CN1294706C - Iterative demodulating-decoding method and apparatus for code modulation system having output external information - Google Patents

Abstract

The present invention relates to an iterative demodulation and decoding method and an iterative demodulation and decoding device for an encoding and modulation system with external information output. A starting phase estimation module estimates the starting value of a phase, and an M2M4 estimator obtain the coarse estimation of amplitude and noise variance at the same time; a phase and frequency deviation tracing module is started by the coarse estimation and the starting phase; the tracing module outputs a tracing phase to a signal corrector of a communication channel; the signal corrector of a communication channel outputs signal to an amplitude and noise variance estimation module, and starts the amplitude and noise variance estimation module to obtain the more accurate estimation of amplitude and noise variance; the output of the signal corrector of a communication channel and the accurate estimation result of amplitude and noise variance are sent to an iterative decoder to complete a primary iterative demodulation and decoding process; iterative decoding external information can start a phase and frequency deviation estimation module and the amplitude and noise variance estimation module to carry out iterative demodulation and decoding of a next time. The iterative demodulation and decoding problems of Turbo/LdPC coding BPSK/QPSK modulation system in an unknown channel of a frequency deviation and a phase are solved, and the present invention can be widely used in the technical field of a wireless and satellite communication technology.

Description

Iterative demodulation coding/decoding method and device for code modulation system with output external information

Technical field: the present invention relates to wireless and technical field of satellite communication, relate in particular to a kind of for have output external information code modulation system iterative demodulation coding/decoding method and device.

Background technology:

In the wireless and satellite communication system, transmitted bit is subjected to the influence of channel random noise and produces random error.Theory and practice proves that providing the error correction/encoding method of transmission reliability by the introducing redundancy is the effective means of a class.And one of Turbo code of introducing in recent years and LDPC sign indicating number encoding scheme that to be the error correcting capability found so far the strongest.

Turbo code equals proposition [C.Berrou in 1993 by C.Berrou, A.Glavieux, and P.Thitimajshima, " Near Shannon limit error-correcting coding and decoding:Turbo-codes, " in ICC ' 93, Geneva, Swithzerland, May, 1993, PP.1064-1070], it is considered to one of maximum progress that coding theory in recent years obtains.Turbo code is 10 in the error rate under white Gaussian noise (Additive White Gaussian Noise-AWGN) channel ^-5The time approach the channel capacity limit with the signal to noise ratio of 0.7dB.LDPC (Low-Density ParityCheck) sign indicating number was at first proposed in 1961 by Gallager, was rediscovered by Mackay etc. but forgotten until 1996 by people for a long time.Well-designed LDPC long code performance even be better than Turbo code.For satellite channel, because power limited, the work signal to noise ratio that reduces satellite communication system is vital beyond doubt, introduces the shortcoming that Turbo (or LDPC) encoding and decoding technique can solve the satellite channel power limited well.

The structure of typical Turbo-code as shown in Figure 1.It is made of two identical recursive system convolution (RSC) (being commonly referred to subcode) codings of structure usually, and RSC1 directly encodes to the information sequence that enters, and gets verification sequence y _1kSimultaneously, with information sequence d _kSequence d after interweaving by interleaver _nBe sent to RSC2 and encode, obtain check digit y _2k, the Turbo-code code word constitutes by connecing the two-way verification sequence behind the information sequence exactly.Check digit (the y that sub-encoders produced _1k, y _2k) can delete through difference and cut matrix and delete the Turbo code that obtains different code checks after getting.

Turbo-code iterative decoding structure as shown in Figure 2, it mainly softly goes into softly to go out module (sub-decoder of Turbo-code) and form by two, sub-decoder is used for the RSC subcode among the selected Turbo-code is deciphered.The information bit d that sub-decoder 1 obtains sub-decoder 2 _kExternal information

As d _kPrior information comes RSC1 is deciphered, and obtains about d _kImproved external information Λ _1e(d _k), after interweaving, obtain Prior information as 2 pairs of RSC2 decodings of sub-decoder.Sub-decoder 2 usefulness and sub-decoder 1 same method produces the improved external information Λ of information bit once more _2e(d _j), after deinterleaving, obtain

As the soft value of the priori of sub-decoder in the next iteration 1.Like this after iteration repeatedly, the output Λ that sub-decoder 2 is produced ₂(d _j) carry out hard decision after deinterleaving, obtain each information bit d _kValuation

Relevant knowledge about the LDPC encoding and decoding sees [T.J.Richardson and R ü diger L.Urbanke for details, " TheCapacity of Low-Density Parity-Check Codes Under Message-Passing Decoding; " IEEETrans.Inform.Theory, vol.47, Feb.2001, pp.599-618], the LDPC decoding adopts confidence level to relay algorithm, carry out iterative decoding based on the bipartite graph structure, and output class is similar to the external information of Turbo decoding.

Present most of Turbo (or LDPC) coding/decoding method has only been considered white Gaussian noise (AWGN) channel, and this is not enough in actual wireless and satellite communication.Under actual channel, after generally need modulating, coded digital signal sends again, and this just requires transmitting-receiving two-end that identical frequency generator is arranged.But because the non-ideal characteristic of practical devices, particularly transmission medium the time become characteristic such as decline, the frequency deviation and the phase place of the modulation signal of feasible transmitting-receiving are unknown in fact to a great extent, and this just requires receiver to carry out channel estimating and follows the tracks of to eliminate unknown frequency deviation and phase place to sending the influence of signal.In addition, the Turbo/LDPC decoding also needs to know received signal amplitude and noise variance.Like this, thus the Turbo/LDPC sign indicating number is decoded is the research focus that paid close attention to by people to various parameters how to estimate Unknown Channel.The difficulty of this problem is: the signal to noise ratio of Turbo/LDPC sign indicating number work will be well below the signal to noise ratio under the normal condition, and in fact how accurately to estimate and to follow the tracks of channel parameter itself to be exactly a great problem under low signal-to-noise ratio.

In satellite and radio communication, typical channel unknown parameter is frequency shift (FS) (frequency deviation) and carrier phase.Traditional channel estimation technique is to adopt PHASE-LOCKED LOOP PLL TECHNIQUE.Because under the utmost point low signal-to-noise ratio (as Turbo code under the AWGN workable signal to noise ratio) the particularity of communication, traditional algorithm does not re-use.Possible solution is a channel estimating and channel decoding characteristic and replace the decoding of associating iterative demodulation independently separately before abandoning, and makes channel estimating that the redundancy of encoding and providing can be provided.

At patent application (application number: 03137079.9), we are based on [I.Bar-David, and A.Elia, " AugmentedAPP (A2P2) Module for a Posteriori Probability Calculation and Channel ParameterTracking, " IEEE Commun.Lett., vol.3, no.1, Jan., 1999, pp.18-20] a kind of iterative demodulation coding/decoding method and device of Turbo coding BPSK modulating system proposed.Its characteristic feature is that channel parameter is followed the tracks of the network utilized the Turbo subcode, and parameter tracking and decoding algorithm must be united and carried out.The shortcoming of this method is that the scope of application is little, promptly only is applicable to the BPSK modulating system that adopts the convolution subcode.Thereby be not suitable for following two kinds of important condition:

(1) Turbo coding QPSK modulating system

(2) the BPSK/QPSK modulating system of LDPC coding.

In addition, in this patent application, do not relate to the estimation of amplitude and noise variance, and estimate that exactly amplitude and noise variance also are very important concerning the Turbo/LDPC decoding.

Summary of the invention:

The purpose of this invention is to provide a kind of Method and circuits device that the simple and practical iterative demodulation of Turbo/LDPC coding BPSK/QPSK communication system is decoded under unknown parameters (inherent spurious frequency deviation of channel, phase place, amplitude, the noise variance the unknown) channel that is used for.

Iterative demodulation coding/decoding method for the code modulation system with output external information of the present invention, its step comprises

1) signal encoded and modulation that receives being carried out initial phase estimates;

2) adopt the M2M4 algorithm for estimating that amplitude and noise variance are carried out rough estimate;

3) adopt non-data auxiliary phase/frequency offset tracking algorithm to carry out phase place and frequency offset tracking, obtain the real-time phase pursuit gain according to above-mentioned initial phase estimated value and above-mentioned amplitude and noise variance rough estimate evaluation;

4) carry out channel signal according to above-mentioned real-time phase pursuit gain and proofread and correct output equivalent AWGN signal;

5), adopt iteration amplitude search algorithm to carry out amplitude and the smart estimation of noise variance according to above-mentioned Equivalent A WGN signal;

6) restarting non-data auxiliary phase/frequency offset tracking algorithm according to above-mentioned initial phase estimated value and above-mentioned amplitude and the smart estimated value of noise variance carries out phase place and frequency offset tracking, obtains the real-time phase pursuit gain;

7) carry out channel signal according to above-mentioned real-time phase pursuit gain and proofread and correct output equivalent AWGN signal;

8) decoder receives the smart estimated value of above-mentioned Equivalent A WGN signal and above-mentioned amplitude and noise variance, finishes an iterative decoding job, and the output external information;

9), start once more according to above-mentioned Equivalent A WGN signal and above-mentioned external information that iteration amplitude search algorithm carries out amplitude and noise variance is smart estimates in the amplitude that does not reach setting as yet with noise variance is smart when estimating number of times;

10) according to above-mentioned initial phase estimated value, smart estimated value of above-mentioned amplitude and noise variance and above-mentioned external information start non-data auxiliary phase/frequency offset tracking algorithm and carry out phase place and frequency offset tracking, obtain the real-time phase pursuit gain;

11) carry out signal correction according to above-mentioned real-time phase pursuit gain, output equivalent AWGN signal;

12) decoder receives above-mentioned Equivalent A WGN signal, and finishes another iterative decoding work according to above-mentioned amplitude and the smart estimated value of noise variance, and the output external information;

13) when iterations reaches predetermined times, the output decoder value, otherwise repeat 9)-12).

Described 1) also can carry out initial frequency deviation to the signal encoded and modulation that receives estimates; Can pass through fft algorithm [D.Taich and I.Bar-David, " Maximum-likelihood estimation of phase and frequency ofMPSK signals; " IEEE Trans.Inform.Theory, vol.45, Nov.1999, pp.2652-2655] carry out initial frequency deviation and estimate.

Described TURBO sign indicating number or the LDPC sign indicating number of being encoded to; Describedly be modulated to BPSK modulation or QPSK modulation.

Iterative demodulation decoding device for code modulation system of the present invention with output external information, comprise iterative decoder, the initial phase estimation module, the M2M4 estimator, the phase tracking module, channel signal adjuster, amplitude and Noise Variance Estimation module, initial phase estimation module are used for that the modulation signals that receives is carried out initial phase and estimate; The M2M4 estimator is used for the amplitude of carrying out to received signal and noise variance rough estimate; The phase tracking module receives the output of initial phase estimation module and the output of M2M4 estimator and carries out phase place and frequency offset tracking to received signal; The channel signal adjuster, the output of receiving phase/frequency-tracking module is proofreaied and correct signal; Amplitude and Noise Variance Estimation module, the output of receive channel signal calibrator are carried out amplitude and the smart estimation of noise variance, and output to iterative decoder and phase tracking module; Iterative decoder is accepted the output of channel signal calibrator and amplitude and Noise Variance Estimation module, carries out iterative decoding; It is characterized in that iterative decoder is also to amplitude and Noise Variance Estimation module and phase tracking module output external information.

Iterative demodulation decoding device for code modulation system of the present invention with output external information, its iterative decoder can be located on the fpga chip; The initial phase estimation module, the M2M4 estimator, the phase tracking module, amplitude and Noise Variance Estimation module and channel signal adjuster are located in the dsp chip; Each several part carries out data preservation and transmission by external RAM.

The initial phase estimation module also comprises and can carry out the submodule that initial frequency deviation is estimated to signal.Specifically can carry out initial frequency deviation by fft algorithm estimates.

Described TURBO coding or the LDPC coding of being encoded to; Described BPSK modulation or the QPSK of being modulated to.

The invention provides the iterative demodulation decoding algorithm of the unknown Turbo/LDPC coding down of channel parameter (channel phase, signal amplitude, noise variance, inherent spurious frequency deviation) BPSK/QPSK modulation.Based on the joint maximum likelihood estimation criterion, proposed the maximal possibility estimation criterion of channel parameter (channel phase, signal amplitude, noise variance, inherent spurious frequency deviation), and proposed the track algorithm of channel phase/frequency deviation in view of the above.Further according to the characteristics of iterative decoding, proposed a kind of channel parameter maximal possibility estimation algorithm that can make full use of the iteration external information, and obtained corresponding channel phase/frequency offset tracking algorithm in view of the above on this basis.If the channel parameter estimation and the track algorithm that utilize the present invention to propose carry out demodulation (" demodulation " speech in fact is the meaning of parameter correction) to received signal here, then the signal after the demodulation can be regarded the awgn channel of known amplitude and noise variance as, thereby can adopt the standard iterative decoding algorithm (can adopt Fig. 2 algorithm to Turbo code) under the awgn channel effectively to decode.

The overall plan of the iterative demodulation decoding algorithm of the unknown Turbo/LDPC coding down of channel parameter (channel phase, signal amplitude, noise variance, inherent spurious frequency deviation) BPSK/QPSK system is seen Fig. 3.As shown in the figure, message bit stream elder generation is through the Turbo/LDPC encoder, and modulation sends on the channel coded-bit through BPSK/QPSK again, supposes the Gray coding that QPSK modulation employing is commonly used here.Suppose channel phase, channel magnitude, noise variance, inherent spurious frequency deviation the unknown, under the bit synchronous hypothesis of ideal, the equivalent low pass complex signal that receives can be expressed as

$r_k={\mathrm{As}}_k{e}^{j(\mathrm{\θ}+\mathrm{k\ω})}+w_k---\left(1\right)$

Wherein, A represents received signal amplitude, s _kBPSK (perhaps QPSK) signal that expression sends, its value in set of signals+1 ,-1} (perhaps { e ^{J π/4},-e ^{J π/4}, e ^{-j π/4},-e ^{J π/4}), θ is the channel unknown phase, w _kThe expression average is that 0 variance is 2 σ ²White complex gaussian noise.Below BPSK and QPSK modulation are discussed respectively, for the sake of simplicity, the situation of no external information is discussed earlier, and then is discussed in Turbo/LDPC iterative decoding external information to assist how to estimate channel parameter down better.

One, the maximum likelihood parameter Estimation under the no external information

The Turbo/LDPC iterative decoding just belongs to this situation before the iteration first.

Suppose the BPSK modulation, then at known A, Δ ω, received signal r under the θ _kProbability density can be expressed as respectively

$p\left(r_k\right|A,\mathrm{\Δ\ω},\mathrm{\θ})=\frac{1}{2}[p\left(r_k\right|s_k=+1,A,\mathrm{\Δ\ω},\mathrm{\θ})+p\left(r_k\right|s_k=-1,A,\mathrm{\Δ\ω},\mathrm{\θ})]$

$={\left({2\mathrm{\π\σ}}^2\right)}^{-1}\exp (-\frac{1}{{2\mathrm{\σ}}^2}({\left|r_k\right|}^2+A^2))\cosh \left(\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)$ (2)

Re[wherein] and Im[] respectively expression get real part and imaginary part, the surplus function that revolves of cosh () expression hyperbolic.

Suppose that the channel unknown parameter remains unchanged, and uses r in N continuous sample ₁ ^NExpression is from the received signal sample in the moment 1 to N.Then Dui Ying conditional probability density is

$p\left({r_1}^N\right|A,\mathrm{\Δ\ω},\mathrm{\θ})=\underset{k=1}{\overset{N}{\mathrm{\Π}}}p\left(r_k\right|A,\mathrm{\Δ\ω},\mathrm{\θ})$

$={\left({2\mathrm{\π\σ}}^2\right)}^{-N}\exp (-\frac{1}{{2\mathrm{\σ}}^2}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}({\left|r_k\right|}^2+A^2))\underset{k=1}{\overset{N}{\mathrm{\Π}}}\cosh \left(\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)$ (3)

Get $\∂p\left({r_1}^N\right|A,\mathrm{\Δ\ω},\mathrm{\θ})/\∂A=0,\∂p\left({r_1}^N\right|A,\mathrm{\Δ\ω},\mathrm{\θ})/\∂\mathrm{\θ}=0,\∂P\left({r_1}^N\right|A,\mathrm{\Δ\ω},\mathrm{\θ})/\∂\mathrm{\Δ\ω}=0,$

Can get: unknown parameter A, Δ ω, the maximal possibility estimation of θ separate uniting of following equation

$A=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)\mathrm{Re}\left[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\right];---\left(4\right)$

$0=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)\mathrm{Im}\left[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\right],---\left(5\right)$

$0=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}k\·\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)\mathrm{Im}\left[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\right].---\left(6\right)$

Consider the QPSK modulation below, correspondingly use following conditional probability density

$p\left(r_k\right|A,\mathrm{\Δ\ω},\mathrm{\θ})=\frac{1}{4}[p\left(r_k\right|s_k=e^{\mathrm{j\π}/4},A,\mathrm{\Δ\ω},\mathrm{\θ})+p\left(r_k\right|s_k={-e}^{\mathrm{j\π}/4},A,\mathrm{\Δ\ω},\mathrm{\θ})]$

$+\frac{1}{4}[p\left(r_k\right|s_k=e^{-\mathrm{j\π}/4},A,\mathrm{\Δ\ω},\mathrm{\θ})+p\left(r_k\right|s_k={-e}^{-\mathrm{j\π}/4},A,\mathrm{\Δ\ω},\mathrm{\θ})]$

$={\left({4\mathrm{\π\σ}}^2\right)}^{-1}\exp (-\frac{1}{{2\mathrm{\σ}}^2}({\left|r_k\right|}^2+A^2))[\cosh \left(\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{\π}/4+\mathrm{k\Δ\ω})}\left]\right)+\cosh \left(\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}-\mathrm{\π}/4+\mathrm{k\Δ\ω})}\left]\right)$

$={\left({2\mathrm{\π\σ}}^2\right)}^{-1}\exp (-\frac{1}{{2\mathrm{\σ}}^2}({\left|r_k\right|}^2+A^2))\cosh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)\cosh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)---\left(7\right)$

Be similar to (3), can obtain

$\ln p\left({r_1}^N\right|A,\mathrm{\Δ\ω},\mathrm{\θ})=\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\ln p\left(r_k\right|A,\mathrm{\Δ\ω},\mathrm{\θ})$

$\≅-\frac{{\mathrm{NA}}^2}{{2\mathrm{\σ}}^2}+\underset{k=1}{\overset{N}{\mathrm{\Σ}}}(\ln \cosh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)+\ln \cosh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}\right[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right))$ (8)

Get $\∂p\left({r_1}^N\right|A,\mathrm{\Δ\ω},\mathrm{\θ})/\∂A=0,\∂p\left({r_1}^N\right|A,\mathrm{\Δ\ω},\mathrm{\θ})/\∂\mathrm{\θ}=0,\∂P\left({r_1}^N\right|A,\mathrm{\Δ\ω},\mathrm{\θ})/\∂\mathrm{\Δ\ω}=0,$

Can get: unknown parameter A, Δ ω, the maximal possibility estimation of θ separate uniting of following equation

$A=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\left\{\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}\right[r_k{e}^{-j{\mathrm{\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Re}\right[r_k{e}^{{-\mathrm{j\θ}}_k}]+\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}\right[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Im}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right\};---\left(9\right)$

$0=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\left\{\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}\right[r_k{e}^{-j{\mathrm{\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Im}\right[r_k{e}^{{-\mathrm{j\θ}}_k}]-\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}\right[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Re}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right\};---\left(10\right)$

$0=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}k\left\{\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}\right[r_k{e}^{-j{\mathrm{\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Im}\right[r_k{e}^{{-\mathrm{j\θ}}_k}]-\tanh \left(\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}\right[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Re}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right\},---\left(11\right)$

Wherein, θ _k=θ+k Δ ω.

Two, the auxiliary maximum likelihood parameter Estimation of iterative decoding external information

Turbo/LDPC iterative decoding (after reaching) iteration for the second time just belongs to this situation.

Providing the external information of each coded-bit if Turbo/LDPC decodes behind the L time iterative decoding is Le ^(l)(d _k), this external information is being used for the L+1 time parameter Estimation and tracking and corresponding channel parameter correction before the iteration as the prior information of parameter Estimation and tracking module again as shown in Figure 3.Auxiliary down in external information, coded-bit such as no longer obeys at general the distribution, and the BPSK system is had:

$\mathrm{Pr}(d_k=+1)=\frac{1}{1+\exp \left({-\mathrm{Le}}^{\left(l\right)}\left(d_k\right)\right)}={[\exp ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2)+\exp ({-\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2)]}^{-1}\exp ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2)\∝\exp ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2)$

$\mathrm{Pr}(d_k=-1)=\frac{1}{1+\exp \left(L{e}^{\left(l\right)}\left(d_k\right)\right)}={[\exp ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2)+\exp ({-\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2)]}^{-1}\exp ({-\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2)\∝\exp ({-\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2)---\left(12\right)$

This unequal probability distributed include the derivation of (2-6) in, can the auxiliary unknown parameter A down of iterative decoding external information arranged, Δ ω, the maximal possibility estimation of θ separate uniting of following equation

$A=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2+\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)\mathrm{Re}\left[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\right];---\left(13\right)$

$0=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2+\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)\mathrm{Im}\left[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\right];---\left(14\right)$

$0=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}k\·\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2+\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\left]\right)\mathrm{Im}\left[r_k{e}^{-j(\mathrm{\θ}+\mathrm{k\Δ\ω})}\right].---\left(15\right)$

For the QPSK modulation, suppose mapping relations (d _2k, d _2k-1) be:

(+1，+1)π/4；(+1，-1)-π/4；(-1，+1)π-π/4；(-1，-1)π+π/4。

That considers the auxiliary prior probability that brings of external information does not wait general distribution character, and the derivation that repeats (7-11) can get: unknown parameter A, Δ ω, the maximal possibility estimation of θ separate uniting of following equation

$A=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\left\{\begin{array}{c}\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k}\right)/2+\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Re}\left[r_k{e}^{{-\mathrm{j\θ}}_k}\right]\\ +\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k-1}\right)/2+\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Im}\left[r_k{e}^{{-\mathrm{j\θ}}_k}\right]\end{array}\right\};---\left(16\right)$

$0=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\left\{\begin{array}{c}\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k}\right)/2+\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Im}\left[r_k{e}^{{-\mathrm{j\θ}}_k}\right]\\ -\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k-1}\right)/2+\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Re}\left[r_k{e}^{{-\mathrm{j\θ}}_k}\right]\end{array}\right\};---\left(17\right)$

$0=\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}k\left\{\begin{array}{c}\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k}\right)/2+\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Im}\left[r_k{e}^{{-\mathrm{j\θ}}_k}\right]\\ -\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k-1}\right)/2+\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Re}\left[r_k{e}^{{-\mathrm{j\θ}}_k}\right]\end{array}\right\},---\left(18\right)$

By (16-18) as can be seen, external information is set as 0, can obtains not having the maximal possibility estimation (9-11) of external information situation.Modulation has above conclusion equally to BPSK.Thereby following maximal possibility estimation that needs discussion to have external information to assist.But the joint maximum likelihood parameter Estimation of BPSK (13-15) or QPSK (16-18) no matter, it is very complicated finding the solution equation discussed above, is impossible in realization substantially.The solution that adopts the present invention to propose has solved above problem well.

The visible Fig. 3 of receiver block diagram of the present invention.Here we always suppose that inherent spurious frequency deviation (with respect to character rate normalization) is smaller, representative value is 100ppm (ppm represent 1,000,000/), can estimate by fft algorithm big frequency deviation, but referenced patent application 03137079.9 and document [D.Taich and I.Bar-David, " Maximum-likelihood estimation of phase and frequency of MPSK signals, " IEEETrans.Inform.Theory, vol.45, Nov.1999, pp.2652-2655].Realize for reducing that complexity, this programme have adopted and lead sign indicating number and carry out initial phase and estimate.Because inherent spurious frequency deviation is less, leading a yard number (being assumed to be P) not quite to the estimated accuracy influence after a little while.

1, initial phase is estimated

Condition: add a small amount of frequency pilot sign.

Supposing the system sends P continuously and leads sign indicating number (Pilot Symbols) before the beginning of every frame, for simplicity, can think all-ones.The signal that then receives is

${\tilde{r}}_v=e^{i(\mathrm{v\ω}+\mathrm{\θ})}+w_v,v=\mathrm{1,2},\mathrm{\Λ},P$

Then initial phase is estimated as

${\mathrm{\θ}}_0=\mathrm{angle}\left(\underset{v=1}{\overset{P}{\mathrm{\Σ}}}{\tilde{r}}_v\right)---\left(19\right)$

Here the phase angle of plural number is asked in angle (x)=arctan (Im (x)/Re (x)) expression.

In actual channel, channel phase may be slowly to change in a frame, and this just need follow the tracks of.Careful investigate (14) and (17) as can be known, when the true phase place of phase deviation of estimation, function (the right of (14) or (17) equation) is the S curve-like with departing from phase place, and discovers that this curve is to amplitude Estimation A and Noise Variance Estimation σ ²Insensitive.Thereby the present invention adopts following non-data auxiliary phase/frequency offset tracking algorithm.

1. non-data auxiliary phase/frequency offset tracking algorithm

(1) BPSK situation

${\mathrm{\θ}}_{k+1}={\mathrm{\θ}}_k+{\mathrm{\ω}}_k+\mathrm{\α}\left\{\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2+\frac{A}{{\mathrm{\σ}}^2}\mathrm{Re}[r_k{e}^{-j{\mathrm{\θ}}_k}\left]\right)\mathrm{Im}\right[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right\},---\left(20\right)$

ω _k+1＝ω _k+β{θ _k+1-θ _k}。(21)

(2) QPSK situation

${\mathrm{\θ}}_{k+1}={\mathrm{\θ}}_k+{\mathrm{\ω}}_k+\mathrm{\α}\left\{\begin{array}{c}\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k}\right)/2+\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Im}\left[r_k{e}^{{-\mathrm{j\θ}}_k}\right]\\ \tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k-1}\right)/2\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}[r_k{e}^{{-\mathrm{j\θ}}_k}\left]\right)\frac{1}{\sqrt{2}}\mathrm{Re}\left[r_k{e}^{{-\mathrm{j\θ}}_k}\right]\end{array}\right\},---\left(22\right)$

ω _k+1＝ω _k+β{θ _k+1-θ _k}。(23)

Wherein, α, β generally can determine its numerical value by emulation for following the tracks of coefficient.For iteration first,, can directly corresponding external information be set as 0, θ owing to there is not external information ₀Initial inherent spurious frequency deviation ω is provided by leading sign indicating number ₀=0.By (20,22) as can be known, this track algorithm also needs to know parameter A/σ ², the estimation of this parameter is discussed below.

2. amplitude and Noise Variance Estimation

For the purpose of receiving conveniently, receiver carries out energy normalized at first to received signal and handles promptly feasible:

$1/N{\mathrm{\Σ}}_{k=1}^N{\left|r_k\right|}^2=1.$

Amplitude and Noise Variance Estimation are discussed below on this basis.Before iteration first,, thereby adopt a kind of amplitude that does not rely on channel phase and frequency deviation and Noise Variance Estimation shortcut calculation greatly because phase place/frequency offset tracking discussed above needs amplitude and Noise Variance Estimation.And because phase place/frequency offset tracking algorithm that the present invention adopts is insensitive to amplitude and Noise Variance Estimation, thereby it is not high to the required precision of this estimation, adopt [D.R.Pauluzzi and N.C.Beaulieu below, " A Comparison of SNR Estimation Techniques for the AWGN Channel; " IEEETrans.Commun., vol.COM-16, Oct.2000, pp.1681-1691] the so-called M2M4 algorithm for estimating that adopts in the literary composition:

$M2=\frac{1}{L}\underset{k=1}{\overset{L}{\mathrm{\Σ}}}{\left|r_k\right|}^2,M4=\frac{1}{L}\underset{k=1}{\overset{L}{\mathrm{\Σ}}}{\left|r_k\right|}^4.---\left(24\right)$

$A_0={({2M}_2^2-M_4)}^{\frac{1}{4}},---\left(25\right)$

${\mathrm{\σ}}_0^2=(M_2-{A_0}^2)/2.---\left(26\right)$

Like this, the above rough estimate of utilization can excute phase/frequency offset estimating and track algorithm.Be further to improve systematic function, also can utilize iterative algorithm that equation (13,16) is found the solution accurate estimation in the hope of amplitude and noise variance.As shown in Figure 3, calculate the rough estimate (A of amplitude and noise variance by M2M4 method of estimation (25,26) ₀, σ ₀ ²) can be effective to the tracking of channel phase/frequency deviation, follow the tracks of good parameter and send and lead the channel signal adjuster and carry out the phase operation of writing to: promptly ${\tilde{r}}_k=r_k\exp \left({-\mathrm{j\θ}}_k\right)),$ Real-time phase θ wherein _kObtain by phase place/frequency offset tracking module.For obtaining more accurate amplitude/Noise Variance Estimation (so that decoder can better be worked), can adopt following step to make iterative computation.

Amplitude/Noise Variance Estimation iterative search algorithm flow (abbreviates iteration amplitude search algorithm as: AmVarEstimProg (A _Min, A _Max, I)):

Minimum and maximum possible value A that the step 1) selecting range is estimated _MinAnd A _Max, putting maximum iteration time is I, and puts primary iteration sequence number i=0.

Step 2) calculates A _m=(A ₁+ A ₂)/2 reach ${\mathrm{\σ}}_m^2=(1-A_m^2)/2.$

Step 3) is calculated $\mathrm{BPSK}:f\left(A_m\right)=A_m-\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_k\right)/2+\frac{A_m}{{\mathrm{\σ}}_m^2}\mathrm{Re}[{\tilde{r}}_k\left]\right)\mathrm{Re}\left[{\tilde{r}}_k\right];$

$\mathrm{QPSK}:f\left(A_m\right)=A_m-\frac{1}{N}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}\left\{\begin{array}{c}\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k}\right)/2+\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Re}[{\tilde{r}}_k\left]\right)\frac{1}{\sqrt{2}}\mathrm{Re}\left[{\tilde{r}}_k\right]\\ +\tanh ({\mathrm{Le}}^{\left(l\right)}\left(d_{2k-1}\right)/2\frac{A}{{\mathrm{\σ}}^2}\frac{1}{\sqrt{2}}\mathrm{Im}[{\tilde{r}}_k\left]\right)\frac{1}{\sqrt{2}}\mathrm{Im}\left[{\tilde{r}}_k\right]\end{array}\right\}.$

If step 4) f (A _mA is then upgraded in)＞0 ₂=A _m. otherwise upgrade A ₁=A _m

If step 5) is put i=i+1. i=I, then export A _mAs final amplitude Estimation and output ${\mathrm{\σ}}_m^2=(1-A_m^2)/2$

As final Noise Variance Estimation. otherwise forward step 2 to).

Above amplitude/noise variance iteration estimator can move before each iteration (Turbo or LDPC) decoding, can make full use of the external information that decoding provides like this.If can increase the complexity of algorithm but all move this amplitude/noise variance iteration estimator when decoding iteration at every turn.When realizing, reality can choose the opportunity and the number of times of operation this amplitude/noise variance iteration estimator flexibly.In the example of back, only before iterative decoding, moved once this module, and final decoding performance can be satisfied with.

Like this, utilize above-described amplitude/noise variance iteration estimator can accurately estimate amplitude and noise variance, this estimated result is passed to channel phase/frequency offset tracking module again, thereby obtain better phase estimation, final Phase Tracking result delivers to the channel signal adjuster, and the Equivalent A WGN signal after the correction can softly go into softly to go out decoding algorithm and carry out iterative decoding with existing.

The present invention is based on the auxiliary joint maximum likelihood parameter Estimation down of free of data and proposed a kind of phase place/frequency offset tracking module that BPSK/QPSK modulates, its characteristic feature is to make full use of the external information that iterative decoding provides, and insensitive to amplitude and variance evaluated error.The present invention proposes the amplitude and the noise variance iteration estimation module of BPSK/QPSK modulating system, its characteristic feature is to make full use of the external information that iterative decoding provides, and can select number of applications according to concrete applying flexible, thereby it is compromise to reach good performance/complexity.In addition, the present invention has adopted a kind of simple amplitude and noise variance initial estimation module, this module can be worked under channel phase and the frequency deviation not knowing, amplitude and noise variance rough estimate that using this module provides can be used to provide start-up parameter to phase place/frequency offset tracking module.

The present invention has mainly solved the iterative demodulation decoding problem of coding BPSK/QPSK modulating system under frequency deviation and the phase place Unknown Channel, and especially, it is applicable to Turbo/LDPC coding BPSK/QPSK modulating system.

The used concrete channel coding method of phase place/frequency offset tracking module that the present invention proposes and system has nothing to do, thereby widely applicable.Applicable to the good sign indicating number of typical cases such as convolution code, Turbo code, LDPC sign indicating number.If decoding can provide external information, then this phase place/frequency offset tracking module can make full use of this external information raising performance.In addition, phase place of the present invention/frequency offset tracking module does not need to know modulation intelligence (free of data is auxiliary), has improved channel utilization greatly.

Channel phase of the present invention/frequency offset tracking module has very strong parameter tracking ability, thereby has loosened the requirement to initial channel parameter Estimation precision, and available existing DSP technology or FPGA technology hardware are realized.

The amplitude that the present invention proposes/Noise Variance Estimation module adopts the iterative search algorithm to find the solution the maximal possibility estimation equation, realizes simple, advantage of high precision thereby have.Amplitude of the present invention/Noise Variance Estimation module does not need to know modulation intelligence (free of data is auxiliary), has improved channel utilization.

The present invention utilizes a spot of frequency pilot sign, has provided the initial estimation algorithm of channel phase.The estimated accuracy of parameter generally can guarantee that phase place of the present invention/frequency offset tracking module, amplitude and noise variance module can be at energy reliably workings on the typical satellite channel.

Advantage of the present invention and good effect are summarized as follows:

1. the characteristics of iterative demodulation decoding scheme provided by the invention are that the demodulation sign indicating number is independent substantially, and both carry out information interaction by external information and improve constantly performance;

2. scheme provided by the invention is simple and practical, is easy to Digital Implementation;

3. the used concrete channel coding method of phase place provided by the invention/frequency offset tracking module and system has nothing to do, thereby widely applicable, applicable to the good sign indicating number of typical cases such as convolution code, Turbo code, LDPC sign indicating number;

4. phase place provided by the invention/frequency offset tracking module can provide in decoding and can make full use of this external information under the condition of external information and improve performance;

5. phase place provided by the invention/frequency offset tracking module does not need to know modulation intelligence (free of data is auxiliary), has improved channel utilization greatly;

6. phase place provided by the invention/frequency offset tracking module has very strong parameter tracking ability, thereby has loosened the requirement to initial channel parameter Estimation precision;

7. phase place provided by the invention/frequency offset tracking module is insensitive to amplitude and variance evaluated error, and the amplitude and the Noise Variance Estimation that adopt simple M2M4 method to provide can start this module operation well;

8. amplitude provided by the invention/Noise Variance Estimation module adopts iterative algorithm to find the solution the maximal possibility estimation equation, realizes simple, advantage of high precision thereby have;

9. amplitude provided by the invention/Noise Variance Estimation module does not need to know modulation intelligence (free of data is auxiliary), has improved channel utilization greatly.

The present invention is applied to typical permanent ginseng satellite channel, the performance that can obtain to approach Turbo code under the desirable awgn channel.

Description of drawings:

Fig. 1 Turbo-code coder structure;

Fig. 2 Turbo-code iterative decoder structure;

The iterative decoding algorithm block diagram of Turbo/LDPC coding BPSK/QPSK modulation under Fig. 3 frequency deviation and phase place the unknown;

The DSP+FPGA development structure of Fig. 4 example;

The rsc encoder that Fig. 5 example is used;

The BPSK receiver system performance of Fig. 6 example;

Abscissa is bit signal to noise ratio Eb/N0 among the figure, ordinate is bit error rate (BER), 4 from top to bottom expressions respectively of curve among the figure: algorithm the 4th iteration of the present invention, desirable coherent reception the 4th iteration, the 12nd iteration of algorithm of the present invention, the performance behind the 12nd iterative decoding of desirable coherent reception;

The QPSK receiver system performance of Fig. 7 example;

Abscissa is bit signal to noise ratio Eb/N0 among the figure, and ordinate is bit error rate (BER), 4 from top to bottom expressions respectively of curve among the figure: the performance behind the 1st, 4,8,12 iterative decodings of algorithm of the present invention;

Embodiment:

Here based on the CPLD chip EP20K400EBC652-2X of Alter company, be given in the developing example of the demodulating and decoding algorithm that adopts the Turbo code of advising in the 3-G (Generation Three mobile communication system) IMT-2000 high-rate service under typical frequency deviation and the phase place Unknown Channel.The FPGA+DSP implementation structure of this example is seen Fig. 4, and the realization master clock is 32MHz, and the decode rate of permission is 115.2kbps.System has 2 external RAMs and a dsp chip (TMSC5402) in addition, and Turbo decoder module hardware in the CPLD chip is realized, and all software implementation realizations in DSP of other modules of receiver shown in Figure 3.Fig. 4 is made further specify before, we provide the specific coding parameter of the Turbo code of selecting for use in the example earlier.

This example is selected the chnnel coding of 3-G (Generation Three mobile communication system) IMT-2000 high-rate service for use.Select for use the Turbo code in the suggestion to be made of two identical recursive convolution subcodes, the generator polynomial of recursive convolution subcode is (13,15,17) 8, information frame length 2280, weaving length 2298, length overall 2304, code rate 1/2.By IMT2000 suggestion as can be known, this 1/2 speed Turbo code be to get through cascading delete by the female sign indicating number of two RSC.Female sign indicating number that this example is selected for use as shown in Figure 5.

The code check adjustment is to adjust code check by deleting some check bits in the Turbo code encoder.In table 1, " 1 " expression output, " 0 " expression deletion; (X, Y ₀, Y ₁) represent the output of first rsc encoder, wherein X is an information bit, Y ₀And Y ₁Be check digit, (X ', Y ' ₀, Y ' ₁) expression second rsc encoder output, X ' is an information bit, Y ' ₀And Y ' ₁It is check digit.

Note, for the Turbo code of 1/2 speed, Y1 and Y ' 1 bit vacant need not, in the following narration of this example, be not counted in the category of deletion.The deletion situation (p=0 represents that this bit is deleted, and it is deleted that p=1 represents that this bit does not have) that deletion indication p only is used for refering in particular to Y0 and Y ' 0 bit.

The code check adjustment of table 1 non-return-to-zero bit

	Code check R
				Output
	1/2	1/3	1/4
				X	11	11	11

Y 0	10	11	11
				Y 1	00	00	10
X	00	00	00
				Y 0	01	11	01
Y 1	00	00	11

Annotate:, should from left to right read this table from top to bottom every kind of code check

The hypothesis channel is typical satellite constant-parameter channel in this example, and initial normalization frequency deviation (with respect to symbol transmission speed) the Δ fT=1e-4 of channel promptly be equivalent to the poorest situation of 100ppm, and the channel unknown phase is the phase place between picked at random (0,2 π).

Algorithm parameter is selected:

(1) be estimated initial phases, the number of pilots that is sent is 20 of every frames.

(2) symbol lengths of M2M4 estimator employing is 2048;

(3) parameter of phase place/frequency offset tracking device is selected: α=0.5e-1, β=1e-4;

(4) parameter of amplitude/noise variance estimator is selected: A _Min=0.001, A _Max=0.999;

Demodulating and decoding general frame to above situation is seen Fig. 3.Fig. 4 has provided the iterative demodulation decoding algorithm hardware of Turbo coding BPSK/QPSK modulating system under FPGA+DSP realization frequency deviation and phase place the unknown and has realized schematic block diagram, in order to save resource in the FPGA sheet, and according to the actual conditions of algorithm---two soft goes into softly to go out the sub-decoder (see figure 2) and can not work simultaneously, the same soft soft module that goes out of going into of time division multiplexing hard-wired the time.

Work of the present invention is mainly reflected in the Processing Algorithm in the dsp chip.The main chip that this example adopts is by a slice EP20K400EBC652-2X of altera corp, and a slice dsp chip TMSC5402 of TI company and 2 plug-in RAM form.Wherein, dsp chip is used to finish the function with lower module: initial phase estimation module, M2M4 estimator, phase place and frequency-tracking module, amplitude/Noise Variance Estimation module, channel signal adjuster, its core are phase place and frequency-tracking module, amplitude/Noise Variance Estimation module.

Each functional module that hard-wired Turbo decoder of FPGA and DSP realize is undertaken transmitting data necessary mutually alternately by external RAM-A.

And Turbo decoding algorithm soft goes into softly to go out in the module four major parts are arranged: forward state metric calculation module (FSMC), back are to state metric calculation module (RSMC), logarithm likelihood ratio calculating module (LLRC) and path metric computing module (BMC).And 2 external RAMs (A ,-B) be used for storing the results of intermediate calculations of FSMC module and RSMC module.Wherein, external RAM-A is used for storing the frame data (receiving sampling) of self-channel, and dsp chip calculates the initial parameter of gained and estimates also to be stored on this sheet RAM.

The computation structure of concrete FSMC and RSMC is very similar, sees patent application (application number: 03137079.9) for details.

Receiver scheme among Fig. 3 realizes in dsp chip all that except that the Turbo decoder module its specific implementation flow process can be summed up as following step:

1. initialization system, and system parameters is set: P=20, L=2048, α=0.5e-1, β=1e-4, A _Min=0.001, A _Max=0.999, I=10, AmVarStart=0 (number of times that sign amplitude/Noise Variance Estimation module starts), t=0 (characterizing Turbo decoding iterations); The external information that all coded-bits are set is 0 (promptly without any priori).

2. each frame is extracted and lead sign indicating number (Pilot Symbols) and utilize formula (19) calculating channel initial phase θ ₀, specific implementation can be handled in digital number on (DSP) chip and carry out (generally can adopt look-up table to realize function arctan ()), and the result who calculates delivers to phase place/frequency offset tracking module.

3.M2M4 estimator calculates and obtain the rough estimate of amplitude and noise variance according to formula (24-26), this calculating can directly be finished in dsp chip, and phase place/frequency offset tracking module is delivered in the rough estimate that calculates.

4. system testing parameter A mVarStart, if detect AmVarStart=0, phase place/frequency offset tracking module is selected following input parameter: the amplitude of M2M4 estimator output and noise variance rough estimate evaluation, initial phase estimated value; If detect AmVarStart=1, phase place/frequency offset tracking module is selected following input parameter: the amplitude of amplitude/Noise Variance Estimation module output and the smart estimated value of noise variance, initial phase estimated value; This module after having selected parameter to the external information of channel received signal and input according to formula (20,21) (BPSK) or (22,23) (QPSK) handle and obtain in the frame all estimated value (being the real-time phase pursuit gain) θ of channel phases constantly _k, k=1,2 ..., N also sends the channel signal correction module to.

5. the channel signal correction module receives the real-time phase pursuit gain θ that phase place/frequency offset tracking module transmits _k, k=1,2 ..., N does the signal correction operation: promptly ${\tilde{r}}_k=r_k\exp (-{\mathrm{j\θ}}_k),$ K=1,2 ..., N starts amplitude/Noise Variance Estimation module simultaneously.

6. system testing parameter A mVarStart, if detect AmVarStart=1, then amplitude/noise variance module is not restarted, system continues execution in step 7.If detect AmVarStart=0, then amplitude/Noise Variance Estimation module is started working after receiving enabling signal, employing amplitude/Noise Variance Estimation iterative search algorithm flow AmVarEstimProg (A _Min, A _Max, I) processing is upgraded AmVarStart=AmVarStart+1, and is jumped to step 4 from the accurate estimated value of the signal and the output amplitude/noise variance of signal correction module.

7.Turbo decoder receives from the Equivalent A WGN signal of channel signal correction module and begins an iterative decoding job, and the output external information.

8. if iterations t equals predefined iterations, then export decoding value, otherwise put t=t+1, forward 4 to and continue to carry out.

This example is seen Fig. 6 and Fig. 7 in the performance under 12 iteration under the typical constant-parameter channel, by Fig. 6 and 7 as can be known:

1. adopt the BPSK modulation, under 4 to 12 iteration, this paper algorithm and the performance under the desirable AWGN under frequency deviation and phase place the unknown only differ in the 0.2dB, adopt the receiver under the desirable AWGN of QPSK modulation ratio to differ about 0.4dB.

2. algorithm complex: algorithm complex of the present invention is compared the AWGN decoding algorithm and is more or less the same, and is no more than 1 times.

Though reference wherein particular specific embodiment illustrates and illustrates the present invention, but, those of ordinary skill in the art will be understood that, under the situation that does not break away from the defined spirit and scope of claim that the present invention adds, can carry out various modifications on form and the details to the present invention.

Claims (10)

1, for the iterative demodulation coding/decoding method of the code modulation system with output external information, its step comprises

1) signal encoded and modulation that receives being carried out initial phase estimates;

2) adopt the M2M4 algorithm for estimating that amplitude and noise variance are carried out rough estimate;

3) adopt non-data auxiliary phase/frequency offset tracking algorithm to carry out phase place and frequency offset tracking, obtain the real-time phase pursuit gain according to above-mentioned initial phase estimated value and above-mentioned amplitude and noise variance rough estimate evaluation;

4) carry out channel signal according to above-mentioned real-time phase pursuit gain and proofread and correct output equivalent AWGN signal;

5), adopt iteration amplitude search algorithm to carry out amplitude and the smart estimation of noise variance according to above-mentioned Equivalent A WGN signal;

6) restarting non-data auxiliary phase/frequency offset tracking algorithm according to above-mentioned initial phase estimated value and above-mentioned amplitude and the smart estimated value of noise variance carries out phase place and frequency offset tracking, obtains the real-time phase pursuit gain;

7) carry out channel signal according to above-mentioned real-time phase pursuit gain and proofread and correct output equivalent AWGN signal;

8) decoder receives the smart estimated value of above-mentioned Equivalent A WGN signal and above-mentioned amplitude and noise variance, finishes an iterative decoding job, and the output external information;

9), start once more according to above-mentioned Equivalent A WGN signal and above-mentioned external information that iteration amplitude search algorithm carries out amplitude and noise variance is smart estimates in the amplitude that does not reach setting as yet with noise variance is smart when estimating number of times;

10) according to above-mentioned initial phase estimated value, smart estimated value of above-mentioned amplitude and noise variance and above-mentioned external information start non-data auxiliary phase/frequency offset tracking algorithm and carry out phase place and frequency offset tracking, obtain the real-time phase pursuit gain;

11) carry out signal correction according to above-mentioned real-time phase pursuit gain, output equivalent AWGN signal;

12) decoder receives above-mentioned Equivalent A WGN signal, and finishes another iterative decoding work according to above-mentioned amplitude and the smart estimated value of noise variance, and the output external information;

13) when the iterative decoding number of times reaches predetermined times, the output decoder value, otherwise repeat 9)-12).

2, the iterative demodulation coding/decoding method for the code modulation system with output external information as claimed in claim 1 is characterized in that described 1) also the signal encoded and modulation that receives is carried out the initial frequency deviation estimation.

3, the iterative demodulation coding/decoding method for the code modulation system with output external information as claimed in claim 2 is characterized in that described 1) carry out the initial frequency deviation estimation by fft algorithm.

4, the iterative demodulation coding/decoding method for the code modulation system with output external information as claimed in claim 1 is characterized in that the described TURBO of being encoded to sign indicating number or LDPC sign indicating number.

5, the iterative demodulation coding/decoding method for the code modulation system with output external information as claimed in claim 1 is characterized in that the described BPSK of being modulated to modulation or QPSK modulation.

6, a kind of iterative demodulation decoding device for code modulation system with output external information, comprise iterative decoder, the initial phase estimation module, the M2M4 estimator, the phase tracking module, channel signal adjuster, amplitude and Noise Variance Estimation module, initial phase estimation module are used for that the modulation signals that receives is carried out initial phase and estimate; The M2M4 estimator is used for the amplitude of carrying out to received signal and noise variance rough estimate; The phase tracking module receives the output of initial phase estimation module and the output of M2M4 estimator and carries out phase place and frequency offset tracking to received signal; The channel signal adjuster, the output of receiving phase/frequency-tracking module is proofreaied and correct signal; Amplitude and Noise Variance Estimation module, the output of receive channel signal calibrator are carried out amplitude and the smart estimation of noise variance, and output to iterative decoder and phase tracking module; Iterative decoder is accepted the output of channel signal calibrator and amplitude and Noise Variance Estimation module, carries out iterative decoding; It is characterized in that iterative decoder is also to amplitude and Noise Variance Estimation module and phase tracking module output external information.

7, as claimed in claim 6 for the iterative demodulation decoding device of code modulation system with output external information, it is characterized in that iterative decoder is located on the fpga chip; The initial phase estimation module, the M2M4 estimator, the phase tracking module, amplitude and Noise Variance Estimation module and channel signal adjuster are located in the dsp chip; Each several part carries out data preservation and transmission by external RAM.

8, the iterative demodulation decoding device for the code modulation system with output external information as claimed in claim 6 is characterized in that the initial phase estimation module also comprises the submodule that signal is carried out the initial frequency deviation estimation.

9, the iterative demodulation decoding device for the code modulation system with output external information as claimed in claim 8 is characterized in that the submodule that initial frequency deviation is estimated is the fft algorithm module.

10, the iterative demodulation decoding device for the code modulation system with output external information as claimed in claim 6 is characterized in that the described TURBO of being encoded to coding or LDPC coding; Described BPSK modulation or the QPSK of being modulated to.

Images