A low-precision phase shifter codebook design method suitable for near-field MIMO
By designing the CS-SCHT codebook and virtual subarray structure, the problems of high hardware cost, high power consumption and insufficient robustness caused by high-precision phase shifters in near-field MIMO systems are solved, achieving efficient beam coverage and channel adaptability under low precision.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UESTC (SHENZHEN) ADVANCED RES INST
- Filing Date
- 2026-02-11
- Publication Date
- 2026-06-02
AI Technical Summary
In near-field MIMO systems, the traditional requirement for high-precision phase shifters leads to high hardware costs, high power consumption, and high computational complexity, and also results in insufficient robustness in non-stationary channels in the spatial domain.
A low-precision phase shifter codebook design method based on conjugate symmetric sequence complex Hadamard transform (CS-SCHT) is adopted. Combined with virtual subarray structure, the near-field spherical wave problem is transformed into plane wave selection in the angle domain at the subarray scale. A 2-bit phase shifter is used to realize analog domain beamforming, and the beam coverage density is improved by overcomplete codebook expansion.
It reduces the accuracy requirements of the phase shifter, lowers hardware costs and power consumption, improves beam coverage and robustness, adapts to non-stationary channels in the spatial domain, and is easy to implement in engineering.
Smart Images
Figure CN122133471A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of wireless communication and array signal processing technology, specifically relating to a low-precision phase shifter codebook design method suitable for near-field MIMO. Background Technology
[0002] With the development of sixth-generation (6G) wireless communication systems, high-frequency communications such as millimeter waves and terahertz waves are gradually being introduced into very large-scale antenna arrays (VMAs). When the antenna array aperture increases significantly, the communication system will be in the near-field region within the conventional coverage area, and the electromagnetic wave propagation will change from a plane wave model to a spherical wave model. In near-field MIMO systems, the array response depends on both angle and distance, making traditional far-field codebooks based on Discrete Fourier Transform (DFT) unsuitable. Existing near-field precoding schemes typically employ high-resolution codebooks in the angle-range domain. These codebooks are not only huge in size and computationally complex, but also place high demands on the phase resolution of the analog domain phase shifters in hybrid analog-digital precoding architectures, typically requiring 4-bit or even higher precision phase shifters, leading to a significant increase in system hardware cost, power consumption, and implementation complexity. Therefore, how to reduce the precision requirements of phase shifters while ensuring near-field precoding performance has become a key issue restricting the engineering implementation of near-field MIMO systems. Summary of the Invention
[0003] This invention aims to address the problems in near-field MIMO hybrid precoding, where the introduction of high-dimensional codebooks and continuous phase control to accommodate the angle-range coupled spherical wave array response leads to a strong dependence of analog domain beamforming on high-precision phase shifters. This results in excessively high hardware costs and power consumption, high implementation complexity, and insufficient robustness in spatially non-stationary channels. The invention provides a low-precision phase shifter codebook design method based on conjugate symmetric sequence complex Hadamard transform (CS-SCHT), ensuring that analog domain codeword elements are taken only from... This enables support for 2-bit phase shifters; and further, an ultra-complete CS-SCHT codebook is constructed to improve beam coverage density; at the same time, combined with a virtual subarray structure, the near-field spherical wave problem is transformed into angle-domain plane wave beam selection at the subarray scale, achieving precoding performance close to that of a high-precision codebook under low-precision hardware conditions.
[0004] The technical solution adopted in this invention is:
[0005] A low-precision phase shifter codebook design method suitable for near-field MIMO includes the following steps:
[0006] S1. Determine the codebook dimension N, satisfying... n is a positive integer;
[0007] S2. Construct an N-dimensional Walsh-Hadamard matrix. Its elements are {+1,-1}, and it satisfies recursion: , ;
[0008] S3, in Based on this, complex rotation and conjugate symmetry constraints are introduced to generate the intermediate matrix of the conjugate symmetric complex Hadamard transformation. ,make Elements only ;
[0009] S4, to The codewords are sorted in sequence so that they are arranged in ascending order of equivalent spatial frequency, resulting in the sorted codebook. ;
[0010] S5, will Normalize the codewords by column or row to obtain the CS-SCHT codebook. ,in .
[0011] Furthermore, S3 generation The specific method is: through block-diagonal phase matrix , right Apply different blocks Rotation, in and To construct phase rotation matrices using the same block partitioning method, the phase rotation factors at corresponding positions appear in pairs to ensure that the final constructed CS-SCHT matrix satisfies a conjugate symmetric structure. Then, recursive block construction is used to obtain... Maintain column and row orthogonality and constant model properties.
[0012] Furthermore, in S4... The specific method for sorting the sequence is as follows: construct a permutation matrix by reversing the bits. ,make ,in Swap the row or column order according to the index binary bit reversal rule, thereby... The code words are sorted from low to high according to their equivalent spatial frequency.
[0013] Furthermore, codebook Each codeword element corresponds to a 2-bit phase shifter control signal, with phase values of 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 10 ... , and .
[0014] Furthermore, it also includes the following steps:
[0015] S6. Set the over-completeness factor to 𝛼, 𝛼≥2;
[0016] S7. While keeping the codeword elements derived from the set Under the premise of changing the initial Hadamard kernel or complex rotation structure, multiple sets of conjugate symmetric sequence complex Hadamard transformation matrices are generated: ;
[0017] S8, satisfying multiple conditions The element-constrained and sorted CS-SCHT matrices are concatenated column-wise to form an overcomplete codebook. The corresponding number of supercomplete codebook columns is 𝐾=𝛼𝑁;
[0018] S9. Perform beam uniformity and correlation screening on the supercomplete codebook: calculate the pairwise correlation coefficients of codewords, remove highly correlated codewords, and constrain the coverage of the equivalent angular domain to make the final codebook cover the target angular domain as uniformly as possible.
[0019] Furthermore, it also includes the following steps:
[0020] S10, Transmitter The root antenna is divided into Each virtual subarray contains [number] virtual subarrays. The antenna uses a plane wave approximation at the subarray scale, while preserving the near-field spherical wave phase difference between subarrays;
[0021] S11. Configure an independent set of analog beam candidates for each virtual subarray, the candidate set being generated by a codebook. or Obtained by truncation / mapping, the method is as follows ,in ;
[0022] S12. In each beam selection iteration In this process, based on the current residual matrix or equivalent channel projection, for each subarray... Independently select the largest projected codeword as the subarray analog wave: .in, Indicates the first The first virtual subarray corresponding to the _th One candidate analog beam codeword vector; Indicates codeword index; This indicates that the current residual channel matrix or residual projection matrix is in the subarray The corresponding row block. The above formula selects the codeword with the largest projected energy as the optimal simulated beam for the subarray by calculating the projection amplitude between each candidate codeword and the residual signal.
[0023] S13. Combine the selected codewords from each subarray into a full-array analog precoding vector using a block diagonal / concatenation method: ;
[0024] S14. Repeat steps S12-S13 to select M RF chain beams to form an analog precoding matrix. Subsequently, the digital precoding is solved at the baseband end. This satisfies the transmit power constraint;
[0025] S15, Output near-field hybrid pre-encoder .
[0026] Furthermore, invisible virtual subarrays are either masked or assigned fixed reference codewords.
[0027] Furthermore, the codebook or It is generated and stored offline, and the phase shifter control signal is output by looking up a table when the system is running.
[0028] The beneficial effects of this invention are as follows:
[0029] (1) Low hardware cost and low power consumption: only codeword elements are taken This can be achieved with a 2-bit phase shifter, significantly reducing the accuracy requirements of the phase shifter and the power consumption of the analog front end;
[0030] (2) Low complexity: The CS-SCHT codebook can be generated quickly through a recursive structure, and codeword multiplication can be implemented by sign flipping and real-virtual swapping, reducing online computation and control overhead;
[0031] (3) Strong beam coverage capability: The number of beams and the sampling density of the angular domain are increased by the super-complete extension, and the main lobe pointing capability can still be maintained even at low precision.
[0032] (4) Adapting to near-field and non-stationary channels: Combining the virtual subarray structure, the near-field spherical wave precoding is transformed into subarray angle domain selection, avoiding the dependence of the whole array angle-distance codebook on the whole array consistent structure, thus making it more robust under the condition of spatial non-stationarity (different visible areas);
[0033] (5) Easy to implement in engineering: The codebook is naturally matched with hardware control and can be seamlessly integrated with existing hybrid precoding frameworks, sparse recovery / greedy search / learning methods, making it easy to integrate into base stations or array terminals. Attached Figure Description
[0034] Figure 1 This is a schematic diagram of low-precision hybrid precoding for near-field MIMO based on the CS-SCHT codebook and virtual subarray structure. Detailed Implementation
[0035] The method of the present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0036] The method of this invention proposes a codebook construction, and further proposes an overcomplete extended codebook based on the codebook construction, as well as a near-field precoding method combined with a virtual subarray structure, specifically including:
[0037] (1) CS-SCHT basecode construction method
[0038] Step S1: Determine the codebook dimension N, satisfying (n is a positive integer), let the codebook matrix .
[0039] Step S2: Construct the N-dimensional Walsh-Hadamard matrix The Walsh-Hadamard Transform (WHT) has elements in the range {+1, -1} and is recursive. , .
[0040] Step S3: In Based on this, complex rotation and conjugate symmetry constraints are introduced to generate the intermediate matrix of the conjugate symmetric complex Hadamard transformation. ,make Elements only One implementation method is to use a block-diagonal phase matrix. , right Apply different blocks Rotate and use recursive block construction to obtain Maintain column / row orthogonality (or near-orthogonality) and constant model properties.
[0041] Step S4: For Sequencing is performed to obtain the increasing "frequency" property of a DFT-like sequence. Specifically, bit-reversal permutation can be used: constructing a permutation matrix. ,make ,in The rows (or columns) are swapped according to the index binary bit reversal rule, thereby... The code words are sorted from low to high according to their equivalent spatial frequency.
[0042] Step S5: Normalize by column (or row) to obtain a constant modulus codeword set, thus obtaining the CS-SCHT basecode. ,in .
[0043] (2) A method for constructing an overcomplete CS-SCHT codebook
[0044] Step S6: For codebooks with the same number of beams and antennas To address the issue of insufficient beam coverage density, an ultra-complete codebook is constructed. Set the number of supercomplete codebook columns. , satisfy: .
[0045] Step S7: Keeping the 2-bit value constraint unchanged, multiple sets of CS-SCHT matrices are generated by changing the initial WHT kernel or the superposition structure.
[0046] Step S8: Combine multiple satisfying The element-constrained and sorted CS-SCHT matrices are concatenated column-wise to form an overcomplete codebook. .
[0047] Step S9: Perform beam uniformity and correlation screening on the supercomplete codebook: calculate the pairwise correlation coefficients of codewords, remove highly correlated codewords, and constrain the coverage of the equivalent angular domain to make the final codebook cover the target angular domain as uniformly as possible.
[0048] (3) Near-field hybrid precoding method combined with virtual subarray structure
[0049] Step S10: Connect the transmitter The root antenna is divided into Each virtual subarray contains [number] virtual subarrays. A single antenna. A plane wave approximation is used at the subarray scale, while the near-field spherical wave phase difference is preserved between subarrays.
[0050] Step S11: Configure an independent set of simulated beam candidates for each virtual subarray, the candidate set being truncated / mapped from the CS-SCHT codebook (basic codebook or super-complete codebook): ,in .
[0051] Step S12: In each beam selection iteration In this process, based on the current residual matrix or equivalent channel projection, for each subarray... Independent selection of maximum projection codeword: ,in Indicates the first The first virtual subarray corresponding to the _th candidate analog beam codeword vectors, This indicates that the current residual channel matrix or residual projection matrix is in the subarray The corresponding line block.
[0052] Step S13: Combine the selected codewords from each subarray into a full-array analog precoding vector using a block diagonal / concatenation method: .
[0053] Step S14: Repeat steps S12-S13 to select M RF chain beams and form an analog precoding matrix. Subsequently, the digital precoding is solved at the baseband end. (e.g., least squares / orthogonalization / water filling, etc.) to satisfy the transmit power constraint.
[0054] Step S15: Output near-field hybrid pre-encoder Furthermore, it can perform subarray-level activation / masking of the visible region (VR) based on spatial nonstationarity to improve robustness and energy efficiency.
[0055] Example 1:
[0056] In this embodiment, a subarray-level simulated beamforming scenario for a near-field MIMO system is considered. This embodiment focuses on the construction of the CS-SCHT codebook and the mapping of low-precision phase shifters.
[0057] First, set the number of subarray antennas to be... ,in , For example, positive integers. or Select codebook dimension .
[0058] According to the steps described in the technical solution, first construct Vivorsh-Hadamard transformation matrix Its elements are taken from the set And satisfy the recursive construction relation. Subsequently, in the... Based on this, by introducing complex rotation factors to different sub-blocks and applying conjugate symmetry constraints, a conjugate symmetric sequence complex Hadamard transformation matrix is constructed. All matrix elements are taken from the set .
[0059] Furthermore, regarding the aforementioned The rows (or columns) are sorted sequentially, preferably using a bit-reversed arrangement, so that the resulting codewords exhibit monotonical changes in the equivalent spatial frequency dimension, thereby obtaining beam sorting characteristics similar to Discrete Fourier Transform. After normalization, a CS-SCHT base codebook is formed, where each codeword is a constant modulus vector.
[0060] In the hardware implementation, each codeword element can be directly mapped to a 2-bit phase shifter control signal. For example, phase 0 corresponds to the value 1, phase λ / 2 corresponds to the value λ, phase λ corresponds to the value -1, and phase 3λ / 2 corresponds to the value -λ.
[0061] Therefore, the CS-SCHT codebook can perform analog beamforming using only a 2-bit phase shifter.
[0062] Example 2:
[0063] In this embodiment, in order to improve beam coverage density and beam pointing capability under low precision conditions, an ultra-complete codebook is constructed based on the CS-SCHT codebook.
[0064] Specifically, the overcompleteness multiple is set to , where ≥ 2, and the corresponding number of overcomplete codebook columns is = . This is while maintaining the fact that codeword elements are still taken from... Under the premise of changing the initial Hadamard kernel or complex rotation structure, multiple different CS-SCHT matrices can be generated.
[0065] For example, different initial kernel matrices can be used when constructing the Walsh-Hadamard matrix, or different combinations of rotation factors can be applied to different sub-blocks during the complex rotation stage, thereby obtaining multiple sets of distinguishable CS-SCHT matrices. Subsequently, these multiple sets of CS-SCHT matrices are concatenated column-wise to form an overcomplete CS-SCHT codebook.
[0066] To avoid excessive correlation between codewords, correlation detection is preferably performed on the codewords in the overcomplete codebook. For codeword pairs with correlation coefficients exceeding a preset threshold, one of them is retained or replaced to ensure that the final codebook achieves relatively uniform beam coverage in the target angular domain.
[0067] Example 3:
[0068] In this embodiment, a near-field hybrid precoding method combining a virtual subarray structure is proposed, considering an ultra-large-scale near-field MIMO downlink system. The transmitter is equipped with... One antenna and 𝑀 RF links, of which :
[0069] First, the transmitting antenna array is divided into n virtual subarrays, each containing... A single antenna. Because the physical aperture of each subarray is small, a plane wave model can be approximated at the subarray scale, while the spherical wave phase difference caused by near-field effects is still preserved between subarrays.
[0070] Configure a CS-SCHT codebook (basic codebook or supercomplete codebook) for each virtual subarray, such as Figure 1As shown, the CS-SCHT codebook, after overcomplete expansion, is configured in each virtual subarray. Beam selection is performed at the subarray scale, and the selected codewords from each subarray are concatenated to form a full-array analog precoding vector, thereby completing the hybrid precoding of near-field MIMO. An iterative beam selection strategy is adopted in the analog precoding design process. Specifically, in each beam selection iteration, for each virtual subarray, the projection intensity between its codeword and the current residual channel or residual matrix is calculated, and the codeword with the largest projection amplitude is selected as the optimal beam for that subarray in this iteration.
[0071] Subsequently, the codewords selected from each virtual subarray are concatenated in a predetermined order to form the analog precoding vector of the entire array, and then normalized. This process is repeated until 𝑀 analog beams are selected, thus forming the analog precoding matrix.
[0072] After constructing the analog precoding matrix, the digital precoding matrix is calculated at the baseband end, for example, by using the least squares method, orthogonalization method or water-filled power allocation method, to meet the system's transmit power constraints and maximize the system's spectral efficiency.
[0073] Example 4:
[0074] In this embodiment, the case of a non-stationary channel in the spatial domain is considered in the near-field MIMO system, that is, different propagation paths have different visible areas on the array.
[0075] In traditional precoding schemes based on full-array angle-range domain codebooks, invalid phase is easily applied in invisible regions because each beam covers the entire array, leading to performance degradation. However, in the method described in this invention, simulated beam selection is performed at the virtual subarray scale.
[0076] Specifically, when a propagation path is invisible on a portion of the virtual subarray, the corresponding subarray can be directly shielded during the beam selection phase, or a fixed reference codeword can be assigned to it, thereby avoiding the application of invalid phase control to the invisible region. This approach effectively improves system robustness and energy efficiency under non-stationary channel conditions in the spatial domain.
[0077] Example 5:
[0078] In this embodiment, the CS-SCHT codebook is generated offline and stored in the memory of the baseband processing unit or controller. During system operation, a 2-bit phase shifter control signal is generated by looking up a table based on the codeword index output by the beam selection algorithm.
[0079] In traditional precoding schemes based on full-array angle-range domain codebooks, invalid phase is easily applied in invisible regions because each beam covers the entire array, leading to performance degradation. However, in the method described in this invention, simulated beam selection is performed at the virtual subarray scale.
[0080] Specifically, when a propagation path is invisible on a portion of the virtual subarray, the corresponding subarray can be directly shielded during the beam selection phase, or a fixed reference codeword can be assigned to it, thereby avoiding the application of invalid phase control to the invisible region. This approach effectively improves system robustness and energy efficiency under non-stationary channel conditions in the spatial domain.
Claims
1. A low-precision phase shifter codebook design method suitable for near-field MIMO, characterized in that, Includes the following steps: S1. Determine the codebook dimension N, satisfying... n is a positive integer; S2. Construct an N-dimensional Walsh-Hadamard matrix. Its elements are {+1,-1}, and it satisfies recursion: , ; S3, in Based on this, complex rotation and conjugate symmetry constraints are introduced to generate the intermediate matrix of the conjugate symmetric complex Hadamard transformation. ,make Elements only ; S4, to The codewords are sorted in sequence so that they are arranged in ascending order of equivalent spatial frequency, resulting in the sorted codebook. ; S5, will Normalize the codewords by column or row to obtain the CS-SCHT codebook. ,in .
2. The low-precision phase shifter codebook design method suitable for near-field MIMO according to claim 1, characterized in that, S3 generation The specific method is: through block-diagonal phase matrix , right Apply different blocks Rotate and use recursive block construction to obtain Maintaining column / row orthogonality and constant model property, where and This is a phase rotation matrix constructed using the same block division method.
3. The low-precision phase shifter codebook design method suitable for near-field MIMO according to claim 1, characterized in that, S4 The specific method for sorting the sequence is as follows: construct a permutation matrix by reversing the bits. ,make ,in Swap the row or column order according to the index binary bit reversal rule, thereby... The code words are sorted from low to high according to their equivalent spatial frequency.
4. The low-precision phase shifter codebook design method suitable for near-field MIMO according to claim 1, characterized in that, codebook Each codeword element corresponds to a 2-bit phase shifter control signal, with phase values of 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 10 ... , and .
5. A low-precision phase shifter codebook design method suitable for near-field MIMO according to claim 1, characterized in that, It also includes the following steps: S6. Set the over-completeness factor to 𝛼, 𝛼≥2; S7. While keeping the codeword elements derived from the set Under the premise of changing the initial Hadamard kernel or complex rotation structure, multiple sets of conjugate symmetric sequence complex Hadamard transformation matrices are generated: ; S8, satisfying multiple conditions The element-constrained and sorted CS-SCHT matrices are concatenated column-wise to form an overcomplete codebook. The corresponding number of supercomplete codebook columns is 𝐾=𝛼𝑁; S9. Perform beam uniformity and correlation screening on the supercomplete codebook: calculate the pairwise correlation coefficients of codewords, remove highly correlated codewords, and constrain the coverage of the equivalent angular domain to make the final codebook cover the target angular domain as uniformly as possible.
6. A low-precision phase shifter codebook design method suitable for near-field MIMO according to claim 5, characterized in that, It also includes the following steps: S10, Transmitter The root antenna is divided into Each virtual subarray contains [number] virtual subarrays. The antenna uses a plane wave approximation at the subarray scale, while preserving the near-field spherical wave phase difference between subarrays; S11. Configure an independent set of analog beam candidates for each virtual subarray, the candidate set being generated by a codebook. or Obtained by truncation / mapping, the method is as follows ,in ; S12. In each beam selection iteration In this process, based on the current residual matrix or equivalent channel projection, for each subarray... Independently select the largest projected codeword as the subarray analog wave: ,in Indicates the first The first virtual subarray corresponding to the _th candidate analog beam codeword vectors, This indicates that the current residual channel matrix or residual projection matrix is in the subarray The corresponding line block; S13. Combine the selected codewords from each subarray into a full-array analog precoding vector using a block diagonal / concatenation method: ; S14. Repeat steps S12-S13 to select M RF chain beams to form an analog precoding matrix. ; Digital precoding is then solved at the baseband end. This satisfies the transmit power constraint; S15, Output near-field hybrid pre-encoder .
7. A low-precision phase shifter codebook design method suitable for near-field MIMO according to claim 6, characterized in that, Invisible virtual subarrays can be masked or assigned fixed reference codewords.
8. A low-precision phase shifter codebook design method suitable for near-field MIMO according to claim 6, characterized in that, The codebook or It is generated and stored offline, and the phase shifter control signal is output by looking up a table when the system is running.