Method for obtaining high-order qam soft information and related device

Abstract

This invention discloses a method and related equipment for acquiring soft information in high-order QAM. The method includes: determining the corresponding target constellation sub-region and candidate constellation points based on the in-phase and quadrature component coordinates of the high-order QAM baseband complex symbol signal and a preset high-order QAM constellation mapping rule; calculating the distance difference of each bit carried by the signal based on the minimum Euclidean distance; dividing each bit into at least two bit groups and determining the compensation weight corresponding to each bit group; using the compensation weight corresponding to each bit group, weighting and correcting the initial log-likelihood ratio of each bit belonging to the corresponding bit group, and outputting the soft information sequence corresponding to the high-order QAM baseband complex symbol signal. This method solves the problem that traditional fixed-mode soft demapping algorithms cannot adapt to the dynamic channel characteristics of satellites and cannot meet the engineering requirements for high-speed and reliable reception of high-order QAM signals from satellites.

Description

Technical Field

[0001] The embodiments of this application relate to the aerospace field, and more specifically, the present invention relates to a high-order QAM soft information acquisition method and related equipment. Background Technology

[0002] To meet the transmission requirements of high-speed data transmission and high spectrum utilization, broadband satellite communication systems generally adopt a transmission system that combines high-order modulation with channel coding. Among them, 128QAM high-order orthogonal amplitude modulation can significantly improve the transmission rate within a limited bandwidth. Combined with the soft-input iterative decoding mechanism of LDPC (Low Density Parity Check) check codes, it can achieve highly reliable communication transmission under satellite channels and is the core technical architecture of current medium and high-speed satellite receivers.

[0003] LDPC decoding is a typical soft-input iterative decoding algorithm. Its decoding performance, iteration convergence speed, and error rate performance are highly dependent on the accuracy of the bit soft information LLR (Log-Likelihood Ratio) output by the demodulation module. High-precision soft information that closely reflects the actual channel conditions can effectively lower the LDPC decoding threshold, reduce the number of iterations, and suppress error planes. Conversely, distorted soft information and excessive confidence bias will directly lead to deterioration in decoding performance, and may even result in iteration non-convergence and data decoding failure.

[0004] In satellite mobile communication and low-Earth orbit high-speed satellite data transmission scenarios, the channel exhibits significant Doppler residual frequency offset, random phase noise, small-amplitude multipath fading, and additive white Gaussian noise, which can easily cause distortion in the 128QAM standard rectangular constellation. Each 128QAM symbol carries 7 modulation bits, and the sensitivity of high and low bits to constellation distortion differs greatly. Traditional fixed-mode soft demapping algorithms cannot adapt to the dynamic channel characteristics of satellites, resulting in problems such as accuracy imbalance, high computational overhead, and poor adaptability to low signal-to-noise ratios, making it difficult to meet the engineering requirements for high-speed and reliable 128QAM satellite reception. Summary of the Invention

[0005] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. The summary section of this invention is not intended to limit the key features and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0006] To address the problem that traditional fixed-mode soft demapping algorithms cannot adapt to the dynamic channel characteristics of satellites and thus fail to meet the engineering requirements for high-speed and reliable high-order QAM reception, this invention proposes a high-order QAM soft information acquisition method, comprising: Acquire the high-order QAM baseband complex symbol signal after carrier synchronization, symbol synchronization and channel equalization; Based on the in-phase component coordinates and quadrature component coordinates of the high-order QAM baseband complex symbol signal, and the preset high-order QAM constellation mapping rules, the target constellation sub-region corresponding to the high-order QAM baseband complex symbol signal is determined, so as to obtain candidate constellation points configured for the target constellation sub-region. For each bit carried by the high-order QAM baseband complex symbol signal, a first candidate constellation point with a first bit value and a second candidate constellation point with a second bit value are determined from the candidate constellation points, so as to calculate the distance difference of the corresponding bit based on the minimum Euclidean distance; Based on the decision boundary geometric characteristics of each bit in the higher-order QAM constellation mapping rule, each bit is divided into at least two bit groups, and the compensation weight corresponding to each bit group is determined respectively. By employing the compensation weights corresponding to each bit group, the initial log-likelihood ratio of each bit belonging to the corresponding bit group is weighted and corrected, and the soft information sequence corresponding to the high-order QAM baseband complex symbol signal is output.

[0007] Secondly, the present invention also proposes a high-order QAM soft information acquisition device, comprising: The signal acquisition unit is used to acquire the high-order QAM baseband complex symbol signal after carrier synchronization, symbol synchronization and channel equalization; The candidate constellation point determination unit is used to determine the target constellation sub-region corresponding to the high-order QAM baseband complex symbol signal based on the in-phase component coordinates and quadrature component coordinates of the high-order QAM baseband complex symbol signal and the preset high-order QAM constellation mapping rules, so as to obtain candidate constellation points configured for the target constellation sub-region. The distance difference calculation unit is used to determine, for each bit carried by the high-order QAM baseband complex symbol signal, a first candidate constellation point marked with a first bit value and a second candidate constellation point marked with a second bit value from the candidate constellation points, so as to calculate the distance difference of the corresponding bit based on the minimum Euclidean distance. The compensation weight determination unit is used to divide each bit into at least two bit groups according to the decision boundary geometric characteristics of each bit in the high-order QAM constellation mapping rule, and determine the compensation weight corresponding to each bit group respectively. The soft information output unit is used to apply the compensation weights corresponding to each bit group to perform weighted correction on the initial log-likelihood ratio of each bit belonging to the corresponding bit group, and output the soft information sequence corresponding to the high-order QAM baseband complex symbol signal.

[0008] In summary, the high-order QAM soft information acquisition method proposed in this application first determines local candidate constellation points using the positions of the received symbols in the constellation plane, and then calculates the distance difference corresponding to each bit using these local candidate constellation points. Simultaneously, considering that the decision boundaries corresponding to different bits differ in number, direction, spacing, and local morphology, appropriate compensation weights are applied to bits with similar decision boundary geometric characteristics. Therefore, while reducing the computational load of soft information, the non-uniform impact of local candidate calculation on different bits can be corrected, making the output soft information more suitable for subsequent channel decoding. Thus, when the number of high-order QAM constellation points is large, limiting the local candidate range by the received symbol positions and compensating for the soft information differences of different bits by the geometric characteristics of the decision boundaries achieves a good balance between computational complexity and soft information reliability. Attached Figure Description

[0009] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A schematic flowchart of a high-order QAM soft information acquisition method provided in this application embodiment; Figure 2 A schematic diagram of a high-order QAM soft information acquisition device provided in this application embodiment; Figure 3 This is a schematic diagram of a high-order QAM soft information acquisition electronic device provided in an embodiment of this application. Detailed Implementation

[0010] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The technical solutions of the embodiments of this application will now be clearly and completely described in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0011] Currently, the mainstream 128QAM soft information demapping schemes in the industry are the traditional ML (Maximum Likelihood) global traversal demapping algorithm and the fixed threshold approximation LLR (Log-Likelihood Ratio) demapping algorithm. These are the common technical solutions for existing high-order QAM demodulation and LDPC decoding. The specific implementation is as follows: Traditional maximum likelihood demapping scheme: For the received QAM baseband I / Q signal, it traverses all 128 standard constellation points in the 128QAM constellation diagram, calculates the Euclidean distance between the received signal and each constellation point, and calculates the LLR soft information of each bit by comparing the minimum Euclidean distances corresponding to all bits 0 and bits 1, providing soft input for LDPC decoding.

[0012] Fixed threshold approximate LLR demapping scheme: In order to reduce the computational load of global traversal, the full constellation point calculation method is abandoned. A fixed I / Q axis decision threshold is set in advance. Based on the fixed interval where the received I / Q signal is located, a unified correction coefficient is used to approximate the bit LLR. This simplifies the operation logic and reduces the hardware implementation overhead. It is a lightweight 128QAM soft demapping technology scheme.

[0013] When the above two types of existing technologies are applied to satellite 128QAM receivers, they have the following core defects and cannot adapt to the complex dynamic channel scenarios of satellites: 1. The traditional global maximum likelihood traversal algorithm has a huge computational load, requiring traversing all 128 constellation points for distance calculation. When implemented on FPGA baseband, it has high resource overhead, limited throughput, and large latency, making it difficult to meet the engineering requirements of high-speed real-time demodulation for satellites, and has poor hardware adaptability.

[0014] 2. The fixed threshold approximation LLR algorithm has extremely poor adaptability. The threshold parameter remains unchanged and cannot be adaptively adjusted to keep up with the dynamic changes in the signal-to-noise ratio of the satellite channel. In constellation distortion scenarios, soft information distortion is severe, which in turn leads to a decrease in LDPC decoding performance.

[0015] 3. In Gray-mapped 128QAM, the reliability of the 7 bits exhibits stratification. For example, high-order bits have large decision boundary intervals and strong noise resistance, while low-order bits have small decision boundary intervals and weak noise resistance. Using a uniform correction coefficient to calculate the LLR of all bits leads to large bit confidence deviations, high false positive rates, and an imbalance of soft information between high and low-order bits, significantly degrading LDPC decoding performance.

[0016] 4. In low signal-to-noise ratio scenarios, the LLR output by existing algorithms is prone to numerical overflow, which leads to slow convergence of LDPC decoding iterations and a significant increase in the bit error rate, seriously reducing the transmission reliability of satellite communication systems.

[0017] To address the issue that traditional fixed-mode soft demapping algorithms cannot adapt to the dynamic channel characteristics of satellites, thus failing to meet the engineering requirements for high-speed and reliable high-order QAM reception on satellites. Please refer to... Figure 1 , Figure 1 This is a flowchart illustrating a high-order QAM soft information acquisition method provided in an embodiment of this application, which may specifically include steps S110 to S150.

[0018] S110 acquires the high-order QAM baseband complex symbol signal after carrier synchronization, symbol synchronization, and channel equalization.

[0019] S120, based on the in-phase component coordinates and quadrature component coordinates of the high-order QAM baseband complex symbol signal, and the preset high-order QAM constellation mapping rules, determine the target constellation sub-region corresponding to the high-order QAM baseband complex symbol signal, so as to obtain candidate constellation points configured for the target constellation sub-region.

[0020] S130, for each bit carried by the high-order QAM baseband complex symbol signal, determine a first candidate constellation point with a first bit value and a second candidate constellation point with a second bit value from the candidate constellation points, and calculate the distance difference of the corresponding bit based on the minimum Euclidean distance.

[0021] S140, based on the decision boundary geometric characteristics of each bit in the higher-order QAM constellation mapping rule, divide each bit into at least two bit groups, and determine the compensation weight corresponding to each bit group respectively.

[0022] S150, using the compensation weights corresponding to each bit group, the initial log-likelihood ratio of each bit belonging to the corresponding bit group is weighted and corrected, and the soft information sequence corresponding to the high-order QAM baseband complex symbol signal is output.

[0023] Understandably, higher-order QAM carries multiple bits through different complex coordinates in the constellation plane. For example, a higher-order QAM symbol can be represented as a complex signal with in-phase and quadrature components. The in-phase component determines the symbol's position in the horizontal direction of the constellation plane, while the quadrature component determines its position in the vertical direction. Actual received signals are affected by factors such as carrier frequency deviation, carrier phase deviation, symbol timing deviation, channel fading, and noise during transmission. If the Euclidean distance between the unprocessed received signal and the standard constellation points is calculated directly, the resulting distance may primarily reflect synchronization errors or channel distortion, rather than accurately reflecting the possibility of different transmitted bit values. Therefore, the signal after carrier synchronization, symbol synchronization, and channel equalization can be used as the object of soft information calculation. Carrier synchronization reduces the overall rotation of the constellation diagram and the phase drift that varies over time; symbol synchronization brings the sampling position closer to the effective decision time of the symbol, reducing mutual interference between adjacent symbols; and channel equalization compensates for the distortion caused by the channel on the symbol amplitude and phase. After these processes, the position of the received symbol in the constellation plane can more accurately reflect its positional relationship relative to the standard constellation points.

[0024] For example, the receiver can sequentially acquire synchronized and equalized complex symbols according to the symbol period. For any symbol to be processed, its in-phase component coordinates and orthogonal component coordinates are read, and this complex symbol is used as input for subsequent target constellation sub-region determination and distance calculation. For example, the ideal transmitted constellation point is located in the first quadrant of the constellation plane. Due to noise, the equalized received symbol may not fall precisely on this constellation point, but rather in a nearby location. In this case, there is no need to make a unique hard decision first; instead, the actual coordinates of the received symbol are retained for comparison with the distances between it and candidate constellation points with different bit tags. As another example, when multiple high-order QAM symbols are received consecutively, subsequent processing can be performed symbol by symbol. For each equalized complex symbol acquired, its target constellation sub-region is determined, candidate constellation points are extracted, and the corresponding soft information is calculated, ultimately forming a soft information sequence according to the symbol reception order.

[0025] Considering that traditional maximum likelihood soft demodulation typically requires calculating the distance between a received symbol and a large number of constellation points in the constellation graph for each received symbol, and that the number of constellation points increases significantly with modulation order (e.g., a 64th-order QAM contains 64 constellation points, and a 256th-order QAM contains 256), traversing all constellation points for each received symbol increases computational and storage requirements. However, the transmitted constellation points most likely to correspond to a received symbol are usually concentrated near its current location. Constellation points far from the received symbol contribute little to the minimum distance result. Therefore, based on the in-phase component coordinates, orthogonal component coordinates, and constellation mapping rules of the received symbol, the entire constellation plane can be divided into multiple constellation sub-regions, and a set of candidate constellation points associated with each sub-region can be pre-configured. When a received symbol falls into a constellation sub-region, only the candidate constellation points corresponding to that region need to be called, without searching the entire constellation graph. The constellation sub-region is not equivalent to the hard decision region of a single constellation point. Its role is to determine the local search range required for subsequent soft information calculations. Therefore, candidate constellation points configured for a target constellation sub-region can include constellation points located within that region, or constellation points located near the boundary of that region and in adjacent regions.

[0026] For example, a correspondence between constellation sub-regions and candidate constellation points can be pre-established based on the positions of decision boundaries in a preset constellation diagram. Upon receiving a complex symbol, the coordinates of its in-phase component are compared with the boundaries of each horizontal region, and its orthogonal component coordinates are compared with the boundaries of each vertical region, thereby determining the target constellation sub-region where the symbol resides. Subsequently, based on the identifier of the target constellation sub-region, candidate constellation points and their corresponding bit markers are read from the preset correspondence. For instance, a received symbol may be located in the upper right part of the constellation diagram, close to the inner decision boundary. When configuring candidate constellation points for the constellation sub-region corresponding to this location, it can simultaneously include constellation points that are relatively close within the upper right region, as well as several neighboring constellation points on the other side of the decision boundary. Thus, even if noise causes the received symbol to cross local boundaries, the candidate point set may still contain the actually transmitted constellation points. As another example, a received symbol may be located at the outer edge of the constellation diagram. Since there are no standard constellation points outside it, candidate points for this target constellation sub-region can be mainly selected from the inner side and constellation points adjacent along the outer edge direction. Compared to the area near the center of the constellation map, the spatial distribution of candidate constellation points in the outer region may not be perfectly symmetrical, but it should still be ensured that candidate constellation points corresponding to the first and second bit values ​​can be found for each bit. For received symbols near the intersection of multiple sub-regions, the configured candidate constellation points can cover multiple directions around the intersection, so that the local candidate set can reflect the main decision probability near that location.

[0027] Understandably, a high-order QAM constellation point has a complete bit tag, but the output of soft demodulation is individual bits. Therefore, the same received symbol needs to calculate soft information for each bit it carries. For a specific bit being processed, candidate constellation points can be divided into two categories according to the tag value of that bit: one category is the first candidate constellation point with the bit tagged as the first bit value, and the other category is the second candidate constellation point with the bit tagged as the second bit value. The first bit value and the second bit value can correspond to binary zero and binary one, respectively, or they can correspond in reverse according to the soft information symbol convention adopted by the system, but they should remain consistent throughout the processing. The distance between the received symbol and the nearest constellation point in the first candidate constellation point can reflect the most favorable symbol interpretation under the assumption that "the current bit is the first bit value"; the distance between the received symbol and the nearest constellation point in the second candidate constellation point can reflect the most favorable symbol interpretation under the assumption that "the current bit is the second bit value". If the minimum distance of the first category is significantly smaller than the minimum distance of the second category, it means that the received symbol supports the first bit value more; otherwise, it supports the second bit value more. If the two minimum distances are close, it indicates that the received symbol is near the decision boundary of that bit, the support for the two bit values ​​is similar, and the absolute value of the corresponding soft information should be small.

[0028] For example, for all candidate constellation points in a target constellation sub-region, the in-phase component coordinates, orthogonal component coordinates, and complete bit markers for each candidate constellation point can be pre-stored. For the current bit, the marker for each candidate constellation point at that bit is read. Candidate constellation points marked with a first bit value are assigned to a first candidate set, and those marked with a second bit value are assigned to a second candidate set. Then, the Euclidean distances between the received symbol and each candidate constellation point in both sets are calculated, and a first minimum Euclidean distance is determined from the first candidate set, and a second minimum Euclidean distance is determined from the second candidate set. A distance difference is formed based on the two minimum Euclidean distances, and the initial log-likelihood ratio for that bit is obtained accordingly. In actual calculations, the Euclidean distance is determined by the coordinate difference between the received symbol and the candidate constellation point in the in-phase component direction and the coordinate difference in the orthogonal component direction. Using squared distances for comparison avoids additional square root operations while still maintaining the distance magnitude relationship. For example, suppose a received symbol is close to a candidate constellation point whose marker at the current bit is the first bit value. Meanwhile, the nearest candidate constellation point for the second bit value is located on the other side of the decision boundary and is significantly farther from the received symbol. In this case, the difference between the two minimum distances is large, and the initial log-likelihood ratio should have a large absolute value, indicating a high decision confidence for this bit. If the received symbol is exactly between two neighboring constellation points, and these two constellation points have different labels on the current bit, then the first minimum Euclidean distance is close to the second minimum Euclidean distance. In this case, the distance difference is close to zero, indicating that a high-confidence judgment for this bit is not advisable. For another bit, the classification results of the same group of candidate constellation points may differ. Two spatially adjacent constellation points may have different labels on the previous bit, but the same label on the next bit. Therefore, each bit must undergo candidate point classification and minimum distance comparison separately; soft information for all bits cannot be generated all at once based solely on the nearest single constellation point of the received symbol. For example, the nearest and second-nearest constellation points of a received symbol have the same label on the current bit. In this case, the minimum distance corresponding to the second bit value may need to be obtained from another, more distant candidate constellation point. The resulting large distance difference reflects the high stability of the bit at its current position.

[0029] Understandably, the geometric characteristics of the decision boundaries can be reflected in the following properties directly related to constellation mapping: First, the direction of the decision boundaries. For example, a bit mainly determined by the interval containing the in-phase component may have its decision boundary extending primarily along the direction of the orthogonal component; another bit mainly determined by the interval containing the orthogonal component may have its decision boundary extending primarily along the direction of the in-phase component. Second, the number of decision boundaries. Bits used to distinguish large constellation regions may have fewer main decision boundaries; bits used to distinguish local points may have more decision boundaries. Third, the spacing between decision boundaries. Bits with larger boundary spacing usually maintain the same bit label over a wider spatial range; bits with smaller boundary spacing are more likely to cross decision boundaries due to noise disturbances. Fourth, the positional relationship between the decision boundaries and the constellation outer edge, constellation center, and constellation sub-regions. After local candidate point reduction, the decision relationships near the constellation outer edge and those within the constellation may be affected differently.

[0030] For example, when determining the preset high-order QAM constellation mapping rules, the distribution of constellation point markers corresponding to each bit can be analyzed, and a correspondence between bits and bit groups can be pre-established accordingly. For instance, bits with larger decision regions, mainly used to distinguish the overall position of the constellation map, can be assigned to one bit group; bits with denser decision boundaries, mainly used to distinguish local adjacent points, can be assigned to another bit group. Alternatively, based on the actual constellation mapping, bits with the same decision boundary direction and similar spacing can be assigned to the same bit group. For constellation mappings where the in-phase component direction and the orthogonal component direction have a symmetrical relationship, if the decision boundaries of two bits correspond only in direction, while the number of boundaries, spacing, and local structure are basically the same, they can also be assigned to the same bit group. After completing the bit grouping, compensation weights are determined for each bit group. The compensation weights can be determined based on the distance difference scale of the corresponding bit group, the deviation caused by the approximation of local candidate points, and the desired soft information confidence scale. For bit groups whose initial log-likelihood ratio is easily amplified, a relatively small compensation weight can be configured; for bit groups whose initial log-likelihood ratio is easily compressed, a relatively large compensation weight can be configured.

[0031] For example, the sign of the initial log-likelihood ratio indicates whether the currently received symbol supports the first bit value or the second bit value more, and its absolute value indicates the degree of support. Weighting the initial log-likelihood ratio according to the compensation weights configured for the bit groups can adjust the confidence scale without typically changing its bit bias. When the compensation weights are positive, the weighted correction usually does not change the positive or negative attribute of the initial log-likelihood ratio, thus not arbitrarily reversing the more likely bit value, but it can change the absolute value of the soft information. When the compensation weights are small, the absolute value of the corrected soft information decreases accordingly, indicating a decrease in the confidence level of the current bit judgment; when the compensation weights are large, the absolute value of the corrected soft information increases accordingly, indicating an increase in the influence of the bit judgment in subsequent decoding. For the current high-order QAM baseband complex symbol, the initial log-likelihood ratio of each bit it carries is obtained sequentially. Based on the pre-established correspondence between bits and bit groups, the bit group to which the current bit belongs is determined, and the compensation weight corresponding to that bit group is read. Subsequently, the initial log-likelihood ratio of the current bit is weighted and corrected using this compensation weight. After processing all bits of the current symbol, soft information is output according to a preset bit order. Subsequent received symbols are then processed, and the soft information corresponding to each symbol is arranged according to the symbol reception order, thus forming a soft information sequence. For example, suppose the initial log-likelihood ratio of a certain bit is very large, but the bit group to which this bit belongs has a large decision region, and the calculation of local candidate points can easily overestimate its confidence. After using a relatively small compensation weight, the correction result still supports the original bit value, but the confidence level is more moderate. This can reduce the adverse effects of erroneous high-confidence information on the subsequent decoding process. As another example, another bit is located in a bit group with dense decision boundaries. Although its initial log-likelihood ratio can correctly represent the bit tendency, its absolute value is generally small. After using a relatively large compensation weight, the distinguishability of the soft information of this bit can be enhanced, allowing the subsequent decoder to make fuller use of the reliable information provided by this bit. If a received symbol happens to be close to the decision boundary of the current bit, its initial log-likelihood ratio is close to zero. Even with appropriate compensation weights for correction, the result will still retain a small absolute value, thus continuing to reflect the high uncertainty of that bit. If a received symbol is significantly close to the candidate constellation point corresponding to the first bit value and far from the candidate constellation point corresponding to the second bit value, its initial log-likelihood ratio will have a large absolute value. After weighted correction, the soft information can still reflect the high decision reliability of that bit, only adjusted to a more reasonable numerical scale based on the statistical characteristics of the bit group.

[0032] In summary, based on the above scheme, the number of Euclidean distance operations required for high-order QAM soft information computation can be reduced. The higher the modulation order and the more symbols to be processed, the more significant the reduction in computational load brought by local candidate search compared to full constellation search. It can reduce storage access pressure. The receiver can read the corresponding candidate constellation points and bit markers according to the target constellation sub-region, without having to repeatedly read and compare all constellation points. It can preserve soft decision information. This scheme does not directly convert the received symbol into a unique hard decision result, but compares the minimum distance under two bit assumptions for each bit, thus outputting soft information reflecting the decision direction and confidence level. It can reduce the non-uniform error between bits caused by the reduction of local candidate points. Since the decision boundary geometry of bits at different positions is different, using a uniform computation scale may lead to soft information imbalance. By determining the compensation weight by grouping, the soft information scale of different types of bits can be more coordinated. It can also improve the effectiveness of the soft information used in subsequent decoding. Appropriately suppressing overestimated bit confidence can reduce the misleading effect of erroneous high-confidence information on the decoding result; appropriately enhancing underestimated bit confidence can prevent effective information from playing an insufficient role in the decoding process.

[0033] Next, assume that the higher-order QAM baseband complex symbols after carrier synchronization, symbol synchronization, and channel equalization are: in, Represents the coordinates of the in-phase components. Represents the orthogonal component coordinates. Candidate constellation points are represented as: The squared Euclidean distance between the received symbol and the candidate constellation point can be expressed as: It is understandable that the example uses the first bit as binary zero, the second bit as binary one, and the square of the second minimum Euclidean distance minus the square of the first minimum Euclidean distance as the positive direction. In actual implementation, the opposite sign convention can also be used, as long as the positive and negative meanings of the soft information are consistent with the input convention of the low-density parity-check code decoder.

[0034] In some examples, calculating the distance difference between corresponding bits includes: Calculate the first minimum Euclidean distance between the high-order QAM baseband complex symbol signal and the first candidate constellation point, and the second minimum Euclidean distance between the high-order QAM baseband complex symbol signal and the second candidate constellation point, so as to determine the distance difference of the corresponding bit based on the square difference between the first minimum Euclidean distance and the second minimum Euclidean distance.

[0035] For example, for the current bit, the in-phase component coordinates and orthogonal component coordinates of the first candidate constellation point are read sequentially. The squared Euclidean distance between the received symbol and each first candidate constellation point is calculated, and the minimum value is retained through successive comparisons. Subsequently, the same process is performed on the second candidate constellation point to obtain the square of the second minimum Euclidean distance. Finally, the difference between the two is calculated according to a predetermined symbol convention to obtain the distance difference for the corresponding bit. For example, the received symbol is... The nearest first candidate constellation point corresponding to a certain bit is... Then the square of the first least Euclidean distance is: The corresponding second most recent candidate constellation point is Then the square of the second least Euclidean distance is: Distance difference is This indicates that the received symbol clearly supports the first bit value. If the received symbol is located between two different marker constellation points, such that the square of the first minimum Euclidean distance is 1.02 and the square of the second minimum Euclidean distance is 1.08, then the distance difference is only 0.06. Although its symbol still tends towards the first bit value, the confidence level is low.

[0036] In some examples, it also includes: Estimate the signal-to-noise ratio (SNR) of the current channel, determine the noise variance compensation factor based on the SNR, and obtain the initial log-likelihood ratio of each bit based on the noise variance compensation factor and the distance difference corresponding to each bit.

[0037] Understandably, the distance difference calculated solely from geometric distance does not fully reflect the strength of the current channel noise. Under the same constellation positional relationships, distance differences in low-noise channels generally have higher reliability, while the same distance difference in high-noise channels may simply be caused by random noise perturbations. Under additive Gaussian noise conditions, the approximate value of the bit log-likelihood ratio is proportional to the difference in the least squared distances corresponding to the two bit hypotheses and inversely proportional to the noise variance. Therefore, a noise variance compensation factor can be determined based on the signal-to-noise ratio, and then this compensation factor can be applied to the distance difference of each bit to obtain the initial log-likelihood ratio. in, Let represent the initial log-likelihood ratio of the i-th bit. This represents the noise variance compensation factor. When the signal-to-noise ratio (SNR) is high, the position of the received symbol is less affected by noise, and the same distance difference can be assigned a larger absolute value of soft information. When the SNR is low, the received symbol may cross the decision boundary due to noise, requiring a reduction in the compensation factor to avoid over-interpreting distance differences.

[0038] For example, the receiver first estimates the signal-to-noise ratio (SNR) of the current channel, and then determines a noise variance compensation factor based on the estimated SNR. Subsequently, for each bit carried by the current symbol, the corresponding distance difference is multiplied by the compensation factor. For example, the distance difference for a certain bit is 4. Under lower SNR conditions, the noise variance compensation factor is 0.3, and the initial log-likelihood ratio is 1.2. Under higher SNR conditions, the compensation factor is 0.8, and the initial log-likelihood ratio is 3.2. The symbols of the soft information are the same in both cases, indicating the same preferred bit value, but the higher SNR condition provides a higher level of confidence. As another example, the distance difference for a certain bit is only 0.1. Even with a high channel SNR, the initial log-likelihood ratio obtained after compensation is still relatively small. This is because the symbol itself is located near the decision boundary of the corresponding bit, and cannot be interpreted as a high-reliability bit simply because the overall channel quality is good.

[0039] In some examples, it also includes: After acquiring the high-order QAM baseband complex symbol signal after carrier synchronization, symbol synchronization and channel equalization, the amplitude of the high-order QAM baseband complex symbol signal is normalized so that the magnitude of the normalized high-order QAM baseband complex symbol signal matches the magnitude of the standard rectangular constellation diagram stored locally. Determining the target constellation sub-region corresponding to the higher-order QAM baseband complex symbol signal includes: The target constellation sub-region is determined based on the in-phase and quadrature component coordinates of the normalized high-order QAM baseband complex symbol signal.

[0040] Understandably, the target constellation sub-region is determined based on the local standard rectangular constellation diagram. If the overall amplitude scale of the received symbol is inconsistent with the standard constellation diagram, even if the signal has undergone channel equalization, it may be assigned to the wrong region due to residual gain error. For example, the coordinate of a certain in-phase component in the standard constellation diagram should be close to 7, but the residual gain of the receiving link reduces it to 4.9. If the region is determined directly based on 4.9, the receiver may assign the symbol to an inner region with a smaller amplitude, thus calling an inappropriate candidate constellation point. The purpose of amplitude normalization is to adjust the overall scale of the received symbol so that its magnitude distribution matches the magnitude scale of the local standard rectangular constellation diagram. Normalization does not project every symbol onto the unit circle, but rather makes the overall amplitude scale of the constellation diagram in which the received symbol is located correspond to the standard constellation diagram. The normalized symbol can be represented as: in, This represents the amplitude normalization coefficient.

[0041] For example, after acquiring synchronized and equalized high-order QAM baseband complex symbols, normalization coefficients are determined based on the current amplitude scale of the received signal and the amplitude scale of the local standard rectangular constellation diagram. The in-phase and quadrature components are simultaneously multiplied by the normalization coefficients to obtain the normalized complex symbols. Subsequently, the normalized in-phase and quadrature component coordinates are compared with the locally stored region boundaries to determine the target constellation sub-region. For example, if the magnitude scale of the corresponding point in the standard constellation diagram is approximately 1.25 times the magnitude of the received symbol, then 1.25 can be used as the normalization coefficient. The original received symbol is... After normalization, it becomes Subsequent region determination uses This avoids misjudgment of the region due to residual amplitude reduction. If the overall amplitude of the received signal is too large, the normalization coefficient can be less than one. For example, if the received symbol is... However, under standard scales it should be close to The constellation sub-regions are then determined after scaling them up by the appropriate ratio.

[0042] In some examples, the higher-order QAM is a 128th-order quadrature amplitude modulation, where each higher-order QAM baseband complex symbol signal carries seven bits. Dividing these bits into at least two bit groups includes: The three bits used to determine the region to which the constellation point belongs are divided into the high-order bit group, and the remaining four bits used to distinguish the sub-points within the region are divided into the low-order bit group.

[0043] For example, the receiver pre-determines, based on a preset 128-order QAM constellation mapping rule, which of the seven bits are used to mark the large area and which are used to mark the sub-points within the large area. Three bits used to determine the large area are registered as the high-order bit group, and the remaining four bits are registered as the low-order bit group. Here, high-order and low-order bits represent their role in the constellation mapping hierarchy and are not required to be mechanically divided according to binary numerical order. After calculating seven initial log-likelihood ratios for a received symbol, the corresponding compensation weight is applied according to the bit group to which each bit belongs. For example, the seven bits are represented sequentially as follows: to If the preset mapping rules stipulate... , and To determine a large region, it is assigned to the high-order bit group; , , and Assign it to the low-order bit group. When the mapping rule uses other bit positions to determine the region, the grouping should be based on the actual mapping relationship, rather than assuming that the three bits with the larger numbers necessarily belong to the high-order bit group.

[0044] In some examples, determining the target constellation sub-region corresponding to the high-order QAM baseband complex symbol signal based on the in-phase component coordinates and quadrature component coordinates of the high-order QAM baseband complex symbol signal, and a preset high-order QAM constellation mapping rule, includes: Based on the positive and negative attributes of the in-phase component and the quadrature component, the 128th order quadrature amplitude modulation constellation diagram is divided into four primary regions, and each primary region is further divided into multiple amplitude regions based on the amplitude of the in-phase component and the quadrature component. According to the decision boundary corresponding to the low-order bit group, each amplitude region is divided into multiple constellation sub-regions; Based on the in-phase component coordinates, quadrature component coordinates, and amplitude of the high-order QAM baseband complex symbol signal, the target constellation sub-region into which the high-order QAM baseband complex symbol signal falls is determined.

[0045] For example, for a normalized received symbol, the signs of the in-phase and quadrature components are first determined. When both the in-phase and quadrature components are positive, it is assigned to the first-level region; when both are negative, it is assigned to the second-level region; when both are negative, it is assigned to the third-level region; and when both are positive, it is assigned to the fourth-level region. After determining the first-level regions, the absolute values ​​of the in-phase and quadrature components are compared with preset amplitude boundaries. For example, smaller amplitude, intermediate amplitude, and larger amplitude can correspond to different amplitude regions. The specific boundaries should be consistent with the local 128-order QAM constellation coordinates and mapping rules. Subsequently, the received symbol coordinates are compared with the local decision boundary corresponding to the lower bit group to determine the constellation sub-region it falls into within the current amplitude region. For example, if both the in-phase and quadrature components of a received symbol are positive, it is first located in the first-level region. Its in-phase component has a larger absolute value, while the quadrature component's absolute value is in the middle range. Therefore, it is located in the corresponding amplitude region within this first-level region. If it is also located to the right of a low-bit decision boundary, it is further determined to be the constellation sub-region corresponding to that side. For example, two symbols are both located in the first-level region, but one symbol is closer to the coordinate axis, while the other symbol is closer to the outer edge of the constellation. By dividing the amplitude region, different candidate constellation points can be assigned to them respectively, avoiding the introduction of too many irrelevant candidate points based solely on quadrants.

[0046] In some examples, each target constellation sub-region is configured with eight to sixteen candidate constellation points, and obtaining the initial log-likelihood ratio for each bit includes: For each bit, calculate the Euclidean distance between the high-order QAM baseband complex symbol signal and the corresponding eight to sixteen candidate constellation points, and determine the first minimum Euclidean distance and the second minimum Euclidean distance from them; Calculate the difference between the square of the first minimum Euclidean distance and the square of the second minimum Euclidean distance to obtain the distance difference amount; The product of the distance difference and the noise variance compensation factor is determined as the initial log-likelihood ratio of the corresponding bit, so as to avoid traversing all constellation points in the high-order QAM constellation diagram.

[0047] For example, the coordinates and bit markers of eight to sixteen candidate constellation points can be pre-stored for each constellation sub-region. After the received symbol falls into the target constellation sub-region, the candidate points corresponding to that region are read. For the first bit, these candidate points are divided into two categories according to the bit marker, and the first minimum Euclidean distance and the second minimum Euclidean distance are calculated respectively. The difference between the squares of the two is calculated and multiplied by the noise variance compensation factor to obtain the initial log-likelihood ratio of the first bit. Subsequently, the above process is repeated for the remaining bits. The same candidate constellation point may be classified into different categories under different bit positions, but its coordinate distance depends only on the position of the received symbol and the candidate point. Therefore, in the specific implementation, the squared Euclidean distance between the received symbol and each candidate point can be calculated first, and then the minimum value is compared according to the bit marker. For example, a target constellation sub-region is configured with twelve candidate constellation points. For a certain bit, seven candidate points are marked as the first bit value, and five candidate points are marked as the second bit value. The minimum value is selected from the seven distances and the five distances respectively to form the distance difference of that bit. If the received symbol is close to the boundary of a constellation sub-region, the candidate points can include neighboring constellation points on the other side of the boundary. This way, even if noise causes the received symbol to fall into an adjacent sub-region, the comparison object related to the actual transmission point can still be retained in the current candidate set.

[0048] In some examples, estimating the signal-to-noise ratio (SNR) of the current channel and determining a noise variance compensation factor based on the SNR includes: The noise variance of the current channel is obtained by statistically averaging the time-domain noise samples within a preset statistical period, and the signal-to-noise ratio of the current channel is determined based on the noise variance. Determine the signal-to-noise ratio range in which the signal-to-noise ratio is located, and read the noise variance compensation factor corresponding to the signal-to-noise ratio range from the preset segmentation parameter lookup table; The noise variance compensation factor is negatively correlated with the noise variance and positively correlated with the signal-to-noise ratio.

[0049] Understandably, directly performing division based on the real-time noise variance requires continuous high-precision calculations, which is not advantageous for fixed-point processors or resource-constrained hardware implementations. This example first performs statistical averaging on time-domain noise samples within a preset statistical period to obtain a more stable noise variance estimate; then, it determines the signal-to-noise ratio (SNR) of the current channel based on the noise variance; finally, it determines the SNR interval and reads the corresponding compensation factor from a preset segmentation parameter lookup table. A larger noise variance indicates greater uncertainty in the received symbol position, and the smaller the compensation factor should be. A higher SNR indicates a higher degree of reliability in the distance difference, and the compensation factor can be larger. Therefore, the compensation factor is negatively correlated with the noise variance and positively correlated with the SNR.

[0050] For example, multiple time-domain noise samples are acquired within a preset statistical period. The squared amplitudes of the noise samples are statistically averaged to obtain an estimate of the noise variance of the current channel. For instance, N complex noise samples are acquired within the statistical period. The noise variance can then be obtained based on the average energy of these noise samples. The signal-to-noise ratio (SNR) is then determined by combining this with the power scale of the currently received signal. After determining the SNR, it is compared with the boundaries of a preset interval. For example, it is determined whether it belongs to a low, medium, or high SNR interval. Based on the determination, the corresponding compensation factor is read and applied to the distance difference of each received symbol within that time period. The role of statistical averaging is to reduce the impact of individual instantaneous noise spikes on the noise variance estimation. For example, if the amplitude of a noise sample suddenly increases, adjusting the compensation factor only based on that sample may cause unnecessary drastic changes in the soft information scale; by averaging the samples within the preset statistical time period, the compensation factor mainly reflects the overall noise level of that time period.

[0051] In some examples, the segmented parameter lookup table includes at least a low signal-to-noise ratio (SNR) level, a medium SNR level, and a high SNR level; The noise variance compensation factor for the low signal-to-noise ratio setting is 0.3, the noise variance compensation factor for the medium signal-to-noise ratio setting is 0.5, and the noise variance compensation factor for the high signal-to-noise ratio setting is 0.8.

[0052] For example, the receiver first determines the current signal-to-noise ratio (SNR) level according to the preset interval boundaries. If it belongs to the low SNR level, it reads 0.3; if it belongs to the medium SNR level, it reads 0.5; and if it belongs to the high SNR level, it reads 0.8. For example, the distance difference of a certain bit is 10. In the low SNR level, its initial log-likelihood ratio is 3; in the medium SNR level, it is 5; and in the high SNR level, it is 8. As another example, the distance difference is -6. In the three levels, it can be -1.8, -3, and -4.8, respectively. The compensation factor only adjusts the absolute value and does not change the direction of the bit values ​​it supports.

[0053] In some examples, determining the compensation weights corresponding to each bit group includes: The compensation weight corresponding to the high-order bit group is set to 0.6 to 0.8 to suppress the initial log-likelihood ratio of each bit in the high-order bit group from becoming saturated. The compensation weights corresponding to the low-order bit group are set to 1.2 to 1.5 to enhance the bit confidence represented by the initial log-likelihood ratio of each bit in the low-order bit group.

[0054] For example, after calculating the initial log-likelihood ratio for each of the seven bits, it is determined whether the current bit belongs to the high-order bit group or the low-order bit group. If it belongs to the high-order bit group, a preset compensation weight is selected from the range of 0.6 to 0.8; if it belongs to the low-order bit group, a preset compensation weight is selected from the range of 1.2 to 1.5. Then, the initial log-likelihood ratio is multiplied by the corresponding weight. For example, if the initial log-likelihood ratio of the high-order bits is 80, using 0.7 as the compensation weight results in a corrected result of 56. This result still supports the original bit value but reduces the likelihood of reaching the amplitude limit later. As another example, if the initial log-likelihood ratio of the low-order bits is 8, using 1.4 as the compensation weight results in a corrected result of 11.2, thereby enhancing its distinguishability from low-confidence soft information that is close to zero. If the initial log-likelihood ratio of a certain low-order bit is close to zero, for example, 0.02, even with a weight of 1.5, the corrected value is only 0.03, which can still reflect that the bit is near the decision boundary and will not be arbitrarily transformed into high-confidence information.

[0055] In some examples, before outputting the soft information sequence corresponding to the higher-order QAM baseband complex symbol signal, the method further includes: The mean and variance of the weighted log-likelihood ratio corresponding to each bit are normalized so that the variance of the log-likelihood ratio corresponding to each bit converges to the target variance interval. The normalized log-likelihood ratio corresponding to each bit is compared with a preset threshold range. The log-likelihood ratio that is greater than the upper limit of the preset threshold range is truncated to the upper limit, and the log-likelihood ratio that is less than the lower limit of the preset threshold range is truncated to the lower limit. According to the receiving order of each high-order QAM baseband complex symbol signal and the arrangement order of each bit, the output clipped soft information sequence is used for iterative decoding by the low-density parity-check code decoder.

[0056] For example, digital receivers and decoders typically use finite bit widths to represent soft information. If the log-likelihood ratio exceeds the representable range, numerical overflow may occur; even without overflow, excessively large absolute values ​​of soft information may cause the decoder to have overly strong confidence in a particular erroneous bit. Therefore, the normalized log-likelihood ratio is compared with a preset threshold range. Values ​​greater than the upper limit are truncated to the upper limit, values ​​less than the lower limit are truncated to the lower limit, and values ​​within the range retain their original size. For example, the threshold range is -64 to 63. When the normalized log-likelihood ratio is 85, the output is 63; when it is -72, the output is -64; and when it is 21, the output is still 21. This processing does not turn all large soft information into the same value, but only truncates the portion that exceeds the representable range or the predetermined confidence range. Soft information within the range retains its relative size.

[0057] In some examples, the preset amplitude limiting threshold range is dynamically determined based on the current channel signal-to-noise ratio, including: When the signal-to-noise ratio is below 12 dB, the preset amplitude limiting threshold range is determined to be -32 to 31. When the signal-to-noise ratio is greater than or equal to 12 dB and less than 15 dB, the preset amplitude limiting threshold range is determined to be -64 to 63. When the signal-to-noise ratio is greater than or equal to 15 dB, the preset amplitude limiting threshold range is determined to be -128 to 127.

[0058] For example, when the signal-to-noise ratio (SNR) of the current channel is below 12 dB, the clipping threshold range is determined to be -32 to 31. The normalized log-likelihood ratio is truncated to 31 when it exceeds 31 and to -32 when it is below -32. When the SNR is greater than or equal to 12 dB and less than 15 dB, the clipping threshold range is determined to be -64 to 63. When the SNR is greater than or equal to 15 dB, the clipping threshold range is determined to be -128 to 127. For example, the normalized log-likelihood ratio is 70. When the SNR is 10 dB, this value is truncated to 31; when the SNR is 13 dB, it is truncated to 63; and when the SNR is 17 dB, it can be retained as 70. As another example, the normalized log-likelihood ratio is -90. The output is -32 under low signal-to-noise ratio (SNR) conditions, -64 under medium SNR conditions, and -90 can be retained under high SNR conditions. When the SNR is exactly 12 dB, the range of -64 to 63 is applicable; when the SNR is exactly 15 dB, the range of -128 to 127 is applicable. This clarifies the processing rules at the boundaries of these ranges.

[0059] In some cases, considering that in high-order QAM constellation diagrams, the location of the received symbol is usually first assigned to a target constellation sub-region, and then a small number of candidate constellation points are selected from this target constellation sub-region to reduce the amount of subsequent distance calculations. This approach usually yields good results when the received symbol is far from the region boundary, because the actual transmitting constellation point corresponding to the received symbol, as well as the competing heterograph constellation points, are likely located within the target constellation sub-region. However, when the received symbol is simultaneously close to the region boundary in both the in-phase component direction and the quadrature component direction, the received symbol may actually be near the boundary formed by multiple constellation sub-regions. In this case, even small noise perturbations on the in-phase or quadrature components may cause the received symbol to be assigned to an adjacent region. For example, the received symbol may be located near the common boundary of four constellation sub-regions. When the in-phase component is subjected to positive noise perturbation and the quadrature component is subjected to negative noise perturbation, the received symbol may be assigned to the lower right region; when the directions of the two noise perturbations change, the received symbol corresponding to the same transmitting constellation point may be assigned to the upper left region. If candidate constellation points are extracted from only one selected region each time, the actual transmitted constellation points or those competing with it may be excluded from the candidate set. Therefore, in some examples, determining the target constellation sub-region corresponding to the received symbol and selecting candidate constellation points from the target constellation sub-region includes: Determine the boundary distances from the received symbol to the boundaries of each relevant region in the in-phase component direction and the quadrature component direction, respectively; When the boundary distance from the received symbol to the boundary of at least two different directions is less than the corresponding boundary uncertainty, multiple constellation sub-regions that are jointly associated with the boundary of the at least two different directions are identified as candidate sub-regions, and candidate constellation points are extracted from the multiple candidate sub-regions. For each bit carried by the received symbol, check whether the candidate constellation points contain valid comparison points located on both sides of the nearest decision boundary corresponding to that bit and having different bit tags; In the absence of a valid comparison point on either side of the nearest decision boundary, the nearest heteromarked constellation point on the other side of the nearest decision boundary is added, and candidate constellation points that cannot form the minimum distance point for each bit are deleted from the added candidate constellation points.

[0060] Understandably, the boundary distances from the received symbol to the relevant region boundary are calculated separately. For received symbols represented by in-phase component coordinates and orthogonal component coordinates, the absolute values ​​of the differences between their in-phase component coordinates and the boundaries of adjacent in-phase directions, and the absolute values ​​of the differences between their orthogonal component coordinates and the boundaries of adjacent orthogonal directions, can be calculated separately. Boundary uncertainty is used to indicate whether the region assignment of the received symbol is sufficiently deterministic under current noise, phase error, amplitude normalization error, or fixed-point quantization error conditions. Boundary uncertainty can be a pre-set fixed value or dynamically determined based on the signal-to-noise ratio, residual error, or historical region misclassification probability. When the received symbol is close to only one region boundary, the two constellation sub-regions on both sides of that boundary can be identified as candidate sub-regions. When the received symbol is close to two region boundaries in different directions simultaneously, the multiple constellation sub-regions associated with these region boundaries are all identified as candidate sub-regions. For cases where lateral and longitudinal boundaries intersect, candidate sub-regions can include the four constellation sub-regions around the intersection. However, simply merging constellation points in multiple constellation sub-regions still cannot guarantee that every bit obtains a valid soft information comparison object. The reason is that soft information generation requires not only the existence of constellation points with bit markers of zero and one in the candidate set, but also that these constellation points with different bit markers represent the two most predominant competing hypotheses near the current receiving position. For each bit, the nearest decision boundary near the current receiving position can be determined. This decision boundary separates the constellation points with the bit markers of zero and one. Only when the candidate set contains valid comparison points on both sides of this boundary can the difference between the two minimum distances accurately represent the tendency of the received symbol for the two bit values. If the candidate set lacks a valid comparison point on one side of the decision boundary, the nearest hetero-marked constellation point to the received symbol on the other side of the decision boundary is searched and added to the candidate set. Adding along the nearest decision boundary does not mean that only constellation points located on the decision boundary can be selected, but rather that the nearest hetero-marked constellation point is selected from the other side of the boundary based on the competing relationship between the two types of bit hypotheses corresponding to the decision boundary. After the addition is completed, the number of candidates may exceed the upper limit preset by the hardware or algorithm. Therefore, candidate constellation points that cannot be the minimum distance points in the minimum distance comparison of all bits can be deleted. In other words, if a candidate point is not the closest candidate point to the received symbol for any bit under any bit assumption, then deleting the candidate point will not change the minimum distance calculation result for any bit.

[0061] For example, in one implementation, the receiver first performs amplitude normalization on the received symbols to obtain normalized in-phase components. and normalized orthogonal components The region boundary sets for the in-phase component direction and the orthogonal component direction are pre-defined based on the 128-order QAM constellation structure. Assume the in-phase direction boundary adjacent to the currently received symbol is... The adjacent orthogonal direction boundaries are Then they can be calculated separately: in, This represents the boundary distance from the received symbol to the in-phase direction boundary. This represents the boundary distance from the received symbol to the orthogonal direction boundary. Further determination of the in-phase direction boundary uncertainty is needed. Boundary uncertainty in orthogonal directions In a simplified implementation, the two boundary uncertainties can be set as a predetermined proportion of the distance between adjacent points in the constellation. In a dynamic implementation, they can be determined based on the current noise standard deviation. For example, as the current noise standard deviation increases, the boundary uncertainties increase accordingly; as the current channel quality improves, the boundary uncertainties decrease accordingly. Not less than ,and Not less than When the received symbol is far from the boundaries of the regions in both directions, the unique target constellation sub-region can be determined in a conventional manner. Less than ,but Not less than When, the two constellation sub-regions on either side of the boundary in the same phase direction are identified as candidate sub-regions. Not less than ,but Less than When this is the case, the two constellation sub-regions on either side of the orthogonal direction boundary are determined as candidate sub-regions. Less than ,and Less than In this process, multiple constellation sub-regions around the intersection of the two boundaries are identified as candidate sub-regions. Under the rectangular region partitioning method, four candidate sub-regions are typically obtained. Constellation points are then extracted from these candidate sub-regions and deduplication is performed to form an initial candidate set. For each bit carried by a 128-order QAM received symbol, the following processing is performed: First, according to the bit's label, candidate constellation points are divided into candidate points labeled zero and candidate points labeled one. Second, the nearest decision boundary for this bit near the current receiving position is determined according to the constellation mapping rules, and it is determined whether the candidate set contains valid comparison points located on both sides of the decision boundary. Third, if the candidate set only contains valid comparison points on one side of the decision boundary, or if the distance of an existing candidate point on the other side is significantly greater than the distance of the nearest dissimilar constellation point obtainable along the decision boundary, then the nearest dissimilar constellation point is added from the other side of the decision boundary. Finally, a dominance check is performed on the candidate set after point addition. If a candidate constellation point cannot be the minimum distance point in either type of label distance comparison for all seven bits, it can be deleted. During dominance checks, the minimum distance points marked as zero and one corresponding to each bit can be recorded first, and then candidate constellation points selected by at least one bit can be retained. To prevent the minimum distance points from switching after slight changes in the received symbol, the second smallest distance point corresponding to each bit can also be retained.

[0062] In some cases, the ordinary squared Euclidean distance is usually expressed as: in, and These represent the in-phase and quadrature components of the received symbol, respectively. and Let represent the in-phase and quadrature components of the k-th candidate constellation point, respectively. This distance calculation method implicitly assumes two basic assumptions: the noise variances in the in-phase and quadrature directions are the same; and the noise in the two directions is uncorrelated. When these assumptions are met, the received symbol forms an approximately circular probability contour line around the transmitting constellation point, and two positions with the same Euclidean distance have approximately the same probability of occurrence. In actual receiving systems, inconsistent analog front-end gain, different filtering characteristics of the in-phase and quadrature branches, sampling clock errors, narrowband interference, and residual image components can all cause the in-phase and quadrature noise to exhibit different variances or correlations. In this case, the probability contour line formed by the received symbol around the transmitting constellation point may be elliptical, and the principal axis of the ellipse may be tilted relative to the in-phase coordinate axis. In this situation, the same geometric distance does not necessarily mean the same probability of occurrence. For example, a received symbol deviating significantly from the constellation point along a direction with stronger noise propagation may still have a high probability; a small deviation along a direction with weaker noise propagation may already be a low-probability event. Ordinary Euclidean distance cannot reflect this difference. Based on this, in some examples, the calculation of the distance between the received symbol and each candidate constellation point includes: Based on the noise samples obtained within the preset statistical period, the variance of in-phase noise, the variance of orthogonal noise, and the correlation between in-phase noise and orthogonal noise are determined respectively, and a two-dimensional noise covariance matrix is ​​constructed based on the variance of in-phase noise, the variance of orthogonal noise, and the correlation. The principal axis direction of the noise distribution is determined based on the two-dimensional noise covariance matrix, and the selection range of candidate constellation points is extended along the principal axis direction with larger variance in the noise distribution. For each candidate constellation point, the corresponding statistical distance is calculated based on the two-dimensional coordinate difference between the received symbol and the candidate constellation point and the inverse of the two-dimensional noise covariance matrix. For each bit, the distance difference is determined based on the minimum statistical distances corresponding to candidate constellation points marked with zero and the minimum statistical distances corresponding to candidate constellation points marked with one.

[0063] For example, pilot symbols can be used to determine noise samples. After the receiver obtains the pilot received symbols, it subtracts the equalized pilot received symbols from the locally known pilot constellation points to obtain the in-phase noise samples. and orthogonal noise samples Within a preset statistical period, the mean values ​​of in-phase noise and quadrature noise are calculated separately. After eliminating the corresponding means, the variances of in-phase noise, quadrature noise, and the covariance between them are calculated. To avoid insufficient noise samples leading to an invertible or unstable covariance matrix, regularization can be applied to the covariance matrix. For example, a small positive number can be added to each of the two diagonal elements of the covariance matrix. This positive number can be determined based on the fixed-point calculation precision, sample size, or minimum allowable noise variance. Subsequently, eigenvalue decomposition is performed on the covariance matrix to obtain the first principal axis direction, the second principal axis direction, and their corresponding eigenvalues. The direction corresponding to the larger eigenvalue is determined as the principal axis direction with stronger noise diffusion. The candidate constellation point selection range no longer uses a fixed square or fixed circular range centered on the received symbol, but instead uses an extended range adapted to the noise principal axis direction. Along the direction of stronger noise diffusion, constellation points at greater distances can be selected; along the direction of weaker noise diffusion, a narrower candidate range is maintained. For example, when the major axis of the noise ellipse is tilted at 45 degrees relative to the in-phase coordinate axis, constellation points located in the lower left and upper right directions of the received symbol can be preferentially added, instead of adding an equal number of constellation points in the upper left and lower right directions. After the candidate point selection is completed, a two-dimensional coordinate difference vector is calculated for each candidate constellation point, and then the statistical distance is calculated using the inverse of the covariance matrix. For each bit carried by the received symbol, the candidate point with the smallest statistical distance is selected from the candidate points marked as zero for that bit, and the candidate point with the smallest statistical distance is selected from the candidate points marked as one for that bit, and the distance difference for that bit is determined based on the difference between the two smallest statistical distances. When the noise statistics change, the covariance matrix can be re-estimated. To prevent a sudden change in the principal axis direction due to a small number of outliers, smoothing can be performed between the current covariance matrix and the covariance matrix of the previous statistical period. Alternatively, the candidate point selection direction can be updated after the pilot noise statistics have changed for several consecutive periods.

[0064] In some cases, amplitude normalization is primarily used to correct scale differences between the received symbol and the standard constellation diagram, but it cannot eliminate residual phase errors. Residual phase errors cause constellation points to rotate around the center of the constellation diagram. For candidate constellation points... When the residual phase error is When rotated, its position can be represented as: The same phase error causes different coordinate offsets for constellation points with different radii. If the distance from a candidate constellation point to the center of the constellation diagram is... When the phase error is small, its arc length displacement is approximately the same as... The coordinate offset is proportional to the product of the absolute values ​​of the phase errors. Therefore, the coordinate offset experienced by the outer constellation points is usually greater than that experienced by the inner constellation points. If the same area boundary protection width is used for all constellation points, two problems can easily arise: when the protection width is set according to the inner constellation points, it may not be sufficient to cover the rotational displacement of the outer constellation points; when the protection width is set according to the outer constellation points, it will introduce too many irrelevant candidate points to the inner constellation points. Based on this, in some examples, the calculation of the distance between the received symbol and each candidate constellation point includes: The residual phase center estimate output by the carrier synchronization module and the estimation uncertainty corresponding to the residual phase center estimate are obtained, and the residual phase search interval corresponding to the current received symbol is determined based on the residual phase center estimate and the estimation uncertainty. For each candidate constellation point, the candidate constellation point is rotated within the residual phase search interval to determine the candidate phase that minimizes the distance between the rotated candidate constellation point and the received symbol. The extended distance corresponding to the candidate constellation point is determined based on the minimum distance and the degree to which the candidate phase deviates from the estimated value of the residual phase center. The corresponding regional boundary protection width is determined based on the modulus value of each candidate constellation point. The larger the modulus value of the candidate constellation point, the larger the corresponding regional boundary protection width. Candidate constellation points that may cross the regional boundary due to residual phase error are then added according to the regional boundary protection width.

[0065] Understandably, the residual phase center estimate and estimation uncertainty output by the carrier synchronization module can be used to determine the residual phase search interval for the current received symbol. The residual phase center estimate represents the most likely residual phase, and the estimation uncertainty represents the range within which the actual residual phase may deviate from this center value. For each candidate constellation point, different candidate phases are examined within the residual phase search interval, and the candidate phase that minimizes the distance between the rotated candidate constellation point and the received symbol is determined. However, if only the candidate phase with the smallest geometric distance is selected, the algorithm may be biased towards larger phase shifts at the edges of the search interval. Therefore, this scheme introduces a penalty term into the distance calculation to address the deviation of the candidate phase from the residual phase center estimate: in, This represents the extended distance corresponding to the k-th candidate constellation point; Indicates the received symbol; This represents the k-th candidate constellation point; Indicates the residual phase search interval; This represents the estimated value of the residual phase center; This represents the phase deviation penalty coefficient. The first term reflects the degree of geometric matching between the rotated candidate constellation point and the received symbol, while the second term reflects the reasonableness of using the candidate phase. The greater the deviation of the candidate phase from the center estimate, the larger the penalty. Furthermore, the boundary protection width of the region is determined based on the magnitude of the candidate constellation point. The larger the magnitude of the candidate constellation point, the greater the arc length displacement that residual phase error may cause; therefore, a larger boundary protection width is set accordingly.

[0066] For example, the carrier synchronization module obtains the residual phase center estimate by means of pilot phase, decision feedback phase, or phase changes of adjacent symbols. The estimation uncertainty is determined based on the statistical distribution of the phase estimation residuals. The residual phase search interval can be determined as follows: For each candidate constellation point, a discrete search method can be used to set multiple candidate phases within the interval. If the search interval is small, several equally spaced phase values ​​can be set; if higher accuracy is required, a coarser phase search can be performed first, followed by a finer local search near the optimal candidate phase. For each candidate phase, the candidate constellation point is rotated to the corresponding position, the squared distance between it and the received symbol is calculated, and the penalty value corresponding to the candidate phase's deviation from the center estimate is calculated. Combining the two, the candidate phase with the smallest extended distance is selected. The phase deviation penalty coefficient can be determined based on the residual phase estimation uncertainty. When the estimation uncertainty is small, it indicates that the center estimate given by the carrier synchronization module is relatively reliable, and a larger penalty coefficient can be set to limit the candidate phase from significantly deviating from the center estimate. When the estimation uncertainty is large, the penalty coefficient can be appropriately reduced to give the candidate constellation point a larger phase adjustment space. The area boundary protection width can be determined based on the candidate constellation point magnitude and the phase estimation uncertainty. For example, the in-phase protection width and the orthogonal protection width can be determined based on the maximum lateral coordinate change and the maximum longitudinal coordinate change that the candidate constellation point may produce within the residual phase search interval, respectively. For inner-layer candidate constellation points, their magnitudes are smaller, resulting in smaller coordinate displacements under the same phase uncertainty, allowing for a narrower protection width. For outer-layer candidate constellation points, their magnitudes are larger, necessitating a wider protection width. When it is determined, based on the protection width, that a candidate constellation point might cross a region boundary after rotation, relevant constellation points on the other side of that boundary are added to the candidate set. This addition does not include all constellation points on the other side of the boundary, but rather those related to the rotation trajectory of the current candidate point or the decision relationship of the current bit. After supplementing the candidate set, the expansion distance of each candidate constellation point is calculated, and the corresponding minimum expansion distance is determined according to the two markings for each bit, generating a distance difference.

[0067] In some cases, conventional soft demodulation typically adjusts the distances between candidate constellation points using a uniform scale based on the additive noise variance. This approach assumes that the uncertainty of the received symbols after equalization primarily originates from additive noise, and that this uncertainty does not change significantly with the positions of the candidate constellation points. However, there may be an error between the channel estimate used by the equalizer and the actual channel response. This channel estimation error is a multiplicative error, and its impact on the received symbols depends on the amplitude of the transmitted constellation points.

[0068] Let the actual channel response be The channel estimate is The channel estimation error is ,but: The received signal can be represented as: in, Indicates sending constellation points, This represents additive noise. After equalization using the channel estimate, the residual term formed by the channel estimation error and the transmitted constellation point... Related. The larger the magnitude of the constellation point, the larger the possible magnitude of the residual term is usually. Therefore, even if two candidate constellation points have the same geometric distance from the received symbol, it is not appropriate to directly assume that they have the same likelihood. The outer candidate constellation points are more affected by channel estimation errors, and their corresponding uncertainties should be greater than those of the inner candidate constellation points. Based on this, in some examples, the compensation for the distances corresponding to each candidate constellation point according to the channel noise includes: Obtain the additive noise variance and the channel estimation error variance, respectively; For each candidate constellation point, the equivalent uncertainty corresponding to the candidate constellation point is determined based on the magnitude of the candidate constellation point, the additive noise variance, and the channel estimation error variance. The equivalent uncertainty corresponding to the candidate constellation point increases as the magnitude of the candidate constellation point increases. The distance between the received symbol and the corresponding candidate constellation point is scaled by using the equivalent uncertainty corresponding to each candidate constellation point; For each bit, the minimum distance corresponding to the candidate constellation point with a bit mark of zero and the minimum distance corresponding to the candidate constellation point with a bit mark of one are determined from the scale-adjusted distances, and the soft information of the bit is generated based on the difference between the two minimum distances.

[0069] Understandably, the additive noise variance and the channel estimation error variance can be obtained separately, and the equivalent uncertainty can be determined for each candidate constellation point: in, This represents the equivalent uncertainty corresponding to the k-th candidate constellation point; Indicates the variance of additive noise; This represents the modulus of the k-th candidate constellation point; This represents the variance of the channel estimation error. This expression shows that the fundamental uncertainty caused by additive noise applies to all candidate constellation points, while the additional uncertainty caused by the channel estimation error increases with the square of the magnitude of the candidate constellation point. By adjusting the distance using the equivalent uncertainty corresponding to each candidate constellation point, soft information computation can simultaneously consider additive noise and multiplicative channel estimation error, instead of using the same noise scale for all candidate constellation points.

[0070] For example, the additive noise variance can be estimated from the pilot residuals. The channel estimation error variance can be obtained from the channel estimation differences between adjacent pilot positions, the channel interpolation residuals, the dispersion between repeated pilot estimation results, or the error covariance of the channel estimator output.

[0071] If the channel estimator can directly output the channel estimation confidence level, then this confidence level can be converted into the channel estimation error variance. When the channel estimation confidence level is low, a larger channel estimation error variance is determined; when the channel estimation confidence level is high, a smaller channel estimation error variance is determined.

[0072] For each candidate constellation point in the candidate set, read its in-phase coordinates and orthogonal coordinates, and calculate the square of the corresponding modulus: Subsequently, the squared magnitudes of the candidate constellation points, the variance of the additive noise, and the variance of the channel estimation error are substituted into the equivalent uncertainty expression. When adjusting the range scale, the squared distance between the received symbol and the candidate constellation point can be divided by the corresponding equivalent uncertainty: in, This represents the distance after scale adjustment. In implementations that further consider probability density normalization, a logarithmic term related to the equivalent uncertainty can be introduced into the distance metric to prevent outer candidate constellation points with large equivalent uncertainties from gaining an unreasonable advantage simply because of a large denominator. In this case, the scale-adjusted distance term and the uncertainty normalization term can be used together as the metric for candidate constellation points, based on a specific likelihood model. For each bit, the candidate point with the smallest adjusted distance is selected from the candidate constellation points marked as zero, and the candidate point with the smallest adjusted distance is selected from the candidate constellation points marked as one. Soft information for that bit is generated based on the difference between the two minimum distances. If the channel estimation errors differ between different subcarriers or different time slots, the corresponding channel estimation error variances are saved separately. During soft demodulation, the corresponding error variance is retrieved based on the subcarrier, time slot, or channel estimation interval to which the currently received symbol belongs.

[0073] In some cases, dividing the seven bits of a 128-order QAM into high-order bit groups and low-order bit groups, and setting different compensation weights for different groups, can address the issue of different decision boundary scales for different bits to some extent. However, this grouping method usually assumes that the decision geometry of the same bit is basically consistent across different constellation subregions.

[0074] For irregular rectangular 128-order QAM, constellation structures with some corner points removed, or constellation diagrams using special bit marking rules, the spacing between the opposite-marked constellation points corresponding to the same bit may differ in different spatial locations. For example, the spacing between the nearest opposite-marked constellation points of a certain low-order bit in the inner region may be two coordinate units, while in the outer region it may be four coordinate units. In this case, even if two received symbols are in similar positions relative to their respective decision boundaries, their minimum squared distance difference may have significantly different absolute values ​​due to different local constellation scales. If the same low-order bit compensation weight is uniformly used, regions with larger local decision spacing may produce soft information with larger absolute values, while regions with smaller local decision spacing may produce soft information with smaller absolute values. Therefore, the absolute magnitude of the distance difference cannot be directly used as a uniform confidence scale across all constellation sub-regions. Based on this, in some examples, the weighted correction of the corresponding distance difference based on the bit group to which each bit belongs includes: For each bit in the target constellation sub-region, determine the nearest hetero-labeled constellation point pair with different bit labels and located on both sides of the corresponding decision boundary, and determine the local decision interval of the bit in the target constellation sub-region based on the distance between the nearest hetero-labeled constellation point pairs; The local decision margin of the bit in the target constellation sub-region is determined based on the ratio between the minimum squared difference of the distance corresponding to the bit and the square of the local decision interval. The compensation weight corresponding to the target constellation sub-region and the bit is determined based on the local decision margin, and the distance difference corresponding to the bit is weighted and corrected using the compensation weight, so that the same bit uses different compensation weights in constellation sub-regions with different local decision intervals.

[0075] For example, during the receiver initialization phase, local decision interval index information can be pre-generated based on the 128th-order QAM constellation coordinates and bit marking rules. For each constellation sub-region and each bit, constellation point pairs with different bit markings and forming the closest competing relationship on both sides of the corresponding decision boundary are found. The Euclidean distance between these constellation point pairs is calculated and used as the local decision interval for the corresponding region and the corresponding bit. For example, for the sixth bit, the local decision intervals corresponding to the inner region, middle region, and outer region can be obtained respectively. Even if these regions belong to the same low-order bit group, their local decision intervals can be different. When processing the current received symbol, the target constellation sub-region is first determined, and then the corresponding local decision interval is read from the local decision interval index information based on the target constellation sub-region identifier and the current bit identifier. Subsequently, the minimum distances corresponding to the candidate constellation points with the current bit marking as zero and the current bit marking as one are determined respectively, and the minimum distance squared difference is calculated. The minimum distance squared difference is divided by the square of the local decision interval to obtain the local decision margin. To avoid numerical instability caused by abnormally small local decision intervals, a minimum allowable value can be set for the local decision interval, or a stabilizing amount appropriate to the fixed-point accuracy can be added to the denominator. Compensation weights can be determined through a preset mapping relationship. For example, when the local decision interval is large and the original distance difference is also large, but the local decision margin is only within a medium range, a smaller compensation weight is used to avoid generating excessively strong soft information due to the large overall geometric scale between constellation points. When the local decision interval is small, and the received symbol has shown a significant bias relative to the decision boundary, a compensation weight that can preserve this bias can be used to avoid the compression of reliable information due to excessively small original distance differences. In another implementation, a compensation weight table can be pre-established for each constellation sub-region and each bit. The input to the weight table includes the target constellation sub-region, bit identifier, and local decision margin interval, and the output is the corresponding compensation weight. This reduces real-time division and complex mapping operations, making it more suitable for fixed-point hardware implementation. When the received symbol is near the boundary of two constellation sub-regions, the local decision intervals corresponding to the two regions can be obtained respectively, and the transition local decision interval can be determined according to the relative position of the received symbol to the center or boundary of the two regions, so as to reduce the sudden change in compensation weight caused by region handover.

[0076] Please see Figure 2 One embodiment of the high-order QAM soft information acquisition device in this application includes: Signal acquisition unit 21 is used to acquire high-order QAM baseband complex symbol signals after carrier synchronization, symbol synchronization and channel equalization; The candidate constellation point determination unit 22 is used to determine the target constellation sub-region corresponding to the high-order QAM baseband complex symbol signal based on the in-phase component coordinates and orthogonal component coordinates of the high-order QAM baseband complex symbol signal and the preset high-order QAM constellation mapping rules, so as to obtain candidate constellation points configured for the target constellation sub-region. The distance difference calculation unit 23 is used to determine, for each bit carried by the high-order QAM baseband complex symbol signal, a first candidate constellation point marked with a first bit value and a second candidate constellation point marked with a second bit value from the candidate constellation points, so as to calculate the distance difference of the corresponding bit based on the minimum Euclidean distance. The compensation weight determination unit 24 is used to divide each bit into at least two bit groups according to the decision boundary geometric characteristics of each bit in the higher-order QAM constellation mapping rule, and determine the compensation weight corresponding to each bit group respectively. The soft information output unit 25 is used to apply the compensation weights corresponding to each bit group to perform weighted correction on the initial log-likelihood ratio of each bit belonging to the corresponding bit group, and output the soft information sequence corresponding to the high-order QAM baseband complex symbol signal.

[0077] like Figure 3 As shown, this application embodiment also provides an electronic device 300, including a memory 310, an FPGA hardware architecture module 320, and an FPGA configuration bitstream 311 stored in the memory and used to configure the FPGA hardware architecture module 320. After the FPGA hardware architecture module 320 is configured by the FPGA configuration bitstream 311, it implements any of the steps of the above-mentioned high-order QAM carrier synchronization method.

[0078] Since the electronic device described in this embodiment is the device used to implement a high-order QAM soft information acquisition device in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in the embodiments of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiments of this application is within the scope of protection of this application.

[0079] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

Claims

1. A high-order QAM soft information acquisition method, characterized in that, include: Acquire the high-order QAM baseband complex symbol signal after carrier synchronization, symbol synchronization and channel equalization; Based on the in-phase component coordinates and quadrature component coordinates of the high-order QAM baseband complex symbol signal, and the preset high-order QAM constellation mapping rules, the target constellation sub-region corresponding to the high-order QAM baseband complex symbol signal is determined, so as to obtain candidate constellation points configured for the target constellation sub-region. For each bit carried by the high-order QAM baseband complex symbol signal, a first candidate constellation point with a first bit value and a second candidate constellation point with a second bit value are determined from the candidate constellation points, so as to calculate the distance difference of the corresponding bit based on the minimum Euclidean distance; Based on the decision boundary geometric characteristics of each bit in the higher-order QAM constellation mapping rule, each bit is divided into at least two bit groups, and the compensation weight corresponding to each bit group is determined respectively. By employing the compensation weights corresponding to each bit group, the initial log-likelihood ratio of each bit belonging to the corresponding bit group is weighted and corrected, and the soft information sequence corresponding to the high-order QAM baseband complex symbol signal is output.

2. The method as described in claim 1, characterized in that, The distance difference between corresponding bits is calculated based on the minimum Euclidean distance, including: Calculate the first minimum Euclidean distance between the high-order QAM baseband complex symbol signal and the first candidate constellation point, and the second minimum Euclidean distance between the high-order QAM baseband complex symbol signal and the second candidate constellation point, so as to determine the distance difference of the corresponding bit based on the square difference between the first minimum Euclidean distance and the second minimum Euclidean distance.

3. The method as described in claim 2, characterized in that, Also includes: Estimate the signal-to-noise ratio (SNR) of the current channel, determine the noise variance compensation factor based on the SNR, and obtain the initial log-likelihood ratio of each bit based on the noise variance compensation factor and the distance difference corresponding to each bit.

4. The method as described in claim 1, characterized in that, Also includes: After acquiring the high-order QAM baseband complex symbol signal after carrier synchronization, symbol synchronization and channel equalization, the amplitude of the high-order QAM baseband complex symbol signal is normalized so that the magnitude of the normalized high-order QAM baseband complex symbol signal matches the magnitude of the standard rectangular constellation diagram stored locally. Determining the target constellation sub-region corresponding to the higher-order QAM baseband complex symbol signal includes: The target constellation sub-region is determined based on the in-phase and quadrature component coordinates of the normalized high-order QAM baseband complex symbol signal.

5. The method as described in claim 4, characterized in that, The higher-order QAM is a 128th-order quadrature amplitude modulation, and each higher-order QAM baseband complex symbol signal carries seven bits. Dividing these bits into at least two bit groups includes: The three bits used to determine the region to which the constellation point belongs are divided into the high-order bit group, and the remaining four bits used to distinguish the sub-points within the region are divided into the low-order bit group.

6. The method as described in claim 5, characterized in that, The step of determining the target constellation sub-region corresponding to the high-order QAM baseband complex symbol signal based on the in-phase component coordinates and quadrature component coordinates of the high-order QAM baseband complex symbol signal, and a preset high-order QAM constellation mapping rule, includes: Based on the positive and negative attributes of the in-phase component and the quadrature component, the 128th order quadrature amplitude modulation constellation diagram is divided into four primary regions, and each primary region is further divided into multiple amplitude regions based on the amplitude of the in-phase component and the quadrature component. According to the decision boundary corresponding to the low-order bit group, each amplitude region is divided into multiple constellation sub-regions; Based on the in-phase component coordinates, quadrature component coordinates, and amplitude of the high-order QAM baseband complex symbol signal, the target constellation sub-region into which the high-order QAM baseband complex symbol signal falls is determined.

7. The method as described in claim 3, characterized in that, Each target constellation sub-region is configured with eight to sixteen candidate constellation points, and obtaining the initial log-likelihood ratio for each bit includes: For each bit, calculate the Euclidean distance between the high-order QAM baseband complex symbol signal and the corresponding eight to sixteen candidate constellation points, and determine the first minimum Euclidean distance and the second minimum Euclidean distance from them; Calculate the difference between the square of the first minimum Euclidean distance and the square of the second minimum Euclidean distance to obtain the distance difference amount; The product of the distance difference and the noise variance compensation factor is determined as the initial log-likelihood ratio of the corresponding bit, so as to avoid traversing all constellation points in the high-order QAM constellation diagram.

8. A high-order QAM soft information acquisition device, characterized in that, include: The signal acquisition unit is used to acquire the high-order QAM baseband complex symbol signal after carrier synchronization, symbol synchronization and channel equalization; The candidate constellation point determination unit is used to determine the target constellation sub-region corresponding to the high-order QAM baseband complex symbol signal based on the in-phase component coordinates and quadrature component coordinates of the high-order QAM baseband complex symbol signal and the preset high-order QAM constellation mapping rules, so as to obtain candidate constellation points configured for the target constellation sub-region. The distance difference calculation unit is used to determine, for each bit carried by the high-order QAM baseband complex symbol signal, a first candidate constellation point marked with a first bit value and a second candidate constellation point marked with a second bit value from the candidate constellation points, so as to calculate the distance difference of the corresponding bit based on the minimum Euclidean distance. The compensation weight determination unit is used to divide each bit into at least two bit groups according to the decision boundary geometric characteristics of each bit in the high-order QAM constellation mapping rule, and determine the compensation weight corresponding to each bit group respectively. The soft information output unit is used to apply the compensation weights corresponding to each bit group to perform weighted correction on the initial log-likelihood ratio of each bit belonging to the corresponding bit group, and output the soft information sequence corresponding to the high-order QAM baseband complex symbol signal.

9. The soft information acquisition device as described in claim 8, characterized in that, The distance difference calculation unit specifically includes: Calculate the first minimum Euclidean distance between the high-order QAM baseband complex symbol signal and the first candidate constellation point, and the second minimum Euclidean distance between the high-order QAM baseband complex symbol signal and the second candidate constellation point, so as to determine the distance difference of the corresponding bit based on the square difference between the first minimum Euclidean distance and the second minimum Euclidean distance.

10. The soft information acquisition device as described in claim 8, characterized in that, It also includes an initial log-likelihood ratio acquisition unit, used to estimate the signal-to-noise ratio of the current channel, determine the noise variance compensation factor based on the signal-to-noise ratio, and obtain the initial log-likelihood ratio of each bit based on the noise variance compensation factor and the distance difference corresponding to each bit.

Images