KR100282347B1 - Decoding Apparatus and Method Using Channel State Information - Google Patents

Abstract

The present invention relates to an apparatus and a method for decoding convolutionally coded received data, the method comprising: calculating a distance between input data and each branch, outputting a branch metric, and channel state information (CSI) in the branch metric obtained in the step; Multiplying the new branch metric to decode the data before the convolutional encoding using the new branch metric to decode the information distorted by the channel when decoding the convolutional coded data at the receiving end. By applying the channel state information (CSI) indicating to the decoding process, it is possible to improve the overall performance of the system.

Description

Decoding Apparatus and Method Using Channel State Information

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a receiver for decoding data received by convolutional encoding. In particular, the present invention relates to a decoding apparatus using CSI that applies channel state information (CSI) representing distortion information on a channel to a decoding process of received data. And to a method.

Digital TV transmission can be classified into a single carrier modulation method using a single carrier and a multicarrier modulation method for transmitting desired data using a plurality of multiple carriers. . That is, the transmission method of the digital TV is classified into a residual side band (VSB) method using one single carrier and a coded orthogonal frequency division (COFDM) method using a plurality of carriers.

The COFDM scheme using a plurality of multi-carriers, among others, has a feature of easily recovering signal damage caused by a multi-path channel, and a single frequency network (SFN) is possible, unlike a conventional single carrier.

In addition, the COFDM maps data in a manner called quadrature amplitude modulation (QAM) and transmits the modulation schemes that are mainly used for quadrature phase shift keying (QPSK), 16-QAM, and 64-QAM.

That is, in order to transmit desired data by the COFDM method, the data is first mapped by one of the three modulation methods described above, subjected to an Inverse Fast Fourier Transform (IFFT), and then inserted by a guard interval. do.

In this case, one of the methods mainly used for transmitting digital data to be transmitted without error is channel coding. In such channel coding, the code is divided into two types, a block code and a convolutional code, of which the convolutional code has a structure as shown in FIG.

In other words, a typical (n, k) convolutional encoder is a linear sequential circuit with m memories, k inputs, and n outputs, where n = 2, k = 1 1) That is, rate 1/2 convolutional code is mainly used.

Fig. 1 described above shows a convolutional encoder when m = 3.

Where m represents the constraint length, which is the length of the output sequence in which one input bit is affected, and the output of this code is the current output U1, U2 and the current input m and the past m Determined by the input of two dogs. At this time, the polynomial g1 (x) = 1 + x of the top side (U1), a polynomial of a lower side (U2) is a g2 (x) = 1 + x 2.

2 shows an encoder trellis diagram of data generated by the convolutional coding method described above.

Here, since the constraint field is 3, there are 4 (2 ^(K-1) ) states, that is, states of 00, 01, 10, and 11, and the output values for each input and the values of the next state are shown.

For example, when 0 is input in the initial state t1 of 00, the output is 00, and when the input is 1, the output is 11 and the output is 11 and the state transitions to 10. If 0 is input in the state of 10 again, the output becomes 10 and transitions to the state of 01. If 1 is input, it outputs 01 and transitions to the state of 11. Thus, each value is output according to each input value, and the transition continues to another state or the same state. At this time, a value inputted and outputted is called a codeword branch.

Based on the encoder trellis diagram created at the transmitter, the receiver trellis diagram can be made as shown in FIG. 3 in the same manner according to the received value.

FIG. 3 shows a trellis diagram of a receiver. Assuming that data is input in the order of 11011 .. to a convolutional encoder of a transmitter, data encoded at a rate of 1/2 is 11,01,01,00,01 Are transmitted in order of. The transmitted data is inputted to the decoder through a channel. At this time, an error may occur due to the transmitted channel. 3 shows an example in which the encoded value 00 of the fourth bit is changed to 10 due to an error on the channel.

At this time, since the data input in the initial state 00 is 11, the difference between the distance between the data of 11 and 00 or 11 in the trellis diagram of the transmitter is obtained. In other words, the difference between the distance between the received data 11 and 00 on the transmitting side becomes 2, and the difference between the distance between the transmitting side 11 and 00 becomes 00. In the next step, 01 is received, where the difference in distance from the transmitter's trellis diagram is 1 for states 00 to 00 (that is, the difference between 01 and 00) and 1 for transitions from state 00 to 10. (Ie, the difference between 01 and 11), and 2 (that is, the difference between 01 and 10) when transitioning from state 10 to 01, and 0 (that is, the difference between 01 and 01) when transitioning from state 10 to 11, respectively. Is calculated. This distance difference is called a branch metric. Therefore, the branch metric of the data input from the receiving end can be calculated as shown in FIG.

Then, the new branch metric and the accumulated old state value are added for each transition to obtain the current state value, and the state values of the two passes are compared for each state, the smaller value is taken out, and the larger value is excluded. If you take a path from each release diagram for each state, only one pass remains after a certain time. Therefore, the value input from the transmitter may be restored according to the transition state of the selected path.

Decoding in this manner is called Viterbi decoding or maximum likelihood decoding and is widely used to decode convolutional encoded data.

3 illustrates a case of a hard decision in which data transmitted from a transmitter is represented by 1 and 0 at a receiver, and FIG. 4 is represented by 1 as 0 to 7 and 0 as -8 to -1. The case of soft decision is shown.

That is, FIG. 4 shows a trellis diagram in the case where data transmitted from a transmitter is received distorted by various environments on a channel. For example, if the code word transmitted in the initial state is 11, the received data should be 77. Since the error value is -2, -3, the received data becomes 5,4. The same principle applies in other states.

In this case, the method of calculating the branch metric of FIG. 4 is also the same as that of the hard decision shown in FIG. 3. However, the branch metric is obtained in more detail by considering various values from -8 to 7, not two cases of 0 and 1. That is, the distance difference between the data received at the receiving end and the original data transmitted at the transmitting end is obtained, respectively, and soft decision is performed with the desired number of bits according to the difference of the distance. It is known that such a decoding method using a soft decision can generally obtain a coding gain of about 3 dB than a decoding method using a hard decision.

In addition, the decoding method by the soft decision is also performed in the same manner as in the case of FIG. 3.

However, FIG. 4 described above is also a method of soft decision of data input through a channel and then decoding with this data. In other words, it is not a method of decoding considering the distortion of each input data by the channel, but a method of decoding using only the soft decision data itself as an input of decoding. In other words, the decoding method based on the soft decision has been widely used in a transmission method using a single carrier, and does not reflect how much of each channel of the received constellation is distorted by the channel. Decode without.

At this time, in a transmission scheme such as a COFDM using multiple carriers, since each QAM symbol transmitted corresponds to one transmitted carrier, decoding does not reflect how much of each symbol is distorted by the channel. Will degrade the performance. In particular, even when interference occurs on the channel and the interference level of each carrier is different, the conventional method does not decode by tracking the change of each carrier, so the performance of the system is further deteriorated in this case.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to apply information (CSI) indicating how much the data received by convolutional coding is distorted by a channel to the decoding process of the received data. The present invention provides a decoding apparatus and method using CSI to improve the performance.

1 is a block diagram of a general convolutional encoder

2 is a trellis diagram of a typical encoder

3 is a hard decision trellis diagram of a typical Viterbi decoder.

4 is a soft decision trellis diagram of a typical Viterbi decoder.

5 is a block diagram illustrating a decoding apparatus using CSI according to the present invention.

6 is a trellis diagram applying CSI to a decoder according to the present invention.

7 is a BER graph showing a performance improvement state according to the present invention

Explanation of symbols for main parts of the drawings

51: buffer 52: branch metric calculation unit

53: Add-Compare-Select (ACS) section 54: Pass Memory

55: trace back part

The decoding apparatus using the CSI according to the present invention for achieving the above object, the branch metric calculation unit for calculating the branch metric between the input data and each branch, and the CSI to the branch metric obtained by the branch metric calculation unit A new branch metric calculation unit that multiplies and calculates a new branch metric, and adds the previous state value accumulated in the new branch metric for each transition, and then compares the pass for each state and selects and outputs the minimum state value among them. After detecting an optimal state signal having a minimum state value among an ACS unit and an overall state value output from the ACS unit, the optimal path is traced using the optimal state signal to survive survival value of the first section of the observation section. Characterized in that it comprises a decoding unit for outputting the.

The decoding method using the CSI according to the present invention comprises the steps of outputting a branch metric by calculating the Euclidean distance between the input data and each branch, and multiplying the branch metric obtained in the step by the CSI to obtain a new branch. Calculating a metric and decoding the data before the convolutional coding using the new branch metric.

Other objects, features and advantages of the present invention will become apparent from the following detailed description of embodiments taken in conjunction with the accompanying drawings.

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

The present invention is to improve the performance of the decoder by applying the CSI to the decoding to know the degree of distortion of the channel for each input data.

That is, the data input to the decoder is input as two data, i.e., I, Q pairs, respectively. Information indicating how much data input to such a decoder is distorted with respect to a channel to be transmitted is referred to as CSI.

In this case, when the transmitting end inserts information such as a pilot and the like, the receiving end can easily use the pilot to obtain CSI for each input data, one of which has been filed in Korea by the present applicant. (Application No. 98-16633). In addition, even when a pilot is not inserted, it is possible to obtain CSI by a method such as a decision directed method. The CSI thus obtained can be applied at a demapping stage when the transmission scheme is QAM.

FIG. 5 is a block diagram of the present invention in which the above-described CSI is applied to a decoding apparatus, and includes a buffer 51 for dividing soft-decision data x into two data Z1 and Z2 (or I and Q). The branch metric between the data Z1 and Z2 output from the buffer 51 and each branch, that is, the branch metric calculation unit 52 for calculating the Euclidean distance, and the output of the branch metric calculation unit 52 are multiplied by CSI, and An Add-Compare-Select (ACS) unit 53 for adding the previous state value accumulated in each transition to each transition and comparing the paths for each state and selecting and outputting the minimum state value thereof. A pass memory 54 which outputs a path having a minimum state value, that is, an optimum state signal among all state values output from the 53, and feeds back the minimum state value to the ACS unit 53, and the optimum state signal. The first pass of the observation interval And a trace back unit 55 that calculates a survival pass value of the liver and decodes the bits before being encoded by the convolutional encoder.

In this case, in order to estimate the most optimal survival path, the waveform of infinite interval should be observed, but to simplify the hardware structure, only a certain interval is observed, and the optimum path is selected from the observation interval and the input value is output.

In the present invention configured as described above, the data finally transmitted by demapping or other method is determined according to the transmission method, and the determined data is often performed by soft decision of 3 or 4 bits. The soft-decision data x is divided into two data Z1 and Z2 (or I and Q) in the buffer 51 and then input to the branch metric calculation unit 52.

The branch metric calculation unit 52 calculates a branch metric BM with four values, such as a pair of encoded bits 00, 01, 10 and 11, using the soft decision data Z1 and Z2. In other words, calculate the distance between the input and each branch.

The branch metric values BM calculated by the branch metric calculation unit 52 are input to the ACS unit 53 and multiplied by the previously obtained CSI. That is, the branch metric is multiplied by information of how much the input data is distorted by the channel.

Here, when the CSI multiplied by the branch metric has a small value, the input data is distorted much by the channel, and the new branch metric is much smaller than the original value. In addition, when the CSI is large, that is, when the input data is not distorted by the channel, there is almost no change in the pre-calculated branch metric.

FIG. 6 shows a trellis diagram at the receiving end according to this method, and it can be seen that a new branch metric is obtained by multiplying the branch metric obtained as shown in FIG. 4 by CSI. That is, the branch metric of the first case in FIG. 3 is calculated as 25. At this time, since the input CSI is 0.81, the branch metric 20 is obtained by multiplying 25 and 0.81.

Similarly, when the branch metric is 5, we can also multiply CSI to get a new branch metric of 4. Hereinafter, the branch metric shown in FIG. 6 is a new branch metric reflecting the CSI obtained by the above method.

In the step of multiplying the CSI by the branch metric, assuming that the CSI value generated when there is no noise or other influence on the transmission channel is y, when the CSI is greater than y, the CSI may be saturated and multiplied by the branch metric. It is also possible to multiply the obtained CSI by the branch metric without saturation.

By multiplying the calculated branch metric by CSI, it is reflected in the process of decoding how much each input data is distorted by the channel, thereby improving overall performance as compared to the conventional method.

That is, in general, data transmitted from a transmitter in a digital communication is distorted by various situations of a channel. At this time, one part of the transmitted data may be severely distorted much more than the other part. In this case, demodulating the data received at the receiving end may cause a so-called catastrophic error that causes a large number of errors in the demodulated data. Therefore, in this case, if the receiver uses CSI for data that is much distorted in comparison with other parts in the decoding stage, it is possible to reduce the phenomenon such as the catastrophic error described above and to improve the performance of the overall system.

In each state of each section, since two passes meet each other, the ACS unit 53 adds the previous state value accumulated in each new branch metric multiplied by the CSI for each transition to obtain the current state value, and then for each state. The state values of the two passes are compared, and the smallest state value among them is output to the pass memory 54. That is, the survival path is obtained for each state of each section in the observation section, the optimum path of the observation section is selected, and the input value at that time is output.

The pass memory 54 compares each state value of one section and outputs the state of the path having the smallest value, that is, the optimum state signal, to the trace back unit 55, and for each state as the optimum state value at this time. After normalizing the selected minimum state value of the feedback to the ACS unit 53 is accumulated to the previous state value. The trace back unit 55 outputs the decoded data, which is the data of the survival pass of the first section of the observation section, that is, the data before being coded by the convolutional encoder, by tracking the optimum path for each section by using the optimum state signal.

Decoding in this way can improve the performance compared to the conventional method of decoding convolutional coded data at the receiver, and this performance improvement is more than that of additive white gaussian noise (AWGN). This is more apparent in the presence of path fading.

That is, when digital data is transmitted over the terrestrial wave, the transmission channel becomes a fading channel by multiple paths, and interference by adjacent channels or the same channel also exists, thereby degrading the overall receiver performance. Especially in the case of notch of channel characteristics due to interference caused by external factors, severely distorted carriers are received, and the distortion degree must be sufficiently reflected to improve overall system performance.

7 shows that the performance is improved as a result of the experiment according to the present invention. In this case, the transmitted channel is a Rayleigh fading channel having 20 multipaths, and a data transmission method is compared with that of 16-QAM. In FIG. 7, the upper BER curve represents the BER curve according to the conventional method, and the lower BER curve represents the BER curve according to the present invention. As can be seen in FIG. 7, it can be seen that the decoding performance using CSI is superior to the conventional decoding result of simply input data.

Meanwhile, the present invention is applicable to a transmission scheme such as QAM and PSK or a scheme such as OFDM, which can apply CSI to Viterbi decoding.

As described above, according to the decoding apparatus and method using the channel state information according to the present invention, the decoding process of the channel state information (CSI) indicating the information distorted by the channel when decoding the convolutional coded data at the receiving end By applying to, the overall performance of the system can be improved. In particular, even when interference is caused to the channel and the interference level of each carrier is different, since the change of each carrier is tracked and decoded, the performance of the entire system can be further improved. In addition, the present invention is more effective when used in a TV transmission method such as a digital TV that uses COFDM as the transmission method.

Claims (8)

A decoding apparatus for receiving and decoding convolutional coded data, A branch metric calculation unit for calculating a branch metric, which is a distance value between the input data and each branch; A new branch metric calculation unit configured to calculate a new branch metric by multiplying the branch metric obtained by the branch metric calculation unit with channel state information (CSI); An add-compare-selector for adding the previous state value accumulated in the new branch metric for each transition, comparing the paths for each state, and selecting and outputting a minimum state value among them; After detecting the optimal state signal having the minimum state value among the total state values output from the add-comparison-selection unit, the optimal path is traced using the optimum state signal to calculate the survival pass value of the first section of the observation interval. A decoding apparatus using channel state information, comprising a decoder. The decoding apparatus of claim 1, wherein the data input to the branch metric calculator is data that is soft-decited with a predetermined bit. The method of claim 1, wherein the new branch metric calculation unit And decoding the channel state information (CSI) for each input data using the transmitted pilot information. The method of claim 1, wherein the channel state information (CSI) is Decoding apparatus using channel state information, characterized in that the information indicating how much the input data is distorted by the channel. The method of claim 3, wherein the new branch metric calculation unit When the obtained channel state information (CSI) value is greater than the channel state information (CSI) value generated when there is no noise of the transmission channel, the obtained channel state information value (CSI) value generated when there is no noise And saturates the multiplier to multiply with the branch metric. The method of claim 3, wherein the new branch metric calculation unit When the obtained channel state information (CSI) value is greater than the channel state information (CSI) value generated when there is no noise of a transmission channel, the channel state information (CSI) value is multiplied by a branch metric, characterized in that the channel state Decoding device using information. The method of claim 1, wherein the decoding unit Compare each state value of one section, normalize the selected minimum state value of each state to the state value of the path having the smallest value, and then feed back to the add-compare-selector to accumulate the previous state value. Decoding apparatus using the channel state information. A decoding method for receiving and decoding convolutional coded data, Calculating a branch metric between the incoming data and each branch; Calculating a new branch metric by multiplying the branch metric obtained in the step by the channel state information (CSI) indicating the degree of distortion caused by the channel; And decoding the data before the convolutional coding by using the new branch metric.