CN103888148B - A kind of LDPC code Hard decision decoding method of dynamic threshold bit reversal - Google Patents

Abstract

The invention discloses a kind of LDPC code Hard decision decoding method of dynamic threshold bit reversal, determine an initial turn threshold T, use dynamic threshold, namely dynamic turn threshold, turn threshold is become dynamic threshold and can reduce the probability of upset by mistake, decreasing the iterations of decoding, the bit-flipping decoding making LDPC code is more efficient；Also improving the error correcting capability of LDPC code, the present invention has strengthened than the error correcting capability of the bit reversal method of standard simultaneously.

Description

A kind of LDPC code Hard decision decoding method of dynamic threshold bit reversal

Technical field

The present invention relates to a kind of LDPC (Low Density Parity Check, low density parity check code) hard decision translate The bit reversal building method of code device.

Background technology

In the various application needing to carry out signal transmission, often can use error correcting code (ECC, Error Correcting Code), when can make signal transmission errors, righted the wrong at receiving terminal and obtained correct signal.

Error correcting code can apply in many systems, in a communications system, and may be by channel effect during signal transmission And the interference of interchannel noise, thus cause the data stored in flash memory device the most incorrect.In flash memory device The data stored are the data after error correcting code device code, and for flash memory storage controls device, error correcting code is must The functional unit needed.

Along with the technique of memorizer is increasingly advanced, memory cell volume is more and more less, and memory element is stored Data the most gradually increase, cause flash memories read during produce error probability constantly raise.Therefore, flash memory The error correcting code error correcting capability of storage control device is to determine the key factor that storage control device is the most qualified.But, Flash memory storage controls to use stronger error correcting code to need higher operand and relatively Long operation time in device, thus increases flash memory The application of storage control device is very restricted.

LDPC code is that Robert Gallager has sparse check matrix in the one that 1962 propose in thesis for the doctorate Packeting error-correcting code.Being suitable for almost all of channel, its performance approaches shannon limit, and describes and realize simple, decoding letter Single and practicable parallel work-flow, is suitable for hardware and realizes.

LDPC code has huge application potential, at deep space communication, fiber optic communication, satellite communication, satellite digital video, number Word watermark, magnetic optical/Hologram Storage, movement and fixed radio communication, cable modulating/demodulating and digital subscriber obtain extensively should With.

Technique according to flash memory device is increasingly advanced, and flash memories controls the error correction energy of the error correcting code in device Power is also required to strengthen.In current flash memories controls device, main error correcting code is BCH code, along with the increasing of error probability Height, BCH code space requirement and operational capability are also gradually increased；Along with the raising of flash memory storage technique, the error correction energy of BCH code Power is the most gradually not suitable for the development of flash technology, so needing the higher error correcting code of error correcting capability.LDPC code is selected to replace BCH code is that comparison is appropriate.

In the Hard decision decoding device of LDPC code, the setting of bit reversal threshold value T, rises the Hard decision decoding of LDPC code Vital effect, it affects error correcting capability and its decoding rate of LDPC code.

If what bit reversal threshold value T set crosses conference increases the iterations of decoding, thus affects decoding rate.

If the too small meeting that bit reversal threshold value T sets causes upset by mistake thus to increase the iterations of decoding, Thus affect decoding rate.

For irregular check matrix, choose fixing bit reversal threshold value T, the biggest mistake can be caused like that Overturn probability thus considerably increase the iterations of decoding, cause decoding failure in severe cases.

When in sequence Z, the error rate of code element reaches certain value, fixing bit reversal threshold value T is used to decode, No matter decoder iteration is big all without successfully decoding often.

Summary of the invention

Based on dynamic threshold, the LDPC code hard decision that it is an object of the invention to provide a kind of dynamic threshold bit reversal is translated Code method, improves decoding rate and error correcting capability.

The present invention is by the following technical solutions:

A kind of LDPC code Hard decision decoding method of dynamic threshold bit reversal, determines an initial turn threshold T, bag Include following steps:

1) read codeword sequence, calculate syndrome according to codeword sequence；

2) judge whether syndrome is zero, if zero, then stop iteration, output codons sequence；Otherwise enter step 3)；

3) each symbol bits to code word each in codeword sequence, calculate the participation of this symbol bits is unsatisfactory for verification The number of equation；

4) turn threshold is calculated: along with increasing of iterations, turn threshold T successively decreases, and obtains current turn threshold；

5) if the number of ungratified check equations is more than current turn threshold, corresponding symbol bits overturns, And calculate the syndrome of codeword sequence corresponding after upset；

6) repeat step 1), 2), 3), 4) and 5), until successfully decoded, or when reaching maximum iteration time, output is translated Code failure.

The LDPC code Hard decision decoding method of above-mentioned dynamic threshold bit reversal, current turn threshold T:

T=j (b-c d)/a；

In formula: d represents current for which time iteration, j be corresponding to symbol bits when prostatitis row weight, a, b, c are experience Normal number.

The LDPC code Hard decision decoding method of above-mentioned dynamic threshold bit reversal, determines an initial turn threshold T, repeatedly In Dai, the step of upset is as follows:

A) according to hard decision sequence Z=[z₀, z₁..., z_N-1] calculate syndrome S=[s₀, s₁..., s_M-1]:

$s_m={\mathrm{\Σ}}_{n=0}^{N-1}{H}_{mn}{z_n}^T\mathrm{mod}2,m=0,1,...,M-1;$

Wherein H_mnTest matrix, if syndrome S=0, stops iteration, and output hard decision sequence Z also shows successfully decoded, Otherwise enter b)；

B) to each symbol bits z_n, calculate number f of its ungratified check equations participated in_n:

$f_n={\mathrm{\Σ}}_{m=0}^{M-1}{s}_m{H}_{mn},n=0,1,...,N-1$

If f_n＞ T, then overturn z_n, obtain new hard decision sequence Z；

C) repeat a) and b) until successfully decoded, or reach maximum iteration time and show decoding failure；

Hard decision sequence is step 5) codeword sequence that obtains after upset；

Wherein M represents the line number of test matrix, and N represents the columns of test matrix.

According to the present invention, using dynamic threshold, namely dynamic turn threshold, turn threshold is become dynamic threshold can To reduce the probability of upset by mistake, decreasing the iterations of decoding, the bit-flipping decoding making LDPC code is more efficient；The most also Improve the error correcting capability of LDPC code, the present invention has strengthened than the error correcting capability of the bit reversal method of standard.

Accompanying drawing explanation

Fig. 1 is the theory diagram of flash memories memory control device.

Fig. 2 is LDPC code unit bit-flipping decoding flow chart.

Detailed description of the invention

LDPC code is the one of linear block codes, and it has all of characteristic of linear block codes.LDPC code can be divided into rule The then big class hypothesis check matrix H of (regular-LDPC) and irregular (irregular-LDPC) two₀For m × n rank matrix, rule LDPC code can be denoted as (n, j, k), wherein n is code length, j be check matrix each column weight (i.e. row in 1 number, be called for short row Weight (column weight), k is the weight (at once, the number of 1, is called for short row weight (row weight)) that check matrix is often gone, And typically have j > 2, k > j and the number of the 1 of the check matrix each row and column of irregular LDPC codes is incomplete same.

The iterative decoding method of LDPC code is broadly divided into two kinds: one is hard-decision method, and one is soft decision method. What hard-decision bits method for turning transmitted in an iterative process is binary system hard information, and soft decision method passes in an iterative process Pass is the real number Soft Inform ation with probability correlation.Hard-decision method is simple to operate, it is easy to hardware realizes, but error-correcting performance one As；Soft decision method better performances, but implementation complexity is higher.This programme improves mainly for Hard decision decoding device and carries Its error correction energy ability high.The detailed process of the hard-decision bits upset interpretation method of LDPC code is introduced as follows.

Bit reversal method is often being taken turns in iteration, first calculates the value of syndrome according to last round of hard decision sequence.As The most all of syndrome is 0, then stop iteration, and show successfully decoded, otherwise calculates the syndrome that each bit participates in It is the number of the check equations of 1, participation is unsatisfactory for check equations number and turns over more than the bit of certain predetermined threshold threshold value T Turn, obtain a new hard decision sequence, subsequently into next round iteration, until successfully decoded or reach maximum iteration time And show decoding failure.Properly select the size of threshold value T, can reach optimal decoding performance.

Hard decision sequence is the sequence of the needs decoding that decoder receives sequence after bit reversal method overturns.

Fig. 1 is the functional block diagram after an embodiment simplification of flash memory storage control device.In flash memory storage controller Comprise LDPC code decoder.Flash memory storage controller is mainly responsible for the read-write of data and the storage of data and other function.

Flash memory storage controller obtains data from main frame and carries out encoding operation through LDPC code decoder and the data that produce Store in flash memories.Flash memory storage controller is needed to come from sudden strain of a muscle if main frame goes for the data in flash memories Deposit in memorizer and read out, be decoded computing through LDPC code decoder and produce data and input to main frame.

Fig. 2 is dynamic threshold (dynamic threshold, turn threshold, dynamic turn threshold and bit reversal threshold value in literary composition Representing that same threshold value uses the title under environment in difference, those skilled in the art is easy to understand) LDPC of bit reversal Code coding method flow chart.Obtaining codeword sequence by flash memory storage controller from flash memories, ldpc code decoder reads This codeword sequence.The whole decoding process of LDPC code as can be seen from Figure 2.Determine an initial turn threshold T, thus its process As follows:

(1), after reading codeword sequence, syndrome is calculated according to codeword sequence；

(2) judging whether syndrome S is zero, stop iteration if zero, output codons sequence also shows successfully decoded, Otherwise enter (3)；

(3) each symbol bits to each code word in each codeword sequence, calculates being unsatisfactory for of its participation The number of check equations；

(4) mode successively decreased is used to determine new turn threshold；

(5) if ungratified check equations number is more than new turn threshold, corresponding code word overturns；

(6) repeat (1), (2), (3), (4) and (5) until successfully decoded, or reach maximum iteration time and display is translated Code failure.

Code element: usually represent a binary digit with the symbol that time interval is identical in digital communication, such Signal in time interval is referred to as (binary system) code element.And this interval is referred to as Baud Length.Code word (code word) is then It is made up of some code elements.

It is the Essential Terms of this area about used term, it is not explained in detail, such as syndrome, according to communication (n, k) syndrome of code, if the error correcting capability of this yard is t, then weight is little for the general explanation of principle, i.e. Linear codes In or have unique syndrome (syndrome) corresponding therewith equal to all error patterns of t.

About turn threshold, using given formula to successively decrease, amount of calculation is smaller, but representative bad, it is impossible to have Effect represents the present situation of decoding.

On the basis of the method for the bit-flipping decoding of LDPC code, the LDPC code of a kind of dynamic threshold bit reversal of proposition Interpretation method.Principle is, LDPC code is when carrying out bit-flipping decoding, and its turn threshold is a fixed value, is decoding Journey easily causes the upset of mistake, thus the iterations adding decoder makes decoding time long.In order to more effectively Utilizing the bit-flipping decoding of LDPC code, this case proposes the Hard decision decoding side of the LDPC code of a kind of dynamic threshold bit reversal Method.Idiographic flow is as shown in Figure 2.

In the Hard decision decoding device of LDPC code, the setting of bit reversal threshold value T, rises the Hard decision decoding of LDPC code Vital effect, it affects error correcting capability and its decoding rate of LDPC code.

If what bit reversal threshold value T set crosses conference increases the iterations of decoding, thus affects decoding rate.

If the too small meeting that bit reversal threshold value T sets causes upset by mistake thus to increase the iterations of decoding, Thus affect decoding rate.

For irregular check matrix, choose fixing bit reversal threshold value T, the biggest mistake can be caused like that Overturn probability thus considerably increase the iterations of decoding, cause decoding failure in severe cases.

When in sequence Z, the error rate of code element reaches certain value, fixing bit reversal threshold value T is used to decode, No matter decoder iteration is big all without successfully decoding often.

The Hard decision decoding method of the LDPC code of a kind of dynamic threshold bit reversal that this programme proposes can effectively be kept away Exempt from above-mentioned problem.The interpretation method of this patent can reduce the iterations of decoding, thus adds decoding rate；Can also carry High error correcting capability, can entangle more mistake in sequence Z.

The formula of dynamic threshold: T=j (b-c d)/a.Wherein T is threshold value, and which time iteration d is, j is row weight, a, b, c For experience normal number (mistake that constant entangles with the technique of flash memories and needs is relevant).

Wherein a, b, c are empirical, for constant, do not affect dynamic threshold " dynamically " characteristic in formula, and as initial Threshold value is i.e. regarded as the fixed threshold under normal condition, the fixed threshold that those skilled in the art determines therefrom that, by with The dependency of row weight and iterations obtains the dynamic threshold directly related with iterations and row weight, it is thus possible to effectively drop Low amount of calculation, improves decoding efficiency.Owing to experience Changshu is on " dynamically " not impact, therefore, do not repeat them here.

The row of dynamic threshold T and check matrix weigh j, which time number of times of iteration decoder is, empirical is closely bound up. The row of check matrix are heavily the ultimate values of dynamic threshold, and decoder iteration number of times more Larger Dynamic threshold value T change space is the least, experience Constant is needed the erroneous decision corrected by flash memories technique and sequence Z.The iteration of decoder each time, the code in sequence Z Unit can carry out judgement according to the dynamic threshold of each code element and decide whether to overturn, and carries out multiple code element every time and turns over Turn.

Syndrome is check matrix H_mnFormula s is seen below with the result of sequence Z transposition product_m。

Bit reversal method specifically comprises the following steps that

A () is according to hard decision sequence Z=[z₀, z₁..., z_N-1] calculate syndrome S=[s₀, s₁..., s_M-1]:

$s_m={\mathrm{\Σ}}_{n=0}^{N-1}{H}_{mn}{z_n}^T\mathrm{mod}2,m=0,1,...,M-1$

Wherein H_mnTest matrix, if syndrome S=0, stops iteration, and output hard decision sequence Z also shows successfully decoded, Otherwise enter (b).

B () is to each symbol bits z_n, calculate number f of its ungratified check equations participated in_n:

$f_n={\mathrm{\Σ}}_{m=0}^{M-1}{s}_m{H}_{mn},n=0,1,...,N-1$

If f_n＞ T, then overturn z_n, obtain new hard decision sequence Z.

C () repeats (a) and (b) until successfully decoded, or reach maximum iteration time and show decoding failure；

Wherein M represents the line number of test matrix, and N represents the columns of test matrix.

It can be seen that bit reversal method is simple hard-decision method, it is only necessary to logical operations, it is achieved the simplest.

Claims (3)

1. a LDPC code Hard decision decoding method for dynamic threshold bit reversal, determines an initial turn threshold T, and it is special Levy and be, comprise the following steps:

1) read codeword sequence, calculate syndrome according to codeword sequence；

2) judge whether syndrome is zero, if zero, then stop iteration, output codons sequence；Otherwise enter step 3)；

3) each symbol bits to code word each in codeword sequence, calculate the participation of this symbol bits is unsatisfactory for check equations Number；

4) turn threshold is calculated: along with increasing of iterations, turn threshold T successively decreases, and obtains current turn threshold；

5) if the number of ungratified check equations is more than current turn threshold, corresponding symbol bits overturns, and counts The syndrome of codeword sequence corresponding after calculating upset；

6) repeat step 1), 2), 3), 4) and 5), until successfully decoded, or when reaching maximum iteration time, output decoding failure.

The LDPC code Hard decision decoding method of dynamic threshold bit reversal the most according to claim 1, it is characterised in that when Front turn threshold T:

；

In formula: d represents current for which time iteration, j be corresponding to symbol bits when prostatitis row weight, a, b, c are experience normal number.

The LDPC code Hard decision decoding method of dynamic threshold bit reversal the most according to claim 1 and 2, determines at the beginning of one The turn threshold T begun, it is characterised in that in iteration, the step of upset is as follows:

A) according to hard decision sequenceCalculate syndrome:

；

WhereinTest matrix, if syndrome, stopping iteration, output hard decision sequence Z also shows successfully decoded, no Then enter b)；

B) to each symbol bits, calculate the number of its ungratified check equations participated in:

If, then overturn, obtain new hard decision sequence Z；

C) repeat a) and b) until successfully decoded, or reach maximum iteration time and show decoding failure；

Hard decision sequence is the codeword sequence obtained after step 5) overturns；

Wherein M represents the line number of test matrix, and N represents the columns of test matrix.