Design of regular (2,d/sub c/)-LDPC codes over GF(q) using their binary images
read more
Citations
Efficient error-correcting codes in the short blocklength regime
Code Design for Short Blocks: A Survey.
Stochastic Decoding of LDPC Codes over GF(q)
Trellis-Based Extended Min-Sum Algorithm for Non-Binary LDPC Codes and its Hardware Structure
Non-Binary Protograph-Based LDPC Codes: Enumerators, Analysis, and Designs
References
Low-Density Parity-Check Codes
The Theory of Error-Correcting Codes
Good error-correcting codes based on very sparse matrices
Design of capacity-approaching irregular low-density parity-check codes
Related Papers (5)
Frequently Asked Questions (12)
Q2. What is the upper bound for a code with parameters?
since the upper bound in [2] for a code with parameters (Mss, Nss) does not provide an analytical expression of dmin as a function of ds,min, the authors apply the Elias upper bound for a code with parameters (Mss = (ds,min − 1).p,Nss = ds,min.p) [21]:dmin ≤ 2A.
Q3. What is the minimum symbol weight of a stopping set?
For all stopping sets with symbol weights ds ≥ ds,min, the equivalent binary matrix is no longer a square matrix: its binary representation is at most a (ds − 1)p × dsp rectangular matrix Hss.
Q4. how can the authors obtain an upper bound on the maximum achievable binary minimum distance for a minimal stopping?
Using the maximum achievable minimum distance given by [2] for a code with the preceding parameters (Mss, Nss), the authors can obtain an upper bound on the maximum achievable binary minimum distance for that minimal stopping set with weight ds,min.
Q5. What is the effect of the optimization on the error floor?
For high rate codes, the waterfall gain decreases with the rate for a given field order, but the optimization can improve drastically the error floor.
Q6. What is the relationship between the binary minimum distance and the topology of the Tanner graph?
Since the binary minimum distance is defined by the minimum number of independent columns of Hb, it is also strongly related to the topology of the Tanner graph associated with H , noted GH .
Q7. What is the minimum distance of a dc-tuple?
Since finding good dc-tuples can be computationally expensive, nextwe provide some guidelines to accelerate the search procedure of the primitive set of rows:• dmin(i) is the minimum number of columns of Hi that are dependent, thus the minimumdistance of a dc-tuple is at most the minimum distance associated with any two sub-matrices Hij and Hij′ of Hi.
Q8. What is the motivation of using the binary image of the LDPC code?
The motivation of using the binary image of the LDPC code is essentially that the authors address and illustrate the optimization process for the non zero values in the case of binary input additive white gaussian noise (BI-AWGN).
Q9. What is the effect of the waterfall gain for B and DM methods over R method?
As predicted by the theoretical thresholds, the waterfall gain for B and DM methods over R method vanishes when field order increases.
Q10. What is the way to ensure a good minimum distance for the whole code?
unlike for cycles, there is no way to “cancel” the influence of such stopping sets by proper symbol assignments: since each stopping set has a minimum distance associated with it, the only way to ensure a good minimum distance for the whole code is to try to maximize the minimum distance over all stopping sets (practically over the most exhaustive set of stopping sets the authors can enumerate).
Q11. How can The authordetermine the minimum distance of a stopping set?
Note that for a graph GH with minimum variable node degree dv = 2, it is quite simple to identify the set of stopping sets with minimum weight ds,min: this can be achieved in conjunction with the PEG construction by adding a procedure which tests if a group of nodes contains 3 imbricated cycles.
Q12. What is the shortest distance of the binary LDPC code?
This can be related to a previous result from [9], where it is shown that the minimum distance of the binary (2, dc)-regular LDPC codes can increase at most logarithmically with the codeword length N : this emphasizes the need for efficient methods to design codes with good minimum distance properties.