JPH022277A - Multi-value qam communication system - Google Patents

Abstract

PURPOSE:To improve the coding efficiency and coding gain by dividing a base band signal into a coding bit and a non-coding bit, using the non-coding bit as a partial set, allocating it at a point where an inter-signal distance is maximized and sending the result and reproducing the signal at the demodulator side by the opposite operation to above. CONSTITUTION:The sender side forms 2<m> sets of subset signals subject to error correction coding through independent block coding of r-bit (2<=r<m) among m-bit base band signals of multi-value 2<m> (m is a positive integer). Then the non-coding signal in (m-r) bits not subject to error correction is used as a subset in the r-bit and modulated in a way that the inter-signal distance, of (m-r) bits is maximized in the signal space after modulation and the result is sent. The reception side reproduces the received signal in the opposite operation to that at the sender side. Thus, the bit number possible for error correction is increased by nearly a multiple of m/r and the coding gain of the r-bit series is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a multi-1i1QAM communication system for transmitting digital signals using multilevel orthogonal amplitude modulation.

Quadrature Amplitude
(Quadrature amplitude modulation)
The communication method is to transmit a multilevel QAM signal that is a composite of nWi amplitude modulation signals orthogonal to I and Q channels, and this multilevel QAM signal has n2 (=2) signal points.

For example, if it is n-n-8(・6), it becomes a 64-value QAM signal with 64 signal points, and n=16<mm=8
), it becomes a 256-value QAM signal having 256 signals.

On the receiving side, a carrier wave is regenerated from this multilevel QAM signal,
Demodulation is performed using regenerated carrier waves whose phases are orthogonal to each other, and a digital signal of a total quantity series of lQ channel signals is obtained by multi-level identification.

In such a multi-level QAM communication system, the transmission capacity can be increased by increasing the number of levels, but on the other hand, as the number of levels increases, internal noise in each part of the device and imperfections in the circuit may increase. Residual bit errors occur due to Therefore, various methods have been proposed to reduce this.

[Conventional technology]

It has been proposed to use random error correction codes as a useful means to reduce the bit error rate. That is, the bit error rate is improved by performing bit error correction encoding on the transmitting side and performing bit error correction decoding on the receiving side. As this type of system, the BCH code used in the 25eQAM system is known. For example, a bit error correction code/bit error correction code is added outside the differential logic.
A method in which a decoder is provided and a method in which a bit error correction code/decoder is provided inside the differential logic have been proposed.

Furthermore, as a method for realizing the [3 C11 code] used in the multi-level QAM communication system, Inventor Akina et al.
No. 2151, title of the invention, [Multilevel QAM communication system 1) proposes a configuration in which an independent error gf correct code/decoder is provided for an m-bit data sequence constituting a 2H-value multi-level modulation signal.

FIG. 9 is a block configuration diagram thereof. In the same figure, D
1 to D are m-bit baseband signal sequences constituting a binary modulation signal, 11 to 1□ are encoders with block length n, 2 is a modulator, 3 is a demodulator, and 41 to 4IIl are decoders. be. m-bit baseband signal! D −D is converted into a random error correction block code by the corresponding encoders 11 to m 1 , respectively. The converted block code is subjected to D/A conversion, then subjected to i-alternative modulation by a modulator 2, and transmitted as a binary QAMAM signal. On the other hand, on the demodulation side, the received binary QAMAM
After demodulating and Δ/D conversion with the signal modulator 3, the decoders 41 to
At step 41, decoding corresponding to error correction coding on the transmitting side is performed to correct bit errors. In this way, the illustrated configuration decomposes a multi-component code into a plurality of binary codes, secures a transbearing range against the uncertainty of the detection phase, and also secures the transmission range when a single code error becomes a multi-bit error. This prevents the error correction ability (code information WI) from deteriorating.

Furthermore, a coded modulation method has been proposed that improves the error correction ability, that is, the coding gain, without expanding the spectrum at the cost of expanding the signal space (signal point: 214 → 2H"l) ( For example, G.

IJ ngerboeck, Channel c
odina with multi-evcl /ph
ase signals”, I E E E
l”rans.

IT, VAT, 28. N(11, pll,
55-67, Jan.

1981 Niggi, Ungerboeck, "Dennel coding using multilevel/phase signals"). This method uses convolutional codes and maximum likelihood decoding, and applies the maximum likelihood decoding method to the received sequence. It selects the transmission sequence closest to
, t) The block code has a block length of n bits, k
It has pit information and can correct up to t bit errors in one block. At this time, the redundant bits used for error correction are (n - k) bits, and the coding efficiency is 1 (1 x m = / n.) The CH code is converted into a multilevel QAM as shown in Figure 9. When applied to a communication system, there will be (n-k)·m redundant bits for the number of transmission bits (n-m bits).

As is well known, in fields such as wireless communication using multilevel QAM communication systems, it is important to improve frequency usage efficiency, so the number of redundant bits for error correction is minimized (bringing coding efficiency close to 1). It is necessary to set it up accordingly. That is, in error correction using block codes applied to multilevel QAM communication systems, it is necessary to increase coding gain without reducing coding efficiency.

[Problem to be solved by the invention]

However, although the configuration in which an independent error correction code/decoder is provided for an m-bit data sequence can improve the code size, the redundant bits for error correction are (n
-k)·m bits, and it is difficult to increase the encoding efficiency.

Furthermore, although the configuration using convolutional codes and maximum likelihood decoding can improve the coding gain, it is necessary to increase the constraint length (number of states) of the encoder. In this case, maximum likelihood decoding is performed according to the Viterbi algorithm (expands the received sequence in a grid pattern and determines for each state which is the most likely of the two paths that meet in a certain state). Then, △C, which is the basic arithmetic circuit for determining the most likely bus,
3(Add, Compare and 5elec
t) The circuit grows in proportion to the number of states. Furthermore, as the number of multi-values increases, the number of bits to be applied increases. For this reason,
There is a problem in that the circuit scale of the signal generator increases significantly.

Therefore, an object of the present invention is to solve the above problems and improve both coding efficiency and coding gain without significantly increasing the circuit scale.

[Means to solve the problem]

The iflQAM communication system of the present invention is configured as follows.

On the sending side, the number of multivalues is 2 (m is a positive integer)
The baseband of rn bits (
Of the IT code, r bits (2≦r<m) are independently block encoded to obtain two error correction encoded subset signals. Then, the uncoded signal of (m−r) bits without error correction is set as a subset of r bits (within the subhit of r bits (i?)), and the (m−r) The bits are modulated and transmitted so that the distance between the signals is maximized.

On the receiving side, the received signal is subjected to multi-value identification to obtain an m-bit signal, and each of the error-correction encoded r-bit signals is independently decoded to reproduce a rib set signal.

Also, the (m−r) bit uncoded signal is subsensor l−
In a time slot in which error correction has been performed during signal reproduction, the signal point closest to the received signal point among the corrected R-pit subset signals is selected from the subset and included. (m-r) bits are reproduced as a non-coded signal. On the other hand, at time [1] where error correction was not performed, the (m-r) bits obtained by multi-level identification are reproduced as they are as a non-coded signal.

[Effect]

By adding redundant bits to only the r bits of the m-bit baseband signal, the signal of one pit sequence among the r pit sequences can be reduced to the maximum P of the n transmission bits.
A block code is formed in which =(nk)·m/r bits are redundant bits.

Therefore, the number of error-correctable bits increases by about m/r times,
The coding gain of the r-bit sequence is improved.

An uncoded signal of (m−r) bits without error correction, that is, without adding redundant bits, is treated as a subset of r bits (an r-bit signal in the navset) and is The configuration in which the modulation is performed so that the distance between signals of (m-r) bits is maximized is to enable identification of uncoded signals of (m-r) bits with the minimum error rate. It is. Therefore, this improves the error rate characteristics of the uncoded signal. .

On the receiving side, the error correction encoded r-pit signals are each independently decoded to reproduce a subset signal. In the time slot in which error correction has been performed during this decoding, one sub-hit signal after error correction is selected. Then, by selecting from the subset one signal point that is closest to the received signal point among the two corrected subset signals, the error correction of the non-coded bits can be performed with high accuracy. In a time slot in which error correction is not performed during reproduction of one subset signal, the (m−r) bits obtained by multi-value identification may be used as the reproduction signal as is.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings. 1st
The figure is a block diagram of an embodiment of the present invention, in which (a) shows a modulation section (transmission side) and (b) shows a demodulation section. encoder 1
11-11. is the r bits of D to D of the m-bit baseband signal constituting the binary modulated signal (
2≦r<m) 1−r+I In are independently block encoded (block length [1)]. Each 1-bit sequence signal forms a block code having a maximum of (n-k)·m/ bits as redundant bits. The encoders 111-11. The r-bit sequence signal of
2 (r bits) subset signals are formed. The code converter 15 converts the m-bit baseband signal into
~Input the (m-r) bits of DIll-1 as they are. That is, the (m-r) bits of D1 to DII-r are non-coded signals ((two nth points) that do not involve error correction.

The code converter 15 inputs an (rn-r) bit non-coded signal and an r-bit subset signal, and maps these input signals onto a signal space.

This mapping is done by treating the 31 encoded signal of (m-r) pitch i- as a signal within a subset of r bits (a subset of r bits), and the signal dark path fit (Euclidean distance) between them is the maximum. This is the process of placing the point at the point where . Note that this mapping will be explained in detail later.

The modulator 12 has m bits < 1 output from the code converter 15.
．． 0 channel signal), performs known orthogonal modulation, and outputs a multilevel QAM signal.

The demodulator is configured as follows. , the demodulator 13 demodulates the received multilevel QAM signal by performing an operation opposite to modulation. The demodulated signal is converted to (m+
s) is identified as a bit signal. However, S is a soft decision bit and indicates the direction of shift of the received signal point with respect to the signal point on the signal space. The code inverse converter 16 performs an operation (de-wappi) inverse to the mapping of the code converter 15 on the transmitting side.
No: De-mapping). The output m bits of the code inverse converter 16 are given to the non-coded signal reproducing unit 118 via the digital delay circuit 17 along with the soft decision bits S from the decoder 13. The digital diversion circuit 17 is connected to the decoder 141.14. The output m bits and soft decision bits S of the code inverse conversion ♂16 are
are delayed and these signals and decoders 141-14. Align position 9 on the time axis with the signal from. Further, a subset signal of r bits out of the m bits output from the sign inverse converter 16 (containing errors at this point) is sent to the decoders 141 to 14.
, respectively. , ti @ v 141~1
4 and 4 perform operations opposite to those of the encoders 111 to 11r, respectively, and output error-corrected Dm-r+1 to omg)r bits. The error correction pulse (syndrome locator) obtained during this error correction is transmitted to the non-coded signal regenerator 18.
given to. The non-coded signal regenerator 18 inputs each of the above-mentioned signals, and in the time slot in which error correction was performed when reproducing the first set signal, the most received signal point among the decoded r-bit subset signals is selected. Select signal points that are close to each other. On the other hand, in time slots in which error correction is not performed when reproducing the subset signal, the (m−r) bits obtained by multi-value identification are output as they are. In this way, the non-coded signal regenerator 18 outputs D
The unary-encoded bits of "Dm-r" to l are extrapolated.

Next, in order to clarify the mapping performed by the code converter 15 in this actual case, a specific example will be given in which m=4. The case of r-2 will be explained.

Figure 2 shows 2IIl (=1
6) A value modulation signal, where A to D indicate coded bits (subsets), and Zafix O to 3 indicate non-coded bits. FIG. 3 shows the mapping rules performed by the code converter 15 on the modulation side.

FIG. 3(a) shows non-encoded bits (X: Δ, B, C.

D), and FIG. 3(b) shows the encoded bit (i: 0.1
．． 2.3) is shown. The code representation (α, β, γ, ζ) on the signal space (two-dimensional plane) in Fig. 2 is α.

β is J1 encoded bit 1. Indicates Q and γ.

ζ is each encoded bit 1. As shown in @), the mapping performed by the code converter 15 is to map an uncoded signal of (m−r) bits without error correction onto the signal space after modulation within a subset of r bits. The signals are arranged at the point where the distance between these signals is maximum. m=4. r
= 2, 22-4 subset signals A, B, C,
If D is obtained and the 2-bit non-coded signals treated as a subset of these signals are arranged on the modulated signal space so that the distance between them is maximized, the result will be as shown in FIG. As illustrated, Ao to A3. Bo-B3°C-C and. -D3 are arranged every other channel in the I channel and Q channel directions of the signal space. The code converter 15 that performs the above mapping may be configured with a D/A converter that receives (α, γ), (β, ζ) as input, or
It may also be configured with OM (Read-Only-Memory).

When configured with ROM, the ROM contains the information shown in Figure 3(a).
, (b) are held in the form of a table.

Note that m=4. In the case where m = 2, the equal codes of the ■ channel and the Q channel are obtained by performing simple natural binary addition of 1 encoded bit and 1 non-encoded bit, as shown in Fig. 2. Sixteen signals on the signal space are obtained. Therefore, in this case, the code converter 15 is unnecessary. However, in the case of a higher number of values such as 64 values or 256 values, or when the encoders 111 to 11. If an odd number of 1,
Next, the operation of this embodiment will be explained when m=4. This will be explained by taking the case of r=2 as an example.

Of the 4-bit baseband signal of Do-03, Do
and Dl are sent directly to the code converter 15, and D and D3
are block encoded by encoders 111 and 112, respectively, and sent to code conversion 515. These signals are subjected to the above-mentioned mapping processing in the code converter 15, and then orthogonally modulated in the modulator 12, and a multilevel QAM signal is output.

The II modulator 13 on the demodulation side demodulates the received multi-1flQAM signal and outputs a total of 6 bits including 2 soft decision bits S. This 6-bit signal is sent to a non-encoded signal 4 regenerator 18 via a digital delay circuit 17.

(If m=4.r=2, the data mapping process by the code inverse converter 16 is unnecessary). Further, the 2-bit subset signal is error-corrected by decoders 141 and 142.

Now, as shown in Fig. 4, the transmission code sequence is...Bo
, Dl, [33, A2. Do, Co, . . . (see figure (a)), time (time slot), n+2 and t,
A code error occurs at +4, and B3 and so on in the transmission code sequence. Suppose that they are incorrectly assigned to A3 and A1, respectively (see figure (b)
)). Further, the position of the received signal at this time is represented by M and ■ in the signal space shown in FIG. 5, respectively. Furthermore, the fifth
In the figure, eI and eq represent the error bits of the ■ channel and Q channel, respectively, and el and e. Soft decision bit S
is formed. At the time when error correction is performed, the non-coded signal regenerator 18 adds error bits e (and e) obtained by soft decision to two bits of each of the channel (1) and the Q channel, and The upper bits are error-corrected non-coded bits.

For example, time t. +2, uncoded signal regenerator 18
calculates A3+e (e is e (and the soft decision bit S represented by eg). At this time, B3 is mistaken for A3, so it is an error only in the I channel. In this case, the soft decision bit is e (=1 (received signal point M in Fig. 5 is A
The error bit e■ is shifted in the direction of 1 with respect to the signal point 3). Also, the 2 bits of the ■ channel of signal point A3 are “1”.
o', the two bits of the Q channel are "10'".

Therefore, A3+e −83 Ich 10+et ” 11 QCh 10 =10, and A3 is the signal point (1, 1
, 1, O), that is, the error is corrected to B3. Note that the erroneous subset signal A (γ=0.ζ−〇) is sent to the decoder 14.
Error correction is performed in step 142, and an error-corrected subset signal B (γ=O0ζ=·O) is obtained.

Also, time t. , 4, the uncoded signal regenerator 18
performs the operation [], -)e. At this time, since D0 was mistakenly set to A1, both the I channel and Q channel contain errors. In this case, signal point A1 in FIG.
From the relationship between
, eo =・O. Also, 2 bits of 1 channel at signal point A1 are "10" and 2 bits of Q channel are 00.
”. Therefore, A +e =D0 1ch 10+(3I==O'I Qch OO+eQ==01, and A3 is the signal point (0,0
, 1, 1), that is, the error is corrected to Do. Note that the error in subset number 41 is corrected by decoders 141 and 142, and the error-corrected 1-knob set signal D (γ=1.ζ
=1) is obtained.

In the above manner, error J[correction is performed.

Here, m=4. Non-coded signal regenerator 1 when r=2
A specific configuration for executing the above-mentioned error correction algorithm performed in step 8 is shown in FIG. As shown in the figure, the non-coded signal regenerator 18 consists of a channel part 18 and a Q channel part 18o, and since they are the same, the channel part 18. Only the circuit configuration of is shown. Ibu 1? Channel error 51 correct (, 1, inverters 20, 23 and 24, exclusive OR circuit 21, AND circuits 22, 25 and 26, and OR circuit 27 are connected as shown. Medium 1.01 is non-i1″ encoded bit (α, β)
, 12. Q2 is the encoded bit (γ.

ζ), 11, Ql are error corrected uncoded bits, 2. ．． 2o is an error correction pulse (syndrome [1 digit) obtained by the decoders 141 and 142, and becomes 1 when an error occurs. In operation, for example, at time t. At +4, ■ channel ttl −1, l2=
Since e1=0, the output of the exclusive OR circuit 21 is 1, and therefore the outputs of the AND circuits 25 and 26 are also O, which is 1'-0. On the other hand, the Q channel is Q,
=0°Q -0 and eo=1, so Q,'=0
The circuit shown in FIG. 6 is a 2-bit + 1-bit full adder that outputs the upper 1 bit, but it can generally be constructed from a multi-bit adder.

(1) An embodiment of the present invention has been described above. The above example is 1
．． This was an example in which the 4-value n code of the Q channel was expressed in natural binary, but as is well known, in the case of a general multi-value m code, a Gray code is used in which 1 code error corresponds to 1 bit error. Using this method has better error rate characteristics. In particular, when error correction is performed, it is possible to reduce the occurrence of multiple errors in non-encoded bit sequences.

For example, in a 2-bit transmission system using an encoder and decoder capable of single error correction, even if errors occur in two codes in one block, in the case of a Gray code, it will only result in a 2-bit error. If a 2-bit error is distributed to two streams, a single error has occurred in each series, which can be corrected. On the other hand, in the case of natural codes, there are 2 to 4 bit errors, which may not be correctable.

For non-encoded bits, the effect is particularly noticeable in the case of error correction. For example, J shown in Figure 7, sea urchin, 0-7
Consider the case where a transmission code (3 bits 1-) is transmitted using 2 non-encoded bits and 1 encoded bit. In this case, it is assumed that the multi-level discrimination in the demodulator operates in natural binary. If the transmission code 3 is incorrectly corrected as 4, 1 is added to the non-encoded bits, so 011 → 100 (
natural binary operation), and the unencoded bits are 01-+10
This results in a 2-bit error correction. However, if it is Gray-encoded, it will be Gray-transformed (10→1
1), the non-encoded bits change from 01 to 11, and only one bit of error needs to be corrected.

When using this Gray code, the r-bit coded signal and the (m')-bit non-coded signal of the m-bit baseband signal are code-converted so that they become Gray codes in the signal space. On the other hand, on the demodulation side, the multi-level identification result is gray coded for only r bits of the encoded signal, and then decoding processing is performed.On the other hand, based on the result of this decoding processing,
) After selecting the uncoded signal of bits, perform Gray transformation.

The system configuration when using this Gray code may be the same as that shown in FIG.

Next, experimental results will be explained in order to clarify the effects of the present invention.

FIG. 8 is a diagram showing error rate characteristics of a 256Q AM signal. The horizontal axis of the figure is the carrier-to-noise ratio (, (3) (CN
R: Carrier-to-No1se Rat
io>, and the 'Ii axis represents the bit error rate (BER:B
ftErrorRate ). As an error correction code, BCH (31, 26, 1> code (block length = 31, information bits: 26. Number of error correctable pits: 1)
) is applied to the conventional configuration shown in FIG. When applied to the configuration of the invention, - For example, when m = 8 and r = 2, the BCH (31, 11, 5) code can be used with the same coding efficiency (26/31) as the conventional one, so the error correction ability is improved. As shown in Fig. 8, it is possible to obtain a coding gain of about 5 cE.
1, 5) When applying the code (coding efficiency: 11/3
1), the reason why the error rate characteristics deteriorate is that when the coded bits are corrected, if the polarity of the soft-decision bits is wrong, the non-coded bits will be wrong, and the coded bits will be wrong. Due to error propagation that occurs during correction.

〔Effect of the invention〕

As explained above, according to the present invention, a baseband signal is divided into coded bits and non-coded bits, and the non-coded bits are modulated into a subset signal in which the coded bits are block-coded. On the demodulation side, in the time slot where error correction was performed during demodulation of the subset signal, the non-encoded bits are placed at the point where the distance between signals is maximum on the signal space of Since the signal closest to the received signal point is selected for reproduction, it provides a multi-level QAM communication system with high coding efficiency and high code density without increasing the scale of the circuit. can do.

[Brief explanation of the drawing]

Figure M1 is a block diagram of an embodiment of the present invention, Figure 2 is a diagram showing a signal space for explaining mapping in an embodiment of the present invention, and Figure 3 is a diagram of encoded bits in an embodiment of the present invention. FIG. 4 shows the transmission code and reception code used to explain the error correction of the non-coded signal performed by the non-coded signal regenerator in FIG. 1. Figure 5 is a diagram showing a signal space for indicating the position of a received signal when a code error occurs, and Figure 6 is a circuit diagram of an example configuration of a non-coded signal regenerator that performs error correction of non-coded bits. , FIG. 7 is an explanatory diagram of an embodiment using Gray codes, FIG. 8 is a diagram showing error rate characteristics for detailed explanation of the present invention, and FIG. 9 is a diagram showing the currently proposed multilevel QA
It is a block diagram of M communication system. 111-11... Encoder, 12... Modulator, 13
... Demodulator, 14-14. ...Decoder, 15...
- Code converter, 16... code inverse converter, 17... digital delay circuit, 18... non-coded signal regenerator. (b) Block diagram of an embodiment of the present invention =・tl, tn*l, ! ne2, jr+4
, tn*4, tn's ・=(α)...8G. B Do. Co... Fig. 4 shows the codes used to explain error correction of non-coded signals Fig. 4 shows the position of the received signal when a code error occurs Fig. 5 An explanatory diagram of mapping in the embodiment of the present invention Figure 2: Circuit diagram for correcting errors in unencoded bits. Figure 7 - Transmission line return 1 - Explanation of an embodiment using encoded bits/togeray codes. Figure 2: Detailed explanation of the present invention. Diagram showing error characteristics Diagram between blocks of the proposed multi-level QAM borrowing system

Claims (1)

[Claims] On the transmitting side, r
The bits (2≦r<m) are each independently block encoded to obtain 2^r subset signals encoded with error correction, and the uncoded signal of (m−r) bits without error correction is r. As a subset of bits, it is modulated and transmitted so that it is located at the point where the distance between signals of (m-r) bits is maximum on the signal space after modulation, and on the receiving side, the received signal is multivalued. identify and obtain an m-bit signal, independently decode the error-correction encoded r-bit signal to reproduce a subset signal, and
In the time slot in which error correction was performed when reproducing the subset signal, the (m-r) bit uncoded signal is located closest to the received signal point among the corrected r-bit subset signals. A signal point is selected from a subset and the (m-r) bits contained therein are reproduced as a non-coded signal, and in time slots in which error correction was not performed, the (m -r) A multilevel QAM communication system characterized by reproducing bits as they are as a non-coded signal.