Device and method for reducing bit error coefficients and frame error coefficients using turbo-decoding in digital communication system
 Device and method for reducing bit error coefficients and frame error coefficients using turbo-decoding in digital communication system / 2263397Proposed device has composite turbo-code decoder incorporating first adder that computes logarithmic ratio of received code character similarity by calculating difference between character probability equal to 1 and character probability equal to 0 in arbitrary state of turbo-coding lattice; second adder that functions to add transfer information and a priori code character information; third adder that computes difference between outputs of first and second adders as peripheral information; first multiplier that multiplies result obtained from output of third adder by predetermined weighting coefficient; correcting value computer that computes correcting value using difference between best metric and second best metric of code character; fourth adder that adds correcting value to result obtained from first multiplier output. |
 Turbo-decoding device and method / 2273093High-speed buffer memory is disposed between receiver buffer memory and turbo-decoding device and operates at same frequency as turbo-decoding device; decoding device reads out information bits stored in receiver buffer memory through high-speed buffer memory, enters delay for read-out information-bits for time required by turbo-decoding device, and then sends delayed information bits to decoding device including flexible input and flexible output of data incorporated in turbo-decoding device; information bit output is effected by receiver buffer memory at operating or clock frequency of turbo-decoding device. |
 Mode and an arrangement for definition of logarithmical likelihood ratio with preliminary coding / 2304352The technical result is reduction of the effect of multiplication of errors achieved because multitude of signal elements is received. At that the signal element contains the totality of modulation symbols out of the totality of coded bits, the first subset of signal elements for which the bit has the first meaning and the second subset of signal elements for which the bit has the second meaning are defined. At that the first and the second subsets are signal elements out of an extended signal group. The probability that the bit equals the first meaning or the second meaning depending from the received signal element is defined, then the symbol of flexible decision of possibility that the bit equals the first meaning or the second meaning is defined. At that the symbols of flexible decision may be represented by the logarithmical likelihood ratios. |
 High correcting capacity decoder / 2438252High correcting capacity decoder has a receiving unit, a codeword storage, a unit for correcting deletions, a signal analyser and an estimate storage, as well as a random number sensor, a unit for generating deletions, an estimate ordering unit, an equivalent code unit, a linearity control unit, a comparator and inversion unit. Due to randomisation of deletion solutions, the number of false deletions is reduced and by using transformation of equivalent codes, full use of redundancy introduced into the code is ensured. |
 Device to decode reed-solomon codes / 2441318Device comprises a buffer data memory, a unit of syndromes calculation, a Galois processor, a unit of discrete Fourier transformation, a unit for searching of weight error positions t+1, a unit of symbol positions sorting, a unit of error values calculation, a summator of Galois field elements. |
 Method for soft-decision decoding systematic block codes / 2444127For all resolved code combinations of a systematic (n,k)-code, based on f most significant bits thereof, the cluster number is determined, the value of f lying in the range 1<f≤k/2. Vector V is obtained and the accuracy of receiving the cluster number is estimated based on reliability gradations and the parity bit. In case of a positive solution, the (n-f,k-f)-code is then processed using ordered statistics by first creating a recourse vector Wyk by multiplying f bits by the first f rows of the code generating matrix G and temporary removal therefrom of cluster bits, after which vector Wyk is subtracted from vector V, for which cluster bits are also neglected, while assigning the least significant bit of the newly formed vector Vyr the lowest reliability gradation estimate. The remaining symbols arranged in descending order of their estimates. The error vector E for the shortened code is found using an equivalent code and after subtraction thereof, the decoder performs bitwise summation of vectors Wyk, Vyk and E, forming the transferred vector by repositioning the most significant bits of the cluster number, where is parity check of the cluster number is not performed, the decoder inverts the most unreliable bit from the checked bits. |
 Soft decision apparatus and method / 2458462Soft decision apparatus and method are used to output a soft decision value for each bit of each symbol, used when decoding each symbol, as a value which corresponds to a function value obtained by applying a predetermined function for each bit to the sampling value of each symbol in accordance with the demodulated signal such that probability distribution of the sampling value in each symbol point is a Gaussian distribution. The function for each bit approaches a curve which expresses probability that each bit is equal to 1 or 0 for the sampling value of each symbol of the demodulated signal, and is also defined using a quadratic function. |
 Decoding device with soft decisions for double-stage cascade code / 2485683Device comprises a correction device, a circuit of bit quality identification, a demodulator, a circuit of symbol quality detection and selection of least valid symbols, a circuit of code cycle synchronisation with integrated soft and hard decisions, a circuit of error vectors generation for least valid symbols, a unit of summators by module two, a unit of decoders of a BCH code, a circuit for detection of least weight, a controller of PC code decoder. |
FIELD: advanced correction of bit and frame error coefficients using turbo-decoding in communication system.
SUBSTANCE: proposed device has composite turbo-code decoder incorporating first adder that computes logarithmic ratio of received code character similarity by calculating difference between character probability equal to 1 and character probability equal to 0 in arbitrary state of turbo-coding lattice; second adder that functions to add transfer information and a priori code character information; third adder that computes difference between outputs of first and second adders as peripheral information; first multiplier that multiplies result obtained from output of third adder by predetermined weighting coefficient; correcting value computer that computes correcting value using difference between best metric and second best metric of code character; fourth adder that adds correcting value to result obtained from first multiplier output.
EFFECT: improved turbo-decoding algorithm using less intricate hardware.
8 cl, 16 dwg
The present invention relates, generally, to a device and method advanced error correction (EYE) in the digital communication system and, in particular, to a device and method of turbodecoding.
In General, turbocode used for high-speed data transmission, especially in 1xEV-DO (evolution, data only) or 1xEV-DV (evolution data and speech). The Burr (Berrou) and others have proposed turbocode in 1993 Turbocode is a parallel connection of two constituent recursive systematic convolutional (PCC) encoders with a random interleaver between them. Thus, the turbo code obtained by coding information bits and perenesennyj information bits in the composite RCA-coders. Turbodecoding includes a serial connection of two component decoders, each for iterative decoding, exchanging his external information with the other constituent decoder. There are three algorithms used for each constituent decoder algorithm the Log-mAh, Max-Log-MAV and the Viterbi algorithm with soft decision (AMR).
The algorithm Log-MAV is an exercise in the logarithmic region algorithm MAV (algorithm of maximum a posteriori probability), which is optimal for decoding a data word in the lattice. Algorithm Max-Log-MAV easily obtained the C algorithm the Log-MAV by approximation calculation of metrics. Despite the advantage of simple implementation compared to the algorithm the Log-MAV, the algorithm Max-Log-MAV leads to performance deterioration when the receiver is probably accurate signal-to-noise ratio (SNR).
Algorithm Log-MAV is calculated metric status and logarithmic likelihood ratio (LOP). Metric status α and β state (s and s') in the lattice at time k time decoding are recursive relation, expressed as
(1)
where γ - the branch metric defined by the symbol received on the channel. Using the metric of the state and the branch metric, LOP the k-th symbol is calculated by the formula
,
where(2)
In equation (2) Mn(i) represents the i-th metric in descending order metrics (log(αk-1(s')γk(s',s)βk(s))) for the information of the symbol n (0 or 1) in the set of States (s, s') at time k time. Therefore, M0(0) and M1(0) are the best metrics for information symbols 1 and 0 at time k time, and fcrepresents the correction value determined by the difference between the best metric and other metrics for each information symbol. Therefore, LOP is updated using the differential is the best metrics between information symbols 0 and 1 at time k and time correction value of f c.
So, the algorithm Log-MAV generates all metrics state in the lattice for each constituent decoder according to equation (1) and calculates the LOP code symbol in the grid, using its metric condition on equation (2). Each constituent decoder takes the external information obtained from LOP, on the other constituent decoder for iterative decoding. So you turbodecoding.
Algorithm Max-Log-mAh is a simplified version of the algorithm the Log-MAV by managing the metrics are calculated condition equation (1) to maximum efficiency, expressed as
(3)
Similarly, the LOP of the k-th symbol decoding is simply calculated by maximum efficiency. LOP is updated using only the difference between the best metric, assuming fc0. Thus,
(4)
So, the algorithm Max-Log-MAV searches for all metrics state in the lattice for each constituent decoder through maximum efficiency according to equation (3) and calculates the LOP code symbol in the grid, using the difference between the best metric between information symbols 0 and 1 in equation (4). The external information received from the LOP, served n the other constituent decoder for iterative decoding. So you turbodecoding.
The so-called algorithm, Max-Log-MAV with feedback factor (KOS) takes into account an additional factor, obtained from LOP, calculated according to equation (4), to improve the performance of the decoding algorithm, Max-Log-MAV. The weighting factor is multiplied as the feedback factor, is approximately 0,588235 and applies only to external information from the second constituent decoder.
Since the algorithm the Log-MAV is an exercise in the logarithmic region of the optimal character of the decoding algorithm MAV, it works like algorithm MAV. However, when the algorithm is Log-MAV is the hardware, the function log(1+e-Δ)defining each metric should be implemented in hardware or in the form of a reference table. Algorithm Max-Log-MAV, on the other hand, does not require a reference table, but it works worse than the algorithm the Log-MAV. Advantages and disadvantages of the algorithm the Log-MAV and algorithm, Max-Log-MAV following.
(1) Algorithm the Log-MAV: as it is the optimal character of the decision making algorithm is the best algorithm turbodecoding. However, the implementation of log(1+e-Δ) increases the complexity of the hardware. In addition, log(1+e-Δis a nonlinear function, and, therefore, require an accurate estimate of the SNR of the received symbol for the calculation of branch metrics by which is determined by the Δ. If the evaluation of the SNR includes errors, then this mismatch SNR significantly degrades performance.
(2) the Algorithm Max-Log-MAV: Calculate log() is not required to calculate the metrics, since all metrics are calculated by maximum efficiency. Therefore, there is a problem of high complexity hardware that occurs when the algorithm the Log-MAV. In addition, the calculation of metrics through maximum efficiency eliminates the need for a non-linear function log(1+e-Δ), which means that there are no problems associated with misalignment of the SNR. However, since the algorithm Max-Log-MAV is an approximation algorithm Log-MAV, it works about 0.3 to 0.4 dB worse than the algorithm the Log-MAV.
As described above, the algorithm Log-MAV and the algorithm, Max-Log-MAV cause increased hardware complexity and performance deterioration in the quality of their respective shortcomings.
The present invention is therefore to create a device and method turbodecoding that work better than the algorithm, Max-Log-MAV when turbodecoding.
Another object of the present invention is to provide devices and str is both turbodecoding, which are less complex than the algorithm the Log-MAV.
The above problem is essentially solved by composite decoder for decoding turbo code and its constituent decoding. The best metric and the second best metric values are calculated for the received code symbol in an arbitrary state lattice turbodecoding during turbodecoding code symbol. Calculated external information needed to turbodecoding code symbol. Calculates a difference between the external information and the difference between the best metric and the second best metric. Updated LOP code symbol by multiplying the calculated difference by a predetermined weighting factor and a decision on the value of the code symbol.
External information is calculated using the difference between the two metrics: input character, reflecting the SNR and the a priori information of the input character.
The weighting factor is less than 1 and approaches 1. Preferably it is more than 0,588235. Preferably it is 1/2+1/4+1/16.
If the SNR can be estimated accurately, then the weighting factor is calculated using the logarithmic function. If the SNR cannot be estimated accurately, then the weighting factor is calculated using the approximated linear fu the options.
In the composite decoder for decoding turbo code first adder calculates LOP received code symbol by calculating the difference between the probability of a code symbol, equal to 1, and the probability of a code symbol 0 in an arbitrary state lattice turbodecoding during turbodecoding code symbol. The second adder adds the information transfer and the a priori information of the code symbol. The third adder calculates the difference between the outputs of the first and second adders as external information. The first multiplier multiplies the output of the third adder by a predetermined weighting factor as a ratio of the feedback. The transmitter correction values calculates the correction value using the difference between the best metric and the second best metric code symbol. The fourth adder adds the correction value with the output of the first multiplier.
The transmitter correction value includes a fifth adder for calculating the difference between the best metric and the second best metric for the information of the symbol 0 as the value of the code symbol, the sixth adder for calculating the difference between the best metric and the second best metric for the information symbol 1 as the value code is about character and reference table to store based on a logarithmic function of the correction values for the outputs of the fifth and sixth adders and to output the correcting values for the outputs of the fifth and sixth adders. The transmitter corrective value additionally includes a seventh adder for calculating the difference between the correction value, the second multiplier for multiplying the output of the seventh adder for a predetermined weighting factor, the eighth adder for calculating the difference between the outputs of the fifth and sixth adders, a third multiplier for multiplying the output of the eighth adder on the slope of a linear function is approximated from logarithmic functions, and a selector for selecting one of outputs of the second and third multipliers according to the reliability of the SNR of the code symbol. The weight ratio is preferably 1/2+1/4+1/16.
Reliability SNR is determined in accordance with what is possible or not accurate estimate of the SNR. The selector outputs the value adopted from the second multiplier, if possible accurate estimate of the SNR, and the value adopted from the third multiplier, if there is no accurate estimate of the SNR.
The above and other objectives, features and advantages of the present invention will become more apparent from the subsequent detailed description, taken together with the attached drawings, on which:
figure 1 presents a block diagram illustrating an example of turbodecoding, using the modified algorithm, Max-Log-MAV in accordance with the option run nastoyascheevremya;
figure 2 presents the precedence diagram illustrating example operations for identifying the best metrics Mn(0) and the second best metric Mn(1) at time k time decoding in accordance with a variant implementation of the present invention;
figure 3 presents the precedence diagram illustrating example operations for calculating the LOP and external information for iterative decoding according to the modified algorithm, Max-Log-MAV in accordance with a variant implementation of the present invention;
4 shows a block diagram of example functional blocks for simultaneous finding the best and the second best metric for LOP in any time of decoding in accordance with a variant implementation of the present invention;
figure 5 presents a block diagram of example functional block for obtaining external information for an information symbol at an arbitrary point in time decoding in accordance with a variant implementation of the present invention;
figure 6 presents a block diagram of example functional blocks for calculating adjustment values used for receiving external information, in accordance with a variant implementation of the present invention;
7 and 8 show graphs illustrating the reamers performance ratio of bit errors (UNCCD) and the factor of human error (CE) algorithms turbodecoding, when the packet size coding (PC) is 3864 and overall encoding speed is 1/2, in accordance with a variant implementation of the present invention;
figure 9 and 10 shows graphs illustrating examples of performance CCD and KCO for the log2 Maxlogname, mod. Maxlogname, Maxlogname with KOS and Maxlogname relatively iterations at Eb/N0equal to 1.3 dB, in accordance with a variant implementation of the present invention;
figure 11 and 12 shows graphs illustrating examples of performance CCD and KCO algorithms turbodecoding, when the size of the PC is 792 and efficient encoding speed is 1/5, in accordance with a variant implementation of the present invention;
on Fig and 14 shows graphs illustrating examples of performance CCD and KCO relatively iterations at Eb/N00.7 dB when the PC is 3864, in accordance with a variant implementation of the present invention; and
on Fig and 16 show graphs illustrating examples of performance CCD and KCO with respect to the mismatch SNR at Eb/N0equal to 1.2 dB, when the size of the PC is 3864 and efficient encoding speed is 1/2, in accordance with a variant implementation of the present invention.
The following describes embodiments of us who Otsego of the invention with reference to the accompanying drawings. In the following description, for brevity, omitted the well-known functions or constructions.
The present invention is intended to create an improved algorithm, Max-Log-mAh, which, by inoculation update LOP existing algorithm, Max-Log-mAh, works only about 0.1 dB or less worse than the algorithm the Log-mAh and offers the best performance turbodecoding than the algorithm, Max-Log-MAV and the algorithm, Max-Log-MAV with KOS. Improved algorithm, Max-Log-MAV mainly represents the algorithm turbodecoding-based algorithm, Max-Log-MAV, it advantageously provides a slight increase in the complexity of hardware without mismatch SNR.
The characteristics of the present invention are presented briefly as follows.
(1) To update the LOP at any time decoding are considered the second best metric for information symbols 0 and 1, as well as the best metric. It is noteworthy that the second best metric excluded from consideration for updates LOP in the existing algorithm, Max-Log-MAV. From these simulation results it is obvious that this update LOP of the present invention leads to the performance of Turbomachinery, which is also good, and the performance of the algorithm the Log-MA is.
(2) If the adjustment value of fc, which is calculated using the second best metric for information symbols 0 and 1 in any time of decoding, as defined, is a nonlinear function, the SNR mismatch leads to changes in performance. Therefore, fcapproximated by a linear function. The simulation results also make clear that the approximation of fca linear function leads to excellent performance of turbodecoding regardless of SNR mismatch.
Therefore, described by a linear approximation of fcin accordance with the present invention. In addition, the performance of turbodecoding evaluated in the case when fcis defined in terms of its initial logarithmic functions, and explores the applicability of this definition of fc.
Figure 1 presents a block diagram illustrating an example of turbodecoding, using the modified algorithm, Max-Log-MAV in accordance with a variant implementation of the present invention. As described above, the modified algorithm, Max-Log-mAh refers to the algorithm, Max-Log-MAV, which updates LOP, using the best and the second best metric for the information symbol at the time of decoding in accordance with variantvalue of the present invention.
Modified algorithm for Max-Log-MAV is applied to each constituent decoder (DEC1 and DEC2). The controller feedback factor (CCCS) for weighting the external information is also applied to each constituent decoder.
As shown in figure 1, the first and second constituent decoders DEC1 and DEC2) 101 and 104 respectively receive external information and LOP for an information symbol using a modified algorithm, Max-Log-MAV. I.e., each constituent decoder 101 and 104 corresponds to one of the constituent encoders of turbochager. Interleaver 102 punctuates the signal received from the first constituent decoder 101. By incorporating interleave data between the constituent codes of the turbo code interleaver 102 changes the order of the sequence data, so that the output of the first constituent decoder 101 is consistent with the input of the second constituent decoder 104. The first CCCS 103 multiplies perenesennyj signal by the weighting coefficient obtained from the external information computed in the first constituent decoder 101 in accordance with the modified algorithm, Max-Log-MAV. The weighting factor is an empirical value. It is more in the algorithm, Max-Log-MAV than in the algorithm, the Log-MAV. Given this, the external information to the information symbol is multiplied by a weighting factor less than 1, thereby achieving the best slave is Chiyo characteristics. The second constituent decoder 104 decodes the output of the first CCCS 103. Departmental 105 performs deteremine, so that the output of the second constituent decoder 104 is consistent with the input of the first constituent decoder 101. Second CCCS 106 multiplies departmeny signal by the weighting coefficient obtained from the external information computed in the second constituent decoder 101 in accordance with the modified algorithm, Max-Log-MAV. The output of the second CCCS 106 is input to the first constituent decoder 101.
The adders 107 and 108 summarize the reliability of the transmission and the a priori probability (AR) accepted code symbol to generate LOP for the information signal, using the external information acquired from the second constituent decoder 105. A priori information is LOP probability information symbol 0, the probability of an information symbol, equal to 1. In the General theory of encoding information symbols 0 and 1 are equiprobable. So the initial a priori information is always 0. As a continuation of the iterative decoding external information from each constituent decoder is used as a priori information of the character for the other constituent decoder. Therefore, a priori information is no longer equal to 0. Crucial unit 109 decides to sign LOP. If mn is to LOP positive, the decision making unit 109 generates information symbol 0, and, if the sign LOP negative, it generates information symbol 1. The magnitude of the decision is served as the output buffer 110, and the unit 111 checks cyclic redundancy code (CEC). In an embodiment of the present invention, the output buffer 110 may be a memory to store the value of decision-making 0 or 1. Unit 111 checks the CEC verifies priori entered the CEC to detect errors in the decoded frame of information symbols.
Now, the following describes the implementation of the modified algorithm, Max-Log-MAV in the composite decoders.
Modified algorithm for Max-Log-MAV evolved from the algorithm, Max-Log-MAV through modification of its update process LOP. Therefore, for ease of implementation and preservation of the insensitivity of turbodecoding to misalignment SNR is selected yet equation (3) to calculate the metrics of the state α and β for the second modified algorithm, Max-Log-MAV. Also equation (2) is used with a correction value of fcis approximated to determine the LOP for the modified algorithm, Max-Log-MAV.
Approximation of fcincludes the definition of fcusing the best metric Mn(0) and the second best metric Mn() for information symbols 0 and 1 of the total number of metric M n(i)the components of fcin equation (2). In other words, the algorithm turbodecoding of the present invention updates the LOP at any time decoding, given the second best metric for information symbols 0 and 1 are removed when you upgrade LOP in the algorithm, Max-Log-MAV, as well as their best metric.
To update the LOP in the modied algorithm, Max-Log-MAV fcapproximated as follows:
(5)
As can be seen from equation (5), fcis determined using the best metric Mn(0) and the second best metric Mn(1) for the information of the symbol n in the time of decoding. Metric Mn(i) (i>1) is smaller than the second best metric Mn(1)dropped approximation, since they have a negligible effect on the fc. While the algorithm, Max-Log-MAV searches over all sets of States (s', s) on the grid at any time decoding and calculates the best metric Mn(0) for an information symbol, updating the metrics for each set of conditions, the modified algorithm, Max-Log-MAV calculates the second best metric Mn(1) in addition to the best metric Mn(0) and at the same time without increasing the time of decoding. To this end p. is there a metric for state s is equal to m(s). Then Mn(0) and Mn(1) are computed simultaneously as depicted in Table 1.
| Table 1 |
| (1) initialization: s=0, Mn(0)=MIN, Mn(1)=MIN |
| (2) the presence of m(s) |
| (3) if Mn(0)<m(s), then Mn(1)=Mn(0) and Mn(0)=m(s) |
| otherwise, if Mn(1)<m(s), then Mn(1)=m(s) |
| (4) if s=S-1, then stop. Otherwise move on (5) |
| (5) increase s by 1. The transition to (2) |
In Table 1, MIN is a very small value, equivalent -∞ to initialize metrics state, and S is the total number of States in the lattice constituent convolutional codes.
Figure 2 presents the precedence diagram illustrating example operations for calculating the best metric Mn(0) and the second best metric Mn(1) at time k time decoding in accordance with a variant implementation of the present invention.
As shown in figure 2, the state of the lattice and the best and the second best metric for information symbols 0 and 1 are set to the initial value at the time k time decoding at step 200, as shown by (1) T is the blitz 1. In step 202 calculates the metric for the information of the symbol n (0 or 1), each time increasing the state 1. Therefore, the operation of figure 2 is the process of determining the current state s. The calculated metric is compared with the current best metric for the information of the symbol n in step 204. If the current metric is greater than the current best metric, then the process moves to step 206. Otherwise it jumps to step 208. In step 206, the current metric is set as the best metric and the current best metric is set as the second best metric.
On the other hand, in step 208, the current metric is compared with the existing second best metric. If the current metric is greater than the current second best metric, then the second best metric is updated to the current metrics in step 210. If the current metric is equal to or less than the current second best metric at step 206 or 210, or in step 208, the procedure goes to step 212.
In step 212 determines whether the current state is the last state. If it is, then the procedure ends. If it is not, the state is incremented by 1 at step 214.
Thus, the best and the second best metric Mn(0) and Mn(1) are obtained simultaneously at any point in time, de is tiravanija. Using these metrics, the adjustment value of fcapproximated in equation (5).
However, nonlinear approximation of fcin equation (5) has an impact on the performance of the decoding in accordance with the absolute value of the input symbol in turbodecoding. I.e., the receiver cannot estimate the exact SNR, leading to SNR mismatch. As a result, if you change the input symbol decoder, then also change the operating characteristics of turbodecoding. It is therefore necessary to approximate fcwith a logarithmic function to a linear function.
Below is a description of the approximation of the logarithmic function than a linear function.
When the representation of fcthrough logarithmic functions as in the algorithm the Log-MAV or through information table corresponding to the logarithmic function, the error term SNR changes Es/N0that is multiplied by the input symbol decoder, despite the constant SNR for the input symbol, which is fine modifies the operating characteristics of turbodecoding. In order to maintain the performance of the decoding regardless of the value of the input symbol, the logarithmic function must be modified. Equation (6) represents the approximation of the logarithmic function.
(6)
Function with the metric as the ratio must be linear with respect to the metric to obtain LOP so, which makes the performance of the decoding independent of changes of the input symbol. If fcmodified non-linear way depending on the metric, changing with the value of the input symbol, then fcalso changes in a linear fashion relative to LOP according to the changing input symbol, despite the same SNR. Therefore, not provided with regular performance.
When the approximation expressed as equation (6), the constant C is negligible. She shifted through the approximation l(x) as a function of the first order with constant C, since fcis determined by the difference between the functions l(x)having a metric for information symbols 0 and 1 as factors.
This rough approximation. Due to errors caused by rough approximation, the modified algorithm, Max-Log-MAV of the present invention works worse than the modified algorithm, Max-Log-MAV with l(x), defined as the logarithmic function. However, the definition of l(x) in the form of nonlinear functions can lead to good performance, if provided accurate estimate of the SNR, but the performance of the decoding changes to the GDS mismatch SNR changes the value of the input symbol.
In the approximation LOP is updated in the modied algorithm, Max-Log-MAV through
where
(7)
The second best metric Mn(1) is calculated simultaneously with the best metric of Mn(0) according to equation (7) in the approximation algorithm.
Now describe the weighting coefficients applied to the external information. External information about the information symbol can be obtained using the LOP of the upgrade process LOP modified algorithm, Max-Log-MAV. Since the algorithm Max-Log-MAV creates external information in repeated approximations, the external information has a relatively large value compared with external information in the algorithm, the Log-MAV. In order to reduce this influence, the external information to the information symbol is multiplied by a weighting factor. In the conventional algorithm, Max-Log-MAV with KOS predetermined weighting factor, for example 0,588235, is multiplied by the external information from the second constituent decoder for each iteration. However, in the modied algorithm, Max-Log-MAV in LOP is a corrective value of fcreflecting the second best metric, and thus, the weighting factor for external information should be closer to 1 than fc. Given in the owl coefficient W fexternal information is formed as
,
where
or(8)
In equation (8) K'=Kf. Lcyksignal reflecting the reliability of the channel at the entrance of turbodecoding, and La(Uk) - priori information of the current information symbol. The formula is obtained by subtracting the external information from the difference between the best metric and the second best metric, and then summing the new correction values of fc' with the resulting difference. Below fc' called the corrective value.
There is the following definition description LOP and external information for iterative decoding in the modied algorithm, Max-Log-MAV.
Figure 3 presents the precedence diagram illustrating example operations for calculating the LOP for the information symbol and the external information used for iterative decoding in the modied algorithm, Max-Log-MAV in accordance with a variant implementation of the present invention.
As shown in figure 3, is calculated metric γ branch to an arbitrary state transition in the lattice at step 400, and metrics α and β status updates for all sets of States (s, s') with respect to the state transition in step 402. Nusage 404 simultaneously discovered the best metric M n(0) and the second best metric Mn(1) to obtain the LOP in the procedure in figure 2, updating the metric condition. LOP is calculated using the difference between the Mn(0) and Mn(1), the input of the decoder to take into account in the SNR and the a priori information of the information symbol according to equation (8) in step 406. This step is performed in the functional blocks 601, 602 and 603 shown in figure 5. In step 408, the external information is multiplied by a weighting factor Wfthat is performed in block 604, shown in figure 5. Correcting the value of fc' is selected as one of the two values defined in equation (8), depending on, approximated if the logarithmic function is a linear function or not. If the receiver can accurately estimate the SNR, fc' is selected as the initial logarithmic functions. Otherwise, it is selected as approximated by a linear function. Thus, if at step 410 the possible accurate estimate of the SNR, then the procedure goes to step 412, and, if this is not possible, then the procedure goes to step 414. In step 414, the logarithmic function is used as the fc'and at step 412 the linear function is used as fc'. The logarithmic function is selected when function blocks 701, 702, 703, 705 and 707, depicted in Fig.6 output FLAG(flag) in the form 0, while a linear function is selected when function blocks 701, 702, 704, 706 and 708 depicted in Fig.6 output FLAG 1.
4 shows a block diagram of example functional blocks to find the best metric and the second best metric in respect of LOP in any time of decoding in accordance with a variant implementation of the present invention.
As shown in figure 4, the bold solid line denotes the section finding the second best metric, i.e., functional blocks 511-514. Therefore, other functional blocks 501, 502 and 503 are according to the algorithm, Max-Log-MAV. These functional blocks update metrics for all States of the lattice, each time increasing the index by 1. Here, the signal SELO equal to 0 for the first state and 1 for the subsequent States. The signal SEL1 is equal to 0 for the first and second States and 1 for subsequent States.
Functional blocks 502, 503, 511, 513 and 514 are selectors to output the input port 0, if the select signal is 0, and the input port 1, if the select signal is equal to 1. Functional blocks 501 and 512 are selectors to output 1 if the signal at the port and less than the signal on port b, and 0 if the signal on port a is equal to or greater than the signal on port b.
Figure 5 presents a block diagram of example functional blocks for genericoviagra information about the information symbol at an arbitrary point in time decoding in accordance with a variant implementation of the present invention.
As shown in figure 5, the first adder 601 outputs the difference between the best metric for 0 and 1 as the information LOP information about the symbol. The second adder 602 summarizes the information transfer and the a priori information of the received symbol. The third adder 603 subtracts the amount adopted from the second adder 602, information LOP adopted from the first adder 601. The output of the third adder 603 is the external information identified in the existing algorithm, Max-Log-MAV. The multiplier 604 multiplies the external information on the weighting factor, as is done in the existing algorithm, Max-Log-MAV with KOS. If the weighting factor is equal to 1, it executes the algorithm, Max-Log-MAV. The fourth adder 605 adds the correction value of fc'obtained by the functional blocks depicted in Fig.6, the output of the multiplier 604. Thus, it turns out final external information for the modified algorithm, Max-Log-MAV.
I.e., external information is obtained for the modified algorithm, Max-Log-MAV through additional use of the multiplier 604 associated with a weight Wfand adder 605 associated with the corrective value of fc'compared with the algorithm, Max-Log-MAV. Also, compared with the algorithm, Max-Log-MAV with KOS is also used su is motor 605.
Figure 6 presents a block diagram of example functional blocks for calculating adjustment values of fc' for use in the calculation of external information in accordance with a variant implementation of the present invention.
As shown in Fig.6, the first adder 701 calculates the difference between the best metric and the second best metric for the information of the symbol 0, and the second adder 702 calculates the difference between the best metric and the second best metric for the information symbol 1. The informational table 703 (ST) finds the correction value of the logarithmic function defined in equation (8)using the difference. The third adder 707 calculates the difference between the adjustment values. The first multiplier 707 multiplies the difference by a weighting factor, thereby making a decision about the final adjustment value.
The fourth adder 704 calculates the difference between the outputs of the first and second adders 701 and 702. The second multiplier 706 multiplies the difference by the amount of tilt, thus taking the decision on the correction value is approximated to a linear function.
One of the correcting values, defined in equation (8)is selected according to the FLAG signal. To select logarithmic function selector 708 selects the input port 0. In contrast, to select l the nonlinear function selector 708 selects the input port 1. For the first case you want ART, whereas for the second case, you just need the adders and a multiplier. It is noteworthy that when the FLAG is 0, it can be ensured accurate estimate of the SNR at the receiver. This structure is depicted in Fig.6, additionally hardware for the modified algorithm, Max-Log-MAV. If the weighting factor Wfand the value of K' can be expressed as exponents of 2, the multipliers on 5 and 6 may be implemented in the form of simple selectors bits or adders, including them.
In order to assess the performance of turbodecoding modified algorithm, Max-Log-MAV in accordance with the present invention, simulations were performed under the following conditions.
All simulations were floating point, and the performance of the decoding was assessed in terms of the ratio of bit errors (UNCCD) and the factor of human error (KCO). To study the effect of mismatch SNR also evaluated the performance of the decoding in moving Eb/N0. Used turbocodes speed 1/5 provided CDMA2000 1xEV-DV, and interacting with quasimomentum the turbo code (CDTC) to convert the total velocity encoding value other than 1/5. The frame size is one of the R is Smurov PC, as defined in the specification 1xEV-DV. Used by the modulation scheme was dip FMN, and assuming a channel with additive white Gaussian noise. For turbodecoding maximum number of decoding iterations was 8. The BWC and KCO were measured by running the simulation until then, until I got 50 human errors.
The weighting factor Wfand the value of K' was determined empirically. As the iteration decoding for turbodecoding are mainly operations suboptimal decoding but not decoding by the maximum likelihood method, there is a likelihood that performance deteriorates during the iterative decoding. Modeling mismatch SNR showed that the best performance is achieved for offset Eb/N0approximately -1 dB, than without the offset Eb/N0. This is due to the fact that the performance deterioration, possibly generated during the decoding iterations, is shifted by erroneous weighing at -1 dB. Thus, the weighting factor Wfempirically determined as follows:
(9)
In the representation of Wfas the sum of the exponents of 2, multiplication by a weighting factor easily ASU is coming hardware.
K' in equation (8) represents the product of the slope K in equation (4) and the weighting factor Wf. K in equation (6) is defined as the average slope of the tangent function l(x)=log(1+e-x). So
(10)
where a is set to the maximum meaningful value. If more than 9, then l(x) is less than 10-4. Thus, for equal 9, K is defined as follows:
(11)
Some modeling has shown that the definition of K given by equation (11) leads to high performance. K' in equation (11) can also be expressed as:
(12)
K' can be simply obtained bitwise choice, which is a simplified hardware implementation of multiplication.
The simulation results with the approximation and without approximation are compared with reference to Fig.7-16. 7 and 8 shows the performance of Turbomachinery in the sense of the BWC and KCO for size PC, equal 3864, and the overall coding rate (R)equal to 1/2. 7 and 8 Log mAh means algorithm the Log-MAV, the log2 Maxlogname means algorithm, Max-Log-MAV using fcdefined as a logarithmic function l(x), mod. Maxlogname means algorithm, Max-Log-MAV using fcdefined as approximarely the I function of the first order, Maxlogname with KOS denotes an existing algorithm, Max-Log-MAV with KOS and Maxlogname denotes an existing algorithm, Max-Log-MAV. As shown, the log2 Maxlogname is close to Logman on performance decoding, but these performance characteristics are not provided in the case of SNR mismatch. Mod. Maxlogname works simply by about 0.1 dB worse than Logmap at KCO equal to 10-2while he works about 0.5 dB better than Maxlogname with KOS. Mod. Maxlogname runs continuously regardless of SNR mismatch.
Figure 9 and 10 shows the operating characteristics of the BWC and KCO for the log2 Maxlogname, mod. Maxlogname, Maxlogname with KOS and Maxlogname relatively iterations at Eb/N0=1,3 dB. It is noted from figures 9 and 10 that the log2 Maxlogname has the best performance regarding the iterations. Mod. Maxlogname works no better than Maxlogname with KOS, but receives performance KCO for Maxlogin with KOS in 8 iterations 7 iterations.
Figure 11 and 12 shows the operating characteristics of the BWC and KCO for size PC, 792, and the effective coding rate equal to 1/5. Similarly for the case when the size of the PC equal 3864, no change in the ranking performance of the five algorithms. Still, compared with the case where the size of the PC equal 3864, mod. Maxlogname works when is Erno 0.1 dB better than Maxlogname with KOS.
On Fig and 14 shows the operating characteristic of the BWC and KCO five algorithms, when Eb/N0=0,7 dB and size PC=792, and Fig and 16 shows the operating characteristic of the BWC and KCO five algorithms for size PC=3864, effective coding rate=1/2 and the mismatch SNR at Eb/N0equal to 1.2 dB, i.e. when the error is equivalent to the offset of the Eb/N0generated when estimating the SNR of the input symbol decoder assuming that accurate estimation SNR is achieved when the offset Eb/N00. As shown, mod. Maxlogname operates independently of the mismatch SNR, since l(x) is approximated by the function of the first order. However, the log2 Maxlogname showing variable performance characteristics according to SNR mismatch, since l(x) is defined as a non-linear function log(), and fcmodified non-linear way depending on the change of the metric in the log()function. However, changes in fcare not large compared to Lugmav. Therefore, as guaranteed estimate the SNR in the range of about -6 dB to +6 dB, then the log2 Maxlogname can be used as algorithm turbodecoding.
From the simulation it is noted that the modified algorithm, Max-Log-MAV works only about 0.1 dB worse than the algorithm the Log-MAV regardless of the size of PC, oz is ACA, these performance characteristics better than the performance of the algorithm, Max-Log-MAV (with KOS or without it). Despite some errors when estimating the SNR of the input symbol, the modified algorithm, Max-Log-mAh is a high performance, regardless of the errors of estimation of the SNR, which is obvious from the simulation results.
As described above, the modified algorithm, Max-Log-MAV performs better than the algorithm, Max-Log-MAV with a small hardware addition in comparison with the algorithm, Max-Log-MAV and a simplified structure in comparison with the algorithm the Log-MAV. Therefore, the modified algorithm, Max-Log-MAV is applicable to a channel decoder in a mobile terminal for universal mobile communication system (USPS) and high speed packet access lines "down" (VPDL), and channel decoder for the system and terminal CDMA2000 1xEV-DV. It is works well with a simplified structure.
Although the invention has been shown and described with reference to some embodiments of the, for the specialist in this field of technology it is clear that it can be made various changes in form and detail within the essence and scope of the invention defined in the attached claims.
1. The method of the composite decoding for decoding the turbo code, with steps: (1) calculation of the best and second is alecxih metrics from the metrics, represents the sum of the metrics of the state and branch metrics for the received information symbol in the lattice of turbodecoding at any time during turbodecoding information symbol; (2) calculating the difference between the best metric for the information of the symbol 0, and the best metric for the information symbol, equal to 1; (3) calculating the difference between the second best metric for the information of the symbol 0, and the second best metric for the information symbol, equal to 1; (4) calculating the difference between the difference between the best metric and the difference between the second best metric, and multiplying the calculated difference by a weighting factor, so the metrics that represent the sum of the metrics of the state and branch metrics are linear; and (5) update the logarithmic likelihood ratio (LOP) information symbol, using the difference between the best metric, obtained in step (2), and the product obtained in step (4), and a decision on the value of the information symbol based on the updated LOP.
2. The method of the composite decoding according to claim 1, which also includes the step of calculating the external information using the updated LOP, input symbol, reflecting the SNR (signal-to-noise), and the a priori information input character after step (5).
3. Way sostav the th decoding according to claim 1, in which the weighting factor is determined as follows:
Weighting factor = K·Wf,
where Wfless than 1 and close to 1 and To - the average slope of the tangent logarithmic function l(x)=log(1+e-x).
4. The method of the composite decoding according to claim 3, in which Wfmore 0,588235.
5. The method of the composite decoding according to claim 1, in which the weighting factor is obtained from the function linearizovannogo of logarithmic functions, using the average slope of the tangent logarithmic functions, and logarithmic function is represented by the difference between the best metric and the second best metric.
6. The method of the composite decoding according to claim 3, in which the average slope of the tangent is an integer from 0 to 9.
7. Composite decoder for decoding turbo code containing a first adder for calculating the difference between the best metric for the received information symbol, equal to 1, and the best metric for the information of the symbol 0 in the lattice of turbodecoding at any time during turbodecoding information symbol; a second adder for summing the information transmission and a priori information of the information symbol; a third adder for calculating the difference between the outputs of the first and second adders and in the water of a difference in the quality of external information; a first multiplier for multiplying an output of the third adder by a predetermined weighting factor as a ratio of the feedback; the transmitter correction value to calculate a correction value using the difference between the best metric and the second best metric of the received information symbol; and a fourth adder for summing the correction value from the output of the first multiplier.
8. Composite decoder according to claim 7, in which the transmitter corrective value contains the fifth adder for calculating the difference between the best metric and the second best metric for data symbol is equal to 0; the sixth adder for calculating the difference between the best metric and the second best metric for the information symbol, equal to 1; the reference table to store based on a logarithmic function of the correction values for the outputs of the fifth and sixth adders and the output of the correction values for the outputs of the fifth and sixth adders; the seventh adder for calculating the difference between the correction value; a second multiplier for multiplying the output of the seventh adder for a predetermined weighting factor; eighth adder for calculating the difference between the outputs of the fifth and sixth adders; a third multiplier to multiply the I output of the eighth adder on the slope of a linear function, approximated from logarithmic functions; and a selector for selecting one of outputs of the second and third multipliers according to the reliability of the signal-to-noise ratio (SNR) of the received information symbol.
9. Composite decoder of claim 8, in which if the weighting factor and the slope of the linear function can be expressed as exponents of 2, each of the multipliers provided in the form of selector bits.