CN1665232A - Method for fast realization of soft-demodulation and soft-modulation in high-order quadrature amplitude modulation - Google Patents

Abstract

The invention discloses a method for fast implementing soft demodulation and soft modulation in the high-order quadrature amplitude modulation (QAM), relating to a mobile communication method of high speed data transmission adopting the QAM. And its soft demodulation implementing method adopts the fast table looking-up method, containing the below steps: firstly generating and storing the look-up table used by the soft demodulation, normalizing receipt signal and noise variance according to channel corresponding coefficients, quantizing the normalized receipt signal and making table look-up or table look-up combined with interpolation for the quantized signal, multiplying the value obtained after the table look-up by the normalized noise variance to obtain the needed bit likehood ratio; the soft modulating method includes the following steps: making converting calculation on the bit likehood ratio and according to the linear relation between character and bit mapped by natural QAM, and according to mean value and variance of the bit, calculating the mean value and variance of the character.

Description

The Fast implementation of soft demodulation soft system in the high-order orthogonal amplitude modulation(PAM)

Technical field

The present invention relates to a kind ofly, relate in particular to the high speed data transfer method of mobile communication of a kind of employing high-order QAM (quadrature amplitude modulation) modulation by using a plurality of send/receive antennas to come the wide-band mobile communication system of transmitting high speed data.

Background technology

Future broadband wireless communication systems not only requires high speed data transfer, and demanding spectrum utilization efficiency.High order modulation is a kind of effective means that wireless communication system is realized the high availability of frequency spectrum.And strong error control codings such as Turbo code, LDPC sign indicating number can be strengthened the antijamming capability of communication system widely, so the Bit Interleave coded modulation becomes an important baseband processing method of wireless communication system.In this type systematic, decoder need come the bit likelihood ratio of self-detector (demodulator), and especially under the situation of high order modulation, demodulation still is the process of a relative complex.

In some communication systems, comprise wired communication system and wireless communication system, adaptive coding and modulating is to improve an important technology of system works efficient.The operation principle of adaptive coding and modulating is according to channel condition, adjusts modulation and coded system adaptively to improve the power efficiency and the spectrum efficiency of system.But at receiving terminal, the demodulation method of various modulation is not quite similar, and complexity is difference to some extent also, and traditional demodulation method is unfavorable for realizing adaptive demodulation flexibly.The present invention uses the method table look-up or table look-up in conjunction with heterogeneous formula interpolation to realize quick realization to soft demodulation having reduced complexity effectively.

In a lot of iterative detection decoding systems, need mean variance, can be used for having the MMSE detection of prior information or the detection of Interference Cancellation according to the soft information reconstruction signal of decoder feedback.But when order of modulation is too high, from the complexity of bit likelihood ratio reconstruction signal mean variance still than higher.

Summary of the invention

Technical problem: the objective of the invention is to propose the low complex degree implementation method of soft demodulation in a kind of high-order orthogonal amplitude modulation(PAM) (promptly calculating the bit likelihood ratio) and soft system (promptly according to bit likelihood ratio compute sign statistic), be suitable for use in the qam mode of inphase/orthogonal according to signal measured value and noise variance.

Technical scheme: the implementation method of the soft demodulation in the high-order orthogonal amplitude modulation(PAM) of the present invention in the Fast implementation of soft demodulation soft system adopts fast zoom table implementation method, and it comprises following steps:

Step 1), generate and preserve the used look-up table of soft demodulation earlier,

Step 2), carry out normalized with noise variance to received signal, according to the channel corresponding coefficient

Step 3), the received signal of normalized is quantized,

Step 4), the signal after quantizing is tabled look-up or tabled look-up in conjunction with interpolation, the likelihood ratio that promptly obtains required bit with normalized noise variance on duty that obtains after tabling look-up.

Soft system method in the high-order orthogonal amplitude modulation(PAM) comprises following steps:

Step 4), the bit likelihood ratio is carried out conversion Calculation,

Step 5), according to the linear relationship of natural QAM mapped symbol and bit, according to the average of bit and the average and the variance of variance compute sign, if this QAM mapping mode is the nature mapping, then step 4) can be omitted.

Normalized comprises following steps:

Step 1), signal normalization: with the received signal value divided by the channel gain coefficient,

Step 2), variance normalization: with noise variance value divided by the channel gain coefficient.

The Fast implementation of tabling look-up in conjunction with interpolation of soft demodulation comprises following steps:

1), carries out normalized with noise variance to received signal, according to the channel corresponding coefficient

2), normalized received signal is quantized, estimate quantization error simultaneously,

3), the signal after quantizing is tabled look-up, obtain the error term coefficient simultaneously, it revised with the method for interpolation,

4), revised value multiply by normalized noise variance again and promptly obtains required bit likelihood ratio.

When the signal after quantizing is tabled look-up, obtain the error term coefficient simultaneously, the interpolation method in the method for interpolation it being revised is to use linear interpolation, or reaches the more interpolation of high-order with quadratic interpolation.

1. high-order QAM is modulated soft demodulation

Modulation is one group of Bit data a ₀a ₁A _2n-1Be mapped to complex symbol s, we use the QAM modulation of inphase/orthogonal, i.e. the real part s of s usually _RWith imaginary part s ₁Respectively by bit a ₀a ₁A _N-1And a _nA _2n-1Mapping forms.Signal through channel equalization or interference-cancelled signals detection etc., always can be expressed as the equivalent transmission channel of signal at receiving terminal at transmission and channel (comprising various transmission meanss and various channel) through communication system:

R=ρ s+n [1] wherein, ρ represents the gain coefficient of real signal terms, n represents remaining interference and noise.Usually the decoder of error control coding needs the likelihood ratio of bit of demodulator as input.We can separate the real part of signal and imaginary part, and formula [1] becomes:

Re(r)＝ρ·Re(s)+Re(n)

[2]

Im(r)＝ρ·Im(s)+Im(n)

According to following formula, we can do demodulation independently to two paths of signals and calculate likelihood ratio.We are example with the in-phase component, the process of the demodulation of deriving.If bit a ₀A _N-1Be mapped to real symbol s, corresponding real received signal is r not, gain coefficient ρ, and multiple noise variance is σ ²The likelihood ratio of definition demodulator output is

$L\left(a_i\right|r)=\ln \left(\frac{P(a_i=+1|r)}{P(a_i=-1|r)}\right)=\ln \left(\frac{\underset{s:a_i=+1}{\mathrm{\Σ}}P\left(r\right|s)\·\underset{j\≠i}{\mathrm{\Π}}P\left(a_j\right)}{\underset{s:a_i=-1}{\mathrm{\Σ}}P\left(r\right|s)\·\underset{j\≠i}{\mathrm{\Π}}P\left(a_j\right)}\right)+L\left(a_i\right)---\left[3\right]$

L (a in the following formula _j) and P (a _j) represent bit a respectively _jPrior information and prior probability.Be without loss of generality, suppose that bit is 0 and 1 etc. general, by Gauss's hypothesis of noise, we can obtain again:

$L\left(a_i\right|r)=\ln \left(\frac{\underset{\mathrm{\α}:a_i=+1}{\mathrm{\Σ}}\exp \left(\frac{|r-\mathrm{\ρ}\·\mathrm{\α}{|}^2}{{\mathrm{\σ}}^2}\right)}{\underset{\mathrm{\α}:a_I=-1}{\mathrm{\Σ}}\exp \left(\frac{{|r-\mathrm{\ρ}\·\mathrm{\α}|}^2}{{\mathrm{\σ}}^2}\right)}\right)---\left[4\right]$

According to $\ln (e^{x_1}+e^{x_2})=\max *(x_1,x_2)=\max (x_1,x_2)+\ln (1+e^{-|x_1-x_2|}),$ Obtain:

$L\left(a_i\right|r)=\underset{\mathrm{\α}:a_i=+1}{\max }*\{-\frac{{|r-\mathrm{\ρ}\·\mathrm{\α}|}^2}{{\mathrm{\σ}}^2}\}-\underset{\mathrm{\α}:a_i=-1}{\max }*\{-\frac{{|r-\mathrm{\ρ}\·\mathrm{\α}|}^2}{{\mathrm{\σ}}^2}\}---\left[5\right]$

Usually we calculate the likelihood ratio of each bit with formula [5], finish the function of demodulation.

2. high-order QAM is modulated the realization fast of tabling look-up of soft demodulation

We see, calculate the bit likelihood ratio with formula [5], and complexity is exponential increase with order of modulation, and operand is very big.We use max (x ₁, x ₂) ≈ max* (x ₁, x ₂) formula [5] is simplified, further can obtain:

$L\left(a_i\right|r)=\underset{\mathrm{\α}:a_i=+1}{\max }\{-\frac{{|r-\mathrm{\ρ}\·\mathrm{\α}|}^2}{{\mathrm{\σ}}^2}\}-\underset{\mathrm{\α}:a_i=-1}{\max }\{-\frac{{|r-\mathrm{\ρ}\·\mathrm{\α}|}^2}{{\mathrm{\σ}}^2}\}$

$=\frac{{\mathrm{\ρ}}^2}{{\mathrm{\σ}}^2}\·\left\{\underset{\mathrm{\α}:a_i=+1}{\max }\right\{-|r/\mathrm{\ρ}-\mathrm{\α}{|}^2\}-\underset{\mathrm{\α}:a_i=-1}{\max }\{-{|r/\mathrm{\ρ}-\mathrm{\α}|}^2\}\}---[6]$

$=\frac{{\mathrm{\ρ}}^2}{{\mathrm{\σ}}^2}\·f(r/\mathrm{\ρ})$

Wherein, f (x) can realize that the soft demodulation of different modulation systems can be made different tables with the method for tabling look-up.If according to the p bit quantization, f (x) realizes doing one 2 with tabling look-up so with normalized received signal r '=r/ ρ for we ^pThe table of length.Different modulation systems only need be made different tables in advance, is placed in the memory, is worth the soft demodulation that can realize the different modulating mode accordingly according to the different modulating form that sends signal from corresponding table.Fig. 1 has described the basic process of the soft demodulation of high order modulation, at first will carry out normalization to received signal, then it is tabled look-up after quantizing, and can obtain the bit likelihood ratio after multiply by corresponding coefficient again.

We can see that here the size of table is exponential increase with quantizing exponent number.If the quantification exponent number of r ' is higher, Biao size will increase greatly so, to such an extent as to the raising of complexity.If quantize to table look-up and comparatively accurate r ' is carried out low order, can cause the loss of performance again, for the performance that obtains with low complex degree, we use to table look-up and in conjunction with the method for polynomial interopolation soft demodulating method are improved, its basic principle is that r ' is carried out quantizing according to the precision of look-up table, utilizes the value after polynomial interopolation obtains upgrading after tabling look-up.

Fig. 2 tables look-up to realize the fundamental block diagram of soft demodulation in conjunction with linear interpolation the following narration of cardinal principle: if r ' precision is higher than the quantification exponent number and drops between adjacent two point of quantification, with linear interpolation f (r ') ≈ f (q (r '))+f ₁(q (r ')) (r '-q (r ')) obtain being worth more accurately.Similarly, also can carry out 2 rank or the interpolation of high-order more.Quantizer can adopt further the dwindling of size of its acquisition table of method of nonlinear quantization.

3. high-order QAM is modulated soft system

In some iterative detection decoding methods, require to carry out the reconstruction of average and variance according to the bit likelihood ratio of decoder feedback through regular meeting.Just be+1 or-1 probability according to each bit, calculating and sending signal is the probability of each symbol in the assemble of symbol, thereby calculates the average and the variance of signal.Mean variance according to bit likelihood ratio signal calculated is as follows:

$E\left[s\right]=\underset{\mathrm{\α}}{\mathrm{\Σ}}\mathrm{\α}\·P(s=\mathrm{\α})---\left[7\right]$

$\mathrm{cov}\left[s\right]=\underset{\mathrm{\α}}{\mathrm{\Σ}}{\left|\mathrm{\α}\right|}^2\·P(s=\mathrm{\α})-E{\left[s\right]}^2---\left[8\right]$

If symbol is by bit d ₀, d ₁... d _Mc-1Mapping forms, then:

$P(s=\mathrm{\α})=\underset{k=0}{\overset{M_C-1}{\mathrm{\Π}}}\frac{1}{2}[1+{\tilde{d}}_k\·\tanh (L\left(d_k\right)/2)]---\left[9\right]$

Wherein, This bit was worth accordingly when expression s was symbol.

Can see, under the situation of high order modulation, need to calculate the probability of each symbol, the average and the variance of ability reconstruction signal, complexity is along with the bit number of each symbol is exponential increase.Complexity rises widely when order of modulation is high, and the difficulty that makes hardware realize increases greatly.

The present invention is directed to the QAM modulation of nature mapping or gray mappings, the mean variance computational methods of having invented low complex degree.This method is regarded general QAM modulation as two stages: the first step is a linearity of bit sequence or the process of non-linear block encoding, such as identical transformation is arranged, Gray codes etc., second step was a Linear Mapping device, i.e. the expression of the modulation symbol bit sequence linearity that can be obtained by the first step.Accordingly, soft system process can be divided into two steps, the first step is carried out the conversion of bit likelihood ratio with the log-likelihood ratio of bit according to the rule of conversion, and the bit log-likelihood ratio of second step after by conversion obtains average and the variance according to the reconstruction of relations signal of Linear Mapping.This method has avoided calculating according to bit log-likelihood ratio this step of probability of all symbols, can reduce the complexity that mean variance calculates effectively.

Beneficial effect: advantage of the present invention is to use to quantize/table look-up to be implemented in fast under the situation that does not reduce performance and makes complexity reduce widely, especially under the situation of high order modulation.And only need change in the table value accordingly to different modulation systems, flexibility strengthens widely.And soft system is separated into the conversion of bit likelihood ratio and two processes of statistic calculating make the computation complexity of sign statistic reduce effectively.

The fast zoom table of the soft demodulation of high-order QAM among the present invention realizes having following advantage:

1. reduced the complexity that QAM modulates soft demodulation widely, especially under the situation of high-order QAM modulation.

Property retention suffer a loss hardly.

3. use under the situation of different modulation systems at transmitting terminal, the soft demodulation of receiving terminal can be carried out under same framework and have identical complexity, helps realizing the reception of Adaptive Modulation.

4. make that the non-regular QAM mapping of high complexity has with the identical complexity of regular QAM mapping originally, make the selection of mapping mode more diversified.

The low complex degree that high-order QAM symbol mean variance among the present invention calculates is realized having the following advantages:

Greatly reduced the complexity of mean variance computational process.

2. shining upon among the QAM naturally, complexity is reduced to linear growth with bit number by exponential increase.

Soft demodulation realizes fast with quantification/look-up method among the present invention, not only effectively reduces complexity, and the realization of different modulating mode only need to change mapping table just passable, can realize Adaptive Modulation comparatively neatly.This method is suitable for use in the communication system of any qam mode, especially in inphase/orthogonal QAM system.Soft system of the present invention is that the calculating of sign statistic (generally including the average and the variance of symbol) is divided into the conversion of bit likelihood ratio and mean variance calculates two steps, this method can effectively reduce complexity, is suitable for use in the qam mode of inphase/orthogonal.

Description of drawings

Fig. 1 tables look-up to realize the fundamental block diagram of soft demodulation.

Fig. 2 tables look-up to realize the fundamental block diagram of soft demodulation in conjunction with linear interpolation.

Fig. 3 is a gray mappings soft system block diagram.

Embodiment

The implementation method of soft demodulation adopts fast zoom table implementation method, and it comprises following steps:

Step 1), according to formula [6], calculate earlier, generate and preserve the used look-up table of soft demodulation according to certain quantization digit.

Step 2), carry out normalized with noise variance to received signal, as shown in Figure 1, received signal r obtains the ρ reciprocal of normalization received signal r '=r/ ρ and normalization noise variance at first by channel gain coefficient ρ normalization according to the channel corresponding coefficient ²/ σ ²

Step 3), the received signal of normalized is quantized, quantification can be used uniform quantization and non-uniform quantizing.

Step 4), the signal q (r ') after quantizing is tabled look-up, the inverse with normalized noise variance on duty that obtains after tabling look-up promptly obtains the log-likelihood ratio of soft demodulation corresponding bits, can or send into decoder for decoding for judgement.

The Fast implementation of tabling look-up in conjunction with interpolation of soft demodulation comprises following steps:

1), carries out normalized with noise variance to received signal, according to the channel corresponding coefficient

2), normalized received signal is quantized, estimate quantization error simultaneously,

3), the signal after quantizing is tabled look-up, obtain the error term coefficient simultaneously, it revised with the method for interpolation,

4), revised value multiply by normalized noise variance again and promptly obtains required bit likelihood ratio.

When the signal after quantizing is tabled look-up, obtain the error term coefficient simultaneously, the interpolation method in the method for interpolation it being revised is to use linear interpolation, or reaches the more interpolation of high-order with quadratic interpolation.

High-order QAM is soft separates the quick implement device of application form/interpolation as shown in Figure 2, and received signal r obtains normalization and receives coefficient r ' at first by channel gain coefficient ρ normalization, then it is quantized, and quantification can be used uniform quantization and non-uniform quantizing.In look-up table, find corresponding value according to the value q (r ') after quantizing, calculate quantization error r '-q (r ') then, and in another look-up table, find corresponding coefficient f ₁(q (r ')) is with this coefficient f ₁(q (r ')) and quantization error are revised the value of f (q (r ')).

In actual hardware was realized, data all were that form with quantized data exists in hardware usually.If the number of significant digit of received signal and corresponding coefficient lower (close) own with the look-up table quantization digit, look-up table can't be too big so, therefore first kind of mode is proper, if the number of significant digit higher (much larger than the look-up table quantization digit) of received signal and corresponding coefficient itself, quantification can cause the loss of data precision so by force, in this case, use to table look-up/method of interpolation combination can overcome above-mentioned shortcoming, obtain high-precision performance with lower complexity.

Soft system method in the high-order orthogonal amplitude modulation(PAM) comprises following steps:

Step 4), the bit likelihood ratio is carried out conversion Calculation,

Step 5), according to the linear relationship of natural QAM mapped symbol and bit, according to the average of bit and the average and the variance of variance compute sign, if this QAM mapping mode is the nature mapping, then step 4) can be omitted.

Be embodiment with mapping and gray mappings naturally below, specify the job step that soft system (promptly according to bit likelihood ratio compute sign statistic) is installed.

1, mapping naturally

Common natural mapped bits b _k(get ± 1) can describe with a linear representation with the relation between the symbol s, as follows:

$s=\underset{k=0}{\overset{M_C-1}{\mathrm{\Σ}}}{c}_k\·b_k---\left[10\right]$

Then owing to the independence between the bit, the average of symbol and variance can be expressed as follows:

$E\left[s\right]=\underset{k=0}{\overset{M_C-1}{\mathrm{\Σ}}}{c}_k\·E\left[b_k\right]---\left[11\right]$

$\mathrm{cov}\left[s\right]=\underset{k=0}{\overset{M_C-1}{\mathrm{\Σ}}}{\left|c_k\right|}^2\·E\left[b_k^2\right]-E{\left[s\right]}^2---\left[12\right]$

E[b wherein _k]=tanh (L (b _k)/2), $E\left[b_k^2\right]=1.$

2, gray mappings

Because gray mappings can not be described the relation between bit and the symbol with a linear relationship, but it can be separated into a Gray encoder and a natural mapper, we can change likelihood ratio by Gray encoder, again by shining upon the average and the variance of compute sign naturally.

Fig. 3 has described the mean variance calculation process of gray mappings, mainly is divided into two parts, at first its bit likelihood ratio is changed according to Gray code, and then according to formula [10], [11], [12] calculate the average and the variance of nature mapping.

Usually the bit sequence Gray code can be expressed with following formula:

$\left\{\begin{array}{c}{a}_0=b_0\\ a_k=b_k\⊕b_{k-1}\⊕1k\≠0\end{array}\right.---\left[13\right]$

Represent b in the formula _kBit before the Gray code, a _kBit after the expression Gray code.According to this relation, we can calculate the relation between the likelihood ratio

$\left\{\begin{array}{c}L\left(a_0\right)=L\left(b_0\right)\\ L\left(a_k\right)=1/2\·\max *\{-L\left(b_k\right)-L\left(b_{k-1}\right),L\left(b_k\right)+L\left(b_{k-1}\right)\}\\ -1/2\·\max *\{-L\left(b_k\right)+L\left(b_{k-1}\right),L\left(b_k\right)-L\left(b_{k-1}\right)\}k\≠0\end{array}\right.---\left[14\right]$

In the formula $\max *(x_1,x_2)=\max (x_1,x_2)+\ln (1+e^{-|x_1-x_2|}),$ Can use max (x ₁, x ₂) approach.Mathematical function tanh (x) realizes with the mode of tabling look-up in engineering usually.

The quick realization of the high-order QAM soft system among the present invention has reduced the complexity of rebuilding symbol average and variance according to the bit likelihood ratio widely, especially when order of modulation height or use Linear Mapping.Its performance near or equal optimum computational methods.

The fast zoom table of the soft demodulation of high-order QAM among the present invention realizes being applied in the system of all kinds of use high-order QAM modulation, be fit to be applied to all kinds of use high-order QAM mapping communication systems, comprise wired communication system and wireless communication system, be suitable for various transmission means communication systems, comprise transmission systems such as single antenna, many antennas, single carrier, multicarrier, both can be used for regular mapping graph, also can be used for the mapping graph of non-rule.

The quick realization of the high-order QAM soft system among the present invention can be applied in the modulation of all kinds of use high-order QAMs and need the system of compute sign average and variance, comprises the iterative detection decoding that uses linear detector in the interference cancellation multi-user detector, multiaerial system of MMSE in the MMSE Turbo equilibrium that has prior information, the cdma system or matched filtering etc.

Claims (4)

1. the Fast implementation of soft demodulation soft system in the high-order orthogonal amplitude modulation(PAM) is characterized in that the implementation method of soft demodulation adopts fast zoom table implementation method, and it comprises following steps:

Step 1), generate and preserve the used look-up table of soft demodulation earlier,

Step 2), carry out normalized with noise variance to received signal, according to the channel corresponding coefficient

Step 3), the received signal of normalized is quantized,

Step 4), the signal after quantizing is tabled look-up or tabled look-up in conjunction with interpolation, the likelihood ratio that promptly obtains required bit with normalized noise variance on duty that obtains after tabling look-up;

Soft system method in the high-order orthogonal amplitude modulation(PAM) comprises following steps:

Step 4), the bit likelihood ratio is carried out conversion Calculation,

Step 5), according to the linear relationship of natural QAM mapped symbol and bit, according to the average of bit and the average and the variance of variance compute sign, if this QAM mapping mode is the nature mapping, then step 4) can be omitted.

2. the Fast implementation of soft demodulation soft system in the high-order orthogonal amplitude modulation(PAM) according to claim 1 is characterized in that normalized comprises following steps:

Step 1), signal normalization: with the received signal value divided by the channel gain coefficient,

Step 2), variance normalization: with noise variance value divided by the channel gain coefficient.

3. the Fast implementation of soft demodulation soft system in the high-order orthogonal amplitude modulation(PAM) according to claim 1 is characterized in that the Fast implementation of tabling look-up in conjunction with interpolation of soft demodulation comprises following steps:

1), carries out normalized with noise variance to received signal, according to the channel corresponding coefficient

2), normalized received signal is quantized, estimate quantization error simultaneously,

3), the signal after quantizing is tabled look-up, obtain the error term coefficient simultaneously, it revised with the method for interpolation,

4), revised value multiply by normalized noise variance again and promptly obtains required bit likelihood ratio.

4. the Fast implementation of soft demodulation soft system in the high-order orthogonal amplitude modulation(PAM) according to claim 3, when it is characterized in that the signal after quantizing tabled look-up, obtain the error term coefficient simultaneously, interpolation method in the method for interpolation it being revised is to use linear interpolation, or with the quadratic interpolation and the interpolation of high-order more.

Images