A variable length turbo code decoder device

By using sliding window MAX-LOG-MAP algorithm and variable length interleaver module in the Turbo code decoder, the adaptive frame length adjustment of the Turbo code decoder is realized, and the decoding performance in the channel environment in traditional design is solved. The problem of difficulty in balancing time delay improves the universality and application value of the system.

CN113872615BActive Publication Date: 2024-11-19THE 54TH RESEARCH INSTITUTE OF CHINA ELECTRONICS TECHNOLOGY GROUP CORPORATION
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202111176822.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-09
Publication Date
2024-11-19
Estimated Expiration
2041-10-09

AI Technical Summary

Technical Problem

The traditional Turbo code decoder design lacks universality and cannot adaptively adjust the frame length according to different channel environments, making it difficult to achieve the optimal balance of decoding performance and delay.

Method used

A variable-length Turbo code decoder was designed, using sliding window MAX-LOG-MAP algorithm and variable-length interleaver module to realize adaptive subdivision and iterative decoding of data frames through the controller, combined with the 3GPP standard The interleaver generation method allows users to set the frame length by themselves to optimize decoding performance and delay.

Benefits of technology

It achieves the optimal balance of decoding performance and delay in different channel environments, reduces resource occupation and hardware complexity, and has wide application value and practicality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113872615B_ABST
    Figure CN113872615B_ABST
Patent Text Reader

Abstract

The present invention discloses a variable length Turbo code decoder device, belonging to the field of Turbo code decoders. The device comprises a variable length interleaver module, a controller, a SISO component decoder module, and a data storage module. The device can be set by the user according to the different index requirements of the system for information reliability and data real-time performance, and can adapt to different application scenarios, and has a wide range of application value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of Turbo code decoders, and in particular to a variable-length Turbo code decoder device. Background Art

[0002] Turbo code was proposed by C.Berrou and Alain Glavieux in 1993. It has been widely used in communication systems due to its excellent performance close to the Shannon theoretical limit. Turbo code is a parallel cascade convolutional code that combines component codes and random interleavers and adopts a parallel cascade structure to realize the idea of ​​random coding. Turbo decoding adopts the iterative decoding idea of ​​soft input and soft output, achieving performance close to the Shannon theoretical limit.

[0003] The common Turbo code decoder principle is as follows Figure 1 As shown in the figure, it consists of two soft-input soft-output (SISO) component decoders DEC1 and DEC2 serially cascaded. The interleaver used in the decoder must be the same as the interleaver in the encoder. DEC1 and DEC2 respectively perform optimal decoding on the two component codes in the encoder, and send the external information obtained by each to the other component decoder as its prior information. After multiple iterations, the external information of DEC1 and DEC2 tends to be stable, and the likelihood ratio approaches the maximum likelihood decoding. Then, a hard decision is made on this likelihood ratio to complete the decoding process.

[0004] The error performance and delay of Turbo code depend largely on the information frame length. The longer the frame length, the longer the interleaver length, the closer the Turbo code is to a random long code, the better the decoding performance, and the greater the decoding delay. Traditional Turbo code decoder hardware design selects a single parameter based on simulation results for fixed design. When the indicators change, the program needs to be significantly modified and reconfigured, or even the system solution needs to be replaced, which lacks versatility. Summary of the invention

[0005] In view of this, the purpose of the present invention is to avoid the deficiencies in the above-mentioned background technology and provide a variable-length Turbo code decoder device, in which the frame length can be set by the user according to the needs, so that the decoding delay and decoding performance can achieve the best balance in different channel environments, and the device has wide application value.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A variable length Turbo code decoder device comprises a variable length interleaver module, a controller, a SISO component decoder module, and a data storage module; wherein:

[0008] After the turbo code decoder device receives a frame of data to be decoded, it stores the data in the data storage module;

[0009] The controller sends a start flag to start the variable length interleaver module, and after completing the interleaving operation according to different data lengths, the mapping pattern is written into the data storage module for storage;

[0010] After the interleaving address is generated, the controller sends out the first frame data decoding start flag to control the SISO component decoder module to start iterative decoding;

[0011] The SISO component decoder module uses a component decoder circuit to realize the functions of two component decoders by multiplexing under the control of the controller; the SISO component decoder module adopts a sliding window MAX-LOG-MAP decoding algorithm to evenly subdivide the input data frame into small blocks of data frames, and decode each subframe separately until all subframes are decoded;

[0012] When the number of iterations reaches the program setting value, the controller performs hard decision on the likelihood ratio output by the SISO component decoder module to obtain a decoding result.

[0013] Furthermore, the variable-length interleaver module adopts the interleaver generation method recommended in the 3GPP standard, the interleaver is variable-length, and the length of the interleaver is controlled by user input; the variable-length interleaver module calculates the mapping pattern of interleaving and deinterleaving and writes it into the data storage module, which is read and used by the controller when it is used to complete the interleaving and deinterleaving.

[0014] Furthermore, the SISO component decoder module includes a branch transfer metric γ calculation module, a training backward metric train-β calculation module, an effective backward metric β calculation module, a forward metric α and a log-likelihood ratio llr calculation module; wherein,

[0015] A frame of data is evenly divided into sub-frames of the same length. Each sub-frame becomes an independent window, and each sub-frame is decoded separately. The SISO component decoder module uses the sliding window MAX-LOG-MAP algorithm, and the decoding process is as follows:

[0016] (1) Calculate the branch transfer metric γ and save it to memory;

[0017] (2) Assume that the initial value of the training backward metric train-β of each window is 0 or an equal probability value, and use the reverse recursive calculation of the previous window to the first bit of the window. This bit state information is the train-β of the window;

[0018] (3) The train-β value is used as the initial value of the recursive calculation of the effective backward metric β of the current window, and γ is read to solve β;

[0019] (4) Read γ and β to solve the forward metric α and the log-likelihood ratio llr;

[0020] (5) Continuously recursively calculate according to the grid time until a complete frame of data is solved;

[0021] (6) Repeat the above steps until the number of iterations meets the value set in the program.

[0022] Compared with the background technology, the present invention has the following advantages:

[0023] (1) The present invention adopts a sliding window MAX-LOG-MAP algorithm and combines it with a storage table to implement a variable length interleaver, which effectively solves the problem of excessive resource occupation and prolonged decoding time, and is novel and creative.

[0024] (2) The present invention is suitable for decoding variable-length data frames, so that the decoding performance and decoding delay achieve the best balance. It has the characteristics of multiple application scenarios and strong practicality, and has wide application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a structural block diagram of a common Turbo decoding principle in the background technology.

[0026] Figure 2 It is a structural block diagram of a Turbo decoding device in an embodiment of the present invention.

[0027] Figure 3 yes Figure 2 The specific structural block diagram of the SISO module. DETAILED DESCRIPTION

[0028] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0029] like Figure 1-3 As shown, a variable length Turbo code decoder device includes a variable length interleaver module, a controller, a SISO component decoder module, and a data storage module. The data storage module in this example uses RAM; wherein,

[0030] After the Turbo code decoder device receives a frame of data to be decoded, it stores the data in RAM; the controller sends a start mark to start the variable length interleaver module, and after completing the interleaving operation according to different data lengths, the mapping pattern is written into the RAM for storage; after the interleaving address is generated, the controller sends the first frame data decoding start mark to control the SISO component decoder module to start iterative decoding. The SISO component decoder module uses a component decoder circuit to realize the functions of two component decoders through multiplexing under the control of the controller. The SISO component decoder module uses a sliding window MAX-LOG-MAP decoding algorithm to evenly subdivide the input data frame into small blocks of data frames, and decode each subframe separately until all subframes are decoded. Repeat the decoding process, and when the number of iterations reaches the program setting value, a hard decision is made on the likelihood ratio output by the component decoder to obtain the decoding result.

[0031] Furthermore, the variable length interleaver module adopts the interleaver generation method recommended in the 3GPP standard, and the interleaver is variable length. In order to reduce hardware complexity and save hardware resources, a C program is designed to calculate the mapping pattern of interleaving and deinterleaving during implementation, and the mapping pattern is written into RAM, which is read and used by the controller when it is used to complete interleaving and deinterleaving.

[0032] Furthermore, the controller mainly completes the control of the following parts:

[0033] 1) Control the read and write operations of RAM;

[0034] 2) Control the multiplexing of component decoders and decide whether to use them as DEC1 or DEC2;

[0035] 3) Decide to interleave or deinterleave external information;

[0036] 4) Iteration number control;

[0037] 5) Start hard decision output.

[0038] The specific process is as follows:

[0039] 1) Wait for external input data, start the variable length interleaver module, calculate the mapping pattern according to the input data length, and store it in RAM;

[0040] 2) Sending an identification signal indicating that the window starts calculating to each decoder module;

[0041] 3) Sending the first iteration start signal and the signal of whether to perform interleaving and deinterleaving selection to each module;

[0042] 4) Sending a start signal for the next iteration to each decoder module;

[0043] 5) When all modules have finished running, wait for the window to finish running.

[0044] 6) Output the decoded data according to the received window end identification signal.

[0045] Furthermore, the SISO component decoder module uses a component decoder circuit, and realizes the functions of two component decoders by multiplexing under the action of the controller. This example adopts the sliding window MAX-LOG-MAP decoding algorithm to evenly subdivide the input data frame into small blocks of data frames, and decode each subframe separately. For a complete component decoding, it includes a branch transfer metric γ calculation module, a training backward metric train-β calculation module, an effective backward metric β calculation module, a forward metric α and a log-likelihood ratio llr calculation module. Among them,

[0046] The solution of the branch transfer metric γ is the key to solving other modules. The forward metric α, the backward metric β and the log-likelihood ratio llr all need to use the γ value to calculate. The branch transfer metric calculation formula is:

[0047]

[0048] In the above formula, M K (s',s) is the logarithm of the branch transfer metric γ; are the information bit and check bit input by the decoder at time k respectively; u k , They are the information bit and check bit output by the encoder respectively; L a (u k ) is the prior information, which comes from the external information generated by the previous component decoder after interleaving or deinterleaving. In the first iteration of decoding, L a (u k ) is 0; L C is the channel confidence, and its value is 2.

[0049] After the branch transfer metric is solved, the branch metric is saved separately in two hardware memories according to whether the input information bit is "1" or "0". The memory stores three identical branch metrics, which are used to solve the training backward metric train-β module, the effective backward metric β module, the forward metric α and the log-likelihood ratio llr module respectively.

[0050] The calculation process of the training backward metric train-β is:

[0051] Assume that the initial value train-β of the backward metric of each window is 0 or an equal probability value, and use the reverse recursive calculation of the previous window to the first bit of the window. At this time, the bit state information train-β is used as the initial value of the recursive calculation of the effective backward metric β of the current window. In order to find the accurate initial state of the effective backward metric, it is necessary to train 5 to 8 times the constraint length of the convolutional code encoder. In this example, the encoder is a (13, 15) component code, and the constraint length is 4, so the length of the training window is set to 32 bits. The calculated train-β value is stored in RAM and provided to the calculation of the effective backward metric module as the initial value of its recursive calculation.

[0052] The calculation formula of effective backward metric β is:

[0053] B k (s) = lnβ k (s')=max(B k+1 (s)+M K (s',s))

[0054] In the formula, B k (s) is the logarithm of the effective backward metric β, M K (s',s) is the logarithm of the branch transfer metric γ.

[0055] The calculation of the effective backward metric is a recursive operation from the back to the front based on the initial value. The β calculation at the current moment requires the β value at the next moment, and the calculation result needs to be stored in the memory to calculate the β value at the previous grid moment.

[0056] The forward metric α is calculated as:

[0057] A k (s) = lnα k (s)=max(A k-1 (s')+M K (s',s))

[0058] In the formula, A k (s) is the logarithm of the forward metric α, M K (s',s) is the logarithm of the branch transfer metric γ.

[0059] The forward metric α is calculated by forward recursion based on the initial value of the forward state. The calculation of α at the current moment uses the α value at the previous moment, and the calculation result needs to be stored in the memory to calculate the α value at the next grid moment.

[0060] The log-likelihood ratio llr calculation formula is:

[0061]

[0062] In the formula, A k-1 (s') is the logarithm of the forward metric, M K (s',s) is the logarithmic value of the branch transfer metric, B k (s) is the logarithm of the backward metric.

[0063] When calculating the log-likelihood ratio llr, the α value, β value, and γ value must be calculated first. In the implementation process, the order of sequentially outputting external information is adopted, and the β value is calculated first, and then the α value and llr are calculated for decision output. The order of calculating the log-likelihood ratio llr is:

[0064] 1) Calculate the γ value and save it as 3 backups;

[0065] 2) Solve the β value based on the γ value that has been calculated and save it;

[0066] 3) According to the γ value, find the α value;

[0067] 4) Calculate the value of the log-likelihood ratio llr based on the α and β values.

[0068] In short, the device can allow users to set the input data frame length according to the system's different requirements for information reliability and data real-time performance, can adapt to different application scenarios, and has wide application value.

[0069] It should be noted that, in addition to the above-mentioned implementation steps, the present invention may also have other implementation modes. Any technical solution using equivalent replacement or equivalent transformation forms falls within the protection scope required by the present invention.

Claims

1. A variable length Turbo code decoder device, characterized in that: It includes a variable length interleaver module, a controller, a SISO component decoder module, and a data storage module; wherein, After the turbo code decoder device receives a frame of data to be decoded, it stores the data in the data storage module; The controller sends a start flag to start the variable length interleaver module, and after completing the interleaving operation according to different data lengths, the mapping pattern is written into the data storage module for storage; After the interleaving address is generated, the controller sends out the first frame data decoding start flag to control the SISO component decoder module to start iterative decoding; The SISO component decoder module uses a component decoder circuit to realize the functions of two component decoders by multiplexing under the control of the controller; the SISO component decoder module adopts a sliding window MAX-LOG-MAP decoding algorithm to evenly subdivide the input data frame into small blocks of data frames, and decode each subframe separately until all subframes are decoded; When the number of iterations reaches the program setting value, the controller makes a hard decision on the likelihood ratio output by the SISO component decoder module to obtain the decoding result; The variable length interleaver module adopts the interleaver generation method recommended in the 3GPP standard. The interleaver is variable in length, and the length of the interleaver is controlled by user input. The variable length interleaver module calculates the mapping pattern of interleaving and deinterleaving and writes it into the data storage module. When it is used, it is read and used by the controller to complete the interleaving and deinterleaving. The SISO component decoder module includes a branch transfer metric γ calculation module, a training backward metric train-β calculation module, an effective backward metric β calculation module, a forward metric α and a log-likelihood ratio llr calculation module; wherein, A frame of data is evenly divided into sub-frames of the same length. Each sub-frame becomes an independent window, and each sub-frame is decoded separately. The SISO component decoder module uses the sliding window MAX-LOG-MAP algorithm, and the decoding process is as follows: (1) Calculate the branch transfer metric γ and save it to memory; (2) Assume that the initial value of the training backward metric train-β of each window is 0 or an equal probability value, and use the reverse recursive calculation of the previous window to the first bit of the window. This bit state information is the train-β of the window; (3) The train-β value is used as the initial value of the recursive calculation of the effective backward metric β of the current window, and γ is read to solve β; (4) Read γ and β to solve the forward metric α and the log-likelihood ratio llr; (5) Continuously recursively calculate according to the grid time until a complete frame of data is solved; (6) Repeat the above steps until the number of iterations meets the value set in the program.

Citation Information

Patent Citations

  • Generation method of variable length S random interleaver

    CN102227097A

  • Turbo code coding device and method

    CN104092470A

  • Compilation method for variable-length Turbo codes

    CN104378122A

  • Method and device for source decoding a variable-length soft-input codewords sequence

    CN1682450A

  • Method and apparatus for parallel turbo decoding in long term evolution system (LTE)

    US20120106683A1