SU1711337A1 - Coder of block-signal-code design - Google Patents

Description

 The invention relates to communication technology and computer technology and can be used in discrete message transmission equipment.

The purpose of the invention is to improve the noise immunity of the codec

FIG. 1 shows a block diagram of a codec in FIG. 2 - matrix of the generated signals, a general view; in FIG. 3 — 16-point alphabetical splitting for amplitude-phase modulation; on. FIG. 4 shows the matrices in FIG. 2 for a specific example; Figures 5-8 are lattice diagrams explaining the decoding process.

A codec consists of transmitting and receiving sides 1 and 2, and communication channel 3. Transmit side 1 includes encoders

4 external codes, a block 5 for selecting a signal subset, a signal encoder 6, and a block 7 for RAM. Receiving side 2 contains a RAM 8, a metric calculator 9, the first and second switches 10 and 11, the permanent memory block 12, the first and second buffer memory blocks 13 and 14, adders 15, a comparison block 16 and a driver .17 read addresses. In Fig. 1, the first and second groups 18 and 19 of the information inputs are indicated, the first - the third clock inputs 20-22, the first third inputs 23-25 synchronization and the output 26 of the codec.

Encoders 4 and 6 and block 5 of the transmitting side can be performed on the ROM. HIGH C4 1

numerator 9, metrics can also be implemented on ROM. readout address generator 17 is, for example, a shift register.

The codec block signal code design (CCM) works as follows.

The inputs 18 encoders O - k.TH receives N blocks of information on L bits. After coding with the i-th code (, N), the output of each encoder .1 appears in a block of information a long n bits (AT || blocks at the outputs of the encoders k forms a matrix

aT

1711337. barking in channel 3 communication. Under the influence of interference in the communication channel 3, the signals are distorted and, in a distorted form, the digital translations are recorded into the block 8 of the receiving side 2.

Block 8 converts a sequence of n signal points sent to its input at a frequency nF into a sequence of n signal points read at frequency F. From the outputs of block 8 and the inputs of the evaluator 9 metrics, the sequence of

J5 of the received signal points comes in the form of 2n real numbers, where each pair of real numbers corresponds to the coordinates of each of the n signal points on the complex plane 7.1

20

BUT

AG

aijl

, N, n

thirty

S

The columns of this matrix are successively served in block 5 for selecting a signal subset, in which one of 2 signal subsets is selected, into which the complete signal alphabet is divided. The full signal constellation is divided into 35 signal subsets in such a way as to maximize the minimum Euclidean distance between adjacent signal points within the subset, ..

At the inputs 19 are received P blocks of information of no n bits. These blocks of uncoded information form a matrix.

AT

lib nil

t, p,

The columns of this matrix are successively fed to the signal coder 6, where one of the 2 ° signal points of the selected subset is selected. Thus, the full signal 50 alphabet contains signal, then-1 check, which are divided into 2N signal subsets of up to 2P points in each subset. At the output of the signal encoder 6, a sequence of 55 signal points is formed, which is recorded in block 7 of the transmitting side 1 and with a frequency nF of the post

The n received signal points come in the form of 2n real numbers, where each pair of real numbers corresponds to the coordinates of each of the n signal points on the complex plane 7.1.

The calculator of 9 metrics includes a coordinate sensor of 2 signal points belonging to the full signal alphabet. The calculator 9 metrics calculates the square of the distance from each of the signal points to the received one with coordinates 9t | and X. signal point, repeat this procedure for each of the n received signals.

The square is the following:

calculated by

, n,, 2

N + P

where wi and jy are a pair of real numbers corresponding to the coordinates of the kth signal point on the complex plane, k 1,2 1,2.

For each of the n received signals, the output of the calculator of 9 metrics contains a set of 2 squares of distances Ag, called the metrics of the ES signal. This set of metrics goes to a group of blocks 10 - 16 that implement the decoding algorithm of the received sequence.

Viterbi.

i

The switch 10 controls the block 12 by means of the information stored therein. The control process of the switch 10 is as follows.

In block 12 a lattice Viterbi decoder diagram is recorded. Here, each memory element contains information about which of the k edges (k

where r

C

 1,2) comes from which node of the diagram and which node it belongs to. The total number of memory elements is equal to: p 5

,

ie Js

the number of edges included in the jth node of the diagram is in the ith section, each section having the number of nodes equal to 5. The memory elements of block 12 are divided into groups, each of which controls the process of calculating the metrics of the paths included in one of the SJ nodes in the section, p. Block 12 provides a control signal with a frequency of nSF and switch 10 switches 2N + f inputs (each of which receives a metric D,,) to SM outputs, each of which is connected to the inputs of adders 15.1 - 15.S. S maxS | - maximum number of nodes

in one of, n cross-sections of the trellis diagram.

At the jth step, the Viterbi decoder summarizes 15 metrics of surviving up to the 1st cross section of paths and metrics of received signals in adders in accordance with the trellis diagram recorded in the memory of block 12. Then, the comparison block 16 selects the smallest and via a control input, instructs the switch 11 to switch the smallest of the calculated path metrics to the input of the first memory block 13, where it is stored for summation over the (1 + 1) -th cross-section of the trellis diagram. At the same time, the comparator unit 16 generates a signal to the second memory unit, the metric of which (combined with the metrics of the surviving paths recorded in the first memory unit 13) is the smallest at the j-th step of the 1st section of the trellis diagram. Block Ik remembers the node number, from which the path with the smallest metric comes (for a given j

All the above described turnout procedure. S

torus

&

time and last

section in the last step in the second memory block 1, a sequence of decoded signals is formed, the reading of the decoded

five.

ten

15

20

25

0

thirty

35

0

five

The sequence is occupied by the t-read address shaper 16, which begins reading from the last decoded signal. The remaining signals are read into the output of the Viterbi decoder, using information about the node numbers from which the edges come out, corresponding to the decoded signals on the j section of the trellis diagram. The decoding is completed when the driver 17 reads the address of the first signal from the block containing n signals.

The codec of the block CCM works as follows.

Let be , . Then there are two 1.8 inputs, information from which is encoded, and two uncoded inputs 19. Let the codes Reader-Muller (PM codes) C, 3,1) and (1}, 1, be used as codes in encoders k and 6 ) respectively. The code (k, 3,1) is a parity check code, and the d code, 1,0 is a repetition code. Then, 12 information bits arrive at the CCM input every F (Fig. 2a). After encoding the PM codes, the matrix of the transmitted signal B is formed, shown in Fig. 28. In total, the signal alphabet should contain 2 24 16 signal points, which should be divided into 2 k subsets of 2 points each. Fig. 3 shows such a division of the 1b-point of the ч chny alphabet of amplitude-phase modulation (AFM). If the minimum distance D between adjacent signal points was U 2, then after splitting the minimum distance between adjacent signal points within the subset increased 2 times: dw; n 2i 4. Each column of matrix B determines the choice of signal point sent to the channel, moreover, the pair of coded bits determines the choice of the signal subset, and the pair of uncoded bits the signal point in this subset.

Let the following 12 information bits arrive at the CCK codec inputs: 0,1,1,1,0,0 0,1,0,0,0,1 (fig.Aa). The matrix B obtained after encoding with the PM codes is shown in FIG. 45. In the first column of the matrix B, the coded bits 01 determine the choice of a subset of signals, equivalent to a triangle in FIG. 3 (the lower bit is considered the highest). The remaining pair of bits 11 determines the choice of a specific signal point within the subset that has the prefix 01, namely, points with coordinates (The remaining signal points are similarly selected. In channel 3, the transmitted signal is inevitably subjected to distortions, and after the inverse digitization the distorted signal points are recorded in the RAM block 8. In the example considered, the signal-to-noise ratio was 16 dB. The correspondence between the columns of matrix B, the transmitted signals and the distorted signals , obtained with the help of the implementation of noise, is given in table 1.

Decoding of the received codeword in the codec is performed using the Viterbi algorithm, which was proposed for decoding convolutional codes. The use of this algorithm in this codec is based on the fact that the block code can be represented in the form of a trellis diagram. Such a diagram is a right directed graph starting at a single point and converging to a single point. Each edge of the graph is associated with a label corresponding to an element of the code word. Any path from the starting point to the final point passes through n edges (n is the length of the code) and the corresponding label set defines one codeword. The decoding task in this case is reduced to the problem of finding a path along a trellis diagram with the help of some decoding rules

Consider the construction of lattice: diagrams in more detail. An uncoded sequence of n bits can be viewed as a (n, h, 1) code. Lattice diagram With ,, 1) code is presented in fig.5d. In this diagram, 16 different paths can be obtained, leading from the initial node to the final node, corresponding to different sets of four bits.

The code PM (and, 3,2) corresponds to the re-, schematic diagram presented in figure 5, and the code PM C,) - sieves

50

5 o d d

0 5

five

The chat diagram in Figure 5Ј. When combining codewords of different codes that correspond to strings of bits, by columns to obtain the code word of the CCM, the lattice diagram of this codeword is the product of the trellis diagrams of the original codewords. The construction of the trellis diagram for considered in the example of the CCM shown in Fig.6. Combining non-coded other words doubles each edge of the graph. If the graph of code С, 1) contains two ;, parallel paths between neighboring nodes, the combination of two such codes gives a trellis diagram with four parallel paths between neighboring nodes. Combining the lattice diagram of two coded words with the above-mentioned lattice, having four parallel paths between adjacent nodes, leads to the fact that the resulting lattice diagram preserves the topology of the coded lattice, but each edge of it is split into four parallel paths. In a more convenient for analysis version, the resulting grid diagram is shown in Fig.7; Here, each thickened edge corresponds to four parallel edges. The first two bits located at each edge are coded bits that determine the choice of signal subset. The last two bits, labeled XX, indicate that in their place can be any of the four possible combinations of two bits, defining a particular point in the signal subset, which corresponds to the vector of one of the four edges.

Decoding of the received signal occurs as follows.

The calculator 9 metrics for each signal point obtained from channel 3 calculates the square of the Euclidean distance. The set of metrics calculated for the four signal points of this example is given in Table 2. The choice of the path of the trellis diagram (Fig.8) with the minimum metric is as follows. For each node of the graph, a comparison is made of all metrics of the paths entering it. A path with a minimum metric, called a survivor, is chosen, the remaining paths are discarded.

In the first step of decoding, four parallel paths leading from the starting point lead to each point of the graph. Therefore, the metric of this edge is equal to the metric of the path. On figs | surviving paths after the first decoding step are presented. Merika at each edge corresponds to the binary representation of the signal point, the metric of the path is shown in brackets. For the point of the first Russ, the edge metrics in the subset of the signal are compared with the prefix 00 (the numbers of the signals L – k in Table 2).

The smallest metric, equal to 3.237, has an edge with metric 0010, which is left as a surviving path. Similarly, the choice is made at other points in this section of the graph.

In the second decoding step (Fig. 8), eight paths enter the point of the first Russ (with prefixes 00 and 01). The minimum metric has a path consisting of edges labeled with 0010 (metric 3.237) and 0000 - (met: re-vka 0.115). The total metric of this path is 3.237 + 0.115.3.352.

The surviving paths in the third decoding step are shown in Fig. 8.

After the fourth step of decoding, the only surviving path remains (Fig. 82) with a metric of 0.97. The labels of the corresponding edges, written in the form of columns, give the decoded matrix B, which in this case is equal to the transmitted matrix B, i.e. no errors. Information bits can be read from matrix B when moving from the bottom.

up left to right, excluding redundant bits.

This codec implements a maximum likelihood decoding algorithm — this corresponds to a soft decoding of the internal codes of the generalized concatenated code. When soft decoding external codes, the probability P (X /) is used a posteriori.

The well-known codec uses the hard value of the received signal X, which corresponds to hard decoding. In this case, the whole internal signals are received and error correction by external codes. In this case, the accepted word Y (y, y., ..., yn)

five.

15

25

Jq

20

0

five

0

five

0

five

Ј Y will be decoded correctly if the square of the Euclidean distance between the transmitted and received words satisfies the inequality

P (y, y) -Ј min d; P; / 8n,, m

where y € Y - transmitted through the channel

additive Nassian noise word; 57 received that word distorted

noise, th - order of the generalized cascade code.

The noise immunity of hard decoding of external codes may not be sufficient, therefore, a codec with a soft decoding algorithm is proposed.

In this case, decoding will be correct if the square of the Euclidean distance between the transmitted and received word satisfies the inequality

P (y, y) "i min d; P; / 4nl. m

Thus, the use of a soft decoding algorithm compared to hard decoding of external codes results in an energy gain of 3 dB, which indicates an increase in noise immunity.

Claims (1)

Formula invented The codec of the block signal-code design, consisting of the transmitting and receiving sides and the communication channel, the transmitting side contains the first — N-th coders of the outer code (2 — the number of subsets into which the signal code is broken), the information inputs of the first — N-ro the outer code encoders are the corresponding inputs of the first codec group, the clock inputs of all the outer code encoders are combined with the clock input of the RAM memory block and are the first clock input of the codec, the synchronization input of the RAM memory block is the first codec synchronization, the receiving side contains a memory block, a clock input and a synchronization input which are the second codec inputs of the same name, the comparison unit, the first and second blocks of the buffer memory, the output of the operating side block of the transmitting side are connected through the communication channel with the information input of the random access memory block of the receiving side, the clock input of the communication channel the third clock input of the codec, which differs from the fact that, in order to increase the noise immunity of the codec, a signal subset selection unit and a signal coder are entered on the transmitting side, the first of which is the Pth information inputs of which (2 ° is the number of The rotary points in each subset are the corresponding inputs of the second codec group, the clock input of the signal encoder is connected to the first clock input of the codec, the outputs of the first are N-ro external code encoders connected to the corresponding inputs of the signal subset selection unit, the outputs of which are connected with the control inputs of the signal coder, the outputs of which are connected to the information inputs of the operating memory unit, on the receiving side there are entered the calculator of metrics, the switches, the fixed memory unit, adders and A read address starter whose clock input and the clock input of the metrics calculator are combined and connected to the second codec synchronization input, the outputs of the RAM block are connected to the information inputs of the metrics calculator, the output of which is connected to the information inputs of the first switch, the clock input of o6vwii.n with clock numbers the inputs of the second switch, all adders, buffer memory blocks and the fixed memory block and is the third codec sync input, the outputs of the fixed memory block are connected with the control inputs of the first switch, the first —S.M outputs of which (S is the maximum number of nodes of the trellis diagram) are connected to the first inputs of the first, respectively, S-ro adders, the outputs of which are connected to the corresponding information inputs of the second switch and the comparison unit, clock input which is connected to the second clock input of the codec, the first and second outputs of the comparison unit are connected respectively to the information inputs of the second block of the buffer memory and the control inputs of the second switch, output Which are connected to the information inputs of the first block of buffer memory, the first - S-e outputs of which are connected to the second inputs of the first, respectively, S-ro adders, the outputs of the read address generator are connected to the address inputs of the second block of buffer memory, the output of which is the codec output. Table 1 (3 ,, -0.) (-0.373; -0.258) (-0.390; 2.723) (-1.112; -3.29) R w table 2 I fiSF Gu 5 F nSF f F f T T nSF fiSF. About -QQXXi + - 01 u Figz A - 0 / L7; - ///// but 6 FIG. five Schig.4 (W t) (W) r "v ch h, f WigaaaaU & i 1st .j . % cJ Tier 1711337. 01 2 10010 OQOO (3,3S2) i