DE4136091A1 - Synchronising code words of cyclic, non-binary, corrected code - dividing symbol sequence to be decoded into code words with determined number of symbols - Google Patents

Abstract

The symbol sequence to be decoded is divided into code words, consisting of a number of successive symbols, determd. by the used code. For each code word, or some of them, a fault pattern is determd. in which a number value represents faulty code word symbols, while another number value represents fault-free symbols. Each fault pattern is linked to a reference pattern of a series of preset such patterns, each reference pattern precisely corresp. to the fault pattern formed with a certain synchronisation fault appearance. The link is such that its result value represents a measuring number for approximation between the fault pattern and each reference pattern. USE/ADVANTAGE - For digital transmission systems, without need of extra synchronisation signals, but using only fault correction properties of cyclic code.

Description

The present invention relates to a method for Synchronization of the code words of a cyclic, non-binary, error correcting codes.

In the field of digital transmission technology is one increasing importance of the so-called FEC (forward error correction) to determine coding method. With these The process becomes a serial symbol stream to be transmitted initially in data words consisting of a certain number K divided by successive symbols. Any of these Data words are made according to certain assignment rules (Coding rules) converted into a so-called code word, which consists of N symbols. Here N is greater than K. The code word sequence thus formed is called a serial symbol sequence transfer. At the reception end there is now the task of serial symbol sequence again in the same sequence of Codewords as on the transmission side of N symbols each to divide. This task is commonly called Code word synchronization. Only with the right one Codeword synchronization is using appropriate decoding techniques an error-correcting image of the received Code word sequence in the data word sequence or the original serial symbol sequence possible.

In (1) S.Y. Tong, "Synchronization Recovery Techniques for Binary Cyclic Codes, "The Bell System Technical Journal, April 1966, pages 561 et seq. Describes how one can already be found in Transmission system for other synchronization tasks existing synchronization device, e.g. B. for a  Frame synchronization, also for code word synchronization is also used. This method has the disadvantage that a corresponding synchronization device already elsewhere must be present.

Further methods for the described in (1) Codeword synchronization is the additional transmission of a special synchronization word and the use of special codes (for example so-called "comma-free" codes), the detection and correction of synchronization errors allow.

These two methods have the disadvantage that in addition Redundancy only for recognizing or correcting Synchronization errors must be introduced. The the latter method also has the disadvantage that special codes must be used.

A synchronization process that without additional Synchronization information can be found in the US PS 46 80 765 (2). The method described there increases however, the synchronization time for more disturbed signals very strong. The synchronization circuit must also be connected to the selected code can be adjusted.

The present invention is based on the object Specify code word synchronization method that only the Exploits error correction properties of a cyclic code, d. H. no additional synchronization information required.

The solution to this problem is specified in claim 1. Advantageous further developments describe the subclaims.

Cyclic codes have the property that a cyclic  Shifting a permissible code word is again a permissible one Is codeword. Is so

= [x _n-1,. . ., x₁, x₀]

a permissible code word, then there is also the word

[x₀, x _n-1 ,. . ., x₂, x₁]

a valid code word. X _{i is} any symbol of the code word. Now let us say the two words and two successive code words of the data stream to be decoded:

. . . y₄, y₃, y₂, y₁, y₀, x₄, x₃, x₂, x₁, x₀. . .

Is caused by a synchronization error Synchronization frame around a symbol to the left, for example shifted, a word to be decoded is the word

y₀, x₄, x₃, x₂, x₁

When trying to decode this word, the decoder will normally find exactly one error because the symbol y ₀ does not belong to the code word. However, there is a residual probability that the symbol y _{0 is} identical to the symbol x ₀ . In such a case, the decoder cannot detect an error. If you mark a symbol recognized by the decoder as incorrect, for example with e _i = +1, and a symbol that is assumed to be correct with e _i = -1, you get the so-called error pattern for the above-mentioned (incorrectly synchronized) word:

 = [+1, -1, -1, -1, -1]

As long as the synchronization error persists this error pattern occurs frequently.

Accordingly, you get as an error pattern

 = [+1, +1, -1, -1, -1]

if the synchronization frame left by two symbols is moved. Based on the error pattern, not only recognize the state synchronously / non-synchronously, but also how many symbols and in which direction the Synchronization frame is shifted. The shift of the The synchronization frame may have a maximum of t symbols. Here t is the correctability of the code used, d. H. t is the maximum number of bad symbols per Codeword that are corrected for the code used can. If the synchronization frame is more than t symbols then contain the code words to be decoded usually more than t errors. As long as not too many mistakes are contained in the code word, the decoder can do this at least still recognize, but no more information about it provide which symbols are correct and which are wrong. Is the synchronization frame by more than t symbols postponed, so the state can still "out of sync" are recognized, but not by how many Icons the frame is moved.

The following is an example of the invention Synchronization procedure described in more detail.

The symbol sequence to be decoded is divided into code words, consisting of a number of successive symbols determined by the code used in each case. As described above, an error symbol e _{i is} assigned to each symbol of a code word. The entirety of the error symbols for a code word is referred to as an error pattern.

In practice, a shortened error pattern is evaluated, only from the t first and t last error symbols exists because the other error symbols are not direct Information regarding the status synchronous / non-synchronous deliver.

The possible due to synchronization errors Error patterns are known in advance. You are here as Reference pattern called. The error pattern of each code word is correlated with each of these reference patterns. Designated one with the error patterns

= [e _2t-1,. . ., e₁, e₀]

and the 2t + 1 reference pattern (including the Reference pattern ₀ for the synchronous case) with

_i = [r ⁱ _2t-1,. . ., r ⁱ ₁, r ⁱ ₀],

where i = -t, -t +1,. . ., + t, so the product

for all i = -t,. . ., + t formed. Be

ξ = [ξ _-t,. . ., ξ₀,. . ., ξ _{+ t} ],

then for a fixed or adaptively adapted number β of  evaluated error patterns the sum

d. H.

formed, where ν is the time numbering of individual Represents code words. The index of the maximum marked with s Total values then indicate in which direction and by how many Symbols of the synchronization frame must be moved:

In practice, the search for the maximum total value means that all total values have to be compared with one another. This comparison can be simplified somewhat. This exploits the fact that at least the maximum value σ _s (β) increases with an increasing number of evaluated error patterns. You are not looking for the maximum value of σ _i (β) but you are looking for the total value σ _i (ν), which is the first to reach a limit σ _lim that is either predefined or adaptively adjustable. If the value of this barrier is not set too low and the symbol stream to be decoded does not contain too many symbol errors, then the index of the sum value σ _i (ν) which exceeds the barrier clearly indicates in which direction and by how many symbols the synchronization frame is shifted got to:

s = [i | σ _i (ν) σ _lim ]

In the previous description of the invention Synchronization procedure was a maximum value in each case searched. With an appropriate definition of error patterns and A minimum value search can of course also be used for reference samples to be necessary.

In the examples, the values e _i = -1 and e _i = +1 were used as numerical values in the error pattern. Depending on the application, it can be advantageous if these numerical values are still weighted. In some receivers there is a drop of information that the received signal is being severely disturbed. In such a case, the numerical values can e.g. B. with 1/2 weights, so that the error patterns obtained during a strong disturbance are taken into account less than in the undisturbed case.

Claims (6)

1. Procedure for synchronizing the code words of a cyclic, non-binary error-correcting code, with the following procedure steps:

• a) the symbol sequence to be decoded is divided into code words consisting of a number of successive symbols determined by the code used in each case;
• b) an error pattern is determined for each code word or for some of the code words, in which symbols of the code word recognized as incorrect by the decoder are represented by a numerical value, symbols of the code word recognized as error-free are represented by another numerical value in the error pattern;
• c) each of these error patterns is linked to each reference pattern of a series of predetermined reference patterns, each of these reference patterns corresponding exactly to the error pattern that arises in the presence of a specific synchronization error, and the linkage is such that the result value of the linkage is a measure of the similarity between Error pattern and the respective reference pattern;
• d) the result values of the links corresponding to the same reference patterns are summed up over a predeterminable number of error patterns;
• e) the results of this summation for each reference pattern are compared with one another;
• f) a correction signal for the synchronization is derived from the sum result which differs the most from the other sum results.

2. The method according to claim 1, wherein the number of Error patterns over which the results of the links be added up, is fixed. 3. The method according to claim 1, wherein the number of Error patterns over which the results of the links be added up, depending on the frequency of errors the symbol sequence to be decoded is set in Way that the greater the frequency of errors, the greater the Number of evaluated error patterns. 4. The method according to claim 1, in which a limit value for the sum of the results of the links is predetermined and after reaching the set limit by one of the Totals derived a correction signal for synchronization becomes. 5. The method according to claim 4, wherein the limit value for the sums of the results of the links depending on the frequency of errors of the symbol sequence to be decoded is set in such a way that the larger the Frequency of errors, the greater the number of evaluated Is error patterns. 6. The method according to any one of the preceding claims, where the numerical values chosen for the error patterns for incorrect and error-free symbols depending on the Error frequency of the symbol sequence to be decoded set in such a way that the greater the frequency of errors, the less consideration of each Error pattern in the evaluation is.