A PAM4 link quality evaluation method for 800G data center interconnection

Abstract

This invention discloses a PAM4 link quality assessment method for 800G data center interconnection, relating to the field of high-speed data transmission link technology. The method includes: collecting the average four-level value, standard deviation of the four-level values, equalization residual, number of pre-FEC bit errors, number of decoded input bits, number of corrected symbols, number of uncorrectable codewords, and number of burst error events for eight physical channels; for each 100G physical channel, calculating level separability, pre-FEC bit error rate, and FEC error correction pressure, and weighting these values ​​to generate channel degradation characteristic values; calculating the average degradation value and inter-channel dispersion from the channel degradation characteristic values ​​of the eight channels to generate a port-level link quality index; comparing the port-level link quality index with a link quality index threshold to output the 800G PAM4 link quality level and link degradation type, thus achieving joint quantification of single-channel degradation and channel inconsistency.

Description

Technical Field

[0001] This invention relates to the field of high-speed data transmission link technology, and in particular to a PAM4 link quality assessment method for 800G data center interconnection. Background Technology

[0002] With the exponential growth of traffic within data centers, 800G Ethernet interfaces, with their high bandwidth density, are gradually becoming the core solution for interconnecting hyperscale data centers. An 8×100G parallel architecture based on four-level pulse amplitude modulation (PAM4) achieves a doubling of port bandwidth with a limited number of channels. However, the inherent three-eye diagram characteristics of PAM4 modulation make the link more sensitive to noise, nonlinear distortion, and inter-channel crosstalk. Existing link quality assessments typically rely on discrete metrics such as bit error rate before forward error correction (FEC) or received signal amplitude, judging link health by setting empirical thresholds. Such methods have low implementation complexity and can provide a preliminary indication of channel transmission impairments.

[0003] However, conventional evaluation methods have limitations in multi-physical dimension fusion and degradation pattern identification. On the one hand, they often rely on single bit error rate or level statistical characteristics, failing to effectively integrate physical layer PAM4 four-level separability, equalization residuals, and link layer FEC multi-level error correction statistics, resulting in a lack of comprehensive characterization of the link's true carrying capacity. On the other hand, conventional methods typically perform independent threshold judgments on eight parallel channels, lacking joint analysis of degradation dispersion between channels, making it difficult to distinguish between overall transmission environment aging and single-channel local faults, and also unable to quantify the potential impact of sudden bit error events on link stability. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a PAM4 link quality assessment method for 800G data center interconnection to solve the problems of existing technologies, such as the single assessment dimension, difficulty in distinguishing between local physical channel degradation and overall link aging, lack of joint quantification of FEC error correction pressure and sudden bit error events, and inability to provide port-level fine-grained early warning.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a PAM4 link quality assessment method for 800G data center interconnects, comprising: during the operation of the 800G PAM4 link, collecting the average value of PAM4 four-level signals, the standard deviation of PAM4 four-level signals, the equalization residual, the number of pre-FEC bit errors, the number of FEC decoding input bits, the number of FEC corrected symbols, the number of FEC uncorrectable codewords, and the number of burst error events for eight physical channels at monitoring cycles; for each 100G physical channel, calculating the level separability based on the sum of the difference between adjacent PAM4 average values ​​and the standard deviation of PAM4 signals; calculating the pre-FEC bit error rate based on the number of pre-FEC bit errors and the number of FEC decoding input bits; and calculating the FEC correction rate based on the pre-FEC correction rate. The FEC error correction pressure is calculated based on the number of symbols, the number of FEC uncorrectable codewords, and the number of FEC decoding input bits. Channel degradation characteristic values ​​are generated by weighting the level separability, pre-FEC bit error rate, FEC error correction pressure, equalization residual, and the number of burst error events. The average degradation value and inter-channel dispersion are calculated from the channel degradation characteristic values ​​of eight 100G physical channels. A port-level link quality index is generated based on the average degradation value, inter-channel dispersion, and the number of FEC uncorrectable codewords. The port-level link quality index is compared with a link quality index threshold to output the 800G PAM4 link quality level and link degradation type.

[0008] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the PAM4 four-level average value, PAM4 four-level standard deviation, and equalization residual include: after the 800G PAM4 link completes link training, clock recovery, and FEC alignment, a periodic pulse is generated by the local management clock, and the periodic pulse is used as the acquisition trigger signal for the monitoring cycle; when each periodic pulse arrives, the link monitoring logic reads the receiving side statistical data of the first to eighth physical channels sequentially according to the physical channel number, and forms the acquisition results for the current monitoring cycle respectively; for any 100G physical channel, after the receiving end completes clock data recovery, it performs level judgment on the equalized PAM4 sample value, and classifies the PAM4 sample value into level 0, level 1, level 2, and level 3 according to the judgment threshold; for any PAM4 level among level 0, level 1, level 2, and level 3, the PAM4 sample value belonging to the same PAM4 level in the current monitoring cycle is classified as follows: The summation of all sampled values ​​at the 4-level is divided by the number of samples corresponding to the current PAM4 level to obtain the average PAM4 level in the current monitoring period. For any PAM4 level among the 0th, 1st, 2nd, and 3rd levels, the deviation between each sampled value and the corresponding average PAM4 level is calculated, and the square root of the average of the squared deviations is taken to obtain the standard deviation of the current PAM4 level in the current monitoring period. For any 100G physical channel, after the receiver completes feedforward equalization, decision feedback equalization, or DSP equalization, it obtains the residual signal between the equalization output value and the corresponding decision level, and calculates the square mean of the residual signal in the current monitoring period to obtain the equalization residual of the current 100G physical channel.

[0009] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the number of pre-FEC error bits, the number of FEC decoding input bits, the number of FEC corrected symbols, the number of FEC uncorrectable codewords, and the number of burst error events include: for any 100G physical channel, setting an FEC pre-FEC error counter and an FEC decoding input bit counter on the input side of the FEC decoder, counting the number of error bits detected before entering FEC decoding, and counting the total number of bits entering the FEC decoding process; at the end of the current monitoring period, reading the current 100G physical... The system reads the cumulative FEC error rate and cumulative FEC decoding input bit value of the channel at the end of the current cycle, and also reads the cumulative FEC error rate and cumulative FEC decoding input bit value of the current 100G physical channel at the end of the previous cycle. It then subtracts the cumulative FEC error rate at the end of the current cycle from the cumulative FEC error rate at the end of the previous cycle to obtain the number of FEC errors before the current monitoring cycle. Finally, it subtracts the cumulative FEC decoding input bit value at the end of the current cycle from the cumulative FEC decoding input bit value at the end of the previous cycle to obtain the number of FEC errors before the current monitoring cycle. Input bit count; for any 100G physical channel, read the cumulative value of FEC corrected symbols and the cumulative value of FEC uncorrectable codewords at the output side of the FEC decoder; subtract the cumulative value of FEC corrected symbols at the end of the current cycle from the cumulative value of FEC corrected symbols at the end of the previous cycle to obtain the number of FEC corrected symbols in the current monitoring cycle; subtract the cumulative value of FEC uncorrectable codewords at the end of the current cycle from the cumulative value of FEC uncorrectable codewords at the end of the previous cycle to obtain the number of FEC uncorrectable codewords in the current monitoring cycle; divide each monitoring cycle into sub-windows, and respectively... Calculate the pre-FEC bit error rate of any 100G physical channel in each sub-window; calculate the average bit error rate of the sub-window in the current monitoring period based on the pre-FEC bit error rate in each sub-window; for any sub-window, compare the pre-FEC bit error rate in the current sub-window with three times the average bit error rate of the sub-window; when the pre-FEC bit error rate in the current sub-window is greater than three times the average bit error rate of the sub-window, the current sub-window is determined to have experienced a burst bit error event; sum up the number of sub-windows determined to have experienced burst bit error events in the current monitoring period to obtain the number of burst bit error events of the current 100G physical channel in the current monitoring period.

[0010] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the step of calculating the level separability for each 100G physical channel based on the sum of the difference between adjacent PAM4 electrical average values ​​and the standard deviation of PAM4 levels involves the following steps: After the current monitoring period ends, the link monitoring logic sequentially reads the acquisition results from the first to the eighth physical channels; for any 100G physical channel, it reads the four PAM4 electrical average values ​​and four PAM4 level standard deviations for that 100G physical channel within the current monitoring period; and it combines the PAM4 four-level average values ​​and PAM4 four-level standard deviations of the same 100G physical channel... The standard deviations are arranged in ascending order of level amplitude. The first PAM4 level and the second PAM4 level form the first adjacent level pair, the second PAM4 level and the third PAM4 level form the second adjacent level pair, and the third PAM4 level and the fourth PAM4 level form the third adjacent level pair. For any adjacent level pair, the difference between the average values ​​of the two adjacent PAM4 levels is calculated, and the sum of the standard deviations of the two adjacent PAM4 levels is calculated. The difference is divided by the sum of the standard deviations to obtain the local separability value of the adjacent level pair. The average of the local separability values ​​of the three adjacent level pairs is used to obtain the level separability of the current 100G physical channel in the current monitoring period.

[0011] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the specific steps for calculating the pre-FEC bit error rate based on the pre-FEC bit error count and the FEC decoding input bits are as follows: After the current monitoring period ends, the link monitoring logic reads the pre-FEC bit error count and the FEC decoding input bits of any 100G physical channel within the current monitoring period; divides the pre-FEC bit error count of the current 100G physical channel by the FEC decoding input bits to obtain the pre-FEC bit error rate of the current 100G physical channel; when the FEC decoding input bits are zero, the link monitoring logic marks the data of the current 100G physical channel within the current monitoring period as invalid sampling.

[0012] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the specific steps for calculating the FEC error correction pressure based on the number of FEC corrected symbols, the number of FEC uncorrectable codewords, and the number of FEC decoding input bits are as follows: After the current monitoring period ends, the link monitoring logic reads the number of FEC corrected symbols, the number of FEC uncorrectable codewords, and the number of FEC decoding input bits for any 100G physical channel during the current monitoring period; multiplies the number of FEC uncorrectable codewords by the uncorrectable penalty coefficient to obtain the uncorrectable penalty amount; adds the uncorrectable penalty amount to the number of FEC corrected symbols and divides it by the number of FEC decoding input bits to obtain the FEC error correction pressure of the current 100G physical channel; and after performing level separability, FEC pre-error rate, and FEC error correction pressure calculations on the first to eighth physical channels during the current monitoring period, the basic quality characteristic record of each 100G physical channel is obtained.

[0013] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the step of generating a channel degradation characteristic value by weighting level separability, pre-FEC bit error rate, FEC error correction pressure, equalization residual, and burst bit error event count is as follows: Within the current monitoring period, the link monitoring logic sequentially reads the level separability, pre-FEC bit error rate, FEC error correction pressure, equalization residual, and burst bit error event count for the first to eighth physical channels; using level separability as the denominator, the reciprocal of level separability is taken to obtain the level separation degradation amount; the pre-FEC bit error rate is multiplied by the bit error rate amplification factor. Add 1 to the result and take the logarithm to the base 10 to obtain the pre-FEC bit error degradation. Multiply the FEC error correction pressure by the error correction pressure amplification factor, add 1, and take the logarithm to the base 10 to obtain the FEC error correction pressure degradation. Divide the equilibrium residual by the sum of the equilibrium residual and 1 to obtain the equilibrium residual degradation. Divide the number of burst bit error events by the sum of the number of burst bit error events and 1 to obtain the burst bit error degradation. Weight the level separation degradation, pre-FEC bit error degradation, FEC error correction pressure degradation, equilibrium residual degradation, and burst bit error degradation according to a fixed weight to obtain the channel degradation characteristic value of each 100G physical channel in the current monitoring period.

[0014] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the calculation of the average degradation value and inter-channel dispersion from the channel degradation characteristic values ​​of eight 100G physical channels involves the following steps: After calculating the channel degradation characteristic values ​​of the first to eighth physical channels, the channel degradation characteristic values ​​of the eight 100G physical channels are summed and divided to obtain the average degradation value within the current monitoring period; after obtaining the average degradation value, the difference between the channel degradation characteristic value and the average degradation value of each 100G physical channel is calculated; the differences of the eight 100G physical channels are squared, and the average and square root of the squared results are taken to obtain the inter-channel dispersion within the current monitoring period; the channel degradation characteristic values, average degradation value, and inter-channel dispersion of the first to eighth physical channels are used as the channel degradation statistics for the current monitoring period.

[0015] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the specific steps for generating a port-level link quality index based on the average degradation value, inter-channel dispersion, and FEC uncorrectable codewords are as follows: Within the current monitoring period, after obtaining the average degradation value and inter-channel dispersion of eight 100G physical channels, the link monitoring logic reads the FEC uncorrectable codewords from the first to the eighth physical channels; sums the FEC uncorrectable codewords from the first to the eighth physical channels within the current monitoring period, and performs bounded normalization on the summation result to obtain the port-level uncorrectable risk quantity; the average degradation value, inter-channel dispersion, and port-level uncorrectable risk quantity are combined according to a fixed penalty coefficient to generate the port-level link quality index for the current monitoring period.

[0016] As a preferred embodiment of the PAM4 link quality assessment method for 800G data center interconnection described in this invention, the specific steps of comparing the port-level link quality index with a link quality index threshold to output the 800G PAM4 link quality level and link degradation type are as follows: Setting a first link quality index threshold, a second link quality index threshold, and a third link quality index threshold, wherein the first link quality index threshold is less than the second link quality index threshold, and the second link quality index threshold is less than the third link quality index threshold; comparing the port-level link quality index within the current monitoring period with the first, second, and third link quality index thresholds sequentially; when the port-level link quality index is less than the first link quality index threshold, outputting an 800G PAM4 link quality level as excellent; when the port-level link quality index is greater than or equal to the first link quality index threshold and less than the second link quality index threshold, outputting an 800G PAM4 link quality level as usable; when the port-level link quality index is greater than or equal to the second link quality index threshold and less than the third link quality index threshold, outputting an 800G PAM4 link quality level as usable. PAM4 link quality level is at a degradation warning level; when the port-level link quality index is greater than or equal to the third link quality index threshold, 800G is output. The PAM4 link quality level is unstable. When the sum of the number of uncorrectable FEC codewords in the eight 100G physical channels is greater than 0, the output link degradation type is FEC uncorrectable risk type. When the sum of the number of uncorrectable FEC codewords in the eight 100G physical channels is equal to 0, the channel dispersion is greater than the dispersion threshold, and the channel degradation characteristic value of any 100G physical channel is greater than the average degradation value and twice the sum of the channel dispersion, the output link degradation type is local physical channel degradation type. When the sum of the number of uncorrectable FEC codewords in the eight 100G physical channels is equal to 0, the channel dispersion is not greater than the dispersion threshold, and the average degradation value is greater than the average degradation threshold, the output link degradation type is overall degradation type. When the sum of the number of uncorrectable FEC codewords in the eight 100G physical channels is equal to 0, the channel dispersion is not greater than the dispersion threshold, and the average degradation value is not greater than the average degradation threshold, the output link degradation type is low-risk stable type.

[0017] The beneficial effects of this invention are as follows: By collecting PAM4 statistics, balanced residuals, FEC counts, and burst errors from eight channels and calculating level separability, a systematic collection of multi-dimensional damage is achieved, overcoming the one-sidedness of single-index evaluation; by generating channel degradation feature values ​​through weighted fusion and calculating the average degradation value and dispersion, joint quantification of single-channel degradation and channel inconsistency is achieved, solving the problem of difficulty in distinguishing between overall aging and local faults; by generating port-level link quality index and outputting quality levels based on comparison thresholds, and outputting degradation types based on parameters, refined grading and degradation mode diagnosis are achieved, improving the accuracy of early warning; by determining burst events based on separability of the ratio of adjacent level difference to standard deviation and comparison of sub-window errors, sensitive quantification of eye diagram closure and burst errors is achieved, compensating for the deficiency of delayed response to burst degradation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the PAM4 link quality assessment method for 800G data center interconnects.

[0020] Figure 2 A flowchart generated for the current period's data collection results.

[0021] Figure 3 A flowchart for generating channel degradation feature values.

[0022] Figure 4 A flowchart showing the output of port-level link quality index and link degradation type.

[0023] Figure 5 This is a comparative monitoring chart of eight 100G physical channels under a scenario of local physical channel degradation.

[0024] Figure 6 This is a comparison chart of continuous monitoring of port-level link quality index and bit error rate before traditional FEC in an overall degraded scenario. Detailed Implementation

[0025] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0026] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0027] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0028] Reference Figures 1-6 As one embodiment of the present invention, this embodiment provides a PAM4 link quality assessment method for 800G data center interconnection, including the following steps:

[0029] S1. During the operation of the 800G PAM4 link, the average value of PAM4 four-level, standard deviation of PAM4 four-level, equalization residual, number of FEC pre-errors, number of FEC decoding input bits, number of FEC corrected symbols, number of FEC uncorrectable codewords, and number of burst error events of the eight physical channels are collected at fixed monitoring cycles.

[0030] Furthermore, in this embodiment, the 800G PAM4 link is an 800G Ethernet link for data center interconnection. The 800G PAM4 link is carried in parallel by eight 100G physical channels, which are respectively referred to as the first 100G physical channel to the eighth 100G physical channel.

[0031] After the 800G PAM4 link enters normal transmission status, the link monitoring logic initiates a fixed monitoring cycle of 1 second. After each 1-second monitoring cycle, the link monitoring logic performs a data latch operation on each of the eight 100G physical channels and writes the collected PAM4 four-level average value, four-level standard deviation, equalization residual, number of pre-FEC bit errors, number of FEC decoding input bits, number of FEC corrected symbols, number of FEC uncorrectable codewords, and number of burst error events into the corresponding monitoring register or cache address for subsequent link quality calculations.

[0032] Specifically, after the 800G PAM4 link completes link training, clock recovery, and FEC alignment, a 1-second periodic pulse is generated using a local management clock. Each time a 1-second periodic pulse arrives, the current time is recorded as... The end time of the previous cycle is recorded as .

[0033] The link monitoring logic reads the receiver-side statistical data of eight 100G physical channels sequentially according to their physical channel numbers, and obtains the data for the first physical channel. The acquisition results generated by a 100G physical channel within the current 1-second monitoring cycle.

[0034] The collected results include the average value of PAM4 four-level signals, the standard deviation of PAM4 four-level signals, the equalization residual, the number of FEC pre-errors, the number of FEC decoding input bits, the number of FEC corrected symbols, the number of FEC uncorrectable codewords, and the number of burst error events for the 100G physical channel.

[0035] Furthermore, regarding the first A 100G physical channel performs level judgment on the equalized PAM4 sampled values ​​after the clock data is recovered at the receiving end.

[0036] Since the PAM4 signal contains four amplitude levels, in this embodiment, the four PAM4 levels are respectively denoted as level 0, level 1, level 2, and level 3.

[0037] Within the current 1-second monitoring period, the receiver classifies each PAM4 sample value into its corresponding level category based on the decision threshold.

[0038] Sum all sampled values ​​belonging to the same PAM4 level within the current 1-second monitoring period, and then divide by the number of samples corresponding to the level to obtain the average amplitude of the PAM4 level within the current period, i.e., the average value of the PAM4 level.

[0039] Meanwhile, the deviation between each sampled value and the corresponding average value of PAM4 is first calculated, and then the average and square root of the squared deviation are taken to obtain the amplitude dispersion of PAM4 level, that is, the standard deviation of PAM4 level.

[0040] Furthermore, for the first A 100G physical channel is used. After the receiver completes feedforward equalization, decision feedback equalization, or DSP equalization, it obtains the residual signal between the equalized output value and the corresponding decision level.

[0041] Within the current 1-second monitoring cycle, the first... The equalized output sample values ​​are denoted as The corresponding decision level amplitude is recorded as The number of samples used for residual statistics in the current period is denoted as . .

[0042] No. Equalized residuals of 100G physical channels , represented as: ;

[0043] Furthermore, for the first A 100G physical channel is provided, with an FEC pre-error counter and an FEC decoder input bit counter set on the input side of the FEC decoder.

[0044] The FEC pre-error counter is used to count the number of erroneous bits detected before entering the FEC decoding process, while the FEC decoding input bit counter is used to count the total number of bits entering the FEC decoding process.

[0045] At the end of the current 1-second monitoring cycle, read the... A 100G physical channel at time The cumulative FEC error value and the cumulative FEC decoding input bit value are read simultaneously, along with the cumulative FEC error value and the cumulative FEC decoding input bit value at the end of the previous cycle.

[0046] Then, the first A 100G physical channel at time The difference between the cumulative FEC bit error rate before the end of the previous cycle and the cumulative FEC bit error rate before the end of the previous cycle is used as the number of FEC bit errors before the end of the current 1-second monitoring cycle.

[0047] Then, the first A 100G physical channel at time The difference between the cumulative value of FEC decoding input bits and the cumulative value of FEC decoding input bits at the end of the previous cycle is used as the number of FEC decoding input bits in the current 1-second monitoring cycle.

[0048] Furthermore, for the first A 100G physical channel is used to read the cumulative value of FEC correction symbols and the cumulative value of FEC uncorrectable codewords at the output side of the FEC decoder.

[0049] The number of FEC corrected symbols represents the number of erroneous symbols that the FEC decoder has corrected within the current 1-second monitoring period; the number of FEC uncorrectable codewords represents the number of codewords that the FEC decoder has determined cannot be correctly recovered within the current 1-second monitoring period.

[0050] At the end of the current 1-second monitoring cycle, read the... A 100G physical channel at time FEC correction sign cumulative value and FEC uncorrectable codeword cumulative value And read the cumulative value of the FEC correction symbol at the end of the previous cycle. and FEC uncorrectable codeword cumulative value .

[0051] Then, the first A 100G physical channel at time The difference between the cumulative value of FEC correction symbols and the cumulative value of FEC correction symbols at the end of the previous cycle is used as the number of FEC correction symbols in the current 1-second monitoring cycle.

[0052] Then, the first A 100G physical channel at time The difference between the cumulative value of FEC uncorrectable codewords and the cumulative value of FEC uncorrectable codewords at the end of the previous cycle is used as the number of FEC uncorrectable codewords in the current 1-second monitoring cycle.

[0053] Furthermore, in order to identify the concentrated occurrence of bit error degradation in a short period of time, the 1-second monitoring period is further divided into several fixed equal-length sub-windows within each 1-second fixed monitoring period.

[0054] In this embodiment, it is preferable to divide the 1s monitoring period into 100 10ms sub-windows.

[0055] For the A 100G physical channel was used to count the number of FEC-preceding bit errors in each sub-window.

[0056] Based on the statistical count of the number of bit errors before FEC in each sub-window, calculate the first... The average number of bit errors per sub-window within the current 1-second monitoring period for a 100G physical channel.

[0057] Then, based on the error judgment rules, determine the sudden error event.

[0058] Specifically, the error rate determination rule is as follows: For any sub-window, the number of bit errors before FEC in the sub-window is compared with the "three times the average bit error rate of the sub-window". When the number of bit errors before FEC in a sub-window is greater than the three times the average bit error rate of the sub-window, it means that the number of bit errors in the current sub-window is higher than the average bit error rate in the current 1-second fixed monitoring period. Therefore, the current sub-window is determined to have experienced a sudden bit error event, and the sudden bit error determination result of the current sub-window is recorded as 1. When the number of bit errors before FEC in a sub-window is less than or equal to the three times the average bit error rate of the sub-window, it means that the number of bit errors in the current sub-window has not significantly deviated from the average bit error rate in the current 1-second fixed monitoring period. Therefore, the current sub-window is determined to have not experienced a sudden bit error event, and the sudden bit error determination result of the current sub-window is recorded as 0.

[0059] Finally, the number of all sub-windows that were identified as sudden bit errors within the current 1-second monitoring period is added together to obtain the number of sudden bit error events.

[0060] Furthermore, after the current 1-second monitoring cycle ends, the link monitoring logic will use the average value of PAM4 four levels, the standard deviation of PAM4 four levels, the equalization residual, the number of bit errors before FEC, the number of FEC decoding input bits, the number of FEC corrected symbols, the number of FEC uncorrectable codewords, and the number of burst error events of each 100G physical channel as the current cycle acquisition results of the current 100G physical channel.

[0061] After collecting data from all eight 100G physical channels, the complete link status of the 800G PAM4 link within the current 1-second monitoring period is obtained.

[0062] S2. For each 100G physical channel, calculate the level separability based on the sum of the difference between the average values ​​of adjacent PAM4 electrical values ​​and the standard deviation of PAM4 levels. Calculate the FEC error rate based on the number of bit errors before FEC and the number of FEC decoding input bits. Calculate the FEC error correction pressure based on the number of FEC corrected symbols, the number of FEC uncorrectable codewords, and the number of FEC decoding input bits.

[0063] Furthermore, after the current 1-second fixed monitoring cycle ends, the link monitoring logic sequentially reads the acquisition results from the first physical channel to the eighth physical channel.

[0064] For any 100G physical channel, read the four PAM4 electrical average values, four PAM4 level standard deviations, number of FEC pre-errors, number of FEC decoding input bits, number of FEC corrected symbols, and number of FEC uncorrectable codewords for the 100G physical channel within the current 1s fixed monitoring period.

[0065] Specifically, the four average PAM4 voltage levels correspond to the first PAM4 level, the second PAM4 level, the third PAM4 level, and the fourth PAM4 level, respectively; the standard deviations of the four PAM4 voltage levels correspond to the first PAM4 level, the second PAM4 level, the third PAM4 level, and the fourth PAM4 level, respectively.

[0066] After the reading is completed, the link monitoring logic arranges the average value and standard deviation of the PAM4 four-level signals of the same 100G physical channel from low to high according to the level amplitude.

[0067] The first PAM4 level and the second PAM4 level are combined to form a first adjacent level pair, the second PAM4 level and the third PAM4 level are combined to form a second adjacent level pair, and the third PAM4 level and the fourth PAM4 level are combined to form a third adjacent level pair.

[0068] For each adjacent level pair, first calculate the difference between the average values ​​of two adjacent PAM4 levels, then calculate the sum of the standard deviations of the two adjacent PAM4 levels, and finally divide the difference by the sum of the standard deviations to obtain the locally separable value of the adjacent level pair.

[0069] Locally separable values ​​are represented as: ;

[0070] in, Indicates the first The 100G physical channel is in the current 1-second fixed monitoring cycle. Locally separable values ​​of adjacent level pairs This represents the electrical average of the higher PAM4 level in an adjacent level pair. This represents the electrical average of the lower PAM4 level in an adjacent level pair. This represents the standard deviation of the higher PAM4 level in adjacent level pairs. This represents the standard deviation of the lower PAM4 level in adjacent level pairs.

[0071] It should be noted that the greater the difference in average values ​​of adjacent PAM4 levels and the smaller the sum of the standard deviations of adjacent PAM4 levels, the easier it is for the receiver to distinguish adjacent PAM4 levels, and the larger the corresponding local separability value. Conversely, when the interval between adjacent PAM4 levels decreases or the level noise spread increases, the local separability value decreases, indicating that there is a risk of eye diagram closure for adjacent level pairs.

[0072] After obtaining the locally separable values ​​of three adjacent level pairs, the average of the three locally separable values ​​is obtained to get the first... Level separability of a 100G physical channel within the current 1s fixed monitoring period.

[0073] It should be noted that the local separable values ​​of the first adjacent level pair, the second adjacent level pair, and the third adjacent level pair are averaged to obtain the overall reflection of the first adjacent level pair. The level separation capability of a 100G physical channel PAM4 four-level signal is an indicator. The greater the level separation capability, the clearer the spacing between the four PAM4 levels and the less noise spread; the smaller the level separation capability, the more likely there are overlap, compression, or noise spread issues between the four PAM4 levels.

[0074] Furthermore, for any 100G physical channel, the link monitoring logic reads the first... The number of FEC-pre-errors and the number of FEC decoding input bits for each 100G physical channel within the current 1-second fixed monitoring period. Dividing the number of FEC-pre-errors by the number of FEC decoding input bits yields the result of the first... FEC-preceding bit error rate of a 100G physical channel.

[0075] It should be noted that within the current fixed 1-second monitoring period, the ratio of the number of erroneous bits detected before entering FEC decoding to the total number of bits entering the FEC decoder is calculated to obtain the original bit error rate of the link before FEC error correction. The higher the bit error rate before FEC, the more severe the original bit error at the input of the FEC decoder.

[0076] In practical implementation, if the number of FEC decoding input bits is zero, it indicates that the first bit... If no valid FEC input data is generated within the current 1-second fixed monitoring period of the 100G physical channel, the link monitoring logic will then... The current periodic data of a 100G physical channel is marked as invalid sampling and will not participate in the subsequent port-level link quality index calculation to avoid misjudgment due to no service or no valid bit stream.

[0077] For any 100G physical channel, the link monitoring logic reads the first... The number of FEC correction symbols, the number of FEC uncorrectable codewords, and the number of FEC decoding input bits for each 100G physical channel within the current fixed 1-second monitoring period. Since uncorrectable FEC codewords better reflect the risk of link instability than corrected symbols, a higher penalty weight is applied to the number of uncorrectable FEC codewords than to the number of FEC correction symbols when calculating FEC error correction pressure.

[0078] Specifically, the FEC error correction pressure is obtained by multiplying the number of uncorrectable codewords by the uncorrectable penalty coefficient, adding it to the number of FEC corrected symbols, and then dividing by the number of FEC decoding input bits.

[0079] FEC's pressure to correct errors is expressed as: ;

[0080] in, Indicates the first The FEC error correction pressure of a 100G physical channel within the current fixed 1-second monitoring period Indicates the first The number of FEC correction symbols per 100G physical channel within the current 1-second fixed monitoring period. Indicates the first The number of uncorrectable FEC codewords per 100G physical channel within the current 1-second fixed monitoring period. Indicates the first The number of FEC decoding input bits per 100G physical channel within the current 1-second fixed monitoring period. This represents the uncorrectable penalty coefficient, with the preferred value being... .

[0081] It should be noted that the uncorrectable penalty coefficient is determined based on the maximum number of correctable symbols in a single codeword of the adopted FEC code type, preferably 10 times the maximum number of correctable symbols, and in this embodiment, it is taken as... This is used to make any FEC uncorrectable codeword contribute more to the FEC error correction pressure than a normal corrected symbol.

[0082] It should be noted that when the number of FEC correction symbols increases, it indicates that the link is consuming more FEC error correction pressure; when the number of FEC uncorrectable codewords appears, it indicates that the link has exceeded the risk of FEC error correction pressure. Therefore, by increasing the uncorrectable penalty coefficient, the influence of FEC uncorrectable codes on FEC error correction pressure can be improved, so that FEC error correction pressure can more sensitively reflect the uncorrectable risk.

[0083] Within the current fixed 1-second monitoring cycle, the level separability, bit error rate before FEC, and FEC error correction pressure are calculated for all physical channels from the first to the eighth physical channel, resulting in the level separability, bit error rate before FEC, and FEC error correction pressure for each 100G physical channel.

[0084] For the A 100G physical channel will be used for the first The basic quality characteristics of each 100G physical channel are recorded within the current 1-second fixed monitoring period to obtain the basic quality characteristic records.

[0085] The basic quality characteristics recorded include level separability, bit error rate before FEC, and FEC error correction pressure.

[0086] S3. Weight the level separability, FEC pre-error rate, FEC error correction pressure, equalization residual and number of burst error events to generate channel degradation characteristic values. Calculate the average degradation value and inter-channel dispersion from the channel degradation characteristic values ​​of the eight 100G physical channels.

[0087] Furthermore, in this embodiment, within each fixed 1-second monitoring cycle, the link monitoring logic generates the first [previous] error based on the level separability, pre-FEC bit error rate, FEC error correction pressure, equalization residual, and number of burst error events already obtained for each 100G physical channel. Channel degradation characteristics of a 100G physical channel.

[0088] Then, the link monitoring logic statistically analyzes the channel degradation characteristics of the eight 100G physical channels to obtain the average degradation value and channel dispersion of the 800GPAM4 link within the current 1-second fixed monitoring period.

[0089] Among them, the channel degradation characteristic value is used to characterize the overall degradation degree of a single 100G physical channel within the current fixed 1s monitoring period; the average degradation value is used to characterize the overall degradation level of the eight 100G physical channels; and the channel dispersion is used to characterize whether the degradation degree among the eight 100G physical channels is consistent.

[0090] Specifically, after the current 1-second fixed monitoring cycle ends, the link monitoring logic sequentially reads the level separability, bit error rate before FEC, FEC error correction pressure, equalization residual, and number of burst error events for the first to eighth physical channels.

[0091] Since a higher level separability means that the four PAM4 levels are easier to distinguish and the link quality is better, while a higher channel degradation characteristic value means more severe degradation, it is necessary to first convert the level separability into the level separation degradation amount.

[0092] In this embodiment, the level separability is used as the denominator, and the reciprocal of the level separability is taken to obtain the level separation degradation amount.

[0093] It should be noted that when the level separability is high, the level separation degradation is small; when the level separability decreases, the level separation degradation increases, thus keeping the level separation degradation consistent with the direction of link degradation.

[0094] Since the bit error rate before FEC is usually a small value, if it is directly included in the weighted calculation, it is easy for the numerical value to be masked by other indicators.

[0095] In order to maintain the effective contribution of the pre-FEC bit error rate to the channel degradation characteristic value, this embodiment performs logarithmic compression on the pre-FEC bit error rate.

[0096] Specifically, multiply the bit error rate before FEC by the bit error rate amplification factor, add 1, and then take the logarithm to the base 10 to obtain the bit error rate degradation before FEC.

[0097] The pre-FEC bit error degradation is expressed as: ;

[0098] in, Indicates the first The pre-FEC bit error rate degradation of a 100G physical channel within the current 1-second fixed monitoring period. Indicates the first The bit error rate before FEC for a 100G physical channel within the current fixed 1-second monitoring period. This indicates the magnification factor.

[0099] It should be noted that when the bit error rate before FEC is low, the change in bit error rate degradation before FEC is relatively gradual; when the bit error rate before FEC increases, the bit error rate degradation before FEC increases accordingly. Using logarithmic scaling compression can avoid the abnormal amplification of channel degradation characteristics due to extreme bit error rates, while preserving the trend of the bit error rate before FEC increasing with link degradation.

[0100] Similar to the bit error rate before FEC, the error correction pressure of FEC is usually also a small value.

[0101] To ensure that the FEC error correction pressure makes a stable contribution to the overall evaluation, this embodiment performs logarithmic compression on the FEC error correction pressure.

[0102] Specifically, the FEC error correction pressure is multiplied by the error correction pressure amplification factor, then 1 is added, and the logarithm to the base 10 is taken to obtain the FEC error correction pressure degradation.

[0103] The degradation of FEC error correction pressure is expressed as: ;

[0104] in, Indicates the first The degradation of FEC error correction pressure on a 100G physical channel within the current 1-second fixed monitoring period. Indicates the first The FEC error correction pressure of a 100G physical channel within the current 1-second fixed monitoring cycle.

[0105] It should be noted that when the FEC error correction pressure is low, the FEC error correction pressure degradation is small; when the FEC error correction pressure increases or an uncorrectable FEC codeword appears, the FEC error correction pressure degradation increases accordingly, so that the error correction load of the FEC decoder can be included in the channel degradation characteristic value.

[0106] The equalization residual is used to reflect the degree to which the equalization output value at the receiver deviates from the ideal PAM4 decision level.

[0107] To avoid inconsistencies in the weighting results due to different chip quantization scales, this embodiment performs bounded normalization on the balanced residuals.

[0108] Specifically, the equilibrium residual is divided by the sum of the equilibrium residual and 1 to obtain the degradation amount of the equilibrium residual.

[0109] The degradation of the equilibrium residual is expressed as: ;

[0110] in, Indicates the first The amount of balanced residual degradation of a 100G physical channel within the current fixed 1-second monitoring period. Indicates the first The balanced residual of a 100G physical channel within the current 1-second fixed monitoring period.

[0111] It should be noted that when the equilibrium residual is small, the degradation of the equilibrium residual is close to 0; when the equilibrium residual increases, the degradation of the equilibrium residual gradually approaches 1, thereby limiting the equilibrium residual to between 0 and 1 and improving the comparability between different 100G physical channels.

[0112] The number of sudden error events is used to reflect the frequency of concentrated error occurrences within a short period of time within the current 1-second fixed monitoring cycle.

[0113] To avoid excessive impact on channel degradation characteristics when the number of sudden bit error events is too large, this embodiment performs bounded normalization on the number of sudden bit error events.

[0114] Specifically, the burst error rate is obtained by dividing the number of burst error events by the sum of the number of burst error events and 1. The burst error rate is expressed as: ;

[0115] in, Indicates the first The amount of sudden bit error degradation in a 100G physical channel within the current fixed 1-second monitoring period. Indicates the first The number of sudden bit error events in a 100G physical channel within the current fixed 1-second monitoring period.

[0116] It should be noted that when no sudden bit error event occurs, the sudden bit error degradation is 0; as the number of sudden bit error events increases, the sudden bit error degradation gradually approaches 1, so that the number of sudden bit error events can stably participate in the comprehensive weighted calculation.

[0117] After obtaining the level separation degradation, FEC pre-error degradation, FEC error correction pressure degradation, equalization residual degradation, and burst error degradation, a weighted sum is performed according to a fixed weight to obtain the channel degradation characteristic value of each 100G physical channel within the current 1s fixed monitoring period.

[0118] Channel degradation eigenvalues ​​are represented as follows: ;

[0119] in, Indicates the first The channel degradation characteristic value of a 100G physical channel within the current fixed 1-second monitoring period. These represent the weights for level separation degradation, pre-FEC bit error degradation, FEC error correction pressure degradation, balanced residual degradation, and burst error degradation, respectively.

[0120] It should be noted that the weights for level separation degradation, pre-FEC bit error degradation, FEC error correction pressure degradation, equalization residual degradation, and burst bit error degradation are determined by offline calibration of historical monitoring samples of the 800G PAM4 link under normal, mild, moderate, and severe degradation conditions, and normalized by the contribution of each degradation level to the link quality level judgment result. Among them, the weight of level separation degradation typically ranges from 0.20 to 0.35, the weight of pre-FEC bit error degradation typically ranges from 0.20 to 0.35, the weight of FEC error correction pressure degradation typically ranges from 0.20 to 0.35, the weight of equalization residual degradation typically ranges from 0.10 to 0.20, and the weight of burst bit error degradation typically ranges from 0.05 to 0.15, and the sum of the five is 1.

[0121] It should be noted that the weighting method allows PAM4 level closure, original bit error rate before FEC, FEC error correction load, residual damage after equalization, and burst errors to all participate in the comprehensive degradation judgment of a single 100G physical channel. Among them, PAM4 level closure, original bit error rate before FEC, and FEC error correction load are the main factors, while equalization residual and burst errors are the supplementary factors, thereby avoiding misjudgment of link degradation caused by relying solely on a single bit error rate.

[0122] Furthermore, after the channel degradation characteristic values ​​of the first to eighth physical channels have been calculated, the channel degradation characteristic values ​​of the eight 100G physical channels are summed and divided by 8 to obtain the average degradation value within the current 1-second fixed monitoring period.

[0123] It should be noted that the overall degradation level of the 800GPAM4 link is obtained by averaging the channel degradation characteristic values ​​of the eight 100G physical channels. The higher the average degradation value, the worse the overall link quality of the eight 100G physical channels.

[0124] Furthermore, after obtaining the average degradation value, the difference between the channel degradation characteristic value and the average degradation value of each 100G physical channel is calculated, and the square root of the average of the squared differences is taken to obtain the channel dispersion within the current 1s fixed monitoring period.

[0125] It should be noted that if the channel degradation characteristic values ​​of the eight 100G physical channels are close to each other, the dispersion between channels is small, indicating that the 800G PAM4 link is more likely to exhibit overall link quality changes; if the channel degradation characteristic value of one or a few 100G physical channels deviates significantly from the other channels, the dispersion between channels increases, indicating that the 800G PAM4 link may have local physical channel degradation.

[0126] After the current 1-second fixed monitoring cycle ends, the link monitoring logic uses the channel degradation characteristic value, average degradation value, and inter-channel dispersion of the first to eighth physical channels as the channel degradation statistics for the current cycle.

[0127] The channel degradation statistics include the channel degradation characteristic value of each 100G physical channel, the average degradation value of the eight 100G physical channels, and the channel dispersion of the eight 100G physical channels.

[0128] S4. Based on the average degradation value, inter-channel dispersion, and number of uncorrectable codewords in FEC, generate a port-level link quality index. Compare the port-level link quality index with the link quality index threshold to output the 800G PAM4 link quality level and link degradation type.

[0129] Furthermore, in this embodiment, within each fixed 1-second monitoring cycle, after obtaining the average degradation value and inter-channel dispersion of the eight 100G physical channels, the link monitoring logic further reads the number of FEC uncorrectable codewords for the eight 100G physical channels and generates a port-level link quality index based on the average degradation value, inter-channel dispersion, and the number of FEC uncorrectable codewords. Subsequently, the port-level link quality index is compared with a link quality index threshold to output an 800G PAM4 link quality level, and the link degradation type is output based on the average degradation value, inter-channel dispersion, number of FEC uncorrectable codewords, and the channel degradation characteristic value of each 100G physical channel.

[0130] Among them, the port-level link quality index is used to characterize the overall carrying quality of the 800G PAM4 link within the current 1-second fixed monitoring period; the link quality level is used to classify whether the current 800G PAM4 link is suitable for stable service carrying; and the link degradation type is used to describe whether the current link degradation is mainly manifested as overall degradation, local physical channel degradation, or FEC uncorrectable risk.

[0131] Specifically, since the occurrence of an uncorrectable FEC codeword in any 100G physical channel may affect the port-level service reliability of the 800G PAM4 link, this embodiment summarizes the number of uncorrectable FEC codewords in the eight 100G physical channels into a port-level uncorrectable risk quantity.

[0132] Specifically, the number of uncorrectable FEC codewords in the first to eighth physical channels within the current 1-second fixed monitoring period is summed, and the summation result is subjected to bounded normalization to obtain the port-level uncorrectable risk quantity.

[0133] It should be noted that when none of the eight 100G physical channels contain FEC uncorrectable codewords, the port-level uncorrectable risk is 0. When any 100G physical channel contains an FEC uncorrectable codeword, the port-level uncorrectable risk increases with the total number of FEC uncorrectable codewords, gradually approaching 1. Bounded normalization avoids an abnormal imbalance in the port-level link quality index caused by an excessively large number of FEC uncorrectable codewords, while ensuring that FEC uncorrectable risk is clearly included in the evaluation results.

[0134] After obtaining the port-level uncorrectable risk quantity, the link monitoring logic combines the average degradation value, inter-channel dispersion, and port-level uncorrectable risk quantity according to a fixed penalty coefficient to generate the port-level link quality index for the current 1-second fixed monitoring period.

[0135] The port-level link quality index is expressed as: ;

[0136] in, This represents the port-level link quality index within the current fixed 1-second monitoring period. This represents the average degradation value within the current fixed monitoring period of 1 second. This indicates the inter-channel dispersion within the current fixed 1-second monitoring period. This represents the amount of uncorrectable risk at the port level within the current fixed 1-second monitoring period. This represents the inter-channel discrete penalty coefficient. This represents the penalty coefficient for uncorrectable risk.

[0137] It should be noted that the inter-channel discrete penalty coefficient and the uncorrectable risk penalty coefficient are determined by offline fitting of historical samples collected from the 800G PAM4 link under three calibration conditions: normal transmission, single-channel degradation, and FEC uncorrectable. The optimal approach is to make the contribution of the inter-channel discrete term of the single-channel degradation sample lower than the contribution of the uncorrectable risk term of the FEC uncorrectable sample.

[0138] In this embodiment, the inter-channel discrete penalty coefficient is preferably 0.3, and the uncorrectable risk penalty coefficient is preferably 5.

[0139] It should be noted that the average degradation value reflects the overall quality degradation of the eight 100G physical channels; the channel dispersion reflects whether a local 100G physical channel deviates significantly from other channels; and the port-level uncorrectable risk reflects whether FEC has reached an unrecoverable risk. By combining these three factors, the port-level link quality index can not only characterize the overall degradation but also enhance the response to single-channel anomalies and uncorrectable FEC risks.

[0140] In practice, a higher port-level link quality index indicates a worse 800G PAM4 link quality, while a lower port-level link quality index indicates a better 800G PAM4 link quality.

[0141] Furthermore, in order to convert the port-level link quality index into a link quality level that can be directly used for operation and maintenance display or alarm judgment, three link quality index thresholds are set, namely the first link quality index threshold, the second link quality index threshold, and the third link quality index threshold.

[0142] It should be noted that the link quality index threshold is determined by collecting port-level link quality index samples under four operating conditions—stable bearer, available bearer, degradation warning, and unstable bearer—during the experimental calibration phase of the 800G PAM4 link, and taking the index value at the boundary between adjacent sample distributions as the corresponding threshold. Specifically, the first link quality index threshold ranges from 1.0 to 2.0, preferably 1.5; the second link quality index threshold ranges from 2.5 to 3.5, preferably 3.0; and the third link quality index threshold ranges from 4.5 to 6.0, preferably 5.0. Furthermore, all three thresholds satisfy the condition that the first link quality index threshold is less than the second link quality index threshold, which is less than the third link quality index threshold.

[0143] The link monitoring logic compares the port-level link quality index within the current fixed 1-second monitoring period with the first link quality index threshold, the second link quality index threshold, and the third link quality index threshold in sequence, and outputs the 800G PAM4 link quality level based on the comparison results.

[0144] Specifically, the judgment is as follows: when the port-level link quality index is less than the first link quality index threshold, the 800G PAM4 link quality level is output as "Excellent"; when the port-level link quality index is greater than or equal to the first link quality index threshold and less than the second link quality index threshold, the 800G PAM4 link quality level is output as "Available"; when the port-level link quality index is greater than or equal to the second link quality index threshold and less than the third link quality index threshold, the 800G PAM4 link quality level is output as "Degradation Warning"; when the port-level link quality index is greater than or equal to the third link quality index threshold, the 800G PAM4 link quality level is output as "Unstable Bearing".

[0145] After outputting the 800G PAM4 link quality level, the link monitoring logic further determines the link degradation type based on the average degradation value, inter-channel dispersion, number of FEC uncorrectable codewords, and channel degradation characteristic values ​​of each 100G physical channel.

[0146] In this embodiment, the link degradation types include overall degradation, local physical channel degradation, FEC uncorrectable risk, and low-risk stable.

[0147] Specifically, when the sum of the number of FEC uncorrectable code words in the eight 100G physical channels is greater than 0, the output link degradation type is "FEC uncorrectable risk type".

[0148] The FEC uncorrectable risk type indicates that at least one 100G physical channel has an FEC uncorrectable codeword within the current 1-second fixed monitoring period, and there is a risk that the service data on the link is unrecoverable.

[0149] When the sum of the number of uncorrectable FEC codewords for the eight 100G physical channels is equal to 0, and the channel dispersion is greater than the dispersion threshold, it is further determined whether a target physical channel exists. If the channel degradation characteristic value of any 100G physical channel is greater than the sum of the average degradation value and twice the channel dispersion, then the current 100G physical channel is determined as the target physical channel, and the link degradation type is output as "local physical channel degradation type".

[0150] It should be noted that the dispersion threshold is determined by collecting channel dispersion samples under balanced deterioration conditions of eight 100G physical channels and under local deterioration conditions of a single 100G physical channel during the 800G PAM4 link experimental calibration phase. The channel dispersion value at the junction of the two types of sample distributions is taken as the dispersion threshold. The value range is usually 0.5~1.2, preferably 0.8.

[0151] When the sum of the number of FEC uncorrectable codewords in the eight 100G physical channels is equal to 0, the dispersion between channels is not greater than the dispersion threshold, and the average degradation value is greater than the average degradation threshold, the output link degradation type is "overall degradation type".

[0152] The overall degradation type indicates that all eight 100G physical channels show a similar trend of quality decline, which usually corresponds to an increase in overall link noise, increased transmission impairment, or worsening of common transmission conditions.

[0153] When the sum of the number of FEC uncorrectable codewords in the eight 100G physical channels is equal to 0, the dispersion between channels is not greater than the dispersion threshold, and the average degradation value is not greater than the average degradation threshold, the output link degradation type is "low-risk stable".

[0154] It should be noted that the average degradation threshold can be obtained by collecting average degradation value samples of eight 100G physical channels under overall stable and overall degraded conditions during the experimental calibration phase of the 800G PAM4 link, and taking the average degradation value at the junction of the two types of sample distributions as the average degradation threshold; the value range is usually 1.5~3.0, preferably 2.0.

[0155] After completing the calculation of the port-level link quality index, the judgment of the link quality level, and the judgment of the link degradation type, the link monitoring logic generates the link evaluation result for the current fixed monitoring cycle of 1 second.

[0156] The link evaluation results include the port-level link quality index, the 800G PAM4 link quality level, the link degradation type, and the target physical channel number corresponding to the local physical channel degradation type.

[0157] The link evaluation results can be written to the link status register, management interface cache, or operation and maintenance display cache for device-side reading, log recording, or link status display.

[0158] The link evaluation results are only used to reflect the quality of the 800G PAM4 link during the current 1-second fixed monitoring period, and do not perform update operations on the receive equalization parameters, FEC parameters, modulation parameters, or transmission path.

[0159] In this embodiment, Figure 5 The comparative monitoring results of eight 100G physical channels under a scenario of local physical channel degradation are shown.

[0160] As shown in the figure, the channel degradation characteristic values ​​of CH1 to CH5 and CH8 remain between 0.28 and 0.35, and the bit error rate before FEC also remains at a low level, indicating that these channels are in a relatively stable transmission state; while the channel degradation characteristic values ​​of CH6 and CH7 increase to 0.82 and 0.71 respectively, corresponding to a synchronous increase in the bit error rate before FEC, showing obvious local anomaly characteristics.

[0161] The results demonstrate that this invention does not rely solely on a single bit error rate threshold to determine link status. Instead, it comprehensively weights PAM4 level separability, pre-FEC bit error rate, FEC error correction pressure, equalization residual, and the number of burst error events to form a channel degradation characteristic value that reflects the true degree of degradation of a single channel. By further calculating the average degradation value and inter-channel dispersion, this invention can accurately identify local physical channel degradation types characterized by "a few channels being abnormal, but the overall port not yet being completely unstable," thereby improving the precision and accuracy of 800G PAM4 link fault location and early warning.

[0162] In this embodiment, Figure 6 The results show the continuous monitoring comparison between the port-level link quality index and the bit error rate before traditional FEC under the overall degradation scenario.

[0163] As shown in the figure, the port-level link quality index continued to rise as the monitoring time increased, and crossed the second link quality index threshold at about 16 minutes, entering the degradation warning zone; in contrast, the bit error rate before traditional FEC only reached the set threshold at about 24 minutes.

[0164] Comparative results show that this invention generates a port-level link quality index by combining the average degradation value, inter-channel dispersion, and FEC uncorrectable risk. This index can anticipate the decline in link capacity caused by PAM4 level closure, increased equalization residuals, and accumulated FEC error correction pressure before the bit error rate significantly exceeds the limit. Compared to traditional methods that rely solely on the bit error rate before FEC, this invention can output the link quality level and overall degradation type earlier, providing early warning for 800G interconnect links in data centers. This avoids passively discovering anomalies only after FEC uncorrectable codewords appear, thereby improving link operational reliability and maintenance response efficiency.

[0165] In summary, this invention achieves systematic acquisition of multi-dimensional damage by collecting PAM4 statistics, balanced residuals, FEC counts, and burst error data from eight channels and calculating level separability, overcoming the limitations of single-index evaluation. By generating channel degradation characteristic values ​​through weighted fusion and calculating average degradation values ​​and dispersion, it achieves joint quantification of single-channel degradation and channel inconsistency, solving the problem of difficulty in distinguishing between overall aging and local faults. By generating port-level link quality indices and comparison thresholds to output quality levels, and outputting degradation types based on parameters, it achieves refined grading and degradation pattern diagnosis, improving early warning accuracy. Through burst event judgment based on separability of adjacent level differences and standard deviations and sub-window error comparisons, it achieves sensitive quantification of eye diagram closure and burst errors, compensating for the lag in response to burst degradation.

[0166] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A PAM4 link quality assessment method for 800G data center interconnection, characterized in that, include: During the operation of the 800G PAM4 link, the average value of PAM4 four-level, standard deviation of PAM4 four-level, equalization residual, number of FEC pre-errors, number of FEC decoding input bits, number of FEC correction symbols, number of FEC uncorrectable codewords, and number of burst error events were collected for eight physical channels at monitoring cycles. For each 100G physical channel, the level separability is calculated based on the sum of the difference between the average values ​​of adjacent PAM4 electrical values ​​and the standard deviation of PAM4 levels. The FEC error rate is calculated based on the number of bit errors before FEC and the number of FEC decoding input bits. The FEC error correction pressure is calculated based on the number of FEC corrected symbols, the number of FEC uncorrectable codewords, and the number of FEC decoding input bits. The channel degradation characteristic value is generated by weighting the level separability, bit error rate before FEC, FEC error correction pressure, equalization residual and number of burst bit error events. The average degradation value and the dispersion between channels are calculated from the channel degradation characteristic values ​​of eight 100G physical channels. Based on the average degradation value, inter-channel dispersion, and number of uncorrectable codewords in FEC, a port-level link quality index is generated. The port-level link quality index is compared with the link quality index threshold to output the 800G PAM4 link quality level and link degradation type.

2. The PAM4 link quality assessment method for 800G data center interconnection as described in claim 1, characterized in that, The PAM4 four-level average value, PAM4 four-level standard deviation, and equalization residual include: After the 800G PAM4 link completes link training, clock recovery and FEC alignment, the local management clock generates periodic pulses, which are used as the acquisition trigger signal for the monitoring cycle. When each periodic pulse arrives, the link monitoring logic reads the receiving side statistical data of the first to eighth physical channels in sequence according to the physical channel number, and forms the collection results for the current monitoring period respectively; For any 100G physical channel, after the receiver completes clock data recovery, it performs level decision on the equalized PAM4 sample value and assigns the PAM4 sample value to level 0, level 1, level 2 and level 3 according to the decision threshold. For any PAM4 level among level 0, level 1, level 2 and level 3, sum all sampled values ​​belonging to the same PAM4 level within the current monitoring period, and then divide by the number of samples corresponding to the current PAM4 level to obtain the average PAM4 level within the current monitoring period. For any PAM4 level among level 0, level 1, level 2 and level 3, calculate the deviation between each sampled value and the corresponding average PAM4 level, and take the square root of the average of the squared deviations to obtain the standard deviation of the current PAM4 level in the current monitoring period. For any 100G physical channel, after completing feedforward equalization, decision feedback equalization, or DSP equalization, the receiver obtains the residual signal between the equalization output value and the corresponding decision level, and calculates the square mean of the residual signal in the current monitoring period to obtain the equalization residual of the current 100G physical channel.

3. The PAM4 link quality assessment method for 800G data center interconnection as described in claim 1, characterized in that, The number of FEC pre-errors, the number of FEC decoding input bits, the number of FEC correction symbols, the number of FEC uncorrectable codewords, and the number of burst error events include: For any 100G physical channel, set an FEC pre-error counter and an FEC decoding input bit counter on the input side of the FEC decoder, count the number of erroneous bits detected before entering the FEC decoding process, and count the total number of bits entering the FEC decoding process. At the end of the current monitoring cycle, read the cumulative FEC bit error rate and cumulative FEC decoding input bit value of the current 100G physical channel at the end of the current cycle, and read the cumulative FEC bit error rate and cumulative FEC decoding input bit value of the current 100G physical channel at the end of the previous cycle. The difference between the cumulative FEC bit error rate at the end of the current cycle and the cumulative FEC bit error rate at the end of the previous cycle is used to obtain the number of FEC bit errors in the current monitoring cycle. The difference between the cumulative value of FEC decoded input bits at the end of the current cycle and the cumulative value of FEC decoded input bits at the end of the previous cycle is used to obtain the number of FEC decoded input bits in the current monitoring cycle. For any 100G physical channel, read the cumulative value of FEC corrected symbol and the cumulative value of FEC uncorrectable codeword at the output of the FEC decoder. The number of FEC correction symbols in the current monitoring period is obtained by subtracting the cumulative value of FEC correction symbols at the end of the current period from the cumulative value of FEC correction symbols at the end of the previous period. The number of FEC uncorrectable codewords in the current monitoring period is obtained by subtracting the cumulative value of FEC uncorrectable codewords at the end of the current period from the cumulative value of FEC uncorrectable codewords at the end of the previous period. Each monitoring cycle is divided into sub-windows, and the number of FEC-preceding bit errors for any 100G physical channel in each sub-window is counted separately. Calculate the average number of bit errors in the current monitoring period based on the number of bit errors before FEC in each sub-window; For any sub-window, compare the number of pre-FEC bit errors in the current sub-window with the average number of bit errors three times that of the sub-window. When the number of pre-FEC bit errors in the current sub-window is greater than the average number of bit errors three times that of the sub-window, the current sub-window is determined to have experienced a burst bit error event. The number of sub-windows that are identified as having sudden bit error events within the current monitoring period is added together to obtain the number of sudden bit error events in the current 100G physical channel within the current monitoring period.

4. The PAM4 link quality assessment method for 800G data center interconnection as described in claim 1, characterized in that, For each 100G physical channel, the level separability is calculated based on the sum of the difference in average PAM4 voltage levels between adjacent channels and the standard deviation of PAM4 voltage levels. The specific steps are as follows: After the current monitoring cycle ends, the link monitoring logic reads the acquisition results from the first physical channel to the eighth physical channel in sequence; For any 100G physical channel, read the four PAM4 electrical average values ​​and four PAM4 level standard deviations for any 100G physical channel within the current monitoring period; The average value and standard deviation of the PAM4 four-level signals of the same 100G physical channel are arranged from low to high according to the level amplitude; The first PAM4 level and the second PAM4 level are combined to form a first adjacent level pair; the second PAM4 level and the third PAM4 level are combined to form a second adjacent level pair; and the third PAM4 level and the fourth PAM4 level are combined to form a third adjacent level pair. For any adjacent level pair, calculate the difference between the average values ​​of two adjacent PAM4 levels, and calculate the sum of the standard deviations of the two adjacent PAM4 levels. Divide the difference by the sum of the standard deviations to obtain the locally separable value of the adjacent level pair. The level separability of the current 100G physical channel within the current monitoring period is obtained by averaging the local separability values ​​of three adjacent level pairs.

5. The PAM4 link quality assessment method for 800G data center interconnects as described in claim 1 or 4, characterized in that, The specific steps for calculating the FEC pre-error rate based on the number of bit errors before FEC and the number of FEC decoding input bits are as follows: After the current monitoring period ends, the link monitoring logic reads the number of FEC pre-errors and the number of FEC decoding input bits for any 100G physical channel during the current monitoring period. Divide the number of FEC-pre-errors in the current 100G physical channel by the number of FEC decoding input bits to obtain the FEC-pre-error rate of the current 100G physical channel. When the number of FEC decoding input bits is zero, the link monitoring logic marks the data of the current 100G physical channel in the current monitoring period as invalid sampling.

6. The PAM4 link quality assessment method for 800G data center interconnection as described in claim 5, characterized in that, The specific steps for calculating the FEC error correction pressure based on the number of FEC correction symbols, the number of FEC uncorrectable codewords, and the number of FEC decoding input bits are as follows: After the current monitoring period ends, the link monitoring logic reads the number of FEC correction symbols, the number of FEC uncorrectable codewords, and the number of FEC decoding input bits for any 100G physical channel during the current monitoring period. Multiply the number of uncorrectable codewords in FEC by the uncorrectable penalty coefficient to obtain the uncorrectable penalty amount; Add the uncorrectable penalty to the number of FEC correction symbols, and divide by the number of FEC decoding input bits to obtain the FEC error correction pressure of the current 100G physical channel; During the current monitoring period, the basic quality characteristics of each 100G physical channel are recorded after calculating the level separability, bit error rate before FEC, and FEC error correction pressure for all physical channels from the first to the eighth physical channel.

7. The PAM4 link quality assessment method for 800G data center interconnection as described in claim 6, characterized in that, The specific steps for generating channel degradation characteristic values ​​by weighting level separability, pre-FEC bit error rate, FEC error correction pressure, equalization residual, and number of burst error events are as follows: During the current monitoring period, the link monitoring logic sequentially reads the level separability, bit error rate before FEC, FEC error correction pressure, equalization residual, and number of burst error events from the first physical channel to the eighth physical channel. Using level separability as the denominator, and taking the reciprocal of level separability, we obtain the level separation degradation. Multiply the bit error rate before FEC by the bit error rate amplification factor, add 1, and then take the logarithm to the base 10 to obtain the bit error rate degradation before FEC. Multiply the FEC error correction pressure by the error correction pressure amplification factor, add 1, and then take the logarithm to the base 10 to obtain the FEC error correction pressure degradation. Divide the equilibrium residual by the sum of the equilibrium residual and 1 to obtain the degradation of the equilibrium residual; Divide the number of burst error events by the sum of the number of burst error events and 1 to obtain the burst error degradation. The degradation values ​​of level separation, pre-FEC bit error rate, FEC error correction pressure, equalization residual, and burst bit error rate are weighted and summed according to fixed weights to obtain the channel degradation characteristic value of each 100G physical channel in the current monitoring period.

8. The PAM4 link quality assessment method for 800G data center interconnection as described in claim 7, characterized in that, The calculation of the average degradation value and inter-channel dispersion based on the channel degradation characteristic values ​​of eight 100G physical channels involves the following steps: After the channel degradation characteristic values ​​of the first to eighth physical channels have been calculated, the channel degradation characteristic values ​​of the eight 100G physical channels are summed and divided to obtain the average degradation value in the current monitoring period. After obtaining the average degradation value, the difference between the channel degradation characteristic value and the average degradation value for each 100G physical channel is calculated. The differences between the eight 100G physical channels are squared, and the average of the squared results is taken to obtain the channel dispersion within the current monitoring period. The channel degradation characteristic value, average degradation value, and inter-channel dispersion of the first to eighth physical channels are used as the channel degradation statistics for the current monitoring period.

9. The PAM4 link quality assessment method for 800G data center interconnection as described in claim 1, characterized in that, The process of generating a port-level link quality index based on the average degradation value, inter-channel dispersion, and the number of uncorrectable codewords in FEC involves the following steps: During the current monitoring period, after obtaining the average degradation value and inter-channel dispersion of the eight 100G physical channels, the link monitoring logic reads the number of FEC uncorrectable code words from the first physical channel to the eighth physical channel. The number of uncorrectable FEC codewords in the first to eighth physical channels during the current monitoring period is summed, and the summation result is bounded and normalized to obtain the port-level uncorrectable risk quantity. The average degradation value, inter-channel dispersion, and port-level uncorrectable risk are combined using a fixed penalty coefficient to generate the port-level link quality index for the current monitoring period.

10. The PAM4 link quality assessment method for 800G data center interconnects as described in claim 9, characterized in that, The specific steps for comparing the port-level link quality index with the link quality index threshold and outputting the 800G PAM4 link quality level and link degradation type are as follows: Set a first link quality index threshold, a second link quality index threshold, and a third link quality index threshold, wherein the first link quality index threshold is less than the second link quality index threshold, and the second link quality index threshold is less than the third link quality index threshold; Compare the port-level link quality index within the current monitoring period with the first link quality index threshold, the second link quality index threshold, and the third link quality index threshold in sequence. When the port-level link quality index is less than the first link quality index threshold, the 800G PAM4 link quality level is output as excellent. When the port-level link quality index is greater than or equal to the first link quality index threshold and less than the second link quality index threshold, the 800G PAM4 link quality level is output as available. When the port-level link quality index is greater than or equal to the second link quality index threshold and less than the third link quality index threshold, the 800G PAM4 link quality level is output as a deterioration warning. When the port-level link quality index is greater than or equal to the third link quality index threshold, the 800G PAM4 link quality level is output as unstable. When the sum of the number of FEC uncorrectable code words in the eight 100G physical channels is greater than 0, the output link degradation type is FEC uncorrectable risk type. When the sum of the number of FEC uncorrectable codewords of the eight 100G physical channels is equal to 0, the channel dispersion is greater than the dispersion threshold, and the channel degradation characteristic value of any 100G physical channel is greater than the average degradation value and twice the sum of the channel dispersion, the output link degradation type is local physical channel degradation. When the sum of the number of FEC uncorrectable codewords of the eight 100G physical channels is equal to 0, the dispersion between channels is not greater than the dispersion threshold, and the average degradation value is greater than the average degradation threshold, the output link degradation type is overall degradation type. When the sum of the number of FEC uncorrectable codewords in the eight 100G physical channels is equal to 0, the dispersion between channels is not greater than the dispersion threshold, and the average degradation value is not greater than the average degradation threshold, the output link degradation type is low-risk stable.

Images