CN102404260A - Decision Feedback Type Equalizer - Google Patents

Abstract

The invention relates to a decision feedback type equalizer, capable of better reducing error dissemination. The decision feedback type equalizer comprises a storage device, a logarithm likelihood ratio calculating part, a logarithm likelihood ratio evaluation part, a decrement calculating part, and a logarithm likelihood ratio attenuation part, wherein the storage device is used for keeping input signals that make the decision responding to multiple signs to be repeatedly operated; the logarithm likelihood ratio calculating part is used for calculating the logarithm likelihood ratio according to the signals obtained by deducting the output of the feedback filter from the input signals kept in the storage device; the logarithm likelihood ratio evaluation part, according to the absolute value of the logarithm likelihood ratio, carries out confirmation over the signals transmitting the bit that is evaluated with low reliability; the decrement calculating part calculates and keeps the logarithm likelihood ratio attenuation amount corresponding to the whole or part of bits of the signals affected by the feedback filter; and the logarithm likelihood ratio attenuation part makes the logarithm likelihood ratio carry out attenuation according to the logarithm likelihood ratio attenuation amount calculated in the previous decision feedback cycle.

Description

Decision-feedback formula equalizer

Quoting of related applications such as basis for priority application

The application is with the Japanese patent application 2010-202098 (applying date: on September 9th, 2010) be the basis, enjoy preferential interests according to this application.The application is through comprising the full content of this application with reference to this application.

Technical field

Execution mode of the present invention relates to decision-feedback formula equalizer.

Background technology

In receiving system of communication/broadcast singal etc.; As resulting from intersymbol interference (the inter-symbol interference of multipath attenuation; ISI) countermeasure technology; Sometimes a mode using the nonlinear equalization with the characteristic that is superior to linear equalization is decision-feedback formula equalizer (decision feedback equalizer) (below, DFE representes decision-feedback formula equalizer).

In the typical type of decision-feedback formula equalizer, symbol with the feedback hard decision is arranged, remove the decision-feedback formula equalizer that moves to the mode of the intersymbol interference of successive character.But, in the type, when in the symbol hard decision, producing mistake, not only can not remove intersymbol interference, but also new interference component is produced to successive character, the error rate of successive character is increased.This generally is called error propagation, and it becomes the reason of the mis-behave of decision-feedback formula equalizer.

As the decision-feedback formula equalizer of other types, duplicate the decision-feedback formula equalizer that the influence that makes erroneous transmissions alleviates as the presumed value of the probability of symbolic point soft thereby have through feedback.For decision-feedback formula equalizer, hope to compare with the soft type in the past of duplicating of feedback further alleviate erroneous transmissions influence ground to improve performance.

Summary of the invention

The problem that the present invention will solve provides a kind of decision-feedback formula equalizer that alleviates error propagation more.

The feedback filter that utilizes of execution mode comprises the decision-feedback formula equalizer based on soft duplicating (soft replica) signal feedback of the log-likelihood ratio of bit: storage device, log-likelihood calculations portion, log-likelihood ratio evaluating part, attenuation calculating part and log-likelihood specific damping portion.Storage device is kept for making the input signal that can carry out repeatedly to the decision-feedback circulation of a plurality of symbols.The log-likelihood ratio of a plurality of bits calculates based on the signal of the output gained that from the input signal of being read by above-mentioned storage device, deducts above-mentioned feedback filter in log-likelihood calculations portion.The log-likelihood ratio evaluating part is estimated the reliability of a plurality of bits based on the absolute value of above-mentioned log-likelihood ratio, exports at least having transmitted the notation index that the symbol that is be evaluated as the low bit of reliability is confirmed.The attenuation calculating part calculates the log-likelihood ratio attenuations of whole or a part of bit of bringing the symbol of influence with the symbol of confirming for via above-mentioned feedback filter the above-mentioned notation index that utilizes output, and circulates and it is kept for next time decision-feedback.Log-likelihood specific damping portion makes in the circulation of the decision-feedback of this time the above-mentioned log-likelihood ratio that calculates through above-mentioned log-likelihood calculations portion, decays according to the above-mentioned log-likelihood ratio attenuation that in previous decision-feedback circulation, calculates through above-mentioned attenuation calculating part.

Decision-feedback formula equalizer according to above-mentioned formation can more alleviate error propagation.

Description of drawings

Fig. 1 is the routine block diagram of formation of the decision-feedback formula equalizer of expression first execution mode.

Fig. 2 is the routine flow chart of action of the decision-feedback formula equalizer of expression first execution mode.

Fig. 3 is used for the figure that first example to the symbol ad hoc approach describes.

Fig. 4 is used for the figure that second example to the symbol ad hoc approach describes.

Fig. 5 is used for the figure that the 3rd example to the symbol ad hoc approach describes.

Fig. 6 is used for the figure that the 4th example to the symbol ad hoc approach describes.

Fig. 7 is used for the figure that the 5th example to the symbol ad hoc approach describes.

Fig. 8 is used for the phasor (phasor) of feedback tap coefficient is shown the figure that describes.

Fig. 9 is the routine block diagram of formation of the decision-feedback formula equalizer of expression second execution mode.

Figure 10 is the routine flow chart of formation of the decision-feedback formula equalizer of expression second execution mode.

Figure 11 is the routine block diagram of formation of the decision-feedback formula equalizer of expression comparative example.

Figure 12 is the routine block diagram of another formation of the decision-feedback formula equalizer of expression comparative example.

Figure 13 is the another block diagram that constitutes example of the decision-feedback formula equalizer of expression comparative example.

Figure 14 is used for the figure that the IQ composition to the symbolic point of R-QAM describes.

Embodiment

Below, on one side with reference to accompanying drawing, the decision-feedback formula equalizer to execution mode of the present invention at length describes on one side.In addition, in the execution mode below,, carry out identical action, and omit the explanation of repetition for the part of additional phase with numbering.

FF representes feedforward (feed-forward), and FB representes feedback (feed-back).

Decision-feedback formula equalizer general with the feedforward Filtering Processing of stipulating feedforward filter portion (FF filter section) and to carry out the feedback filter portion (FB filter section) of regulation feedback filtering processing relevant.

In the implementation method of FF filter section; For example just like more research in the past such service time the zone equalization filter the method (below, FDE representes that frequency field is balanced) of balanced (the frequency domain equalizer) filter in frequency of utilization zone of method and research in nearer several years.In the explanation below; For the principle of decision-feedback formula equalizer is described; Sometimes will use frequency field decision-feedback formula equalizer (the frequency domain decision feedback equalizer of the latter's FDE in the FF filter section; FD-DFE) as an example, but this execution mode is not limited to frequency field decision-feedback formula equalizer.This execution mode comes acquired character to improve through the change of feedback action; Irrelevant with the concrete formation of FF filter section, can be applied in the decision-feedback formula equalizer (also can be applied to certainly to use in the former the decision-feedback formula equalizer of time zone equalization filter) prevailingly.

Likewise, this execution mode is irrelevant with the concrete formation of FB filter section, can be applied to prevailingly in the decision-feedback formula equalizer.

At first, on one side with reference to Figure 11～Figure 13, the decision-feedback formula equalizer to comparative example describes on one side.

Shown in figure 11, the decision-feedback formula equalizer of comparative example is relevant in feedback filter portion (the FB filter section) 102 of this feedforward filter portion 100 with continued access with feedforward filter portion (FF filter section) 100.Deduct the signal that obtains after the output of FB filter section 102 from the output of FF filter section 100, given symbol hard decision portion 103 from subtraction portion 108, the output of symbol hard decision portion 103 gives FB filter section 102.

Below, on one side with reference to Figure 12, use on one side through the data symbol { s of frequency field decision-feedback formula equalizer to being transmitted by single carrier with symbol rate 1/T _m(m=0,1 ..., (M-1), E{s _m}=0, E{|s _m| ²The method of }=1) restoring describes comparative example.

DFT representes DFT (discrete Fouriertrans form), and IDFT representes contrary DFT (perhaps contrary DFT) (inverse DFT).

The frequency field decision-feedback formula equalizer of Figure 12 uses the FF filter section 100 of Figure 11 to comprise the FDE of DFT portion 1001, tap (tap) coefficient multiplying portion 1002 and contrary DFT portion 1003 and constitutes.

As I (I is the integer more than 1), MI that is input among the FDE receives sample { r to the symbolic number of each DFT block that will in FDE, use as the sample number of M (M is the integer more than 1), each symbolic number _m(m=0,1 ..., (MI-1)) represent by following formula (1).

[mathematical expression 1]

$r_m=\underset{k=0}{\overset{M-1}{\mathrm{\Σ}}}{s}_{k}h(\mathrm{mT}/I-\mathrm{kT})+n(\mathrm{mT}/I)...\left(1\right)$

At this, h (t) is the impulse response of route of transmission, and n (mT/I) is average 0, variances sigma ²The additivity white noise.

In DFT portion 1001, will receive sample r _mAfter the DFT that utilizes MI to order converts the signal of frequency field into, in tap coefficient multiplying portion 1002, multiply by the balanced tap coefficient { W of frequency field _I(I=0,1 ..., (MI-1)), and, in contrary DFT portion 1003, the regional signal of IDFT return time that utilizes MI to order.And then, extract (decimation) and with its signal by the mode of corresponding 1 sample of 1 symbol as the M sample.

The feedback filtering that FB filter section 102 is stipulated is handled.At this as an example, have B tap, with " { f _k ^*, k ∈ F _B" represent their tap coefficient.At this, the key element of FB is the integer more than 1 of the time of delay of expression tap, and its unit is a symbol time.

Under the situation of B=0, F _BBe null set, become linear equalizer with FB filter.

[mathematical expression 2]

Under the situation of B＞0, carried out the symbol behind the hard decision

in the past and be input to FB filter section 102.

And in subtraction portion 108, the output of FB filter section 102 is deducted from FDE output (under the situation of Figure 12, the output of contrary DFT portion 1003).

Therefore, m sample of DFE output (that is the output of decision-feedback formula equalizer) is as shown in the formula shown in (2).

[mathematical expression 3]

$z_m=\frac{1}{\mathrm{MI}}\underset{l=0}{\overset{\mathrm{MI}-1}{\mathrm{\&Sigma;}}}{W}_l{R}_l\exp \left(\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000068491080000021.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">j</figure-callout>}\frac{2\mathrm{\&pi;}}{M}\mathrm{lm}\right)-\underset{k\&Element;F_B}{\mathrm{\&Sigma;}}{f}_k^{*}{\hat{s}}_{m-k}...\left(2\right)$

Wherein, m=0,1 ..., (M-1),

$R_l=\underset{m=0}{\overset{\mathrm{MI}-1}{\mathrm{\&Sigma;}}}{r}_m\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000068491080000021.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">j</figure-callout>}\frac{2\mathrm{\&pi;}}{\mathrm{MI}}\mathrm{lm}),l=\mathrm{0,1},...,(\mathrm{MI}-1),$

Asterisk () ^*The expression complex conjugate.

[mathematical expression 4]

At this, the impulse response of establishing the route of transmission through what MI point DFT was transformed to transfer function is:

$H_l=\underset{m=0}{\overset{\mathrm{MI}-1}{\mathrm{\&Sigma;}}}h(\mathrm{MT}/I)\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000068491080000021.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">j</figure-callout>}\frac{2\mathrm{\&pi;}}{\mathrm{MI}}\mathrm{lm}).$

So, will send symbol s _mWith DFE output z _mMean Square Error | e _m| ²}=E{|z _m-s _m| ²FDE tap coefficient { W in the minimized MMSE standard _IAnd FB tap coefficient { f _k, utilize following formula (3) expression.

[mathematical expression 5]

$W_l=\frac{H_l^{*}(1+\underset{k\&Element;F_B}{\mathrm{\&Sigma;}}{f}_k^{*}\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000068491080000021.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}\frac{\mathrm{kl}}{M}))}{{\mathrm{\&sigma;}}^2+{\left|{\hat{H}}_l\right|}^2}=\frac{H_l^{*}{F}_l}{{\mathrm{\&sigma;}}^2+{\left|{\hat{H}}_l\right|}^2}$

Wherein, ${\left|{\hat{H}}_l\right|}^2=\underset{k=0}{\overset{I-1}{\mathrm{\&Sigma;}}}{\left|H(l+\mathrm{KM})\mathrm{Mod}\mathrm{MI}\right|}^2$

Vf＝-v，

Wherein,

$f={({\mathrm{<figure-callout\; id="1"\; label="f\; k"\; filenames="HDA0000068491080000021.png,HDA0000068491080000041.png"\; state="\{\{state\}\}">f</figure-callout>}}_{k_{}},{\mathrm{<figure-callout\; id="2"\; label="f\; k"\; filenames="HDA0000068491080000021.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">f</figure-callout>}}_{k_{}},...,f_{k_B})}^{\&prime;},v={({\mathrm{<figure-callout\; id="1"\; label="v\; k"\; filenames="HDA0000068491080000021.png,HDA0000068491080000041.png"\; state="\{\{state\}\}">v</figure-callout>}}_{k_{}},{\mathrm{<figure-callout\; id="2"\; label="v\; k"\; filenames="HDA0000068491080000021.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">v</figure-callout>}}_{k_{}},...,v_{k_B})}^{\&prime;},$

$V=\left[\begin{array}{cccc}{v}_0& v_{k_1}-{\mathrm{<figure-callout\; id="2"\; label="v\; k"\; filenames="HDA0000068491080000021.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">v</figure-callout>}}_{k_{}}& .& v_{k_l}-v_{k_B}\\ v_{k_2}-v_{k_1}& v_0& v_{k_2}-{\mathrm{<figure-callout\; id="3"\; label="v\; k"\; filenames="HDA0000068491080000021.png,HDA0000068491080000061.png"\; state="\{\{state\}\}">v</figure-callout>}}_{k_{}}& .\\ .& .& .& .\\ v_{k_B}-v_{k_l}& .& .& {\mathrm{<figure-callout\; id="0"\; label="v"\; filenames="HDA0000068491080000031.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">v</figure-callout>}}_0\end{array}\right]$

$v_k=\frac{{\mathrm{\&sigma;}}^2}{M}\underset{l=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\frac{\exp (-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HDA0000068491080000021.png,HDA0000068491080000032.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}\frac{\mathrm{lk}}{M})}{{\mathrm{\&sigma;}}^2+{\left|{\hat{H}}_l\right|}^2}...\left(3\right)$

In addition, single order () ' expression transposition.

[mathematical expression 6]

If the modulation symbol point is used s _j, j=1,2 ..., R (R is the number of modulation symbol point) expression, then hard decision symbol

By following expression.

${\hat{s}}_m=\underset{s_j}{\arg }\min {|z_m-s_j|}^2...\left(4\right)$

Below, on one side with reference to Figure 13, use the action of the situation of feedback soft duplicating (soft replica) signal on one side, comparative example is described.

SR representes soft duplicating, and LLR representes log-likelihood ratio (log likelihood ratio).

The decision-feedback formula equalizer of Figure 13 comprises: according to DFE output (promptly; The output of this decision-feedback formula equalizer; Being that output from FF filter section 200 deducts the signal that obtains after the output of FB filter section 202, is the signal that subtraction portion 208 is exported at this) calculate the log-likelihood calculations portion (LLR calculating part) 203 of the log-likelihood ratio of each bit; Generate the soft generation portion (SR generation portion) 204 of duplicating of soft reproducing signals according to log-likelihood ratio; And FB filter section 202.

FF filter section, the FB filter section with Figure 11 or Figure 12 is identical respectively for FF filter section 200, FB filter section 202, but is not restricted to this.

DFE output is outputed in the processing block (not shown) of back level such as hard decision or soft-decision for example.

LLR calculating part 203 is obtained LLR by next the output according to DFE of following content.

Observing DFE output Z _mThe time, it is put S as modulation symbol _jAnd the probability of the tape spare that sends is obtained through calculating following formula (5).

[mathematical expression 7]

$P\left[s_j\right|z_m]=\frac{1}{\sqrt{{2\mathrm{\&pi;\&sigma;}}^2}}\exp \left(\frac{-{|z_m-s_j|}^2}{{2\mathrm{\&sigma;}}^2}\right)...\left(5\right)$

Use this result, n the bit b of symbol m _nLLR calculate by following formula (6).

[mathematical expression 8]

$L(m;n)=\ln \left(\frac{P[b_n=+1|z_m]}{P[b_n=-1|z_m]}\right)=\ln \left(\frac{\underset{\&ForAll;s_j\mathrm{where}{b}_n=+1}{\mathrm{\&Sigma;P}\left[s_j\right|z_m]}}{\underset{\&ForAll;s_j\mathrm{where}{b}_n=-1}{\mathrm{\&Sigma;P}\left[s_j\right|z_m]}}\right)...\left(6\right)$

[mathematical expression 9]

Next; SR generation portion 204 exports as soft reproducing signals

according to the statistical presumed value that LLR calculates the IQ composition of symbol.Shown in figure 14, when the IQ composition with the R-QAM symbolic point is expressed as s respectively _{I; i}, s _{Q; i},

The time, the presumed value of IQ composition is obtained by following formula (7).

${\hat{s}}_{m,I}=E\left\{s_{m,I}\right\}=\underset{i=0}{\overset{\sqrt{R}-1}{\mathrm{\&Sigma;}}}{s}_{I;i}\&CenterDot;P\left[s_{I;i}\right|z_m]$

${\hat{s}}_{m,Q}=E\left\{s_{m,Q}\right\}=\underset{i=0}{\overset{\sqrt{R}-1}{\mathrm{\&Sigma;}}}{s}_{Q;i}\&CenterDot;P\left[s_{Q;i}\right|z_m]...\left(7\right)$

[mathematical expression 10]

And then, if the value of supposition bit is independent with the value of other bits, then become following formula (8).

${\hat{s}}_{m,I}=E\left\{s_{m,I}\right\}=\underset{i=0}{\overset{\sqrt{R}-1}{\mathrm{\&Sigma;}}}{s}_{I;i}\&CenterDot;\left[{\mathrm{Ps}}_{I;i}\right|z_m]=\underset{i=0}{\overset{\sqrt{R}-1}{\mathrm{\&Sigma;}}}{s}_{I;i}\&CenterDot;\underset{j\&Element;{\mathrm{\&Psi;}}_I}{\mathrm{\&Pi;}}P[b_j\left|z_m\right]$

${\hat{s}}_{m,Q}=E\left\{s_{m,Q}\right\}=\underset{i=0}{\overset{\sqrt{R}-1}{\mathrm{\&Sigma;}}}{s}_{Q;i}\&CenterDot;P\left[s_{Q;i}\right|z_m]=\underset{i=0}{\overset{\sqrt{R}-1}{\mathrm{\&Sigma;}}}{s}_{Q;i}\&CenterDot;\underset{j\&Element;{\mathrm{\&Psi;}}_Q}{\mathrm{\&Pi;}}P[b_j\left|z_m\right]...\left(8\right)$

At this, Ψ _I, Ψ _QBe respectively the set of mapping (mapping) for the index of the bit of the IQ composition of symbolic point.

[mathematical expression 11]

According to above-mentioned $L(m;n)=\mathrm{Ln}\left(\frac{P[b_n=+1|z_m]}{P[b_n=-1|z_m]}\right)$ Relation, P [bn=+l|z _m] and P [bn=-l|z _m] use LLR to obtain by following formula (9) that kind.

$P[b_n=+1|z_m]=\frac{\exp \left(L(m;n)\right)}{1+\exp \left(L(m;n)\right)}$

$P[b_n=-1|z_m]=\frac{1}{1+\exp \left(L(m;n)\right)}...\left(9\right)$

(first execution mode)

So far comparative example is illustrated, below, first execution mode is described.

In addition, in decision-feedback formula equalizer, known meeting produces error propagation.That is, under the situation that the reliability of certain bit reduces in equalizer output, may the symbol that transmit this bit correctly not carried out removing of intersymbol interference.This means and at a plurality of symbols and to transmit between the reliability of bit of this symbol, find relational via feedback tap.This execution mode utilizes the reliability of this relational bit to upgrade, and realizes that repetition is balanced.

First execution mode is that example describes with the decision-feedback formula equalizer that feeds back soft type of duplicating.

Fig. 1 is the block diagram of an example of the decision-feedback formula equalizer of this execution mode of expression.

As shown in Figure 1, the decision-feedback formula equalizer of this execution mode comprises: storage device 11, the output of maintenance feedforward filter portion (FF filter section) 10; LLR according to DFE output (that is, deducting the signal that obtains after the output of feedback filter portion (FB filter section) 12 from the signal of being read by storage device 11, also is the signal that subtraction portion 18 is exported), calculates in log-likelihood calculations portion (LLR calculating part) 13; The soft generation portion (SR generation portion) 14 of duplicating generates soft reproducing signals according to LLR; FB filter section 12 is carried out feedback filtering to soft reproducing signals and is handled; Log-likelihood ratio evaluating part (LLR evaluating part) 15 uses the reliability of LLR bit to estimate; The index of bringing the symbol of influence for the symbol that transmits the low bit of reliability via FB filter section 12 obtained in index/attenuation calculating part (attenuation calculating part) 16, and whole perhaps LLR attenuations of a part of bit of this symbol are calculated/kept; And log-likelihood specific damping portion (LLR decay portion) 17, to the bit that calculates the LLR attenuation, respectively the LLR that is calculated by LLR calculating part 13 multiply by this LLR attenuation.

FF filter section 10, FB filter section 12, LLR calculating part 13, SR generation portion 14, subtraction portion 18 respectively can be identical with FF filter section 200, FB filter section 202, LLR calculating part 203, SR generation portion 204, the subtraction portion 208 of Figure 13, but also be not limited to this.

Fig. 2 is the routine flow chart of action of the decision-feedback formula equalizer of this execution mode of expression.

At this, will be called " DFE circulation " or " decision-feedback circulation " to the processing action of whole symbols execution LLR calculating, SR generation, FB filter.

In this example, decision-feedback formula equalizer is carried out at least 1 time, maximum N _ItInferior DFE circular treatment.N _ItIt for example can be predetermined times.In addition, also can make the DFE circular treatment necessarily carry out N _ItInferior, but, end its later DFE circular treatment and no longer execution said preferred under the situation that rated condition is set up like the back.

In this example, the DFE input signal in each circulation is the output of above-mentioned storage device 11.DFE output after the DFE circular treatment finishes for example also can be input in the processing block (not shown) of back level.The processing block of back level for example is the balanced processing of processing, (secondary) of hard decision, the processing of error correction decoding etc., but is not limited to this.

In this execution mode; Repeatedly implement the decision-feedback circulation; Symbol in (i+1) inferior decision-feedback circulation is soft to be duplicated when generating; Bring the symbol of influence if circulate to the low symbol of reliability, then reduce, thereby can alleviate the influence of error propagation through making at this soft log-likelihood ratio that uses in the generation that duplicates in the preceding once decision-feedback of (the i time).

Below, when explanation DFE circular treatment, with the i time (i=1,2 ..., N _It) m symbols Z after the DFE circulation _{M, i}N bit b _nLLR utilize L _i(m; N) expression.

In addition, the processing based on LLR decay portion 17 of step S3 is not carried out in the DFE circulation in the first time, in each for the second time later DFE circulation, carries out (with reference to 30 among the figure) respectively.

Like this, in step S1, relatively m and M (M is symbolic number (integer 1 or more)) if m＜M, then get into this step S2, otherwise (that is, as if to this processing of whole sign-offs), the then processing of entering step S1 next time.

In each time of DFE circulation, at first, in step S2, LLR calculating part 13 calculates and output LLR.For example, can likewise calculate LLR with the LLR calculating part 203 of Figure 13.

Next, carry out a series of processing (wherein being the processing of step S4, S5 for the first time) of step S3～S5 and a series of processing of step S6～S8.In addition, the processing of the processing of step S3～S5 and step S6～S8 can make arbitrary side carry out earlier, also can the while executed in parallel.

At first, the processing of the step S6～S8 that does not have in the comparative example to Figure 13 describes.

LLR evaluating part 15 and index/attenuation calculating part 16 moves in the last circulation in addition of DFE circulation.This decision-feedback formula equalizer is when reaching N _ItInferior before at the Ns time (Ns＜N _It) when ending the DFE circular treatment, the processing of step S6～S8 is at i=1 ..., action (is being carried out N in each DFE circulation of Ns-1 time _ItUnder the situation of inferior DFE circular treatment, the processing of step S6～S8 is at i=1 ..., N _ItAction in each DFE circulation of-1 time).

In step S6, LLR evaluating part 15 utilizes stipulated standard to estimate the absolute value of LLR, and the bit low to reliability detects.

When 15 pairs of low bits of reliability of LLR evaluating part detect,, for example can use following benchmark (still, being not limited to these) as above-mentioned stipulated standard.

In first benchmark, detecting becomes | L _i(m; N) |＜L _ThBit, with it as the low bit of reliability.Wherein, L _ThIt is predetermined threshold value.

In second benchmark, detecting becomes | L _i(m; N) |＜MA (| L _i(m; N) |, N _MA) bit, with it as the low bit of reliability.Wherein, MA (, N _MA) be the predetermined long N in interval _MAMoving average (moving average).

LLR evaluating part 15 is exported index (bit index) that this bit is confirmed and the index (notation index) that the symbol that transmits this bit is confirmed to detected bit.

In addition; In the i time DFE circulation; In its finish time, under the low all nd situation of bit of reliability, preferably do not carry out the DFE circulation of (after the i+1 time) after it; This DFE of the i time is input to the processing block (for example, hard decision, secondary equilibrium, error correction decoding etc.) of back level.

Next; In DFE circulation, detect under the situation of the low bit of reliability at the i time; In step S7, index/attenuation calculating part 16 is obtained the index of bringing the symbol of influence for the symbol that transmits this detected bit via FB filter section 12, and; About whole or a part of bit of this symbol, to calculating in the soft attenuation of duplicating the LLR that uses in the generation of next DFE circulation ((i+1) is inferior).

The LLR attenuation that calculates for next DFE circulation, remains in the memory (for example, the memory of the inside of index/attenuation calculating part 16 or outside memory etc.) (not shown) of regulation in step S8.

In addition, index/attenuation calculating part 16 confirms to be judged as the symbols Z of the low bit of reliability to transmission through stipulated standard _{M, i}(wherein, m=0 ..., M-1) bring the symbol of influence.

Below, for several examples of the above-mentioned stipulated standard of the notation index of the reason that is used to confirm to become error propagation, with reference to Fig. 3～Fig. 7 (still, be not limited to these) on one side be described on one side.In addition, in Fig. 3～Fig. 7, Z is shown respectively _{M, i}Past of output time the m soft reproducing signals and the FB tap that generate.For the convenience of explaining, the time shaft of FB tap reverses.

In first benchmark, as shown in Figure 3, will be with Z _{M, i}Output time m be index m-k (the k ∈ F that there is whole moment of FB tap in starting point _B), confirm as the notation index that brings above-mentioned influence.

In second benchmark, as shown in Figure 4, from the absolute value of FB tap coefficient | f _k| N is selected on descending ground successively _tTap (N _tBe the preset upper limit number), will be with Z _{M, i}Output time m be that (k is for from | f for the starting point index m-k that obtains the moment that their exist _k| the N that a big side begins _tIndividual), confirm as the notation index that brings above-mentioned influence.

In the 3rd benchmark, as shown in Figure 5, will be with Z _{M, i}Output time m be the absolute value that starting point is obtained the FB tap coefficient | f _k| than defined threshold f _ThThe index m-k in the moment that big tap exists (| f _k|＞f _Th), confirm as the notation index that brings above-mentioned influence.

In the 4th benchmark, as shown in Figure 6, select N time of delay from small to large successively from the FB tap _tTap (N _tBe the preset upper limit number), will be with Z _{M, i}Output time m be that (k is for from being worth the N of little beginning for the starting point index m-k that obtains the moment that their exist _tIndividual), confirm as the notation index that brings above-mentioned influence.

In the 5th benchmark, as shown in Figure 7, will be with Z _{M, i}Output time m be that starting point is obtained in the FB tap time of delay less than defined threshold k _ThThe index m-k (k＜k in moment of existing of tap _Th), confirm as the notation index that brings above-mentioned influence.

In addition, said reference also can wait suitably and select according to the hardware size of the path of frequency band, request.Among said reference, first benchmark can be expected to obtain maximum performance, but exists under the situation in overriding path, in the 3rd benchmark, also can expect to obtain high-performance.The second and the 4th benchmark has the little advantage of hardware size.

Next, object bit and the computational methods of LLR attenuation in index/attenuation calculating part 16 are described.In addition, the FB tap coefficient f that uses as shown in Figure 8 _k=| f _k| exp (j θ _k) expression.

In first method, among the bit that transmits by the notation index m-k that obtains as stated, be chosen as the bit on the identical IQ axle of the low bit of reliability for being mapped in LLR evaluating part 15, calculate 1-|f _k|| cos (θ _k) | as the LLR attenuation, be chosen as the bit on the different IQ axle of the low bit of reliability, calculate 1-|f for being mapped in LLR evaluating part 15 _k|| sin (θ _k) | as the LLR attenuation.

In second method, among the bit that transmits by the notation index m-k that obtains as stated, be chosen as the bit on the identical IQ axle of the low bit of reliability for being mapped in LLR evaluating part 15, calculating 1/exp (| f _k|| cos (θ _k) |) as the LLR attenuation, be chosen as the bit on the different IQ axle of the low bit of reliability for being mapped in LLR evaluating part 15, calculating 1/exp (| f _k|| sin (θ _k) |) as the LLR attenuation.

According to these first and second methods, can consider to have shone upon arbitrary axle of the I that is chosen as the low bit of reliability, Q axle and, decide the I axle of soft reproducing signals, the attenuation rate of Q axis signal to this contribution degree of FB tap.

In third party's method,, calculate 1-|f for the whole bits that transmit by the notation index m-k that obtains as stated _k| as the LLR attenuation.

In cubic method, for the whole bits that transmit by the notation index m-k that obtains as stated, calculating 1/exp (| f _k|) as the LLR attenuation.

These the 3rd, in the cubic method, can decide attenuation rate to constitute more easily than first and second method.

As the 5th method, among the bit that transmits by the notation index m-k that obtains as stated, be chosen as the bit on the identical IQ axle of the low bit of reliability for being mapped in LLR evaluating part 15, calculate 1-|f _k|| cos (θ _k) |; Then with its quantification, for example 1,1/2,1/4,1/8 ..., hardware such as 0 realizes being easy among the coefficient, uses coefficient near this quantized value as the LLR attenuation; Be chosen as the bit on the different IQ axle of the low bit of reliability for being mapped in, calculate 1-|f with LLR evaluating part 15 _k|| sin (θ _k) |, then with its quantification, for example 1,1/2,1/4,1/8 ..., hardware such as 0 realizes among the easy coefficient, uses coefficient near this quantized value as the LLR attenuation.

As the 6th method, the whole bit for being transmitted by the notation index m-k that obtains as stated calculates 1-|f _k|, then with its quantification, for example 1,1/2,1/4,1/8 ..., hardware such as 0 realizes among the easy coefficient, uses coefficient near this quantized value as the LLR attenuation.

And then as the 7th method; Among the bit that transmits by the notation index m-k that obtains as stated; Be chosen as the bit on the identical IQ axle of the low bit of reliability for being mapped in LLR evaluating part 15, for example use 1,1/2,1/4,1/8 ..., hardware such as 0 realize easy coefficient as the LLR attenuation (with | f _k| irrelevant), be chosen as the bit on the different IQ axle of the low bit of reliability, coefficient of utilization 1 for being mapped in LLR evaluating part 15.In addition, for the bit that is mapped on the above-mentioned identical IQ axle, the LLR attenuation of use for example can be set during fabrication.

As the method from all directions, for the whole bit that transmits by the notation index m-k that obtains as stated, for example use 1,1/2,1/4,1/8 ..., hardware such as 0 realize easy coefficient as the LLR attenuation (with | f _k| irrelevant).In addition, the LLR attenuation for example can be set during fabrication.

These the five～the from all directions methods particularly the FB tap coefficient be under the situation of real number value effectively.

Through above processing, about being chosen as the low symbols Z of reliability _{M, i}Bit b _n, obtain the set { I of index of the bit of the object that in circulation next time, becomes LLR decay _{M, n}}=(I _{M, n}(0), I _{M, n}(1) ..., I _{M, n}(L _{M, n}-1)) (I _{M, n}() is 0 to the integer of MR-1) and to the set { g of the attenuation of this bit _{M, n}}=(g _{M, n}(0), g _{M, n}(1) ..., g _{M, n}(L _{M, n}-1)).

Whole bits (to index k=0,1 ..., the LLR attenuation of MR-1 g}=(g (0), g (1) ..., g (MR-1)), for example can be by definite as getting off.Repeatedly be designated as in same bits under the situation of LLR decay object (calculating object of LLR attenuation), the LLR attenuation of this bit adopts amassing a plurality of LLR attenuations of this bit calculation.{ g} is retained to till next DFE circulation the LLR attenuation.

[mathematical expression 12]

$g\left(k\right)=\mathrm{l\; for\; k}\&NotElement;\underset{\&ForAll;m,n}{U}\left\{l_{m,n}\right\}$

$g\left(k\right)=\underset{\&ForAll;\mathrm{j\; where}{l}_{m,n}\left(j\right)=k}{\mathrm{\&Pi;}{g}_{m,n}\left(j\right)}\mathrm{for\; k}\&Element;\underset{\&ForAll;m,n}{U}\left\{l_{m,n}\right\}...\left(10\right)$

Also can replace said method, and for example decide LLR attenuation { g} by following formula (11).Repeatedly be designated as in same bits under the situation of LLR decay object (calculating object of LLR attenuation), the LLR attenuation of this bit adopts value minimum among a plurality of LLR attenuations of this bit calculation.{ g} is retained to till next DFE circulation the LLR attenuation.

[mathematical expression 13]

$g\left(k\right)=\mathrm{l\; for\; k}\&NotElement;\underset{\&ForAll;m,n}{U}\left\{l_{m,n}\right\}$

$g\left(k\right)=\underset{\&ForAll;\mathrm{j\; where}{l}_{m,n}\left(j\right)=k}{\min }{g}_{m,n}\left(j\right)\mathrm{for\; k}\&Element;\underset{\&ForAll;m,n}{U}\left\{l_{m,n}\right\}...\left(11\right)$

Next, the processing to step S3～S5 describes.The processing of step S3 is the processing that in the comparative example of Figure 13, does not have.

Circulate at for the second time later DFE (i=2 ..., N _It) in, in step S3, LLR decay portion 17 move (with reference to figure in 30).Through LLR Li (m to importing from LLR calculating part 13; N) { g} obtains L to multiply by the LLR attenuation that in previous circulation, calculates _i' (m; N).

L _i′(m?；n)＝g(m?·log ₂(R)+n)L _i(m；n)

In step S4, SR generation portion 14 generates soft reproducing signals according to LLR.For example, also can generate soft reproducing signals with the SR generation portion 204 of Figure 13 identically.

But, in this execution mode, in the DFE circulation that primary DFE circulates (i=1) and the second time is later, different to the input of SR generation portion 14.

That is, in primary DFE circulation (i=1), do not make 17 actions of LLR decay portion as yet, SR generation portion 14 from LLR calculating part 13 with reference to L ₁(m; N)=L (m; N) generate soft reproducing signals.

[mathematical expression 14]

In for the second time later DFE circulation, from LLR decay portion 17 with reference to Li ' (m; N) generate soft reproducing signals

The soft reproducing signals that generates is input in the FB filter section 12.

In addition, the inside that the function of 30 switch can be in LLR calculating part 13 among the figure also can be in the outside.

Next, in step S5,12 pairs of soft reproducing signals that generate through SR generation portion 14 of FB filter section carry out feedback filtering and handle.FB filter section 12 can be identical with FB filter section 102, but be not limited thereto.

The output of FB filter section 12 is administered in the subtraction portion 18.Be from the signal of reading by storage device 11, to deduct the signal that obtains after the output of FB filter section 12 (promptly from the signal of subtraction portion 18 output as stated; DFE output); This signal is imported in the LLR calculating part 13, and enters into DFE circular treatment next time, perhaps; Be imported in the processing block (for example, hard decision, secondary equilibrium, error correction decoding etc.) of back level.

(second execution mode)

Below, second execution mode is described.

In this execution mode; Also identical with first execution mode; Symbol in (i+1) inferior decision-feedback circulates is soft to be duplicated when generating; If in the decision-feedback circulation of preceding once (the i time), bring the symbol of influence, then reduce, thereby the influence of error propagation is reduced through making at this soft log-likelihood ratio that uses in the generation that duplicates to the low symbol of reliability.At this moment, in this execution mode, details is of the back, adjudicates through in LLR decay portion 17, importing threshold value, thereby even be designated as decay object bit, under the sufficiently high situation of the reliability of this bit, also can avoid making the LLR decay.That is, in noise circumstance, even removed intersymbol interference well, also can produce the bit of low reliability because of The noise, the bit that its place ahead possibly take place is designated as the situation of decay object bit.In this execution mode,, import the judgement of reliability under such situation, can not make the mode of removing mitigation of intersymbol interference.

Identical with first execution mode, second execution mode is to be that example describes with the decision-feedback formula equalizer that feeds back soft type of duplicating.

Second execution mode with the first execution mode difference be that the center describes.

Fig. 9 is the block diagram of an example of the decision-feedback formula equalizer of this execution mode of expression.

As shown in Figure 9, the decision-feedback formula equalizer of this execution mode comprises: storage device 11, the output that keeps FF filter section 10; LLR calculating part 13 according to DFE output (that is, from the signal of being read by storage device 11, deducting the signal that obtains after the output of FB filter section 12, also is the signal that subtraction portion 18 is exported), calculates LLR; SR generation portion 14 generates soft reproducing signals according to LLR; FB filter section 12 is carried out feedback filter to soft reproducing signals and is handled; LLR evaluating part 15 uses LLR to estimate the reliability of bit; The index of bringing the symbol of influence for the symbol that transmits the low bit of reliability via FB filter section 12 obtained in index/attenuation calculating part 16, the LLR attenuation of whole or a part of bit of this symbol of calculating/maintenance; And LLR decay portion 17, for the bit of the LLR that calculates among the bit that calculates the LLR attenuation, at LLR calculating part 13, respectively this LLR multiply by this LLR attenuation less than defined threshold.

The processing block of the FF filter section 10 of this execution mode, storage device 11, FB filter section 12/LLR calculating part 13, SR generation portion 14, LLR evaluating part 15, index/attenuation calculating part 16, subtraction portion 18, back level is identical with first execution mode.

The computing according to LLR decay portion 17 of this execution mode itself is identical with first execution mode.In addition, in the DFE circulation first time (i=1), do not make 17 actions of LLR decay portion as yet, SR generation portion 14 is with reference to L _i(m; N)=L (m; N) generate soft reproducing signals, this point is also identical with first execution mode.

Circulate at for the second time later DFE (i=2 ..., N _It) in; The LLR decay portion 17 of first execution mode is to the bit that calculates the LLR attenuation; Respectively this LLR multiply by this LLR attenuation; Relative therewith, this LLR attenuation (with reference to 32 among the figure) less than the bit of defined threshold, multiply by to this LLR to this LLR among the bit that calculates the LLR attenuation respectively in the LLR decay portion 17 of this execution mode.

Figure 10 is the routine flow chart of action of the decision-feedback formula equalizer of this execution mode of expression.

This flow chart is with respect to the flow chart of Fig. 2, is different with the corresponding part of difference about above-mentioned LLR decay portion 17.

That is, in this execution mode also be, LLR decay portion 17 for the second time later DFE circulation (i=2 ..., N _It) in action, but in this execution mode, in each for the second time later DFE circulation, then the calculating of step S2 according to the LLR of LLR calculating part 13, before the processing of step S3, in step S21,17 couples of LLRL of LLR decay portion from 13 inputs of LLR calculating part _i(m; N) and defined threshold L _{Th, att}Magnitude relationship compare, whether judgement satisfies following conditions.

|L _i(m；n)|＜L _th.att

And, if | L _i(m; N) |＜L _{Th, att}, then get into step S3, otherwise skips steps S3 gets into step S4 (with reference to 32 among the figure).

That is, LLR decay portion 17 is through multiply by LLR attenuation { thereby the g} acquisition L that in previous circulation, calculates to the bit that satisfies above-mentioned condition _i' (m; N).

L _i′(m；n)＝g(m·log ₂(R)+n)L _i(m?；n)

Therefore, in this execution mode, SR generation portion 14 is in the DFE circulation first time (i=1), with reference to L ₁(m; N)=L (m; N) generate soft reproducing signals, circulate at for the second time later DFE (i=2 ..., N _It) in, | L _i(m; N) |＜L _{Th, att}Under the situation about setting up, with reference to L _i' (m; N)=g (mlog ₂(R)+n) L _i(m; N) generate soft reproducing signals, | L _i(m; N) |＜L _{Th, Att}Under the invalid situation, with reference to L ₁(m; N)=L (m; N) generate soft reproducing signals.

In addition, the inside that the function of 30 switch can be in LLR calculating part 13 among the figure also can be in the outside.

In addition, the inside that the function of 32 switch can be in LLR decay portion 17 among the figure also can be in the outside.

In addition; In this execution mode also be, preferably in the i time DFE circulation, in its finish time; Under the low all nd situation of bit of reliability; Do not carry out the DFE circulation of (after the i+1 time) after it, in the processing block that this DFE of the i time is input to afterwards grade (for example, hard decision, secondary equilibrium, error correction decoding etc.).

According to this execution mode; Except the effect that obtains through first execution mode; And then can also not be through making from the error propagation of symbol in advance but through noise under the situation that the bit reliability reduces, what do not make symbol in advance softly duplicates that employed log-likelihood ratio reduces in the generation.

In addition, in the indication shown in the processing sequence shown in the above-mentioned execution mode, can carry out based on program as software.General calculation machine system stores this program in advance, through reading in this program, thereby also can obtain and the identical effect of effect according to the decision-feedback formula equalizer of above-mentioned execution mode.The indication of in above-mentioned execution mode, recording and narrating; Program as computer is carried out is recorded in disk (floppy disk, hard disk etc.), CD (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD ± R, DVD ± RW etc.), semiconductor memory or the recording medium similar with it.So long as the recording medium that computer or embedded system can read, then its file layout can be any way.As long as computer from this recording medium read-in programme, makes the indication of recording and narrating in the CPU executive program based on this program, just can realize the identical action of decision-feedback formula equalizer with above-mentioned execution mode.Certainly, under computer is obtained the situation of situation or read-in programme of program, also can obtain or read in through network.

The MW (middleware) of the OS (operating system) that turns round on computers based on the indication that is installed to the program computer, the embedded system from recording medium in addition,, database management language, network etc. etc. can carry out each part handled that is used to realize this execution mode.

And then the recording medium in this execution mode is not limited to and computer or embedded system medium independently mutually, comprises also that download is stored through the program of transmission such as LAN, internet or the temporary transient recording medium of storage.

In addition, recording medium is not limited to one, and the situation of carrying out the processing this execution mode from a plurality of media also is contained in the recording medium in this execution mode, and the formation of medium can be any formation.

In addition; Computer in this execution mode or embedded system be used for based on the program that is stored in recording medium is carried out in this execution mode each handle, can be personal computer, microcomputer etc. by a device that constitutes, multiple arrangement be undertaken, and system that network connects etc. is any to be constituted.

In addition, the computer in this execution mode is not limited to personal computer, also comprises the calculation processing apparatus that is contained in the messaging device, microcomputer etc., is to utilize program to realize the general name of the unit of the function in this execution mode.

Although clear several embodiments of the present invention, but these execution modes point out as an example, are not intended scope of invention is limited.These execution modes can be implemented with other variety of ways, in the scope of the main idea that does not break away from invention, can carry out various omissions, displacement, change.These execution modes, its distortion and be contained in scope of invention, main idea in identical ground, be contained in the scope of invention and equalization thereof of claims record.

Claims (21)

1. a decision-feedback formula equalizer utilizes feedback filter to make the soft reproducing signals feedback based on the log-likelihood ratio of bit, it is characterized in that possessing:

Storage device is kept for making the input signal that can carry out repeatedly to the decision-feedback circulation of a plurality of symbols;

The log-likelihood ratio of a plurality of bits based on using the input signal of reading from above-mentioned storage device to deduct the signal of the output gained of above-mentioned feedback filter, calculates in log-likelihood calculations portion;

The log-likelihood ratio evaluating part based on the absolute value of above-mentioned log-likelihood ratio, is estimated the reliability of a plurality of bits, exports at least having transmitted the notation index that the symbol that is be evaluated as the low bit of reliability is confirmed;

The attenuation calculating part; Calculate the whole or corresponding log-likelihood ratio attenuation of a part of bit with following symbol; And for the circulation of next time decision-feedback and it is kept, wherein this symbol is to give the above-mentioned notation index that utilizes output and definite symbol brings the symbol of influence via above-mentioned feedback filter; And

Log-likelihood specific damping portion; The above-mentioned log-likelihood ratio that will be in this decision-feedback circulation be calculated by above-mentioned log-likelihood calculations portion is decayed according to the above-mentioned log-likelihood ratio attenuation that in previous decision-feedback circulation, is calculated by above-mentioned attenuation calculating part.

2. decision-feedback formula equalizer according to claim 1; It is characterized in that; The above-mentioned log-likelihood ratio attenuation that in previous decision-feedback circulation, calculates multiply by in above-mentioned log-likelihood specific damping portion to this log-likelihood ratio under the situation of absolute value less than defined threshold of the above-mentioned log-likelihood ratio that is calculated by above-mentioned log-likelihood calculations portion.

3. decision-feedback formula equalizer according to claim 1; It is characterized in that; The above-mentioned notation index that above-mentioned attenuation calculating part will be exported is as starting point, obtains whole notation indexs in the past that the tap of above-mentioned feedback filter exists, as the object that calculates above-mentioned log-likelihood ratio attenuation.

4. decision-feedback formula equalizer according to claim 1; It is characterized in that; The above-mentioned notation index that above-mentioned attenuation calculating part will be exported is as starting point; Obtain in the notation index in the past that the tap of above-mentioned feedback filter exists, begin successively the notation index till the preset upper limit number from the big notation index of the absolute value of its tap coefficient, as the object that calculates above-mentioned log-likelihood ratio attenuation.

5. decision-feedback formula equalizer according to claim 1; It is characterized in that; The above-mentioned notation index that above-mentioned attenuation calculating part will be exported is as starting point; Obtain the notation index of absolute value in the notation index in the past that the tap of above-mentioned feedback filter exists, its tap coefficient, as the object that calculates above-mentioned log-likelihood ratio attenuation greater than threshold value.

6. decision-feedback formula equalizer according to claim 1; It is characterized in that; The above-mentioned notation index that above-mentioned attenuation calculating part will be exported is as starting point; Obtain in the notation index in the past that the tap of above-mentioned feedback filter exists, from its time of delay little notation index begin successively the notation index till the preset upper limit number, as the object that calculates above-mentioned log-likelihood ratio attenuation.

7. decision-feedback formula equalizer according to claim 1; It is characterized in that; The above-mentioned notation index that above-mentioned attenuation calculating part will be exported is as starting point; Obtain in the notation index in the past that the tap of above-mentioned feedback filter exists, its time of delay is less than the notation index of threshold value, as the object that calculates above-mentioned log-likelihood ratio attenuation.

8. according to each described decision-feedback formula equalizer in the claim 2 to 7; It is characterized in that; Above-mentioned attenuation calculating part does not rely on the log-likelihood ratio attenuation of the tap coefficient ground computational rules of above-mentioned feedback filter to the bit that utilizes the symbol determined by the notation index in the above-mentioned past of obtaining to transmit.

9. according to each described decision-feedback formula equalizer in the claim 2 to 7; It is characterized in that; Above-mentioned attenuation calculating part in the bit that utilizes the symbol determine by the notation index in the above-mentioned past of obtaining to transmit, be mapped in to be evaluated as the bit on the identical IQ axle of the low bit of reliability and to be mapped in and be evaluated as the bit on the different IQ axle of the low bit of reliability with above-mentioned log-likelihood ratio evaluating part with above-mentioned log-likelihood ratio evaluating part, do not rely on the log-likelihood ratio attenuation of the tap coefficient ground computational rules of above-mentioned feedback filter respectively.

10. according to each described decision-feedback formula equalizer in the claim 2 to 7; It is characterized in that; The bit that the symbol that above-mentioned attenuation calculating part is determined by the notation index in the above-mentioned past of obtaining to utilization transmits; Based on the tap coefficient of above-mentioned feedback filter or with the value after its quantification, the log-likelihood ratio attenuation of computational rules.

11. according to each described decision-feedback formula equalizer in the claim 2 to 7; It is characterized in that; Above-mentioned attenuation calculating part in the bit that utilizes the symbol determined by the notation index in the above-mentioned past of obtaining to transmit, be mapped in to be evaluated as the bit on the identical IQ axle of the low bit of reliability and to be mapped in and be evaluated as the bit on the different IQ axle of the low bit of reliability with above-mentioned log-likelihood ratio evaluating part with above-mentioned log-likelihood ratio evaluating part; Value after quantizing based on the tap coefficient of above-mentioned feedback filter or with it respectively, the log-likelihood ratio attenuation of computational rules.

12. decision-feedback formula equalizer according to claim 8; It is characterized in that; Above-mentioned attenuation calculating part repeatedly becomes under the situation of the object that calculates above-mentioned log-likelihood ratio attenuation in same bits; A plurality of log-likelihood ratio attenuations that employing goes out this bit calculation long-pending is as the log-likelihood ratio attenuation to this bit.

13. decision-feedback formula equalizer according to claim 9; It is characterized in that; Above-mentioned attenuation calculating part repeatedly becomes under the situation of the object that calculates above-mentioned log-likelihood ratio attenuation in same bits; A plurality of log-likelihood ratio attenuations that employing goes out this bit calculation long-pending is as the log-likelihood ratio attenuation to this bit.

14. decision-feedback formula equalizer according to claim 10; It is characterized in that; Above-mentioned attenuation calculating part repeatedly becomes under the situation of the object that calculates above-mentioned log-likelihood ratio attenuation in same bits; A plurality of log-likelihood ratio attenuations that employing goes out this bit calculation long-pending is as the log-likelihood ratio attenuation to this bit.

15. decision-feedback formula equalizer according to claim 11; It is characterized in that; Above-mentioned attenuation calculating part repeatedly becomes under the situation of the object that calculates above-mentioned log-likelihood ratio attenuation in same bits; A plurality of log-likelihood ratio attenuations that employing goes out this bit calculation long-pending is as the log-likelihood ratio attenuation to this bit.

16. decision-feedback formula equalizer according to claim 8; It is characterized in that; Above-mentioned attenuation calculating part repeatedly becomes under the situation of the object that calculates above-mentioned log-likelihood ratio attenuation in same bits; Minimum value in a plurality of log-likelihood ratio attenuations that employing goes out this bit calculation is as the log-likelihood ratio attenuation to this bit.

17. decision-feedback formula equalizer according to claim 9; It is characterized in that; Above-mentioned attenuation calculating part repeatedly becomes under the situation of the object that calculates above-mentioned log-likelihood ratio attenuation in same bits; Minimum value in a plurality of log-likelihood ratio attenuations that employing goes out this bit calculation is as the log-likelihood ratio attenuation to this bit.

18. decision-feedback formula equalizer according to claim 10; It is characterized in that; Above-mentioned attenuation calculating part repeatedly becomes under the situation of the object that calculates above-mentioned log-likelihood ratio attenuation in same bits; Minimum value in a plurality of log-likelihood ratio attenuations that employing goes out this bit calculation is as the log-likelihood ratio attenuation to this bit.

19. decision-feedback formula equalizer according to claim 11; It is characterized in that; Above-mentioned attenuation calculating part repeatedly becomes under the situation of the object that calculates above-mentioned log-likelihood ratio attenuation in same bits; Minimum value in a plurality of log-likelihood ratio attenuations that employing goes out this bit calculation is as the log-likelihood ratio attenuation to this bit.

20. decision-feedback formula equalizer according to claim 1 and 2; It is characterized in that; Above-mentioned log-likelihood ratio evaluating part is under the situation of absolute value less than defined threshold of the above-mentioned log-likelihood ratio that is calculated by above-mentioned log-likelihood calculations portion, and the reliability that is evaluated as this bit is low.

21. decision-feedback formula equalizer according to claim 1 and 2; It is characterized in that; Above-mentioned log-likelihood ratio evaluating part is under the situation of absolute value less than the moving average of the absolute value of the log-likelihood ratio that in this decision-feedback circulation, calculates of the above-mentioned log-likelihood ratio that is calculated by above-mentioned log-likelihood calculations portion, and the reliability that is evaluated as this bit is low.

Images