An automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter

Abstract

This invention discloses an automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter, specifically relating to the field of bias control technology for optical communication transmitters. The method includes: acquiring the transmission symbol stream output by the encoding or modulation element within a transmission cycle of the optical transmission transmitter; XORing adjacent symbols to obtain transition bit strings; writing them into run-length segments of the same value according to the zero-value transition start and end; and outputting a transmission segment base table; reading the real-time bias value at the bias end; generating a central particle based on the real-time bias value; generating a revisiting particle from the historical convergence bias position in reverse order of writing; generating a mirror particle around the real-time bias value; and outputting a candidate bias particle swarm; fragmenting the transmission symbol stream within the transmission cycle into run-length segments of the same value; selecting the power-on segment based on a bidirectional greedy submodulus optimization algorithm; dividing the applied particle and shadow particle based on a relaxation-induced neighborhood search algorithm; and generating a target bias value based on the differential response driven by the particle swarm algorithm between the main output and the complementary output.

Description

Technical Field

[0001] This invention relates to the field of optical communication transmitter bias control technology, and more specifically, to an automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter. Background Technology

[0002] In the external modulation transmission link of the optical transmission transmitter, the bias point control of the electro-optic modulator is mainly used to offset the deviation of the operating point caused by temperature drift, device aging and changes in drive state. Existing processing mostly judges the bias state based on output optical power, backlight monitoring current, low-frequency disturbance response or error signal, and adjusts the bias voltage through table lookup correction, disturbance locking, small step search or particle swarm optimization. In scenarios such as data center interconnection, metropolitan area wavelength division transmission and high-speed coherent optical module transmission, the transmitter is in a long-term service-bearing state, the transmission of optical signals cannot be paused, and the bias end cannot be subjected to large-scale voltage sweeping. If multiple candidate bias voltages are applied to the electro-optic modulator in sequence, they will directly change the intensity, extinction ratio and eye diagram opening of the output optical signal. Therefore, under such in-service launch conditions, conventional particle swarm optimization is difficult to apply directly. When all particles actually act on the bias end, the monitoring end can observe optical power jitter, modulation depth decrease, receiver bit error rate increase or lock margin decrease. However, when only a single very small step size is allowed to be tested, the particle swarm will degenerate into local follow adjustment, and it will be difficult to return to the correct bias region in time after rapid temperature changes or long-term device aging. The technical problem to be solved by this application is: how to enable the particle swarm optimization algorithm to retain the multi-candidate search capability and complete the automatic control of the bias point of the electro-optic modulator without disrupting the in-service transmission state of the optical transmission transmitter. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter. This method involves fragmenting the transmitted symbol stream within each transmission cycle into segments with the same run value, selecting the powered-on segments based on a bidirectional greedy submodulus optimization algorithm, dividing the applied particles and shadow particles based on a relaxation-induced neighborhood search algorithm, and generating the target bias value based on the differential response between the main output and the complementary output using a particle swarm optimization algorithm. This method aims to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter, comprising: S1. Obtain the transmission symbol stream output by the encoding or modulation element within the transmission round of the optical transmission transmitter, XOR adjacent symbols to obtain the transition bit string, and write it into the same value run segment according to the zero value transition start and end, and output the transmission segment base table. S2. Read the real-time bias value at the bias end, generate the center particle with the real-time bias value, generate back visit particles from the historical convergence bias position in reverse order of writing, and generate mirror particles around the real-time bias value, and output the candidate bias particle group. S3. Using a bidirectional greedy sub-modulus optimization algorithm, the jump bit coverage difference is written in increments from the empty set. The particle number fragments are taken in descending order of coverage difference to form a forward fragment set. The run-length occupancy difference is deleted in descending order from the transmission fragment base table. The particle number fragments are taken in ascending order of occupancy difference to form a reverse fragment set. The intersection is taken to generate the power-on fragment sequence according to the sending order. S4. Based on the relaxation-induced neighborhood search algorithm, the candidate bias particles are arranged according to the bias distance from the real-time bias value. The number of particles of the powered segment is written into the applied particle group, and the remaining particles are inherited in order of parity from the near-end applied direction or the far-end return direction to generate the shadow particle group. S5. During the transmission period of the optical transmission transmitter, a bias micro-shift is applied to the bias terminal according to the sequence order of the applied particle group and the powered segment, and adjacent sampling differential is performed from the main output terminal to the complementary output terminal to generate the applied particle response sequence.

[0005] In a preferred embodiment, it further includes: S6. When the signs of the differentials at both ends are opposite, the individual migration direction is written using the particle swarm algorithm. The migration direction of the population is formed by the number of times the direction appears and the reverse order of writing. The shadow particle group is driven to move forward one position by the offset distance to generate the target offset value. The target offset value is applied to the offset end and written to the historical convergence offset position. The transmitter offset control result is output.

[0006] In a preferred embodiment, S1 includes: S1-1. Within each transmission cycle of the optical transmission transmitter, the symbol bit number is written bit by bit into the transmission symbol stream according to the transmission clock of the encoding or modulation element, and the transmission symbol pointed to by the previous symbol bit number and the transmission symbol pointed to by the next symbol bit number are used as XOR inputs to output the adjacent symbol operation chain. S1-2. For adjacent symbol operation chains, when the XOR result is one, write the next symbol bit number into the transition bit string; when the XOR result is zero, merge the next symbol bit number into the same value bit string formed by continuing the previous symbol bit number, and output the transition bit string and the same value bit string. S1-3. Based on the transition bit string, perform a chain break and write segment on the same value bit chain. Write the same value bit chain between adjacent transition bits in the transition bit string into the same value run segment, and generate a transmission segment base table according to the transmission order of the same value run segment in the transmission round.

[0007] In a preferred embodiment, S2 includes: S2-1. During the transmission cycle of the optical transmission transmitter, read the real-time bias value of the bias terminal of the electro-optic modulator according to the bias sampling clock, write the real-time bias value into the first position of the particle table, and set the bias difference field of the first position of the particle table to zero to generate the central particle. S2-2. Starting from the last bit of the historical convergence bias position, read back the historical bias value bit by bit, subtract the real-time bias value from the historical bias value to obtain the signed back visit difference, and connect the signed back visit difference back to the real-time bias value to generate back visit particles. S2-3. Perform sign flipping on the signed return difference of the return particles, connect the flipped signed return difference back to the real-time bias value to generate mirror particles, and write them into the particle table in the order of pairing the return particles and mirror particles to output the candidate bias particle swarm.

[0008] In a preferred embodiment, S3 includes: S3-1. Read the transmission segment base table within the transmission round of the optical transmission transmitter. For each run segment with the same value, scan from the start symbol bit number to the end symbol bit number. When the scan bit hits the transition bit string, write 1 in the corresponding bit of the segment cover word. When the scan bit does not hit the transition bit string, write 0 in the corresponding bit of the segment cover word. Then, generate the segment occupancy word by subtracting the start symbol bit number from the end symbol bit number and adding 1. S3-2. Using the bidirectional greedy sub-module optimization algorithm, the empty set is used as the positive fragment set. Fragments with the same value that have not entered the positive fragment set are selected one by one. The fragment cover word and the positive cover word are bitwise ORed. The positive gain word is generated by subtracting the number of bits of the previous value from the number of bits of the value after the operation. The candidate order is determined first according to the positive gain word from high to low. When the positive gain words are the same, the candidate order is rewritten according to the number of bits occupied by the fragment from short to long. The first fragment of the candidate order is written into the positive fragment set until the fragment count of the positive fragment set is consistent with the particle count of the candidate bias particle swarm.

[0009] In a preferred embodiment, S3 further includes: S3-3. Using the transmission segment base table as the reverse segment set, falsely delete the same run-length segments in the reverse segment set one by one and recalculate the reverse overlay word. Generate the reverse loss word by subtracting the number of bits of the false deletion from the number of bits of the previous value. First, determine the deletion order according to the reverse loss word from low to high. When the reverse loss words are the same, rewrite the deletion order according to the number of bits occupied by the segment from long to short. Move the segment with the first deletion order out of the reverse segment set until the segment count of the reverse segment set is consistent with the particle count of the candidate bias particle swarm. S3-4. Search the reverse fragment set item by item for the fragment identifiers in the forward fragment set. Write the matching fragments into the consensus fragment table. Store the missing fragments as forward candidate fragments according to the sending order. Then, store the fragments in the reverse fragment set that did not match the forward fragment set as reverse candidate fragments according to the reverse sending order. Use the forward candidate fragments and reverse candidate fragments to fill the gaps in the consensus fragment table. Output the power-on fragment sequence according to the sending order.

[0010] In a preferred embodiment, S4 includes: S4-1. Write the segment number in the above electric segment sequence into the applied resource bit table. For each particle, subtract the real-time bias value from the candidate bias value to obtain the signed bias difference. When the signed bias difference is negative, flip the sign and write the bias distance. When the signed bias difference is positive, write the bias distance directly. When the signed bias difference is zero, write the zero distance. Generate the distance arrangement chain in ascending order of bias distance and particle generation order. S4-2. Starting from the first position of the distance arrangement chain, occupy the same sequential resource position in the resource position table one particle at a time. The candidate bias particles whose resource positions are occupied are written into the applied particle group. When the signed bias difference is positive, the positive application direction is written. When the signed bias difference is negative, the reverse application direction is written. When the signed bias difference is zero, the hold direction is written. After all resource positions are occupied, the unoccupied particle chain is output.

[0011] In a preferred embodiment, S4 further includes: S4-3. The relaxation-induced neighborhood search algorithm is used to perform direction induction on the unoccupied particle chain. In the unoccupied particle chain, the odd-order particles read the direction of the adjacent particles in the distance arrangement chain that are ahead of themselves and have been written into the applied particle group. The even-order particles read the historical convergence bias position and write the return direction in the reverse order that is the same as their own order. When the return direction is missing, the first return direction in the reverse order is read and written to generate the induced direction chain. S4-4. Drive the unoccupied particle chain to perform relaxation and placeholder operations using the induced direction chain. Write the relaxation position after the current particle in the forward induced particle and write the relaxation position before the current particle in the reverse induced particle. Keep the induced particle written to the relaxation position of the current particle in the reverse induced particle. When the relaxation position is occupied, search for empty relaxation positions one by one along the induced direction and write the corresponding unoccupied particle. S4-5. Write the unoccupied particles that have completed the relaxation and occupancy into the shadow particle group, and mark the output end of the shadow particle group as the migration read end, so that the shadow particle group participates in the subsequent particle swarm algorithm migration calculation and does not output bias micro-shift to the bias end of the electro-optic modulator.

[0012] In a preferred embodiment, S5 includes: S5-1. Read the same sequence segment in the power-on segment sequence according to the particle order of the applied particle group, write the bias micro-shift of the applied particle into the starting symbol position of the same sequence segment, and generate a particle power-on mapping table. S5-2. When the optical transmission transmitter sends the start symbol position number, the real-time bias value is added to the bias micro-shift amount to form the applied bias value and written to the bias terminal of the electro-optic modulator. When the end symbol position number is sent, the real-time bias value is written back and the bias application record in the segment is output. S5-3. Based on the bias application record within the segment, read the sampled values ​​of the main output terminal and the complementary output terminal respectively, subtract the previous sampled value from the current sampled value to generate a double-ended differential symbol, and write the double-ended differential symbol into the applied particle response sequence according to the order of the applied particles.

[0013] In a preferred embodiment, S6 includes: S6-1. Read the response sequence of applied particles according to the order of applied particles using the particle swarm algorithm. Perform a symbolic product operation between the differential symbol of the main output terminal and the differential symbol of the complementary output terminal. When the symbolic product is negative, write the bias micro-shift direction of the applied particles in the same order into the individual migration table. When the symbolic product is positive or zero, write the position number of the applied particles in the same order into the mismatch migration position. S6-2. Scan from the first position to the last position of the individual migration table. Each time the positive bias micro-shift direction appears, the positive count is incremented by one. Each time the negative bias micro-shift direction appears, the negative count is incremented by one. If the positive count and the negative count are not equal, the direction with the larger count value is written into the group migration direction. If the positive count and the negative count are equal, the first non-empty direction is read from the last position of the individual migration table and written into the group migration direction. S6-3. Using the migration direction of the population as the positioning direction, subtract the real-time bias value from the bias value of the shadow particles in the shadow particle group to obtain the signed shadow difference. Form a same-direction shadow chain by sorting the shadow particles with the same sign as the migration direction of the population in ascending order of bias distance. Read the bias value of the first and last shadow particles in the same-direction shadow chain to generate the target bias value. Apply the target bias value to the bias end of the electro-optic modulator and write it to the historical convergence bias position.

[0014] The technical effects and advantages of this invention are as follows: 1. This scheme reduces optical signal disturbances in in-service transmission by applying particle and shadow particle separation, allowing a small number of particles to perform real bias micro-shifts and the majority of particles to participate in migration calculations. 2. After writing the transmission symbol stream into run-length segments of the same value, select the power-on segment so that the bias probe follows the actual transmission sequence, thereby relatively reducing the impact of random probes on service transmission. 3. Perform bidirectional greedy sub-modulus optimization on the transmission segment base table, so that transition coverage and run length occupancy participate in the segment picking, which relatively improves the representativeness of the offset trial segment; 4. Perform differential sign product judgment between the main output terminal and the complementary output terminal to prevent changes in power in the same direction from entering the migration count, thereby relatively suppressing misleading regulation by non-bias factors; 5. Generate back-visit particles and mirror particles based on historical convergence bias positions, so that particle search can inherit existing convergence results, thereby improving the bias callback capability after temperature drift and aging. Attached Figure Description

[0015] Figure 1 This is a flowchart outlining the method steps of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Refer to the instruction manual appendix Figure 1 The present invention provides an automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter, comprising: S1. Obtain the transmission symbol stream output by the encoding or modulation element within the transmission round of the optical transmission transmitter, XOR adjacent symbols to obtain the transition bit string, and write it into the same value run segment according to the zero value transition start and end, and output the transmission segment base table. In this embodiment, the fragmentation processing of the transmission symbol stream is used to convert the service symbols output by the optical transmission transmitter through the encoding or modulation elements within a transmission round into the transmission segment base table read by the subsequent bidirectional greedy sub-mode optimization algorithm. The transmission round starts with the bias control trigger bit and ends with the completion bit of the current round's transmission symbol stream writing. The symbol bit number within the transmission round is generated by incrementing the transmission clock of the encoding or modulation elements. Single-bit transmission symbols directly enter the XOR operation, while multi-bit transmission symbols are XORed bit by bit, with any bit XORed to one as the basis for transition writing. The implementation process includes the following steps: In S1-1, to ensure that the transmitted symbol stream enters the adjacent symbol operation according to the transmitter's transmission timing, the transmitted symbol stream output by the encoding or modulation element within the transmission round is read, and a symbol bit number is written for each transmitted symbol according to the transmission clock. During processing, the first transmitted symbol is written to the starting symbol bit number of the transmission round, and the symbol bit number of the subsequent transmitted symbol is obtained by adding one to the previous symbol bit number. Then, the transmitted symbol pointed to by the previous symbol bit number and the transmitted symbol pointed to by the subsequent symbol bit number are used as a set of XOR inputs, and adjacent symbol operation terms are generated group by group along the transmission clock. All adjacent symbol operation terms are written to the adjacent symbol operation chain in order of symbol bit number. If there is only one transmitted symbol in the transmission round, no adjacent symbol operation terms are generated, and the symbol bit number of the single transmitted symbol is written to the single symbol run record for S1-3 to write into the transmission segment base table. In S1-2, to separate the transmission state change position and the transmission state hold position from the adjacent symbol operation chain, the XOR inputs in the adjacent symbol operation chain are read item by item and the XOR operation is performed. When the XOR result is one, the bit number of the next symbol is written into the transition bit string, and the bit number of the next symbol is used as the starting bit number of the new same value bit chain. When the XOR result is zero, the bit number of the next symbol is merged into the same value bit chain formed by the continuation of the bit number of the previous symbol. For multi-bit transmission symbols, when all bit XOR results are zero, the bit number of the next symbol is merged into the same value bit chain. When any bit XOR result is one, the bit number of the next symbol is written into the transition bit string. If the transition bit string is empty at the end of the transmission round, the same value bit chain formed from the starting bit number of the transmission round to the ending bit number of the transmission round is output as the same value bit chain of the whole round. In S1-3, to convert the transition bit string and the same value bit chain into the segment record to be selected and read in the subsequent power-on segment, the same value bit chain is broken and written according to each transition bit number in the transition bit string; the first same value run segment is started by the starting symbol bit number of the transmission round when the previous transition bit does not exist, and terminated by the symbol bit number before the first transition bit number; the middle same value run segment is started by the previous transition bit number, and terminated by the symbol bit number before the next transition bit number; the last same value run segment is started by the last transition bit number and terminated by the ending symbol bit number of the transmission round; when two transition bit numbers are adjacent, no empty segment is written, and the next transition bit number is used as the starting bit of the next same value run segment; after the break and write segment is completed, the same value run segment is written into the transmission segment base table according to the transmission order of the transmission round, and the starting symbol bit number, the ending symbol bit number, the transmission order, and the adjacent transition bit numbers are used as segment record fields for S3 to read; Through the above implementation process, the transmission symbol stream is first converted into a transmission segment base table with transmission order, transition boundaries and run length occupancy relationship, and then enters the subsequent power-on segment selection process, so that the timing of bias micro-shift application comes from the actual transmission timing of the optical transmission transmitter. In practical applications: When the optical transmission transmitter outputs a transmission symbol stream of 1, 1, 0, 0, 0, 1 from the encoding or modulation element in one transmission round, the transmission clock sequentially writes six symbol bits. After XORing adjacent symbols, a transition bit string is written to the third and sixth symbol bits. The first to second symbol bits, the third to fifth symbol bits, and the sixth symbol bit are written to the same run length segment, thereby forming a transmission segment base table for subsequent bias particle power-on segments to select and read.

[0018] S2. Read the real-time bias value at the bias end, generate the center particle with the real-time bias value, generate back visit particles from the historical convergence bias position in reverse order of writing, and generate mirror particles around the real-time bias value, and output the candidate bias particle group. In this embodiment, the construction of the candidate bias particle swarm is used to convert the real-time bias value, historical convergence bias position, and mirror bias position of the electro-optic modulator bias terminal within a transmission round into a particle table for selection and reading in subsequent power-on segments; the transmission round uses the transmit counter of the encoding or modulation element, the bias sampling clock is provided by the bias control channel of the optical transmission transmitter, the historical convergence bias position is read back in reverse order according to the target bias value written to the tail bit at the end of the previous transmission round, and the number of particles in the candidate bias particle swarm does not exceed the number of valid run-length segments with the same value in the transmission segment base table; this implementation process includes the following steps: In S2-1, to ensure that the candidate bias particle swarm has a reference particle with the current bias state, the real-time bias value of the electro-optic modulator bias terminal is read according to the bias sampling clock during the transmission round of the optical transmission transmitter, and the real-time bias value is converted into the real-time bias code value in the code field of the bias terminal driver circuit register. During processing, the real-time bias code value is written to the first position of the particle table, the bias difference field of the first position of the particle table is written to zero, and the generation source field of the first position of the particle table is written to the center source to generate the center particle. If the bias sampling clock does not return a sampling value in this transmission round, the target bias value written to the tail of the historical convergence bias position in the previous transmission round is read as the real-time bias value, and the reuse source field is written to the first position of the particle table for subsequent revisit particle calculation. In S2-2, to convert the historical convergence bias position into a candidate position for revisiting the real-time bias value, the historical bias value is read back bit by bit in reverse order of writing, starting from the last bit of the historical convergence bias position. The upper limit of the revisit is the number of valid run segments with the same value minus one in the base table of transmission segments. During processing, the historical bias value is subtracted from the real-time bias value to obtain the signed revisit difference. When the signed revisit difference is positive, it means that the historical bias value is on the positive side of the real-time bias value. When the signed revisit difference is negative, it means that the historical bias value is on the negative side of the real-time bias value. When the signed revisit difference is zero, it means that the historical bias value is the same as the real-time bias value. Then, the signed revisit difference is connected back to the real-time bias value to generate a revisit particle, and the revisit particle is written to the next particle position after the center particle in the particle table. If the historical convergence bias position is empty, no revisit particle is generated. The particle table only retains the center particle and writes the number of the center particle into the particle count field of the candidate bias particle group. In S2-3, to form candidate particles on both sides of the real-time bias value without amplifying the actual write operation at the bias end, the sign of the signed back-visit difference of each back-visit particle is flipped, and the flipped signed back-visit difference is connected back to the real-time bias value to generate a mirror particle. During processing, when the bias value of the mirror particle is higher than the upper boundary of the code field of the bias end driver circuit register, the upper boundary code value is written; when the bias value of the mirror particle is lower than the lower boundary of the code field of the bias end driver circuit register, the lower boundary code value is written; when the bias value of the mirror particle falls into the code field of the register, the calculated bias value of the mirror particle is written. When the center particle, back-visit particle, and mirror particle have the same bias value, the first particle is retained according to the generation order, and subsequent particles with the same bias value are not written to the particle table. The number of particles already written in the particle table is written back as the particle count of the candidate bias particle group, and the candidate bias particle group is output for S3 to read. Through the above implementation process, the real-time bias value first forms the central particle, the historical converged bias position then forms the return particle, the return particle forms the mirror particle after the sign is flipped, and the candidate bias particle group is carried by the particle table and the particle count is passed to the subsequent power-on segment selection process. In practical applications: When the optical transmission transmitter reads the real-time bias code value of 512 in the current transmission round, and reads back to 520 and 508 in the reverse order of writing the historical convergence bias position, 520 minus 512 yields the forward signed return difference 8 and generates the return particle 520. 508 minus 512 yields the reverse signed return difference negative 4 and generates the return particle 508. Then, the sign is flipped on the forward signed return difference 8 to generate the mirror particle 504, and the sign is flipped on the reverse signed return difference negative 4 to generate the mirror particle 516. The particle table is written in the order of generation of the center particle 512, the return particle 520, the mirror particle 504, the return particle 508, and the mirror particle 516, and the number of written particles is passed to S3 as the basis for retrieving the number of power-on segments.

[0019] S3. Using a bidirectional greedy sub-modulus optimization algorithm, the jump bit coverage difference is written in increments from the empty set. The particle number fragments are taken in descending order of coverage difference to form a forward fragment set. The run-length occupancy difference is deleted in descending order from the transmission fragment base table. The particle number fragments are taken in ascending order of occupancy difference to form a reverse fragment set. The intersection is taken to generate the power-on fragment sequence according to the sending order. In this embodiment, the generation of the power-on segment sequence is used to convert same-value run-length segments in the transmission segment base table into bias micro-shift application opportunities, so that the actual applied particles are only bound to the transmission segments that can carry bias probes within the transmission round. This embodiment uses the particle count field of the candidate bias particle swarm as the basis for segment count, and uses the transmission order, transition bit, and segment occupancy length of the same-value run-length segments as the calculation objects. The power-on segment sequence is generated through bidirectional calculation of forward write gains and reverse delete losses. This implementation process includes the following steps: In S3-1, to ensure that the same-value run segments have the bit-level representation required for subsequent sub-mode gain calculation, each same-value run segment in the transmission segment base table is read, and the symbol bits in the transmission round are scanned bit by bit from the start symbol bit to the end symbol bit. During processing, the transmit bits inside the same-value run segment are written with 1 in the corresponding bit of the segment overlay word, the previous and next transition bits adjacent to the same-value run segment are written with 1 as transition action bits, and the remaining scan bits are written with zero. Then, the segment occupancy word is generated by subtracting the start symbol bit from the end symbol bit and adding 1. If the same-value run segment is located at the beginning of the transmission round and there is no previous transition bit, only the transmit bits inside the segment and the next transition action bit are written. If the same-value run segment is located at the end of the transmission round and there is no next transition bit, only the previous transition action bit and the transmit bits inside the segment are written. The transmission segment base table with the segment overlay word and the segment occupancy word is output. In S3-2, to select the same-value run-length segments that make a significant contribution to the coverage of the transmission round transition from the empty set direction, a bidirectional greedy submodal optimization algorithm is used, with the empty set as the positive segment set and the positive coverage word initialized to an all-zero word. During processing, the same-value run-length segments that have not yet entered the positive segment set are read one by one, and a bitwise OR operation is performed between the segment coverage word and the positive coverage word. The number of bits in the result of the bitwise OR operation is counted, and then the number of bits in the positive coverage word before the bitwise OR operation is subtracted to obtain the positive gain word of the same-value run-length segment. Then, the positive gain word is first sorted from high to low. The candidate order is arranged from lowest to highest. When the positive gain words are the same, the candidate order is rewritten according to the segment occupancy words from shortest to longest. When the segment occupancy words are the same, the candidate order is rewritten according to the transmission order from first to last. The same run-length segment with the first position of the candidate order is written into the positive segment set. At the same time, the positive overlay word is updated with the segment overlay word of the same run-length segment until the segment count of the positive segment set reaches the particle count of the candidate bias particle swarm. If the number of same run-length segments that have not entered the positive segment set is exhausted before the particle count, the positive segment set that has been written is used to enter the S3-4 filling process. In S3-3, to remove run-length segments with low impairment to transmission round transitions from the whole set, all run-length segments with the same value in the transmission segment base table are used as the reverse segment set, and the reverse overlay word is obtained by bitwise ORing all segment overlay words. During processing, one run-length segment with the same value in the reverse segment set is falsely deleted one by one. The bitwise OR operation of the segment overlay word is re-performed on the remaining run-length segments with the same value after the false deletion to obtain the overlay word after the false deletion. Then, the reverse loss word of the run-length segment with the same value is obtained by subtracting the number of bits of the one-value overlay word after the false deletion from the one-value overlay word before the false deletion. First, the deletion order is arranged in ascending order of reverse loss words. If the reverse loss words are the same, the deletion order is rewritten in descending order of fragment occupied words. If the fragment occupied words are the same, the deletion order is rewritten in ascending order of transmission order. The same run-length fragment with the same deletion order is removed from the reverse fragment set. At the same time, the reverse overwrite word is recalculated using the remaining run-length fragments with the same value until the fragment count of the reverse fragment set reaches the particle count of the candidate bias particle swarm. If the fragment count of the reverse fragment set is already equal to the particle count of the candidate bias particle swarm, the false deletion operation is no longer performed, and the reverse fragment set is directly sent to S3-4. In S3-4, to combine the coverage gain of the forward fragment set and the deletion retention result of the reverse fragment set into the same power-on fragment source, the fragment identifiers in the forward fragment set are searched item by item in the reverse fragment set. Matching same-value run segments are written to the consensus fragment table, while missing same-value run segments are written to the forward candidate fragment table according to their transmission order. Then, same-value run segments that missed the forward fragment set are searched from the reverse fragment set and written to the reverse candidate fragment table in reverse transmission order. When handling missing segments, the fragment count in the consensus fragment table is subtracted from the particle count of the candidate bias particle swarm. Once the missing number of bits is obtained, one bit is added starting from the first position of the forward candidate fragment list, and then one bit is added starting from the first position of the reverse candidate fragment list. This alternating filling continues until the missing number of bits reaches zero. If the forward candidate fragment list is exhausted first, the bits are added from the reverse candidate fragment list in reverse order of transmission. If the reverse candidate fragment list is exhausted first, the bits are added from the forward candidate fragment list in order of transmission. After filling is completed, the consensus fragment list is rearranged according to the order of transmission, and the rearranged consensus fragment list is written as a power-on fragment sequence for S4 to allocate and apply particle groups according to the fragment count of the power-on fragment sequence. Through the above implementation process, the same-value run-length segments in the transmission segment base table are first written as segment overlay words and segment occupancy words, and then two sets of segment selection results are formed by positive gain writing and reverse loss deletion respectively. Finally, the power-on segment sequence corresponding to the particle count of the candidate bias particle swarm is generated through consensus retrieval and candidate filling. In practical applications: If four run segments with the same value are formed within a transmission round, three particles are written into the candidate bias particle swarm. The bidirectional greedy sub-modulus optimization algorithm first selects three run segments with the same value in the forward segment set according to the coverage gain, and then retains three run segments with the same value in the reverse segment set according to the deletion loss. When there are two identical segments in the two sets of results, the two identical segments are first written into the consensus segment table. The remaining missing segment is then written from the forward candidate segment table and the reverse candidate segment table according to the alternation substitution rule. Subsequently, three power-on segments are output according to the transmission order, so that the subsequent applied particles enter the bias micro-shift application process according to the actual transmission order of the transmitter.

[0020] S4. Based on the relaxation-induced neighborhood search algorithm, the candidate bias particles are arranged according to the bias distance from the real-time bias value. The number of particles of the powered segment is written into the applied particle group, and the remaining particles are inherited in order of parity from the near-end applied direction or the far-end return direction to generate the shadow particle group. In this embodiment, the allocation of the applied particle group and the shadow particle group is used to divide the candidate bias particle swarm into particles that actually act on the bias end of the electro-optic modulator and particles that only participate in subsequent migration calculations; the power-on segment sequence provides the actual number of applications, the real-time bias value provides the distance calculation benchmark, the historical convergence bias position provides the source of the revisit direction, and the relaxation-induced neighborhood search algorithm, through resource bit occupation, direction induction, and relaxation bit occupation, allows unoccupied particles to retain migration basis in a state where they are not written to the bias end; this implementation process includes the following steps: In S4-1, to ensure that candidate bias particles enter the resource allocation process according to their proximity to the real-time bias value, the segment number of each power-on segment in the power-on segment sequence is first read and written into the applied resource bit table according to the transmission order of the segment number. During processing, the candidate bias value in the candidate bias particle group is read one by one, and the signed bias difference is obtained by subtracting the real-time bias value from the candidate bias value. When the signed bias difference is negative, the sign of the value is flipped and written into the bias distance. When the signed bias difference is positive, the signed bias difference is directly written into the bias distance. When the signed bias difference is zero, the bias distance is written into zero. Then, the candidate bias particles are arranged in ascending order of bias distance. When the bias distances are the same, the candidate bias particles are arranged in ascending order of particle generation order to generate a distance arrangement chain for the applied resource bit occupancy step to read. In S4-2, to ensure that the actual number of bias micro-shifts is constrained by the power-on segment sequence, candidate bias particles are read starting from the first position of the distance arrangement chain, and the read candidate bias particles occupy the same-order resource bits in the applied resource bit table. During processing, candidate bias particles with occupied resource bits are written into the applied particle group. When the signed bias difference of a candidate bias particle is positive, it is written in the positive application direction; when the signed bias difference is negative, it is written in the reverse application direction; and when the signed bias difference is zero, it is written in the hold direction. When all resource bits in the applied resource bit table are occupied, the writing of candidate bias particles into the applied particle group is stopped, and candidate bias particles in the distance arrangement chain that have not yet occupied resource bits are written into the unoccupied particle chain according to their original arrangement order. If the power-on segment sequence has no segment bit number, the applied resource bit table is empty, all candidate bias particles are written into the unoccupied particle chain, and no bias end writing action is generated in subsequent steps. In S4-3, to generate an induced direction that does not act on the bias end for particles in the unoccupied particle chain, a relaxation-induced neighborhood search algorithm is used to perform directional induction according to the particle order of the unoccupied particle chain. During processing, unoccupied particles with odd-numbered orders search for neighboring particles that have been written into the applied particle group before them within the distance arrangement chain, and read the applied direction of the neighboring particle as the proximal induced direction. If there are no neighboring particles that have been written into the applied particle group before them, the applied direction of the first particle of the applied particle group is read as the proximal induced direction; even-numbered orders... Unoccupied particles are read from the historical convergence bias position and written to the historical bias value in reverse order that is in the same order as themselves. The return direction is obtained by subtracting the real-time bias value from the historical bias value. When the difference is positive, the positive return direction is written; when the difference is negative, the reverse return direction is written; when the difference is zero, the return direction is kept. If there is no historical bias value in the same order as itself, the first historical bias value in reverse order is read and written to generate the return direction. If the historical convergence bias position is empty, the return direction is kept. After the direction induction is completed, the induced direction chain is generated according to the order of the unoccupied particle chain. In S4-4, to ensure that each particle in the unoccupied particle chain occupies a unique relaxation position, a relaxation position table is first generated according to the number of particles in the unoccupied particle chain, with the relaxation position number corresponding one-to-one with the sequence of the unoccupied particle chain. During processing, the induced direction chain is read to drive the unoccupied particle chain to perform relaxation placement. Positive induced particles are written to the relaxation position after the current particle sequence, and negative induced particles are written to the relaxation position before the current particle sequence, keeping the induced particles written to the relaxation position of the current particle sequence. When a positive induced particle is at the last position, an empty relaxation position is searched from the last position to the first position. When a negative induced particle is at the first position, an empty relaxation position is searched from the first position to the last position. When the target relaxation position has been occupied, an empty relaxation position is searched position by position along the induced direction, and the first empty relaxation position found is written to the corresponding unoccupied particle. When the relaxation position table is full, the unoccupied particle chain that has completed relaxation placement is output. In S4-5, to ensure that unoccupied particle chains participate in subsequent particle swarm migration calculations without affecting the electro-optic modulator bias terminal, unoccupied particles that have completed relaxation are written into the shadow particle group according to their relaxation position numbers. During processing, a migration read-end marker is written into the particle table field of the shadow particle group, along with the candidate bias value, bias distance, induction direction, and relaxation position number of the shadow particles. The migration read-end marker serves as the record field for reading the bias value and induction direction of the shadow particles in S6, and is not connected to the electro-optic modulator bias terminal. The shadow particle group only provides migration calculation input in subsequent particle swarm algorithms and does not output bias micro-shifts to the electro-optic modulator bias terminal. Only when the particle group is applied does it enter the bias micro-shift application process in S5. Through the above implementation process, the candidate bias particle swarm first forms a distance arrangement chain according to the bias distance, and then the number of actual applied particles is limited by the power-on segment sequence. Unoccupied particles are written into the shadow particle group after being induced by the near-end application direction and the far-end return direction, thus forming a particle allocation structure with a small number of applications and a large number of migration reads. In practical applications: If the power-on segment sequence contains three segment bits and the candidate bias particle group contains seven candidate bias particles, then the first three candidate bias particles in the distance arrangement chain occupy three application resource bits in sequence and are written into the application particle group. The remaining four candidate bias particles are written into the unoccupied particle chain. The first and third particles in the unoccupied particle chain read the adjacent application particle direction to form the proximal induction direction, and the second and fourth particles read the historical convergence bias position to form the far-end return direction. The four unoccupied particles then occupy four relaxation bits according to the induction direction and are written into the shadow particle group for S6 to read when generating the target bias value.

[0021] S5. During the transmission transmission period of the optical transmission transmitter, a bias micro-shift is applied to the bias end according to the sequence order of the applied particle group and the powered segment, and adjacent sampling differential is performed from the main output end to the complementary output end to generate the applied particle response sequence. In this embodiment, the generation of the applied particle response sequence is used to convert the candidate bias values ​​in the applied particle group into bias micro-movements within the emission round, and to write the changes in the optical response of the main output and complementary output after the bias micro-movement into the data sequence read by the particle swarm algorithm; the main output is the output optical path monitoring point of the electro-optic modulator entering the optical transmission link, and the complementary output is the complementary optical output monitoring point of the electro-optic modulator. The samples at both ends are aligned by the same transmit counter, and the bias write completion bit is obtained by converting the write response sample number of the bias control channel. The write response sample number comes from the bias end drive circuit register record or the factory calibration record; this implementation process includes the following steps: In S5-1, to ensure that each applied particle acts on only one corresponding power-on segment, segments in the same order of the power-on segment sequence are read according to the particle order of the applied particle group, and the candidate bias value and real-time bias value of the applied particle are read. During processing, the bias micro-shift amount is obtained by subtracting the real-time bias value from the candidate bias value of the applied particle. When the bias micro-shift amount is positive, a positive micro-shift is written; when the bias micro-shift amount is negative, a reverse micro-shift is written; when the bias micro-shift amount is zero, a hold micro-shift is written. Then, the bias micro-shift amount is written to the starting symbol position of the segment in the same order to generate a particle power-on mapping table. If the particle count of the applied particle group is less than the segment count of the power-on segment sequence, the power-on segments exceeding the particle count of the applied particle group are not written into the particle power-on mapping table. If the particle count of the applied particle group is greater than the segment count of the power-on segment sequence, the applied particles exceeding the segment count of the power-on segment sequence do not enter the bias micro-shift application process of this launch round. In S5-2, to convert the bias micro-shift in the particle power-on mapping table into the actual write action of the electro-optic modulator bias terminal, when the optical transmission transmitter sends the start symbol bit number recorded in the particle power-on mapping table, the real-time bias value is read and added to the bias micro-shift to form the applied bias value. During processing, the applied bias value is written to the electro-optic modulator bias terminal, and the bias write completion bit is calculated from the start symbol bit number according to the write response sampling number of the bias control channel. Then, when the stop symbol bit number is sent, the real-time bias value is written back to the electro-optic modulator bias terminal to form the bias application record within the segment. If the bias write completion bit number is later than the stop symbol bit number, the corresponding power-on segment will not enter the sampling differential process between the main output terminal and the complementary output terminal, and a non-response mark will be written in the bias application record within the segment for S5-3 to skip reading. In S5-3, to convert the bias-shifted dual-end optical response into symbolic data that can be directly computed by the particle swarm optimization algorithm, the sampled values ​​of the main output end and the complementary output end are read from the bias write completion bit according to the bias application record within the segment, and the sampled values ​​at both ends are aligned according to the transmission counter sequence. During processing, the same-end differential value is generated by subtracting the previous sampled value from the current sampled value. The sampled value before the first valid sampled item in the segment is read from the same-end sampled value of the transmission sequence before the segment start symbol bit number. If the previous transmission sequence does not exist, the same-end reference sampled value before the real-time bias value is applied is read. The differential values ​​of the main output end and the complementary output end are both written with positive difference written as 1, negative difference written as negative 1, and zero difference written as zero to generate dual-end differential symbols, and the dual-end differential symbols are written into the applied particle response sequence according to the applied particle sequence. Power-on segments containing non-response markers are not written into the applied particle response sequence, and the corresponding applied particle bit number is written into the non-response particle bit table for S6 not to participate in the generation of individual migration direction. Through the above implementation process, the applied particle group first establishes a sequence mapping with the powered segment sequence, and the candidate bias value of the applied particle is then converted into the bias micro-shift amount within the segment. The main output terminal and the complementary output terminal then form a double-ended differential symbol after the bias writing is completed, so that the subsequent particle swarm algorithm can determine the bias micro-shift direction based on the double-ended differential symbol. In practical applications: if the candidate bias value of the first particle in the applied particle group is 520 and the real-time bias value is 512, then the bias micro-shift is positive 8; when the optical transmission transmitter sends the start symbol bit number of the same sequence of powered segments, the bias terminal writes the applied bias value 520, and after the bias writing is completed, reads the sampled values ​​of the main output terminal and the complementary output terminal; if the sampled value of the main output terminal increases relative to the previous sampled value, and the sampled value of the complementary output terminal decreases relative to the previous sampled value, then the differential symbol of the main output terminal is written as 1, the differential symbol of the complementary output terminal is written as negative 1, and the applied particle response sequence is written according to the applied particle sequence.

[0022] S6. When the signs of the differentials at both ends are opposite, the individual migration direction is written using the particle swarm algorithm. The migration direction of the population is formed according to the number of times the direction appears and the reverse order of writing. The shadow particle group is driven to move forward one position by the offset distance to generate the target offset value. The target offset value is applied to the offset end and written to the historical convergence offset position. The transmitter offset control result is output. In this embodiment, the output of the transmitter bias control result is used to convert the two-end difference symbols in the applied particle response sequence into the migration direction of the particle swarm algorithm, and to generate the target bias value through the shadow particle group; the differential symbols at the main output end and the differential symbols at the complementary output end are encoded with the values ​​of one, negative one, and zero. A negative product of symbols indicates that the two-end response directions are opposite, a positive product indicates that the two-end response directions are the same, and a zero product indicates that at least one end has not changed. Only applied particles with opposite two-end response directions enter the individual migration table. This implementation process includes the following steps: In S6-1, to extract particle responses related to bias migration from the applied particle response sequence, the particle swarm optimization algorithm reads the differential symbols of the main output and complementary output in sequence according to the applied particle order, and performs a symbolic product operation on the differential symbols of the main output and complementary output. During processing, when the symbolic product is negative, the bias micro-shift direction of the applied particles in the same order is written into the individual migration table; when the symbolic product is positive or zero, the position number of the applied particles in the same order is written into the mismatch migration bit. The mismatch migration bit only records the position numbers of applied particles that do not enter the migration count, and is used in subsequent firing rounds to exclude the mapping combination of the same power-on segment and the same applied particle, and does not participate in the generation of the swarm migration direction. If there are unresponsive particle position numbers in the applied particle response sequence, the unresponsive particle position numbers do not enter the symbolic product operation and are directly written into the mismatch migration bit. In S6-2, to converge multiple bias micro-shift directions in the individual migration table into the group migration direction for this transmission round, the individual migration table is scanned from the first position to the last position, and forward and reverse count numbers are set respectively. During processing, the forward count number is incremented by one each time a forward bias micro-shift direction appears, and the reverse count number is incremented by one each time a reverse bias micro-shift direction appears. When the forward and reverse count numbers are not equal, the direction with the larger count number is written into the group migration direction. When the forward and reverse count numbers are equal, the first non-empty direction is read from the last position of the individual migration table in reverse order from the first position, and this non-empty direction is written into the group migration direction. If the individual migration table is empty, the real-time bias value is directly written into the target bias value, this transmission round is marked as no migration round, and the same-direction shadow chain bit-taking process is skipped. In S6-3, to ensure that shadow particles not actually applied to the bias terminal of the electro-optic modulator participate in the generation of the target bias value, the population migration direction is used as the bit-taking direction. The bias values ​​of shadow particles in the shadow particle group are read one by one, and the signed shadow difference is obtained by subtracting the real-time bias value from the shadow particle bias value. During processing, shadow particles with the same sign as the population migration direction are written into the same-direction shadow chain, and the same-direction shadow chain is arranged in ascending order of bias distance. When the same-direction shadow chain is empty, the real-time bias value is used as the target bias value, and the same-direction shadow chain only... When there is one shadow particle, the bias value of the first shadow particle in the same direction shadow chain is read as the target bias value. When there are two or more shadow particles in the same direction shadow chain, the bias value of the shadow particle one position after the first shadow particle in the same direction shadow chain is read as the target bias value. The process of reading the bias value of the shadow particle one position after the first shadow particle in the same direction shadow chain and reading the first bias value after the shadow particle group is moved forward by one offset distance belongs to the same value taking process. After the target bias value is applied to the bias end of the electro-optic modulator, it is written to the write tail bit of the historical convergence bias position for the next transmission round to be read back in reverse order of writing. Through the above implementation process, the particle response sequence is first screened out by double-ended sign product operation to select the individual migration direction, and then the population migration direction is generated by forward and reverse counting. Subsequently, the target bias value is obtained by the same-direction shadow chain, so that the real applied particles provide response evidence, the shadow particles provide migration candidates, and the target bias value is then written back to the historical convergence bias position. In practical applications: if the double-ended differential sign product of the three applied particle responses is negative, positive, and negative respectively, then the bias micro-shift direction of the first and third applied particles is written into the individual migration table, and the position number of the second applied particle is written into the mismatch migration position; if the positive bias micro-shift direction appears twice and the negative bias micro-shift direction appears zero times in the individual migration table, then the population migration direction is written into the positive direction. The particle swarm algorithm then selects shadow particles with positive signed shadow differences from the shadow particle group to form a co-directional shadow chain, and reads the bias value of the first and last shadow particles of the co-directional shadow chain as the target bias value and writes it into the bias terminal of the electro-optic modulator.

[0023] Furthermore, the present invention also includes an optical transmission transmitter using an external modulation transmission link, where the encoding or modulation element outputs a transmission symbol stream within one transmission round, and the bias terminal of the electro-optic modulator uses register code values ​​to write the bias voltage, with the bias terminal code field ranging from 0 to 1023; in this transmission round, the real-time bias value 512 is read, and the historical convergence bias positions 508 and 520 are read out in reverse order of writing; the bias control channel registers 3 applied resource bits in this transmission round, and the sampling order of the main output terminal and the complementary output terminal is aligned by the transmitter's transmit counter; In S1, the encoding or modulation element outputs the transmission symbol stream 1, 1, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0 within this transmission round, and the transmission clock is sequentially written to the first to twelfth symbol bits; after adjacent symbols are XORed, the third, sixth, eighth, tenth, and twelfth symbol bits are written to the transition bit string; after the link break write is performed according to the transition bit string, the first to second symbol bits are written to the first same-value run segment, the third to fifth symbol bits are written to the second same-value run segment, the sixth to seventh symbol bits are written to the third same-value run segment, the eighth to ninth symbol bits are written to the fourth same-value run segment, the tenth to eleventh symbol bits are written to the fifth same-value run segment, and the twelfth symbol bit is written to the sixth same-value run segment, and a transmission segment base table is formed according to the transmission order; In S2, the real-time bias value 512 is first written to the first position of the particle table and a center particle is generated. The bias difference field of the center particle is written to zero. The historical convergence bias position is read in reverse order of writing, first 508 is read, 508 is subtracted from 512 to get the signed backtracking difference -4, the negative 4 is connected to the real-time bias value 512 to generate the backtracking particle 508, then the sign of the negative 4 is flipped to get the positive 4, the positive 4 is connected to the real-time bias value 512 to generate the mirror particle 516; the historical bias value 512 continues to be read. 20, 520 minus 512 yields a signed back-visit difference of 8. The back-visit particle 520 is generated by connecting 8 back to the real-time bias value 512. Then, the sign of 8 is flipped to obtain a negative 8. The negative 8 is then connected back to the real-time bias value 512 to generate a mirror particle 504. The five particles are written into the candidate bias particle group in the generation order of 512, 508, 516, 520, and 504. No particles cross the boundary of the bias code field from 0 to 1023, and no particles with duplicate bias values ​​appear. In S3, each run-length segment with the same value in the transmission segment base table is first written as a segment overlay word and a segment occupancy word. The segment overlay word is written into the internal transmit bits and adjacent transition bits of the segment. The segment occupancy word is obtained by subtracting the start symbol bit from the end symbol bit number and then adding one. The bidirectional greedy submodulus optimization algorithm uses the number of applied resource bits (3) as the number of power-on segments to be selected in this transmission round. During the forward calculation, it calculates the number of newly added one-value bits after incorporating the segment overlay word item by item starting from the empty set. The second run-length segment with the same value, the first run-length segment with the same value, and the segment occupancy word are selected in order of increasing number of newly added one-value bits from high to low and increasing number of segments from short to long. The fourth run-length segment forms a forward segment set; during reverse calculation, run-length segments with the same value are falsely deleted item by item from the transmission segment base table and the overlay word is recalculated. The second, fourth, and fifth run-length segments with the same value are retained in order of deletion loss from low to high and segment occupied word from long to short to form a reverse segment set; the second and fourth run-length segments with the same value enter the consensus segment table, and the missing part is filled by the first run-length segment with the same value as the forward candidate. The power-on segment sequence composed of the first, second, and fourth run-length segments with the same value is output according to the transmission order. In S4, the three segment numbers of the applied segment sequence are written into the applied resource bit table, and the offset distance is calculated for each candidate offset particle group; the offset distance of the central particle 512 is 0, the offset distance of the revisiting particle 508 is 4, the offset distance of the mirror particle 516 is 4, the offset distance of the revisiting particle 520 is 8, and the offset distance of the mirror particle 504 is 8; after the distance arrangement chain is formed according to 512, 508, 516, 520, and 504, the first three candidate offset particles occupy three applied resource bits and are written into the applied particle group, of which 512 is written into the holding position. Write the reverse application direction to 508 and the forward application direction to 516; write the unoccupied particles 520 and 504 into the unoccupied particle chain; 520 reads the forward application direction of the adjacent applied particle 516 in odd order and writes the forward induction direction; 504 reads the historical convergence bias position in even order and writes the same historical bias value 520 in reverse order and generates the forward return direction; the two unoccupied particles occupy the relaxation position in the induction direction and write it into the shadow particle group; the shadow particle group only writes the migration read end mark and does not output the bias micro-shift to the bias end of the electro-optic modulator. In S5, the particle group and the power-on segment sequence are sequentially used to establish a particle power-on mapping table. The central particle 512 is bound to the first run segment of the same value, the revisiting particle 508 is bound to the second run segment of the same value, and the mirror particle 516 is bound to the fourth run segment of the same value. The bias shift of the central particle 512 is 0, and the real-time bias value 512 is written in the first run segment of the same value and written back to the terminating symbol position. The bias shift of the revisiting particle 508 is -4, and the applied bias value 508 is written in the third symbol position and written back to the fifth symbol position. The bias shift of the mirror particle 516 is +4, and the applied bias value 516 is written in the eighth symbol position and written back to the ninth symbol position. The bias write completion bit is obtained by converting the number of write response samples from the bias control channel. The second and fourth run segments of the same value complete the write response before the terminating symbol position, thus entering the double-ended sampling differential process. When S5 continues to perform dual-ended sampling, the main output and complementary output read the same-order sampled values ​​according to the transmission counter; the center particle 512 does not form a bias micro-shift in the first run segment with the same value, and the differential symbols of the main output and complementary output are both written to zero; in the second run segment with the same value corresponding to the return particle 508, the current sampled value of the main output minus the previous sampled value is negative and written to negative one, and the current sampled value of the complementary output minus the previous sampled value is positive and written to one; in the fourth run segment with the same value corresponding to the mirror particle 516, the current sampled value of the main output minus the previous sampled value is positive and written to one, and the current sampled value of the complementary output minus the previous sampled value is negative and written to negative one; the three sets of dual-ended differential symbols are written to the applied particle response sequence according to the order of the applied particles; In S6, the particle swarm optimization algorithm reads the applied particle response sequence according to the applied particle order and performs a symbolic product operation on the differential symbols of the main output terminal and the complementary output terminal. The symbolic product corresponding to the central particle 512 is zero, and the particle position number is written into the mismatch migration bit. The symbolic product corresponding to the returning particle 508 is negative, and the reverse application direction is written into the individual migration table. The symbolic product corresponding to the mirror particle 516 is negative, and the forward application direction is written into the individual migration table. The reverse bias micro-shift direction appears once and the forward bias micro-shift direction appears once in the individual migration table, and the count is the same. Therefore, the first non-empty direction is read from the tail of the individual migration table, and the group migration direction is written into the positive direction. The mismatch migration bit only records the particle position number corresponding to the central particle 512, which is used to exclude the mapping combination of the same power-on segment and the same applied particle in the next launch round. It does not participate in the group migration direction counting in this launch round. When S6 continues to generate the target bias value, the bias value of the shadow particle in the shadow particle group is read with the positive direction of the group migration direction as the bit-taking direction. The shadow particle 520 minus the real-time bias value 512 yields the signed shadow difference 8, which has the same sign as the group migration direction and is written into the same-direction shadow chain. The shadow particle 504 minus the real-time bias value 512 yields the signed shadow difference negative 8, which has a different sign than the group migration direction and does not enter the same-direction shadow chain. When there is only one bit in the same-direction shadow chain, the bias value 520 of the first shadow particle in the same-direction shadow chain is read as the target bias value. The target bias value 520 is written into the bias terminal of the electro-optic modulator. At the end of this transmission round, 520 is written into the write tail bit of the historical convergence bias position for the next transmission round to be read back in reverse order of writing.

[0024] Working principle: This scheme first obtains the transmission symbol stream within the transmission round from the encoding or modulation element of the optical transmission transmitter. It then identifies the transition positions by XORing adjacent symbols and writes the intervals where the symbols remain unchanged into run segments of the same value, forming a transmission segment base table. Next, it reads the real-time bias value at the bias end of the electro-optic modulator and, combined with historical convergence bias positions, generates center particles, revisiting particles, and mirror particles to obtain a candidate bias particle swarm. Subsequently, a bidirectional greedy sub-mode optimization algorithm selects a suitable power-on segment sequence from the real transmission segments for power-on testing. Then, a relaxation-induced neighborhood search algorithm divides the candidate bias particles into an application particle group and a shadow particle group, ensuring that some particles are actually applied to the bias end while others only participate in migration calculations. During the transmitter's sustained transmission, the application particles complete the bias micro-shift according to the power-on segment sequence and read the differential responses of the main output and complementary output ends. Finally, the particle swarm algorithm generates the swarm migration direction based on the bidirectional differential symbols, extracts the target bias value from the shadow particle group, and writes it back to the bias end. In practical applications, when high-speed optical modules in data centers or metropolitan area optical transmission equipment are transmitting service data, the transmitter cannot be shut down, nor can the bias voltage be swept over a wide range to find the bias point. This solution first finds suitable segments for short-term probing from the service symbols being transmitted, and then only allows a small number of particles close to the real-time bias value to perform tiny bias changes within these segments, while observing the optical response at the main output and complementary output. If the response directions at both ends are opposite, it indicates that the bias micro-shift better reflects the change in the operating point of the electro-optic modulator, and this direction will be written into the migration basis. Shadow particles that are not actually applied participate in the generation of the target bias value according to this direction. In this way, the transmitter completes automatic bias point adjustment while continuing to transmit service optical signals, reducing optical power jitter and modulation quality fluctuations caused by all candidate bias values ​​directly impacting the transmission link.

[0025] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter, characterized in that, include: S1. Obtain the transmission symbol stream output by the encoding or modulation element within the transmission round of the optical transmission transmitter, XOR adjacent symbols to obtain the transition bit string, and write it into the same value run segment according to the zero value transition start and end, and output the transmission segment base table. S2. Read the real-time bias value at the bias end, generate the center particle with the real-time bias value, generate back visit particles from the historical convergence bias position in reverse order of writing, and generate mirror particles around the real-time bias value, and output the candidate bias particle group. S3. Using a bidirectional greedy sub-modulus optimization algorithm, the jump bit coverage difference is written in increments from the empty set. The particle number fragments are taken in descending order of coverage difference to form a forward fragment set. The run-length occupancy difference is deleted in descending order from the transmission fragment base table. The particle number fragments are taken in ascending order of occupancy difference to form a reverse fragment set. The intersection is taken to generate the power-on fragment sequence according to the sending order. S4. Based on the relaxation-induced neighborhood search algorithm, the candidate bias particles are arranged according to the bias distance from the real-time bias value. The number of particles of the powered segment is written into the applied particle group, and the remaining particles are inherited in order of parity from the near-end applied direction or the far-end return direction to generate the shadow particle group. S5. During the transmission period of the optical transmission transmitter, a bias micro-shift is applied to the bias terminal according to the sequence order of the applied particle group and the powered segment, and adjacent sampling differential is performed from the main output terminal to the complementary output terminal to generate the applied particle response sequence.

2. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 1, characterized in that: Also includes: S6. When the signs of the differentials at both ends are opposite, the individual migration direction is written using the particle swarm algorithm. The migration direction of the population is formed by the number of times the direction appears and the reverse order of writing. The shadow particle group is driven to move forward one position by the offset distance to generate the target offset value. The target offset value is applied to the offset end and written to the historical convergence offset position. The transmitter offset control result is output.

3. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 2, characterized in that: S1 includes: S1-1. Within each transmission cycle of the optical transmission transmitter, the symbol bit number is written bit by bit into the transmission symbol stream according to the transmission clock of the encoding or modulation element, and the transmission symbol pointed to by the previous symbol bit number and the transmission symbol pointed to by the next symbol bit number are used as XOR inputs to output the adjacent symbol operation chain. S1-2. For adjacent symbol operation chains, when the XOR result is one, write the next symbol bit number into the transition bit string; when the XOR result is zero, merge the next symbol bit number into the same value bit string formed by continuing the previous symbol bit number, and output the transition bit string and the same value bit string. S1-3. Based on the transition bit string, perform a chain break and write segment on the same value bit chain. Write the same value bit chain between adjacent transition bits in the transition bit string into the same value run segment, and generate a transmission segment base table according to the transmission order of the same value run segment in the transmission round.

4. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 3, characterized in that: S2 includes: S2-1. During the transmission cycle of the optical transmission transmitter, read the real-time bias value of the bias terminal of the electro-optic modulator according to the bias sampling clock, write the real-time bias value into the first position of the particle table, and set the bias difference field of the first position of the particle table to zero to generate the central particle. S2-2. Starting from the last bit of the historical convergence bias position, read back the historical bias value bit by bit, subtract the real-time bias value from the historical bias value to obtain the signed back visit difference, and connect the signed back visit difference back to the real-time bias value to generate back visit particles. S2-3. Perform sign flipping on the signed return difference of the return particles, connect the flipped signed return difference back to the real-time bias value to generate mirror particles, and write them into the particle table in the order of pairing the return particles and mirror particles to output the candidate bias particle swarm.

5. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 4, characterized in that: S3 includes: S3-1. Read the transmission segment base table within the transmission round of the optical transmission transmitter. For each run segment with the same value, scan from the start symbol bit number to the end symbol bit number. When the scan bit hits the transition bit string, write 1 in the corresponding bit of the segment cover word. When the scan bit does not hit the transition bit string, write 0 in the corresponding bit of the segment cover word. Then, generate the segment occupancy word by subtracting the start symbol bit number from the end symbol bit number and adding 1. S3-2. Using the bidirectional greedy sub-module optimization algorithm, the empty set is used as the positive fragment set. Fragments with the same value that have not entered the positive fragment set are selected one by one. The fragment cover word and the positive cover word are bitwise ORed. The positive gain word is generated by subtracting the number of bits of the previous value from the number of bits of the value after the operation. The candidate order is determined first according to the positive gain word from high to low. When the positive gain words are the same, the candidate order is rewritten according to the number of bits occupied by the fragment from short to long. The first fragment of the candidate order is written into the positive fragment set until the fragment count of the positive fragment set is consistent with the particle count of the candidate bias particle swarm.

6. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 5, characterized in that: S3 also includes: S3-3. Using the transmission segment base table as the reverse segment set, falsely delete the same run-length segments in the reverse segment set one by one and recalculate the reverse overlay word. Generate the reverse loss word by subtracting the number of bits of the false deletion from the number of bits of the previous value. First, determine the deletion order according to the reverse loss word from low to high. When the reverse loss words are the same, rewrite the deletion order according to the number of bits occupied by the segment from long to short. Move the segment with the first deletion order out of the reverse segment set until the segment count of the reverse segment set is consistent with the particle count of the candidate bias particle swarm. S3-4. Search the reverse fragment set item by item for the fragment identifiers in the forward fragment set. Write the matching fragments into the consensus fragment table. Store the missing fragments as forward candidate fragments according to the sending order. Then, store the fragments in the reverse fragment set that did not match the forward fragment set as reverse candidate fragments according to the reverse sending order. Use the forward candidate fragments and reverse candidate fragments to fill the gaps in the consensus fragment table. Output the power-on fragment sequence according to the sending order.

7. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 6, characterized in that: S4 includes: S4-1. Write the segment number in the above electric segment sequence into the applied resource bit table. For each particle, subtract the real-time bias value from the candidate bias value to obtain the signed bias difference. When the signed bias difference is negative, flip the sign and write the bias distance. When the signed bias difference is positive, write the bias distance directly. When the signed bias difference is zero, write the zero distance. Generate the distance arrangement chain in ascending order of bias distance and particle generation order. S4-2. Starting from the first position of the distance arrangement chain, occupy the same sequential resource position in the resource position table one by one. The candidate bias particles whose resource positions are occupied are written into the applied particle group. When the signed bias difference is positive, the positive application direction is written. When the signed bias difference is negative, the reverse application direction is written. When the signed bias difference is zero, the hold direction is written. After all resource positions are occupied, the unoccupied particle chain is output.

8. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 7, characterized in that: S4 also includes: S4-3. The relaxation-induced neighborhood search algorithm is used to perform direction induction on the unoccupied particle chain. In the unoccupied particle chain, the odd-order particles read the direction of the adjacent particles in the distance arrangement chain that are ahead of themselves and have been written into the applied particle group. The even-order particles read the historical convergence bias position and write the return direction in the reverse order that is the same as their own order. When the return direction is missing, the first return direction in the reverse order is read and written to generate the induced direction chain. S4-4. Drive the unoccupied particle chain to perform relaxation and placeholder operations using the induced direction chain. Write the relaxation position after the current particle in the forward induced particle and write the relaxation position before the current particle in the reverse induced particle. Keep the induced particle written to the relaxation position of the current particle in the reverse induced particle. When the relaxation position is occupied, search for the empty relaxation position along the induced direction and write the corresponding unoccupied particle. S4-5. Write the unoccupied particles that have completed the relaxation and occupancy into the shadow particle group, and mark the output end of the shadow particle group as the migration read end, so that the shadow particle group participates in the subsequent particle swarm algorithm migration calculation and does not output bias micro-shift to the bias end of the electro-optic modulator.

9. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 8, characterized in that: S5 includes: S5-1. Read the same sequence segment in the power-on segment sequence according to the particle order of the applied particle group, write the bias micro-shift of the applied particle into the starting symbol position of the same sequence segment, and generate a particle power-on mapping table. S5-2. When the optical transmission transmitter sends the start symbol position number, the real-time bias value is added to the bias micro-shift amount to form the applied bias value and written to the bias terminal of the electro-optic modulator. When the end symbol position number is sent, the real-time bias value is written back and the bias application record in the segment is output. S5-3. Based on the bias application record within the segment, read the sampled values ​​of the main output terminal and the complementary output terminal respectively, subtract the previous sampled value from the current sampled value to generate a double-ended differential symbol, and write the double-ended differential symbol into the applied particle response sequence according to the order of the applied particles.

10. The automatic control method for the bias point of an electro-optic modulator in an optical transmission transmitter according to claim 9, characterized in that: S6 includes: S6-1. Read the response sequence of applied particles according to the order of applied particles using the particle swarm algorithm. Perform a symbolic product operation between the differential symbol of the main output terminal and the differential symbol of the complementary output terminal. When the symbolic product is negative, write the bias micro-shift direction of the applied particles in the same order into the individual migration table. When the symbolic product is positive or zero, write the position number of the applied particles in the same order into the mismatch migration position. S6-2. Scan from the first position to the last position of the individual migration table. Each time the positive bias micro-shift direction appears, the positive count is incremented by one. Each time the negative bias micro-shift direction appears, the negative count is incremented by one. If the positive count and the negative count are not equal, the direction with the larger count value is written into the group migration direction. If the positive count and the negative count are equal, the first non-empty direction is read from the last position of the individual migration table and written into the group migration direction. S6-3. Using the migration direction of the population as the positioning direction, subtract the real-time bias value from the bias value of the shadow particles in the shadow particle group to obtain the signed shadow difference. Form a same-direction shadow chain by sorting the shadow particles with the same sign as the migration direction of the population in ascending order of bias distance. Read the bias value of the first and last shadow particles in the same-direction shadow chain to generate the target bias value. Apply the target bias value to the bias end of the electro-optic modulator and write it to the historical convergence bias position.

Images