Title of Invention	"APPARATUS FOR GENERATING A SUB CODES IN COMMUNICATIONS SYSTEM"
Abstract	There are provided an apparatus and method for generating an initial puncturing matrix from which a first sub-code is produced in a communication system having a turbo encoder for generating information symbols, first parity symbols, and second parity symbols for the input of an information bit stream and a sub-code generator for generating sub-codes from the information symbols, the first parity symbols, and the second parity symbols using puncturing matrices, a number of the sub-codes being identical to a number of the puncturing matrices, comprising the steps of selecting as many information symbols as a number of columns in the initial puncturing matrix from the information symbols output from the turbo encoder.

Title of Invention

"APPARATUS FOR GENERATING A SUB CODES IN COMMUNICATIONS SYSTEM"

Abstract

There are provided an apparatus and method for generating an initial puncturing matrix from which a first sub-code is produced in a communication system having a turbo encoder for generating information symbols, first parity symbols, and second parity symbols for the input of an information bit stream and a sub-code generator for generating sub-codes from the information symbols, the first parity symbols, and the second parity symbols using puncturing matrices, a number of the sub-codes being identical to a number of the puncturing matrices, comprising the steps of selecting as many information symbols as a number of columns in the initial puncturing matrix from the information symbols output from the turbo encoder.

Full Text	GENERATING CODES IN A COMMUNICATION SYSTEM BACKGROUND OF THE INVENTION I Field of the Invention The present invention relates generally to an apparatus and method for generating codes in a data communications system, and in particular, to an apparatus and method for generating complementary codes with use of turbo codes in a communications system employing a retransmission scheme. 2. Description of the Related Art In general, a system using a retransmission scheme (e.g., hybrid ARQ [Automatic Repeat Request]) performs soft combining to improve transmission throughput. The soft combining techniques are divided into packet diversity combining and packet code combining. These two combining schemes are usually called soft packet combining. Although the packet diversity combining scheme is sub-optimal in performance relative to the packet code combining scheme, it is favorable due to easy implementation when performance loss is low. A packet transmission system uses the packet code combining scheme to improve transmission throughput. That is, a transmitter transmits a code with a different code rate at each packet transmission. If an error is detected from the received packet, a receiver requests a retransmission and performs soft combining between the original packet and a retransmitted packet. The retransmitted packet may have a different code from the previous packet. The packet code combining scheme is a process of combining received N packets with a code rate R to a code with an effective code rate of R/N prior to decoding, to thereby obtain a coding gain. With regard to the packet diversity combining scheme, on the other hand, the transmitter transmits a code with the same code rate R at each packet transmission. If an error is detected from the received packet, the receiver requests a retransmission and performs soft combining between the original packet and the retransmitted packet. The retransmitted packet has an identical code to that in the previous packet. In this sense, the packet diversity combining scheme can be considered symbol averaging on a random channel. The packet diversity combining scheme reduces noise power by averaging the soft outputs of received symbols and achieves such a diversity gain as offered by a multipath channel because the same code is repeatedly transmitted on a fading channel. However, the packet diversity combining scheme does not provide such an additional coding gain as obtained according to a code structure in the packet code combining scheme. Due to implementation simplicity, most packet communication systems have used the packet diversity combining scheme, which is under study for application to the synchronous IS-2000 system and the asynchronous UMTS system. The reason is that the existing packet communication systems have used convolutional codes and even packet code combining does not offer a great gain when convolutional codes with a low code rate are used. If a system with R=l/3 supports retransmission, there is not a wide difference in performance between the packet code combining scheme and the packet diversity combining scheme. Thus, the packet diversity combining scheme is selected considering implementation complexity. However, the use of turbo codes as forward error correction codes (FEC) requires a different packet combining mechanism because the turbo codes are designed as error correction codes to have performance characteristics very close to the "Shannon Channel Capacity Limit" by iterative decoding and their performance varies obviously with the coding rates unlike convolutional codes. Therefore, it can be concluded that packet code combining is feasible for a packet communication system using turbo codes in a retransmission scheme to achieve the goal of optimum performance. SUMMARY OF THE INVENTION It is, therefore, an object of the present invention to provide an apparatus and method for generating sub-codes for optimal code combination in a retransmission system using turbo codes. It is another object of the present invention to provide an apparatus and method for generating complementary codes using turbo codes in a communications system. The above and other objects can be achieved by providing an apparatus and method for generating sub-codes from turbo codes in a communications system. In the sub-code generating apparatus, a turbo encoder generates information symbols, first parity symbols, and second parity symbols for the input of an information bit stream and a sub-code generator generates sub-codes from the information symbols, the first parity symbols, and the second parity symbols using puncturing matrixes. Here, the number of the sub-codes is identical to the number of the puncturing matrixes. To generate an initial puncturing matrix for the first sub-code, the sub-code generator selects as many information symbols as the number of columns in the initial puncturing matrix from the information symbols from the turbo encoder, if the difference between the number Ns of selected symbols in the initial puncturing matrix and the number of the columns in the initial puncturing matrix is equal to or greater than the number of component encoders in the turbo encoder. Then, the sub-code generator selects as many first and second parity symbols as the difference in such a way that the number of the selected first parity symbols is equal to or greater than the number of the selected second parity symbols. BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which: FIG. 1 is a graph showing the performance difference between packet code combining and packet diversity combining in a packet data system using turbo codes; FIG. 2 is a block diagram of a sub-code generating apparatus according to an embodiment of the present invention; FIG. 3 is a graph showing the performances of a retransmission scheme without using sub-codes, a retransmission scheme implementing diversity the embodiment of the present invention. DETAILED DESCRIPTION OF f HE PREFERRED EMBODIMENT A preferred embodiment of the present invention will be described hereinbelow with reference to the accompanying drawings. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail. The present invention proposes a method of generating sub-codes for optimal packet combining in a retransmission system using turbo codes and a system that selectively employs a packet code combining scheme and a packet diversity combining scheme according to the data rates. The advantages and performance gain of the proposed system will be described. First, there will be given a description of the operation of the system that selectively uses the packet code combining scheme and the packet diversity scheme combining according to the data rates. In a system using R=l/5 turbo codes, for example, packet code combining is applied until the overall code rate of codes produced by soft-combining of retransmitted packets reaches 1/5. For the subsequent retransmitted packets, packet diversity combining and then packet code combining are performed. If the first packet is transmitted at a data rate of 1/3, the required redundancy symbols are provided at a retransmission request to make the overall code rate 1/5. Thus, when a receiver receives both packets, the overall code rate becomes 1/5. Each of the following packets is repeated prior to transmission and the receiver performs packet diversity combining and then packet code combining on the retransmitted packets at the code rate 1/5. FIG. 1 is a graph showing the performance difference between packet code combining and packet diversity combining in the case of turbo codes. As shown in FIG. 1, a turbo code with a low data rate of 1/6 exhibits a greater performance gain than a turbo code with a high code rate of 1/3 with the same symbol energy Es and obtains a performance gain of 3dB from the packet code combining. Consequently, generation of R=l/3 turbo codes by packet code combining of R=l/6 sub-codes produces a gain that turbo codes with a code rate lower than 1/3 exhibits and a gain that code combining of different codes offers, contemporaneously. More specifically, for the same code symbol energy Es and the same given code rate, turbo codes provide performance close to the "Shannon Channel Capacity Limit" according to the code rates only if iteration decoding is fully implemented, unlike convolutional codes. It is known that a turbo code with a low code rate offers a greater performance gain than a turbo code with a high code rate with the same code symbol energy Es. For example, when R=l/3 is reduced to R=l/6, the performance difference can be estimated by analyzing a change in the "Shannon Channel Capacity Limit". The reason for assuming the same symbol energy irrespective of R=l/3 or 1/6 for the curves of FIG. 1 is that the same symbol energy Es is used for each retransmission in a hybrid ARQ system. If an R=l/3 code is repeated once and the two codes are packet-diversity-combined on an AWGN (Additive White Gaussian Noise) channel, a maximum gain of 3dB is obtained in terms of a symbol energy-to-noise ratio (Es/No). The same is the case with an R=l/6 code. Thus, a performance curve for the R=l/3 turbo code shifts to the left in parallel by a +3dB scale due to a packet diversity combining gain and a performance curve for the R= 1/6 turbo code also shifts to the left in parallel by a +3dB scale when the same symbol energy is given. Here, the performance curves are derived with respect to the energy-to-noise ratio (Eb/No) which is measured to compare code performances according to the code rates. As a consequence, the difference between the turbo code performance curves is equivalent to the performance difference between the packet diversity combining and the packet code combining. The performance difference according to the code rates can be estimated from the "Shannon Channel Capacity Limit" and a minimum performance difference can be obtained using a minimum required signal-to-noise ratio (SNR). In a system using turbo codes with a code rate R and a very large encoder block size L, a minimum Eb/No required to provide an error-free channel is expressed as (Equation Removed) According to the above equation, the minimum required Eb/No in AWGN at each code rate for the turbo codes is listed in Table 1 below, hi Table 1, a typical Eb/No indicates a required Eb/No for a bit error rate (BER) below 0.00001 when the encoding block size L of the turbo codes is 1024. (Table 1) (Table Removed) As shown in Table 1, required Eb/No are 0.86, 0.57, 0.0, -0.414, -0.55, -0.82, -0.975, and -1.084dB respectively, for the code rates of 3/4, 2/3, 1/2, 3/8, 1/3, 1/4, 1/5, and 1/6. An at least 0.53dB performance difference exists between a system using R=l/3 code and a system using R=l/6 code. This is a minimum performance difference based on the "Shannon Channel Capacity Limit". Considering implementation of a real decoder and system environment, the difference becomes wider. From a simulation, approximately 1.12dB performance difference was observed between a system using packet code combining for R=2/3 codes and a system using packet diversity combining for the R=2/3 codes. Table 2 shows the performance difference between packet code combining and packet diversity combining after one retransmission in a system with a sub-code code rate of 2/3. As shown in Table 2, a minimum performance difference is 1.12dB and the packet code combining scheme produces a higher performance gain in the system using turbo code. (Table 2) (Table Removed) As described above, the packet code combining scheme shows excellent performance in the retransmission system using turbo code. Therefore, the present invention provides a sub-code generating method for optimal packet code combining in the retransmission system using turbo code . Generation of sub-codes for packet code combining according to a predetermined rule produces the aforementioned code combining gain and maximizes the performance of a system requesting sub-codes of the same size for each retransmission. FIG. 2 is a block diagram, of a sub-code generating apparatus using turbo codes according to an embodiment of the present invention. As shown in FIG. 2, the sub-code generating apparatus includes a turbo encoder, a sub-code generator 204, and a controller 205. First, with regard to the turbo encoder, a first component encoder (or a first constituent encoder) 201 encodes an input information bit stream and outputs first code symbols, i.e., information symbols X and first parity symbols YO and Yl. An interleaver 202 interleaves the input information bit stream according to a predetermined rule. A second component encoder (or a second constituent encoder) 203 encodes the interleaved information bit stream and outputs second code symbols, i.e., information symbols X' and second parity symbols YO and Yl. Thus, the output symbols of the turbo encoder are the first and second code symbols. Since the information symbols X' generated from the second component encoder 203 are not transmitted in reality, the code rate of the turbo encoder is 1/5. The sub-code generator 204 generates sub-codes from the first and second code symbols received from the first and second component encoders 201 and 203 by puncturing and repetition under the control of the controller 205. The controller 205 stores puncturing (and repetition) matrixes generated from algorithms shown in FIGs. 4, 5 and 6 and outputs symbol selection signals for the puncturing matrixes to the sub-code generator 204. Then, the sub-code generator 204 selects a predetermined number of code symbols within a predetermined puncturing range according to the symbol selection signals. The reference characters as used here, X, YO, Yl, Y'O, and Y'l are defined as follows.* X: systematic code symbol or information symbol YO: redundancy symbol from the upper component encoder of the turbo encoder Yl: redundancy symbol from the upper component encoder of the turbo encoder Y'O: redundancy symbol from the lower component encoder of the turbo encoder Y' 1: redundancy symbol from the lower component encoder of the turbo encoder FIGs. 4, 5 and 6 are flowcharts illustrating sub-code (or puncturing matrix) generating procedures according to the embodiment of the present invention. Specifically, FIG. 4 illustrates a procedure of generating the first sub-code C0 in a sub-code set, FIG. 5 illustrates a procedure of generating middle sub-codes Q to Cs_2 in the sub-code set, and FIG. 6 illustrates a procedure of generating the last sub-code C in the sub-code set. Hereinbelow, ENC1 (referred to as first code symbols) indicate the information symbols X and the first parity symbols YO and Yl output from the first component encoder 201 and ENC2 (referred to as second code symbols) indicate the second parity symbols Y'O and Y'l output from the second component encoder 203. Referring to FIG. 4, a maximum code rate (Rmax) available to a transmitter is set in step 401. This value is mostly given according to the data rate used in the system. A minimum code rate (Rmin) is set to be an integer-multiple of Rmax (=k/n). Here, k is the number of input symbols and n is the number of output symbols. Although Rmin can be determined arbitrarily, it is usually 1/6, 1/7 or lower Because a coding gain is saturated due to the decrease of code rates at or below R=l/7 in turbo codes. In addition, the real code rate, i.e., mother code rate (R) of a decoder in a receiver, is determined. R is set to be greater than Rmin. In real system implementation, Rmax and Rmin are preset. In some sense, Rmax is the code rate of the sub-codes to be generated and Rmin is a target code rate after code combining of the sub-codes. In general, Rmin is the code rate of an encoder in the transmitter. In step 403, the number of sub-codes (S) is calculated by the following equation using Rmax and Rmin. Here, the number of the sub-codes or the number of puncturing matrixes is a minimum integer exceeding the ratio of Rmax to Rmin. (Equation Removed) where 11 represents a minimum integer equal to or larger than . A variable m is set to an initial value of 1 in step 405 and C (= mx k) is determined in step 407. C is the number of the columns of each puncturing matrix, determined by Rmax. For example, for Rmax = 3/4, C can be 3, 6, 9, ... and is set to a minimum available value for the first sub-code to be transmitted. Here, C is set to 3 for Rmax=3/4. In step 407, the number of symbols to be selected from the puncturing matrix ,Ns, is calculated by multiplying the variable m by the code length i.e., the number of code symbols n from Rmax=k/n. Ns is the number of selected symbols or the number of selected positions in each puncturing matrix and calculated by C/Rmax. In step 409, (Ns-C) is compared with the number of the component encoders of the turbo encoder in the transmitter. The present turbo encoder is generally provided with two component encoders. Thus, it is supposed that two component encoders are used. It is determined whether (Ns—C) is 2 or greater in step 409 because the turbo encoder has two component encoders connected in parallel with the interleaver interposed as shown in FIG. 2 unlike conventional encoders using other single codes. In other words, at least one parity symbol from each component encoder must be transmitted after the information symbols are all transmitted in order to preserve the characteristics inherent in the turbo encoder. If (Ns-C) is less than 2, only one symbol is selected from either the first parity symbol set or the second parity symbol set. From the perspective of turbo codes, either case may face problems. In the first case, sub-codes generated without second parity symbols are not turbo codes but convolutional codes with constraint length K=4 from an encoder having only the first component encoder and offers no interleaver gain that is available in the turbo encoder. On the other hand, in the second case , transmission of only systematic symbols without parity symbols from the first component encoder results in sub-codes with a code rate of 1. This is equivalent to an uncoded system without any coding gain. Accordingly, (Ns-C) must be equal to or greater than 2 to provide turbo encoder performance. If (Ns-C) is equal to or greater than 2 in step 409, C systematic information symbols are selected from the puncturing matrix in step 411 and the other symbols are selected according to a predetermined type . For type 1, the other symbols are selected from the first and second parity symbols by Eq. (3) in step 413. The number of selected first parity symbols is equal to or greater than that of selected second parity symbols. For example, if the number of the other symbols , (Ns-C), is 3, first and second parity symbols are selected by Eq. (3) and then one more symbol is selected from the first parity symbols. (Equation Removed) where LJ represents a maximum integer equal to or less than . For type 2, the other symbols are selected from the first and second parity symbols by Eq. (4) in step 415. If a and b are given as symbol distribution rates for the first parity symbols and the second parity symbols, respectively, as many symbols as a minimum integer equal to or larger than the ratio of a(Ns-C) to (a+b) are selected from the first parity symbols and as many symbols as a maximum integer equal to or less than the ratio of b(Ns-C) to (a+b) are selected from the second parity symbols. (Equation Removed) where a+b =1 and a and b indicate the symbol distribution ratios for ENCl and ENC2, respectively. If the condition given in step 409 is not satisfied, that is, (Ns-C) is less than 2, the variable m is incremented by 1 in step 417 and the procedure returns to step 407. The purpose of step 409 is to determine whether sub-codes capable of preserving the nature of turbo codes can be generated within the present puncturing range (the size of the puncturing matrix). If the nature of the turbo codes cannot be preserved, the puncturing range is enlarged in step 417. As described above, the initial puncturing matrix is so constructed that all information symbols are selected and at least one symbol is selected from each of the first and second parity symbol sets in the turbo encoder. Now there will be given a description of a middle puncturing matrix generation method referring to FIG. 5. By repeating the procedure of FIG. 5, puncturing matrices C, to Cs_2 are generated. Referring to FIG. 5, step 501 or 503 is performed according to a predetermined type. For type 1, Ns symbols are selected from the first and second parity symbol sets by Eq. (5) in step 501. Ns is the product of m and n given from Rmax (=k/n). The number of selected first parity symbols is equal to or greater than that of selected second parity symbols. Here, unselected symbols from the previous puncturing matrices are selected. (Equation Removed) For type 2, Ns symbols are selected from the first and second parity symbol sets according to predetermined ratios by Eq. (6) in step 503. If a and b are given as symbol distribution ratios for the first parity symbols and the second parity symbols, respectively, as many symbols as a minimum integer equal to or greater than the ratio of a(Ns) to (a+b) are selected from the first parity symbols and as many symbols as a maximum integer equal to or less than the ratio of b(Ns) to (a+b) are selected from the second parity symbols. Here, unselected symbols from the previous puncturing .matrices are selected. (Equation Removed) A last puncturing matrix Cs_! generating method will be described below referring to FIG. 6. Referring to FIG. 6, the remaining unselected symbols from the previous puncturing matrices are all selected in step 601. The number of the selected symbols is defined as Ns2. In step 603, a new Ns is defined by (Ns-Ns2). Since symbols at all positions are selected from the, puncturing matrices in the process of the operations shown in FIGs. 4, 5 and 6, the new Ns is the number of symbols to be repeatedly selected, hi step 605, it is determined whether the new Ns is greater than 0. If the new Ns is 0, the procedure ends. If it is greater than 0, as many symbols as the new Ns are selected repeatedly from the information symbols, hi other words, the selected symbols are transmitted repeatedly.. The above-described sub-code generation method according to the present invention will be made clear below with specific numerical citations. For Rmax=3/4 and R=l/5, Rmin=l/6 and S=6/(4/3)=4.5-»5. Thus, five puncturing matrices are produced. (Equation Removed) Since the code rate of sub-codes is 3/4 and the number of sub-codes is 5, the sub-codes have a code rate 3/20 ((l/S)xRmax=(l/5)x(3/4)=3/20) after code combining. This implies that for 3 information bits, a receiver receives 20 code symbols. However, since 15 symbols are generated from Sxn=5x4=20 and Sxk=5x3=15, 5 symbols among the 15 symbols are repeatedly transmitted. The repeated symbols are preferably information symbols. In the above example, if an information symbol X is repeated once in each sub-code, a decoder receives turbo codes with R=l/5 in which information symbols occur twice for each of S sub-codes when S sub-codes are all received. The resulting sub-codes from the procedures shown in FIGs. 4, 5 and 6 are a kind of complementary codes, but they are not in a strict sense of the term because repeated symbols exist and each sub-code exhibits a different characteristic. In view of the sub-codes being produced from turbo codes, they will be called quasi-complementary turbo codes (QCTCs). A hybrid ARQ system using QCTCs employs the following retransmission scheme The hybrid ARQ system is the major example using packet code combining. Packet code combining is available to the present hybrid ARQ systems, HARQ type I, Type II, and Type III. In these systems, a retransmission technique can be implemented using QCTCs. If a transport unit (TU) is defined as an information bit block being a basic unit for packet transmission, one sub-code Q is selected for each TU transmission in the hybrid systems. A retransmission unit and an initial transmission TU can be the same or different in size. For an initial transmission and each retransmission, the following QCTC set is used. From a QCTC Cq having a code set size S, a mother code C can be reconstructed, or a new code Cq with a lower code rate than the mother code rate Rm can be generated by combining (or code-combining) sub-codes Q (i = 0, 1, 2, . . . ,S-1). Here, the mother code has a minimum code rate available in the encoder. Then, the QCTC is defined as Original code C with code rate R=Rm or code C with code rate R=Rm (Equation Removed) where S is the number of sub-codes with a code rate of Ri and Rm is the mother code rate. The operation of a system transmitting TUs of the same size for an initial transmission and each retransmission using a QCTC will be described. Needless to say, the transmission scheme using different TUs can also be supported in the present invention. Here, the number of sub-codes S is 4 and the mother code rate R is 1/5. (Step 1) Transmission is performed on a TU basis and a sub-code Q of the QCTC is transmitted at the initial transmission and each retransmission. (Step 2) When the overall code rate of codes produced by soft combining of the initially transmitted and retransmitted packets is greater than 1/5, each sub-code Q of the QCTC is transmitted in the order of C0, C], Co, . . ., C at each retransmission request. This is packet code combining. (Step 3) When the overall code rate of codes produced by soft combining of the initially transmitted and retransmitted packets is less than or equal to 1/5, each sub-code Q of the QCTC is repeatedly transmitted in the order of C0, Cb C2, . . ., Cs_! at each retransmission request. This is packet diversity combining. (Step 4) The QCTC set size can be an arbitrary value, determined by the mother code rate. For R=l/5 and a sub-code code rate of 2/3 for retransmission, up to four sub-codes can be used. Table 3 below lists QCTC sets for forward traffic channel packet data rates that are expected to be available in the present IS-2000 1XEVDV system. Here, a mother code rate R=l/5 and a sub-code code rate R=2/3, 1/3, or 1/6. (Table 3) (Table Removed) As shown in Table 3, for a sub-code code rate of 1/6 less than the mother code rate 1/5, the same code C0 is used at each transmission. For a sub-code code rate of 1/3, greater than the mother code rate 1/5, different code C0 and C] are used at each transmission. In this case, the code set size S is 2. For a subcode code rate of 2/3, greater than the mother code rate 1/5, different code C0, Cl5 C2, C3 are used at each transmission. The code set size S is 4. When S subcodes are all transmitted, the receiver can recover the mother code rate R and obtain the maximum coding gain offered by the encoder. Table 4 below illustrates examples of puncturing matrices for each sub-code code rate. (Table 4) (Table Removed) As shown in Table 4, when a rate 1/5 turbo code is used as a mother code and a rate 2/3 sub-code is generated with code symbols output for 4 information bits, 20 code symbols are generated from the 4 information bits. The rate 2/3 sub-code is generated by puncturing 14 symbols among the 20 symbols. For packet diversity combining of such sub-codes, C0 produced from the above puncturing matrices is repeatedly transmitted at each retransmission request. On the other hand, for packet code combining, a different code symbol is transmitted at each retransmission request. After transmitting all sub-codes C0, C], C2, C3 in the set, the packet diversity combining is executed. For HARQ Type III using packet code combining, full code symbols of the mother code are decoded after four transmissions occur. Meanwhile, "l"s in the puncturing matrices of Table 4 indicate that the symbols at the positions are selected or transmitted and "0"s indicate that the symbols at the positions are punctured. A "2" indicates that the symbol at the position occurs twice. The puncturing (and repetition) matrices are designed to satisfy the following conditions. (Condition 1) An information symbol X is repeated in a sub-code of a QCTC when repetition is used. (Condition 2) If the information symbol X is repeated in a sub-code of the QCTC using repetition, the repeating period is set to be a minimal constant in the QCTC having all sub-codes in combination. (Condition 3) If puncturing is used, redundancy symbols except the information symbol X are punctured if possible in the sub-codes of the QCTC. (Condition 4) If puncturing is used, redundancy symbols except the information symbol X are uniformly punctured if possible in the sub-codes of the QCTC. A puncturing and repetition matrix with R=l/6 satisfying the above conditions will be described. Decoding is carried out after the symbols X that occur twice are soft-combined and so the real code rate for the decoder is 1/5. The rate 1/5 code, having the energy of the information symbol X increased, has an improved performance, as compared to a rate 1/5 code having a uniform symbol energy across the symbols. In other words, the most appropriate symbol to be repeated is an information symbol. It can be said that the pimcturing and repetition matrices shown in Table 4 are constructed in such a way that information symbol energy is increased through uniform repetition of the information symbols. In Table 4, for R=l/6, the sequence of transmission code symbols is given as C0: X, X, YO, Yl, Y'O, Y'l, X, X, YO, Yl, Y'O, Y'l, . . . Because six code symbols are generated for the input of one information symbol, the code rate of the sub-code is 1/6. For R=l/3, the sequences of transmission code symbols are given as C0: X, YO, Y'O, X, YO, Y'O, X YO, Y'O, X, YO, Y'O,... C,: X, Yl, Y' 1, X, Yl, Y' 1, X, Yl, Y' 1, X, Yl, Y' 1, . . . Because three code symbols are generated for the input of one information symbol, the code rate of the sub-code is 1/3. A different code is transmitted at each transmission due to the use of a different puncturing matrix. After soft combining of C0 and Cb X occurs twice and each of YO, Yl, Y'O and Y' 1 occurs once. Thus, the decoder with a code rate 1/5 can be used in this case and the puncturing matrices satisfy the above-described conditions, ensuring performance. In the first case with R=2/3 of Table 4, the sequence of transmission code symbols are given as (Sequence Removed) Because three code symbols are generated for the input of two information symbols, the code rate of the sub-code is 2/3. A different code is transmitted at each transmission due to the use of a different puncturing matrix. After soft combining of C0, Cb C2, Q, X occurs twice and each of YO, Yl, Y'O and Y' 1 occurs once. Thus, the decoder with a code rate 1/5 can be used also in this case as for R=l/6 and the puncturing matrices satisfy the above-described conditions, ensuring performance. In the second case with R=2/3 of Table 4, the sequence of transmission code symbols are given as (SequenceRemoved) Because six code symbols are generated for the input of four information symbols, the code rate of the sub-code is 2/3. A different code is transmitted at each transmission due to the use of a different puncturing matrix. After soft combining of C0, Cb C2, C3, X occurs twice and each of YO, Yl, Y'O and Y' 1 occurs once. Thus, the decoder with a code rate 1/5 can be used also in this case as for R=l/6 and the puncturing matrices satisfy the above-described conditions, ensuring performance. FIG. 3 is a graph showing a comparison between the performance of HARQ using packet code combining and the performance of HARQ using packet diversity combining in terms of data throughput for a QCTC with R=2/3 and S=4 according to the present invention. As shown in FIG. 3, an HARQ 301 using packet code combining for the QCTC and an HARQ 302 using packet diversity combining for the QCTC show better performance than an HARQ 303 without the QCTC. For the same real time data throughput (e. g., 0.25), about -4dB of Es/No is required in the HARQ 301, about -1.3dB in the HARQ 302, and about IdB in the HARQ 303. Consequently, the use of the QCTC according to the present invention ensures a higher data throughput with less symbol energy. As described above, sub-codes generated for optimum packet code combining according to a predetermined rule remarkably improves transmission throughput in a packet retransmission system using turbo codes according to the present invention. While the invention has been shown and described with reference to a certain preferred embodiment thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Claim 1. A method of generating an initial puncturing matrix from which a first sub-code is produced in a communication system having a turbo encoder for generating information symbols, first parity symbols, and second parity symbols for the input of an information bit stream and a sub-code generator for generating sub-code from the information symbols, the first parity symbols, and the second parity symbols using puncturing matrices, a number of the cub-codes being identical to a number of the puncturing matrices, comprising steps of: selecting as many information symbols as a number of columns in the initial puncturing matrix from the information symbols output from the turbo encoder, if a different between a number Ns of selected symbols in the initial puncturing matrix and the number of the columns in the initial puncturing matrix is equal to or greater than a number of component encoders in the turbo encoder ; and selecting as many first and second parity symbols as the difference , a number of the selected first parity symbols being equal to or greater than a number of the selected second arity symbols. 2. The method as claimed in claim 1, further comprising the step of increasing the number of the columns in the puncturing matrix by an integer multiple if the difference is less than the number of the component encoders. 3. The method as claimed in claim 1, further comprising the step of selecting the Ns symbols from unselected first and second parity symbols in the initial puncturing matrix in generating a second puncturing matrix, wherein the number of the selected second parity symbols is equal to or greater than the number of the selected second parity symbols. 4. The method as claimed in claim 3, further comprising the step of selecting the Ns symbols from unselected first and second parity symbols in the other puncturing matrices except for a last puncturing matrix and repeating (Ns-Ns2) information symbols, Ns2 being a number of the other unselected parity symbols in the other puncturing matrices. 5. The method as claimed in claim 1, wherein the communication system user the sub code to hybrid ARQ (Automatic Repeat Request). 6. The method as claimed in 1, wherein the number Ns of selected symbols is determined by C/Rmax, C being the number of the columns and Rmax being a code rate of the sub-code. 7. A method of generating a first sub-code to be transmitted by selecting a predetermined number of symbols from information symbols and first and second parity symbols within a predetermined puncturing range in a communication system having a turbo encoder for generating information symbols, first parity symbols, and second parity symbols for the input of an information bit stream, the predetermined number of symbols being Ns symbols, comprising the steps of: selecting all the information symbols within the puncturing range, if a difference between Ns and a number of the information symbols is equal to or greater than a number of component encoders in the turbo encoder : and selecting as many first and second parity symbols as the difference, number of the selected first parity symbols being equal to or greater than a number of the selected second parity symbols. 8. The method as claimed in claim 7, further comprising the step of increasing the puncturing range by an integer multiple if the difference is less than the number of the component encoders. 9. The method as claimed in claim 7, further comprising the steps of selecting the Ns symbols from unselected first and second parity symbols in the first sub-code in generating a second sub-code, wherein the number of the selected first parity symbols is equal to or greater than the number of the selected second parity symbols. 10. The method as claimed in claim 9, further comprising the step of selecting remaining unselected first and second parity symbols in the other sub-codes except for a last sub-code and repeating (Ns-Ns2) information symbols, Ns2 being a number of the other unselected parity symbols in generating the last sub-code. 11. The method as claimed in claim 7, wherein the communication system uses the sub code to hybrid ARQ (Automatic Repeat Request). 12 The method as claimed in claim 7, wherein the number Ns of selected symbols is determined by C/Rmax, C being a number of columns in a puncturing matrix and Rmax being a code of the sub-code. 13. A method of generating puncturing matrices by which information symbols and first and second parity symbols are punctured in a communication system having a turbo encoder for generating information symbols, first parity symbols, and second parity symbols for the input of an information bit stream, and a sub-code generator for generating sub-codes from the information symbols, the first parity symbols, and the second parity symbols by puncturing, comprising the steps of: determining a number of S of the puncturing matrices by calculating a minimum integer equal to of greater than a ratio of Rmax to Rmin, Rmax being a given maximum code rate and Rmin being a given minimum code rate for the turbo encoder ; determining a number Ns of symbols to be selected from each puncturing matrix by C/Rmax, C being a number of columns of the puncturing matrix; and generating a first puncturing matrix for which C information symbols are selected, as many first parity symbols as a minimum integer equal to or greater than a ratio of a(Ns-C) to (a+b) are selected, and as many second parity symbols as a maximum integer equal to or less than a ratio of b(Ns-C) to (a+b) are selected, if a difference between Ns and C, (Ns-C), is equal to or greater than a number of component encoders in the turbo encoder, a and b being symbol distribution ratios for the first and second parity symbols. 14. The method as claimed in claim 13, further comprising the steps of increasing the number of the columns in the puncturing matrix by an integer multiple if the difference is less than the number of the component. 15. The method as claimed in claim 13, further comprising the step of generating middle puncturing matrices except for the first and a last puncturing matrices by selecting as many first parity for the first parity symbols as a minimum integer equal to or greater than a ratio of aNs to (a+b) and as many second parity symbols as a maximum integer equal to or less than a ratio of bNs to (a+b) without selecting any information symbols, so that the first and second parity symbols differ in the first and other puncturing matrices. 16 The method as claimed in claim 15, further comprising the step of generating the last puncturing matrix by selecting remaining unselected first and second parity symbols in the other puncturing matrices except for the last puncturing matrix and repeating (Ns-Ns2) information symbols, Ns2 being a number of the other unselected parity symbols in the other puncturing matrices. 17. The method as claimed in claim 13, wherein the communication system uses the sub-codes to hybrid ARQ (Automatic Repeat Request). 18. A method of generating an initial puncturing matrix from which a first sub-code is produced in a communication system substantially as herein described with reference to the foregoing description, and the accompanying drawings. 19. A method of generating a first sub-code to be transmitted by selecting a predetermined number of symbols from information symbols and first and second parity symbols within a predetermined puncturing range in a communication system substantially as herein described with reference to the foregoing description, tables and the accompanying drawings. 20. A method of generating puncturing matrices by which information symbols and first and second parity symbols are punctured in a communication system substantially as herein described with reference to the foregoing description', tables and the accompanying drawings.

Full Text

GENERATING CODES IN A COMMUNICATION SYSTEM
BACKGROUND OF THE INVENTION
I Field of the Invention
The present invention relates generally to an apparatus and method for generating codes in a data communications system, and in particular, to an apparatus and method for generating complementary codes with use of turbo codes in a communications system employing a retransmission scheme.
2. Description of the Related Art
In general, a system using a retransmission scheme (e.g., hybrid ARQ [Automatic Repeat Request]) performs soft combining to improve transmission throughput. The soft combining techniques are divided into packet diversity combining and packet code combining. These two combining schemes are usually called soft packet combining. Although the packet diversity combining scheme is sub-optimal in performance relative to the packet code combining scheme, it is favorable due to easy implementation when performance loss is low.
A packet transmission system uses the packet code combining scheme to improve transmission throughput. That is, a transmitter transmits a code with a different code rate at each packet transmission. If an error is detected from the received packet, a receiver requests a retransmission and performs soft combining between the original packet and a retransmitted packet. The retransmitted packet may have a different code from the previous packet. The packet code combining scheme is a process of combining received N packets with a code rate R to a code with an effective code rate of R/N prior to decoding, to thereby obtain a coding gain.
With regard to the packet diversity combining scheme, on the other hand, the transmitter transmits a code with the same code rate R at each packet transmission. If an error is detected from the received packet, the receiver requests a retransmission and performs soft combining between the original packet and the retransmitted packet. The retransmitted packet has an identical code to that in the previous packet. In this sense, the packet diversity
combining scheme can be considered symbol averaging on a random channel. The packet diversity combining scheme reduces noise power by averaging the soft outputs of received symbols and achieves such a diversity gain as offered by a multipath channel because the same code is repeatedly transmitted on a fading channel. However, the packet diversity combining scheme does not provide such an additional coding gain as obtained according to a code structure in the packet code combining scheme.
Due to implementation simplicity, most packet communication systems have used the packet diversity combining scheme, which is under study for application to the synchronous IS-2000 system and the asynchronous UMTS system. The reason is that the existing packet communication systems have used convolutional codes and even packet code combining does not offer a great gain when convolutional codes with a low code rate are used. If a system with R=l/3 supports retransmission, there is not a wide difference in performance between the packet code combining scheme and the packet diversity combining scheme. Thus, the packet diversity combining scheme is selected considering implementation complexity. However, the use of turbo codes as forward error correction codes (FEC) requires a different packet combining mechanism because the turbo codes are designed as error correction codes to have performance characteristics very close to the "Shannon Channel Capacity Limit" by iterative decoding and their performance varies obviously with the coding rates unlike convolutional codes. Therefore, it can be concluded that packet code combining is feasible for a packet communication system using turbo codes in a retransmission scheme to achieve the goal of optimum performance.
SUMMARY OF THE INVENTION
It is, therefore, an object of the present invention to provide an apparatus and method for generating sub-codes for optimal code combination in a retransmission system using turbo codes.
It is another object of the present invention to provide an apparatus and method for generating complementary codes using turbo codes in a communications system.
The above and other objects can be achieved by providing an apparatus and method for generating sub-codes from turbo codes in a communications
system. In the sub-code generating apparatus, a turbo encoder generates information symbols, first parity symbols, and second parity symbols for the input of an information bit stream and a sub-code generator generates sub-codes from the information symbols, the first parity symbols, and the second parity symbols using puncturing matrixes. Here, the number of the sub-codes is identical to the number of the puncturing matrixes. To generate an initial puncturing matrix for the first sub-code, the sub-code generator selects as many information symbols as the number of columns in the initial puncturing matrix from the information symbols from the turbo encoder, if the difference between the number Ns of selected symbols in the initial puncturing matrix and the number of the columns in the initial puncturing matrix is equal to or greater than the number of component encoders in the turbo encoder. Then, the sub-code generator selects as many first and second parity symbols as the difference in such a way that the number of the selected first parity symbols is equal to or greater than the number of the selected second parity symbols.
BRIEF DESCRIPTION OF THE DRAWINGS
The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which:
FIG. 1 is a graph showing the performance difference between packet code combining and packet diversity combining in a packet data system using turbo codes;
FIG. 2 is a block diagram of a sub-code generating apparatus according to an embodiment of the present invention;
FIG. 3 is a graph showing the performances of a retransmission scheme without using sub-codes, a retransmission scheme implementing diversity
the embodiment of the present invention.
DETAILED DESCRIPTION OF f HE PREFERRED EMBODIMENT
A preferred embodiment of the present invention will be described hereinbelow with reference to the accompanying drawings. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail.
The present invention proposes a method of generating sub-codes for optimal packet combining in a retransmission system using turbo codes and a system that selectively employs a packet code combining scheme and a packet diversity combining scheme according to the data rates. The advantages and performance gain of the proposed system will be described.
First, there will be given a description of the operation of the system that selectively uses the packet code combining scheme and the packet diversity scheme combining according to the data rates.
In a system using R=l/5 turbo codes, for example, packet code combining is applied until the overall code rate of codes produced by soft-combining of retransmitted packets reaches 1/5. For the subsequent retransmitted packets, packet diversity combining and then packet code combining are performed. If the first packet is transmitted at a data rate of 1/3, the required redundancy symbols are provided at a retransmission request to make the overall code rate 1/5. Thus, when a receiver receives both packets, the overall code rate becomes 1/5. Each of the following packets is repeated prior to transmission and the receiver performs packet diversity combining and then packet code combining on the retransmitted packets at the code rate 1/5.
FIG. 1 is a graph showing the performance difference between packet code combining and packet diversity combining in the case of turbo codes. As shown in FIG. 1, a turbo code with a low data rate of 1/6 exhibits a greater performance gain than a turbo code with a high code rate of 1/3 with the same symbol energy Es and obtains a performance gain of 3dB from the packet code combining. Consequently, generation of R=l/3 turbo codes by packet code combining of R=l/6 sub-codes produces a gain that turbo codes with a code rate lower than 1/3 exhibits and a gain that code combining of different codes
offers, contemporaneously.
More specifically, for the same code symbol energy Es and the same given code rate, turbo codes provide performance close to the "Shannon Channel Capacity Limit" according to the code rates only if iteration decoding is fully implemented, unlike convolutional codes. It is known that a turbo code with a low code rate offers a greater performance gain than a turbo code with a high code rate with the same code symbol energy Es. For example, when R=l/3 is reduced to R=l/6, the performance difference can be estimated by analyzing a change in the "Shannon Channel Capacity Limit". The reason for assuming the same symbol energy irrespective of R=l/3 or 1/6 for the curves of FIG. 1 is that the same symbol energy Es is used for each retransmission in a hybrid ARQ system.
If an R=l/3 code is repeated once and the two codes are packet-diversity-combined on an AWGN (Additive White Gaussian Noise) channel, a maximum gain of 3dB is obtained in terms of a symbol energy-to-noise ratio (Es/No). The same is the case with an R=l/6 code. Thus, a performance curve for the R=l/3 turbo code shifts to the left in parallel by a +3dB scale due to a packet diversity combining gain and a performance curve for the R= 1/6 turbo code also shifts to the left in parallel by a +3dB scale when the same symbol energy is given. Here, the performance curves are derived with respect to the energy-to-noise ratio (Eb/No) which is measured to compare code performances according to the code rates. As a consequence, the difference between the turbo code performance curves is equivalent to the performance difference between the packet diversity combining and the packet code combining. The performance difference according to the code rates can be estimated from the "Shannon Channel Capacity Limit" and a minimum performance difference can be obtained using a minimum required signal-to-noise ratio (SNR).
In a system using turbo codes with a code rate R and a very large encoder block size L, a minimum Eb/No required to provide an error-free channel is expressed as
(Equation Removed)
According to the above equation, the minimum required Eb/No in AWGN at each code rate for the turbo codes is listed in Table 1 below, hi Table 1, a typical Eb/No indicates a required Eb/No for a bit error rate (BER) below 0.00001 when the encoding block size L of the turbo codes is 1024.
(Table 1)

(Table Removed)
As shown in Table 1, required Eb/No are 0.86, 0.57, 0.0, -0.414, -0.55, -0.82, -0.975, and -1.084dB respectively, for the code rates of 3/4, 2/3, 1/2, 3/8, 1/3, 1/4, 1/5, and 1/6. An at least 0.53dB performance difference exists between a system using R=l/3 code and a system using R=l/6 code. This is a minimum performance difference based on the "Shannon Channel Capacity Limit". Considering implementation of a real decoder and system environment, the difference becomes wider. From a simulation, approximately 1.12dB performance difference was observed between a system using packet code combining for R=2/3 codes and a system using packet diversity combining for the R=2/3 codes.
Table 2 shows the performance difference between packet code combining and packet diversity combining after one retransmission in a system with a sub-code code rate of 2/3. As shown in Table 2, a minimum performance difference is 1.12dB and the packet code combining scheme produces a higher performance gain in the system using turbo code. (Table 2)
(Table Removed)

As described above, the packet code combining scheme shows excellent performance in the retransmission system using turbo code. Therefore, the present invention provides a sub-code generating method for optimal packet code combining in the retransmission system using turbo code . Generation of sub-codes for packet code combining according to a predetermined rule produces the aforementioned code combining gain and maximizes the performance of a system requesting sub-codes of the same size for each retransmission.
FIG. 2 is a block diagram, of a sub-code generating apparatus using turbo codes according to an embodiment of the present invention. As shown in FIG. 2, the sub-code generating apparatus includes a turbo encoder, a sub-code generator 204, and a controller 205.
First, with regard to the turbo encoder, a first component encoder (or a first constituent encoder) 201 encodes an input information bit stream and
outputs first code symbols, i.e., information symbols X and first parity symbols YO and Yl. An interleaver 202 interleaves the input information bit stream according to a predetermined rule. A second component encoder (or a second constituent encoder) 203 encodes the interleaved information bit stream and outputs second code symbols, i.e., information symbols X' and second parity symbols YO and Yl. Thus, the output symbols of the turbo encoder are the first and second code symbols. Since the information symbols X' generated from the second component encoder 203 are not transmitted in reality, the code rate of the turbo encoder is 1/5.
The sub-code generator 204 generates sub-codes from the first and second code symbols received from the first and second component encoders 201 and 203 by puncturing and repetition under the control of the controller 205. The controller 205 stores puncturing (and repetition) matrixes generated from algorithms shown in FIGs. 4, 5 and 6 and outputs symbol selection signals for the puncturing matrixes to the sub-code generator 204. Then, the sub-code generator 204 selects a predetermined number of code symbols within a predetermined puncturing range according to the symbol selection signals.
The reference characters as used here, X, YO, Yl, Y'O, and Y'l are defined as follows.*
X: systematic code symbol or information symbol
YO: redundancy symbol from the upper component encoder of the turbo encoder
Yl: redundancy symbol from the upper component encoder of the turbo encoder
Y'O: redundancy symbol from the lower component encoder of the turbo encoder
Y' 1: redundancy symbol from the lower component encoder of the turbo encoder
FIGs. 4, 5 and 6 are flowcharts illustrating sub-code (or puncturing matrix) generating procedures according to the embodiment of the present invention. Specifically, FIG. 4 illustrates a procedure of generating the first sub-code C0 in a sub-code set, FIG. 5 illustrates a procedure of generating middle sub-codes Q to Cs_2 in the sub-code set, and FIG. 6 illustrates a procedure of generating the last sub-code C in the sub-code set.
Hereinbelow, ENC1 (referred to as first code symbols) indicate the information symbols X and the first parity symbols YO and Yl output from the first component encoder 201 and ENC2 (referred to as second code symbols) indicate the second parity symbols Y'O and Y'l output from the second component encoder 203.
Referring to FIG. 4, a maximum code rate (Rmax) available to a transmitter is set in step 401. This value is mostly given according to the data rate used in the system. A minimum code rate (Rmin) is set to be an integer-multiple of Rmax (=k/n). Here, k is the number of input symbols and n is the number of output symbols. Although Rmin can be determined arbitrarily, it is usually 1/6, 1/7 or lower Because a coding gain is saturated due to the decrease of code rates at or below R=l/7 in turbo codes. In addition, the real code rate, i.e., mother code rate (R) of a decoder in a receiver, is determined. R is set to be greater than Rmin.
In real system implementation, Rmax and Rmin are preset. In some sense, Rmax is the code rate of the sub-codes to be generated and Rmin is a target code rate after code combining of the sub-codes. In general, Rmin is the code rate of an encoder in the transmitter.
In step 403, the number of sub-codes (S) is calculated by the following equation using Rmax and Rmin. Here, the number of the sub-codes or the number of puncturing matrixes is a minimum integer exceeding the ratio of Rmax to Rmin.
(Equation Removed)
where 1*1 represents a minimum integer equal to or larger than *.
A variable m is set to an initial value of 1 in step 405 and C (= mx k) is determined in step 407. C is the number of the columns of each puncturing matrix, determined by Rmax. For example, for Rmax = 3/4, C can be 3, 6, 9, ... and is set to a minimum available value for the first sub-code to be transmitted. Here, C is set to 3 for Rmax=3/4.
In step 407, the number of symbols to be selected from the puncturing matrix ,Ns, is calculated by multiplying the variable m by the code length i.e.,
the number of code symbols n from Rmax=k/n. Ns is the number of selected symbols or the number of selected positions in each puncturing matrix and calculated by C/Rmax.
In step 409, (Ns-C) is compared with the number of the component encoders of the turbo encoder in the transmitter. The present turbo encoder is generally provided with two component encoders. Thus, it is supposed that two component encoders are used. It is determined whether (Ns—C) is 2 or greater in step 409 because the turbo encoder has two component encoders connected in parallel with the interleaver interposed as shown in FIG. 2 unlike conventional encoders using other single codes. In other words, at least one parity symbol from each component encoder must be transmitted after the information symbols are all transmitted in order to preserve the characteristics inherent in the turbo encoder.
If (Ns-C) is less than 2, only one symbol is selected from either the first parity symbol set or the second parity symbol set. From the perspective of turbo codes, either case may face problems. In the first case, sub-codes generated without second parity symbols are not turbo codes but convolutional codes with constraint length K=4 from an encoder having only the first component encoder and offers no interleaver gain that is available in the turbo encoder. On the other hand, in the second case , transmission of only systematic symbols without parity symbols from the first component encoder results in sub-codes with a code rate of 1. This is equivalent to an uncoded system without any coding gain. Accordingly, (Ns-C) must be equal to or greater than 2 to provide turbo encoder performance.
If (Ns-C) is equal to or greater than 2 in step 409, C systematic information symbols are selected from the puncturing matrix in step 411 and the other symbols are selected according to a predetermined type . For type 1, the other symbols are selected from the first and second parity symbols by Eq. (3) in step 413. The number of selected first parity symbols is equal to or greater than that of selected second parity symbols. For example, if the number of the other symbols , (Ns-C), is 3, first and second parity symbols are selected by Eq. (3) and then one more symbol is selected from the first parity symbols.
(Equation Removed)
where L*J represents a maximum integer equal to or less than *.
For type 2, the other symbols are selected from the first and second parity symbols by Eq. (4) in step 415. If a and b are given as symbol distribution rates for the first parity symbols and the second parity symbols, respectively, as many symbols as a minimum integer equal to or larger than the ratio of a(Ns-C) to (a+b) are selected from the first parity symbols and as many symbols as a maximum integer equal to or less than the ratio of b(Ns-C) to (a+b) are selected from the second parity symbols.
(Equation Removed)
where a+b =1 and a and b indicate the symbol distribution ratios for ENCl and
ENC2, respectively.
If the condition given in step 409 is not satisfied, that is, (Ns-C) is less than 2, the variable m is incremented by 1 in step 417 and the procedure returns to step 407. The purpose of step 409 is to determine whether sub-codes capable of preserving the nature of turbo codes can be generated within the present puncturing range (the size of the puncturing matrix). If the nature of the turbo codes cannot be preserved, the puncturing range is enlarged in step 417.
As described above, the initial puncturing matrix is so constructed that all information symbols are selected and at least one symbol is selected from each of the first and second parity symbol sets in the turbo encoder.
Now there will be given a description of a middle puncturing matrix generation method referring to FIG. 5. By repeating the procedure of FIG. 5, puncturing matrices C, to Cs_2 are generated.
Referring to FIG. 5, step 501 or 503 is performed according to a predetermined type. For type 1, Ns symbols are selected from the first and second parity symbol sets by Eq. (5) in step 501. Ns is the product of m and n given from Rmax (=k/n). The number of selected first parity symbols is equal to or greater than that of selected second parity symbols. Here, unselected symbols from the previous puncturing matrices are selected.
(Equation Removed)
For type 2, Ns symbols are selected from the first and second parity symbol sets according to predetermined ratios by Eq. (6) in step 503. If a and b are given as symbol distribution ratios for the first parity symbols and the second parity symbols, respectively, as many symbols as a minimum integer equal to or greater than the ratio of a(Ns) to (a+b) are selected from the first parity symbols and as many symbols as a maximum integer equal to or less than the ratio of b(Ns) to (a+b) are selected from the second parity symbols. Here, unselected symbols from the previous puncturing .matrices are selected.
(Equation Removed)
A last puncturing matrix Cs_! generating method will be described below referring to FIG. 6.
Referring to FIG. 6, the remaining unselected symbols from the previous puncturing matrices are all selected in step 601. The number of the selected symbols is defined as Ns2. In step 603, a new Ns is defined by (Ns-Ns2). Since symbols at all positions are selected from the, puncturing matrices in the process of the operations shown in FIGs. 4, 5 and 6, the new Ns is the number of symbols to be repeatedly selected, hi step 605, it is determined whether the new Ns is greater than 0. If the new Ns is 0, the procedure ends. If it is greater than 0, as many symbols as the new Ns are selected repeatedly from the information symbols, hi other words, the selected symbols are transmitted repeatedly..
The above-described sub-code generation method according to the present invention will be made clear below with specific numerical citations.
For Rmax=3/4 and R=l/5, Rmin=l/6 and S=6/(4/3)=4.5-»5. Thus, five puncturing matrices are produced.
(Equation Removed)

Since the code rate of sub-codes is 3/4 and the number of sub-codes is 5, the sub-codes have a code rate 3/20 ((l/S)xRmax=(l/5)x(3/4)=3/20) after code combining. This implies that for 3 information bits, a receiver receives 20 code symbols. However, since 15 symbols are generated from Sxn=5x4=20 and Sxk=5x3=15, 5 symbols among the 15 symbols are repeatedly transmitted. The repeated symbols are preferably information symbols. In the above example, if an information symbol X is repeated once in each sub-code, a decoder receives turbo codes with R=l/5 in which information symbols occur twice for each of S sub-codes when S sub-codes are all received.
The resulting sub-codes from the procedures shown in FIGs. 4, 5 and 6 are a kind of complementary codes, but they are not in a strict sense of the term because repeated symbols exist and each sub-code exhibits a different characteristic. In view of the sub-codes being produced from turbo codes, they will be called quasi-complementary turbo codes (QCTCs). A hybrid ARQ system using QCTCs employs the following retransmission scheme
The hybrid ARQ system is the major example using packet code combining. Packet code combining is available to the present hybrid ARQ systems, HARQ type I, Type II, and Type III. In these systems, a retransmission technique can be implemented using QCTCs. If a transport unit (TU) is defined as an information bit block being a basic unit for packet transmission, one sub-code Q is selected for each TU transmission in the hybrid systems.
A retransmission unit and an initial transmission TU can be the same or different in size. For an initial transmission and each retransmission, the following QCTC set is used.
From a QCTC Cq having a code set size S, a mother code C can be reconstructed, or a new code Cq with a lower code rate than the mother code rate Rm can be generated by combining (or code-combining) sub-codes Q (i = 0, 1, 2, . . . ,S-1). Here, the mother code has a minimum code rate available in the encoder. Then, the QCTC is defined as
Original code C with code rate R=Rm or code C with code rate R=Rm
(Equation Removed)
where S is the number of sub-codes with a code rate of Ri and Rm is the mother
code rate.
The operation of a system transmitting TUs of the same size for an initial transmission and each retransmission using a QCTC will be described. Needless to say, the transmission scheme using different TUs can also be supported in the present invention. Here, the number of sub-codes S is 4 and the mother code rate R is 1/5.
(Step 1) Transmission is performed on a TU basis and a sub-code Q of the QCTC is transmitted at the initial transmission and each retransmission.
(Step 2) When the overall code rate of codes produced by soft combining of the initially transmitted and retransmitted packets is greater than 1/5, each sub-code Q of the QCTC is transmitted in the order of C0, C], Co, . . ., C at each retransmission request. This is packet code combining.
(Step 3) When the overall code rate of codes produced by soft combining of the initially transmitted and retransmitted packets is less than or equal to 1/5, each sub-code Q of the QCTC is repeatedly transmitted in the order of C0, Cb C2, . . ., Cs_! at each retransmission request. This is packet diversity combining.
(Step 4) The QCTC set size can be an arbitrary value, determined by the mother code rate. For R=l/5 and a sub-code code rate of 2/3 for retransmission, up to four sub-codes can be used.
Table 3 below lists QCTC sets for forward traffic channel packet data rates that are expected to be available in the present IS-2000 1XEVDV system. Here, a mother code rate R=l/5 and a sub-code code rate R=2/3, 1/3, or 1/6.
(Table 3)

(Table Removed)

As shown in Table 3, for a sub-code code rate of 1/6 less than the mother code rate 1/5, the same code C0 is used at each transmission. For a sub-code code rate of 1/3, greater than the mother code rate 1/5, different code C0 and C] are used at each transmission. In this case, the code set size S is 2. For a subcode code rate of 2/3, greater than the mother code rate 1/5, different code C0, Cl5 C2, C3 are used at each transmission. The code set size S is 4. When S subcodes are all transmitted, the receiver can recover the mother code rate R and obtain the maximum coding gain offered by the encoder.
Table 4 below illustrates examples of puncturing matrices for each sub-code code rate.
(Table 4)

(Table Removed)

As shown in Table 4, when a rate 1/5 turbo code is used as a mother code and a rate 2/3 sub-code is generated with code symbols output for 4 information bits, 20 code symbols are generated from the 4 information bits. The rate 2/3 sub-code is generated by puncturing 14 symbols among the 20 symbols. For packet diversity combining of such sub-codes, C0 produced from the above puncturing matrices is repeatedly transmitted at each retransmission request. On the other hand, for packet code combining, a different code symbol is transmitted at each retransmission request. After transmitting all sub-codes C0, C], C2, C3 in the set, the packet diversity combining is executed. For HARQ Type III using packet code combining, full code symbols of the mother code are decoded after four transmissions occur.
Meanwhile, "l"s in the puncturing matrices of Table 4 indicate that the symbols at the positions are selected or transmitted and "0"s indicate that the symbols at the positions are punctured. A "2" indicates that the symbol at the position occurs twice. The puncturing (and repetition) matrices are designed to satisfy the following conditions.
(Condition 1) An information symbol X is repeated in a sub-code of a QCTC when repetition is used.
(Condition 2) If the information symbol X is repeated in a sub-code of the QCTC using repetition, the repeating period is set to be a minimal constant in the QCTC having all sub-codes in combination.
(Condition 3) If puncturing is used, redundancy symbols except the
information symbol X are punctured if possible in the sub-codes of the QCTC.
(Condition 4) If puncturing is used, redundancy symbols except the information symbol X are uniformly punctured if possible in the sub-codes of the QCTC.
A puncturing and repetition matrix with R=l/6 satisfying the above conditions will be described. Decoding is carried out after the symbols X that occur twice are soft-combined and so the real code rate for the decoder is 1/5. The rate 1/5 code, having the energy of the information symbol X increased, has an improved performance, as compared to a rate 1/5 code having a uniform symbol energy across the symbols. In other words, the most appropriate symbol to be repeated is an information symbol. It can be said that the pimcturing and repetition matrices shown in Table 4 are constructed in such a way that information symbol energy is increased through uniform repetition of the information symbols.
In Table 4, for R=l/6, the sequence of transmission code symbols is given as
C0: X, X, YO, Yl, Y'O, Y'l, X, X, YO, Yl, Y'O, Y'l, . . .
Because six code symbols are generated for the input of one information symbol, the code rate of the sub-code is 1/6.
For R=l/3, the sequences of transmission code symbols are given as C0: X, YO, Y'O, X, YO, Y'O, X YO, Y'O, X, YO, Y'O,... C,: X, Yl, Y' 1, X, Yl, Y' 1, X, Yl, Y' 1, X, Yl, Y' 1, . . .
Because three code symbols are generated for the input of one information symbol, the code rate of the sub-code is 1/3. A different code is transmitted at each transmission due to the use of a different puncturing matrix. After soft combining of C0 and Cb X occurs twice and each of YO, Yl, Y'O and Y' 1 occurs once. Thus, the decoder with a code rate 1/5 can be used in this case and the puncturing matrices satisfy the above-described conditions, ensuring performance.
In the first case with R=2/3 of Table 4, the sequence of transmission code symbols are given as
(Sequence Removed)
Because three code symbols are generated for the input of two information symbols, the code rate of the sub-code is 2/3. A different code is transmitted at each transmission due to the use of a different puncturing matrix. After soft combining of C0, Cb C2, Q, X occurs twice and each of YO, Yl, Y'O and Y' 1 occurs once. Thus, the decoder with a code rate 1/5 can be used also in this case as for R=l/6 and the puncturing matrices satisfy the above-described conditions, ensuring performance.
In the second case with R=2/3 of Table 4, the sequence of transmission code symbols are given as
(SequenceRemoved)
Because six code symbols are generated for the input of four information symbols, the code rate of the sub-code is 2/3. A different code is transmitted at each transmission due to the use of a different puncturing matrix. After soft combining of C0, Cb C2, C3, X occurs twice and each of YO, Yl, Y'O and Y' 1 occurs once. Thus, the decoder with a code rate 1/5 can be used also in this case as for R=l/6 and the puncturing matrices satisfy the above-described conditions, ensuring performance.
FIG. 3 is a graph showing a comparison between the performance of HARQ using packet code combining and the performance of HARQ using packet diversity combining in terms of data throughput for a QCTC with R=2/3 and S=4 according to the present invention. As shown in FIG. 3, an HARQ 301 using packet code combining for the QCTC and an HARQ 302 using packet diversity combining for the QCTC show better performance than an HARQ 303 without the QCTC. For the same real time data throughput (e. g., 0.25), about -4dB of Es/No is required in the HARQ 301, about -1.3dB in the HARQ 302, and about IdB in the HARQ 303. Consequently, the use of the QCTC
according to the present invention ensures a higher data throughput with less symbol energy.
As described above, sub-codes generated for optimum packet code combining according to a predetermined rule remarkably improves transmission throughput in a packet retransmission system using turbo codes according to the present invention.
While the invention has been shown and described with reference to a certain preferred embodiment thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claim
1. A method of generating an initial puncturing matrix from which a first sub-code is produced in a communication system having a turbo encoder for generating information symbols, first parity symbols, and second parity symbols for the input of an information bit stream and a sub-code generator for generating sub-code from the information symbols, the first parity symbols, and the second parity symbols using puncturing matrices, a number of the cub-codes being identical to a number of the puncturing matrices, comprising steps of:
selecting as many information symbols as a number of columns in the initial puncturing matrix from the information symbols output from the turbo encoder, if a different between a number Ns of selected symbols in the initial puncturing matrix and the number of the columns in the initial puncturing matrix is equal to or greater than a number of component encoders in the turbo encoder ; and
selecting as many first and second parity symbols as the difference , a number of the selected first parity symbols being equal to or greater than a number of the selected second
arity symbols.
2. The method as claimed in claim 1, further comprising the step of increasing the
number of the columns in the puncturing matrix by an integer multiple if the difference is
less than the number of the component encoders.
3. The method as claimed in claim 1, further comprising the step of selecting the Ns
symbols from unselected first and second parity symbols in the initial puncturing matrix in
generating a second puncturing matrix, wherein the number of the selected second parity
symbols is equal to or greater than the number of the selected second parity symbols.

4. The method as claimed in claim 3, further comprising the step of selecting the Ns
symbols from unselected first and second parity symbols in the other puncturing matrices
except for a last puncturing matrix and repeating (Ns-Ns2) information symbols, Ns2
being a number of the other unselected parity symbols in the other puncturing matrices.
5. The method as claimed in claim 1, wherein the communication system user the sub
code to hybrid ARQ (Automatic Repeat Request).
6. The method as claimed in 1, wherein the number Ns of selected symbols is
determined by C/Rmax, C being the number of the columns and Rmax being a code rate of
the sub-code.
7. A method of generating a first sub-code to be transmitted by selecting a
predetermined number of symbols from information symbols and first and second parity
symbols within a predetermined puncturing range in a communication system having a
turbo encoder for generating information symbols, first parity symbols, and second parity
symbols for the input of an information bit stream, the predetermined number of symbols
being Ns symbols, comprising the steps of:
selecting all the information symbols within the puncturing range, if a difference between Ns and a number of the information symbols is equal to or greater than a number of component encoders in the turbo encoder : and
selecting as many first and second parity symbols as the difference, number of the selected first parity symbols being equal to or greater than a number of the selected second parity symbols.
8. The method as claimed in claim 7, further comprising the step of increasing the puncturing range by an integer multiple if the difference is less than the number of the component encoders.
9. The method as claimed in claim 7, further comprising the steps of selecting the Ns
symbols from unselected first and second parity symbols in the first sub-code in generating
a second sub-code, wherein the number of the selected first parity symbols is equal to or
greater than the number of the selected second parity symbols.
10. The method as claimed in claim 9, further comprising the step of selecting
remaining unselected first and second parity symbols in the other sub-codes except for a
last sub-code and repeating (Ns-Ns2) information symbols, Ns2 being a number of the
other unselected parity symbols in generating the last sub-code.
11. The method as claimed in claim 7, wherein the communication system uses the sub
code to hybrid ARQ (Automatic Repeat Request).
12 The method as claimed in claim 7, wherein the number Ns of selected symbols is determined by C/Rmax, C being a number of columns in a puncturing matrix and Rmax being a code of the sub-code.
13. A method of generating puncturing matrices by which information symbols and first and second parity symbols are punctured in a communication system having a turbo encoder for generating information symbols, first parity symbols, and second parity symbols for the input of an information bit stream, and a sub-code generator for generating sub-codes from the information symbols, the first parity symbols, and the second parity symbols by puncturing, comprising the steps of:
determining a number of S of the puncturing matrices by calculating a minimum integer equal to of greater than a ratio of Rmax to Rmin, Rmax being a given maximum code rate and Rmin being a given minimum code rate for the turbo encoder ;
determining a number Ns of symbols to be selected from each puncturing matrix by C/Rmax, C being a number of columns of the puncturing matrix; and
generating a first puncturing matrix for which C information symbols are selected, as many first parity symbols as a minimum integer equal to or greater than a ratio of a(Ns-C) to (a+b) are selected, and as many second parity symbols as a maximum integer equal to or less than a ratio of b(Ns-C) to (a+b) are selected, if a difference between Ns and C, (Ns-C), is equal to or greater than a number of component encoders in the turbo encoder, a and b being symbol distribution ratios for the first and second parity symbols.
14. The method as claimed in claim 13, further comprising the steps of increasing the
number of the columns in the puncturing matrix by an integer multiple if the difference is
less than the number of the component.
15. The method as claimed in claim 13, further comprising the step of generating
middle puncturing matrices except for the first and a last puncturing matrices by selecting
as many first parity for the first parity symbols as a minimum integer equal to or greater
than a ratio of aNs to (a+b) and as many second parity symbols as a maximum integer
equal to or less than a ratio of bNs to (a+b) without selecting any information symbols, so
that the first and second parity symbols differ in the first and other puncturing matrices.
16 The method as claimed in claim 15, further comprising the step of generating the last puncturing matrix by selecting remaining unselected first and second parity symbols in the other puncturing matrices except for the last puncturing matrix and repeating (Ns-Ns2)
information symbols, Ns2 being a number of the other unselected parity symbols in the other puncturing matrices.
17. The method as claimed in claim 13, wherein the communication system uses the
sub-codes to hybrid ARQ (Automatic Repeat Request).
18. A method of generating an initial puncturing matrix from which a first sub-code is
produced in a communication system substantially as herein described with reference to
the foregoing description, and the accompanying drawings.
19. A method of generating a first sub-code to be transmitted by selecting a
predetermined number of symbols from information symbols and first and second parity
symbols within a predetermined puncturing range in a communication system substantially
as herein described with reference to the foregoing description, tables and the
accompanying drawings.
20. A method of generating puncturing matrices by which information symbols and first
and second parity symbols are punctured in a communication system substantially as
herein described with reference to the foregoing description', tables and the accompanying
drawings.

Documents:

in-pct-2002-558-del-abstract.pdf

in-pct-2002-558-del-assignment.pdf

in-pct-2002-558-del-claims.pdf

in-pct-2002-558-del-correspondence-others.pdf

in-pct-2002-558-del-correspondence-po.pdf

in-pct-2002-558-del-description (complete).pdf

in-pct-2002-558-del-drawings.pdf

in-pct-2002-558-del-form-1.pdf

in-pct-2002-558-del-form-19.pdf

in-pct-2002-558-del-form-2.pdf

in-pct-2002-558-del-form-3.pdf

in-pct-2002-558-del-form-5.pdf

in-pct-2002-558-del-gpa.pdf

in-pct-2002-558-del-pct-101.pdf

in-pct-2002-558-del-pct-210.pdf

in-pct-2002-558-del-pct-220.pdf

in-pct-2002-558-del-pct-304.pdf

in-pct-2002-558-del-pct-308.pdf

in-pct-2002-558-del-petition-137.pdf

« Previous Patent

Next Patent »

Patent Number

210997

Indian Patent Application Number

IN/PCT/2002/00558/DEL

PG Journal Number

45/2007

Publication Date

09-Nov-2007

Grant Date

16-Oct-2007

Date of Filing

03-Jun-2002

Name of Patentee

SAMSUNG ELECTRONICS CO., LTD.

Applicant Address

416, MAETAN-DONG, PALDAL-GU, SUWON-SHI, KYUNGKI-DO, REPUBLIC OF KOREA.

Inventors:

#	Inventor's Name	Inventor's Address
1	KIM, MIN-KOO	973-3, YOUNGTONG-DONG, PALTAL-GU, SUWON-SHI, KYONGGI-DO, REPUBLIC OF KOREA.
2	JANG, JAE-SUNG	1102-203, CHUNGNAG-DONG, KWACHON-SHI, KYONGGI-DO, REPUBLIC OF KOREA.

PCT International Classification Number

H03M 13/00

PCT International Application Number

PCT/KR01/01780

PCT International Filing date

2001-10-20

PCT Conventions:

#	PCT Application Number	Date of Convention	Priority Country
1	2000/62151	2000-10-21	Republic of Korea