RU2713573C1 - Data transmission device based on codes with low density of checks on parity - Google Patents

Abstract

FIELD: data processing.

SUBSTANCE: invention relates to means for encoding with error detection and correction. Device additionally comprises inter-stack summation unit, control input of which is connected to output of control unit, first and second inputs of inter-stack summation unit are connected to outputs of memory unit, output of unit of inter-stack accumulation is connected to third input of multiplexer; output of summation unit is connected to input of memory unit, control input of which is connected to output of control unit, control unit second input, which is input of feedback signal from receiving device transmitted through feedback channel output, and third input of transmitting device. At that, the receiving device additionally includes a soft solutions storage and modification unit, the first output-input of which through serially connected second iterative decoder, second unit for calculating checksum in decoded code word and second error control unit in decoded code word is connected to input of feedback channel.

EFFECT: technical result consists in improvement of efficiency of noise-immune coding.

1 cl, 9 dwg

Description

The invention relates to the field of coding theory, in particular to systems for combined coding with error correction and detection in order to increase the efficiency of use of the spectrum when transmitting data in a digital radio communication system.

равная отношению число информационных символов к кодовым, поскольку кодирование вносит избыточность, скорость кодирования

.The urgent task is not only the correction of errors in channel coding, but also the detection of uncorrected errors and the organization of an automatic repeat request (AZP) in order to correct such errors due to the additional transmission of the minimum possible number of code symbols. The main characteristic of error-correcting coding is the coding rate

equal to the ratio of the number of information symbols to code symbols, since encoding introduces redundancy, encoding speed

.

The parallel convolutional turbo code is widely known [Berrou C. Near Shannon limit error - correcting coding and decoding / C. Berrou, A. Glaviex, P. Thitimajshina // IEEE International Communications Conference: Proc. 1993. - Geneva (Switzerland), 1993. - P. 1064-1070], which allowed to approach the potentially achievable characteristics of noise-resistant coding. The method is characterized by the following. An information sequence is fed to the input of the first encoder, the same sequence is fed to the input of the second, but after the interleaver. Thus, statistical independence is achieved, especially with large code block sizes. An iterative approach is that the results of one decoding with a soft output can be used with another decoding, as a priori information. However, the turbo code is not more efficient than the convolutional code decoded by the Viterbi algorithm with a frame size of less than 200 bits.

An analogue of the proposed method is a product of the company Trellis Ware code F-LDPC [Halford, Thomas & Bayram, Metin & Kose, Cenk & M. Chugg, Keith & Polydoros, Andreas. (2008). The F-LDPC Family: High-Performance Flexible Modern Codes for Flexible Radio. 376 - 380. 10.1109 / ISSSTA.2008.75.]. The advantage of this solution is the deep integration of the automatic retransmission request protocol (ADR) with the error-correcting coding algorithm. It is argued that the solution has the most flexible capabilities, a single encoder / decoder core supports 40 code rates (1 / 2–32 / 33) while the permissible frame sizes are 128, 256, 512, ..., 16384 bits. However, for example, for voice data, the frame sizes produced by the vocoder can be 16, 40, 80, 172 bits, so the inability to work with blocks of 20 bits or more is a drawback of the F-LDPC code.

Classical ARQ (ARQ), which in case of erroneous decoding of a frame involves erasing the received data, complete retransmission of the frame and independent decoding of the newly received frame is not effective, because leads to too much redundant data transfer. Currently, hybrid retransmission is becoming increasingly common [Error Control Coding: Fundamentals and Applications / S. Lin, D. J. Costello, Jr. - NJ: Englewood Cliffs. - 1983. - 603 p.]. The essence of the method is that such a strategy is more effective that provides for error decoding to save the frame data, obtain some additional information from the transmitter, and re-decode the frame using this additional information. If the amount of additional information is less than the volume of the whole frame, then the efficiency of using the communication channel is significantly increased. This approach was investigated by Hagenauer [J. Hagenauer. Rate Compatible Punctured Convolutional Codes (RCPC Codes) and their Applications. // IEEE Trans. Commun. - Vol. 36. - Apr. 1988. - P. 389-400.] And Chase [D. Chase. Code Combining A Maximum Likelihood Decoding Approach for Combining an Arbitrary Number of Noisy Packets. // IEEE Trans. Comm. - Vol. 33. - May 1985. - P. 385393] from different positions.

Consider the classification of the considered schemes AZP (ARQ):

• Hybrid AZP Type 1 (HARQ-1 - Chase summation). This retransmission strategy is characterized in that the same frame is retransmitted.

• Hybrid AZP Type 2 (HARQ-2 - redundancy buildup). This retransmission strategy is characterized in that the number of bits for retransmission is less than in the original frame. A copy for retransmission cannot be decoded on its own.

• Hybrid AZP Type 3 (HARQ-3). A copy for retransmission can be decoded independently. However, the frame for retransmission may differ from the originally transmitted frame, that is, it may be encoded with a different code.

This classification can be summarized as follows:

• Automatic transfer station (ARQ) with full transmission (Type 1 and Type 3 belong to it).

• Automatic transfer circuit (ARQ) with partial transmission (Type 2).

The topic of further research will be a partial transmission - hybrid AZP type 2 (HARQ-2).

Several approaches to the implementation of such a strategy of ARP (ARQ) are possible:

1. Retransmission of part of the previously transmitted code symbols of the frame and their optimal addition to the code symbols stored in the receiver.

2. The use of codes with increasing complexity. In case of erroneous decoding on the transmitting side, new (additional) check symbols are calculated, which are transmitted to the receiver.

The known method [Pat. 2427491 CA (0237743 WO), MKI H04L1 / 18. Automatic Request Protocol Based Packet Transmission Using Punctured Codes / R. Wang, Ming J., A. Garmonov, A. Savinkov, A. Zhdanov. –PST Gazette. - 2002, May 19. - P. 9340], where such a strategy does not lead to a decrease in throughput since additional code symbols or partial retransmission are transmitted as part of the next packet, where additional punching is performed, making room for verification code symbols for a packet in which an error is detected, thus, an error in one package is corrected due to the redundancy of another.

In the class of linear block codes, a subclass of codes with a low density of parity checks [Gallager R. “Low Density Parity Check Codes” // MIT Press 1963], which is superior in its characteristics to a turbo code, can be especially distinguished. A feature of this code is the presence of a simple graphical model of a decoding algorithm, defined as decoding on a graph (Fig. 1). Decoding on a graph should be understood in the sense that the graph defines a partition of the entire codeword into shorter codes, and short codes can overlap with each other. Overlapping data can be avoided by repeating overlapping code characters. Different copies of the received code symbol or soft decision on this symbol are included in different code words of the shorter code, and all code words of the short code can be decoded independently. Such a structure can be represented as a bipartite graph, one type of vertices, which is associated with code symbols, and the number of edges emanating from each vertex corresponds to the number of repetitions of the code symbol (variable vertex is indicated by a rectangle). Another type of vertices corresponds to short codes, and each vertex is a codeword with the number of code symbols equal to the number of incoming edges (the verification vertex is indicated by a circle). A checksum can be assigned to each test vertex as an indicator of errors in the short code. Variable vertices can only be connected with test vertices, and, conversely, test vertices can only be connected with variable vertices, which allows us to consider the graph as bipartite.

Such a bipartite graph is called the graph of Tanner [Tanner R.M. A Recursive Approach to Low Complexity Codes / R.M. Tanner // IEEE Transaction on Information Theory. - 1981. –Vol. IT27, No. 9. - P. 533547]. In fact, for a linear block code, the Tanner graph is a graphical representation of the verification matrix. Error-free decoding is characterized by the equality to zero of all checksums - code constraints.

Code with a low density of parity checks can be decoded in soft solutions using the message exchange algorithm (Fig. 1). The messaging algorithm [Gallager R. “Low Density Parity Check Codes” // MIT Press 1963] initializes each variable vertex with information about the reliability of reception of each symbol (soft decision) received from the channel.

1. The value of the outgoing message from the variable vertex of equal posterior probability for the code symbol is calculated to take a certain value of the code symbol assigned to this variable vertex for each edge of the bipartite graph.

2. At each test vertex, outgoing messages from those variable vertices with which they are connected by the edge of the bipartite graph and for which they are incoming are modified, matching each edge with outgoing messages to the test vertex, which are equal to the second set of probabilities for the code symbol to take a certain value .

3. Update the outgoing message to the variable top with incoming messages, which are outgoing messages to the check top.

4. Repeat iterations a certain number of times or until code restrictions are met.

. Эта функция является решающей функцией или логарифмом отношения правдоподобия. Функция обладает коммутативностью:The sign of the outgoing message is selected so as to satisfy code constraints. The absolute value of the outgoing message is calculated using the function

. This function is the decisive function or the logarithm of the likelihood ratio. The function has commutativity:

. Наиболее простой известен как алгоритм “min-sum”. Абсолютное значение исходящего сообщения равно минимальному абсолютному значению среди входящих сообщений за исключением сообщения от той вершины, для которого предназначено данное исходящее сообщение:There are several methods for defining a function.

. The simplest is known as the min-sum algorithm. The absolute value of the outgoing message is equal to the minimum absolute value among incoming messages, with the exception of the message from the vertex for which this outgoing message is intended:

состоит в определении функции через функцию

, применяемую в LOG MAP алгоритме [Pietrobon S.S. Implementation and Performance of a Turbo/MAP Decoder / S.S. Pietrobon // International J. of Satellite Communication. – 1998. Vol. 16 (Jan-Feb). – P. 23-46]. Функция может быть задана как:Another method of specifying a function

consists in defining a function through a function

used in the LOG MAP algorithm [Pietrobon SS Implementation and Performance of a Turbo / MAP Decoder / SS Pietrobon // International J. of Satellite Communication. - 1998. Vol. 16 (Jan-Feb). - P. 23-46]. The function can be defined as:

, может быть представлена в виде таблицы.Where is the function

may be presented in tabular form.

определено как:There is a modification of the proposed method, where the additional term

defined as:

Modifications of this algorithm are given in [W.K Leung, W.L Lee, A.Wu, L.Ping, “Efficient implementation technique of LDPC decoder” Electron.Lett .., vol. 37 (20), pp.1231-1232, Sept2001]. This method does not require a table, but leads to some deterioration in performance, but it allows you to reduce the number of operations during decoding.

Currently, the class of “turbo-like” codes is widely known, among which are the codes with “repetition-accumulation” known from [H. Jin, A. Khandekar and RJ McEliece, “Irregular repeat-accumulate codes," Proceedings of the Second International Symposium on Turbo Codes and Related Topics, pp. 1-8, Brest, France, September 2000]. “Repeat - by accumulation ”is carried out in the following order: repetition of each information symbol, interleaving of repeated characters, accumulation of the sum modulo 2 interleaved characters. These codes have the properties of both turbo codes and codes with a low density of parity checks and can be decoded in parallel mode - the messaging algorithm, and in serial mode - MAP algorithm. It should be noted that in the latter case, the algorithm is essentially hybrid, consisting of a MAP algorithm for decoding a convolutional code with two states and a message exchange algorithm with unevenly repeated information symbols [J. Li, Low-Complexity, Capacity-Approaching Coding Schemes : Design, Analysis and Applications, Ph.D. dissertation, Texas A&M University, 2002.]. From [H. Jin, A. Khandekar and RJ McEliece, “Irregular repeat-accumulate codes," Proceedings of the Second International Symposium on Turbo Codes and Related Topics, pp. 1-8, Brest, France, September 2000], it is known that the greatest efficiency from application is achieved by irregular or irregular repetition of information symbols.In addition, the characteristics of accumulation repetition codes are characterized to a large extent by an interleaver, which together with an irregular repetition pattern and a checksum accumulation pattern determine the Tanner graph for a given code. ensure sufficient diversity of repeated characters, thereby ensuring the absence or small number of code words with low weight. The law of interleaving or permutation is set by a sequence of non-repeating numbers, each of which is used exactly once, such that they can form a natural series of numbers (each subsequent number on unit larger than the previous one), if arranged in ascending order. Numbers in the form of a natural series is a sequence before interleaving, the same numbers in the form of a pseudo-random sequence is a sequence after interleaving.

An example of such a code is shown in FIG. 8. This is a systematic code consisting of 7 variable nodes, of which 4 are informational, and 3 are test nodes. This code corresponds to a verification matrix of the form:

, где первый массив отражает количество повторений, а второй число повторенных символов. Рассмотрим процедуру кодирования. Пусть информационные символы будут

и индекс соответствует номеру переменного узла. Тогда после повторения последовательность символов примет вид

. Общее число пвторенных символов семь. После перемежения по закону

имеем последовательность вида

. Добиваясь того чтобы контрольная сумма на узле 3 была нулевой, формируем кодовый символ

. Далее следуя "зигзаг" паттерну добиваемся равенству нулю остальных контрольных сумм:

,

.An irregular repetition pattern is a correspondence between the number of characters and the number of repetitions. For the code of FIG. 8 two information symbols are repeated once, one information symbol is repeated two times, one information symbol is repeated three times. A compact repetition pattern can be written in two arrays:

where the first array reflects the number of repetitions, and the second number of repeated characters. Consider the encoding procedure. Let the information symbols be

and the index corresponds to the variable node number. Then after repeating the sequence of characters will take the form

. The total number of repeated characters is seven. After interleaving by law

we have a sequence of the form

. In order to ensure that the checksum on node 3 is zero, we form a code symbol

. Further, following the “zigzag” pattern, we achieve equality to zero of the remaining checksums:

,

.

, где первый массив отражает количество накоплений путем суммирования по модулю 2, а второй число накопленных символов.The pattern of accumulating checksums, similarly to the pattern of repeating code symbols, can be written in the form of two arrays

, where the first array reflects the number of accumulations by summing modulo 2, and the second number of accumulated characters.

символов, где

- коэффициент параллельности перемежителя. Чем он больше, тем больше пропускная способность устройства. Для того чтобы схема параллельного перемежения была реализуемой, работа каждого из процессоров не должна приводить к конфликтам при доступе к памяти. К числу таких конфликтов можно, например, отнести попытку одновременной записи и чтения отдельной ячейки памяти. Память обычно реализуют в виде блоков (“банков”) памяти, таким образом, что к каждому блоку за один квант времени возможен либо доступ на запись (по указанному адресу), либо доступ на чтение. When encoding, interleaving is performed only among repeated information symbols, and when decoding, redundant code symbols obtained according to the zigzag pattern are added to them. Speaking of the interleaver, the interleaver of only information symbols is usually meant, since only it carries a pseudo-random component. Actual is the task of constructing a partially parallel interleaver, in which work is interleaved immediately for one clock cycle

characters where

- coefficient of parallelism of the interleaver. The larger it is, the greater the bandwidth of the device. In order for the parallel interleaving scheme to be feasible, the operation of each processor should not lead to conflicts when accessing memory. Such conflicts include, for example, an attempt to simultaneously write and read a single memory cell. Memory is usually implemented in the form of blocks (“banks”) of memory, so that for each block in one time slot, either write access (at the specified address) or read access is possible.

The method is based on the classic irregular repetition code with accumulations with a low density of parity checks [D. Divsalar, H. Jin, and R. J. McEliece. "Coding theorems for‘ turbo-like ’codes." Proc. 36th Allerton Conf. on Communication, Control and Computing, Allerton, Illinois, Sept. 1998, pp. 201–210.], Where a new effective interleaver is applied, and decoding quality is evaluated, that is, a decision is made on the correct reception

The disadvantage of the classical scheme when decoding according to the algorithm of "distribution of messages" is that it is difficult to coordinate speeds by punching or puncturing. When replacing the punctured character with zero and the arrival of this zero at the decoding check node, all zeros will be issued as an outgoing message, which reduces the decoding efficiency.

The closest analogue in technical essence to the proposed one is the device according to the patent of the Russian Federation 2323520, H03M 13/00 "Method for transmitting voice data in a digital radio communication system", adopted as a prototype.

In FIG. 3 shows a diagram of a transmitting device (transmitting side) of the prototype device, where it is indicated:

1 - block repetition of information symbols;

2 - block interleaving information symbols;

3 - block separation of test characters into groups;

4 - control unit;

5 - block summation;

6 - multiplexer;

8 - memory block;

9 - communication channel;

24 is the key.

The prototype device consists of a transmitting and receiving devices connected by a communication channel 9, which provides data transfer. The transmitting device contains series-connected information symbol repetition unit 1, information symbol interleaver 2, block for separating test characters into groups 3, summing unit 5, key 24, memory unit 8 and multiplexer 6, the output of which is connected to communication channel 9, from the output of which the signal goes to the receiver. In addition, the outputs of the control unit 4 are connected to the control inputs of units 1, 2, 3, 5, 6, and 24, respectively. In this case, the first input of the information symbol repetition unit 1 is the first input of the transmitting device to which the data packets arrive, the same input is connected to the second input of the multiplexer 6. The input of the control unit 4 is the synchronization input, as well as the second input of the transmitting device.

In FIG. 4 shows a diagram of a receiving device, where it is indicated:

10 is a block for the formation of correlation responses and estimates of the noise parameter;

11 - block separation soft solutions;

12 is a block repetition of test characters;

13 - block repetition of information symbols;

14 - block interleaving information symbols;

15 - block forming interleaved soft decisions;

16 - iterative decoder;

17 is a block checksum in a decoded codeword;

18 is an error control unit in a decoded codeword.

The receiving device comprises a series-connected unit for generating correlation responses and estimating a noise parameter 10, a soft decision separation unit 11, an information symbol repetition unit 13, an information symbol interleaving unit 14, an interleaved soft decision generating unit 15, an iterative decoder 16, a checksum calculation unit in decoded the codeword 17 and the error control unit in the decoded codeword 18, the first output of which is the output of the device, and the second output is connected to the second input iterative decoder 16. In addition, the other output of the soft decision separation unit 11 through the check symbol repetition unit 12 is connected to the second input of the interleaved soft decision forming unit 15, the third input of which is connected to the second output of the correlation response generating unit and estimating the noise parameter 10, the input of which is the input of the receiving device, which receives a signal from the transmitting device through the communication channel 9.

The prototype device operates as follows.

путем суммирования по заданному модулю проверочного символа

со всеми перемеженными повторенными информационными символами из группы

. Полученные проверочные символы запоминают в блоке памяти 8. Запомненные проверочные символы поступают в мультиплексор 6, где формируют кодовое слово путем добавления проверочных символов к информационным символам. Через канал связи 9 сформированное кодовое слово передают на приемное устройство. On the transmitting side, the generated data packets, each of which contains an ordered sequence of information symbols, receive the input of the information symbol repetition unit 1 and simultaneously to the second input of the multiplexer 6. In the information symbol repetition unit 1, each information symbol is repeated a certain number of times, depending on its number in order in the sequence of characters, forming a sequence of repeated information symbols, and the number of repetitions is determined in advance . The sequence of repeated information symbols from the output of the information symbol repetition unit 1 is fed to the input of the information symbol interleaver 2, where interleaving is performed in the sequence of repeated information symbols, achieving sufficient diversity of successive repeated information symbols, and forming a sequence of interleaved repeated information symbols. In the block for dividing the check symbols into groups 3, interleaved repeated information symbols are divided into groups, then in the summing unit 5, the first check symbol is obtained by summing the initial value of the summation variable with all interleaved repeated information symbols from the first group by a given module, and the check symbol

by summing the check symbol for the given module

with all interleaved repeated information symbols from the group

. The obtained verification symbols are stored in the memory unit 8. The stored verification symbols are transmitted to the multiplexer 6, where a codeword is formed by adding verification symbols to the information symbols. Through the communication channel 9, the generated codeword is transmitted to the receiving device.

On the receiving side, the code word generated in the transmitting device is input to the block for generating correlation responses and estimating the noise parameter 10, where the received correlation responses are generated for each of the transmitted code symbols and the noise parameter of the signal-plus-noise mixture is estimated, which is approximately equal to the standard deviation equivalent to Gaussian white noise, which approximates the interference in the channel, and then form soft decisions on the received code symbols using th noise parameter estimation mixtures signal plus noise. The resulting soft decisions are fed to the input of the soft decision separation unit 11, where the soft decisions about the adopted code symbols are divided into decisions about information characters and decisions about check characters. Soft decisions about information symbols from the soft decision separation unit 11 are fed to the input of the information symbol repetition block 13, where soft decisions about information symbols are repeated in accordance with the information symbol repetition pattern. Soft decisions about the test characters from the soft decision separation unit 11 are input to the test character repetition unit 12, where soft decisions about the test characters are repeated at least two times, with the exception of the last test character. For soft decisions about information symbols, interleaving is performed similar to interleaving on the transmitting side in the block of interleaving information symbols 14. Then, in the block for generating interleaved soft decisions 15, groups of interleaved soft decisions about information symbols similar to the groups of corresponding information symbols on the transmitting side are formed using signals from the block of verification characters 12 and from the block for generating correlation responses and estimating the noise parameter 10, the first group of interleaved soft solutions are complemented by the soft solution corresponding to the initial value of the summation variable and the first copy of the soft decision on the first check character, forming the first codeword of the generalized parity check code, the nth group of interleaved soft solutions is supplemented by the second copy of the nth check character and the first copy soft decision about the nth check character, forming the nth code word of the generalized parity check code. The generated interleaved soft decisions are fed to the first input of the iterative decoder 16, in which a decoded codeword is generated. Each decoded codeword in the checksum calculation unit in the decoded codeword 17 is multiplied by a check matrix, obtaining an ordered sequence of checksums, if all the checksums are equal to zero, then the code word is considered decoded if the number of non-zero checksums is odd and the last check the sum is nonzero, then replace it with zero, if the number of checksums other than zero is odd and the last checksum in the sequence is zero, then for enyayut any nonzero value, divide the sequence of checksums in pairs, which include elements of a sequence, consecutive, each checksum includes only a single pair, each pair calculated difference numbers checksums are folded all the obtained difference numbers checksums. Then, in the error control block in the decoded codeword 18, the obtained number is compared with a predetermined threshold; if the number exceeds a predetermined threshold, then the information part of the codeword is considered decoded incorrectly; if the received number is less than the specified threshold, then the information part of the codeword is considered decoded correctly.

The objective of the proposed technical solution is the development of a transceiver device that allows you to increase the energy gain of noise-resistant coding through the use of a hybrid request-repeat algorithm.

To solve the problem, a data transmission device based on codes with a low density of parity checks, consisting of transmitting and receiving devices connected via a communication channel that provides data transmission, while the transmitting device contains series-connected information symbol repetition unit, information symbol interleaver , a unit for separating test characters into groups, a summing unit, as well as a series-connected memory unit and a multiplexer, the output of which connected to the communication channel, in addition, the outputs of the control unit are connected to the control inputs of the information symbol repetition unit, information symbol interleaver, group for separating test characters into groups, summing unit and multiplexer, respectively; the input of the information symbol repetition unit is the first input of the transmitting device to which the data packets arrive, the same input is connected to the second input of the multiplexer; the input of the control unit is the input of the synchronization signals, as well as the second input of the transmitting device; the receiving device contains a series-connected unit for generating correlation responses and estimating a noise parameter, a soft decision separation unit, an information symbol repetition unit, information symbol interleaving unit, an interleaved soft decision generation unit, a first iterative decoder, a first checksum calculation unit in a decoded codeword, a first an error control unit in the decoded codeword, the first output of which is the first output of the device, and the second output is connected to the first input of the first iterative decoder, in addition, the second output of the soft decision separation unit through the test symbol repetition unit is connected to the second input of the interleaved soft decision generation unit, the third input of which is connected to the second output of the correlation response generation unit and noise parameter estimation, the input of which is the input the receiving device, according to the invention, a feedback channel is introduced, into the transmitting device is an inter-packet summing unit, the control input of which is connected it is connected to the corresponding output of the control unit, the first and second inputs of the inter-packet summing block are connected to the corresponding outputs of the memory block, the output of the inter-packet summing block is connected to the third input of the multiplexer; the output of the summing unit is connected to the input of the memory unit, the control input of which is connected to the corresponding output of the control unit, the second input of the control unit, which is the input of the feedback signal from the receiving device transmitted through the output of the feedback channel and the third input of the transmitting device; to the receiving device, a soft decision storage and modification unit, the first output-input of which is connected through a second iterative decoder, the second checksum calculation unit in the decoded codeword and the second error control unit in the decoded codeword, connected to one input of the feedback channel, the second output of the second error control unit in the decoded codeword is connected to the second input of the second iterative decoder, and the third output of the second error control unit in the decoded code the word is the second output of the receiving device, while the output of the interleaved soft decision forming unit is connected to the input of the soft decision storage and modification unit, the second output-input of which is connected to the input-output of the first iterative decoder; the third output of the first error control unit in the decoded codeword is connected to another input of the feedback channel.

Distinctive (new) features of the claimed data transmission device regarding the prototype are the following:

- a feedback channel, through which feedback signals for the transmitting device from the first and second error control unit in the decoded word are transmitted, if the information part of the code word is decoded incorrectly, a retransmission request signal is generated on the feedback channel for the transmitting side;

- in the transmitting device: an inter-packet summing block, where on the transmitting side, upon receipt of the retransmission request signal, a packet with a code-conjugate part is formed, for which, in the next encoded packet, a part of the code symbols is replaced by a module sum of two with code symbols of a previously transmitted and erroneously decoded packet , thereby forming a code-conjugate part;

- in the receiving device:

a soft decision storage and modification unit, where soft decisions are modified in the code-conjugate part at those positions in which the code symbols are replaced by the sum of the symbols of different packets, for which the sign is replaced by the opposite for the adopted soft decision, if the recovered code symbol of the current packet corresponds to a logical unit, getting a modified sequence of soft decisions for an erroneous package;

- the second iterative decoder, where the erroneous packet is re-decoded using a modified sequence of soft decisions for the erroneous packet as additional information at the positions in which the code symbols were replaced by the sum of the symbols of different packets, signals are transmitted through the feedback channel characterizing the decoding performance of the code-conjugate packets ;

- the second checksum calculation unit in the decoded codeword, where each decoded codeword is multiplied by a check matrix, obtaining an ordered sequence of checksums, if all the checksums are zero, then the code word is considered decoded if the number of non-zero checksums is odd and if the last checksum is non-zero, then replace it with zero if the number of checksums other than zero is odd and the last checksum in the sequence is zero, then it is replaced with any non-zero value;

- the second error control block in the decoded codeword, where the sequence of checksums is divided into pairs that include sequence elements in a row, and each checksum is included in only one pair, in each pair the difference of the numbers of checksums is calculated, and all the differences obtained are added numbers of checksums, compare the received number with a given threshold, if the number exceeds the set threshold, then the information part of the code word is considered decoded incorrectly, if half If the number is less than the specified threshold, then the information part of the codeword is considered decoded correctly.

Graphic materials used in the description:

FIG. 1 - graph decoding algorithm: messaging;

FIG. 2 - Peterson's graph and the code set on it according to the Tanner method;

FIG. 3 is a diagram of a prototype transmitting device;

FIG. 4 is a diagram of a prototype receiving device;

FIG. 5 is a diagram of the transmitting side of the proposed device;

FIG. 6 is a diagram of the receiving side of the proposed device;

FIG. 7 is a method for detecting errors without CRC;

FIG. 8 is an example of an accumulation interleaving code;

и

.FIG. 9 - simulation results: the dependence of the bit error on the signal-to-noise ratio for

and

.

The inventive device comprises: a transmitting device (transmitting side), the circuit of which is shown in FIG. 5, the receiving device (receiving side) and the feedback channel - in FIG. 6.

In FIG. 5 the following notation is introduced:

1 - block repetition of information symbols;

2 - block interleaving information symbols;

3 - block separation of test characters into groups;

4 - control unit;

5 - block summation;

6 - multiplexer;

7 - block interpacket summation;

8 - memory block;

9 - communication channel.

The transmitter shown in FIG. 5, contains a series of information symbol repetition unit 1, an information symbol interleaver 2, a check symbol separation unit 3, a summing unit 5, a memory unit 8, an inter-packet summing unit 7, and a multiplexer 6, the output of which is connected to the communication channel 9. In addition to Moreover, the outputs of the control unit 4 are connected to the control inputs of the blocks 1, 2, 3, 5, 6, 7, 8, respectively.

The second output of the memory unit 8 is connected to the third inputs of the inter-packet summing unit 7 and the multiplexer 6.

The first input of the information symbol repetition unit 1 is the first input of the transmitting device to which the data packets arrive, the same input is connected to the first input of the multiplexer 6. The first input of the control unit 4 is the input of the feedback signal to automatically request repetition, it is the second input of the transmitting devices.

The signals from the outputs of the control unit 4 are control signals, which, inter alia, provide synchronization of the remaining blocks.

Through communication channel 9, the signal is supplied to the receiving device.

In FIG. 6 is indicated:

10 is a block for the formation of correlation responses and estimates of the noise parameter;

11 - block separation soft solutions;

12 is a block repetition of test characters;

13 - block repetition of information symbols;

14 - block interleaving information symbols;

15 - block forming interleaved soft decisions;

16, 20 - the first and second iterative decoders;

17, 21 - the first and second blocks for calculating checksums in the decoded codeword;

18, 22 - the first and second blocks of error control in the decoded codeword;

19 - block storage and modification of soft solutions;

23 - feedback channels

The receiver of FIG. 4, contains in series a block for generating correlation responses and estimating a noise parameter 10, a soft decision separation unit 11, a check symbol repetition unit 12, an interleaved soft decision generating unit 15, a first iterative decoder 16, a first checksum calculation unit in the decoded codeword 17, the first error control unit in the decoded codeword 18, the output of which is the first output of the receiving device. The output of the soft decision separation unit 11 through series-connected information symbol repetition unit 13 and the information symbol interleaver 14 is connected to the second input of the interleaved soft decision forming unit 15, the output of which is connected to the input of the soft decision storage and modification unit 19, the first input-output of which the second iterative decoder 20, the second checksum calculation unit in the decoded codeword 21, and the second error control unit in the decoded code c catch 22 is connected to the input of the feedback channel 23. In this case, the second output of the error control block in the decoded codeword 22 is connected to the second input of the second iterative decoder 20. The second input-output of the storage unit and the modification of soft decisions 19 is connected to the input-output of the first iterative decoder 16, the second input of which is connected to the third output of the error control unit in the decoded codeword 18, the second output of which is connected to the second input of the feedback channel 23. The second output of the correlation response generation unit s and a noise estimation parameter 10 is connected to the third input of the block forming the interleaved soft decisions 15.

The input of the receiving device is the input of the unit for generating correlation responses and estimating the noise parameter 10, which receives a signal from the transmitting device through the communication channel 9. The second output of the error control unit in the decoded codeword 22 is the second output of the receiving device.

The first 16 and second 20 iterative decoders are identical and executed according to the similar scheme of the iterative decoder of the prototype device with the only difference: an input-output is added that allows you to save restored soft solutions in the storage and modification block of soft solutions 19 (Fig. 6), use modified soft solutions from block storage and modification of soft decisions 19 (Fig. 6) as a priori information, summing it with the restored soft decisions at this iteration.

The device operates as follows.

путем суммирования по заданному модулю проверочного символа

со всеми перемеженными повторенными информационными символами из группы

. Полученные проверочные символы запоминают в блоке памяти 8. Запомненные проверочные символы поступают в мультиплексор 6, где формируют кодовое слово путем добавления проверочных символов к информационным символам. Через канал связи 9 сформированное кодовое слово передают на приемное устройство. В случае получения по каналу обратной связи сигнала запроса на повторную передачу формируют пакет с кодово-сопряженной частью, для чего в блоке межпакетного суммирования 7 суммируют кодовые символы текущего пакета с кодовыми символами ошибочно декодированного пакета, сохраненного в блоке памяти 8. В мультиплексоре 6 мультиплексируют кодово-сопряженную часть пакета, полученную с блока межпакетного суммирования 7 и кодовые символы текущего пакета, которые также сохраняют в блоке памяти 8.On the transmitting side, the generated data packets, each of which contains an ordered sequence of information symbols, receive the input of the information symbol repetition unit 1 and simultaneously to the first input of the multiplexer 6. In the information symbol repetition unit 1, each information symbol is repeated a certain number of times, depending on its number in order in the sequence of characters, forming a sequence of repeated information symbols, moreover, the number of repetitions is determined in advance e. The sequence of repeated information symbols from the output of the information symbol repetition unit 1 is fed to the input of the information symbol interleaving unit 2, where interleaving is performed in the sequence of repeated information symbols, achieving sufficient diversity of successive repeated information symbols and forming a sequence of interleaved repeated information symbols. In the block for dividing the check symbols into groups 3, interleaved repeated information symbols are divided into groups, then in the summing unit 5, the first check symbol is obtained by summing the initial value of the summation variable with all interleaved repeated information symbols from the first group by a given module, and the check symbol

by summing the check symbol for the given module

with all interleaved repeated information symbols from the group

. The obtained verification symbols are stored in the memory unit 8. The stored verification symbols are transmitted to the multiplexer 6, where a codeword is formed by adding verification symbols to the information symbols. Through the communication channel 9, the generated codeword is transmitted to the receiving device. If a retransmission request signal is received via the feedback channel, a packet with a code-conjugate part is formed, for which the code symbols of the current packet with the code symbols of the erroneously decoded packet stored in the memory block 8 are summed in the inter-packet summing block 7, and the code multiplexer 6 is multiplexed - the conjugate part of the packet received from the inter-packet summing block 7 and the code symbols of the current packet, which are also stored in the memory block 8.

On the receiving side, the code word generated in the transmitter is input to the block for generating correlation responses and estimating the noise parameter 10, where the received correlation responses to each of the transmitted code symbols are generated and the noise parameter of the signal plus noise mixture is estimated, which is approximately equal to the standard deviation of the equivalent Gaussian white noise, which approximates the interference in the channel, and then form soft decisions about the received code symbols using the resulting estimate the smart parameter of the signal plus noise mixture. The resulting soft decisions are fed to the input of the soft decision separation unit 11, where the soft decisions about the adopted code symbols are divided into decisions about information characters and decisions about check characters. Soft decisions about information symbols from the soft decision separation unit 11 are fed to the input of the information symbol repetition block 13, where soft decisions about information symbols are repeated in accordance with the information symbol repetition pattern. Soft decisions about the test characters from the soft decision separation unit 11 are input to the test character repetition unit 12, where soft decisions about the test characters are repeated at least two times, with the exception of the last check character. For soft decisions about information symbols, interleaving is performed similar to interleaving on the transmitting side in the block of interleaving information symbols 14. Then, in the block for generating interleaved soft decisions 15, groups of interleaved soft decisions about information symbols similar to the groups of corresponding information symbols on the transmitting side are formed using signals with a block of check symbols 12 and from a block for generating correlation responses and estimating a noise parameter 10, the first group being interleaved soft solutions are complemented by the soft solution corresponding to the initial value of the summation variable and the first copy of the soft decision on the first check character, forming the first codeword of the generalized parity check code, the nth group of interleaved soft solutions is supplemented by the second copy of the nth check character and the first copy soft decision about the nth check character, forming the nth code word of the generalized parity check code. The generated interleaved soft decisions are sent to the first input of the first iterative decoder 16 (to the first switch 30 of Fig. 5), in which the process of exchanging outgoing and incoming messages is iteratively such that the outgoing message of the code word of the generalized repetition code is an incoming message of the code word of the generalized code parity checks, and the outgoing message of the generalized parity check code is an incoming message of the code word of the generalized repetition code forming at each iteration The sequence of reduced soft decisions, and the sequence of the reconstructed soft decisions forming the decoded code words.

Each decoded codeword in the checksum calculation unit in the decoded codeword 17 is multiplied by a check matrix, obtaining an ordered sequence of checksums, if all the checksums are zero, then the code word is considered decoded if the number of non-zero checksums is odd and the last check the sum is nonzero, then replace it with zero, if the number of checksums other than zero is odd and the last checksum in the sequence is zero, then for enyayut any nonzero value, divide the sequence of checksums in pairs, which include elements of a sequence, consecutive, each checksum includes only a single pair, each pair calculated difference numbers checksums are folded all the obtained difference numbers checksums. Then, in the first error control block in the decoded codeword 18, the received number is compared with a predetermined threshold; if the number exceeds a set threshold, then the information part of the code word is considered decoded incorrectly; if the received number is less than the specified threshold, then the information part of the code word is considered decoded correctly.

The reconstructed check symbols of the packet in which the error was detected are stored in the soft decision storage and modification block 19, where they change the sign of the soft decision to the opposite if the recovered soft decisions are in the positions in which the code symbols were replaced by the sum of the symbols of different packets in the code conjugate parts of the packet correspond to a logical unit, obtaining soft solutions of the current packet, which are used as additional information when decoding the current packet by the first iterative decoder 16. Restore The updated sequence of soft decisions of the current packet is stored in the storage and modification block of soft decisions 19, where the soft decision sign is reversed if the restored soft decisions of the current packet are in the positions at which code symbols were replaced by the sum of the symbols of different packets in the code-conjugate part of the packet correspond to a logical unit, obtaining a modified sequence of soft solutions for an erroneous package. The error packet is re-decoded by the second iterative decoder 20, using the modified sequence of soft solutions for the error packet, as additional information at those positions in which the code symbols were replaced by the sum of the symbols of different packets. Transmit on the feedback channel 23 signals characterizing the decoding performance of code-conjugate packets.

The error detection algorithm in the codeword is shown in FIG. 7.

Error detection in the inventive method of transmitting data in a digital communication system can be explained by the following example. Suppose we have a “repeat with accumulation” code given by a bipartite regular graph. Consider the various error configurations. Note the case where erroneously decoded vertex variables exist, but all the checksums are equal to zero. This is a case of an undetectable error. Consider also the case where errors are only in the verification part of the codeword. Note that each error in the verification part causes two neighboring checksums to be zero. However, non-zero checksums with an error in the information part are spaced a distance proportional to the spacing parameter of the interleaver (interleaver). Thus, in the presence of a single error, it is easy to determine which symbol it refers to: informational or verification. To do this, it suffices to consider the difference in the numbers of test vertices corresponding to non-zero checksums. If such a difference is 1, then the erroneous character can be unambiguously attributed to the test characters, if more, then the erroneous character is informational.

Turning to the case of many errors, the algorithm is as follows:

• if the number of non-zero checksums is odd, then the control value of the last sum is changed, achieving an even number of checksums;

• consistently break the checksums into pairs, calculate the sum of the differences of the addresses of the test vertices;

• compare the obtained value with the threshold, if it is less than the threshold, then refer errors only to check characters.

Thus, error detection is carried out without CRC, thereby improving the quality of radio communications by reducing the number of redundant transmitted code symbols, while simultaneously detecting and correcting errors.

All units included in the communication system can be implemented on the processor, FPGA, for example, manufactured by Altera and Xilinx, as well as on a specialized custom microcircuit.

. На Фиг. 9 представлены результаты моделирования: зависимость битовой ошибки от отношения сигнал-шум для

для

. Для сравнения, помимо прототипа и изобретения приведены результаты моделирования аналога данного изобретения по источнику [Halford, Thomas & Bayram, Metin & Kose, Cenk & M. Chugg, Keith & Polydoros, Andreas. (2008). The F-LDPC Family: High-Performance Flexible Modern Codes for Flexible Radio. 376 - 380. 10.1109/ISSSTA.2008.75.]. Моделирование устройства проводили со следующими параметрами: число информационных символов

, шаблон повторения информационных символов

размер перемежителя

, первые

символа перемежителя

, а остальные получают как

, где

, кодово-сопряженную часть составляет каждый 9-ый проверочный символ.Thus, the technical result consists in obtaining an energy gain of encoding of an average of 0.5 dB, which is confirmed by the simulation results presented in FIG. 9, where the indicated gain is obtained at

. In FIG. Figure 9 presents the simulation results: the dependence of the bit error on the signal-to-noise ratio for

for

. For comparison, in addition to the prototype and the invention, the results of modeling an analog of the present invention according to the source [Halford, Thomas & Bayram, Metin & Kose, Cenk & M. Chugg, Keith & Polydoros, Andreas. (2008). The F-LDPC Family: High-Performance Flexible Modern Codes for Flexible Radio. 376 - 380. 10.1109 / ISSSTA.2008.75.]. The device was simulated with the following parameters: number of information symbols

, information symbol repeat pattern

interleaver size

first

interleaver character

and the rest get how

where

, the code-conjugate part is every 9th verification character.

Claims (2)

A data transmission device consisting of a transmitting and receiving device connected via a communication channel for transmitting data, wherein the transmitting device comprises series-connected information symbol repetition unit, information symbol interleaver, a unit for separating test characters into groups, a summing unit, and also in series connected memory unit and multiplexer, the output of which is connected to the communication channel, in addition, the outputs of the control unit are connected to control inputs odes of the information symbol repetition unit, information symbol interleaving unit, the unit for separating test characters into groups, the summing unit and the multiplexer, respectively; the input of the information symbol repetition unit is the first input of the transmitting device to which the data packets arrive, the same input is connected to the second input of the multiplexer; the input of the control unit is the input of the synchronization signals, as well as the second input of the transmitting device, the receiving device contains serially connected unit for generating correlation responses and estimating a noise parameter, a soft decision separation unit, an information symbol repetition unit, an information symbol interleaving unit, an interleaved soft decision generating unit, first iterative decoder, first checksum calculation unit in the decoded codeword, first decode error control unit The specified code word, the first output of which is the first output of the device, and the second output is connected to the second input of the first iterative decoder, in addition, the second output of the soft decision separation unit through the check symbol repetition unit is connected to the second input of the interleaved soft decision forming unit, the third input of which connected to the second output of the block for generating correlation responses and estimating a noise parameter, the input of which is the input of the receiving device, characterized in that the channels are introduced l feedback, to the transmitting device - the inter-packet summing block, the control input of which is connected to the corresponding output of the control block, the first and second inputs of the inter-packet summing block are connected to the corresponding outputs of the memory block, the output of the inter-packet summing block is connected to the third input of the multiplexer; the output of the summing unit is connected to the input of the memory unit, the control input of which is connected to the corresponding output of the control unit, the second input of the control unit is the input of the feedback signal from the receiving device transmitted through the output of the feedback channel and the third input of the transmitting device; to the receiving device, a soft decision storage and modification unit, the first output-input of which is connected through a second iterative decoder, the second checksum calculation unit in the decoded codeword and the second error control unit in the decoded codeword are connected to one feedback channel input, the second the output of the second error control unit in the decoded codeword is connected to the second input of the second iterative decoder, and the third output of the second error control unit in the decoded codeword fishing is the second output of the receiver, the output of the interleaved block forming soft decisions connected to the input of the storage unit and modifying the soft decisions, the second output-input of which is connected to the input-output of the first iterative decoder; the third output of the first error control unit in the decoded codeword is connected to another input of the feedback channel.

Images