system includes: A DAC and an OPU, the input terminal of each DAC is connected to the output terminal of the electronic processor outside the system, and the output terminal of each DAC is connected to the input terminal of the OPU.

The electronic processor can output a digital signal representing a certain value, and the digital signal can be received by multiple DACs in cooperation. Specifically, since the digital signal contains multiple bits (that is, the digital signal is a multi-bit digital signal), the digital signal can be split by bits, thereby splitting the digital signal into multiple sub-digital signals. It can be seen that among the plurality of sub-digital signals, each sub-digital signal corresponds to at least one bit of the digital signal, so the plurality of sub-digital signals can be spliced in sequence to form the digital signal. In this way, the electronic processor can send the digital signal to multiple DACs in the form of multiple sub-digital signals. Then, after the multiple DACs receive the corresponding sub-digital signals, the multiple DACs can be regarded as one Overall, it is equivalent to receiving the digital signal. For example, suppose there is a digital signal X representing 409, and the digital signal X is 110011001. Since the digital signal 011, sub-digital signal X3 is 001), so the electronic processor can send sub-digital signal X1 to DAC1, sub-digital signal X2 to DAC2, and sub-digital signal X3 to DAC3. In this way, DAC1, DAC2 and DAC3 as a whole are equivalent to receiving the digital signal 110011001.

It is worth noting that among these multiple sub-digital signals, although each sub-digital signal corresponds to at least one bit of the digital signal, the value represented by most of the sub-digital signals is smaller than the corresponding bit of the digital signal. the numerical value represented. Still as the above example, the sub-digital signal X1 corresponds to the first three digits of the digital signal X "110", but the value represented by the sub-digital signal The value is 384. In the same way, the sub-digital signal X2 corresponds to the middle three digits "011" of the digital signal X, but the value represented by the sub-digital signal is 24.

For any one of these multiple DACs, the DAC can convert the sub-digital signal it receives into an analog signal and send it to the OPU. Among these multiple DACs, other DACs besides this DAC can also perform similar operations. Therefore, the OPU can receive multiple analog signals, and perform electro-optical conversion on the multiple analog signals respectively, and obtain the analog signals one by one. Corresponding multiple optical signals to be calculated. Then, the OPU performs calculations on the plurality of optical signals to be calculated respectively, and obtains a plurality of calculated optical signals that correspond one-to-one to the multiple optical signals to be calculated.

It should be noted that among the multiple optical signals to be calculated, although each optical signal to be calculated corresponds to at least one bit of the digital signal, the values represented by most of the optical signals to be calculated are less than or equal to it. The value represented by the corresponding bit of the digital signal (because the multiple optical signals to be operated on are in one-to-one correspondence with the multiple sub-digital signals), then, the multiple operations corresponding to the multiple optical signals to be operated on are one-to-one. The final optical signal cannot be directly spliced as the calculation result of the digital signal (because most of the calculated optical signals represent relatively small values). The OPU needs to process the multiple calculated optical signals. Amplification and superposition can be used as the calculation result of the digital signal.

In order to further understand the working process of the above-mentioned optical computing system, the following is a further introduction to the matrix multiplication operation implemented by the optical computing system. It should be noted that the operations that the optical computing system can achieve are not limited to matrix multiplication operations. The following only uses matrix multiplication operations as a schematic explanation, and does not limit the types of operations that can be achieved by the optical computing system. As shown in Figure 3 (Figure 3 is another structural schematic diagram of the optical computing system provided by the embodiment of the present application), the computing task required by the optical computing system is to implement the multiplication operation between matrix A and matrix B. Suppose matrix A is an m×n matrix, and matrix B is an n×k matrix. The two matrices are as shown in formula (3):

The accuracy of the DAC in the optical computing system is d bits, the accuracy of the elements in matrix A is r1 bits, and the accuracy of the elements in matrix B is r2 bits, r1=h1*d+v1, r2=h2*d+v2, where, r1, r2, h1, h2, v1, v2 and d are all integers greater than 0. Since r1>d, r2>d, the elements of matrix A and matrix B satisfy a certain relationship. The element a ₁₁ in matrix A contains bit r1, and the 1st to dth bits of a ₁₁ are a ⁰ ₁₁ . The d+1st to 2dth digits are a ¹ ₁₁ ,…, the (h1-1)*d+1st to h1*dth digits are a ^h1-1 ₁₁ , the h1*d+1 to r1th digits The bit is a ^h1 ₁₁ . Then, a ₁₁ =a ^h1 ₁₁ *2 ^h1*d +a ^h1-1 ₁₁ *2 ^(h1-1)*d +…+a ¹ ₁₁ *2+a ⁰ ₁₁ . For example, assuming d=4, the digital signal representing 153 is 10011001,10011001=1001*2 ⁴ +1001.

Then, matrix A and matrix B can be decomposed as formula (4) when performing matrix multiplication:

In the above formula, the matrix multiplied by 2 ^h1*d+h2*d is:

The matrix multiplied by 2 ⁰ is:

In the result obtained by formula (4), the matrices multiplied by the remaining powers of 2 can also be referred to formula (5) and formula (6), which will not be described again here. Based on formula (4) to formula (6), it can be seen that by excluding the nth power of 2, matrix A can be decomposed into h1+1 m×n matrices, and matrix B can be decomposed into h2+1 n× The matrix of k, the result of multiplying matrix A by matrix B contains (h1+1)×(h2+1) m×k matrices.

Based on the above decomposition process, it can be seen that when the electronic processor outputs the digital signal a ₁₁ , it can set up h1+1 DACs to receive it. Among them, the first DAC receives the digital signal a ⁰ ₁₁ , (the digital signal a ⁰ ₁₁ corresponds to the digital signal a 0 11). The 1st to dth bits of the signal a ₁₁ , that is, the first DAC is used to convert the 1st to dth bits of the digital signal a ₁₁ into the corresponding analog signal), and the second DAC receives the digital signal a ¹ ₁₁ ,..., the h1th DAC receives the digital signal a ^h1-1 , and the h1+1th DAC receives the digital signal a ^h1 ₁₁ (the digital signal a ^h1 ₁₁ corresponds to the h1* _d +1th bit to the r1th bit of the digital signal a 11 bit, that is, the h1+1th DAC is used to convert the h1*d+1th bit to the r1th bit of the digital signal a ₁₁ into the corresponding analog signal). After receiving the corresponding digital signal, these h1+1 DACs can convert their respective digital signals into analog signals and send them to the OPU. That is, the first DAC converts the digital signal a ⁰ ₁₁ into an analog signal corresponding to a ⁰ ₁₁ and sends it to the OPU,..., the h1+1th DAC converts the digital signal a ^h1 ₁₁ into an analog signal corresponding to a ^h1 ₁₁ and sent to OPU.

Similarly, for any one of the remaining elements in matrix A except a ₁₁ , when the electronic processor outputs the digital signal corresponding to the element, it can also use h1+1 DACs to receive it. It can be seen that the reception of matrix A can be completed by setting (m×n) groups of DACs (each group of DACs contains (h1+1) DACs, which are used to receive an element in matrix A). In the same way, (n×k) groups of DACs can also be set up to complete the reception of matrix B (each group of DACs contains (h2+1) DACs, used to receive an element in matrix B).

Based on this, the OPU may include (h1+1)×(h2+1) modulators (for example, microring modulators, electroabsorption external modulators or Mach-Zehnder interferometers, etc.), where the (h1+1 )×(h2+1) modulators are used to obtain the matrix multiplied by 2 ^h1*d+h2*d (i.e. formula (5)). Then, the modulator needs to obtain the following two matrices:

Then, in (m×n) groups of DACs, the modulator can be connected to the h1+1th DAC of each group, so the modulator can receive the analog signal corresponding to a ^h1 ₁₁ , and the analog signal corresponding to a ^h1 ₁₂ The corresponding analog signal,..., the analog signal corresponding to a ^h1 _mn . After the modulator performs electro-optical conversion on these analog signals, the optical signal corresponding to a ^h1 ₁₁ and the optical signal corresponding to a ^h1 ₁₂ can be obtained, ..., the optical signals corresponding to a ^h1 _mn , these optical signals can constitute the optical signals corresponding to the matrix shown in formula (7). Similarly, the modulator is in a (n×k) group of DACs, and the modulator can be connected to the h2+1th DAC of each group, so the modulator can receive the analog signal corresponding to b ^h2 ₁₁ , Analog signals corresponding to b ^h2 ₁₂ ,..., analog signals corresponding to b ^h2 _mn . After the modulator performs electro-optical conversion on these analog signals, an optical signal corresponding to the matrix shown in formula (8) can be obtained. In this way, the modulator can perform matrix multiplication using these two optical signals to obtain the optical signal corresponding to formula (5).

The first modulator is used to obtain the matrix multiplied by 2 ⁰ (i.e. formula (6)). Then, the modulator needs to obtain the following two matrices:

Then, in (m×n) groups of DACs, the modulator can be connected to the first DAC of each group, so the modulator can receive the analog signal corresponding to a ⁰ ₁₁ , and the analog signal corresponding to a ⁰ ₁₂ The analog signal,..., is the analog signal corresponding to a ⁰ _mn . After the modulator performs electro-optical conversion on these analog signals, the optical signal corresponding to the matrix shown in formula (9) can be obtained. Similarly, the modulator is in a (n×k) group of DACs, and the modulator can be connected to the first DAC of each group. Therefore, the modulator can receive the analog signal corresponding to b ⁰ ₁₁ , and b The analog signal corresponding to ⁰ ₁₂ ,..., the analog signal corresponding to b ⁰ _mn . After the modulator performs electro-optical conversion on these analog signals, the optical signal corresponding to the matrix shown in formula (10) can be obtained. In this way, the modulator can use these two optical signals to perform matrix multiplication to obtain the optical signal corresponding to formula (6).

Similarly, for the remaining modulators except the (h1+1)×(h2+1) modulator and the first modulator, the DACs and workflows connected to this part of the modulators can also refer to the previous discussion. The relevant descriptions of the (h1+1)×(h2+1)-th modulator and the first modulator will not be described again here.

After the (h1+1)×(h2+1) modulators complete the operation, (h1+1)×(h2+1) optical signals can be obtained. These (h1+1)×(h2+1) optical signals The signal has a one-to-one correspondence with (h1+1)×(h2+1) m×k matrices, including the optical signal corresponding to formula (5), the optical signal corresponding to formula (6), and so on.

In order to obtain the result as shown in formula (4), the OPU is also equipped with an amplifier and an adder. The amplifier can be used for (h1+1)×(h2+1) modulators to obtain (h1+1)×(h2+ 1) optical signals are amplified to obtain (h1+1)×(h2+1) amplified optical signals. Among them, the matrix shown in formula (5) is the matrix multiplied by 2 ^h1*d+h2*d , so the optical signal corresponding to formula (5) needs to be amplified 2 ^h1*d+h2*d times, formula (6 ) is a matrix multiplied by 2 ⁰ , so the optical signal corresponding to formula (6) needs to be amplified 2 ⁰ times. The amplification multiples of other optical signals also follow the same rule and will not be repeated here. The adder is used to add (h1+1)×(h2+1) amplified optical signals to obtain an optical signal corresponding to the result obtained by formula (4).

It should be understood that the first effective bit of the aforementioned digital signal is the (i-1)*d+1th to i*d bits of the digital signal, and the second effective bit of the aforementioned digital signal is the h*d+th bit of the digital signal. 1 to the r-th bit. For example, assuming that the digital signal is a digital signal a ₁₁ , then, i=1,...,h1, that is, the first effective bit can be the 1st to d-th bit of a ₁₁ , and also It can be the d+1st to 2dth bits of a ₁₁ ,..., it can also be the ( _h1-1 )*d+1st to h1*dth bits of a 11, etc., and the second effective bit It can only be h1*d+1 to r1 of a ₁₁ . Of course, the digital signal can be a digital signal a ₁₂ , b ₁₁ , etc.

Further, since the digital signal can be split into multiple sub-digital signals bit by bit, the sub-digital signal corresponding to the first significant bit of the digital signal can be converted into the first analog signal by the corresponding DAC (ie, the aforementioned first DAC). Signal. Similarly, the sub-digital signal corresponding to the second effective bit of the digital signal can be converted into a second analog signal by a corresponding DAC (ie, the aforementioned second DAC). For example, assume that the first valid bit can be the 1st to d bits of a ₁₁ , and the second valid bit can only be the h1*d+1 to r1th bit of a ₁₁ . Then, the first ADC is the first DAC among the h1+1 DACs used to receive digital signals (used to receive digital signals a ⁰ ₁₁ ), the first analog signal is the analog signal corresponding to a ⁰ ₁₁ , and the second The DAC is the h1+1-th DAC among the h1+1 DACs for receiving digital signals (for receiving the digital signal a ^h1 ₁₁ ), the second analog signal is the analog signal corresponding to a ^h1 ₁₁ , and so on.

Further, based on the above example, the first optical signal is the optical signal corresponding to a ⁰ ₁₁ , the third optical signal is the optical signal corresponding to the matrix shown in formula (6), and the first modulator is the first modulator , the second optical signal is the optical signal corresponding to a ^h1 ₁₁ , the fourth optical signal is the optical signal corresponding to the matrix shown in formula (5), and the second modulator is (h1+1)×(h2+1 ) modulator. Of course, the third optical signal can also be an optical signal corresponding to the remaining matrices. The remaining matrices are obtained based on the matrix shown in formula (9) (by performing matrix multiplication) (at this time, the first modulator is converted into the corresponding modulator ), in the same way, the fourth optical signal can also be an optical signal corresponding to the remaining matrices, and the remaining matrices are obtained based on the matrix (matrix multiplication operation) shown in formula (7) (at this time, the second modulator is transformed into the corresponding modulator).

Further, the third optical signal is amplified by a factor greater than or equal to 2 to the (i-1)*d power, and the fourth optical signal is amplified by a factor greater than or equal to 2 h1*d power. For example, based on the above example, when the third optical signal is an optical signal corresponding to the matrix shown in equation (6), the third optical signal is amplified by a factor of 2 ⁰ , and the fourth optical signal is an optical signal corresponding to the matrix shown in equation (5). When the optical signal corresponding to the matrix shown is shown, the fourth optical signal is amplified by a factor of 2 ^h1*d+h2*d .

It should also be understood that this embodiment only takes r1=h1*d+v1 and r2=h2*d+v2 as an example for schematic introduction. In actual applications, the following situations may also occur: r1=h1*d and r2= h2*d. Then, the first significant bit of the aforementioned digital signal is the (i-1)*d+1 to i*d bits of the digital signal, and the second significant bit of the digital signal is the (j-1)*th bit of the digital signal. d+1 bit to j*d bit. For example, assuming that the digital signal is a digital signal a ₁₁ , then i=1,...,h1,j=1,...,h1,i≠j, that is, the first significant bit can be the 1st to 1st bit of a ₁₁ The dth bit, and the second valid bit can be the d+1th to 2dth bits of a ₁₁ , or the first valid bit can be the (h1-2)*d+1th to (h1-) bits of a ₁₁ 1)*d bit, and the second valid bit can be the (h1-1)*d+1th to h1*d bit of a ₁₁ and so on. Of course, the digital signal can be a digital signal a ₁₂ , b ₁₁ , etc.

Further, since the digital signal can be split into multiple sub-digital signals bit by bit, the sub-digital signal corresponding to the first significant bit of the digital signal can be converted into the first analog signal by the corresponding DAC (ie, the aforementioned first DAC). Signal. Similarly, the sub-digital signal corresponding to the second effective bit of the digital signal can be converted into a second analog signal by a corresponding DAC (ie, the aforementioned second DAC). For example, assume that the first valid bit can be the 1st to dth bits of a ₁₁ , and the second valid bit can be the ( _h1-1 )*d+1th to h1*d bits of a 11. Then, the first ADC is the first DAC among the h1+1 DACs used to receive digital signals (used to receive digital signals a ⁰ ₁₁ ), the first analog signal is the analog signal corresponding to a ⁰ ₁₁ , and the second The DAC is the h1-th DAC among the h1+1 DACs for receiving digital signals (for receiving the digital signal a ^h1-1 ₁₁ ), the second analog signal is the analog signal corresponding to a ^h1-1 ₁₁ , and so on.

Further, based on the above example, the first optical signal is the optical signal corresponding to a ⁰ ₁₁ , the third optical signal is the optical signal corresponding to the matrix shown in formula (6), and the first modulator is the first modulator , the second optical signal is the optical signal corresponding to a ^h1-1 ₁₁ , the fourth optical signal is the optical signal corresponding to the matrix shown in formula (11), and the second modulator is the h1×h2-th modulator. Of course, the third optical signal can also be an optical signal corresponding to the remaining matrices. The remaining matrices are obtained based on the matrix shown in formula (9) (performing matrix multiplication). Similarly, the fourth optical signal can also be an optical signal corresponding to the remaining matrices. of the optical signal, and the remaining matrices are obtained based on the matrix shown in formula (12) (performing matrix multiplication). Formula (11) and formula (12) are:

Further, the third optical signal is amplified by a factor greater than or equal to 2 raised to the (i-1)*d power, and the fourth optical signal is amplified by a factor greater than or equal to 2 raised to the (j-1)*d power. For example, based on the above example, when the third optical signal is an optical signal corresponding to the matrix shown in equation (6), the third optical signal is amplified by a factor of 2 ⁰ , and the fourth optical signal is an optical signal corresponding to the matrix shown in equation (11). When the optical signal corresponding to the matrix shown is shown, the fourth optical signal is amplified by a factor of 2 ^{(h1-1)*d+(h2-1)*d} .

In order to further understand the workflow of the above-mentioned optical computing system, the optical computing system will be further introduced below with two application examples. As shown in Figure 4 (Figure 4 is a schematic diagram of an application example of the optical computing system provided by the embodiment of the present application), assume that the system needs to implement the multiplication operation of matrix A and matrix B. Matrix A and matrix B are as shown in formula (1) , assuming that the precision of the elements in matrix A and the precision of the elements in matrix B are both 8 bits (bit), and the precision of DAC is 4 bits, define a ₁₁ =a ¹ ₁₁ *2 ⁴ +a ⁰ ₁₁ , where, a ¹ ₁₁ corresponds to the 5th to 8th digits of a ₁₁ , and a ⁰ ₁₁ corresponds to the 1st to 4th digits of a ₁₁ . Then, matrix A multiplied by matrix B can be decomposed as follows:

Based on formula (13), it can be seen that by eliminating the n-th power term of 2, matrix A can be decomposed into matrix A1 and matrix A2, and matrix B can be decomposed into matrix B1 and matrix B2, as shown in formula (14):

It can be seen that 32 groups of ADCs need to be set up in the system, each group contains 2 DACs. The 1st to 16th groups of DACs can be used to receive matrix A, that is, each group of DACs in these 16 groups is used to receive an element in matrix A. , the 17th to 32nd groups of DACs can be used to receive matrix B, that is, each group of DACs in these 16 groups is used to receive an element in matrix B (it should be noted that only the 1st to 32nd groups of DACs are shown in Figure 4 One of Group 16, and one of Groups 17 to 32).

Four modulators need to be set up in the OPU. The first modulator is used to implement the calculation of matrix A1 and matrix B1. The second modulator is used to implement the calculation of matrix A1 and matrix B2. The third modulator is used to implement the calculation of matrix A2 and matrix. For the calculation of B1, the fourth modulator is used to implement the calculation of matrix A2 and matrix B2. Therefore, the first modulator is connected to the first DAC in each group from group 1 to group 16, and to the first DAC in each group from group 17 to group 32. The second modulator is connected to the first DAC in each group from group 1 to group 16, and is connected to the second DAC in each group from group 17 to group 32. The third modulator is connected to the second DAC in each group from group 1 to group 16, and is connected to the first DAC in each group from group 17 to group 32. The fourth modulator is connected to the second DAC in each of groups 1 to 16, and to the second DAC in each of groups 17 to 32.

Furthermore, the first modulator can also be connected to a first amplifier, and the first amplifier can amplify the output of the first modulator ²⁸ times. The second modulator can also be connected to a second amplifier, which can amplify the output of the second modulator ²⁴ times. The third modulator can also be connected to a third amplifier, which can amplify the output of the third modulator ²⁴ times. The 4th modulator can also be connected to a 4th amplifier, and the 4th amplifier can amplify the output of the 4th modulator by 1x.

The four amplifiers are connected to the adder, and the adder is used to add the outputs of the four methods to obtain the optical signal corresponding to the result in formula (13).

As shown in Figure 5 (Figure 5 is a schematic diagram of another application example of the optical computing system provided by the embodiment of the present application), the optical computing system needs to implement the multiplication operation of two 4×4 matrices. Assume that the accuracy of the elements in matrix A and the matrix The precision of the elements in B is 16 bits (bit), and the precision of the DAC is 4 bits. Then, matrix A can be decomposed into matrix ①, matrix ②, matrix ③ and matrix ④, and matrix B can be decomposed into matrix ⑤, matrix ⑥, matrix ⑦ and matrix ⑧. Therefore, 32 sets of DACs need to be set up in the system, and each set of DACs contains 4 DACs, and 16 modulators are set in the OPU.

Through the connection relationship shown in Figure 5, the multiplication operation of matrix A and matrix B can be realized, thereby obtaining the corresponding optical signal.

Furthermore, the accuracy of the modulator in the OPU is usually 4 bits. By arranging multiple modulators in the OPU, even if the accuracy of each modulator itself is limited, high-bit digital signal operations can also be implemented. In this way, the optical computing system provided by the embodiments of the present application can implement high-precision optical computing of digital signals by deploying low-precision devices, which is beneficial to improving the computing performance of optical computing without any performance loss compared to electrical computing.

Furthermore, in the optical computing system, the OPU is equipped with a 2 n linear optical amplifier and an adder, which can process the output of the modulator accordingly to obtain accurate calculation results, which is conducive to the realization of low-cost, low-power High-efficiency, low-latency optical computing.

The above is a detailed description of the optical computing system provided by the embodiment of the present application. The optical signal processing method provided by the embodiment of the present application will be introduced below. Figure 6 is a schematic flow chart of an optical signal processing method provided by an embodiment of the present application. As shown in Figure 6, the method is implemented through an optical computing system. The system includes a first DAC, a second DAC and an OPU. The method includes:

1001. Convert the first effective bit of the digital signal into the first analog signal through the first DAC;

1002. Convert the second effective bit of the digital signal into a second analog signal through the second DAC;

1003. The OPU performs an operation based on the first optical signal derived from the first analog signal to obtain a third optical signal;

1004. The OPU performs an operation based on the second optical signal derived from the second analog signal to obtain a fourth optical signal;

1005. Use the OPU to amplify and superimpose the third optical signal and the fourth optical signal to obtain a calculation result of the digital signal.

In a possible implementation, the operation is a matrix multiplication operation.

The above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person familiar with the technical field can easily think of changes or replacements within the technical scope disclosed in the present application, and all of them should be covered. within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

An optical computing system, characterized in that the system includes:

The first digital-to-analog converter DAC is used to convert the first effective bit of the digital signal into the first analog signal;

a second DAC, used to convert the second effective bit of the digital signal into a second analog signal;

The optical computing unit OPU is configured to: perform operations based on the first optical signal derived from the first analog signal to obtain a third optical signal; perform operations based on the second optical signal derived from the second analog signal to obtain the third optical signal. Four optical signals; amplify and superimpose the third optical signal and the fourth optical signal to obtain the calculation result of the digital signal.
The system according to claim 1, characterized in that the accuracy of the first DAC and the accuracy of the second DAC are d bits, the digital signal contains r bits, and the first effective bit is the digital The (i-1)*d+1th to i*dth bits of the signal, the second effective bit is the (j-1)*d+1th to j*dth bits of the digital signal, where, r=h*d, r, h and d are all integers greater than zero, i=1,...,h,j=1,...,h,i≠j.
The system according to claim 1, characterized in that the accuracy of the first DAC and the accuracy of the second DAC are d bits, the digital signal contains r bits, and the first effective bit is the digital The (i-1)*d+1th bit to the i*dth bit of the signal, the second effective bit is the h*d+1th to r-th bit of the digital signal, where r=h*d+ v, r, h, d and v are all integers greater than zero, i=1,...,h.
The system according to claim 2, wherein the amplification factor of the third optical signal is greater than or equal to 2 to the power of (i-1)*d, and the amplification factor of the fourth optical signal is greater than or equal to 2. It is equal to 2 raised to the (j-1)*d power.
The system according to claim 3, wherein the amplification factor of the third optical signal is greater than or equal to 2 to the power of (i-1)*d, and the amplification factor of the fourth optical signal is greater than or equal to 2. It is equal to 2 raised to the h*d power.
The system according to any one of claims 1 to 5, wherein the OPU includes a first modulator, a second modulator, an amplifier and an adder;

The first modulator is used to convert the first analog signal into a first optical signal, and perform operations based on the first optical signal to obtain a third optical signal;

The second modulator is used to convert the second analog signal into a second optical signal, and perform calculations based on the second optical signal to obtain a fourth optical signal;

The amplifier is used to amplify the third optical signal to obtain a fifth optical signal; to amplify the fourth optical signal to obtain a sixth optical signal;

The adder is used to superpose the fifth optical signal and the sixth optical signal to obtain a calculation result of the digital signal.
The system of claim 6, wherein the first modulator and the second modulator are microring modulators, electroabsorption external modulators or Mach-Zehnder interferometers.
The system according to any one of claims 1 to 7, characterized in that the operation is a matrix multiplication operation.
An optical signal processing method, characterized in that the method is implemented through an optical computing system, the system includes a first DAC, a second DAC and an OPU, and the method includes:

Convert the first effective bit of the digital signal into a first analog signal through the first DAC;

Convert the second effective bit of the digital signal into a second analog signal through the second DAC;

The OPU performs operations based on the first optical signal derived from the first analog signal to obtain a third optical signal;

The OPU performs an operation based on the second optical signal derived from the second analog signal to obtain a fourth optical signal;

The OPU amplifies and superimposes the third optical signal and the fourth optical signal to obtain a calculation result of the digital signal.
The method according to claim 9, characterized in that the accuracy of the first DAC and the accuracy of the second DAC are d bits, the digital signal contains r bits, and the first effective bit is the digital The (i-1)*d+1th to i*dth bits of the signal, the second effective bit is the (j-1)*d+1th to j*dth bits of the digital signal, where, r=h*d, r, h and d are all integers greater than zero, i=1,...,h,j=1,...,h,i≠j.
The method according to claim 9, characterized in that the accuracy of the first DAC and the accuracy of the second DAC are d bits, the digital signal contains r bits, and the first effective bit is the digital The (i-1)*d+1th bit to the i*dth bit of the signal, the second effective bit is the h*d+1th to r-th bit of the digital signal, where r=h*d+ v, r, h, d and v are all integers greater than zero, i=1,...,h.
The method according to claim 10, wherein the amplification factor of the third optical signal is greater than or equal to 2 to the power of (i-1)*d, and the amplification factor of the fourth optical signal is greater than or equal to It is equal to 2 raised to the (j-1)*d power.
The method according to claim 11, wherein the amplification factor of the third optical signal is greater than or equal to 2 to the power of (i-1)*d, and the amplification factor of the fourth optical signal is greater than or equal to 2. It is equal to 2 raised to the h*d power.
The method according to any one of claims 9 to 12, wherein the OPU includes a first modulator, a second modulator, an amplifier and an adder, and the steps performed by the OPU include:

Convert the first analog signal into a first optical signal through the first modulator, and perform operations based on the first optical signal to obtain a third optical signal;

Convert the second analog signal into a second optical signal through the second modulator, and perform calculations based on the second optical signal to obtain a fourth optical signal;

The third optical signal is amplified by the amplifier to obtain a fifth optical signal; the fourth optical signal is amplified to obtain a sixth optical signal;

The fifth optical signal and the sixth optical signal are superimposed by the adder to obtain the calculation result of the digital signal.
The method of claim 14, wherein the first modulator and the second modulator are microring modulators, electroabsorption external modulators or Mach-Zehnder interferometers.
The method according to any one of claims 9 to 15, characterized in that the operation is a matrix multiplication operation.