KR100362557B1 - 2-dimensional interleaving apparatus and method - Google Patents

Abstract

The present invention relates to a two-dimensional interleaving apparatus and method. According to the present invention, a first feedback convolutional encoder for encoding n data strings using an m-th order raw polynomial and interleaving and outputting n data taken from each group by dividing the n data into k groups And a second feedback convolutional encoder for encoding the interleaved data from the interleaver using the m-th order raw polynomial, wherein k is a positive integer multiple of 2 ^m −1. And a controller for controlling the selection of the groups such that the period of selecting the particular one group from the k groups and then selecting the particular one group is not an integer multiple of 2 ^m −1.

Description

Two-dimensional interleaving apparatus and method {2-DIMENSIONAL INTERLEAVING APPARATUS AND METHOD}

The present invention relates to an interleaving / deinterleaving apparatus and method, and more particularly to two-dimensional interleaving / decoding for turbocoders used in communication systems (satellite systems, ISDN, digital cellular, W-CDMA, IMT-2000, W-ATM). An interleaving apparatus and method are disclosed.

In general, a turbo coder is a type of error correcting code that is widely involved in improving reliability of a digital communication system. The turbo codes disclosed to date can be divided into parallel turbo code and serial turbo code. The parallel turbo encoder is a system that creates a parity symbol using two simple parallel chain codes from an input consisting of frames of L information bit streams. The parallel coder is a RSC (component coder). Recursive Systematic Convolutional) code. The turbo code has an internal interleaver between the component encoders.

The interleaver is used to randomize a data string input in units of a predetermined size (frame) and to improve a distance property of a code word. In particular, it is used in the supplemental channel used as a data transmission channel in the air interface standardization process of IMT-2000 (CDMA2000), which has attracted much attention recently. In addition, in the Universal Mobile Telecommunication System (UMTS) standardization process promoted by the European Telecommunication Standards Institute (ETSI), it is promising that a turbo coder is used in a channel through which user data is transmitted. Therefore, a specific implementation method of the turbo interleaver for the turbo code is urgently required.

FIG. 1 is a diagram illustrating a structure of a general parallel turbo coder, which is disclosed in detail in a patent number US Pat. No. 5,446,747, issued on August 29, 1995, which is a patent for a turbo code.

The turbo encoder having the configuration as shown in FIG. 1 outputs the first frame encoder 111 without encoding the input frame data sequence, and encodes and outputs the input frame, and the second component encoder 113 by interleaving the input frame data sequence. It consists of an interleaver 112 for outputting and a second component encoder 113 for inputting and encoding the output of the interleaver. The first and second constituent encoders 112 and 113 may use a recursive systematic convolutional (RSC) encoder or a non-recursive systematic convolutional (NSC) encoder that is well known in the art. Such a constituent encoder may have an internal configuration that depends on the coding rate, the finite length k value, and the generated polynomial. The interleaver 112 has the same memory size as the maximum frame size and reduces the correlation between the information bits by changing the input order of the information bits input to the second component encoder 113.

Conventionally, various methods such as PN random interleaver, random interleaver, block interleaver, non linear interleaver, and S-Random interleaver are proposed as the internal interleaver 112 of the turbo coder. It became. However, these interleavers can be seen as algorithms designed with the focus on performance improvement as an academic field rather than an implementation point of view. Therefore, considering the implementation of the actual system in terms of hardware implementation complexity can be said to be the way to reconsider. Hereinafter, a description will be given of the technology, properties and problems of the conventional turbo coder interleaver.

Turbocoders are basically performance dependent on internal interleaver and component encoder. In general, turbocoders increase performance as the input frame size (the number of information bits in one frame) increases. However, as the interleaver size increases, the amount of computation increases, so if the frame size is very large, the actual implementation is impossible.

Experimental results show that the random interleaver outperforms the commonly used block interleaver. However, one of the problems in designing the random interleaver is that as the frame size varies and becomes larger, the size of the memory for storing the interleaver's mapping rule or address becomes too large. That is, the memory space for addressing is greatly increased. Therefore, if the size of the hardware is a problem, an address is generated at every symbol clock by using a certain rule that generates an index rather than a look-up table method of storing an interleaver's index. An address enumeration method of reading data is required.

Taken together, optimal interleaver performance under these limited conditions is needed to implement a turbo interleaver where various interleaver sizes are required and hardware complexity is limited, such as for IMT-2000 or Universal Mobile Telecommunication System (UMTS). It must be designed to ensure this. That is, the required interleaver is required to devise a certain rule for generating an address and then to perform interleaving / deinterleaving based on this. It is also required that the required properties of the turbo internal interleaver be excellent.

In the CDMA-2000 standardization process or the UMTS standardization process, there is no proposed interleaver that is simple in design and has excellent performance as a turbo interleaver. In the case of the forward link and the reverse link given in the current CDMA-2000 standard specification, there are various types of logical channels, and various sizes of interleavers according to various data rates are also included. A large amount of memory is required to reflect faithfully. For example, in the CDMA2000 forward link transmission mode, an interleaver having a wide range of sizes ranging from a minimum of 144 bits / frame to a maximum of 36864 bits / frame is used. The interleaver may be classified into a one-dimensional interleaver that interleaves the entire input frame as one region, and a two-dimensional interleaver that divides the input frame into several partial regions and then interleaves each region.

As described above, the problems of the prior art are summarized as follows.

First, various methods such as PN random interleaver, random interleaver, block interleaver, nonlinear interleaver, and S-R random interleaver have been proposed as internal interleaver for conventional turbo codes. However, these interleavers can be seen as algorithms designed with the focus on performance improvement as an academic field rather than an implementation point of view. Therefore, when considering the actual system implementation, in addition to the above-mentioned conditions, hardware implementation complexity must be considered.

Second, since most of the existing interleaving methods are lookup table methods, the interleaving rules according to each interleaver size must be stored by the control unit (CPU, HOST) of the transceiver, so that a separate storage space in addition to the interleaver buffer is required on the CPU memory side. That is, as the frame size varies and becomes larger among design problems, the size of the memory for storing the interleaver index (mapping rule or address) becomes too large. That is, there is a problem in that the memory space for addressing is greatly increased.

Third, it is not easy to implement an interleaver that satisfies both distance and random properties.

Fourth, when using the two-dimensional interleaver, since the regions were selected without considering the periodicity of the component encoder, a problem of resetting the state of the component encoder occurred.

It is therefore an object of the present invention to provide an interleaver device and method that can solve the problems of the prior art.

Another object of the present invention is to provide an interleaving / deinterleaving apparatus and method capable of maximizing a distance property which is a characteristic of a turbo coder in a communication system.

Another object of the present invention is to provide an apparatus and method for interleaving an input frame data into a predetermined size by an interleaver of a turbo coder in a communication system.

It is still another object of the present invention to provide an apparatus and method for controlling group selection such that a period for selecting a specific group in the 2-dimensional interleaving apparatus and again selecting the specific group is different from the period of the component encoder constituting the coder. In providing.

In order to achieve the above objects, a first feedback convolutional encoder for encoding n data strings using an m-th order raw polynomial and interleaving n data taken from each group by dividing the n data into k groups And a second feedback convolutional encoder for encoding the interleaved data from the interleaver using the m-th order raw polynomial, wherein k is a positive value of 2 ^m −1. And a controller for controlling the selection of the groups such that the period of selecting the particular one group from the k groups and then selecting the particular one group is not an integer multiple of 2 ^m −1. do.

1 is a block diagram showing a general turbo encoder.

2 is a block diagram of an interleaving apparatus of a communication system according to the present invention;

3 is a block diagram of a deinterleaving apparatus of a communication system according to the present invention;

4 is a view for explaining the general operation principle of the two-dimensional turbo interleaver.

Fig. 5 is a diagram showing output for a specific input sequence when K = 4, N _G = 7 and group selection is performed sequentially.

6 is a diagram illustrating a conceptual diagram for group selection of a two-dimensional interleaving apparatus according to the present invention;

7 illustrates a two-dimensional interleaving apparatus according to an embodiment of the present invention.

To find a turbo coder that can perform optimally for various frame sizes, there are many things to consider such as the number of turbo coders (finite length k), the generation polynomial, and the optimal coding rate. In addition, it is very time and effort to experimentally find a turbo coder that performs optimally without theoretically verifying how all these factors affect performance.

Therefore, the method for the actual implementation should implement some criteria that affect the performance and implement the interleaver in a way that satisfies the criteria as much as possible. These standard characteristics are as follows.

Distance property: The distance between adjacent codeword symbols must be maintained to some extent. Since this plays the same role as the codeword distance property of the convolutional code, it is preferable to design the distance as large as possible under the same conditions.

Weight Property: The weight of a codeword corresponding to a non-zero information word should be more than a certain amount. Since it plays the same role as the minimum distance property of the convolutional code, it is desirable to design the weight to be large under the same conditions.

Random Properity: The correlation coefficient between output word symbols after interleaving should be very low compared to the correlation factor between original input information symbols before interleaving. That is, randomization must be faithfully performed between output word symbols. This is a factor that directly affects extrinction information quality generated in iterative decoding.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention describes in detail an optimal interleaving / deinterleaving apparatus capable of satisfying the above-described interleaver properties (distance property, weight property, random property, etc.) of the turbo coder.

In general, a random interleaver is excellent as an internal interleaver of a turbo coder. In addition, the turbocoder has a higher performance with a larger frame size. Therefore, the present invention proposes a 2-dimensional interleaver to improve the complexity while satisfying the performance of the random interleaver. The two-dimensional interleaver means dividing the size of the input frame into a predetermined number of groups, and interleaving by applying a specific interleaving rule for each group. In the present invention, the group selection period for selecting a specific group and reselecting the specific group is shifted from that of the internal component encoder. In addition, in the embodiment of the present invention, a linear feedback shift register (LFSR) for generating a PN sequence is mapped to each group for interleaving, and a random number generated from the LFSR is used as a read address. do. However, the present invention is not limited to the method of interleaving in the group unit, and the main concern is to select a group in consideration of the periodicity of the component encoder in selecting addresses generated for each group. In addition, the first and second constituent encoders of the turbo code exemplified in the present invention can be configured not only with the constituent encoders of the prior art but also by the constituent encoders proposed by CDMA2000 standardized by TIA and UMTS standardized by ETSI. It is not limited by the specific constituent encoders, such as the use of constituent encoders. In addition, the interleaver of the present invention can be applied not only to the internal interleaver of serial turbo code but also to a channel interleaver.

2 and 3 show the basic configuration of the interleaver and deinterleaver according to the present invention.

Referring to FIG. 2, the address generator 211 generates a read address for changing data order according to an input frame data size L and an input clock and outputs the read address to the interleaver memory 212. The interleaver memory 212 stores input data in order in the write mode, and outputs data according to the address provided by the address generator 211 in the read mode. The counter 213 inputs a symbol clock and outputs it to a write address of the interleaver memory 212. As described above, the interleaver stores input data in order in the interleaver memory 212 in the write mode, and is stored in the interleaver memory 212 according to the read address generated by the address generator 211 in the read mode. Output the data. As another example, the data may be changed in the interleaver memory in the write mode in advance, and the data may be read out in order in the read mode and output to the second component encoder.

Referring to FIG. 3, the deinterleaver memory 312 generates a write address for restoring the data order based on the input frame data size L and the input clock. To provide. The deinterleaver memory 312 stores input data according to the write address (write ADDR) provided by the address generator 311 in the write mode, and sequentially outputs the stored data in the read mode. The counter 313 inputs a symbol clock and provides a read address Read ADDR for reading data from the deinterleaver memory 312 to the memory 312.

As described above, the deinterleaver is a reverse process of the interleaver, and all parts of the deinterleaver are completely identical in structure, and only the order of data inputted in the read / write mode is different. Therefore, the following description will focus on the interleaver.

In general, the performance of a turbo code for a given input frame size is determined by the constraint length K and the characteristics of the turbo interleaver. In the case of K, 3 or 4 is generally selected in consideration of the complexity in the decoder design and the degree of improvement in performance. For example, in the case of CDMA2000 or UMTS, K = 4 dominates. On the other hand, the turbo interleaver is very difficult to select because it is almost impossible to find the optimal turbo interleaver due to so many design variables. Therefore, much research is underway on sub-optimal turbo interleaver design instead of optimal turbo interleaver.

First we need to understand the nature of the convolutional code. In the general convolutional code, when W (I), which is a hamming weight of the input sequence I, is low, the code weight is low. On the contrary, when W (I) is high, the code weight is high. This is due to the nature of the feed forward convolutional code. Turbo codes that use recursive convolutional code have an infinite impulse response sequence, unlike convolutional codes. Namely, when K = 4, 100000... Using an input sequence of zero as input generates a component code sequence with a period of seven.

1 shows the configuration of a turbo coder. As shown, a turbo interleaver provided in the turbo coder is present between two component encoders using a recursive convolutionalencoder, wherein the interleaver has a second weight when the weight of the code output from the first encoder is low. The component encoder interleaves the input data to provide data capable of generating a high weight code. As a result, even if the wear of the code output from the first component encoder is low, the weight of the code output from the second component encoder is high, so that the weight of the entire code can be increased. For example, when the input sequence 100000 .. is input to a turbo coder having a restriction length of K = 4, the output of each component coder of the turbo coder is a binary sequence having a periodicity of 2 ^(K-1) -1 = 7. (periodic binary sequence) will be generated. The specific pattern of this sequence is related to the generator polynomial of the turbocoder. That is, a finite code sequence is generated by resetting the state of the component encoder to a zero state by a specific input sequence rather than a zero sequence.

For example, if K = 4, the input sequence is 10000001000000000…. The in sequence resets the state of the first component encoder to zero after coding the second 1. In addition, due to the periodic relationship, the state of the first component encoder is similarly reset in all cases where the distance between two 1s is a multiple of 7. Therefore, all subsequent codes become zero, which causes the minimum distance of the turbo code to be lowered. In this case, the turbo interleaver must know this pattern of the input sequence, so that the second component encoder will have a higher code weight to increase the minimum weight of the entire turbocoder.

Turbo interleaver that serves as described above can be largely divided into two types based on the design method. One and two dimensional turbo interleaver. The one-dimensional interleaver means to consider a frame to be coded as one object and interleave the whole. The two-dimensional interleaver divides an entire frame into sub-frames or a plurality of groups having the same size, and then performs interleaving using an interleaving rule corresponding to each group independently. . In general, the optimization process is easier than the one-dimensional case because the performance of the two-dimensional turbo interleaver is superior to that of the one-dimensional turbo interleaver and has an interleaving rule in each group unit.

4 is a view for explaining the general operation principle of the two-dimensional turbo interleaver (Turbo interleaver).

Referring to FIG. 4, Group0, Group1, Group2, and Group3 represent each group when the input frame is divided into four groups. P0 (k) represents an interleaving rule for interleaving the data of Group0, P1 (k) represents an interleaving rule for interleaving the data of Group1, and P2 (k) represents the data of the Group2 An interleaving rule for interleaving is shown, and P3 (k) represents an interleaving rule for interleaving data of Group3. Meanwhile, P0 (0), P1 (0), P2 (0) and P3 (0) represent addresses of interleaved data in each group when k is '0', and P0 (1), P1 (1). ), P2 (1) and P3 (1) represent addresses of interleaved data in each group when k is '1', and P0 (2), P1 (2), P2 (2) and P3 (2). ) Denotes the addresses of interleaved data in each group when k is '2', and P0 (3), P1 (3), P2 (3) and P3 (3) when k is '3'. Represent the addresses of interleaved data in each group. That is, one data bit is read from each group by the interleaving rule Pg (k) corresponding to each group at each time point k. This method can be regarded as a natural extension when the number of groups, which is a one-dimensional method, is one. Of course, unlike the method of increasing sequential in selecting groups as in the above example, it may give a certain rule. For example, the entire interleaving rule may be changed by bringing the group selection as 0 2 1 3 rather than the order of 0, 1, 2, 3 as shown in FIG. 4 above. That is, FIG. 4 illustrates a case in which the number of data bits of the input frame is 16, and the input frame is divided into four groups each having 4 bits, and each group is interleaved by a corresponding interleaving rule.

Problems may occur when performing 2D turbo interleaving as described above. If the number of groups is exactly equal to the period of the component encoder or a multiple of the period, this will greatly reduce the minimum distance of the turbo code. For example, if k = 4, the period is 7, so if you take the number of groups in multiples of seven or seven, the probability of resetting each component encoder to zero is very high.

To show this phenomenon, in the accompanying drawings, the worst case of the input sequence of the turbo coder is 100000010000 .. The input sequence is maintained even after interleaving. As shown, it can be seen that the pattern of the existing input sequence is maintained even after interleaving. Therefore, it can be seen that the weight of the turbo code is very low because the second component encoder is also reset in the zero state. This is true for all cases where the number of groups is a multiple of 7, but most occurrences of the minimum distance are 7 or 14. The turbo interleaver designed in this way has a very low minimum distance, which degrades the bit error rate (BER) or frame error rate (FER) performance of the entire turbo codes.

Therefore, in the present invention, when designing a two-dimensional turbo interleaver (Turbo interleaver) to solve the above problems by bringing the group selection pattern differently, the cycle of selecting a specific group and the specific group to configure the turbo coder Offset the period of the encoder. That is, the present invention aims to maximize the minimum distance of the turbo code by grasping the characteristics of the coder in group selection of the two-dimensional turbo interleaver.

Hereinafter, a turbo interleaving apparatus for maximizing a minimum distance of the turbo code according to an embodiment of the present invention will be described in detail.

The most efficient way to break the periodicity of group selection is to take a different start of the group to be selected first in every cycle. In the following Table 1, k = 4 and the number of groups is 7, an example of such a scheme is shown. In the following Table 1, group selection is sequential and cyclic shift is described as an example. However, it is also possible to perform random and cyclic shift with a specific pattern in which group selection is predetermined.

shift step Group Selection Order 0 0123456 0123456 0123456 0123456 0123456 0123456 0123456 0123456 One 0123456 1234560 2345601 3456012 4560123 5601234 6012345 0123456 2 0123456 2345601 4560123 6012345 1234560 3456012 5601234 0123456 3 0123456 3456012 6012345 2345601 5601234 1234560 4560123 0123456 4 0123456 4560123 1234560 5601234 2345601 6012345 3456012 0123456 5 0123456 5601234 3456012 1234560 6012345 4560123 2345601 0123456 6 0123456 6012345 5601234 4560123 3456012 2345601 1234560 0123456

For example, when the shift step is 2 in Table 1, the group is selected in the order of "0123456" in the first period, and the group selection is selected in the first period in the second period. Only shift is selected in order of "2345601". That is, in every selection cycle, a group is selected with the selection sequence shifted by two in the group selection sequence in the previous selection cycle. When the group is selected by the cyclic shift method as shown in Table 1, the order of returning the original order is 7 times farther, so that the minimum distance can be taken much larger. That is, if shift step = 0, the group selection pattern of the same selection period is generated immediately in the next selection period, but if shift step = 1,2,3,4,5,6, the first selection pattern occurs much later. do. In other words, the group selection order in units of coder cycles is shifted by the cyclic shift method, so as to deviate from the cycle of the turbo coder itself.

This cyclic shift group selection is very simple. Each group selection is obtained by the following formula.

g = (clk% N _G + shift_step * k)% N _g for clk = 0,... , frame_length-1

Where N _g is the number of groups, clk is a counter that represents the data clock in bits, and k is a counter that increments by 1 for each group selection cycle, with k = 0. It is initialized and shows the number of the first group selected in the group selection period. shift_step is a constant that can be selected to be any number satisfying GCD (2 ^K-1 , shift_step) = 1 and depends on the constraint field K. That is, a cyclic shift method can be implemented by using only two counters and one address.

For example, if N _g = 7 and shift_step = 1, clk is 0,1,2,3,4,5,6, | 7,8,9,10,11,12,13, | .... Then k = 0,0,0,0,0,0,0, | 1,1,1,1,1,1,1, | ..., so g = 0,1,2 , 3,4,5,6, | 1,2,3,4,5,6,0 | ...

Until now, the cyclic area shift method in the case of sequentially increasing group selection order in selecting a group has been described. However, as mentioned above, in order to make the distance property of the turbo interleaver larger, It may be taken randomly when extracting any interleaved address.

For example, assume that k = 4 and the number of regions 7. In this case, even if any group multiplexing method is used, the period of group selection cannot be broken without the cyclic group shift described above. Therefore, the group multiplexing is first increased after the cyclic shift. This predetermined pattern causes the group selection to be randomized again. Here, the group multiplexing means selecting a group at random according to a specific pattern. Here, even if the group selection order is shifted by the cyclic shift, the specific pattern for randomly selecting groups does not change. That is, group selection is performed while the specific pattern is shifted.

FIG. 6 is a group selector including a group multiplexing operation for randomizing group selection and a cyclic shift pore for shifting the group selection order every cycle in order to break a cycle of group selection in the two-dimensional interleaving apparatus according to the present invention. It is shown.

Referring to FIG. 6, first, it is determined whether the number of groups N _G is a multiple of 2 ^(K-1) −1, and in this case, the period of group selection is increased by a cyclic group shift operation. A group multiplexer then randomly selects the groups by a specific pattern based on these.

7 illustrates an address generator for performing two-dimensional interleaving according to an embodiment of the present invention.

Referring to FIG. 7, the address generators 711-71N each generate a read address by a given interleaving rule in the address generation region of the corresponding group. Controller 721 controls the multiplexing of selector 731. The selector 731 selects and outputs addresses generated by the address generators 711-71N under the control of the controller 721. The address thus output is used to read the input frame data sequentially stored in the interleaver memory 212. In this way, the size of the interleaver is divided into a plurality of groups, interleaving is generated for each group, and an address generated for each group is sequentially or randomly selected and used as a read address. In this case, in order to obtain excellent interleaving properties, the group selection order of the selector is important. In the present invention, a method of shifting the group selection order every cycle by a cyclic shift method is basically proposed. In order to bring more distance properties, we propose a method of randomizing group selection order.

Referring to the operation based on the configuration of FIG. 7, the address generators 711-71N each generate an address with an interleaving rule given to them in the corresponding group. For example, the address generator may use a Linear Feedback Shift Register (LFSR) for address generation. The controller 721 provides the selector 731 with a selection signal for selecting addresses from which the address generators 711-71N have occurred. The controller 721 outputs the group selection signal by the cyclic shift method as shown in Table 1 above. For example, when the shift step is 2, a selection signal for shifting the group selection order by '2' is output every cycle in the order of 0123456 in the first period and 2345601 in the second period. . In this case, the controller 721 may control to randomly select groups by a specific pattern in order to further improve the distance property of the interleaver. However, the specific pattern is shifted only every N _g group selection, but the order is not changed. The selector 731 selects and outputs the addresses generated by the address generators based on the selection signal from the controller 721. The selected address is used to randomly read input frame data sequentially stored in the interleaver memory 212.

An example of the controller 721 controlling the group selection is shown in FIG. 8A, and a timing diagram according to the configuration of FIG. 8A is shown in FIG. 8B.

Referring to FIG. 8A, the calculator 803 provides a value calculated by mod 7 in response to clock pulses of the second clock signal clk2 of FIG. 9B as a counter count value of the counter 801. Then, the counter 801 performs counting starting from the counter value provided from the calculator 803 in response to the clock pulses of the first clock signal clk1, and stores the counter value in a group selection order. It is used as the address of the lookup table 805 being used. The counter 801 is a counter for counting [0..6] in a cyclic manner.

9A shows another example of the controller 721, and FIG. 9B shows a timing diagram according to the configuration of FIG. 9A.

Referring to FIG. 9A, the counter 903 outputs the counter signal counter of FIG. 9B in response to clock pulses of the clock signal clk of FIG. 9B. The comparator 909 compares the counter signal from the counter 903 with a predetermined number '6' and applies an activation pulse (ACT) to the load counter 901 and the shift counter 905 when they are the same. Then, the shift counter 905 outputs the shift counter of FIG. 9B to the load counter 901 when the operation pulse is applied. Then, the load counter 901 writes the loadable counter of FIG. 9B in the lookup table 907 in the group selection order, shifted by '1' every cycle in response to the clock pulses of the clock signal. The counter 901 is a counter for counting [0..6] in a cyclic manner.

As described above, the present invention proposes a method capable of satisfying the distance quality of the interleaver when performing 2D interleaving. That is, the present invention prevents the component encoder included in the turbo coder from being reset in the zero state by changing the period for selecting a specific group from the turbo coder when implementing the 2D interleaver. distance).

Claims (18)

a first feedback convolutional encoder for encoding n data strings using an order m polynomial, a two-dimensional interleaver for dividing the n data into k groups, and interleaving and outputting n data taken from each group; A coding apparatus comprising a second feedback convolutional encoder for encoding the interleaved data from the interleaver using the m-th order raw polynomial, Controlling the selection of the data groups such that when the k is a positive integer multiple of 2 ^m −1 and at least one of the data groups has a specific data pattern, the output of the interleaver generates data that does not become the specific data pattern And a controller. The method of claim 1, And the controller shifts the group selection order constantly for each selection cycle. The method of claim 1, And said specific data pattern is a sequence in which the distance between two '1's is a multiple of 2 ^m-1 . In the interleaving method of a communication system comprising a feedback encoder having m memories and a two-dimensional interleaver having groups of integer multiples of 2 ^m −1, each group having addresses of a given size, Selecting addresses so that a group selection cycle of selecting one of the addresses in each group across the groups does not become an integer multiple of 2 ^m −1. The method of claim 4, wherein And the controller shifts the group selection order constantly for each selection cycle. a first feedback convolutional encoder for encoding n data strings using an order m polynomial, a two-dimensional interleaver for dividing the n data into k groups, and interleaving and outputting n data taken from each group; A coding apparatus comprising a second feedback convolutional encoder for encoding the interleaved data from the interleaver using the m-th order raw polynomial, Wherein the k is a positive integer multiple of 2 ^m −1, and the cycle of selecting a particular one of the k groups and then again selecting the particular one is different from the period of the feedback convolutional encoder Apparatus comprising a controller for controlling the. The method of claim 6, And the controller shifts the group selection order constantly for each selection cycle. The method of claim 6, M is '3', and the period of the feedback weaving encoder is characterized in that the '7'. a first feedback convolutional encoder for encoding n data strings using an order m polynomial, a two-dimensional interleaver for dividing the n data into k groups, and interleaving and outputting n data taken from each group; A coding apparatus comprising a second feedback convolutional encoder for encoding the interleaved data from the interleaver using the m-th order raw polynomial, Wherein k is 2 ^m so that the amount of times the integer 1, to select the particular one group of the k groups are cycles in which the re-select a particular one of the group is not a multiple of 2 ^m -1 selection of the group Apparatus comprising a controller for controlling the. The method of claim 9, And the controller shifts the group selection order constantly for each selection cycle. The method according to claim 10, And the group selection order is a sequence randomized by a specific pattern. The method of claim 9, wherein the interleaver, A memory for sequentially storing the n pieces of data; An address generator for generating a read address with a given interfering rule for each of the k groups, And a selector for selecting the groups in a predetermined order under the control of the controller and selecting a read address from the selected group, And output the n pieces of data from the memory with a read address from the selector. a first feedback convolutional encoder for encoding n data strings using an order m polynomial, a two-dimensional interleaver for dividing the n data into k groups, and interleaving and outputting n data taken from each group; A method of interleaving an interleaver in an encoding apparatus comprising a second feedback convolutional encoder for encoding the interleaved data from the interleaver using the m-th order raw polynomial, Wherein k is 2 ^m so that the amount of times the integer 1, to select the particular one group of the k groups are cycles in which the re-select a particular one of the group is not a multiple of 2 ^m -1 selection of the group The method comprising the step of controlling. The method of claim 13, And the control process shifts the group selection order constantly for each selection cycle. The method according to claim 14, And the group selection order is a sequence randomized by a specific pattern. a first feedback convolutional encoder for encoding n data strings using an order m polynomial, a two-dimensional interleaver for dividing the n data into k groups, and interleaving and outputting n data taken from each group; A method of interleaving an interleaver in an encoding apparatus comprising a second feedback convolutional encoder for encoding the interleaved data from the interleaver using the m-th order raw polynomial, Sequentially storing the n pieces of input data in a memory; Generating a read address with a given interfering rule for each of the k groups, Wherein k is a positive integer of 2 times ^m -1, so as to select the particular one of the group of the k groups is the period for re-selecting a particular one of the group is not a multiple of 2 ^m -1 selection of the group To control the process, Reading the input data from the memory with a read address from the selected group. The method of claim 16, The control method is characterized in that for shifting the group selection order constantly for each selection cycle. The method of claim 17, And the group selection order is a sequence randomized by a specific pattern.