Classifications

• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
  • H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
  • H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
  • H03M13/13—Linear codes
  • H03M13/15—Cyclic codes, i.e. cyclic shifts of codewords produce other codewords, e.g. codes defined by a generator polynomial, Bose-Chaudhuri-Hocquenghem [BCH] codes
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
  • H03M13/65—Purpose and implementation aspects
  • H03M13/6502—Reduction of hardware complexity or efficient processing
  • H03M13/6505—Memory efficient implementations
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
  • H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
  • H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
  • H03M13/13—Linear codes
  • H03M13/15—Cyclic codes, i.e. cyclic shifts of codewords produce other codewords, e.g. codes defined by a generator polynomial, Bose-Chaudhuri-Hocquenghem [BCH] codes
  • H03M13/151—Cyclic codes, i.e. cyclic shifts of codewords produce other codewords, e.g. codes defined by a generator polynomial, Bose-Chaudhuri-Hocquenghem [BCH] codes using error location or error correction polynomials
  • H03M13/158—Finite field arithmetic processing
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
  • H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
  • H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
  • H03M13/13—Linear codes
  • H03M13/15—Cyclic codes, i.e. cyclic shifts of codewords produce other codewords, e.g. codes defined by a generator polynomial, Bose-Chaudhuri-Hocquenghem [BCH] codes
  • H03M13/159—Remainder calculation, e.g. for encoding and syndrome calculation
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
  • H03M13/65—Purpose and implementation aspects
  • H03M13/6569—Implementation on processors, e.g. DSPs, or software implementations

Definitions

• the present invention relates to coding data and, more particularly, to methods and apparatus for Reed-Solomon coding.
• the Galois field itself can be constructed as the set of all polynomials of degree ⁇ m with the coefficients from the "ground" binary field GF(2) whose elements are 0 and 1 and two field operations, addition and multiplication, defined as follows: addition is performed as an "exclusive OR" (XOR) operation on the ordered r ⁇ -tuple of polynomial coefficients and multiplication is implemented as polynomial multiplication modulo certain "primitive" binary polynomial of degree m as field multiplication, where the "primitive" binary polynomial is irreducible (can not be factored further into a product of binary polynomials of lower degree) and in addition, does not divide any binary polynomial of the form x d -l for d ⁇ 2 m -l .
• a polynomial G(x) of degree 2t known as a generator polynomial, is defined according to the following formula
• the remainder is a polynomial of degree 2t-l and finally, 2t coefficients of this polynomial are used as check symbols which are appended to k message symbols forming a string of k+2t symbols that comprise a "codeword" in systematic form.
• the length of the codeword should not exceed n-1, the number of non-zero elements in the GF. So the message actually transmitted over a communication channel can be represented as the polynomial
• r m check symbols - redundant symbols that allow detection and correction of up to t errors.
• the task of encoding consists, therefore, in performing the division of polynomials whose coefficients are symbols of some GF.
• a conventional software approach to deal with calculations in the GF is based on using two lookup tables where transformations from the polynomial form of the GF symbols to exponential form and back are stored. This method can be summarized (for our main case of GF(2 8 )) as follows: 1. Symbols (elements of the GF) in the polynomial form are represented as bytes:
• a discrete logarithm table logjable and an exponent table expjable are pre- calculated to facilitate transformations between exponential (power) and .
• the total consumption of MIPS for only the encoder operation is 120-144 MIPS, which is prohibitively high.
• a method for encoding data. The method comprises (a) providing a first table of first component values multiplied by code generator coefficients and a second table of second component values multiplied by the code generator coefficients, (b) determining a Galois field element based on a message symbol and a high order remainder symbol, (c) separating the Galois field element into first and second components, (d) using the first component and a generator coefficient index to access one or more first table values in the first table and using the second component and the generator coefficient index to access one or more second table values in the second table, (e) determining a current remainder symbol based on first and second table values having equal generator coefficient indices, and a previous remainder symbol, (f) performing step (e) for each generator coefficient of the code generator to provide a set of remainder symbols, and (g) performing steps (b) - (f) for each message symbol in a codeword to provide a set of final remainder symbols which constitute check symbols to be transmitted
• apparatus for encoding data.
• the apparatus comprises a first lookup table containing first component values multiplied by code generator coefficients, a second lookup table containing second component values multiplied by the code generator coefficients, and a processor.
• the processor comprises means for determining a Galois field element based on a message symbol and a high order remainder symbol, means for separating the Galois field element into first and second components, means for using the first component and a generator coefficient index to access one or more first table values in the first lookup table and for using the second component and the generator coefficient index to access one or more second table values in the second lookup table, and means for determining a current remainder symbol, based on first and second table values having equal generator coefficient indices, and a previous remainder symbol, for each generator coefficient of the code generator to provide a set of remainder symbols.
• Each message symbol of a codeword is processed to provide a set of final remainder symbols which constitute check symbols to be transmitted with the message symbols of the codeword.
• FIG. 1 is a schematic block diagram of a prior art method for Reed- Solomon encoding
• FIG. 2 is a schematic block diagram that illustrates a method for Reed- Solomon encoding, wherein the Galois field element is separated into halfwords before multiplication by the code generator coefficients
• FIG. 3 is a schematic block diagram that illustrates Reed-Solomon encoding in accordance with an embodiment of the invention
• FIGS. 4A and 4B illustrate embodiments of lookup tables for Reed-
• FIG. 5 is a schematic block diagram that illustrates Reed-Solomon encoding in a digital signal processor in accordance with an embodiment of the invention
• FIG. 6 is a flow chart that illustrates Reed-Solomon encoding in accordance with an embodiment of the invention.
• the remainder r k+ ⁇ (x) after step k+1 can be calculated from the remainder r k (x),at the previous step k and new incoming byte m k+ exactly in the same way
• RH I (X) (m k+] +r k ⁇ 2t . ⁇ )*g(x) + (r k2t . 2 *x 2t ⁇ 1 +.. ⁇ r k:0 *x+0*x 0 )
• Rfri ⁇ x (fk+igit-i + n ⁇ x 21'1 + (f k+I g 2t -2 + r k ⁇ 2t -3)x 2t'2 + ... +fk+ ⁇ go (21)
• FIG. 1 A schematic diagram representing the iteration of the remainder is shown in FIG. 1.
• a message symbol 10 (m s ) is summed with a high order remainder 46 (r 3 ) to provide a Galois field element 14 (f).
• Galois field addition is performed as an exclusive OR operation.
• the Galois field element 14 is multiplied by code generator coefficients 20, 22, 24, and 26 (go, g b g 2 and g ) to provide intermediate values 30, 32, 34 and 36.
• the intermediate values are summed with previous remainder symbols to provide current remainder symbols.
• the product fg 0 is summed with zero to provide remainder symbol 40 (r 0 );
• the product fgi is summed with previous remainder symbol 40 to provide current remainder symbol 42 (ri);
• the product fg 2 is summed with previous remainder symbol 42 to provide current remainder symbol 44 (r 2 );
• the product fg 3 is summed with previous remainder symbol 44 to provide current remainder symbol 46 (r 3 ).
• the process is repeated for each message symbol 10 in a codeword.
• the remainder symbols 40, 42, 44 and 46 represent check symbols for transmission with the message symbols.
• the codeword thus includes message symbols and check symbols.
• the example of FIG. 1 produces four check symbols for transmission with the message symbols and is capable of correcting two errors in the codeword.
• the code generator includes 16 code generator coefficients g ; and produces 16 check symbols for correction of eight errors in the codeword. It can be seen that when the conventional look-up table approach to the calculation of the remainder polynomial is used, the main bottleneck in the flow of computation is the repeated computation of the result of multiplication in the GF.
• a lookup table is constructed to facilitate calculations based on the observation that all the GF operations involved in one iteration of the remainder polynomial are performed on the same set of the second operands, namely, 2t coefficients of the code generator polynomial g(x).
• 2t*n values of the results of the multiplications of all n elements of GF by 2t coefficients of the code generator polynomial g(x) are pre-calculated and stored in the lookup table so that the contents of each entry F(f,i) in the table is defined as:
• g t is a coefficient of the code generator polynomial and sign "®" denotes Galois field multiplication operation. If this approach is used, even when the number of the errors corrected is relatively small, say 8 errors, and the Galois field used is GF(256), the size of the table is 4 Kbytes (16*256*lbyte, 16 is the number of check bytes to correct 8 errors). Often several different options for the RS code should be executed - e.g., when using several latency paths in the ADSL application, the codeword in one latency path has 16 check bytes and in another latency path has 12 check bytes, in this case the memory size will be 7 Kb etc.
• Galois field element 14 is separated into a high halfword 50 (HO) and a low halfword 52 (HL) using high and low masks according to equation (24).
• the high and low halfwords are each multiplied by each of the code generator coefficients 20, 22, 24, and 26.
• the respective products are summed to provide intermediate values.
• the intermediate values are summed with the previous remainder symbols to provide current remainder symbols.
• the architecture of FIG. 2 is combined with lookup tables to achieve advantageous results.
• each table in the case of GF(256) and 16 coefficients of the code generator polynomial is 256 bytes, and the total size of the two tables is 512 bytes - 8 times smaller than in the prior art approach, with the total execution time only ⁇ 50% longer because ⁇ 50% of the operations are not directly connected with the execution of the GFMUL operation.
• code generation using lookup tables TH and TL is illustrated schematically in FIG. 3.
• a first lookup table contains the values of high halfword HO multiplied by each of the code generator coefficients and a second lookup table contains the values of low halfword 0L multiplied by each of the code generator coefficients.
• the high halfword value is used with a coefficient index to access table TH to provide one or more high table values 54
• the low halfword value is used with the coefficient index to access table TL to provide one or more low table values 56.
• Respective high and low table values with the same coefficient index are summed in summing operation 60 to provide intermediate values, and the intermediate values are summed with the respective previous remainder symbols in summing operation 62 to provide current remainder symbols 40, 42, 44 and 46.
• the lookup tables can provide a single set of table values to be summed or two or more sets of table values to be summed, depending on the architecture of the system performing code generation.
• FIGS. 4 A and 4B The content of the lookup tables in accordance with an embodiment of the invention is shown in FIGS. 4 A and 4B.
• FIG. 4A illustrates a high halfword table 70 (TH)
• FIG. 4B illustrates a low halfword table 72 (TL).
• table 70 contains the products of each possible value of the high halfword HO multiplied by each of the code generator coefficients.
• each entry in table 70 contains a value of g,H, where i ranges from 0 to 2t-l and H ranges from 0 to the maximum value of high halfword H.
• each entry in table 72 contains a value of g,L, where i ranges from 0 to 2t-l and L ranges from 0 to the maximum value of low halfword L.
• the high halfword has 16 possible values ranging from 00 to F0 (using hexadecimal notation).
• the number of code generator coefficients depends on the number of desired check symbols and may be 16 for the case of eight check symbols.
• table 72 contains the products of each possible value of the low halfword 0L multiplied by each of the code generator coefficients.
• the low halfword values range from 0 to F.
• each of lookup tables 70 and 72 has 256 entries.
• tables 70 and 72 are organized such that each row contains the products corresponding to one halfword value. As discussed below, this organization facilitates parallel execution in a typical processor.
• the invention has been described in connection with embodiments where an eight-bit Galois field element is divided into high and low halfwords and two lookup tables are utilized. In other embodiments, each Galois field element may have more than eight bits.
• each Galois field element may be separated into more than two components and more than two lookup tables may be utilized.
• a 16-bit Galois field element The approach described in Patent No. 6,360,348 requires a 2 M byte lookup table.
• two lookup tables each of 8 K bytes (16 coefficients x 256 half GF element values x 2 bytes per entry), are required.
• four lookup tables each of 512 bytes (16 coefficients x 16 quarter GF element values x 2 bytes per entry), are required.
• an additional advantage of this embodiment can be achieved simply because of the fact that table accesses can be conducted with a monotonically increasing second index, so the table can be used as one-dimensional array with sequential access, and shift register cells are used in a strictly sequential order. This fact greatly speeds up the execution, first because multiple entries are read in one cycle, and second because (the XOR operation can also be performed on the multiple Galois field symbols simultaneously - e.g., 4 symbols at once when implemented on DSPs with a 32- bit data word path, 8 symbols for DSPs with a 64-bit data word path, etc.
• FIG. 5 A block diagram of a digital signal processor for encoding data according to an embodiment of the invention is shown in Fig. 5.
• FIG. 6 A process for encoding data in a digital signal processor having a word width of w Galois field elements, according to an embodiment of the invention, is shown in Fig. 6.
• a core processor 100 of the digital signal processor receives message symbols m and supplies check symbols r to be transmitted with the message symbols of a codeword.
• the core processor 100 accesses a high halfword lookup table 110 and receives table values gjH, gj + iH, g i+2 Hand gi +3 H.
• the core processor 100 also accesses a low halfword lookup table 112 and receives table values gL, gi + iL, g i+2 Land gi +3 L.
• message symbols m are bytes and core processor 100 has a word width of 32 bits. Accordingly, core processor 100 may process four table values per cycle from each of lookup table 110 and lookup table 112 as shown in Fig. 5 and described below. Core processors having different word widths may process more or fewer table values per cycle. Referring again to Fig.
• a next message symbol m is received and processed in step 150.
• a Galois field element f is calculated from the message symbol and a high order remainder symbol, as shown in Equation (28).
• the Galois field element is separated into first and second components, HO and OL, as shown in Equations (23) and (24). In this embodiment, the Galois field element is separated into high and low halfwords. However, in other embodiments, the Galois field element may be separated into more than two components.
• the high halfword HO and an index i are used to access w table values from lookup table 110 (Fig. 5).
• the index i is an index of generator coefficient gj.
• index i ranges from 0 to 15.
• low halfword 0L and index i are used to obtain w table values from lookup table 112 (Fig. 5).
• an intermediate value Si is calculated in step 160 for each of the w sets of table values as shown in Equation (30).
• a set of table values includes a table value from lookup table 110 and a table value from lookup table 112, the two table values having the same generator coefficient index.
• a current remainder symbol r is calculated by an exclusive OR operation on the intermediate value s; and a previous remainder symbol r ⁇ for each of the w intermediate values s; as also shown in Equation (30).
• w remainder symbols are calculated in step 162.
• step 170 a determination is made as to whether 2t remainder symbols have been calculated, where 2t represents the number of coefficients in the code generator polynomial.
• a determination is made in step 180 as to whether the end of the codeword has been reached.
• a typical codeword may include 188 message symbols. When the end of the codeword has not been reached, the process returns to step 150 to process the next message symbol m.
• 2t check symbols are output in step 182.
• 16 check symbols may be output for transmission with the 188 message symbols of the codeword.
• the parameters of the RS code can be set "on the fly" and can be different for different bearer channels within one ADSL channel.
• two methods of encoding can be realized based on the present invention. According to the first method, only the table for the highest possible number of check bytes is generated, and if a code with less error correction capability is used, the encoder works in two stages as follows.
• K is the number of cycles - usually 10-12 - required for conventional encoding.
• Another option is to generate "right" tables on the fly for any required code parameters which can require additional memory.
• the _ table is generated with a size 32*4k, where 4k is the closest integer multiple of 4 from above to the specified number of check bytes - e.g., 16 for 14, 12 for 10, etc.
• the remaining 64 bytes are filled by zeros, so when these values are read, they don't affect the result of the calculation.
• the same tables used in the encoder may be used in the decoder for calculating the syndromes in the following manner.
• the accumulation of the syndrome can be performed using the conventional approach on a much shorter sequence of 2t bytes than the original codeword of length n bytes.

Landscapes

• Physics & Mathematics (AREA)
• Probability & Statistics with Applications (AREA)
• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Mathematical Physics (AREA)
• Algebra (AREA)
• General Physics & Mathematics (AREA)
• Pure & Applied Mathematics (AREA)
• Error Detection And Correction (AREA)
• Detection And Correction Of Errors (AREA)

Priority Applications (1)

Applications Claiming Priority (2)

Publications (1)

Family

ID=34710449

Family Applications (1)

Country Status (5)

Cited By (4)

Families Citing this family (19)

Citations (5)

Family Cites Families (11)

• 2003
  • 2003-12-12 US US10/735,423 patent/US7162679B2/en not_active Expired - Lifetime
• 2004
  • 2004-11-16 KR KR1020067013905A patent/KR20060125837A/ko not_active Withdrawn
  • 2004-11-16 JP JP2006543834A patent/JP4777258B2/ja not_active Expired - Fee Related
  • 2004-11-16 CN CNA2004800370605A patent/CN1894857A/zh active Pending
  • 2004-11-16 WO PCT/US2004/038272 patent/WO2005062472A1/en not_active Ceased

Patent Citations (5)

Also Published As

Similar Documents

Legal Events