High-speed multi-bit error correction and detection device and method based on Reed-Solomon code for memory

By introducing a column interleaving mapping module and combining it with low-rate RS codes in the memory, the timing bottleneck problem of multi-bit flipping in Reed-Solomon codes is solved, achieving high-speed error correction while reducing encoding and decoding latency and improving the data throughput performance of the memory.

CN123114847APending Publication Date: 2026-11-17BEIJING MICROELECTRONICS TECH INST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611206098.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing Reed-Solomon codes suffer from severe timing bottlenecks when dealing with multi-bit flips, and the encoding and decoding process consumes a large number of clock cycles, failing to meet the low access latency requirements of high-speed memory.

Method used

A column interleaving mapping module is introduced to interleave the physical bit data of the storage array through a multiplexer layer and convert it into logical codewords. Low-rate RS codes are used for error correction in the logical domain to reduce the number of redundant bytes and reduce the encoding and decoding clock cycles.

Benefits of technology

By transforming the physical to logical domains, the encoding and decoding clock cycles are significantly shortened while maintaining the error correction capability, thereby improving the data throughput performance of the memory.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114847A_ABST
    Figure CN123114847A_ABST
Patent Text Reader

Abstract

This invention relates to a high-speed multi-bit error correction and detection device and method for memory based on Reed-Solomon codes, belonging to the field of memory data reliability technology. This invention combines column interleaving technology with a hybrid error correction architecture of low-rate RS codes. By inserting a multiplexer layer between the storage array and the RS encoding / decoding module, and remapping the physical bit lines with a set interleaving distance ID, consecutive multi-bit errors in the physical space are discretized into independent single-bit errors in the logical codeword space. This invention utilizes column interleaving to effectively expand the burst error capture range of low-redundancy RS codes, allowing for the selection of RS code types with a smaller total byte count (n) while maintaining the same consecutive bit error correction capability. For consecutive 4-bit flip errors, this invention can significantly reduce the total error correction and detection delay from 52 clock cycles in traditional schemes to 22 clock cycles, significantly improving the random access performance of high-speed memory.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of memory data reliability technology, and relates to a high-speed multi-bit error correction and detection device and method for memory based on Reed-Solomon codes. Background Technology

[0002] With the stringent requirements for data integrity in aerospace electronic systems, high-energy physics experimental facilities, and nuclear industry control equipment, the hardened design of memory to resist single-event effects has become a research hotspot. When high-energy protons, neutrons, or heavy ions bombard sensitive nodes of a memory, the charge generated by the ionization trajectory is collected by the memory cell. If the collected charge exceeds the critical charge, it will cause the stored logic state to flip, i.e., a single-event flip. As semiconductor process technology enters the nanoscale, the spacing between memory cells shrinks. The incidence of a single high-energy particle often induces flips in multiple adjacent cells around the track. This phenomenon is called multi-bit flip, or MBU. MBU has become one of the main failure modes of high-reliability advanced process memories in spaceborne and ground-based applications.

[0003] To address the MBU threat, the industry widely adopts Reed-Solomon error correction codes. RS codes are cyclic codes based on finite field operations, exhibiting excellent error correction performance against burst errors. For an encoding scheme with parameters RS(n, k, t), k represents the number of original information bytes, t represents the number of redundant check bytes, and the total number of bytes n = k + t. This encoding has the ability to correct errors of t / 2 consecutive bytes. However, existing RS error correction architectures face severe timing bottlenecks when dealing with MBU: the encoding process requires n clock cycles to calculate the check bytes; the decoding process involves complex steps such as synod calculation, solving critical equations, and Chan's search, typically requiring 4n-2k clock cycles. Therefore, completing a full error correction and detection operation requires approximately 5n-2k clock cycles.

[0004] For example, to achieve continuous 4-bit error correction capability, the RS(12,4,8) encoding scheme must be used. This scheme has an encoding delay of 12 cycles and a decoding delay as high as 40 cycles, totaling a delay of 52 clock cycles. If the memory operates at a frequency of 1 GHz, the error correction delay alone is as high as 52 nanoseconds. For high-speed memories requiring low access latency (such as RRAM or PCM modules with DDR interfaces), this will severely reduce system bandwidth. Therefore, how to significantly reduce the clock cycle consumption of RS encoding and decoding while maintaining or even increasing the number of anti-MBU error correction bits has become a bottleneck problem restricting the application of radiation hardening for high-speed memories. Summary of the Invention

[0005] The technical problem solved by this invention is to overcome the shortcomings of the prior art and propose a high-speed multi-bit error correction and detection device and method for memory based on Reed-Solomon codes. The total number of bytes n of low code rate RS codes is significantly reduced, the encoding and decoding clock cycle is greatly shortened, and high-speed error correction and detection is achieved.

[0006] The solution of the present invention is: a high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes, comprising: a memory array, a column interleaving mapping module, a Reed-Solomon encoding and decoding module, and a control logic unit; A storage array consists of several storage units arranged in rows and columns to store data. The column interleaving mapping module is connected to the bit line output of the storage array. On the one hand, it performs bit width remapping with an interleaving distance of ID, outputting several sets of logical codeword data; on the other hand, it performs deinterleaving operation, restoring the logical codeword data and writing it into the corresponding storage cell. The Reed-Solomon codec module is used to receive logical codeword data after column interleaving mapping, perform encoding based on RS(n,k,t) code to generate redundant check bytes, and perform comma calculation and error correction on the read data; where k represents the number of original information bytes, t represents the number of redundant bytes, and the total number of bytes n=k+t; The control logic unit sets the interleaving distance ID according to external configuration instructions, controls the data flow between read and write operations of the column interleaving mapping module, and adjusts the working parameters of the Reed-Solomon encoding and decoding module according to the interleaving distance ID value.

[0007] Furthermore, the Reed-Solomon codec module (103) contains several parallel RS decoder cores, the number of which is equal to the value of the interleaving distance ID.

[0008] Furthermore, the Reed-Solomon encoding / decoding module (103) has an encoding operation delay of n clock cycles, a decoding operation delay of 4n-2k clock cycles, and a total error correction / detection delay of 5n-2k clock cycles. By increasing the interleaving distance ID, the number of redundant bytes t of the required RS(n,k,t) code is reduced while keeping the error correction target unchanged, thereby reducing the value of n and reducing the total error correction / detection delay.

[0009] Furthermore, the interleaving distance ID is the number of bit lines separating two adjacent bits belonging to the same logical codeword in the physical bit line array after column interleaving mapping.

[0010] Furthermore, a single RS(n,k,t) code has the ability to correct t / 2 consecutive byte errors; when the number of physically consecutive flipped bits m caused by high-energy particle bombardment satisfies the condition: m≤ID×(t / 2), the Reed-Solomon codec module (103) converts the physically multiple consecutive flipped bits into several logically independent codeword errors, and completes error correction within each independent codeword. Furthermore, the storage array (101) is any one of resistive random access memory (RRAM), phase change memory (PCM), magnetoresistive random access memory (MRAM), or NAND flash memory (NAND).

[0011] Furthermore, the control logic unit (104) includes a configuration register, used to load preset interleaving distance ID and RS(n,k,t) code parameters during chip initialization; or to dynamically adjust the interleaving distance ID based on error statistics during system operation.

[0012] Furthermore, the column interleaving mapping module (102) includes a multiplexer network configured by control signals, which is connected to the output terminals of several bit lines of the storage array (101) one-to-one through several physical bit lines; during the read operation, a bit width remapping with an interleaving distance of ID is performed to output several sets of logical codeword data; during the write operation, the logical codeword data is restored and written into the storage unit corresponding to the physical bit line.

[0013] Furthermore, the bit width heavy mapping with an interleaving distance of ID is defined by a mapping rule where bits in the physical bit line array whose sequence number interval is an integer multiple of ID are mapped to the same group of logical codewords, specifically: in, Indicates the logical codeword group number; Indicates the physical bit line number.

[0014] High-speed multi-bit error correction and detection methods for memory based on Reed-Solomon codes include: The number of physically consecutively flipped bits m is estimated based on the radiation flux of the application environment, and the interleaving distance ID is determined. During a read operation, several bits of data output from the bit line output terminal of the storage unit are remapped with an interleaving distance of ID to output several sets of parallel logical codeword data. Error correction and detection are performed on several groups of logical codeword data; the RS decoder in each logical group only deals with errors within t / 2 bytes, and the error can be completely corrected when m≤ID×(t / 2); During a write operation, the redundant data generated by the RS encoder passes through the deinterleaving network along with the original data. After restoring the bits in the logical codeword data to the correct physical bit line positions, a programming pulse is applied to write the data into the memory cell.

[0015] The advantages of this invention compared to the prior art are: This invention introduces a column-interleaved architecture based on a "physical-logic domain transformation," adding a multiplexer layer between the storage array bit lines and the RS encoding / decoding logic. This layer "breaks up" physically continuous bit data at a predetermined interleaving distance ID, mapping it to different sets of logic codeword registers. This mechanism causes physically continuous MBUs caused by high-energy particles to manifest as multiple independent single-bit or short burst errors in the logic codeword space. Therefore, the system only needs to configure low-rate RS codes with a small error correction capability t / 2 to effectively achieve error correction for long burst MBUs. Because the total number of bytes n of the low-rate RS code is significantly reduced, the encoding / decoding clock cycle is greatly shortened, achieving high-speed error correction and detection. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the circuit structure of the storage array column interleaving and multiplexer layer in an embodiment of the present invention; Figure 2 This is a schematic diagram comparing the traditional pre-interleaving RS encoding and the post-interleaving RS encoding of the present invention for a line of 1024-bit data under a consecutive 4-bit flip error. Detailed Implementation

[0017] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0018] A high-speed multi-bit error correction and detection method for memory based on Reed-Solomon codes includes the following steps: Step 1: Configure the interleaving distance ID. Based on the estimated maximum number of consecutively flipped bits *m* according to the radiation flux of the application environment, determine the interleaving distance ID of the multiplexer layer. The interleaving distance ID is defined as the number of bit lines separating two adjacent bits belonging to the same logical codeword in the physical bit line array after column interleaving mapping; Step 2: Read Data Column Interleaving Mapping. During the read operation, the physical bit line data read by the sensitive amplifier enters the multiplexer layer. The multiplexer layer rearranges the data according to the formula (Logic Group = Physical Bit Line Number mod ID). For example, when ID=4, physical bit lines 0, 4, 8... belong to the 0th logic group; physical bit lines 1, 5, 9... belong to the 1st logic group.

[0019] Step 3: RS Decoding and Error Correction. Multiple sets of parallel logic codewords are fed into the RS encoding / decoding module. Since physically consecutive m-bit errors have been discretized into different logic groups, the RS decoder within each logic group only deals with errors within t / 2 bytes. When the number of physically consecutive bit flips m caused by high-energy particle bombardment satisfies the condition m≤ID×(t / 2), the Reed-Solomon encoding / decoding module converts the physically consecutive bit flips into multiple logically independent codeword errors and completes error correction within each independent codeword.

[0020] Step 4: Write Data Deinterleaving. During the write operation, the redundant data generated by RS encoding passes through the deinterleaving network along with the original data (i.e., the reverse process of read interleaving). After restoring the bits in the logical codeword to the correct physical bit line positions, a programming pulse is applied to write the data to the memory cell.

[0021] The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes includes four core modules: a memory array 101, a column interleaving mapping module 102, a Reed-Solomon encoding and decoding module 103, and a control logic unit 104. The storage array 101 consists of multiple storage cells used to store raw data and verification data; the storage array 101 is any one of resistive random access memory (RRAM), phase change memory (PCM), magnetoresistive random access memory (MRAM), or NAND flash memory.

[0022] The column interleaving mapping module 102 includes a multiplexer network configured by control signals, connected to the bit line output of the storage array, for performing bit width heavy mapping with an interleaving distance of ID; The Reed-Solomon codec module 103 is used to receive logical codeword data after column interleaving mapping, perform encoding based on RS(n,k,t) code to generate redundant check bytes, and perform comma calculation and error correction on the read data; where k is the number of data bytes, t is the number of redundant bytes, n=k+t is the total number of bytes, and a single RS codeword has the ability to correct errors of t / 2 consecutive bytes.

[0023] The control logic unit 104 includes a configuration register, which is used to load a preset interleaving distance ID during the chip initialization phase and adjust the operating parameters of the Reed-Solomon codec module according to the ID value; or to dynamically adjust the interleaving distance ID according to error statistics during system operation.

[0024] The Reed-Solomon codec module 103 contains multiple parallel RS decoder cores, the number of which is equal to the interleaving distance ID. First, the original data is encoded and redundant check bytes are added. Then, the deinterleaving operation is performed via the multiplexer layer to restore the logical codeword data and write it into the storage unit corresponding to the physical bit line. The encoding operation in the Reed-Solomon codec module 103 has a delay of n clock cycles, the decoding operation has a delay of (4n-2k) clock cycles, and the total delay for a single error correction and detection operation is (5n-2k) clock cycles. By increasing the interleaving distance ID, while keeping the error correction target m unchanged, the number of redundant bytes t of the required RS(n,k,t) code is reduced, thereby reducing the value of n and lowering the total error correction and detection delay.

[0025] Example: like Figure 1 As shown, the memory array 101 is composed of phase-change memory (PCM) cells arranged in rows and columns. Each row of word lines corresponds to 1024 memory cells, meaning that the width of a single row is 1024 bits (128 bytes). After a specific word line is selected by the row decoder, the resistance state of the memory cell is converted into a full-swing digital logic level by a sensitive amplifier group.

[0026] The column interleaving mapping module 102 includes a set of multiplexer (MUX) networks. This multiplexer network is physically placed adjacent to the sensitive amplifier group, and its gating control terminal is connected to the configuration register in the control logic unit 104. During read operations, this module logically rearranges the bit data read from the 1024 physical bit lines according to a preset interleaving distance ID. That is, bits with sequence number intervals that are integer multiples of ID in the physical bit line array are mapped to the same set of logical codewords; the specific mapping rule follows the formula: in, Indicates the logical codeword stream group; Indicates the physical bit line number; Indicates the number of logical codeword stream groups.

[0027] In this embodiment, the configuration register is set to ID=4. Therefore, physical bit line [0] is connected to the 0th bit of the 0th group codeword register; physical bit line [1] is connected to the 0th bit of the 1st group codeword register; physical bit line [2] is connected to the 0th bit of the 2nd group codeword register; physical bit line [3] is connected to the 0th bit of the 3rd group codeword register; physical bit line [4] is connected to the 1st bit of the 0th group codeword register... and so on. After passing through the column interleaving mapping module 102, the originally continuous 1024-bit data is reorganized into 256 independent logical codeword streams, each with a width of 4 bits.

[0028] The Reed-Solomon codec module 103 contains 256 parallel RS(6,4,2) decoder cores. Each decoder core is responsible for processing the corresponding group of 4-bit logic codewords. Here, the parameters are n=6, k=4, and t=2. For a single RS(6,4,2) decoder, its error correction capability is t / 2=1 bit.

[0029] According to the formula m≤ID×(t / 2) of this invention, substituting ID=4 and t / 2=1, we can obtain the maximum number of correctable consecutive physical error bits m=4. That is, the architecture of this invention uses RS codes that can only correct 1 bit error to achieve complete error correction coverage of 4 consecutive MBU bits on the physical array.

[0030] The control logic unit 104 includes a set of memory-mapped configuration registers. After system power-on reset, the bootloader writes the interleaving distance ID value and the corresponding RS decoder enable parameters into the registers based on the radiation intensity predicted by the current orbit. Furthermore, this module is responsible for switching the data flow of the multiplexer network between read and write operations: during a write operation, the multiplexer network operates in "deinterleaving" mode, restoring the logical codeword bits generated by the RS encoder to their physical order.

[0031] Error correction performance and timing comparison analysis, such as Figure 2 As shown, this visually demonstrates the difference in error correction mechanisms between the traditional scheme and this scheme when subjected to high-energy particle bombardment resulting in consecutive 4-bit flips (MBU).

[0032] Reference Figure 2 The upper half (traditional scheme before interleaving): In a row of 1024 bits of data in the storage array, physically adjacent bits 0, 1, 2, and 3 are flipped. In the traditional scheme without interleaving measures, if a single high-rate RS code is used, it must be ensured that the burst correction length of the codeword covers these consecutive 4 bits (i.e., half a byte). Typically, RS(12,4,8) codes are used to cover this area in parallel. The RS(12,4,8) code has n=6, k=2, and t=4. Encoding requires 12 clock cycles, decoding requires 40 clock cycles, and the total error correction / detection delay is 52 clock cycles.

[0033] Reference Figure 2 The lower half (the interleaved scheme of this invention with ID=4): also faces the continuous flipping of bit0, bit1, bit2, and bit3. After passing through the column interleaving mapping module 102 with ID=4, the data distribution undergoes a fundamental change. Bit0 is mapped to the 0th logical codeword, bit1 is mapped to the 1st logical codeword, bit2 is mapped to the 2nd logical codeword, and bit3 is mapped to the 3rd logical codeword. The consecutive 4-bit errors in physical space are completely discretized into single-bit errors in 4 independent logical codewords (each codeword contains only 1 error bit).

[0034] At this point, the Reed-Solomon codec module 103 only needs to utilize 256 parallel RS(6,4,2) decoders. The RS(6,4,2) code has n=6, k=4, t=2, and its error correction capability is t / 2=1 byte. RS(6,4,2) is perfectly adequate for handling single-bit errors. Since n=6, encoding requires only 6 clock cycles, decoding requires 16 clock cycles, and the total error correction / detection delay is only 22 clock cycles.

[0035] Table 1 summarizes the key performance indicators of the two solutions in dealing with the same physical error scenarios.

[0036] Comparison items Traditional solutions (non-interlacing / low-interlacing) The present invention's solution (ID=4 columns interleaved) Physical error type 4 consecutive bit flips 4 consecutive bit flips logical error manifestations Four consecutive bits of error within the same codeword Each of the four independent codewords contains a single-bit error. RS code type selected RS(12,4,8) RS(6,4,2) Total bytes n 12 6 Encoding cycle 12 6 Decoding cycle 40 16 Total delay period 52 22 Latency reduction ratio - Approximately 57.7% As shown in Table 1, this invention achieves nearly 58% error detection and correction timing margin by introducing relatively small hardware overhead (multiplexer wiring layer). This is crucial for improving the memory's read burst length support capability.

[0037] In summary, this invention, by introducing a column interleaving mapping mechanism, breaks the strong coupling between burst error length and encoding / decoding latency in Reed-Solomon error correction codes at extremely low hardware cost. This scheme enables high-speed memories to maintain robustness against long consecutive bit flips while preserving low latency and high bandwidth data throughput performance even in harsh radiation environments, making it particularly suitable for next-generation aerospace-grade solid-state memory controllers and high-reliability industrial embedded systems.

[0038] This invention transforms the correction of continuous burst errors into the correction of discrete independent errors through column interleaving technology, breaking the direct limitation of the traditional RS code's error correction capability t / 2 on the number of consecutive error bits. For the same consecutive error target m, after using ID=4 interleaving, the required t value can be reduced to 1 / 4 of the original (corresponding to a reduction in the n value). Taking a row of 1024 bits of data as an example, under the condition of facing 4 consecutive bit flips, the traditional non-interleaving scheme requires RS(12,4,8) code (n=12, k=4, t=8), which requires 12 cycles for encoding and 40 cycles for decoding, with a total delay of 52 cycles. However, after using the ID=4 interleaving scheme of this invention, only RS(6,4,2) code (n=6, k=4, t=2) is required. Physically consecutive bit 0, bit 1, bit 2, and bit 3 flips are separated into 4 different logical codewords after column interleaving, and each codeword contains only a single bit error, which RS(6,4,2) can completely correct. The scheme requires only 6 cycles for encoding and 16 cycles for decoding, with a total latency of only 22 cycles, which is about 57.7% lower than the traditional scheme.

[0039] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solutions of the present invention by utilizing the methods and techniques disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall fall within the protection scope of the technical solutions of the present invention.

Claims

1. A memory high speed multi-bit error correction and detection device based on Reed Solomon code, characterized in that, include: Storage array (101), column interleaving mapping module (102), Reed-Solomon encoding and decoding module (103), control logic unit (104); The storage array (101) contains several storage units arranged in rows and columns to store data; The column interleaving mapping module (102) is connected to the bit line output terminal of the storage array (101); on the one hand, it performs bit width remapping with an interleaving distance of ID and outputs several sets of logical codeword data; on the other hand, it performs deinterleaving operation to restore the logical codeword data and write it into the corresponding storage unit. The Reed-Solomon codec module (103) is used to receive the logical codeword data after column interleaving mapping, perform encoding based on RS(n,k,t) code to generate redundant check bytes, and perform comma calculation and error correction on the read data; where k represents the number of original information bytes, t represents the number of redundant bytes, and the total number of bytes n=k+t; The control logic unit (104) sets the interleaving distance ID according to the external configuration instructions, controls the data flow between the read and write operations of the column interleaving mapping module (102), and adjusts the working parameters of the Reed-Solomon encoding and decoding module (103) according to the interleaving distance ID value.

2. The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes according to claim 1, characterized in that, The Reed-Solomon codec module (103) contains several parallel RS decoder cores, the number of which is equal to the value of the interleaving distance ID.

3. The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes according to claim 1 or 2, characterized in that, The Reed-Solomon codec module (103) has an encoding operation delay of n clock cycles, a decoding operation delay of 4n-2k clock cycles, and a total error correction and detection delay of 5n-2k clock cycles. By increasing the interleaving distance ID, the number of redundant bytes t of the required RS(n,k,t) code is reduced while keeping the error correction target unchanged, thereby reducing the value of n and reducing the total error correction and detection delay.

4. The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes according to claim 1, characterized in that, The interleaving distance ID is the number of bit lines separating two adjacent bits belonging to the same logical codeword in the physical bit line array after column interleaving mapping.

5. The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes according to claim 1, characterized in that, A single RS(n,k,t) code has the ability to correct t / 2 consecutive byte errors; when the number of physically consecutive flipped bits m caused by high-energy particle bombardment satisfies the condition: m≤ID×(t / 2), the Reed-Solomon codec module (103) converts the physically multiple consecutive flipped bits into several logical independent codeword errors, and completes the error correction in each independent codeword.

6. The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes according to claim 1, characterized in that, The storage array (101) is any one of resistive random access memory (RRAM), phase change memory (PCM), magnetoresistive random access memory (MRAM), or NAND flash memory (NAND).

7. The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes according to claim 1, characterized in that, The control logic unit (104) includes a configuration register, which is used to load a preset interleaving distance ID and RS(n,k,t) code parameters during the chip initialization phase; or to dynamically adjust the interleaving distance ID based on error statistics during system operation.

8. The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes according to claim 1, characterized in that, The column interleaving mapping module (102) includes a multiplexer network configured by control signals, which is connected to the output terminals of several bit lines of the storage array (101) one-to-one through several physical bit lines. During the read operation, a bit width remapping with an interleaving distance of ID is performed to output several sets of logical codeword data. During the write operation, the logical codeword data is restored and written into the storage cell corresponding to the physical bit line.

9. The high-speed multi-bit error correction and detection device for memory based on Reed-Solomon codes according to claim 1 or 8, characterized in that, The bit width heavy mapping with an interleaving distance of ID is defined by the following rule: bits in the physical bit line array whose sequence number interval is an integer multiple of ID are mapped to the same group of logical codewords. Specifically: in, Indicates the logical codeword group number; Indicates the physical bit line number.

10. A method for a high-speed multi-bit error correction and detection device for a memory based on Reed-Solomon codes as described in any one of claims 1 to 9, characterized in that, include: The number of physically consecutively flipped bits m is estimated based on the radiation flux of the application environment, and the interleaving distance ID is determined. During a read operation, several bits of data output from the bit line output terminal of the storage unit are remapped with an interleaving distance of ID to output several sets of parallel logical codeword data. Error correction and detection are performed on several groups of logical codeword data; the RS decoder in each logical group only deals with errors within t / 2 bytes, and the error can be completely corrected when m≤ID×(t / 2); During a write operation, the redundant data generated by the RS encoder passes through the deinterleaving network along with the original data. After restoring the bits in the logical codeword data to the correct physical bit line positions, a programming pulse is applied to write the data into the memory cell.