JPH05300181A - Device and method for encoding modulation for mobile communication - Google Patents

Abstract

PURPOSE: To improve the encoding gain of encoding modulation against the fading of mobile communication, etc. CONSTITUTION: Interleaved block encoding modulation having a built-in time diversity is applied to a fading channel 30. Such various modulated block codes of various dimensions that are generated from an M-PSK arrangement are disclosed. The signal points constituting each code word are rearranged by means of an interleaver 21 by matching interleaving to the block codes so that the effective size of the interleaver 21 can be increased.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION This invention relates to coded modulation techniques, and more particularly to the use of such techniques in channel applications, such as fading cellular mobile communications.

[0002]

BACKGROUND OF THE INVENTION Over the past decade, trellis coded modulation has been shown to be a practical power and bandwidth efficient modulation technique for channels with additive white Gaussian noise (AWGN). This technology is now widely used in commercial telephone line modems, increasing the line rate of these modems to 19.2 Kbit / s.

More recently, the applicability of the trellis coded modulation technique to higher classes of channels, especially fast fading channels, has been detected and therefore accurate because the signal amplitude changes drastically in a short time. Applicability to channels where recovering transmission information is impractical is being investigated. In fact, the signal amplitude is so weak that accurate data recovery is not possible even if the signal amplitude is detected.

Various types of mobile communication channels fall into this category. The use of trellis coding in such channels, as in telephone line modem applications, results in a so-called "coding gain" of signal power (compared to so-called "uncoded" modulation methods). The net result is an enhanced ability to recover accurate information without the need for additional signal bandwidth.

Unfortunately, the improvement in error rate performance achieved with a given amount of coding gain has been found to be much lower in the fast fading channel than in the telephone line channel. For example, a coding gain of 3 dB improves the error rate on the telephone line channel by as much as three orders of magnitude, but only by a third on the fast fading channel. This imbalance mainly results from the very nature of the fast fading channel, the fading nature.

[0006]

In the prior art, a combination of a specific trellis code (a) exhibiting so-called "built-in time diversity" and an interleave / deinterleave technique (b) for rearranging the transmitted signal points is used. It has been recognized that, by using it, it is possible to take into account the occurrence of fast fading in the channel and thereby provide increased coding gain. However, at the same time, generally such prior art solutions are not entirely satisfactory for the particular application. Digital cellular mobile communication (hereinafter simply referred to as "mobile communication") is a famous example of CK.

In particular, in order to realize the possible coding gain of such a method, an interleaver / de-interleaver with the property that a sufficient amount of transmission delay is introduced at both the transmitter and the receiver. The use of riva is required. For example, the real-time nature of mobile communication systems means that such delays have a significant negative impact on system characteristics.

Further, the possible coding gain realizations are:
Systems that use Time Division Multiple Access (TDMA), which is now the North American standard for next generation mobile communications, require even greater delays. (This effect is
In a TDMA system this results from the fact that the signal points originating from one particular source are closer together than in other cases. In addition, there is some additional delay due to the fact that the trellis decoder has to wait until it receives a large number of subsequent signal points in order to output every particular signal point, and / or some It wastes some of the channel capacity in some particular applications. Such applications include systems that include a speech encoder that encodes block by block.

[0009]

According to the present invention, the use of interleaved block coded modulation with built-in time diversity has many advantages and is equivalent or superior to the prior art described above. Achieve coding gain. The effect is that it is less complex to implement,
Achieve system designs with lower interleaver / deinterleaver and encoder delays, higher bandwidth efficiency for some codes, and desired trade-offs in complexity, power efficiency, and bandwidth efficiency Flexibility is increased, and there are few problems as a general system.

In the preferred embodiment, constant amplitude signal constellations are used to compensate for the rapid changes in signal amplitude that are characteristic of mobile communications and other fast fading channels. Furthermore, the use of non-coherent differential detection is preferred due to the fast changes in carrier phase that occur in such channels. Both of these criteria are conveniently met using M-point differential phase shift keying, or M-DPSK.

According to the invention, the particular way in which the signal points are rearranged by the interleaver is matched to the particular block code used, thereby increasing the effective but impractical size of the interleaver. And thus has the effect of reducing the interleaver / deinterleaver delays previously described.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In the transmission system of FIG.
The input data above is applied to the 2N-dimensional block encoder / mapper at a rate of m bits per signal interval of T seconds, where m is an integer or fraction. The block encoder / mapper 13 integrates a block of input data consisting of bits corresponding to N signal intervals, and outputs the selected 2N-dimensional block code to the integrated N × m bits of (m + r) codes. It is used to encode N groups of encoded bits. These groups are subsequently provided on lead 16. Here, the parameter r is the average number of redundant bits per signal interval introduced by the block encoder / mapper 13.
The allowable bit pattern of each of the (m + r) bit groups is 2
It is associated with a particular signal point of the dimension (2D) M (≤2m ^{+ r} ) -PSK constellation. Here, M is a selected integer. In particular,
The signal point on the lead wire 16 during the nth signal interval is designated as P _n . The block code is “2N-dimensional” because each signal point is two-dimensional and each “code word” output by the block encoder / mapper 13 is represented by N signal points.
Called.

To ensure an understanding of the terms and concepts used herein, reference is made to FIG. The 2N-dimensional block encoder / mapper produces a 2N-dimensional “codeword”. Each signal word consists of a succession of N “signal points”. Each signal point is
2 is a point in a given two-dimensional "array", which is shown diagrammatically in FIG. 2 as a phase shift keying array with eight signal points, or 8-PSK.

In this 2N-dimensional codeword, one signal point is sent in each signal interval during N "signal intervals". The set of all 2N-dimensional codewords is called a "2N-dimensional array", where each codeword is an "element" of the 2N-dimensional array. 2N-dimensional arrays are also called codebooks or scripts.
In the following description, each 2N-dimensional codeword is often treated as a concatenation of two N-dimensional components in the constituent N-dimensional array. Here, the configured N-dimensional array is 2N
It arrives at the same time as the dimensional array. This view is N / 2, N /
Iterate over 4 dimensional elements and arrays.

As shown in FIG. 1, the encoder / mapper 1 on the lead 16 corresponds to each group of N × m input bits.
The N consecutive two-dimensional signal points output by 3 are added to the interleaver 21. The function of the interleaver is to rearrange the signal points P _n such that the signal points belonging to a particular codeword are temporally separated from each other during transmission. This method reduces the likelihood that channel fading will affect one or more of the N signal points of the codeword.

In the preferred embodiment, the block code used in the block encoder / mapper 13 has built-in time diversity, as will be described in detail below.
And, as also described below, this time separation of the signal points greatly enhances the ability to accurately recover the transmitted data of such codes. Further in accordance with the present invention, the block code and interleaver rearrangement algorithm are jointly selected to further enhance this capability.

Finally, the rearranged signal points Q _n output by the interleaver on the lid line 24 are applied to a modulator 25 whose output is in turn applied to the fast fading channel 30. The modulator 25 will be described in more detail below.

In the receiver, the demodulator 41 and the deinterleaver 44 perform the reverse functions of the modulator 25 and the interleaver 21, respectively. Therefore, the output of the deinterleaver on lead 45 corresponds to the succession of signal points appearing on lead 16 at the output of encoder / mapper 13, the received but channel corrupted signal point P. It is a sequence of _n . These are applied to the block decoder 51 which restores and supplies the original transmitted input data on the lead 53.

In accordance with the present invention, the block decoder 51, as will be described in more detail below, is a so-called "soft decision", similar to the maximum likelihood decoder conventionally used for trellis coded signals in the AWGN environment. Work based.

At this point, it is instructive to explain the concept of time diversity coding using the description of the first block code modulation scheme, referred to as code I, implemented in the embodiment shown in FIG. Ah

In particular, an 8-PSK signal constellation as shown in FIG. 3 is used to implement a 4-dimensional (4D) code. This means that each codeword generated by that code is 8-P
It is meant to consist of two two-dimensional points in the SK array. These points are transmitted during each signal interval. The eight points in the array are labeled with numbers from 0 to 7. In this case, the parameters m and r are both equal to 1.5 and of course N = 2. Thus, the block encoder / mapper 13 produces a 3-bit word at each of two consecutive signal intervals. Each such word identifies by its bit value a particular one of signal points 0-7.

Lead 11 within the codeword on lead 16
Input to the upper block encoder / mapper 13 (N ×
The coding / mapping of m) is shown in the table of FIG. Here, each of the 3-bit patterns in parentheses is (m + r
=) Indicates the value of the three encoded bits and is simply the binary value of the decimal label of its associated signal point. Two ^three
= 8 bit patterns, and thus 8 code words. Notationally, each codeword is referred to below as x and y.
In the form of (x, y) as the first signal point and the second signal point respectively constituting the code word.

It is important that each of the codewords of code I is different from any other codeword in both of its two-dimensional points. Thus, for example, the first or second signal point of codeword (0,0) is not equal to the first or second signal point of any other codeword. The importance of this property can be understood by considering the case where the amplitude of one of the two constituent signal points is significantly attenuated by channel fading, resulting in the complete loss of the information carried thereby. As long as the other constituent signal point of the codeword is accurately restored, it is possible to nevertheless restore that information.

In particular, if the first signal point of the restored codeword is "3" and the second signal point is lost, then no other codeword has "3" as the first signal point. So
It is determined that the transmitted codeword was nevertheless (3,7). (This analysis oversimplifies how the decoding process is preferably performed, but is useful for illustration.)

Thus, this code provides a high degree of built-in immunity to transmission errors introduced by fading through the mechanism of time diversity. That is, the information about each input data bit appears redundantly in the encoded signal in the time domain. Thus, for example,
Information about each of the three bits of the input bit pattern 010 appears in both the first signal point "3" and the second signal point "7" of the corresponding codeword (3,7).

Generally speaking, if X is an integer greater than 1, each codeword consisting of a series of signal points arranged is
A code is said to have X times the time diversity if it differs from each of the other code words by at least X signal point positions. Therefore, the code I is evaluated to have twice the built-in time diversity. There are actually many ways to combine the signal points of FIG. 3 into a codebook of four-dimensional codewords that shows twice the built-in time diversity. Advantageously, however, the one shown in FIG. 4 has the further advantage of maximizing the least squared Euclidean distance between codewords (given to the requirement of double time diversity). It further enhances the error immunity of the whole coding scheme. The distance of this code is "4".

It is important to emphasize here that what is described is a coded modulation method. This a) increases the size of the signal constellation beyond 2 ^m signal points to reduce the signal bandwidth requirement caused by the introduction of redundant bits, and b) block coding and constellation mapping are interdependent. Has the meaning. This is a)
The introduction of redundant bits is given by the extension of the signal bandwidth,
b) The block coding and array mapping have a significant symmetry with the conventional block coding method in that they have nothing to do with each other.

More specifically, here, a) two-dimensional M
The code of the invention using signal points from the PSK constellation,
b) Block-code the input bits using, for example, a Reed-Solomon code or other block code, each signal point resulting in a sequence of signal points each taken from the same two-dimensional M-PSK constellation. It is appropriate to compare it with the method of transmitting the encoded bit.

In such a case, the resulting set of transmitted signal points will have some of the characteristics of the inventive code, such as X times time diversity and some coding gain. However, such a method is not related to the minimum Euclidean distance between the series of transmitted signal points, unlike the present invention. As a result, the least-squares Euclidean distance between a series of transmitted signal points is as small as the distance defined therein as the "absolute least-squares Euclidean distance" associated with the script in question. The distance is a two-dimensional M-PSK multiplied by the diversity parameter "X".
It is given as the least squares Euclidean distance of every pair of signal points in the array.

Thus, for example, the absolute least square Euclidean distance of a four-dimensional code having twice the time diversity and using the two-dimensional M-PSK array of FIG. 3 is the codeword (0,0), ( 1, 1), (2, 2), ...
The least square Euclidean distance between the code words of the four-dimensional code consisting of (7, 7) is 1.17.

The code of the present invention symmetrically has a least-squares Euclidean distance between consecutive transmission signal points, that is, between codewords, which is larger than the above-mentioned absolute least-squares Euclidean distance. For example, for code I, which also has twice the diversity and also uses the array of FIG. 3, the least square Euclidean distance between codewords is 4.

As a further difference, conventional block coding schemes generally use the same constellation that is used for non-coding schemes. Thus, for example, for transmitting m (m is an integer) information bits per signal interval in an uncoded system, the constellation used has 2 ^m modulation signal points.
Even if r redundant bits per signal interval were introduced by a conventional block code, the same constellation would still be used, and the baud rate would be (m + r) / m times to accommodate these redundant bits. In contrast, a coded modulation method such as the present invention accommodates at least some of the redundant bits by extending the array size to have 2 ^m or more modulation signal points.

Further, there are many ways to assign the input bit pattern to the various codewords of FIG. However, advantageously, the particular allocation method shown in FIG. 3 has the additional advantage of reducing the bit error rate that occurs when the transmitted codeword is incorrectly decoded. In particular, Gray code type methods are used to assign input bit patterns to codewords.

Suppose that the second signal point of the codeword is lost and the information bits are recovered only based on the first received signal point. In FIG. 4, the code word whose first signal point is closest to each other in Euclidean distance is 1
Notice how only one bit position is assigned to a different bit pattern. The underlying concept is that the signal points that are closest to each other are the signal points that are most likely to be confused with each other. As this is the case, the adoption of the previously described Gray coding type method for bit pattern allocation ensures that the minimum number of bit errors is associated with such most probable decoding errors.

Thus, since the signal points 0 and 1 have the minimum distance between them, the bit patterns "000" and "001" are assigned to signal points which differ by one bit position, respectively. (It is not possible for this particular code to simultaneously provide such increased error correction capability for the assumed fading of the first constellation. However, some general advantages are nevertheless only in one case. Is achieved by dealing with.)

Finally, this concept of Gray coding is to be used effectively in the other codes described herein, at any time, even if no special mention is made thereof. Be careful.

If one particular signal point is lost due to fading, it is recognized that there is a high possibility that a signal point close in time is also lost. Therefore, according to principles known to those skilled in the art, the error immunity provided by the code's built-in time diversity is enhanced by temporally separating the two constellation points of each codeword, which results in 2
It is less likely that individual points will fade at the same time. The signal point P _n on the lead wire 16 in FIG.
Is added for this purpose.

Specifically, the interleaver 21 is
Take and store frames of J codewords. The codewords are stored diagrammatically in FIG. 5 where they are stored in respective rows of a storage matrix held in an interleaver. At the times depicted in the figure, the signal points P ₁ and P ₂ of the first codeword are stored in the first row and the signal points P ₃ and P ₄ are stored in the second row.

In the most straightforward form, the interleaver waits until all J codewords have been read. It then arranges the signal points that make up the codeword row by row, i.e. initially the odd-numbered signal points P ₁ , P ₃ ,
…, P _2J-1 , and then the even-numbered signal point P ₂ ,
Read out P ₄ , ..., P _2J . (For more efficient execution, all J codewords are odd before they are read, as long as enough codewords have been read to ensure a synchronous flow of the signal points on lead 24. It is possible for the interleaver to start reading the signal point of the number.)

Note that the signal points are grouped together at the positions of the corresponding signal points of the codeword in the frame. That is, when i = 1, 2, ..., N, each i-th signal point of the codeword of the frame is arranged in each group. Thus, as desired, the two signal points of each codeword appear in a reordered series on the interleaver output lead 24, but now advantageously far apart in time and clearly. J signal intervals apart. This process is obviously repeated for successive frames of J codewords.

Ideally, the effectiveness of the interleaver is maximized when the parameter J is greater than or equal to 1/4 of the minimum velocity of the medium of interest times the signal rate times the carrier wavelength. . (This formula states that, as when Frequency Division Multiple Access (FDMA) schemes are used,
Based on the assumption that there is only one user per mobile communication channel. A consideration for the case of the so-called TDMA scheme, where some users are time-multiplexed on one channel, will be dealt with at a more convenient point below. )

However, in certain applications, a value of J less than this optimum value must be used to reduce the transmission delay introduced by the interleaver / deinterleaver. (This is necessary to ensure the desired level of data throughput, ie to avoid unnatural speech delays that would otherwise be introduced into the conversation.)

Finally, the interleaver leads 2
The rearranged signal points Q _n output on 4 are applied to the π / M shift M-DPSK modulator 25. The phase of the carrier is shifted from the phase during the previous signal interval by 2πQ _n / M increased by the constant π / M radian. According to the present invention, the use of such a π / M shift modulator helps reduce the ratio of peak power to average power and alleviates potential receiver timing recovery problems.

Furthermore, according to the invention, the fact that the transmitted signal points are interleaved advantageously removes in the decoder the correlation between the noise samples introduced by the M-DSPK demodulation process.

Code I has a bandwidth efficiency of 1.5 bits / signal spacing and, for a mobile channel with a medium speed of 20 (60) miles an hour, assuming a bit error rate of 10 ^-3 , It has a coding gain of 8.9 (14.3) dB, which is a so-called uncoded 4-DPSK method with the same information bit rate. These coding gains are 37 ms, which is determined by human factors, especially speech delay.
A particular "interleaving length" (given by the interleaving frame size, in this case 2J, divided by the signal rate). These coding gains assumed a carrier frequency of about 900 MHz. From now on, all coding gains in other codes are noted under these conditions.

Here are some alternative block coded modulation schemes suitable for use in fading channels.
These codes achieve varying degrees of greater bandwidth efficiency and / or coding gain compared to this first code by accepting somewhat higher decoder complexity. (For this purpose, the complexity of the decoder is understood to be the number of additions and / or comparisons per information bit that need to be made in performing the decoding process described below.)

A schematic representation of a second code called II is shown in FIGS. 6 and 7. This is the 8th dimension (8
D) code, which is initially a 2 of the type shown in FIG.
This is accomplished by defining a constituent 4-dimensional 8-PSK array formed by selecting particular elements from a concatenated pair of dimensional 8-PSK arrays. The desired eight-dimensional array is then similarly formed by selecting specific elements from the concatenated pairs of four-dimensional arrays. Each element of this 8-dimensional code is used as a code word of this 8-dimensional code.

In particular, as the elements to be included in the four-dimensional array, 2
³ × 2 ³ = select half of 64 possible 4D elements.
They are elements in the form of (a, b). Here, "a" and "b" are the two-dimensional 8-PSK shown in FIG.
Two-dimensional signal points with both even numbers in the array or two with both odd numbers
It is one of the dimensional signal points. So, for example, (0,
0) and (5, 7) are signal points of a four-dimensional array, but (2,
5) and (7,0) are not. As shown in FIG.
The 32 elements of the four-dimensional array are the four subsets S ₀ ,
..., it is divided into S _3.

Finally, as what is included in the 8-dimensional array,
Select 1/4 of 2 ⁵ × 2 ⁵ = 1024 possible 8-dimensional elements. These 256 elements are the elements of the four 4D subset pairs shown in the codebook of FIG. Specifically, the connection of a specific pair of four-dimensional elements is performed only when the pair of four-dimensional subsets to which the first and second four-dimensional elements belong is one of the four patterns shown in FIG. Is an element of an 8-dimensional array.

Thus, for example, (0,0,1,5) is
a) (0,0) and (1,5) are both elements of the four-dimensional subset S ₀ , and b) the pattern of (S ₀ , S ₀ ) is four allowed four-dimensional parts Since it is one of the patterns of a set pair, it is an element of an 8-dimensional array. On the other hand, (0,
2, 0, 6) is a) (0, 2) and (0, 6) are elements of the subsets S ₀ and S ₃ , and b) (S ₀ , S ₃ ) has 4 allowed patterns. It is not an element of an 8D array because it is not one of the patterns of the 4D subset pair created.

The above method of generating a particular code is generally defined as follows. (A) Select a particular concatenation of the signal points in the first array to make the set of first elements. (B) Group the set of elements defined immediately before into a subset. (C) Select at least some of the elements of the selected concatenation of the subset of step (b).
Each such element is a sequence of elements each taken from a respective subset of the concatenation. (D) Steps (b), (c) and (d) are repeated until a 2N-dimensional element is formed. As a result, those elements become 2N-dimensional codewords.

The eight-dimensional array formed in this way is 2 ⁸ =
Having 256 codewords, this code is capable of communicating 8 information bits for each 8-dimensional codeword. Generally, a method of assigning a bit pattern to a specific code word is used. However, in order to reduce the implementation complexity (as well as exploiting the possibility of using Gray codes as explained above), the coding / mapping process is split into two stages. As shown in FIG. 7, 2 bits are
It is used to select a particular 4D subset pair, while as shown in FIG. 6, 3 bits is 1 out of 8 elements of each of the 2 4D subsets selected. Used to select. A total of 6 bits are used at this stage, and of course all 8 bits are used together.

Like the first code, the second code exhibits twice the built-in time diversity, as will be seen below. First of all, each of the four-dimensional subsets shown in FIG. 6 has itself twice the built-in time diversity. (In fact, it can be seen that the four-dimensional subset S ₀ itself is the same as the code I described above. Furthermore, each of the other three subsets can be used instead.) Thus
Any pair of 4D subsets is guaranteed to show twice the built-in time diversity.

In particular, if two two-dimensional elements belong to the same four-dimensional subset pair, ie come from the description in the same row of FIG. 7, those two eight-dimensional elements are at least at two-dimensional signal point positions. It is easy to see that two, and in some cases all four, are different. In addition, if two eight-dimensional elements belong to different four-dimensional subset pairs, i.e. come from the description in the different rows of Figure 7, then those two eight-dimensional elements are
At least one two-dimensional signal point position in the first constituent four-dimensional element, and at least one in the second constituent four-dimensional element
It is also easy to see that the two two-dimensional signal point positions are different. This results from the fact that each four-dimensional subset appears only once in each of the two columns of the codebook of FIG. Thus, as mentioned above, every two 8
The dimension codewords are guaranteed to differ from each other in at least two two-dimensional signal point positions.

The interleaving of this code is performed using the interleaving method described above, where the codewords are read into the interleaver row by row, one codeword per row, and then first of all codewords. Performed by using the above-described interleaving method, which is conceptualized as being read in a column, the two-dimensional element of 1, then the second two-dimensional element of all codewords, and so on. Instead, FIG.
And the interleaving scheme described below in connection with reference numeral III in FIG. 9 may be used.

In describing the type of code just described, there are many ways in which it is possible to provide twice the built-in time diversity. For example, a) selecting a four-dimensional array, b) selecting a particular number of subsets that will divide the selected four-dimensional array into it, and c) selecting the selected four-dimensional array as a fraction of the aforementioned number. Split into sets,
Or d) a specific 8-dimensional array for the codebook,
That is, there are many possible ways to select a particular 4-dimensional subset pair.

For example, the 32 element four-dimensional array is composed of 32 elements which are not selected in the above example. For example (S
₀ , S ₀ ), (S ₁ , S ₁ ), (S ₂ , S ₃ ), and (S ₃ ,
S ₄ ), such as 4 4 that meet the criteria discussed above
Different pairs of dimensional subsets can be used as codebooks. On the other hand, another method of providing twice the built-in time diversity shows significant differences in coding gain (as described below), bandwidth efficiency, and complexity.

Reference numeral II in FIGS. 6 and 7 indicates 2.0 bits /
It has a bandwidth efficiency of signal spacing and a coding gain of 8.4 (14.5) dB at a medium speed of 20 (60) miles per hour. Thus, it can be seen that this code provides greater bandwidth efficiency while achieving essentially the same coding gain as code I of FIG. However, this "improvement" is
It was achieved at the expense of some complexity as compared to code I and some poorer performance at 20 mph or less (eg, code I was 7.2 dB vs. code II at 5.4 dB at 10 mph).

FIGS. 8 and 9 give the same bandwidth efficiency (1.5 bits / signal spacing) as code I, but again at the cost of some increased complexity to achieve a sufficiently large coding gain. Another 8 with double built-in time diversity
A dimensional code, Code III, is depicted.

As shown in FIG. 8, the code III is initially four subsets S of the 32-element 4-dimensional 8-PSK array of FIG.
Each of ₀ , S ₁ , S _2, and S ₃ is constructed by dividing it into four smaller subsets, each consisting of two elements. The resulting finer subset is T _0.
Through T ₁₅ . As shown in the codebook of FIG. 9, the concatenation of a particular pair of four-dimensional elements can be done in the first and second
The four-dimensional subset pairs to which the four-dimensional elements of
It is an element of an 8-dimensional array only if it is one of the 16 patterns shown in FIG.

There are 64 codewords in total in the 8-dimensional array.
As further shown in FIG. 9, 4 bits are used to select a particular 4-dimensional subset pair, while as shown in FIG.
Two bits are used to select one of the two elements of each four-dimensional subset, for a total of six bits.

Conveniently, this code is 20
It has a code gain of 10.1 (17.1) dB for the mobile channel at a medium speed of (60) miles. It can thus be seen that this code gives a greater coding gain than either of the two codes mentioned above.

To fully understand how these larger coding gains occur, it is useful to define parameters that are a) easy to calculate and b) give an indication of the channel fading performance. . This parameter, called "least squares Euclidean distance at X times time diversity", or MDX, is
It is the least squared Euclidean distance between any two codewords that are useful for designing codes with X times time diversity for fading channels and that differ exactly from each other by X signal point positions.

Generally, the larger this distance, the larger the coding gain. Other factors, including, for example, medium speed, interleaving length, number of code dimensions, and total minimum Euclidean distance between every two codewords also affect the coding gain, resulting in the same MDX. The three codes show somewhat different coding gains. However, in general, there is a high correlation between MDX and coding gain as long as the interleaving length is sufficient to ensure that the signal points for a particular codeword are fading independently.

Given an assumed interleaving length of 37 ms and a medium speed of 20 mph, such independent fading is indeed guaranteed for eg the 4D and 8D codes described thus far. . Thus, for these three codes 4, 4, respectively
MDXs 8 and 8 lead to substantially the same coding gain for the first two codes and a sufficiently large coding gain for the third code. (Throughout this discussion, all quoted values for MDX were calculated assuming an M-PSK constellation of radius 1.)

With respect to the codes of FIGS. 6 and 7, the interleaving of code III is performed using the interleaving scheme described above. However, according to the invention, the particular way in which the signal points are rearranged by the interleaver is
Increase the effectiveness of the interleaver, i.e. the contribution of the interleaving process to the coding gain, for a given interleaving distance, where the consideration of the system as described above forces the interleaver, or b) It is used to allow shorter interleaving lengths for a given level of coding gain, thereby possibly meeting system requirements that would otherwise not be met and matched to a unique block code.

Looking at the symbols in FIGS. 8 and 9, all codewords that differ only in two signal points have different first and second signal point positions or third and fourth signal point positions. Is observed. This is so, according to the logic previously set for the interleaver of FIG. 5, in particular, as shown in FIG. 10, first the first two-dimensional signal points (of FIG. Column A), then the third two-dimensional signal point (column C) of all codewords, then the second two-dimensional signal point (column B), and the fourth two-dimensional signal point (column D). Moreover, it is effective to rearrange the order in which the sequence of signal points is read from the interleaver. This method has the effect of separating the first signal point and the second signal point of every codeword at 2J signal intervals rather than at J signal intervals. The same applies to the third signal point and the fourth signal point.

In the more general case, to maximize the separation of signal points of codewords that differ by X signal point positions, and thus provide the desired X-fold built-in time diversity,
It will be appreciated that various rearrangement methods may be used within the interleaver, depending on the structure of the codeword. More formally, given a particular one of the codewords of the alphabet that differ exactly by X signal point positions and have inter-codeword distance MDX, two of those X signal point positions are k-th. When shown as the q-th signal point position, the interleaver determines the code of each frame such that the signal points at the k-th and q-th signal point positions are separated by J signal points or more in the rearranged sequence. It operates to generate a rearranged sequence of signal points of a word. Given this general principle, it is unnecessary to further describe the structure of the interleaver that is advantageously used in connection with the various other codes described below.

11 to 13, the code is a 16-dimensional code with double built-in time diversity, which has the same MDX (ie, 4) as the codes I and II.
IV is drawn. However, 16-dimensional codes have significantly higher bandwidth efficiencies (2 + 3/8 bits / signal spacing) than codes described to date at the cost of lower coding gain (at the same interleaving length). give. The complexity of the decoder is also somewhat greater than for the codes already described.

First of all, a four-dimensional 8-composed of 64 elements formed from concatenated pairs of the two-dimensional 8-PSK array of FIG.
Define the PSK sequence. The four-dimensional array is then shown in FIG.
As shown in FIG. ₇ , the data is classified into eight subsets S _{0 to} S ₇ . (Note that the subsets S ₀ through S ₃ are the same as the four subsets shown in FIG. 6.) Then, the constituent 8-dimensional array has certain elements from the concatenated pair of the 4-dimensional array. It is formed by selecting. In particular, 32 were selected from 64 possible subset pairs to construct an 8-dimensional array. The eight-dimensional array consists of 2 ¹¹ elements, but as shown in FIG. 12, it is classified into eight subsets each consisting of four 4-dimensional subset pairs. 8 dimensional subset of ^{_{eight, S '0, S' 1}} , ..., indicated as S ^_'7.

Finally, 2 ¹¹ for inclusion in the 16-dimensional array.
× 2 Select 1/8 of ¹¹ possible 16-dimensional arrays.
These 2 ¹⁹ elements are the elements of the 8 8-dimensional subset pairs shown in the codebook of FIG. Specifically, the concatenation of a particular pair of 8-dimensional elements is such that if the pair of 8-dimensional subsets to which the first and second constituent 8-dimensional elements belong, each of the eight patterns shown in FIG. If one
And in that case only, it is an element of a 16-dimensional array.

As further shown in FIGS. 11-13, 3 bits are used to select a particular 8-dimensional subset pair. Here 2 bits are used to select a particular 4 dimensional subset pair from each of the 2 selected 8 dimensional subset pairs, ie 4 bits in total are used at this level, and 4 more Three bits are used to select a particular pair of two-dimensional points from each of the selected four-dimensional subsets, i.e. a total of 12 bits are used at this level, resulting in a total of 19 bits. Since each codeword extends over 8 signal intervals, the bandwidth efficiency is 19/8 = 2 + 3/8 bits / signal interval.

This code is 6.8 (13.9) d in a moving channel at a medium speed of 20 (60) mph.
It has a coding gain of B.

Up to this point, various codes have been described using a quasi-graphic method. Instead of the block encoder /
Each of the codes can be described using some simple Boolean expressions that define the operation of the many logical elements that make up the mapper, eg, the block encoder / mapper 13 of FIG. This latter method is now used to describe the next code because the graphical method for the next code is complicated to represent. It is understood that any prior art code has also been described using this latter method. vice versa,
The Boolean expressions that proceed below can be used to build quasi-graphic representations made for other codes.

In particular, this next code is another 16-dimensional 8-PS with double built-in time diversity.
K code, and this code is the code of FIGS. 8 and 9.
It has the same MDX as III (ie 8). This code is
It has neither the highest bandwidth efficiency nor the highest coding gain, but conveniently provides a useful combination of both of these parameters. Specifically, 8.5 (20 mph)
And a coding gain of 16.7 (60 mph) and 20
Has bandwidth efficiency of bit / signal spacing. The decoder complexity is the highest of all the codes described.

As shown in FIG. 14, the lead wire 11 of FIG.
16 input bits integrated over the 8 signal intervals above
In addition to the circuit, this circuit has leads 1341-2,1
The three bits on each of 351-2, 1361-2 and 1371-2 make up the particular 8 on lead 16 of FIG. 1 which make up the 16-dimensional codeword associated with the 16 input bits.
Finally, the two-dimensional points are specified.

(As in FIGS. 15 and 21 described below), in FIG. 14, the number of bits shown as being carried over a particular set of leads, eg lead 11, is N. Note that this is the number of bits collected over the signal interval, which is, for example, the number of bits indicated as being carried by a particular set of leads is the average number of bits per signal interval. (Symmetrical to FIG. 1.)

As in the case of the codes already described (graphically), the mapping is performed in stages. In particular, the 16 bits on lead wire 11 are output lead wire 131
A 16-dimensional to 8-dimensional array mapper converter providing 10 bits to each of 1 and 1312. These ten
Each of the bits is associated with a particular element of the constituent 8-dimensional array. Only particular combinations of these eight-dimensional elements appear on leads 1311 and 1312. As will become more apparent below, this constraint corresponds, for example, to the constraint labeled IV in FIGS. 11-13, in which a 16-dimensional array consists only of pairs of a particular selected 8-dimensional subset.

In the next step, the leads 1311 and 13
The bits on 12 are applied to 8-D to 4-D array mapper converters 132 and 133, respectively. For example, converter 132 provides 6 bits on each of its output leads 1321 and 1322. Each of these 6 bits is associated with a particular element of the constituent four-dimensional array, and, again corresponding to the way the code IV is constructed, only certain combinations of these four-dimensional elements lead 1321 and Appears on 1322. The converter 133 is the same as the converter 132. Only one such converter is used, depending on the implementation.
The two groups of 10 bits, each on leads 1311 and 1312, are subsequently sent to this one converter.

In the final step, the bits on leads 1321-2 and 1331-2 are added to four-dimensional array mappers 134-137, respectively. For example, mapper 13
4 supplies 3 bits on each of its output leads 1341 and 1342. Each of these 3 bits is associated with a particular point in the 8-PSK constellation of FIG. 3, and all 2 ⁶ = 64 combinations of two 2-dimensional points are used.
The mappers 134 to 137 are the same. As with the converters 132 and 133, only one such mapper is used.

16-dimensional to 8-dimensional array mapper converter 13
Examples of 1-, 8-D and 4-D array mapper converters 132 and 4-D mappers 134 are shown in FIGS. 15-17, respectively.

Referring first to FIG. 15, the 16 bits on lead 11 are divided into three groups. The first group is I0 _{n to} I3 _n added to the 8-dimensional subset pair selector 231.
It consists of 4 bits. I4 _{n to} I9 _n (I10 _{n to} I
15 _n) second group consisting of 6-bit (third group) is simply the output bits X4 _n through X9 _n (X4 _{n + 4} through X9 _{n + 4)} passes. The output bit of the selector 231 has two 4 bits.
Groups of bits, namely X0 _{n to} X3 _n and X0 _{n + 4 to} X3 _{n + 4}
Is.

The 10 bits with the subscript "n" specify the first constituent eight-dimensional element of the codeword consisting of four two-dimensional signal points P _{n to} P _{n +3} (see FIG. 1). Finally used for. Similarly, the subscript “n +
10 ”with“ 4 ”is four two-dimensional signal points P _{n + 4}
Finally used to identify the second constituent 8-dimensional element of the codeword consisting of P _{n +7} . In particular, bit X0 _n
Through X3 _n specify a particular one of 16 eight-dimensional subsets, while bits X4 _n through X9 _n finally identify a particular one element of that eight-dimensional subset. The same applies to the 10 bits with the subscript “n + 4”.

The selector 231 uses the following Boolean expression:
Give output bits.

[Equation 1]

10 bits of X0 _{n to} X9 _n are converted by the converter 1
Added to 32. These bits are also divided into three groups, as shown in FIG. The first group consists of 8 bits X0 _{n to} X7 _n added to the four-dimensional subset pair selector 232. The second group (third group) consisting of 1-bit X8 _n (X9 _n ) is simply passed and output bit Y5 _n (Y
5 _{n + 2} ). The output bits of the selector 232 are 5 each.
There are two groups of bits, Y0 _{n to} Y4 _n and Y0 _{n + 2 to} Y4 _{n + 2} .

The 6 bits with the subscript "n" specify the first constituent four-dimensional element of the first constituent eight-dimensional element of the codeword, which consists of two two-dimensional signal points P _n and P _{n + 1.} Finally used to. Similarly, the 6 bits with the subscript “n + 2” represent two two-dimensional signal points P.
Used to identify the second constituent four-dimensional element of the first constituent eight-dimensional element of the codeword consisting of _{n + 2 and Pn + 3} .

Bits Y0 _{n to} Y4 _n identify a particular one of the 32 four-dimensional subsets, while one bit Y5
_n specifies a specific one element of the four-dimensional subset.
The same applies to the 6 bits with the subscript “n + 2”.

Selector 232 provides its output bit according to the following Boolean equation:

[Equation 2]

The 6 bits Y0 _{n to} Y5 _n are added to the mapper 134 as shown in FIG. Mapper 134
Output bits of two groups of 3 bits each, namely Z0 _n
Through Z2 _n and Z0 _{n + 1} through Z2 _{n + 1} . The 3 bits with the subscript “n” define the two-dimensional signal point P _n . Similarly, the 3 bits with the subscript “n + 1” define the two-dimensional signal P _{n + 1} . Mapper 134 provides its output bits according to the following Boolean expression:

[Equation 3]

The graphical representation of the sixth code, code VI, is
It is shown in FIGS. Compared with the reference numeral II in FIGS. 6 and 7, for example, this is 12 points for 8 points (M = 1
2) Built-in time to achieve high bandwidth efficiency of 2 + 3/8 bits / signal spacing, at the cost of larger M-PSK array size, 2) slightly higher complexity and some reduction in coding gain. It is another eight-dimensional code with diversity.

As with the other eight-dimensional codes described above, the code VI
Is reached by first defining a four-dimensional array formed by selecting specific elements from a concatenated pair of two-dimensional arrays. However, as already noted, this two-dimensional array is a 12-point array, specifically 12 shown in FIG.
It is a PSK sequence. This code is thus the value of M is 2
It is unique in that it is not an integer power of. The desired 8-dimensional array is then formed by similarly selecting a particular element from a concatenated pair of 4-dimensional arrays. Each element of this 8-dimensional array is used as a code word of this 8-dimensional code.

In particular, we use 1 for inclusion in a 4-dimensional array.
2. Select half of 2 ² = 144 possible 4D elements.
Like in VI of FIGS. 6 and 7, they are
Explanatoryly, elements of the form (a, b), where "a" and "b" are both even two-dimensional signal points or both odd two-dimensional 12-PSK arrays shown in FIG. It is one of the two-dimensional signal points. As shown in FIG. 19, 72 elements of the four-dimensional array are divided into 6 subsets S ₀ , ..., S ₅ each having 12 elements.

Finally, we have 768 out of 72 ² = 5184 possible 8D elements to include in the 8D array.
Select the pieces. The reason for using 768 will be described below. These 768 elements are specific among the elements of the 6 four-dimensional subset pairs shown in FIG. Note the difference between this code and all the other codes previously described. In other 8-dimensional codes, every element of each of the selected 4-dimensional subset pairs is used, i.e., a codeword. However, in the case of FIG. 20, this method yields 144 codewords from each 4-dimensional subset pair, for a total of 864 codewords. However, only 768 codewords are desired. Therefore, in this case, FIG. 2 selected as described below.
Only 128 elements of each 4-dimensional subset pair of 0 are used.

From the above, it will be appreciated that 10 bits are needed to identify a particular codeword. Specifically, 3 bits are used to select a particular 4-dimensional subset pair (FIG. 20), and each such pair has 2 ⁷ = 128 elements, so 7 bits of are required to select a particular codeword.

However, not all possible combinations of these 10 bits are allowed, ie 1
Note that only 768 codewords are available, but 024 combinations. In fact, even if all 144 elements of each 4-dimensional subset pair are included in the code, there is still an insufficient number of code words, ie 144 × 6 = 864.
Is. (The use of all 864 codewords results in slightly higher bandwidth efficiency, but requires additional processing and incurs additional processing delay.)

The task at this point is then to provide an easily implemented method of generating 768 10-bit word patterns every 4 signal intervals from the input bit stream. This is accomplished according to the technique of processing 19 bits captured over 8 signal intervals to produce two 8D codewords, which results in the above 19/8 = 2 + 3.
A bandwidth efficiency of / 8 bits / signal spacing is achieved.

Specifically, FIG. 21 shows that the input bit is taken into the input lead 201 and the signal point P _n ,
, _{Pn + 3} is shown on the output lead 219. Specifically, this circuit comprises a precoder 208, a parallel-to-serial group converter 209, and a block encoder / mapper 213 corresponding to the block encoder / mapper 13 of FIG.

The input bits on the lead wire 201 are divided into three groups of 8, 8 and 3 bits, respectively. 8-bit 2
The group simply passes through the precoder. The group of 3 bits is a bit pattern selector 2 which supplies 2 groups of 2 bits each on its output leads 2082 and 2083.
081 added. The 2 bits on the lead wire 2082 takes one of the three patterns "00", "01" and "10", but does not take "11". The same applies to the two bits on the lead wire 2083. In addition, pattern "10" cannot appear on leads 2082 and 2083 at the same time. Thus there are 2 ³ = 8 different possible combinations of the values of the bits on leads 2082 and 2083.

The output bits of the precoder 208 are each 1
It is made into two groups of 0 bits, and those groups are two consecutive 8
Sequential parallel-to-serial converter 2 for identifying the dimension codeword
09 to the encoder / mapper 213. To this end, each group of 10 bits has a lead 2082 or 2
It is divided into a 3-bit group consisting of 2 bits from one of 083 and 1 bit of the other precoder input bits, and a 7-bit group. The 3-bit group can thus take 3 × 2 = 6 possible patterns. They were used to select a particular 4-dimensional subset pair as shown in FIG.
Are possible patterns. The 7-bit group is used to select a particular codeword from the selected group.

Specifically, the 3-bit group and the 7-bit group are the 8-dimensional-4 in the block encoder / mapper 213.
Added to the dimensional array mapper converter. The 3-bit group is a 4-dimensional subset pair that supplies via lead 2121 a 3-bit word on output lids 2134 and 2135, each of which specifies one of the selected 4-dimensional subsets of the pair. It is added to the selector 2133.

The remaining 7 bits are divided into 3 groups of 3, 2, and 2 bits, respectively. The two groups of 2 bits simply pass through the converter 2131. The group of 3 bits is added to a bit pattern selector 2132 shown as identical to the bit pattern selector 2081. Thus there are three possible combinations of 2-bit values on output lead 2136, as well as on lead 2137.

Each output lead 2136 and 2137
The upper 2 bits are the 2 on leads 2138 and 2139.
Are grouped with one of the two 2-bit groups, each 3 ×
Given a 4-bit word estimated to be one of the 4 = 12 bit patterns. Each such 12-bit pattern is used to select a particular element from a particular subset of the four-dimensional subset identified by the bits on leads 2134 and 2135.

Lead wires 2136, 2138 and 2134
The upper 7 bits are applied to a four-dimensional array mapper 2141 which supplies on its output lead two two-dimensional signal points P _n and P _{n + 1} forming the first constituent four-dimensional element of the 8-dimensional codeword. . Similarly, leads 2137, 2139 and 213
The 7 bits on 5 provide the four-dimensional array mapper 2142 which supplies on its output lead the two two-dimensional signal points P _{n + 2} and P _{n + 3} forming the second constituent four-dimensional element of the eight-dimensional codeword. Added to. The four-dimensional array mappers 2141 and 2142 are the same. As with the symbol V, only one such mapper is used.

The allocation of bit patterns to the four-dimensional subset pairs shown in FIG. 20 and the allocation of bit patterns to the elements of each four-dimensional subset shown in FIG. 19 are arbitrary. However, this allocation method shown in FIGS. 19 and 20 ensures that there is no DC energy of code origin in the transmitted signal.

In particular, we find that 4 in a particular 4D subset
We start by recognizing that not all dimensional elements are used with the same probability. This corresponds to the input lead wires 2136 and 2 of the four-dimensional array mapper 2141 (2142).
This results from the fact that the bit patterns on 138 (2137 and 2139) do not occur with the same probability. Thus, in particular, the four bit patterns "1000", "10"
10 "," 1001 ", and" 1011 "are probability 1 respectively.
/ 16, but the other eight bit patterns each occur with a probability of 3/32.

This situation causes a lack of DC balance in the code if care is not taken when assigning a particular bit pattern to a particular element of the four-dimensional subset. However, in the preferred embodiment, the sum of the first coordinates of the four-dimensional elements is zero, and the same for the second, third and fourth coordinates, so that bit patterns with equal probabilities are shown in FIG. Care is taken to ensure that it is assigned to each 4D element of the 4D subset of.

For example, as described above, two bit patterns "100" each of which has a probability of 1/16.
Consider "0" and "1010". Referring again to FIG. 18, further two-dimensional signal point is four-dimensional elements having a radius Assuming that on a circle "A", four-dimensional subset S ₄ which is allocated the bit pattern "1000" (0,0) It can be seen that the coordinates of are (A, 0) and (A, 0). Similarly, it can be seen that the coordinates of the four-dimensional elements (6, 6) of the same four-dimensional subset S ₄ to which the bit pattern “1010” is assigned are (−A, 0) and (−A, 0). . Applying the above criteria, the sum of the first coordinates is (A) + (-
A) = 0, the sum of the second coordinates is (0) + (0) = 0, the sum of the third coordinates is (A) + (− A) = 0, and the sum of the fourth coordinates. Is (0) + (0) = 0.

The criteria described above guarantee long-term DC balance. Imposing further restrictions on the assignment of bit patterns to specific 4D elements across various 4D subsets as shown in FIG. 19 and to the assignment of bit patterns to 4D subset pairs as shown in FIG. This has the effect that short-term DC balance is also achieved.

It should be noted in FIG. 19 that six 4D elements, any of which in different 4D subsets begin with a particular 2D element, are assigned to one unique bit pattern. Therefore, for example, all the four-dimensional elements assigned to the bit pattern “0110” are “8” as the first two-dimensional signal points.
Have.

Furthermore, it should be noted that not all 4-dimensional subsets are used with equal probability. This means that the four-dimensional subset pair selector 2133
Of the input leads 2121 of the above does not occur with equal probability. Thus, in particular, "1
The two bit patterns "00" and "101" each occur with a probability of 1/8, while the other four bit patterns occur with a probability of 3/16.

In FIG. 19, for every four-dimensional element of the subset S ₄ having a particular first two-dimensional point, it has the same first two-dimensional point and a balanced second two-dimensional point. Note that four-dimensional elements are assigned to subset S ₅ . Thus, for example, subset S ₄ contains four-dimensional element (1,7), and thus subset S ₅ contains four-dimensional element (1,1). As a result of the foregoing, the subset S ₅ exhibits more “DC balance” with the subset S ₄ than with the other subsets.

Similarly, the four-dimensional subset S ₀ is a subset S ₃
And “DC balance”, and the four-dimensional subset S ₁ represents “DC balance” with the subset S ₂ . In the preferred embodiment shown in FIG. 20, selecting six 4D subset pairs to form an 8D array, and a new "DC balanced" 4
The dimensional subsets S ₄ and S ₅ are used with equal probability 1/8 and the other four 4-dimensional subsets are equal probability 3/1.
Care is taken to assign the input bit pattern to the selected 4-dimensional subset pair as used in 6.

This code is 6.7 (12 ..) for moving channels at medium speeds of 20 (60) miles per hour.
1) It has a coding gain of dB. MDX of this code is 2
And, in fact, the coding gain of this code is the smallest of all the codes described here.

Referring again to FIG. 1, the preferred structure of block decoder 51 will now be described. The same basic type of structure can be used for all codes described in this document. The decoding process for the assumed channel and modulation / demodulation type, eg mobile channel using M-DPSK with non-coherent detection, is similar to the decoding process used for AWGN channels. This method is not suitable for mobile channel degradation and non-coherent M
-The DPSK demodulation process is not optimal in that it introduces noise that does not fit the AWGN model. However, this method has the advantage of being simple to implement and exhibits at least near-optimal performance.

Decoding proceeds by examining each of the two-dimensional points of the particular received signal corresponding to the transmitted 2N-dimensional codeword. In particular, the so-called "two-dimensional point distance" is
For each 2D receiving point, it is calculated by measuring the squared Euclidean distance between the receiving point and all possible transmitting 2D points. In carrying out this calculation, a particular radius of the transmission constellation is assumed, and the receiver is given an output constellation whose long-term average radius is at least approximately equal to the radius assumed during decoding. An automatic gain control (not shown) is provided for this purpose.

Decoder performance is not sensitive to this exact setting of the receive constellation radius. Further, in order to reduce the contribution of the faded (and hence potentially unreliable) signal points to the decoding process, each of the two-dimensional point distances is
It may be weighted by a factor proportional to the amplitude of the corresponding received two-dimensional point.

Received N 2's corresponding to the transmission codeword
The dimensional received signal points are sequentially classified into N / 2 four-dimensional elements. Each of the N / 2 four-dimensional elements is further processed as follows. For each 4D subset, one of that subset "closest" to the received 4D element being processed
Find the element. This is achieved by taking each element of the four-dimensional subset in order and finding the sum of the two two-dimensional point distances corresponding to that element. 4 corresponding to the minimum of such sums, which is referred to below as the "four-dimensional subset distance"
The dimensional element is identified as the "closest" four-dimensional element. Therefore, as a result so far, for each received four-dimensional element,
Specific 4D elements and their associated 4D subset distances in each of the 4D subsets are identified.

If the code is a four-dimensional code, there is only one four-dimensional subset. Therefore, only one 4D element is identified and that element is taken to be the transmission codeword. If the code is an 8-dimensional or higher code, the above process is repeated for each of the received 8-dimensional elements.

In particular, the received N / 2 four-dimensional receiving elements are sequentially grouped into N / 4 eight-dimensional elements, each of which is then processed in a manner similar to the four-dimensional case. It In particular, for each 8-dimensional subset, find the one 8-dimensional element of that subset that is "closest" to the received 8-dimensional element being processed. This is 4 of the 8D subset
This is accomplished by taking the dimension subset pairs in order and summing the two four-dimensional subset distances corresponding to that pair. The 8 dimensional element in the 4 dimensional subset pair is one element of each 4 dimensional subset obtained from the previous step 4
A pair of dimensional elements, corresponding to the smallest such sum, and further called the "8 dimensional subset distance",
Identified as the "closest" eight-dimensional element.

If the code is an 8-dimensional code, then an 8-dimensional element has now been identified. Otherwise, the process is repeated again for 16 dimensions and so on.

The identified codeword is then, for example, a code
Mapped to data bits based on the bit allocation method for the code as in FIGS. 14 to 17 for V.

The decoding process described above can be performed in a highly parallel manner, and thus a more complex and therefore more effective block coding scheme and / or a practical implementation of higher input data rates. This has the effect of reducing the time required for decoding that enables In particular, the two-dimensional element distances can all be calculated simultaneously for each of the N two-dimensional points of the received 2N-dimensional signal. The same applies to the 4-dimensional and 8-dimensional subset distances. Furthermore, once multiple 2N-dimensional signals are available at the deinterleaver output, each of those signals is decoded in parallel independently of the others. This property of block codes is not possible with trellis codes, which require processing each two-dimensional point in series, for example.

According to the invention, this type of code can be combined with spatial diversity to further increase performance. In particular, receiving stations, such as automobiles, are provided with two or more receiving antennas separated by at least a quarter wavelength, so that the output received from each antenna fades independently of each other.

Then, the first step of the decoding process is performed on the signals received from each antenna. The resulting "preliminary" two-dimensional point distance associated with each point in the array is added to the two-dimensional point distance for points obtained from other antennas. The resulting sum is the "final" two-dimensional point distance used as the two-dimensional point distance in the subsequent decoding process steps.

The foregoing merely describes the gist of the present invention. For example, although the present invention has been described primarily in the context of digital cellular mobile communications, it is equally applicable to other fading signal environments. Further, while the above description assumed that there is only one user per mobile communication channel,
If a frequency division multiple access (FDMA) scheme is used, the invention is equally applicable to the so-called TMDA scheme, in which some users are time multiplexed onto one channel. In such applications, interleaver design and T
The MDA frame formats match each other with the largest possible size.

In particular, the interleaving length is TMDA
It is an integral multiple of the frame length, and is large enough to guarantee an appropriate time interval between signal points of the codeword. This leads to the requirement that signal points from one codeword be spread out in two or more TMDA frames. Instead, the number of time slots assigned to each user in a TDMA frame, which is divided into as many frames as possible, is 1
Increased to a larger number, thereby reducing interleaving length requirements and / or increasing performance. These considerations favor the use of coding schemes with the lowest possible dimensions that allow the achievement of other system goals.

Trellis coding schemes generally require a larger number of time slots per TMDA frame and per user that either require larger interleaving lengths that introduce delays or that do not match other system requirements. So, these same considerations make the trellis coding scheme less attractive than the block coding modulation scheme of the present invention.

A further advantage of the use of block coded modulation, for example when compared to trellis coded modulation, is that once the signal captured in the interleaving frame is received, unlike the case of trellis coded information Is to be decrypted immediately. This property of block coded modulation avoids delays that might otherwise occur with trellis coding schemes if the system has a channel decoder followed by a speech decoder.

Other variations are possible. For example, M = 4
Various values of the parameter M can be selected, such as 6, 8, 12, 16 and so on. Any integer greater than 1 can be used as the parameter N. In each of the M and N combinations, a variety of different codes are reached that give different combinations of bandwidth efficiency, coding gain and implementation complexity. X is 2
Codes with X times time diversity that are other than are similarly constructed. Only a few of the possible codes have been given in this. The particular code used in a particular application selected is a function of the particular needs and restrictions of that application.

As an example, increasing the signal rate, or the reciprocal of the signal spacing, introduces and / or increases intersymbol interference.
Relatively high, even at the expense of a possible increase in complexity and / or a possible decrease in coding gain, in order to maintain a given data rate without encountering an unsuitable amount of intersymbol interference. Choose a code that has bandwidth efficiency, so that the signal rate is kept acceptably low.

Furthermore, the diversity built into the code described in this document is in the time domain (by ensuring that the transmission of the signal points giving diversity takes place in well-spaced signal intervals). However, the diversity can be used in other areas, for example, by transmitting those signal points by physically separated channels and fading independently. Thus, while the codes disclosed therein have been described as having built-in time diversity, more generally, built-in time diversity can be described as one example, "built-in diversity".

The above description relates to one embodiment of the present invention, and those skilled in the art can think of various modified examples of the present invention, all of which are within the technical scope of the present invention. Included in. It should be noted that the reference numbers in the claims are for easy understanding of the invention and should not be construed to limit the technical scope thereof.

As described above, according to the present invention, the block coding modulation technique with interleaving having built-in time diversity is applied to a fading channel. This results in excellent coding gain with low complexity implementation, low delay in interleaver / deinterleaver and encoder and high bandwidth efficiency.

[Brief description of drawings]

FIG. 1 is a block diagram of a data communication system showing an embodiment of the present invention.

FIG. 2 is a diagram illustrating technical terms and concepts.

FIG. 3 is a diagram showing an 8-PSK constellation that forms the basis of many block codes disclosed herein.

FIG. 4 shows a codebook for a first block coded modulation method according to the present invention.

FIG. 5 is a diagram showing the operation of the interleaver shown in FIG. 1 with respect to the first block coded modulation method.

FIG. 6 is a diagram graphically showing a second block coding modulation method according to the present invention.

FIG. 7 is a diagram graphically showing a second block coding modulation method according to the present invention.

FIG. 8 is a diagram graphically showing a third block coding modulation method according to the present invention.

FIG. 9 is a diagram graphically showing a third block coding modulation method according to the present invention.

FIG. 10 is a diagram showing an operation of an interleaver regarding a block code modulation method.

FIG. 11 is a diagram graphically showing a fourth block coding modulation method according to the present invention.

FIG. 12 is a diagram graphically showing a fourth block coding modulation method according to the present invention.

FIG. 13 is a diagram graphically showing a fourth block coding modulation method according to the present invention.

FIG. 14 is a block diagram of the encoder / mapper of FIG. 1 for a fifth block coded modulation method according to the present invention.

FIG. 15 is a detailed view of various portions of the encoder / mapper of FIG.

16 is a detailed view showing various parts of the encoder / mapper of FIG.

FIG. 17 is a detailed view showing various parts of the encoder / mapper of FIG.

FIG. 18 is a diagram showing a 12-PSK array that forms the basis of a sixth block coded modulation method according to the present invention.

FIG. 19 is a diagram graphically showing a sixth block coding modulation method according to the present invention.

FIG. 20 is a diagram graphically showing a sixth block coding modulation method according to the present invention.

FIG. 21 is a circuit block diagram for carrying out a sixth block coded modulation method according to the present invention.

[Explanation of symbols]

 11 Input Data 13 2N-Dimensional Block Coder / Mapper 21 Interleaver 30 Fading Channel 44 Deinterleaver 51 Block Decoder

Claims (26)

[Claims] 1. A receiving means for receiving a succession of blocks of input data and defining the character system of the code word, each of which is associated with one of the possible values of this input data block, and said input data block comprising: Defining and generating means for generating the associated ones of the codewords corresponding to the blocks, wherein each of the codewords of the alphabet is at least two modulation signal points taken from a predetermined array of modulation signal points. And the character system of the codeword has a built-in X times diversity, where X is an integer greater than 1, and the least square Euclidean distance between the codewords is greater than the absolute least square Euclidean distance for the character system. A device characterized by the above. 2. The apparatus according to claim 1, further comprising transmission means for transmitting a signal representing the generated codeword. 3. The input data block is received at a rate of m bits every T seconds, which is a predetermined signal interval, and the array includes 2 ^m modulation signal points. 1. The device according to 1. 4. The apparatus of claim 1, wherein the array of modulation signal points is an M-PSK array, where M is a selected integer. 5. The above-mentioned definition and generation means encodes each of the input data blocks into blocks of coded bits, and also signals each consisting of one codeword generated corresponding to each block of coded bits. By selecting a point,
An apparatus as claimed in claim 1, characterized in that it produces a related one of the code words. 6. The apparatus of claim 1, wherein each codeword of the alphabet differs from each of the other codewords of the alphabet at least at a particular signal point position X. 7. Each of the codewords of the script differs from each of the other codewords of the script at least at position X of a particular signal point, and the apparatus is generated during each signal interval. The apparatus according to claim 1, further comprising a transmission means for transmitting a signal point including a code word. 8. The generated codewords occur in successive frames of codewords, and further the apparatus is such that signal points at corresponding signal point positions of the codewords in each frame are grouped together. The apparatus according to claim 1, further comprising rearrangement means for rearranging the signal points of the codeword of each frame. 9. The apparatus of claim 8 further comprising transmission means for transmitting a signal representative of the rearranged signal points using carrier phase difference. 10. The apparatus according to claim 8, further comprising a transmission means for transmitting a signal representing the rearranged signal points using a carrier phase difference increased by a constant value. 11. The generated codewords occur in successive frames of codewords, each of the codewords consisting of N signal points,
The above device rearranges the signal points of the codewords of each frame so that the i-th signal points of the codewords of the frame are arranged in their respective groups for i = 1, 2, ..., N. 7. The apparatus of claim 6 further comprising rearrangement means for 12. The generated codewords occur in successive frames of codewords, and the apparatus is arranged so that the codewords of each frame are increased so that the separation between signal points in each codeword is increased in a rearranged sequence. 7. The apparatus of claim 6, further comprising generating means for generating a rearranged series of signal points of 13. The generated codewords each occur within a continuous frame of J codewords, particular ones of the codewords of the alphabet having two of them being the kth and qth signal point positions. The above device reorders the signal points of the codeword of each frame so that the signal points at the k-th and q-th signal point positions are separated by J signal points or more in succession. 7. The apparatus according to claim 6, further comprising generating means for generating the generated sequence. 14. A step of receiving a succession of blocks of input data, and a corresponding codeword of a codeword alphabet corresponding to each of the blocks of input data, each codeword being associated with one of the possible values of the input data block. And each of the codewords of the script comprises a combination of at least two modulation signal points taken from a predetermined array of modulation signal points, the script of the codeword having X The coded modulation method is characterized in that it has a built-in X times diversity when an integer larger than 1, and that the least square Euclidean distance between the codewords is larger than the absolute least square Euclidean distance associated with the script. 15. The method of claim 1, further comprising transmitting a signal representing the generated codeword.
4. The method described in 4. 16. The block of input data is received at a rate of m bits every T seconds, which is a predetermined signal interval, and wherein the array includes 2 ^m or more modulated signal points. 16. The method according to claim 15, wherein 17. The method of claim 16, wherein the constellation of modulation signal points is an M-PSK constellation where M is a selected integer. 18. In the step of generating, one of the codewords is configured by encoding each block of input data into a block of coded bits and selecting one of the codewords corresponding to each block of coded bits. 18. The method of claim 17, wherein the method is generated by selecting 19. The method of claim 14, wherein each of the codewords of the alphabet differs from each other codeword of the alphabet at least at a particular signal point position X. 20. Each of the codewords of the character system differs from each of the other codewords of the character system at least at a particular signal point position X, and the method further includes codewords generated at respective signal intervals. 19. The method of claim 18, further comprising the step of transmitting signal points. 21. The generated codewords occur in successive frames of codewords, and the method further comprises grouping signal points at corresponding signal point positions of the codewords in each frame together. The method of claim 19, further comprising the step of rearranging the signal points of the codewords of each frame. 22. The method of claim 21, further comprising transmitting a signal representative of the rearranged signal points using carrier phase difference. 23. The method of claim 21, further comprising the step of transmitting a signal representative of the rearranged signal points using a carrier phase difference increased by a constant. 24. The generated codewords occur in successive frames of codewords, each of the codewords consisting of N signal points and the method, where i = 1, 2, ..., N, 20. The method of claim 19, further comprising rearranging the signal points of the codewords of each frame such that each i-th respective signal point of the codewords of the frame is arranged in a respective group. Method. 25. The generated codewords occur in successive frames of codewords, and the method is such that the separation between signal points in each codeword is increased in a rearranged sequence.
20. The invention of claim 19 including the step of generating a rearranged sequence of signal points of the codewords of each frame. 26. Each of the generated codewords is J
Occurring in successive frames of codewords, the particular ones of the codewords of the above script differ in exactly two X signal point positions, the two of which are the kth and qth signal point positions, and the method Further comprising the step of generating a rearranged sequence of signal points of the codeword of each frame such that the signal points at the th and qth signal point positions are separated by more than J signal points in the sequence. The method according to claim 19.