Patent Grant
8321725
The present invention relates to error correction in coding schemes for digital communication systems, and more particularly to design optimization for Interleavers of any size within a specified wide range used in such error correction. Even more particularly, the present invention relates to optimization of Turbo Interleavers such that smaller optimal Interleavers can be built from larger optimal Interleavers.
Interleaving is a process of reordering a sequence of symbols or bits in a predetermined manner. “Interleaver size” is equal to the size of the sequence. The apparatus performing the interleaving is referred to herein as an Interleaver.
Turbo Interleavers are interleavers used in the construction of turbo codes. In a turbo code built as a parallel concatenation of two constituent recursive convolutional codes, a Turbo Interleaver serves to re-order an input data sequence in a pseudo-random fashion prior to an encoding by a second of the constituent codes. As a result, separate encodings produced by the two constituent encoders are largely uncorrelated, which property allows them to be combined by a turbo encoder to produce a composite encoding with excellent error protection capability.
S-random Interleavers are one of the most widespread forms of turbo Interleavers.
The principle behind S-random Interleavers is to avoid mapping neighbor positions of an original input sequence to another neighbor position of the interleaved sequence within a window of size S. The design goal in S-random Interleavers is to maximize S while preserving the above principle. However, S-random Interleavers have to be re-designed every time the Interleaver size is changed and there is typically no requirement of any resemblance between the Interleavers with similar sizes.
Thus, it is desirable to have a general Interleaver design for Interleavers of any size within a set of sizes, wherein the design methodology is concise and efficient such that the same Interleaver design is near-optimal for all Interleavers within the set of sizes. It is also advantageous to have a design for building a near-optimal Interleaver that can easily be reduced to smaller-sized near-optimal Interleavers without performance degradation.
Therefore, the present invention advantageously addresses the above and other needs.
The present invention advantageously addresses the needs above as well as other needs by providing a method and apparatus for a turbo Interleaver which employs Interleavers of variable length employing one or more permutations.
In one embodiment, the invention is characterized as a method of interleaving blocks of indexed data of varying length. The method includes the steps of: providing a set of basic Interleavers comprising a family of one or more permutations of the indexed data and having a variable length; selecting one of the basic Interleavers based upon a desired Interleaver length L; and adapting the selected basic Interleaver to produce an Interleaver having the desired Interleaver length L.
In another variation, a method of interleaving blocks of indexed data of variable length includes the steps of: providing a family of basic Interleavers comprising “two-dimensional permutations” including computing the “two-dimensional permutations”, further comprising: writing the indexed data into an Interleaver matrix having one or more rows in each of two dimensions; permuting the indexed data in one or more rows in at least one of the two dimensions to produce “constituent permutations”, possibly being different from one row to another row, wherein the constituent permutations are pseudo-random permutations described by a limited number of parameters, wherein an amount of storage required for storing the limited number of parameters is less than that for storing a vector representation of the constituent permutations; reading out the data from the Interleaver matrix; selecting one of the basic Interleavers for use in encoding based upon a desired Interleaver length L; adapting the selected basic Interleaver to produce an Interleaver having the desired Interleaver length L; wherein the selecting includes: identifying a group of the basic Interleavers having a length greater than or equal to the desired Interleaver length L; and selecting one of the basic Interleavers having a length smallest among the identified group of the basic Interleavers; wherein the adapting includes: deleting indexed data having indices higher than required for a permutation of length L; providing an Interleaver device for interleaving blocks of indexed data, the Interleaver device further comprising a memory device for storing descriptions of the basic Interleavers; and storing the descriptions in the memory device.
In another embodiment, a system for interleaving and turbo encoding blocks of indexed data, of varying length, comprises: a parallel concatenation of two or more constituent encoders for recursive convolutional codes of recursion period p; and an Interleaver device coupled to the parallel concatenation for performing the steps of: accessing stored descriptions of basic Interleavers, the basic Interleavers comprising a family of one or more permutations of the indexed data and having a variable length; identifying a group of the basic Interleavers having a length greater than or equal to a desired Interleaver length L; selecting one of the basic Interleavers having a length which is smallest among the group of the basic interleaves; and adapting the selected one of the basic Interleavers to produce an Interleaver having the desired Interleaver length L.
The above and other aspects, features and advantages of the present invention will be more apparent from the following more particular description thereof, presented in conjunction with the following drawings wherein:
Corresponding reference characters indicate corresponding components throughout the several views of the drawings.
The following description of the presently contemplated best mode of practicing the invention is not to be taken in a limiting sense, but is made merely for the purpose of describing the general principles of the invention. The scope of the invention should be determined with reference to the claims.
Referring to
The transmitter receives an analog signal through a Transmitter Interface 102 and performs an A/D conversion at A/D Connector 104. The discrete samples generated from the A/D Converter 104 are fed to Segmentation Processor 106 where fixed-length data units of 44 octets are formed by fragmenting an initial MAC protocol data unit (IMPDU), and then the fixed length data units passed to the Turbo Coder 108 which uses an Interleaver to pseudo-randomize the input between 2 concatenated encoders and encodes the fixed length data units (data streams) and sends encoded data units (data streams) to the Burst Formatter 110.
A burst, a series of repetitive waveforms at a prescribed time and amplitude lasting a short duration, is formed at Burst Formatter 110 and is passed to Modulator 112 where the burst is modulated by mixing with a carrier waveform of known frequency. Transmitter 114 transmits the modulated burst when the switch 120 connects antenna 122 to the Transmitter 114. The Synthesizer and Oscillator 118 keeps track of timing for the transmitter (RF/IF) 114 and for the Timing and Control Processor 116 which controls when bursts are formatted.
When the Antenna 122 receives a burst and the receiver (RF/IF) Preamp Mixer 126 is connected to the antenna through the switch 120, the received burst is amplified by the receiver (RF/IF) Preamp Mixer 126, and then demodulated to remove the carrier waveform frequency. The Equalizer 130 filters the demodulated burst with filters adjusted so as to produce an enhanced digital signal which is next Turbo decoded by Turbo Decoder 132 through a concatenation of decoders and an Interleaver using feedback from each other decoder to decode information from the received burst. Decoded data is converted from digital to analog by D/A Converter 134 and passed through receiver interface 136 to another system for further processing as needed. Since the digital communication system of
Referring to
The two constituent encoders 10, 10′ produce parity bits Y0(t), Y1(t) and Y0′(t), respectively, selected ones of which are removed from an output stream (output) of the two simple constituent encoders 10, 10′ according to a prescribed puncturing pattern by the puncturer 36 in order to achieve a desired overall Turbo code rate. Both the first encoder (encoder #1) 10 and the second encoder (encoder #2) 10′ process the same information bit stream and X(t) (or “information bits”), but the encoder #210′ processes information bits X(t) in a different order than the order in which encoder #110 does since the Interleaver 16 processes the information bits X(t) before they reach encoder #210′. By rearranging an order of presentation of the information bits X(t), the Interleaver 16 serves to decorrelate the outputs of the two simple constituent encoders 10, 10′ so that the information bits X(t) causing encoder #110 to produce a low-Hamming weight output are unlikely to cause encoder #210′ to also produce a low-Hamming weight-output.
In
Both encoders 10, 10′ produce, in addition to the information bits X(t) (or systematic bits), parity bits Y0(t) and Y1(t) which are punctured by puncturer 36 to achieve a desired overall Turbo Code rate.
Information bit stream X(t) is received at switch 12, and is processed in accordance with several modular adders of above and shift registers above which are hard-wired to represent two (2) numerator polynomials and one denominator polynomial.
Referring still to
A first numerator polynomial over a denominator polynomial, representing Turbo Code “1101” is hardwired to return output Y0(t) by combining: X(t) with a result of modulator adder 17 to create a first bit W(t); the modular sum (second bit) of shift register 18 and W(t) from the modular adder 20; another zero bit (third bit) indicated by the lack of connection to the register 21; and the modular sum (fourth bit) of another register 22 and a result of modular adder 20 from modular adder 24. The result is Y0(t)=W(t)+S0(t)+S2(t).
Information bit stream X(t) is presented in its original, uninterleaved order at a switch 12 and processed by the first encoder 10. In
Thus, during an encoding step at time t≧0, a shift register contents of the shift register 18, 21, 22, are S0(t), S1(t), S2(t) and the information bit X(t) is present at the input to binary adder 17. The encoder 10 then produces its two coded output bits (coded bits) Y0(t), Y1(t) according to the following two summations:
Y0(t)=W(t)+S0(t)+S2(t)
Y1(t)=W(t)+S0(t)+S1(t)+S2(t),
wherein
W(t)=X(t)+S1(t)+S2(t).
After the coded bits are output, the current encoding step at time t is completed by shifting the contents of the shift register 18, 21, 22 once to prepare for a next encoding step at time t+1. At time t+1:
S0(t+1)=W(t), S1(t+1)=S0(t) and S2(t+1)=S1(t). At the start of an encoding process at t=0, the shift register contents are initialized to zero, wherein S0(0)=S1(0)=S2(0)=0. The second encoder 10′ operates in the same fashion on an output of Interleaver 16 to produce another two (2) coded output bits.
Since the digital communication system of
The basic interleavers are stored in a memory, which may be located within the Interleaver 16 as in
A couple of simple examples will clarify these concepts. First, consider an interleaver of length 8 using a permutation n=(04261537). This permutation could be used as either a list of write (input sequence) addresses or read addresses. Let dI(0), dI(1), . . . , dI(7) denote input data (input sequence) in their original sequence; and let do(0), do(1), . . . , do(7) denote values of the same input data but in a permuted order. The interleaver could be implemented to write d1(0) to output position 0, dI(1) to output position 4, dI(2) to output position 2, dI(3) to output position 6, etc. In this case, the interleaver action can be expressed mathematically as
do(π(k))=dI(k).
Alternately, the interleaver could be implemented to read data values from the input data according to the permutation n. That is, a first interleaved value do(0) is read from input position 0, a second interleaved value do(1) is read from input position 4, and so on. Mathematically,
do(k)=dI(π(k)).
Neither interpretation is to be preferred; it is merely a matter of convention. For purposes of describing interleaver operations herein, the first interpretation (in which permutations specify write addresses for the interleaver) is used.
It should be noted, however, that in a turbo decoder, such as the turbo decoder 132 in
The permutation π=(04261537) arises from bit-reversal indexing. For example, input position 1 has a 3-bit binary representation 001 and is mapped to output position 4, which has 100 as its 3-bit binary representation. Likewise, input position 3 (binary 011) is mapped to output position 6 (binary 110). In VLSI hardware or on some digital signal processing, which, optionally may be employed within the Interleaver 16 bit-reversed indexing is easily accomplished without special memory storage.
The permutation π=(03614725) can be generated by the simple mathematical recursion:
α=3; π(0)=); π(k)=π(k-1)+α(mod 8).
Different recursions of this type can be described by the two parameters α and π(0). Thus, a family of basic interleavers based on simple recursive formulas could be represented by a small table of parameters stored in the memory. Advantageously, for an interleaver of large block length or a large set of interleavers of various block lengths, the ability to store a small table of parameters rather than the explicit permutations results in a large reduction in memory requirements. Thus, it is advantageous to design turbo interleavers in this way provided a parameterized family of interleavers results in good turbo code performance. These design issues are favorably addressed by the proposed invention.
Given a family of basic interleavers of various block sizes represented and stored in memory in some fashion, a turbo device (either the turbo encoder 108 or the turbo decoder selects one of them for use in implementing an interleaver of specific length L. In one preferred embodiment of the invention, the lengths of the basic interleavers are all different, and the turbo device selects the basic interleaver having a smallest length N among all basic interleavers whose lengths are at least as big as the desired length L.
In other embodiments, it may be desirable to have multiple basic interleavers all of the same length. For example, there may be an implementation advantage in having all basic interleavers have lengths that are integral powers of two. In such a design, there may be multiple basic interleavers of length N=2c, each optimized for a different interval of block sizes between 2c-1 and 2c. In such embodiments, the turbo device (turbo encoder 108, turbo decoder 132) first identifies a set of basic interleavers having a smallest length N among all the basic interleavers whose lengths are at least as big as the desired length L and then selects one of the basic interleavers in the set according to other selection criteria depending on L.
Once a basic interleaver has been selected, it is then adapted to length L by the process referred to herein as pruning. Pruning refers to a discarding of permutation indices that are invalid for a pruned matrix. For example, one prunes the permutation π=(03614725) f length 8 on the integers modulo 8 to a new permutation of length 5 on the integers modulo 5 by ignoring the invalid indices 5, 6 and 7. Thus, a pruned permutation is π*=(03142).
The process of pruning, in accordance herewith, is further explained by an algorithm shown in
The above rules are refined later herein so that Rule 1 and Rule 2 continues to be satisfied for Turbo Interleavers of any size N obtained from a single Interleaver of size 2m by means of puncturing (2m-1<N≦2m). Obtaining an Interleaver of any size N from a mother Interleaver of a larger size via puncturing is one aspect of this invention.
A smaller Interleaver IN, of size Na is formed by using a pre-designed Interleaver matrix, I2
A Smaller Interleaver, IN
Thus, a Smaller Interleaver, IN
This can be accomplished, for example, by a processor modified with a computer program, the steps of which are shown in
1) Initialize a counter i to zero, where i represents a new smaller Interleaver bit position, and j represents an original larger Interleaver bit position. This corresponds to Initialize Counter 310 in
2) For every original bit position I2
If I2
Otherwise reject I2
Pruning is a key aspect of the invention described herein. It is the advantage that the method is easily implemented in either a VLSI or a DSP and so provides an efficient mechanism for providing interleavers of arbitrary lengths without storing separate descriptions for every possible length. The set of basic interleavers are designed to be robust with respect to pruning in accordance with principles to be described in conjunction with a detailed, explicit design illustrative of the invention.
The design of a turbo interleaver should take into account the structure of the constituent recursive convolutional codes in order to ensure that an overall Turbo code has a favorable Hamming weight spectrum (“weight spectrum”) leading to good error correction performance. The weight spectrum of a linear binary code of length N is a tabulation giving a number of code words of each Hamming weight from 0 to N. A Hamming weight is a number of non-zero entries in the code word. Since the constituent code of a Turbo Code is recursive, it takes an input sequence of a Hamming weight of at least two (2) to cause a systematic encoder to leave an all zero state and later to return to the all zero state and therefore to generate a less desirable, low parity Hamming weight sequence. In general, a systematic, recursive encoder would generate a parity sequence of high Hamming weight for input sequences having a Hamming weight of one (1), since the encoded sequence upon leaving the all zero-state can never return to it. For recursive convolutional codes as constituent codes, the probability that both encoders generate encoded sequences that leave the all zero-state and later return to the all zero-state is the highest when the input sequence is of Hamming weight two (2).
It is also observed that when the recursive eight-state constituent encoders have a primitive feedback polynomial of degree 3, an input sequence of Hamming weight two (2) can cause the first constituent encoder to generate a finite error event only if the two “1's” in the input sequence are separated by 6+7n (n is an integer) zeros. It is therefore important that any input sequence consisting of exactly two 1's separated by 6+7n (nεN) zeros should not be mapped by the interleaver to a new sequence with two 1's now separated by 6+7m (mεN) zeros. In that way the second encoder 10′ will generate high parity Hamming weight when the first encoder generates low parity Hamming weight, and vice versa, corresponding to an input sequence of Hamming weight two (2).
Even if the two 1's are separated by the undesirable 6+7n zeros in an input sequence of Hamming weight two, the corresponding parity Hamming weight will grow larger as n grows larger. Thus, it is less crucial to address the cases for n>1, as the most critical values for n are 0 followed by 1, respectively. This is because as n grows, parity Hamming weight grows sufficiently larger.
The rules that are introduced in one embodiment of the invention to design Turbo Interleavers for eight-state Turbo codes are thus:
Rule 1: Minimize the occurrence of events:
|I[x]−I[x−7]|=7 (1)
wherein I[x] denotes the position that x is mapped to by the Interleaver matrix I.
Rule 2: If the first rule is satisfied with zero occurrences of equation (1), minimize the occurrence of event:
|I[x]−I[x−7]|=14, or
|I[x]−I[x−14]|=7, or
|I[x]−I[x−14]|=14 (2)
By following the above created rules, the probability of both of the encoders 10, 10′ generating low-Hamming weight parity sequences is minimized.
An explicit exemplary turbo interleaver design (exemplary design) will not be described in order to more fully illustrate and develop the concepts of the invention. In the exemplary design, each of the basic interleavers implemented by the Interleaver 16 is a two-dimensional block interleaver (or interleaver matrix) of dimension R×C, where R−2r is a number rows and C−2c is a number of columns. Conceptually, the input data (data) are written into the interleaver matrix row by row. Then row and column permutations are performed to randomize data positions. The data are then read out column by column. Specifically, given an input position l=C·i+j, a corresponding output interleaved position will be given by mathematical formula I(l)=R·πi (j)+ρ(i), wherein πi is a column permutation applied to data in row i and wherein ρ is bit-reversed indexing, which is especially simple to implement and requires no additional parameter storage in the memory of the interleaver 16.
One could, of course, make the ρ permutation different for different columns ad the expense of additional implementation complexity and increased storage requirements to specify each of the individual column permutations. In either case, the ρ premutation(s) should perform pseudo-random interlacing of top and bottom halves of the interleaver matrix in order to facilitate on-the-fly implementation of pruning. Such interlacing ensures that, if I(l) is an invalid index for a pruned interleaver, then I(l+1) will be a valid index, assuming that no basic interleaver is pruned to half its length or beyond.
The proposed two-dimensional structure is advantageous for use in turbo interleaving for several reasons. Since the turbo interleaver is built in a structured way from simple constituent permutations that can be described by a small set of parameters, implementation complexity is small. When different constituent permutations are used from row to row, a composite interleaver permutation exhibits sufficient randomness to achieve good turbo code performance despite its low complexity. Furthermore, by choosing R and C appropriately, one can balance the “spreading capability” of the interleaver (how well is separates neighboring positions) and its “randomness” properties. The spreading capability of the interleaver is also important for the turbo interleaver in that it helps to enhance the overall “weight spectrum” of the turbo code. In general, spreading capability increases with increasing R, and randomness increases with increasing C.
Preferably, as a rule of thumb, for building interleaver matrixes in accordance with the invention, one would make R as large as possible without making C so small that the randomness produced by the permutations applied to each row is degraded.
Thus, in the illustrative designs presented below, the set of basic interleavers have the property that, in general, those of larger length use a larger number of rows R. This is an important aspect of the two-dimensional design.
In the illustrative designs, in accordance with one aspect of the invention presented below, the constituent permutations applied to rows of the interleaver matrix are based on a novel class of permutations derived from Galois field arithmetic.
A Galois Field (GF) with pm elements is denoted as GF(pm), wherein p is a prime number and m is any integer greater than one (1). It can be formed from GF(p) using a primitive polynomial p(x) of degree m over GF(p)[x]. In the case of GF(2m), the roots of primitive polynomial p(x) of degree m over GF(2)[x], form a subset of the primitive elements in GF(2m). A primitive element in a Galios field with q elements has order q−1, i.e., the smallest positive integer n such that αn=1 is n=q−1.
If ∝ is a primitive element in GF(2m); all of the other nonzero elements of GF(2m) can be obtained as consecutive powers of α.
GF(2m)={0,α0=1,α,α2,α1, . . . ,α2m-2} (5)
Furthermore, every element of the field GF(2m) can be expressed in terms of 1, α, α2, . . . αm-1;
For example, GF(8) can be constructed from GF(2) using the primitive polynomial p(x)=x3+x+1 over GF(2)(x). Let α be a root of p(n). Multiplication of elements can be performed using the fact that α7=1 (by definition since α is primitive in GF(8)). Addition of elements in GF(8) can be performed using equalities in terms of 1, α, and α2 since α3=α+1 in Galois Field arithmetic (where 1≡−1).
An exemplary multiplication of elements in GF(8) is Equation (4).
α3·α6=α9=α2 (6)
An exemplary addition is Equation (5).
α3+α4=(α+1)+(α2+α)=α2+1=α6 (7)
An Interleaver of size 2m, I2m, is formed by the following four (4) steps:
(1) First, a matrix is filled row by row with bit positions starting with 0 in the upper leftmost position, and ending with (r×c−1, where 2m=r×c as defined above) in the lower rightmost position. This is the conventional manner of filling Interleaver matrices.
Thus, an Interleaver matrix of size 32=4×8 would result in matrix (8):
(2) Secondly, permute each row i (i=0, 1, 2, . . . , r−1) (within itself according to a predetermined rule. One method is to permute) according to the following permutation rules employing Galois Field arithmetic (9):
j←logαib(αi
j←logαib(αi
In Equation (9), ∝ is a root of the primitive polynomial p(n) used to construct GF(c), αi
Furthermore, by definition logαib(o) is set to (c−1) as a result of the second part of Equation (9).
An exemplary permutation would be such as is shown in Equation (10) for each row i (i=0, 1, 2, 3), for an Interleaver of size 32 having 8 columns and 4 rows.
j←logαib(αi
j←logαib(αi
For the sake of demonstration, Table 1 is constructed with constants ib and i0 to permute a Turbo Interleaver matrix of size Ns, within the set N, where 16<N≦32. The values of Table 1 are fabricated herein for the sake of the following example for the construction of I32.
Thus, in accordance with Table 1, each row is shuffled such that:
row i=0: j←logα(αj+1) for j=0,1,2,3,4,5,6
j←logα(1) for j=7 (11)
row i=1: j←logα(αj+α2) for j=0,1,2,3,4,5,6
j←logα(α2) for j=7 (12)
row i=2: j←logα
j←logα
row i=3: j←logα
j←logα
The shuffling of each row results in a pseudo-random order of positions represented by sequences to the right of each of the arrows in equations (15), which represent the original bit positions which map to newly ordered sequences to the left of each arrow:
row 0: (0,1,2,3,4,5,6,7)←(7,3,6,1,5,4,2,0)
row 1: (0,1,2,3,4,5,6,7)←(6,4,7,5,1,3,0,2)
row 2: (0,1,2,3,4,5,6,7)←(6,2,1,3,0,7,5,4)
row 3: (0,1,2,3,4,5,6,7)←(2,5,6,1,7,0,4,3) (15)
The above shuffling results in an Interleaver matrix:
(3) Thirdly, each of the rows of Interleaver matrix I32 are shuffled or re-ordered according to any method that interlaces an upper half of the matrix resulting from the above permutations, with a lower half of the matrix.
One method of doing this is to re-order the rows according to a bit reversal on row index (e.g. as represented by the pattern (00, 01, 10, 11) for a four-row matrix). From the above example matrix, this results in matrix:
(4) Fourthly, the contents of the resulting permuted and re-ordered matrix, or Interleaver matrix, are read out column by column to an encoder as in the case of a Block Interleaver.
In the above example, this results in the bit position sequence:
7 22 14 26 3 18 12 29 6 17 15 30 1 19 13 25 5 16 9 31 4 23 11 24 2 21 8 28 0 20 10 27 (18)
Permutations of the rows within themselves should be done in such a way that Rule 1 and Rule 2 are satisfied for any Interleaver size. N obtained from an original Interleaver of size 2m where 2m-1<N≦2m.
The preferred basic interleaver structure, described previously herein allows the formulation of simple design criteria that help ensure robustness to pruning. The key observation is that, because of the way Interleavers of size 2m are constructed, for any window of size 2W, W an integer, there is at most W indices that must be pruned in order to obtain an Interleaver of size N wherein 2m-1<N≦2m. An Interleaver of 17 elements is obtained by pruning the interleaver of equation (18).
7 14 3 12 6 15 1 13 5 16 9 4 11 2 8 0 10 (19)
resulting in an Interleaver matrix having elements of
Thus, the previously discussed rules for good turbo interleaver design—Rules 1 and 2 given in equations (1) and (2)—can be generalized to provide rules for the design of good turbo interleavers that are robust to pruning. The modified rules are as follows:
Modified Rule 1: Minimize the occurrence of events.
|I[x]−I[x−j]|=7 where 7≦j≦14, jεN (20)
Modified Rule 2: If the first modified rule is satisfied with zero occurrence, minimize the occurrence of events.
|I[x]−I[x−j]|7, or (21)
|I[x]−I[x−j]|=14 where 7≦j≦28, jεN (22)
A third Rule is also introduced: Rule 3: If the Modified Rule 1 and Modified Rule 2 are satisfied with zero occurrence, maximize the variable S such that, neighbor positions within a window size S are not mapped to neighbor positions within a window of size S.
Because of the key observation stated before the modified rules, it can be seen that if an Interleaver of size 2M satisfies Modified Rule 1 and Modified Rule 2, then all of the Interleavers of size N (2m-1<N≦2m) obtained by pruning satisfy Rule 1 and Rule 2 stated at the beginning of this invention.
In designing preferred integer constants io and ib for the Galois Field permutations to achieve pseudo-randomness in a Galois Field Interleaver, performance of constructed Interleaver matrices are measured according to how well they meet the modified rules above.
Interleaver matrices of size 128, 256, 512, 1024, 2048 and 4096 are constructed in accordance with the modified rules to yield near optimal Interleaver matrices of any size N where 64<N≦4096.
The primitive polynomials used to construct GF(c), where c is the number of columns of the Interleaver matrix are as follows:
p(x)=x4+x+1 to construct GF(16)
p(x)=x5+x2+1 to construct GF(32)
p(x)=x6+x+1 to construct GF(64)
p(x)=x7+x3+1 to construct GF(128)
p(x)=x8+x4+x3+x2+x to construct GF(256)
p(x)=x9+x4+1 to construct GF(512) (23)
Table 2 shows best values of ib and io, as defined and determined above, for each row index i for each Interleaver matrix size specified.
In all cases of Interleavers and associated matrices designed from the above constants in accordance with the invention, Modified Rule 1 and Modified Rule 2 is completely satisfied except for a small number of
In one embodiment, a computer search determines the constants such that the Modified Rules are satisfied.
Referring to
Curve 410 of
The illustrative design of Table 2 can be simplified by further restricting the choice of parameters. For example, the hardware implementation as well as storage requirements are reduced if the parameter ib is made constant and equal to 1. In this case, the parameters describing the constituent permutations to be applied within each row R should be re-optimized with respect to Modified Rules 1 and 2.
Table 4 shows the near optimal values of integer constant i0 for each row index i and each Interleaver matrix size, if ib is 1.
While the invention herein disclosed has been described by means of specific embodiments and applications thereof, numerous modifications and variations could be made thereto by those skilled in the art without departing from the scope of the invention set forth in the claims.
For example, in the preferred embodiments, the constituent permutations applied within each row of the interleaver matrix were based on discrete logarithms in a Galois field. It is clear that this is only a particular example of a broad class of permutations based on Galois field arithmetic, which may be employed in accordance herewith. Optimally, another closely related choice would be to take a non-primitive element βeGF(C) of multiplicative order ord(β) and define
πi(J)=log(βio+βj), (j=0,1,2, . . . ,ord(β)−1)
to produce a constituent permutation of length ord (β)−1. Alternatively, a logarithm of a different linear or affine function of βj may be employed. More generally, one could take permutation mapping data at position i=0, 1, 2, . . . , ord(β)−1 to a new position
πi(j)=ƒ(βj),
wherein ƒ is any integer-valued function acting on finite field GF(C) and β is a non-zero element in GF(C) of multiplicative order ord(β).
In yet another alternative embodiment, the finite field(s) need not be binary—that is C need not be a power of 2.
This application is a continuation of U.S. application Ser. No. 11/051,585, filed Mar. 30, 2005, entitled TURBO CODE INTERLEAVER WITH NEAR OPTIMAL PERFORMANCE, by Mustafa Eroz, et al., now U.S. Pat. No. 7,526,687, issued Apr. 28, 2009, which is a continuation of U.S. application Ser. No. 10/024,834, filed Dec. 19, 2001, entitled TURBO CODE INTERLEAVER WITH NEAR OPTIMAL PERFORMANCE, by Mustafa Eroz, et al., now U.S. Pat. No. 6,925,587, issued Aug. 2, 2005, which is a divisional of U.S. application Ser. No. 09/375,067, filed Aug. 16, 1999, entitled TURBO CODE INTERLEAVER WITH NEAR OPTIMAL PERFORMANCE, by Mustafa Eroz, et al., now U.S. Pat. No. 6,334,197, issued Dec. 25, 2001, which claims benefit of U.S. Provisional Application Ser. No. 60/096,807, filed Aug. 17, 1998.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5063533 | Erhart et al. | Nov 1991 | A |
| 5159608 | Falconer et al. | Oct 1992 | A |
| 5687095 | Haskell et al. | Nov 1997 | A |
| 5721745 | Hladik et al. | Feb 1998 | A |
| 5742612 | Gourgue et al. | Apr 1998 | A |
| 5751725 | Chen | May 1998 | A |
| 5761249 | Ben-Efraim | Jun 1998 | A |
| 5822359 | Bruckert et al. | Oct 1998 | A |
| 5859840 | Tiedemann et al. | Jan 1999 | A |
| 5881093 | Wang et al. | Mar 1999 | A |
| 5907582 | Yi | May 1999 | A |
| 5910182 | Dent et al. | Jun 1999 | A |
| 5944850 | Chouly et al. | Aug 1999 | A |
| 5970085 | Yi | Oct 1999 | A |
| 5978414 | Nara | Nov 1999 | A |
| 5983384 | Ross | Nov 1999 | A |
| 5987057 | Song | Nov 1999 | A |
| 5996104 | Herzberg | Nov 1999 | A |
| 6023783 | Divsalar et al. | Feb 2000 | A |
| 6088387 | Gelblum et al. | Jul 2000 | A |
| 6094427 | Yi | Jul 2000 | A |
| 6289486 | Lee et al. | Sep 2001 | B1 |
| 6332209 | Eroz et al. | Dec 2001 | B1 |
| 6334197 | Eroz et al. | Dec 2001 | B1 |
| 6339834 | Crozier et al. | Jan 2002 | B1 |
| 6347385 | Cui et al. | Feb 2002 | B1 |
| 6370669 | Eroz et al. | Apr 2002 | B1 |
| 6430722 | Eroz et al. | Aug 2002 | B1 |
| 6519732 | Li | Feb 2003 | B1 |
| 6530059 | Crozier et al. | Mar 2003 | B1 |
| 6665829 | Eroz et al. | Dec 2003 | B2 |
| 6925587 | Eroz et al. | Aug 2005 | B2 |
| 7526687 | Eroz et al. | Apr 2009 | B2 |
| 7657797 | Eroz et al. | Feb 2010 | B2 |
| 7761750 | Eroz et al. | Jul 2010 | B2 |
| 20020083395 | Eroz et al. | Jun 2002 | A1 |
| 20020166093 | Eroz et al. | Nov 2002 | A1 |
| 20030041297 | Eroz et al. | Feb 2003 | A1 |
| 20030051205 | Eroz et al. | Mar 2003 | A1 |
| Number | Date | Country |
|---|---|---|
| 19520987 | Dec 1996 | DE |
| 19736653 | Dec 1998 | DE |
| 0300139 | Jan 1989 | EP |
| 0952673 | Oct 1999 | EP |
| 62-190932 | Aug 1987 | JP |
| 7-202851 | Aug 1995 | JP |
| 11-68734 | Mar 1999 | JP |
| WO 9637050 | Nov 1996 | WO |
| WO 9848517 | Oct 1998 | WO |
| WO 9907076 | Feb 1999 | WO |
| WO 0013323 | Mar 2000 | WO |
| WO 0041343 | Jul 2000 | WO |
| WO 0048353 | Aug 2000 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 20080065948 A1 | Mar 2008 | US |
| Number | Date | Country | |
|---|---|---|---|
| 60096807 | Aug 1998 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 09375067 | Aug 1999 | US |
| Child | 10024834 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 11051585 | Mar 2005 | US |
| Child | 11980917 | US | |
| Parent | 10024834 | Dec 2001 | US |
| Child | 11051585 | US |
