JP4185557B1 - Data symbol demapping method for decoding in a wired communication system - Google Patents

Abstract

ここでl_s(b_i,k)は、k番目のシンボルのi番目のビットの尤度を、y_k はI-Q平面でのk番目の受信され等価されたシンボルのポイントを、αはI-Q平面での理想シンボルポイントを、S_i ⁽¹⁾ と S_i ⁽⁰⁾は各自、i番目のビットが“1”、“０”である時のシンボルセットを示している。
このようにして取得した尤度(ls)を軟判定値として使用して、後の復号化のために、受信データシンボルのそれぞれについてデータビットを決定する。
【選択図】図１A soft decision demapper used to decode data symbols received over a wired channel. The soft decision demapper receives a series of data symbols of information bits that are mapped in the IQ plane, transmitted over a wired channel, and equalized. The demapper calculates the likelihood of each data symbol received using the following equation:

Where l _s (b _{i, k} ) is the likelihood of the i th bit of the k th symbol, y _k is the point of the k th received symbol in the IQ plane, α is the IQ plane S _i ⁽¹⁾ and S _i ⁽⁰⁾ indicate symbol sets when the i-th bit is “1” and “0”, respectively.
The likelihood (ls) obtained in this way is used as a soft decision value to determine a data bit for each received data symbol for later decoding.
[Selection] Figure 1

Description

The present invention relates to a data symbol demapping method for decoding in a wired communication system.

Tosato, F, Bisagila, P "Simplified soft-output demapper for binary interleaved COFDM for binary interleaved COFDM on application to HIPERLN / 2 withapplication to HIPERLN / 2) ”, Communications 2002, ICC 2002, iTriple International Conference on Volume 2 (Communications, 2002. ICC 2002. IEEE International Conference on Volume 2), 2002, Vol. 2, p. 664 The -668 publication discloses a decoding system for decoding data symbols encoded with an error correction code such as a turbo product code (TPC). This decoding system comprises a demapping technique that requires soft decisions to demap data corresponding to data symbols for transmission data bits prior to decoding.

This soft decision demapping is generally known to have good performance and uses a soft symbol demodulation demapper, i.e., a soft decision demapper, to determine the data bits that match the received data symbol. Create a soft decision value for. The soft decision value gives a determination of how many “1” or “0” bits are transmitted. The soft decision value is based on the log likelihood rate (LLR) of the data bits calculated from the received symbols.
For example, channel state information (CSI) is usually required to calculate the LLR for reasons of channel impairment, such as fading or delay spread. The above soft demapper requires channel state information (CSI) in order to construct a soft decision equation that is known to be theoretically optimal, even if channel equalization is used.
Usually the kth received symbol is

として表される。ここでh_kは、関連するチャネル応答であり、x_kは送信されたデータシンボルであり、そしてn_kは分散σ²を有した対応する付加的白色ガウス雑音(AWGN)である。
われわれは受信機において、ゼロフォーシング(ZF)等化を行うものとする。この場合、ｋ番目の等化された受信シンボルは、以下のように記され、

Represented as: Where h _k is the associated channel response, x _k is the transmitted data symbol, and n _k is the corresponding additive white Gaussian noise (AWGN) with variance σ ² .
We shall perform zero forcing (ZF) equalization at the receiver. In this case, the kth equalized received symbol is written as follows:

ここでの、ノイズ項w_kの分散は

である。

Where the variance of the noise term w _k is

It is.

When decoding symbols into data bits with error correction, soft decision decoding requires soft information. Soft information is obtained from LLR. The LLR for b _{i, k} which is the i th bit in the k th symbol is

として表され、S₁ ⁽¹⁾ と S₁ ⁽⁰⁾ は、i番目のビットがそれぞれ“1”と“0”のシンボルセットである。

S ₁ ⁽¹⁾ and S ₁ ⁽⁰⁾ are symbol sets whose i-th bit is “1” and “0”, respectively.

Equation (3) is a measure of how many bits b _{i, k} are “1” or “0”. From equation (3), it is clear that the more negative the equation λ (b _{i, k} ), the more likely the transmitted bit b _{i, k} is “1”. In contrast, the more positive λ (b _{i, k} ) is, the more likely the transmitted bit b _{i, k} is “0”.
Here, it can be noted that the noise variance σ ² is common to all received symbols, and can be excluded from the calculation formula without affecting the decoding performance. Therefore, the soft formula is

となる。
式（3）またはそれと等価な式（4）から生ずる軟判定は、チャンネル応答h_kが完全に判っている場合には、浮動小数点実装を行うことで、最高の性能をもたらすことが示されている。

It becomes.
The soft decision resulting from equation (3) or equivalent equation (4) has been shown to provide the best performance with floating point implementations when the channel response h _k is fully known. Yes.

Another important point to note from Equation (4) is that the term | h _k | ² is equivalent to the proportionality factor. If | h _k | ² changes significantly for different k, the soft decision range also increases.

However, CSI or h _k is often not readily available at the receiver that receives the transmitted data symbols. Therefore, it is not practical to rely on a soft decision demapper to perform decoding under such conditions. The CSI available at the receiver may not be accurate. Under these conditions, the given CSI is not accurate, so using a conventional soft demapper will result in performance degradation. In addition, if the system has residual intersymbol interference (ISI) due to the channel delay being longer than the length of the cyclic prefix of an orthogonal frequency division multiplexing (OFDM) system, the noise dispersion on each subcarrier Are different. This is especially true when transmitting encoded data symbols over wired communication channels that are susceptible to interference from delayed waves, but in contrast, in wireless communication channels, delayed waves are The interference by is relatively small.

Usually, in a wired communication system, a delayed wave having a level higher than that expected in a wireless communication system is observed. Therefore, a significant inter-symbol interference ( ISI). In other words, for wired communication systems, using the above equation (3) or (4), distributed delay waves used in y _k = x _k + w _k in the formula (2) (s ² / | h _k | ² ) When the conventional equation (3) or (4) is used, the noise variance s ² is obtained from the channel gain of h _k . However, in situations where there is severe ISI, the noise variance s ² will include ISI. When the noise variance based on this type of ISI is denoted as s _g ² , the noise variance including ISI is denoted as s ′ ² = s ² + s _g ² . Since s _g ² does not depend exactly on h _k , assuming that s' ² = s ² + s _g ² = s ² / | h _k | ² + s _g ² / | h _k | ² , the second term “ s _g ² / | h _k | ² ”is an inappropriate estimate. Therefore, the above equation (3) cannot accurately reflect this situation. In this sense, the assumption in equation (2) above is not valid, and the associated equation (3) or (4) will give inaccurate results.
The present invention has been made to solve the above-described problems and enable error-free data transmission in a wired communication system, and an improved demapping method for data symbols received through a wired communication path. I will provide a.

The method according to the invention receives a series of data symbols of information bits mapped in the IQ plane, transmitted over a wired channel and equalized, and using the following equation, It consists of the process of calculating the likelihood (ls) of.

Where l _s (b _{i, k} ) is the likelihood of the i th bit of the k th symbol, y _k is the k th point of the received symbol in the IQ plane, and α is the ideal in the IQ plane Symbol points are represented, and S _i ⁽¹⁾ and S _i ⁽⁰⁾ represent symbol sets in which the i-th bit is “1” and “0”, respectively.
In this method, the likelihood obtained in this way is used as a soft decision value to determine the data bits for each received data symbol and to give a series of demapped data bits. The resulting demapped data bits are then decoded to reproduce the information bits.

In particular, when performing data transmission using quadrature amplitude modulation (QAM), quadrature phase shift keying (QPSK), or two phase shift keying (BPSK), the channel response is widespread in communication via a wired channel. Because it depends on the frequency, we cannot expect uniform channel response or CSI.

In fact, data transmission depending on such a modulation scheme may be affected by intersymbol interference (ISI) based on a delayed wave that may occur in a signal transmitted through a wired communication path. From this viewpoint, in the method of the present invention, when the likelihood (ls) is calculated, the influence of CSI is ignored and good decoding performance is given. In particular, Equation (5) of the present invention assumes that the noise term (w _k ) in Equation (2) is a variance (s ² ) that does not include the channel response (h _k ) and does not depend on the frequency. It is established in the estimated sound. For this reason, the delay wave is eliminated and demapping is accurately performed with high reliability.

In the preferred embodiment, the mapped data symbols are modulated using Orthogonal Frequency Division Multiplexing (OFDM) using subcarriers with a frequency range of 2 MHz to 30 MHz.
In addition to the estimation that the variance of the noise term (w _k ) in equation (2) can be (s ² ), the effect of CSI is deleted, that is, the proportional coefficient | h _k | ² is removed from the conventional equation (4). By doing so, a limited range fixed-point implementation can be used instead of a floating-point implementation to calculate a highly accurate likelihood (ls). The bit width used in the fixed-point implementation should be between -0.1 and 0.1.

FIG. 1 is a block diagram of a system used to implement the method of the present invention. This system is designed for wired data communication having an error correction function, and includes a transmitter 100 and a receiver 200 coupled via a wired communication channel 10.
The transmitter 100 includes an encoder 110, eg, a turbo product code (TPC) encoder that encodes a series of input information data bits, a bit interleaver 120, a 256-value quadrature amplitude modulation (QAM) mapper 130, and an orthogonal frequency. Division multiplexing (OFDM) modulator 140, upsampler 150,
A finite impulse response (FIR) filter 160, a digital IQ modulator 170, and a digital-to-analog converter (DAC) 180 are provided. The interleaver 120 adds bits to change the order of the encoded data bits sent from the TPC 110. Thereafter, the interleaved bits are converted or mapped by the QAM mapper 130 into, for example, 256 ideal signal points arranged in a square lattice in the IQ plane. The OFDM modulator 140 then performs an inverse Fourier transform to produce a modulated signal of symbols carried on the subcarrier, which is upsampled by the upsampler 150 to match its frequency to the DAC converter 180 frequency. . After removing the image frequency with the FIR filter 160, the signal is fed into an IQ modulator 170 where it is converted to a frequency signal and then converted to an analog signal by the DAC 150 to produce an OFDM signal. This OFDM signal uses a carrier wave having a frequency band of 2 MHz to 30 MHz, and a subcarrier spreads within this frequency band. The OFDM signal obtained as a result is transmitted to the receiver 200 through the wired communication path 10.

Receiver 200 is an analog-to-digital converter (ADC) 280, digital IQ modulator 270, FIR filter 260, downsampler 250, and orthogonal frequency division multiplexing for converting received analog data into digital data. (OFDM) demodulator 240, soft 256-QAM demapper 230, bit deinterleaver 220, and turbo product code (TPC) decoder 210 are provided. IQ
The demodulator 270 performs frequency conversion to obtain a complex signal, which is separated through the FIR filter 270 and sent to the down sampler 250, where it is converted back to the original frequency signal. The OFDM demodulator 240 converts the downsampled signal into a corresponding symbol (y _k ) in the frequency domain, performs zero forcing (ZF) equalization, and is represented in the IQ plane defined by the demapper 230. Get the mapped signal. The Kth equalization symbol is described as y _k = x _k + w _k , where the variance of the noise term w _k is s ² .

Also, the OFDM demodulator 230 is configured to remove frequencies carrying null data and assign only valid symbols (y _k ) in the IQ plane. In this connection, the channel response (h _k ) is detected at the receiver from the header included in the transmission signal from the transmitter 100.

The demapper 230 includes 256QAM for performing quadrature amplitude modulation (QAM) to give 256 ideal signal point groups arranged in a square lattice in the IQ plane, and each ideal point is data of each candidate target symbol. Corresponds to the bit. The demapper is also configured to calculate the likelihood (ls) of each received data symbol (y _k ) using the following equation:

Where l _s (b _{i, k} ) represents the likelihood of the i th bit of the k th symbol, y _k represents the point of the symbol received and equalized k th in the IQ plane, α represents an ideal symbol point in the IQ plane, and S _i ⁽¹⁾ and S _i ⁽⁰⁾ represent symbol sets in which the i-th bit is “1” and “0”, respectively.

By using the calculated likelihood (l _s ) as a soft decision value, the demapper 230 determines a data bit for each received symbol and provides a series of demapped data bits. Specifically, the soft decision value (l _s ) for each received symbol is within the data bit set (signal point α) arranged in the IQ plane defined by QAM, as schematically shown in FIG. Specify the closest one. In FIG. 2, 16-bit QAM signal points are shown for simplicity only. The data bit designated in this way, for example, “1110” shown in FIG. 2 is determined as representing the received symbol (y _k ). In this way, the received symbols are demapped to a series of corresponding data bits.

The output of the demapper 230 is then provided to the TPC decoder 210 via the bit deinterleaver 230, where the data bits from the interleaved bits are decoded to extract the input information data bits. In this connection, the demapper 230 has a soft decision value (l _s ).
To determine the data bits for the received data symbols using a fixed point arithmetic in a range limited to -0.1 to 0.1. This range is set, for example, to have a 5-bit resolution for fixed-point calculations. Such fixed point calculations are sufficient to accurately determine the data bits based on the soft decision value (l _s ).
In this connection, the demapper 230 is configured to evaluate the likelihood, ie, the soft decision value (l _s ), with accuracy to a predetermined decimal place within the range of (−b) to (+ b). Here, b is the minimum value of the distance between the adjacent ideal points in the IQ plane. In this embodiment, a distance of 0.1 is used to satisfy −0.1 <l _s <+0.1, and this range is quantized to individual levels with a limited bit width. By using the quantized range in this way, the demapper can accurately determine the data bits based on the soft decision value (l _s ) using fixed point calculations.

In order to show that the method of the present invention is superior to the conventional soft decision demapper, a computer simulation using a system configuration as shown in the block diagram of FIG. 3 was performed. The information data bits are encoded with a TPC encoder.
The TPC encoded bits are then modulated with 256QAM and transmitted using OFDM over the ISI channel. The cyclic prefix period is 0.5usec. The channel response (h) used in the simulation has intersymbol interference (ISI) as shown in FIG. The kth received symbol after inverse OFDM and zero forcing (ZF) equalization is shown below.
y _k = x _k + w _k
Here, the variance of the noise term w _k is s ² .

Based on the received symbol (y _k ), the soft decision 256-QAM demapper performs a fixed-point and floating-point operation using Equation (5) of the present invention while comparing with Equation (4) of the conventional method. Mapping was performed. Data bits obtained after deinterleaving and decoding were compared with the original information data bits to calculate the bit error rate (BER).

Fixed point results Table 1 shows the bit error rate (BER) in a fixed point implementation in a system using equation (5) of the present invention and conventional equations (3) and (4). The TPC decoder uses 32 test patterns and 6 iterations. The 5-bit quantizer is equivalent to the 32 quantization level and is included in the demapper. The input to the quantizer, that is, the soft decision value (l _s ) obtained from Equation (5), and Equations (3) and (4
LLR (l) obtained from) is limited to the ranges -0.1 <ls <0.1 and -0.1 <l <0.1.
This result shows that the decoding system using the formula (5) of the present invention is superior in performance to that using the conventional formula (4). This result completely contradicts the conventional method (4), which was thought to give excellent performance in wired communication systems. Note that the fixed range implementation has limited dynamic range and bit width. If some channel gains | h _k | ² are relatively small, the true value in the conventional method (4) is not reflected since the accuracy is limited in the fixed-point implementation. Therefore, the performance of the system using Equation (4) of the conventional method is worse than that using Equation (5) of the present invention.

Floating point results Figure 5 shows the BER results when the system is implemented in floating point using both equations (5) and (3). The TPC decoder used 64 test patterns and 6 iterations.
In the previous section, performance comparisons based on floating-point implementations contradicted the assumption that equation (4) from the conventional method gives accurate results. Considering that such good performance may be due to the limited resolution in the fixed-point implementation, we also performed a simulation with the floating-point implementation. As can be seen from FIG. 5, the BER using the equation (5) is smaller than that using the equation (4) of the conventional method, and in particular, the equation (5) was used as the SNR increased. It can be seen that the system performs much better than the system using Equation (4).

As shown in FIG. 4, the impulse response of the channel exceeds the length of the cyclic prefix (CP) of the OFDM system. The second peak (echo at approximately 1.5E-6 seconds) causes intersymbol interference (ISI) between transmitted OFDM symbols. This clearly shows that it interferes with received 256-QAM symbols in the frequency domain.
Consider a transmission line with time domain impulse response h = [h (0), h (1),..., H (L + M−1)]. If the CP length is L, it is the length M of the extra channel delay. The corresponding reception time domain signal vector in OFDM transmission is

In general, when m ≠ n, it is clear that σ _g ² (m) ≠ σ _g ² (n). Thus, the ISI power associated with different subcarriers is different. In wire communication systems, the exact ISI distribution is inherently unknown, so using the conventional method (4) may actually degrade the performance of the TPC decoder. To prove this, floating point simulations were performed in the augmented channel. The reinforced channel impulse response is shown in FIG. As can be seen in FIG. 6, the reinforced channel is exactly the same as the previous channel with the response shown in FIG. 4 except that the second peak was removed and ISI beyond the CP period was removed. It is. Thus, there is no residual interference to the received data symbols.
The result of BER is a comparison between the system employing the formula (5) of the present invention and the system employing the formula (4) of the conventional method, and is shown in FIG. The BER results show that the system employing the conventional equation (4) is superior to the one using the equation (5) of the present invention.
Only when there is no unknown interference or interference source, the conventional method (4)
5) It has been proven to function better. From the above simulation, in the situation where channel state information (CSI) is not available, as seen in wired communication systems that are susceptible to interference due to delay waves, in contrast to wireless communication systems, we use equation (5) of the present invention. It can be confirmed that the soft decision demapping is superior to that using the conventional formula. Further, as is clear from comparison with the conventional method (4), since the equation (5) of the present invention does not require the channel gain | h _k | ² itself as a proportional coefficient for determining the symbol, , because there is no taking into account the received symbol _{(y k) (h k)} , the method of the present invention, can be eliminated to estimate overly noise due to delayed waves, giving reliable results Can do.

1 is a block diagram of a system that implements the method of the present invention. The graph which shows the signal point arrangement | positioning in 16 bit point QAM in an IQ plane in order to demonstrate the demapping method of this invention. FIG. 3 is a block diagram illustrating a simulation system used to simulate the method of the present invention for comparison with prior art decoding methods. The graph which shows an example of a channel response. The graph which shows the simulation result about the channel response of FIG. The graph which shows another example of a channel response. The graph which shows the simulation result about the channel response of FIG.

Claims (3)

A data symbol demapping method for decoding in a wired communication system, comprising:
Receiving a series of data symbols for information bits mapped in the IQ plane, transmitted over a wired channel and equalized;
Calculate the likelihood for each data symbol using the following formula:

Where λs (b _{i, k} ) indicates the likelihood of the i-th bit of the k-th symbol, y _k indicates the point of the equalized k-th received symbol in the IQ plane, and α Denotes an ideal symbol point in the IQ plane, and S _i ⁽¹⁾ and S _i ⁽⁰⁾ denote symbol sets whose i-th bit is “1” and “0”, respectively.
And
To determine a data bit for each received data symbol, using the likelihood obtained as described above as a soft decision value and providing a series of demapped data bits to be decoded later A method characterized by. The mapped data symbols The method of claim 1, wherein the modulated by orthogonal frequency division multiplexing (OFDM) with a subcarrier having a frequency band of 2 MHz to 30 MHz. The method according to claim 1, wherein the likelihood is assigned within a limit range of −0.1 to 0.1.

Images