CN1330454A - Turbo grille coding modulation - Google Patents

Abstract

In a method of transmitting data over a communications channel, at least some of the bits of an incoming bit stream are passed through a turbo encoder to generate turbo encoded output bits, and words corresponding to symbol points on a constellation in a trellis code modulation scheme are generated using at least the bits passed through the turbo encoder, possibly in conjunction with other bits that are not passed the through the turbo encoder. Typically, the turbo encoded bits are the least significant bits.

Description

The turbo grille coded modulation

The present invention relates to field of data transmission, be specifically related to a kind of modulation scheme that is used for sending by communication channel data in Discrete Multitone modulation system for example.

In order to improve data-transmission efficiency by communication channel, data are transmitted as code element, each code element is represented a plurality of bits.For example, in QAM (quadrature amplitude modulation) system, code element is by the amplitude and the phase place representative of signal.For example, temporal four bits of 16 unique code elements (being the combination of amplitude and phase place) representative.Code element forms conformation (constellation) a little on phase amplitude figure.Along with the increase of number of symbols, the possibility of transmission error also increases.Adopt the forward error correction coding scheme, to allow receiver to detect mistake and to recover the correct code element that sends.

A kind of optimized encoding form in the data communication is a convolutional encoding, and convolutional encoding is to depend on the bit-level encoding scheme of bit sequence formerly.In trellis coded modulation, increase number of symbols to provide redundant.Only allow specific conversion (transition).Under the situation that mistake takes place, receiver can utilize the knowledge to all possible permission conversion to detect most probable correct conversion.

Different with the block code that sends data with predetermined block, convolution code is bad to deal with burst error.Partly, developed turbo (Te Bo) sign indicating number in order to address this problem.In essence, the turbo sign indicating number comprises two or more convolution codes of being separated by the interleaver that the list entries of first encoder is operated.For example, referring to AT﹠amp; The breadboard Hamid R.Sadjapour of T Shannon is 1996 " the turbo sign indicating number of Discrete Multitone modulation is used " of delivering.

Because the high coding gain of turbo sign indicating number, it receives increasing concern.In DSL (digital subscriber line) system, used the turbo sign indicating number to replace trellis code, obtain the better error rate (BER) performance.But when constellation size increased, the coding gain advantage of turbo sign indicating number began to become unnecessary.This is because redundant bit makes constellation size become bigger.

An object of the present invention is to increase the message transmission rate in the DMT system for example.

According to the present invention, a kind of method by communication channel transmission data is provided, comprising: receive an incoming bit stream; Make at least some described bits export bit to produce the turbo coding by the turbo encoder; And use the described at least bit that passes through described turbo encoder, produce word corresponding to the code element point on the conformation in the trellis code modulation scheme.

In the present invention, preferably use the turbo encoder, come only the least significant bit in the conformation (LSB) to be encoded, because LSB is the most responsive to mistake.Accessible by this way data transfer rate only differs several dB with the Shannon capacity.The present invention preferably combines powerful turbo sign indicating number and trellis coded modulation scheme to improve data transfer rate in DMT (Discrete Multitone) system.

In the DMT system, use a plurality of subchannels to send data, each subchannel has different carrier waves and different QAM conformations, and different QAM conformations comprises the bit of the different numbers on each constellation points.Usually, the bit number on each constellation points is an integer, and if a subchannel can not support a data bit, this subchannel is unavailable so.According to the present invention, the spread-spectrum algorithm can be combined with the turbo trellis coded modulation, the channel that makes carrying be less than a bit information also can use.As a result, can greatly improve overall channel capacity.

Below will only the present invention is described in detail with reference to accompanying drawing with way of example, wherein:

Fig. 1 shows o'clock in accordance with the principles of the present invention encoder of x and y＞1, and wherein x and y are the numbers of the bit in each constellation points (code element);

Fig. 2 shows the coder structure of x=1 and y＞1 o'clock, and wherein the turbo encoding rate is 2/3;

Fig. 3 shows the coder structure of y=1 and x＞1 o'clock;

Encoder when Fig. 4 shows x=y=1, wherein encoding rate is 1/2;

Fig. 5 is the block diagram of decoder;

Fig. 6 shows a conformation, and how expression determines three final bits;

Fig. 7 shows a conformation, represents determining of highest significant position.

By DMT (discrete multisound system) structure the present invention is described below, the DMT system can comprise 1000 subchannels usually, each subchannel can carry the different number code elements of representing the given number bit, promptly, the constellation points number of each subchannel can change, so the bit number of each constellation points also can change.Encoder

As shown in Figure 1, the part of incoming bit stream is fed to encoder data block 10, and encoder data block 10 is addressable memories.Suppose that each code element has 10 bits, comprise a check bit of per two subchannels, 1000 channels once can carry 9500 bits so.Therefore, common 9500 bits with an incoming bit stream are fed in the encoder.With the sub-fraction of these bits, normally 1500 bits are fed to encoder data block 10.

Preferably encoder is embodied as the parallel encoder that we describe in the Britain application No.0010330.9 that submitted on April 18th, 2000, the full content of this application is introduced into this paper, as a reference.

In the example shown, sequentially export three bit u from encoder data block 10 ₁, u ₂, u ₃, and export three bit u ' ₁, u ' ₂, u ' ₃As interleaving data.Data bit u ₂And u ₃Form the component v of the first output word v ₀, v ₁, bit u ₁Form the bit w of the second output word w ₁By behind the shift register 16,18 separately, utilize 12,14 couples of bit u of recursive system convolution coder ₁, u ₂, u ₃And u ' ₁, u ' ₂, u ' ₃Group carry out turbo and encode and form bit w ₀

Employed similar in the conformation coder structure that is adopted and the ADSL system.Binary word u=(u _{Z '}, u _{Z '-1}..., u ' ₁) determine two binary word V=(V _{Z '-y}..., V ₀) and w=(W _Y-1..., w ₀) (wherein z '=x+y-1), use this two binary words searches two constellation points (each constellation points comprises x and y bit respectively) in the encoder look-up table.

Fig. 1 shows the coder structure of x＞1 and y＞1 o'clock, and use therein turbo encoder is that the encoding rate with ratio 1/2 perforation is 3/4 systematic encoder.Turbo encoder 20 comprises two recursive system convolution coders 12,14 (RSC1 and RSC2).Encoder RSC1 receives the alphabetic data from encoder data block 10, and encoder RSC2 receives the interleaving data from identical block 10.

The length of data block depends on the quantity of the data that just are being sent out in each signal frame, is 9500 bits in the above-mentioned example.Usually, send integer number of data blocks in each signal frame.Fig. 2 to 4 shows the coder structure that is used for other x and y value.

Fig. 2 shows the coder structure of x=1 and y＞1 o'clock, and wherein the turbo encoding rate is 2/3.In y=1 and x＞1 o'clock, coder structure as shown in Figure 3 is to shown in Figure 2 similar.

Coder structure when Fig. 3 shows x=y=1, wherein encoding rate is 1/2.When (or x＜1), can use the coder structure similar in y＜1 according to the value of x (or y) to Fig. 1 to 4.Its difference only is and will utilizes an extended code to use K subchannel to send bit, wherein a y=1/K.

If the extended code that is used is [b ₁, b ₂..., b _k], 0 can be sent as [b so ₁, b ₂..., b _k], wherein (k=1,2 ..., K), 1 can be sent as [b ₁, b ₂..., b _k].The conformation of each subchannel in the K groups of subchannels is used a bit of every channel conformation, and k channel sends bit b _kIn a word, need K subchannel to send a data bit.The advantage of this set is if the different modulating demodulator in identical cable core uses different extended codes, can greatly reduce crosstalking of self so.(vol.4, no.3 have described suitable extended code in pp.80-82) at IEEE Communication Letters in March, 2000 at R.V.Sonalkar and R.R.Shively.Decoder

The decoding program of turbo trellis coded modulation may further comprise the steps:

1) least significant bit (LSB) is carried out soft decoding;

2) highest significant position (MSB) is carried out hard decoder;

3) use the turbo decoder algorithm that LSB is decoded; With

4) determine all data bits.

If a N bit conformation is used to the transfer of data in the given subchannel, this conformation position can be represented by two-dimensional vector so: X _b=[b _XM, b _{X (M-1)}..., b _X1, 1] and Y _b=[b _YM, b _{Y (M-1)}..., b _Y1, 1], M=N/2 when wherein N is even number, M=when N is odd number (N+1)/2.Decoder is for X _bAnd Y _bBe identical.

If the data that received be (X, Y).If (2 ^M+ 1+2k)＜X＜(2 ^M+ 1+2 (k+1)), k=0 wherein, 1 ..., 2 ^M-1, and keep X ₁=(2 ^M+ 1+2k) and X ₂=(2 ^M+ 1+2 (k+1)), so final X gets X ₁Still get X ₂, depend on decoder result from LSB.For N＞1, the soft bit of LSB among the X (the logarithm probability that does not have constant) is confirmed as: σ wherein ²It is noise power.By in above-mentioned formula, replacing X can obtain the soft bit of LSB among the Y in a similar manner with Y.

If N=1, soft bit is ${\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A0111565900161.png,A0111565900171.png"\; state="\{\{state\}\}">P</figure-callout>}}_1=\log \left(\mathrm{prob}(b_{x1}=1)\right)=-(\frac{{(X+1)}^2}{2{\mathrm{\&sigma;}}^2}+\frac{{(Y+1)}^2}{2{\mathrm{\&sigma;}}^2})$ $P_0=\log \left(\mathrm{prob}(b_{x1}=0)\right)=-(\frac{{(X-1)}^2}{2{\mathrm{\&sigma;}}^2}+\frac{{(Y-1)}^2}{2{\mathrm{\&sigma;}}^2})$

If N＜I and extended code are [b ₁, b ₂..., b _k], so can the soft bit of following calculating ${\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A0111565900161.png,A0111565900171.png"\; state="\{\{state\}\}">P</figure-callout>}}_1=\log \left(\mathrm{prob}(b_{x1}=1)\right)=-\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}(\frac{(X+{(1-2b_k)}^2)}{2{\mathrm{\&sigma;}}^2}+\frac{(Y+{(1-2b_k)}^2)}{2{\mathrm{\&sigma;}}^2})$ $P_0=\log \left(\mathrm{prob}(b_{x1}=0)\right)=-\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}(\frac{(X-{(1-2b_k)}^2)}{2{\mathrm{\&sigma;}}^2}+\frac{(Y-{\left(2b_k\right)}^2)}{2{\mathrm{\&sigma;}}^2})$

Soft bit is exported the turbo decoder circuit that sends to as shown in Figure 5.This turbo decoder comprises two LOG-MAP decoders 30,32.Each decoder comprises forward direction (α) iteration, and is back to (β) iteration, and carries out final soft bit output and calculate.Unique difference is that this output not only comprises data bit but also comprises the error checking and correction bit of its last iteration.

Needing the reason of this error checking and correction bit of output is to need LSB to come from two possible constellation points X ₁And X ₂(or Y ₁And Y ₂) in determine X (or Y), and some LSB is the error checking and correction bit.The document of the Sadjapour that can speak of in front and C.Berrou and A., (Vol.44 finds the specific example of turbo decoder to the IEEE Trans.on Communications that delivers in October, 1996 in No.10) " near optimal error correction coding and decoding Turbo code ".

The following calculating of soft output error checking and correction bit at time k

P _ck1＝prob(b _ck＝1)＝MAX _(s，s′)[γ _ck1(R _k，s，s′)α _k-1(s′)β _k(s)]

P _Ck0=prob (b _Ck=0)=MAX _{(s, s ')}[γ _Ck0(R _k, s, s ') and α _K-1(s ') β _k(s)] wherein s is the state of turbo encoder at time fk, and s ' is the state at time k-1.R _kRepresent received data.β _k(s) be in state s (time k) back probability, a to iteration _K-1(s ') representative is at the probability of state s ' (time k-1) forward direction iteration.γ _Ck0(R _k, s, s ') and γ _Ck1(R _k, s, s ') and be to be R in received data _kAnd the error checking and correction bit is respectively 0 and transformed to the probability of s at 1 o'clock from state s '.

By behind the turbo decoder, determine LSB, and if N＞1, so still need from two possible constellation points, to determine MSB.With X is example, and it has two probable value X ₁Or X ₂(they are two consecutive points in the conformation).For two adjacent constellation points, X ₁And X ₂LSB must be different.Therefore, by checking that its LSB can be from X ₁And X ₂In determine X _b=[b _XM, b _{X (M-1)}..., b _X1, 1].Similarly, can determine Y _b=[b _YM, b _{Y (M-1)}..., b _Y1, 1].Determining X _bAnd Y _bAfter, obtain the final data bit that receives for following three kinds of situations:

When N was even number, final bit was [b _N, b _N-1... b ₁]=[b _XM, b _YM, b _{X (M-1)}, b _{Y (M-1)}..., b _X1, b _Y1].If N=3, three final bits are determined by conformation as shown in Figure 6, and are as shown in table 1 below.

????b x2b x1	???b y2b y1	???b 3b 2b 1
			??????00	?????00	????000
??????00	?????01	????101
			??????00	?????10	????001
??????00	?????11	????001
			??????01	?????00	????000
??????01	?????01	????101
			??????01	?????10	????111
??????01	?????11	????111
			??????10	?????00	????100
??????10	?????01	????100
			??????10	?????10	????110
??????10′	?????11	????011
			??????11	?????00	????010
??????11	?????01	????010
			??????11	?????10	????110
??????11	?????11	????011

If N is odd number and N＞3, can determine low bit (N-5), i.e. [b in the mode identical so with even number N situation _N-5, b _N-6..., b ₁]=[b _{X (M-3)}, b _{Y (M-3)}, b _{X (M-4)}, b _{Y (M-4)}..., b _X1, b _Y1], determine 5 MSB according to the conformation among Fig. 7, as shown in table 2 below.

????B xMb x(M-1)b x(M-2)	????B yMb y(M-1)b y(M-2)	????b Nb N-1b N-2b N-3b N-4
			????000	????000	????00000
????000	????001	????00001
			????000	????010	????10100
????000	????011	????10100
			????000	????100	????10101
????000	????101	????10101
			????000	????110	????00100
????000	????111	????00101
			????001	????000	????00010
????001	????001	????00011
			????001	????010	????10110
????001	????011	????10110
			????001	????100	????10111
????001	????101	????10111
			????001	????110	????00110
????001	????111	????00111
			????010	????000	????10000
????010	????001	????10001
			????010	????010	????10110
????010	????011	????10110
			????010	????100	????10111
????010	????101	????10111
			????010	????110	????11100
????010	????111	????11101
			????011	????000	????10000
????011	????001	????10001
			????011	????010	????10001
????011	????011	????10001
			????011	????100	????11101
????011	????101	????11100
			????011	????110	????11100

????011	????111	????11100
			????100	????000	????10010
????100	????001	????10011
			????100	????010	????10011
????100	????011	????10011
			????100	????100	????11110
????100	????101	????11110
			????100	????110	????11110
????100	????111	????11111
			????101	????000	????10010
????101	????001	????10011
			????101	????010	????11000
????101	????011	????11000
			????101	????100	????11001
????101	????101	????11001
			????101	????110	????11110
????101	????111	????11111
			????110	????000	????01000
????110	????001	????01001
			????110	????010	????11000
????110	????011	????11000
			????110	????100	????11001
????110	????101	????11001
			????110	????110	????01100
????110	????111	????01101
			????111	????000	????01010
????111	????001	????01011
			????011	????010	????11010
????111	????011	????11010
			????111	????100	????11011
????111	????101	????11011
			????111	????110	????01110
????111	????111	????01111

Should be appreciated that, described turbo sign indicating number is used in combination with trellis code and can obtains more performance than the trellis code of current use.When the spread-spectrum algorithm is combined with the turbo-trellis coded modulation, might use carrying to be less than the channel of a bit information, cause the very big increase of channel capacity.

Can use and well known to a person skilled in the art that standard DSP technology realizes the above-mentioned functions piece in digital signal processor.

Claims (18)

1. method that sends data by communication channel, comprise: receive an incoming bit stream, make at least some described bits export bit to produce the turbo coding by the turbo encoder, and use the described at least bit that passes through described turbo encoder, produce word corresponding to the code element point on the conformation in the trellis code modulation scheme.

2. according to the process of claim 1 wherein that described bit by the turbo encoder is the least significant bit (LSB) of input data, highest significant position is directly transferred to output word.

3. according to the process of claim 1 wherein that the part with incoming bit stream is fed to a memory, and will be fed to the recursive system convolution coder from the interleaving bits group of described memory to produce described turbo coding output bit.

4. according to the described code element of carrying on a plurality of subchannels that the process of claim 1 wherein in Discrete Multitone (DMT) system.

5. according to the process of claim 1 wherein at least some described subchannels, conformation code element point comprises and is less than a bit, and for these subchannels, uses extended code to send a whole bit on K subchannel.

6. according to any one method in the claim 1 to 5, wherein said word comprises the error checking and correction bit.

7. according to the process of claim 1 wherein that described word is used to obtain constellation points from the encoder look-up table.

8. one kind is used for input traffic is encoded with the encoder by traffic channel, comprising: encoder data block, a part that is used to store input traffic; The first recursive system convolution coder is used to receive the alphabetic data from encoder data block; And the second recursive system convolution coder, being used to receive interleaving data from encoder data block, described convolution coder is exported the least significant bit of an output data word at least, and this output data word forms the code element point on the conformation in trellis code modulation scheme.

9. encoder according to Claim 8 also comprises the shift register of the input that is connected to each recursive system convolution coder, and each unit of described shift register receives the corresponding bits from described encoder data block.

10. encoder according to Claim 8, wherein when x=y=1, wherein x and y are two bit numbers in the conformation, the input of described recursive system convolution coder is directly connected to described encoder data block.

11. the method that turbo trellis code modulation signal is decoded comprises:

(i) least significant bit to received signal carries out soft decoding;

(ii) the highest significant position to input signal carries out hard decoder;

(iii) use the turbo decoder algorithm that least significant bit is decoded;

(iv) determine all data bits.

12. according to the method for claim 11, wherein following processing received signal is to determine the soft bit (the logarithm probability that does not have constant) of LSB:

σ wherein ²It is noise power.

13., wherein soft bit is sent to the turbo decoder to produce data bit and error checking and correction bit according to the method for claim 12.

14., wherein after determining LSB, from possible constellation points, determine MSB by checking LSB according to the method for claim 13.

15. according to the method for claim 11, wherein in conjunction with a LSM and MSB decoder to be used for different conformations.

16. one kind is used for decoder device that the turbo trellis code modulation signal that receives is decoded, comprise: a pair of decoder that is connected to receive soft input bit, one interleaver, one deinterleaver, and described decoder is derived a data bit and at least one check bit from described soft input bit.

17. according to the decoder device of claim 16, wherein each decoder is carried out forward direction and the calculating of back to iteration and final soft bit output.

18. according to the decoder device of claim 16, wherein said decoder is the LOG-MAP decoder.

Images