Communication Unit and Method for Soft Decision Demodulation

Abstract

A communication unit for performing soft-decision demodulation is described, comprising a receiver. The receiver is arranged to receive a transmitted signal having a first bit set comprising k bits, the transmitted signal being determined according to the values ​​of the k bits from 2... k The signal is selected from a set of possible signals, and the transmitted signal includes a second set of bits, which includes bits based on a set of 2... Qm The phase rotation of the transmitted signal selected from a set of possible rotations by Q m The receiver includes a demodulator (118), which comprises a set of 2 bits. k A correlator is configured to detect the transmission of each possible transmitted signal and output the 2... k The phases are used as the third input set. The demapper circuit (113) is connected to the demodulator (118) and configured to: receive the third input set; and determine statistics derived from the multiple aggregate correlator output phase distributions of the third input set, and calculate Q based on the statistics. m Output the second a posteriori soft bit set of each soft bit.

Description

Technical Field

[0001] This invention relates to a communication unit for soft-decision demodulation and decoding of communication data packets in a receiver chain. The scope of this invention is applicable to, but not limited to, demodulation for current and future generations of communication standards. Background Technology

[0002] Orthogonal signaling (OS) is a known communication technique. In a transmitter of a communication unit using OS, a binary-to-decimal conversion is used to convert a k-bit sequence (each of the k bits having a value of "0" or "1") into values ​​from "0" to "2". k Integers within the range of -1. This integer is then used as an index value to select 2. k A specific one from a communication sequence is sent. In the OS, 2 k The set of communication sequences was designed to be orthogonal to each other, making them easily distinguishable from one another. 2 k A selected communication sequence is transmitted over a communication channel, and this transmission may be distorted due to noise, interference, dispersion, fading, Doppler effects, and other factors. As a result, the received sequence will often differ from the transmitted sequence to some extent, and the orthogonality of the sequences may not be maintained.

[0003] In the corresponding OS receiver, a set of 2 k There are 2 correlators. Each correlator compares the received signal with 2 known signals from the receiver. k The corresponding sequence in each of the orthogonal communication sequences is compared. 2 k Each of the correlators generates a correlation metric, which takes the received signal and 2 k The magnitude is quantified as the strength of the correlation between corresponding sequences in a set of orthogonal communication sequences. More specifically, a high magnitude indicates a high correlation, therefore the 2 k In a set of orthogonal communication sequences, the corresponding one is the one whose signal is transmitted with high confidence. In contrast, a low amplitude indicates low correlation, therefore the 2 k In a set of orthogonal communication sequences, the corresponding one is the one with the lowest confidence level of the transmitted signal. The receiver compares the results of the set of 2... k The output of each correlator is 2 k Each amplitude is used to identify the one with the highest amplitude. Between 0 and 2... k The index value of the correlator that produces the highest amplitude is identified within the range of -1. Finally, a hard decision on the bit values ​​in the k-bit sequence is obtained by performing a decimal-to-binary conversion on the index value of the correlator that produces the highest amplitude.

[0004] In traditional OS communication schemes, when channel conditions are favorable, it can be expected that the received sequence will be consistent with 2. k One of the selected transmissions in a set of orthogonal communication sequences is similar. In this case, it can be expected that only this set of 2 from the receiver... k The correct correlator among the correlators will produce an output with a high amplitude, and the decimal-to-binary conversion will recognize the correct bit sequence without errors.

[0005] However, when channel conditions are unfavorable, the received sequence may not match 2. k Rather, they are not very similar to any of the orthogonal communication sequences, but may be somewhat similar to 2. k Two or more similar orthogonal communication sequences, which may or may not include 2 k The one selected to be transmitted from a set of orthogonal communication sequences. In this case, the receiver's set of 2... k Two or more correlators in a given correlator will produce outputs with relatively high amplitudes, and the correct one may not have the highest amplitude, especially if the channel does not maintain the orthogonality of the communication sequences. In this case, the decimal-to-binary conversion will identify incorrect bit sequences containing some bit errors. This can be mitigated by employing a channel encoder in the transmitter to encode the bit sequence before mapping it to the OS sequence. In the receiver, the recovered bit sequence can then be passed through the appropriate channel decoder to mitigate bit errors caused by unfavorable channel conditions.

[0006] However, it is well known that the channel decoder's ability to correct bit errors is limited when operating based on hard decision. In the OS scheme, it is assumed that the received sequence is related to 2 k Two of the orthogonal communication sequences are similar, that is, with 2 k The correct one of the orthogonal communication sequences is less similar to the other one, while the correct one is less similar to the other two. k The incorrect one of the orthogonal communication sequences is more similar. In this case, the amplitudes output by the two corresponding correlators will be similar, but the amplitude of the incorrect one will be slightly higher. In the process of generating a hard decision through decimal-to-binary conversion, the only knowledge retained is that a particular correlator among the correlators has the highest amplitude. The knowledge that the two correlators among the correlators have similar amplitudes will be lost, and the channel decoder will not know that another (less likely) k-bit sequence, which is also a candidate, has appeared. To solve this problem, it is known that if the receiver can provide soft decisions, the soft-decision channel decoder provides stronger error correction capabilities. These soft decisions not only express what the most likely value of each bit is (similar to a hard-decision receiver), but also how likely that value is (unlike a hard-decision receiver). More specifically, by the set of 2k The amplitude of the correlator output can be used to determine the corresponding 2 k Each of the orthogonal communication sequences is a correct likelihood, and these likelihoods can then be converted into a likelihood that each of the k bits has a value of 0 or 1.

[0007] However, the inventors of this invention have recognized that the document lacks a mathematically rigorous and robust method for converting the set of 2 k The amplitude of the correlator output is converted into the corresponding 2 k The likelihood of each of the orthogonal communication sequences is calculated, and then these likelihoods are transformed into the likelihood that each of the k bits has a value of 0 or 1. Summary of the Invention

[0008] In some implementations, the present invention provides soft-decision demodulators and methods for soft-decision demodulation using aggregate correlator output distributions, which, for example, have improved soft-decision quality. In particular, examples of the present invention describe the estimation of aggregate correlator output distributions and their use for soft-decision demodulation. These and other aspects of the invention will become apparent and elucidated with reference to the exemplary embodiments described below.

[0009] In a first aspect of the invention, a communication unit for performing soft-decision demodulation is described, comprising a receiver. The receiver is arranged to receive a transmitted signal having a first bit set comprising k bits, the transmitted signal being determined according to the values ​​of the k bits from 2... k The transmitted signal is selected from a set of possible signals, and the transmitted signal includes a second set of bits, the second set of bits including those based on 2... Qm The phase rotation Q of the transmitted signal is selected from a set of possible rotations. m The receiver includes: a demodulator, which comprises a set of 2 bits. k A correlator is configured to detect the transmission of each possible transmitted signal and output the 2-value of the correlator output. k The receiver also includes a demapper circuit connected to the demodulator and configured to: receive the third input set; determine statistics derived from the multiple aggregate correlator output phase distributions of the third input set; and calculate, based on the statistics, Q... m The second a posteriori soft bit set is generated and output. In this way, high-quality soft decisions can be obtained in a robust and practical manner, which does not depend on an excessive number of correlator output phase distributions and is insensitive to such an excessive number of correlator output phase distributions.

[0010] In an optional example, the number of phase distributions output by the aggregation correlator can be one, and where when 2 k When one of the possible signals is selected as the transmitted signal, the phase distribution of the aggregate correlator output is approximately 2. k The output phase of each correlator is 2 k Aggregation of distributions. In this way, the number of correlator output amplitude distributions is reduced, resulting in high practicality and robustness. The aggregated correlator output phase distribution is represented by a second set of distribution parameters.

[0011] In an optional example of the communication unit, the output phase distribution of the aggregate correlator can be represented by a second set of distribution parameters. This approach, instead of using histograms to represent the output phase distribution of each correlator, reduces the amount of information required to represent each correlator's output phase distribution, resulting in high practicality and robustness. Furthermore, the fitting function used to generate the parameters can fill the gaps between samples, thereby improving robustness.

[0012] In an optional example of the communication unit, the second set of distributed parameters may include extended parameters. 相位 In this way, the amount of information required to represent the phase distribution of each correlator output decreases only at one parameter, resulting in high practicality and robustness.

[0013] In an optional example of the communication unit, the demodulator is further configured to, based on the detected possible transmitted signals, be controlled by the set of 2 k The correlator outputs 2. k The magnitudes are used as the first set of inputs. In this way, the magnitudes of the correlator outputs become usable for classifying the correlator outputs in order to help estimate their aggregate distribution.

[0014] In an optional example, the demapping circuit can be configured to perform at least one of the following: estimating a second set of distribution parameters for the aggregated correlator output phase distribution by fitting a fifth probability distribution to the phase error of the correlator output obtained by correlating the received synchronization signal with a known synchronization signal; and fitting a sixth probability distribution to the second set of distribution parameters for the aggregated correlator output phase distribution by fitting a fifth probability distribution to the phase error of the correlator output obtained by correlating the received synchronization signal with a known synchronization signal. k The two correlators obtained 2 k The phase error of the correlator output with the largest amplitude in the set of correlator outputs is used to estimate a second set of distribution parameters for the aggregated correlator output phase distribution; this second set of distribution parameters is then selected from a lookup table connected to the demapper circuit. In this way, the receiver does not need to know the distribution parameters in advance, but can estimate them based on the received signal, resulting in high practicality.

[0015] In an alternative example of the communication unit, at least one of the fifth or sixth probability distributions can be a Gaussian distribution. In this way, a distribution very similar to that of the real signal can be expected, resulting in high-quality soft decision-making and increased robustness.

[0016] In the optional example, 2 k The output phase of each correlator can be combined with a second set of distribution parameters of the aggregate correlator output phase distribution to obtain a set including 2 k+Qm A third set of soft-prior soft signals for each soft phase. In this way, the contribution of each correlator output to soft-decision demodulation can be expected to be commensurate with the appropriate confidence level they provide, resulting in high-quality soft decisions.

[0017] In an optional example, the demapping circuit can be configured to accept Q. m The fourth set of soft-prior soft signals, consisting of 10 soft bits, is used as the fourth input set. In this way, the feedback information provided by the channel decoder can iteratively enhance the soft-decision demodulation, resulting in high-quality soft decisions.

[0018] In an optional example, the second set of prior soft bits can be combined with the soft bits of the fourth set of prior soft signals to obtain a set including Q. m The second set of soft bits is a soft-bit outer set. In this way, a positive feedback loop between the channel decoder and the soft-decision demodulator is avoided, resulting in high-quality soft decisions.

[0019] In an optional example, the communication unit can combine all prior soft signals to generate 2. k+Qm A set of posterior soft signals, and where 2 k+Qm The sets of posterior soft signals can be combined to obtain a first set of posterior soft bits and a second set of posterior soft bits. In this way, all available information contributes appropriately to the soft decision, resulting in high-quality soft decisions.

[0020] In a second aspect of the invention, a method for performing soft-decision demodulation is described, comprising a communication unit having a receiver having a demapper circuit coupled to a demodulator, the demodulator comprising a set of 2 k A correlator. The method performed at the receiver includes: receiving a transmitted signal having a first bit set and a second bit set, the first bit set comprising k bits, the transmitted signal being derived from 2^k bits based on the values ​​of the k bits. k The second bit set is selected from a set of possible signals, including those based on 2... Qm The phase rotation Q of the transmitted signal is selected from a set of possible rotations. m1 bit; detect the transmission of each possible transmitted signal; based on the detected possible transmitted signals, by the set of 2 k The correlator outputs 2. k A set of phases is used as a third input set; a statistic is determined from the distribution of multiple aggregate correlator output phases of the third input set; and a Q-value is calculated based on the statistic. m The second a posteriori soft bit set of 1 soft bit is output. Attached Figure Description

[0021] Further details, aspects, and embodiments of the invention will be described by way of example only with reference to the accompanying drawings. In the drawings, similar reference numerals are used to identify the same or functionally similar elements. Elements in the drawings are shown for simplicity and clarity only and are not necessarily drawn to scale.

[0022] Figure 1 A top-level block diagram of an OS-based transmission system adapted according to an exemplary embodiment of the present invention is shown.

[0023] Figure 2 It shows how to obtain Figure 1 Examples of two aggregation methods, a) the maximum method and b) the two aggregation methods, are given for the correlator output distribution of the OS scheme transmission system.

[0024] Figure 3 The dataset of multipath (MP) channels with an MP-7.5 dB impedance is shown, which has been fitted using histograms and Rice distributions. Figure 1 The correct relevant data for the OS solution.

[0025] Figure 4 This shows the fitting using histograms and Rice distributions. Figure 1 Incorrect related data in the MP -7.5dB dataset of the OS scheme.

[0026] Figures 5(A) to 5(C) show the use of Figure 3 and Figure 4 The dataset of additive white Gaussian noise (AWGN) channels Figure 1 The distribution of correlator outputs in the OS scheme.

[0027] Figures 6(A) to 6(C) show the use of Figure 3 and Figure 4 The multipath communication channel of the dataset Figure 1 The distribution of correlator outputs in the OS scheme.

[0028] Figures 7(A) to 7(B) show the conversion of AWGN channels to OS symbols LLR for the OS scheme datasets in Figures 5(A) to 5(C).

[0029] Figures 8(A) to 8(B) illustrate the conversion of multipath communication channels to OS symbols LLR for the OS scheme datasets in Figures 6(A) to 6(C).

[0030] Figure 9 A top-level block diagram of an alternative OS transmission system adapted according to an exemplary embodiment of the present invention is shown.

[0031] Figure 10 A block diagram of the test platform for the example 16OS scheme (where M=16 orthogonal signaling) is shown.

[0032] Figure 11 The sample external information transfer (EXIT) graphs for an example 16OS scheme dataset with real and estimated channels (both AWGN -7.5dB) are shown.

[0033] Figure 12 EXIT charts are shown for 16OS scheme datasets[5] with different estimated and true signal-to-noise ratio (SNR) values ​​on AWGN channels.

[0034] Figures 13(A) to 13(B) show EXIT charts of 16OS scheme datasets[5] with the same estimated and true SNR values ​​on AWGN channels.

[0035] Figures 14(A) to 14(B) show EXIT charts of a 16OS scheme dataset[5] with the same estimated and true SNR values ​​on a multipath channel.

[0036] Figures 15(A) to 15(B) show top-level block diagrams of an OS-PSK transmission system according to another exemplary embodiment of the present invention.

[0037] Figure 16 The distribution of the amplitudes of the correct and incorrect 16OS-QPSK correlators for the MP_-7.5dB dataset is shown.

[0038] Figure 17 The distribution of phase errors of the correct and incorrect 16OS-QPSK correlators is shown for the MP_-7.5dB dataset.

[0039] Figures 18(A) to 18(B) show the Example 16 OS-PSK test platform.

[0040] Figure 19 Example 16 OS-nPSK channel simulation is shown.

[0041] Figures 20(A) to 20(B) show the EXIT charts for each scheme of the 16OS-QPSK scheme.

[0042] Figures 21(A) to 21(C) show the amplitude distribution of the correlator output for the 16OS-PSK scheme used in AWGN channels.

[0043] Figures 22(A) to 22(C) show the phase error distribution of the correlator output for the 16OS-PSK scheme used in AWGN channels.

[0044] Figures 23(A) to 23(C) show the amplitude distribution of the correlator output for the 16OS-PSK scheme used in multipath channels.

[0045] Figures 24(A) to 24(C) show the phase error distribution of the correlator output for the 16OS-PSK scheme used in multipath channels.

[0046] Figures 25(A) through 25(F) show the EXIT diagrams for the OS-PSK scheme used in multipath channels.

[0047] Figure 26 The AWGN_0db dataset of an example OS scheme according to some example embodiments of the present invention is shown, with the first 20 rows out of 8100 rows shown here.

[0048] Figure 27 The correlator output distribution of the maximum amplitude method according to some exemplary embodiments of the present invention is shown.

[0049] Figure 28 The diagram illustrates 4-bit message combinations and corresponding symbolic LLR parameters according to some example embodiments of the present invention.

[0050] Figure 29 The following are some example embodiments of the invention, showing items in the bit LLR calculation when there is no feedback from the decoder.

[0051] Figure 30 A 4-bit message combination according to some example embodiments of the present invention is shown, having a corresponding symbol LLR parameter and a priori bit LLR combination.

[0052] Figures 31(A) and 31(B) show the AWGN_0db dataset of a 16OS-QPSK scheme according to some example embodiments of the present invention, showing the first 20 rows out of 8100 rows.

[0053] Figures 32(A) and 32(B) illustrate items in the symbol bit LLR conversion for the 6-bit message of Figures 31(A) to 31(B) according to some example embodiments of the present invention.

[0054] Figure 33The diagram illustrates 6-bit message combinations and corresponding LLR parameters according to some example embodiments of the present invention.

[0055] Figure 34 The conversion of the prior bit LLR to the symbol LLR field is illustrated according to some example embodiments of the present invention.

[0056] Figures 35(A) to 35(B) show flowcharts of software demapping methods for OS schemes according to some example embodiments of the present invention.

[0057] Figures 36(A) to 36(B) show flowcharts of soft demapping algorithms for OS-PSK schemes according to some example embodiments of the present invention.

[0058] Figure 37 Typical computing systems that can be employed in electronic devices or wireless communication units to perform software demapping and channel decoding according to some example embodiments of the present invention are shown. Detailed Implementation

[0059] To alleviate the lack of a mathematically rigorous and robust method in the literature, the set of 2 k The amplitude of the correlator output is converted into the corresponding 2 k The inventors of this invention describe the following in the first part of the specification: A communication unit and method for performing soft-decision demodulation are described, comprising a receiver for communicating with another communication unit on a communication channel. The receiver is arranged to receive a transmitted signal having a first set of k bits, which is from 2... k The receiver is selected from a set of possible signals. It includes a demodulator comprising a set of 2... k A set of 2 correlators, configured to: detect the transmission of each possible transmitted signal, and based on the detected possible transmitted signals, by the set of 2 k Each correlator outputs 2 k The amplitudes of the outputs of the multiple correlators are used as the first input set. The receiver also includes a demapper circuit connected to the demodulator and configured to receive the first input set; determine a statistic derived from the amplitude distributions of the multiple aggregate correlator outputs of the first input set, wherein the amplitude distributions of the multiple aggregate correlator outputs are less than 2. 2kThe first solution calculates and outputs a first posterior soft bit set comprising k soft bits based on the statistics. In this way, the first solution provides a soft demapping method for OS correlator outputs only, enabling high-quality soft decisions in a robust and practical manner, independent of and insensitive to an excessive number of correlator output amplitude distributions.

[0060] Furthermore, in the second part of the description, according to some examples of the invention, a soft-decision method is described for demapping the correlator output of the demodulator from an orthogonal signaling-phase-shift keying (OS-PSK) modulation scheme to extrinsic soft bits (also called bit LLRs or soft bits) in the form of log-likelihood ratios (LLRs). Here, a PSK symbol-to-symbol LLR circuit and a symbol-to-symbol LLR circuit can be additionally used, whereby the transmitter chain applies direct sequence spread spectrum (DSSS) technology, and the modulator circuit converts the information bits into OS-PSK modulated data. In this way, the extrinsic soft bits generated for PSK-modulated bits are consistent with the extrinsic soft bits generated for OS-modulated bits, thereby expressing a relatively appropriate confidence level for the bit values.

[0061] Correlator distribution study and symbolic LLR conversion: OS scheme

[0062] Now for reference Figure 1 This diagram illustrates a top-level block diagram of an OS scheme transmission system adapted according to an exemplary embodiment of the present invention. An exemplary embodiment of the invention provides receiver circuitry in receiver 102 of an orthogonal signaling (OS) transmission system 100, which uses a set of correlators to detect each received OS signal and demaps the correlator output values ​​to a soft bit set 103, wherein the soft bits 103 are fed into a soft-decision channel decoder 104. In some examples, the demapper operates in the receiver circuitry to perform soft-decision demodulation in the transmission system 100, which includes a transmitter 105, a channel 106, and a receiver 102, wherein the transmitter signals a bit set 107 comprising k bits by transmitting an OS transmit signal 108, the OS transmit signal 108 being derived from the values ​​of the k bits from 2... k The receiver 102 selects from a set of 2 possible signals. k A correlator is used to detect the transmission of each possible signal, and the correlator outputs 109 2 k A set of amplitudes is provided as input to the demapping circuit 113. The demapping circuit 113 calculates the posterior soft signal set 110 based on statistics and parameters 111 derived from the amplitude distributions of multiple aggregated correlators.

[0063] In some examples, the aggregation operation is not directly within the flow from the received correlator output to the soft demodulation LLR. Instead, in some examples, including the one shown, aggregation can be performed based on channel statistics, for example, beforehand during an offline lookup table process, or, for example, beforehand using an online channel estimation process. For completeness, it is confirmed that only the distribution parameters (e.g., s and sigma) extracted from the aggregation are considered. This is used in a direct flow from input to output. In some examples, the demapper circuit 113 is called a soft demapper because it generates soft signals. These soft signals can then be converted into soft bits 103, according to the example process shown in the flowcharts of Figures 35(A) to 35(B) and described later.

[0064] In this way, examples of the present invention provide a mathematically rigorous and robust circuit structure and method, which will consist of a set of 2 k The amplitude of the correlator output is converted into the corresponding 2 k The problem described in the background section is solved by taking the likelihood of each of the orthogonal communication sequences and then converting these likelihoods into the likelihood of each of the k bits with values ​​of "0" or "1".

[0065] In examples of this invention, the distribution of the correlator outputs (also referred to as "cross-correlation values") of the demodulator was studied to identify the possibility of using the distribution characteristics to softly demap the correlator outputs into a soft signal that can be used to generate soft bits. Furthermore, in one example embodiment, as part of the soft demapping, a circuit (in Matlab) was developed that converts the amplitudes of the correlator outputs into a set of soft amplitudes in the form of log-likelihood ratios (LLRs) by using multiple aggregated correlator output amplitude distributions.

[0066] Examples of the present invention focus on Figure 1 The two parts are: a lookup table 112 that characterizes the “modulation channel-specific library of correlator distribution” (also identified in step 4406 of Figures 35(A) to 35(B)), and a development demapper circuit 113 that is arranged to convert the correlator output into OS symbols LLR (also identified in step 4411).

[0067] Some of the circuits described below, such as CRC, encoders, bit-to-symbol circuits, modulators, demodulators, correlators, and decoders, can be implemented by those skilled in the art using hardware, firmware, or software. Therefore, the various ways in which these circuits can be implemented will not be described in more detail except as necessary for those skilled in the art to reproduce the concepts described herein.

[0068] data

[0069] To model the transmitter and channel, six datasets of correlator output values ​​of the demodulator obtained under different channel conditions were used, provided by [5]. These datasets were based on three signal-to-noise ratio (SNR) values ​​of 0 dB, -3 dB, and -7.5 dB, and on two channel models: additive white Gaussian noise (AWGN) channel and multipath (MP) channel. Data (e.g., M=16 M-ary orthogonal signaling (OS) transmission scheme) was used to generate data. Figure 1 (As shown), where M is a power of 2. This means that every k = log2(M) = 4 bits is converted from bits to symbol circuit 114 in transmitter 105 to become OS symbols. The datasets in this document are represented by the following symbols: AWGN_0dB, AWGN_-3dB, AWGN_-7.5dB, MP_0dB, MP_-3dB, MP_-7.5dB.

[0070] For example, refer to Figure 26 According to some exemplary embodiments of the present invention, the AWGN_0db dataset of the example OS scheme shows the first 20 rows out of 8100 rows. For example, Figure 26 The first 20 rows 3501 of the dataset AWGN_0db are shown, where each dataset has 8100 rows of data. Each row in the dataset has a symbol (with a value in the range [0, M-1] or [0, 15] 3502) and a corresponding one of M=16 transmission codes (with an index in the range [0, M-1] or [0, 15]), where each of the 16 codes has a cross-correlation magnitude 3503. Although this example embodiment uses M=16, it is conceivable that the concepts and definitions described herein can be applied to M=2. k Any value of .

[0071] In the M=16 correlator outputs for each symbol, such as Figure 26 The bold numbers in the diagram indicate that the correlator output with the transmit code index is assumed to be equal to the symbol value of the "correct" correlator output. The remaining M-1=15 correlator outputs in this symbol row are assumed to be "incorrect" correlator outputs, such as... Figure 26 The correlation values ​​are shown in bold. For example, Figure 26 Line 16, 3504, has a sign value of '5' (possibly for the 4-bit message '0101'). For this sign value, the transmit code with index 5 is the correct correlator, with an amplitude of 49.9. For the same sign, codes with indices in the ranges [0, 4] and [6, 15] are incorrect correlator outputs, with amplitudes in the range [6.0, 13.2]. This high SNR example illustrates the analogy between correct and incorrect, where the code with the highest cross-correlation amplitude almost always has the "correct" index.

[0072] Correlator output distribution

[0073] like Figure 1 As shown, the demapping circuit 113 is arranged to receive the amplitude of the correlator output 109 as input, and knows the parameters 111 of the amplitude distribution of the aggregate correlator output (these parameters are in...). Figure 1 (As shown in the diagram as an additional input), and converts the correlator output to OS notation LLR. Hereafter, the term "circuit" is intended to encompass any arrangement, module, hardware, or firmware of electronic circuits or components or logic gates, and any functionality implemented as software operation. It should be noted that various implementations of the concepts described herein can be implemented using hardware, firmware, software, or any combination thereof, depending, for example, on the primary application, purpose, etc., as understood by those skilled in the art. To support the subsequent discussion of the "conversion to OS LLR" circuit, this section details the correlator output distribution. More specifically, according to some exemplary embodiments of the invention, the aggregation of the correlator output distribution is described in detail in the following sections. Following this, according to some exemplary embodiments of the invention, the parameterization of the aggregated correlator output distribution is described in detail, which allows the aggregated distribution to be described using a small number of parameter values. Finally, according to some exemplary embodiments of the invention, the following sections discuss how to estimate the parameter values ​​in practical applications.

[0074] Distribution quantity

[0075] The maximum number of OS LLR distributions that can be used to calculate k-bit transmissions is 2. 2k As described below. Assume that each of the six datasets (each with 8100 rows) contains a relatively large number of symbols (similar to a transmission), with several rows for each symbol value, thus each symbol value has an array of M=16 correlator outputs. For example, in Figure 26 There are 4 rows with the value 3505, whose sign value is '7'. The method to obtain the maximum number of distributions is to, for each sign value, take the distributions of all correlator values ​​with code index '0' from all dataset rows containing that sign value, the distributions of all correlator values ​​with code index '1', and so on, until code index M-1=15. In this way, each sign value will have M=16 correlator output distributions, one corresponding to the correct correlator output, and the other M-1 (or 15) corresponding to the incorrect correlator outputs. This is in... Figure 2 The maximum distribution method (201) is shown in the figure, where Figure 2 The diagram illustrates a) a maximum distribution method 201 and b) two aggregated distribution methods 202 for obtaining the distribution of the correlator output. Figure 2 As shown, since there are M=16 symbolic values, using this method for a dataset will result in a total of M*M or 2... 2k=256 distributions. In this figure, the rows in the top (maximum distribution) method 201 correspond to different sign values, and the columns show different code indices. It is worth noting that the distribution on the diagonal 203 belongs to the correct correlator output, and is determined by... Figure 2 The remaining 2 2k -2 k A distribution of 204 is used to characterize incorrect correlator outputs.

[0076] You can also put 2 2k The distributions are aggregated into only two distributions to compute the symbol LLR: as shown in the figure, (i) the aggregation of the distributions of all correct correlator output values ​​205, and (ii) the aggregation of the distributions of all incorrect correlator output values ​​206, which is referred to as the two-aggregate distribution method 202. Here, in this two-aggregate distribution method 202, the distributions are not specific to a single symbol, but rather show the characteristics of the data corresponding to all symbols. Furthermore, the incorrect distributions are not specific to a single code index, but represent the incorrect correlator outputs for all code indices. This contrasts with the maximum distribution method 201, which has a separate distribution for each pair of “sent symbol-correlator index”. In the two-aggregate distribution method 202, the incorrect distributions 206 and the correct distributions 205 provide a set representation of their corresponding distributions in the maximum distribution method 201, as shown below. Figure 2 As shown. Therefore, compared to the two aggregation distribution methods 202, the maximum distribution method 201 provides a more specific representation of the correlator output value. However, the maximum distribution method 201 includes M 2 =256 distributions, each parameterized by several parameters, which can be considered to introduce excessive complexity in some examples of the invention. Another concern in some examples may be that the maximal distribution method may be overtuned and unrobustible for varying channel conditions. It should be noted that in the two aggregation distribution methods 202, the correct distributions of M=16 symbols are similar to each other, and the remaining incorrect distributions are similar to each other. In experiments, the inventors of the invention found this assumption to be sufficiently valid to achieve good operation.

[0077] In summary, in some examples, two aggregation distribution methods 202 are proposed, where the output amplitude distributions of multiple aggregation correlators are two, and where when 2 k When one of the possible signals is selected as the transmitted signal, the amplitude distribution of the first aggregate correlator output is approximately 2. k The 2 on the diagonal of the output amplitude of each correlator 203 k Aggregates 205 distributions, where when 2 k When one of the possible signals is not selected as the transmitted signal, the amplitude distribution of the second aggregate correlator output is approximately 2.k The amplitude output amplitude of each correlator is 2 2k -2 k Aggregates 206 of distribution 204.

[0078] Fit the distribution to the data

[0079] In some examples of this invention, to fit the distributions of six datasets of correlator output values ​​from a demodulator modeling the transmitter and channel to the correlator outputs, each dataset is divided into two sets of values ​​according to the two aggregation distribution methods described above: the set of all correct correlator outputs and the set of all incorrect correlator outputs. Through investigation, for each of the two sets of data correlators, for all datasets, several distributions (such as the normal distribution, Rayleigh distribution, Rice distribution, etc.) were fitted using Matlab's `fitdist` command. A complete list of the distributions considered can be found at: https: / / www.mathworks.com / help / stats / fitdist.html#btu538h-distname. Based on visual observation, it was concluded that the normal and Rayleigh distributions are good fits to the sets of correct and incorrect correlator data, respectively. In the example investigation, since both the normal and Rayleigh distributions are special cases of the Rice distribution, it was also concluded that the Rice distribution fits the correct and incorrect data. All other distributions accepted by the `fitdist` command (a total of 24 accepted distributions) fit the data poorly unless they are also generalizations of the Rice distribution. More specifically, it was found that the Nakagami and Weibull distributions also fit the data incorrectly, but they also fail to generalize the normal distribution, and therefore they do not fit the correct data well.

[0080] Assuming the Rice distribution is a general case of the normal and Rayleigh distributions, the Rice distribution is chosen in some examples of this invention so that comparisons can be made between the correct and incorrect correlator output sets. The Rice distribution has two parameters: a non-centrality parameter (s) and a scaling parameter (s). The mean and variance parameters of the correlator are similar to those of the normal distribution. Therefore, for each dataset, the correlator output will have four distribution parameters; the correct correlator output distribution has four s... 正确 and 正确 and the incorrect correlator output distribution of s 不正确 and 不正确 It should be noted that the Rice distribution with certain conditions is a special case of certain distributions in the Chi family of distributions. For example, suppose the random variable R is a distribution with a noncentrality parameter s and a scaling parameter s. If R follows a Rice distribution with a value of 1, then R also follows a non-central Ka distribution with two degrees of freedom, and R 2 The chi-square distribution is noncentral, with two degrees of freedom and a noncentral parameter s. 2 Therefore, having The set of correlator output values ​​for a Rice distribution with a value of 1 can be modeled using the chi-square distribution and chi-square distribution described above.

[0081] Now for reference Figure 3 and Figure 4 , Figure 3 The fitting using histograms and Rice distributions illustrates... Figure 1 The OS scheme has an MP -7.5dB multipath (MP) channel dataset with correct correlation data, and Figure 4 The fitting using histograms and Rice distributions illustrates... Figure 1 The MP -7.5dB dataset of the OS scheme has incorrect correlation data. To graphically illustrate the distribution of the correlator output data, two plots were drawn for each dataset. Figure 1 The OS solution transmits MP-7.5dB data sets in the system. Figure 3 The correct data 300 is shown in the image, and in... Figure 4 The text shows incorrect data (400). Each plot has a histogram of the data (301) and a Rice distribution fitted to the histogram (302). The number of histogram bins is automatically selected by the Matlab function `histogram`, and the function `histfit` is used to plot the histogram and distribution.

[0082] Referring now to Figures 5(A) through 5(C) and Figures 6(A) through 6(C), Figures 5(A) through 5(C) show the use of Figure 3 and Figure 4 The dataset of additive white Gaussian noise (AWGN) channels Figure 1 The distribution of the correlator output of the OS scheme, and Figures 6(A) to 6(C) show the distribution of the correlator output for the OS scheme. Figure 3 and Figure 4 The multipath communication channel of the dataset Figure 1 The distribution of correlator outputs for the OS scheme is shown. The distribution of all datasets is included in Figures 5(A) to 5(C) and 6(A) to 6(C), which illustrate the amplitude and phase distribution of correlator outputs for AWGN and multipath channels using the 16OS scheme. It should be noted that the bins are narrower for incorrect data because the incorrect subset contains M-1 = 15 times more samples than the correct subset.

[0083] like Figure 3 and Figure 4As shown, there is an overlap in correlator values ​​between the correct and incorrect distributions in the MP_-7.5dB dataset. The correct distribution values ​​start at approximately "15," which is exactly in the middle of the incorrect distribution—close to its mode. Furthermore, the incorrect distribution ends at approximately "55," which is the middle value of the correct distribution and close to its mode. This overlap between the correct and incorrect distributions means that the correct correlator output value is likely to be close to, and possibly smaller than, the incorrect correlator output value in the same M=16 code set.

[0084] By observing other SNR distributions (as shown in Figures 5(A) to 5(C) and 6(A) to 6(C)), it can be seen that not all SNRs overlap between the correct and incorrect distributions. Because the incorrect correlator always has a minimum value close to zero, the incorrect distribution has a range starting from values ​​close to zero, regardless of the SNR. However, as the SNR decreases from 0 dB, the maximum output value of the incorrect correlator in the distribution increases. Therefore, a smaller SNR will have a wider range of incorrect correlator output values. Furthermore, in the correct correlator distribution, as the SNR decreases, the minimum value decreases and the maximum value increases. In other words, a smaller SNR will also have a larger range of correct correlator values, but all from the small and large sides of the distribution. Therefore, the smaller the SNR, the greater the chance of overlap between the correct and incorrect correlator output values. This confirms the concept that the smaller the SNR, the more difficult it is to select the correct code index.

[0085] Therefore, in some examples of the present invention, representations of two aggregation correlator distribution methods 202 are proposed, wherein the first aggregation correlator output amplitude distribution and the second aggregation correlator output amplitude distribution are each represented by a set of distribution parameters. For example, in some cases, the set of distribution parameters may include Figure 3 Non-centrality parameter s and scale parameter 303.

[0086] Estimation methods

[0087] The analysis of the OS-PSK soft demapping and EXIT graph implementation described later assumes a test platform environment with information about Figure 1 Prior knowledge of the truth values ​​of transmitted bits 107 and OS symbol 115. In practical, known receivers, this knowledge is unavailable, and the distribution parameters s and must be estimated in the absence of this knowledge. 111. To obtain the distribution parameters 111 of the correlator output values, online or offline methods can be used (as shown in 4403 in Figures 35(A) to 35(B)). In the online method, the receiver data analyzer circuit 116 is responsible for analyzing the correlator output received from the demodulator 118 during operation, finding the optimal distribution to fit the data, and calculating the parameters of the distribution, which benefit from real-time tuning of the distribution parameters.

[0088] Offline method

[0089] In offline methods, such as Figure 1 As shown in Figures 35(A) to 35(B), correlator distributions can be computed offline and used to build a library of correlator distributions across all applicable modulation channel combinations. For example, the six datasets mentioned earlier in this document can be used as models for different channel conditions and modulation schemes. The configuration time switch 117 can be configured to select the correct dataset based on the system's modulation scheme and channel type. In some examples, this can be performed as an extension of any other channel estimation task performed by the demodulator 118. For example, the outputs of these channel estimation tasks can be used to index a lookup table of pre-computed correlator output distributions. More specifically, offline estimation of correlator distributions can be used to record correlator outputs for several channel conditions and to build a library of these distributions, for example, in the form of a lookup table (LUT) 112. Then, during actual data transmission, the offline channel estimation method 4405 identifies the current state of the channel and selects the appropriate distribution from the lookup table 112.

[0090] Another example of an offline approach is to use a single set of correlator distributions in all cases, regardless of varying channel conditions. This single set of correlator distributions can be recorded under worst-case conditions, perhaps at the lowest SNR where reliable synchronization is achievable. When channel conditions match this worst-case scenario, using the appropriate correlator distribution will ensure the best possible performance. When channel conditions are better than this worst-case scenario, even if the assumed correlator distribution is pessimistic compared to the true distribution, the chance of successful decoding can be expected to increase.

[0091] Online methods

[0092] Referring now to Figures 35(A) to 35(B), online example methods 4407 and 4408 are described as an alternative to offline distribution estimation using a dataset. These methods compute distribution parameters in the early stages of the real data transmission stream 4400, such as when a channel estimation task is underway. The first online channel estimation method 4407 correlates the received synchronization signal with known synchronization signals. Let's take an example: Suppose that for the transmission of a 4-bit message (k=4), demodulator 118 knows that before transmitting data, the transmitter sends a synchronization sequence '0000', '0001', ..., '1111', which has M=16 messages and corresponds to symbol values ​​from 0 to M-1=15 respectively. Therefore, demodulator 118 expects that among the M=16 correlators, the correlator with index '0' should output the highest cross-correlation amplitude in the first received transmission. Similarly, the correlator with index '1' is expected to output the highest amplitude in the second transmission, and so on, until the last symbol is transmitted.

[0093] Using the first online channel estimation method 4407, since the receiver 102 knows the true symbol values, it can associate each set of M=16 correlator outputs transmitted with its symbol value. Synchronization sequences can be transmitted several times, such that for each symbol value, there will be an M=16 set of correlator outputs (to correspond to...). Figure 26 (Simultaneously shown in the same manner, but with a different time order), thus the distribution can be estimated based on the required sample size. This online technique is sometimes referred to in this paper as correlator distribution estimation of a synchronization sequence using the first online channel estimation method 4407. Using this first online channel estimation method 4407, the following can be used: Figure 2 Maximum distribution method 201 to obtain 2 2k One distribution, or two aggregation distribution methods 202 can be used to obtain two correct aggregation correlator distributions 205 and incorrect aggregation correlator distributions 206, such as... Figure 2 As shown. It is also common to use between two and a maximum of two. 2k Multiple distributions among the values ​​will be explained in later examples. After distribution aggregation, distribution fitting methods (such as Matlab's histfit function) can be used to estimate the distribution for each aggregation. Figure 3 The parameters s and 303.

[0094] Figures 35(A) and 35(B) also identify a second online technique that can operate in parallel with data transmission, referred to as the maximum amplitude correlator distribution estimation of the second online channel estimation method 4408. In this second online channel estimation method 4408, it is assumed that under normal channel conditions, the correlator with the maximum output amplitude corresponds to the transmitted message value. Since any correlator in any transmission outputs either the maximum amplitude or not the maximum amplitude, the value recorded for each of the two cases in all considered received transmissions 4402 provides two distributions for any correlator: the distribution when the correlator provides an output with the maximum amplitude, and the distribution when the correlator provides an output without the maximum amplitude. This is shown in the middle column 3601 of Figure 2726, which illustrates the correlator output distribution of the maximum amplitude method.

[0095] Now for reference Figure 27 According to some exemplary embodiments of the present invention, the correlator output distribution 3600 of the maximum amplitude method exists. There are a total of M*2=32 distributions of maximum and non-maximum correlator output amplitudes, as shown in the middle column 3601 of Figure 2726. As shown, the number of distributions here is between 2 and at most 2. 2k Between. As shown in the figure, it is also possible to aggregate all maximum amplitude distributions (M=16) of 3603 and all non-maximum amplitude distributions (also M=16) of 3604, each set being a separate distribution that produces two distributions, as shown in the right-hand column 3602 in Figure 2726. Similarly, after distribution aggregation, distribution fitting methods (in some examples, such as Matlab's histfit function) can be used to estimate the parameters s and s of each aggregated distribution. Then, assuming the parameters s of the maximum amplitude distribution of the aggregation are... Provides the correct distribution of the aggregation Figure 3 A reasonable estimate of the corresponding parameter 303. Similarly, it can be assumed that the parameters s and of the non-maximum amplitude distribution of the aggregation are... Provides incorrect distribution of aggregation Figure 3 A reasonable estimate of the corresponding parameter 303.

[0096] It should be noted that the maximum amplitude correlator distribution estimation method in the second online channel estimation method 4408 of Figures 35(A) to 35(B) can be further refined through the following steps: completing the parameters s and The first estimate, and then the LLR is provided to Figure 1 Before the channel decoder 104, these parameters are used to calculate the LLR, as described in detail in the "Calculating Symbolic LLR" section below. In some examples, Figure 1The channel decoder 104 can then attempt to remove any errors in the LLR sequence and use, for example, a cyclic redundancy check (CRC) to determine if it has succeeded. If not, the channel decoder 104 can then... Figure 1 The data analyzer circuit 116 provides feedback LLRs. These LLRs can then be considered, and in some cases, they can cause the classification of correlator outputs between the maximum amplitude group and the non-maximum amplitude group to be overridden. More specifically, when the maximum amplitude group is formed, if the feedback LLR provides a sufficiently strong indication that the amplitude does not reflect the correct transmission, then the correlator output with the maximum amplitude for a particular transmission can be swapped for another correlator output that does not have the maximum amplitude.

[0097] Summary

[0098] In summary, three methods for estimating the output amplitude distribution of the correlator have been described with reference to Figures 35(A) to 35(B):

[0099] The offline channel estimation method 4405 using correlator dataset 112 has a) 2 2k a) 2 (maximum) and b) 2 (minimum), the number of distributions;

[0100] The online method using the synchronization sequence of the first online channel estimation method 4407 has a) 2 2k a) 2 (maximum) and b) 2 (minimum), the number of distributions; and

[0101] The second online channel estimation method 4408 is an online method based on the maximum amplitude correlator output, with a) 2M and b) 2, the distribution number of...

[0102] In each of the three methods described above, the method using two distributions includes a first aggregated distribution of all correlator output amplitudes, where, among all M correlator outputs in the same transmission, it is "assumed" that each correlator output corresponds to its transmitted signal in its transmission. Similarly, there exists a second aggregated distribution of all correlator output amplitudes, where it is "assumed" that each correlator output does not correspond to its transmitted signal in its transmission. In other words, regardless of which of the three methods is used, the two aggregated distributions in each method are assumed to refer to the same two sets of correlator outputs in all three methods.

[0103] To interpret the above method differently, the M correlator outputs of any given transmission can be divided into two groups. The first group comprises a single correlator output corresponding to one of the M possible signals, which is selected as the transmitted signal. The second group comprises M-1 correlator outputs corresponding to M-1 possible signals, none of which are selected as the transmitted signal. Assuming several transmissions occur, the amplitude of the first group of aggregate correlator outputs for those transmissions can be represented by a single distribution known as the first aggregate correlator output amplitude distribution. Similarly, the second aggregate correlator output amplitude distribution represents the second group of aggregate correlators for those transmissions. Therefore, using any of the three distribution estimation methods listed above, the method with two distributions refers to the first and second aggregate correlator output amplitude distributions.

[0104] In the inventors' research, as described below, the correlator distribution of the offline channel estimation method 4405 is measured from the available dataset and stored in lookup table 112 to minimize the impact of estimation errors. Furthermore, the two-aggregate distribution method (two first aggregate distributions and a second aggregate distribution) is employed due to its versatility compared to other methods with more distributions and its robustness to varying channel conditions.

[0105] In summary, a method for estimating the parameters of the output amplitude distributions of the first and second aggregation correlators has been described, in which at least one of the following is used:

[0106] In the first online channel estimation method 4407 of Figures 35(A) to 35(B), the first set of distribution parameters of the first aggregate correlator output amplitude distribution is estimated by fitting the first probability distribution to the amplitude of the correlator output obtained by correlating the received synchronization signal with a known synchronization signal.

[0107] In the first online channel estimation method 4407 of Figures 35(A) to 35(B), the first set of distribution parameters of the second aggregate correlator output amplitude distribution is estimated by fitting the second probability distribution to the amplitude of the correlator output obtained by correlating the received synchronization signal with a signal different from the known synchronization signal.

[0108] In the second online channel estimation method 4408 of Figures 35(A) to 35(B), the third probability distribution is fitted to the set of 2 k The two correlators obtained 2 k The first set of distribution parameters for estimating the amplitude distribution of the first aggregated correlator output is obtained by considering the amplitude of the correlator output with the largest amplitude among the set of correlator outputs.

[0109] In the second online channel estimation method 4408 of Figures 35(A) to 35(B), the fourth probability distribution is fitted to the set of 2 k The two correlators obtained 2 k The output group of each correlator does not have 2 with the maximum amplitude. k -1 The amplitude of the correlator output is used to estimate the first set of distribution parameters of the amplitude distribution of the second aggregate correlator output;

[0110] In the offline channel estimation method 4405 shown in Figures 35(A) to 35(B), from Figure 1 Select the first set of distribution parameters for the output amplitude distribution of the first and second aggregate correlators from lookup table 112.

[0111] In the examples of this invention, it is also demonstrated that the Rice distribution can be used to represent a first probability distribution, a second probability distribution, a third probability distribution, or a fourth probability distribution.

[0112] Calculate OS LLR

[0113] The log-likelihood ratio (LLR) is a form of soft decision-making that expresses not only the maximum possible value of an uncertain decision but also how likely that value is. LLRs can have any real value and are commonly used to represent the likelihood of a bit value being 0 or 1. When bits are represented by 0 and 1, they are called "hard bits," and if bits are represented using LLRs, they are called "soft bits," and these LLRs are called "bit LLRs." However, it is well known that, in contrast to digital bits, LLRs can also be used to represent analog signal values ​​or symbols. In this case, those LLRs can be called symbolic LLRs, and the signal represented by a symbolic LLR can be called a "soft signal," as opposed to a "hard signal" when analog signal values ​​are used without forming another representation. More specifically, if a symbol has M possible values, then a set of M LLRs can be used to represent the associated probabilities, where each LLR compares the probability of a particular value occurring with the probability of that value not occurring.

[0114] The inventors of this invention have developed a Matlab function that takes the correlator output amplitudes as a set of inputs to a soft demapper circuit and uses the correlator output distribution to convert them into a set of symbolic LLRs (sometimes referred to as soft amplitudes) as a set of prior soft signals, such as... Figure 1 The conversion to the OS LLR demapping circuit 113 is shown in the figure. Therefore, it can be said that... Figure 1 Middle 2 k The parameters s and the output amplitudes of the first and second aggregate correlators are related to the output amplitude distributions of the first and second aggregate correlators. 111 combined to obtain 2k A set of prior soft signals 110 with soft amplitudes. It should be noted that if a non-Rice distribution is used to model the aggregated output amplitude distribution, the corresponding parameter 111 can be used instead in this LLR calculation operation. The following discussion uses the case where k=4 and M=16 as an example. However, it is envisioned that the concepts described herein can be readily extended to other values.

[0115] exist Figure 1 In the receiver 102, a demapper circuit 113, configured to convert the correlator output into OS symbol LLRs, acquires M=16 correlator output amplitudes 109 (correct and incorrect) for each received symbol and returns M=16 corresponding symbol LLR values ​​as soft signals 110. This operation is performed for all 4402 received symbols in a frame. Each of these M=16 symbol LLRs corresponds to the same received symbol (i.e., a 4-bit message, k=4), and they collectively represent the probabilistic representation of the 4-bit message. Later in the receiver 102, a bit-to-LLR circuit 119 converts the original 4-bit message's M=16 symbol LLRs or soft signals 110 into k=4 bit LLRs as an external soft bit set 103, where each bit LLR corresponds to one bit of the 4-bit message. In summary, each transmitted symbol 115 results in the generation of a set of M=16 symbol LLRs or soft signals 110 in the receiver, which are then converted into a set of k=4 bit LLRs or soft bits 103.

[0116] The description will then focus first on the operation of the demapper circuit 113, which is arranged to convert the correlator output into OS symbol LLRs, with each transmitted symbol 115 outputting M=16 OS LLRs or soft signals 110. The description will then focus on... Figure 1 The operation of the conversion to bit LLR circuit 119 is such that each transmitted symbol outputs k=4 bits of LLR or soft bits 103.

[0117] The developed Matlab function implements the conversion to the OS symbol LLR 4410 shown in Figures 35(A) to 35(B). Instead of accepting only 16 correlator values, it returns a symbol LLR of any number of M correlator values ​​fed to it as a vector, as shown in step 4409 of Figures 35(A) to 35(B). In some examples, the function takes the following values ​​as input: a vector of correlator values, modulation parameter M (M=16 in this run), non-centrality of two correct and incorrect distributions, and a scaling parameter (s). 正确 , 正确 s 不正确 , 不正确The function also outputs a number of symbolic LLRs equal to the length of the input correlator vector.

[0118] Now for reference Figure 34 Plot 700 shows the conversion from AWGN channel to OS symbol LLR for the OS scheme datasets of Figures 5(A) to 5(C). Also referring to Figures 35(A) to 35(B), plot 800 shows the conversion from multipath communication channel to OS symbol LLR for the OS scheme datasets of Figures 6(A) to 6(C). Plot 700 shows the symbol LLR AWGN value 701 dataset, and plot 800 shows the symbol LLR multipath value 701 dataset provided by the developed Matlab function across the entire range of correlator output values; from zero to the maximum value (i.e., the maximum value of the correct correlator output). Similar to the correlator output distribution, the range of correlator output values ​​increases as SNR decreases. For example, for the AWGN channel, when the minimum correlator output value for all SNRs is zero, the maximum correlator output value increases from approximately 70 in the 0dB-SNR dataset to approximately 100 in the -7.5dB-SNR dataset. As expected, the functions represented by the symbol LLR plots are strictly increasing. This is expected, because there shouldn't be an LLR value for more than one correlator output. The sign LLR range at the positive and negative ends decreases as the SNR decreases. This is also expected, because the determinism of the signal representing certain values ​​(as shown by the correlator output) must decrease as the noise power increases relative to the signal power, and this is captured by the probabilistic nature of the LLR value.

[0119] The definition and derivation of the symbol LLR are as follows:

[0120]

[0121] Applying Bayes' theorem, we give

[0122]

[0123] Here, the conditional probability is characterized by the Rician distribution, and we obtain: as well as Perform these replacements and give...

[0124]

[0125] In summary, a method for OS demodulation 118 has been described, in which... Figure 1 2 k The output amplitude 109 of each correlator is combined with a first set of distribution parameters 111 of the output amplitude distributions of the first and second aggregate correlators to obtain a set of parameters including 2 kA set of prior soft signals 110 with soft amplitude, as shown in steps 4410 of Figures 35(A) to 35(B).

[0126] According to a second aspect of the invention, it is described Figure 1 An example implementation of the bit-to-LLR circuit 119 is provided to illustrate how these two conversion functions can be evaluated.

[0127] Conversion to bit LLR and EXIT graph evaluation: OS solution

[0128] Now for reference Figure 9 A top-level block diagram of an alternative OS scheme transmission system 900 according to an exemplary embodiment of the present invention is shown. For testing purposes, an alternative OS scheme transmission system 900 was developed in Matlab to perform soft demapping of the output amplitude 109 of the correlator of the demodulator 118 to the external bit LLR 904 in the receiver chain 901, which is performed in the soft demapping circuit 907, providing compatibility with the iterative decoding principle. The test platform (also developed in Matlab) evaluates the new soft demapping functionality by measuring the quality of the external bit LLR and presents the quality results in the form of an external information transmission (EXIT) graph.

[0129] In this example, the bit-to-LLR circuit 906 is developed to provide compatibility with the iterative decoding setup next to the demapper circuit 113, which is arranged to convert the correlator output into OS symbol LLRs in order to construct the soft demapper circuit 907 or soft demapping function.

[0130] Iterative decoding principle

[0131] Similarly, each group of M=16 OS LLRs (k ​​= 4) 908 generated from the demapper circuit 113 corresponds to a 4-bit message. Assuming we find k=4 bit LLRs for each 4-bit message (one bit LLR per message bit), each group of M=16 OS LLRs needs to be converted into a group of k=4 bit LLRs. Therefore, a total of N bit LLRs 909 can be obtained from the N / 4×16 OS LLRs from the bit-to-LLR conversion circuit 906, as shown below. Figure 9 As shown. Although by default this conversion can be performed without using feedback from the decoder—when Figure 9When the feedback bit LLR910 is removed or hardwired to provide a zero value LLR—however, according to the iterative decoding principle, this conversion can also benefit from the feedback bit LLR910 (as shown in step 4418 of Figures 35(A) to 35(B)). In this example, M=16 is used, although the inventors recognize and understand that the concepts and definitions described herein can be applied to M=2. k Any value of .

[0132] During the conversion to bit LLR (step 4413 in Figures 35(A) to 35(B)), feedback is typically used after decoder 912 causes cyclic redundancy check (CRC) in CRC decoder 911 to fail 4421 at the end of its decoding process 4416. In this case, instead of requesting a Hybrid Automatic Repeat Request (HARQ) retransmission from transmitter 913, decoder 912 may want to attempt further symbol-to-bit LLR conversion in bit-to-LLR conversion circuit 906, hoping that the use of feedback bit LLR 910 (in step 4418 in Figures 35(A) to 35(B)) will cause decoding to pass in subsequent attempts 4422, thus saving time, bandwidth, and power compared to retransmission. The bit-to-LLR circuit 906, or simply the bit-LLR circuit, stores N / 4 × 16 OS LLRs 4411 in a memory cell until the iterative decoding process ends 4422, at which point it clears the memory. After the second round of OS symbol-to-bit LLR conversion, if decoder 912 fails CRC again in CRC decoder 911 4421, it may still want to try more conversion attempts 4421. This is because further iterations of demapping decoding use (in steps 4418 of Figures 35(A) to 35(B)) feedback values ​​based on past LLR history, so decoder 912 is more likely to pass 4422. If CRC decoder 911 continues to fail 4421, decoder 912 may finally decide to request a retransmission from transmitter 913 (HARQ 4422), or even abort the current code circuitry, instead of another iteration of demapping decoding 4421. The process of multiple iterations of decoding 912 and conversion in the symbol-to-bit LLR (in the soft demapper circuit 907) to bit-to-bit LLR circuit 906 (step 4413 in Figures 35(A) to 35(B)) is called the iterative decoding principle or Turbo principle [1].

[0133] Terminology: LLR type

[0134] While the generation of one set of N-bit LLRs may involve multiple iterations of demapping-decoding (due to consecutive CRC failures 4421), the generation of another set of bit LLRs may pass CRC 4422 in the first round and require no iterations. In any round of demapping-decoding, either in the first round without feedback (step 4412 in Figures 35(A) to 35(B)) or in subsequent rounds with feedback (step 4418 in Figures 35(A) to 35(B)) (step 4421 in Figures 35(A) to 35(B)), the signal values ​​between the decoder circuit 912 and the soft demapping circuit 907 can be categorized into three types: a priori LLRs, extrinsic LLRs, and a posteriori LLRs. A priori LLRs are always inputs and contain pre-existing information about the circuits to which they are fed. Extrinsic LLRs are always outputs and contain new information based on the computations performed in the circuits that generated them. The posterior LLR is also an output, but it represents all the information from the first iteration (including pre-existing and new information) as well as all symbols of the same code.

[0135] Decoder core 915 with soft output typically generates a posteriori information in the form of a posteriori LLR 914 as output. This can be done by using internal memory to update the message information state over time, thus collecting all information. Alternatively, the content received by the decoder as input can itself represent all the information, so the decoder's output is in the form of a posteriori LLR 914. Although the decoder typically outputs a posteriori LLR, such as Figure 9 As shown in 914, the data type exchanged between the decoder 912 and the soft demapper circuit 907 is an external LLR. If this is not the case, and the two circuits swap the a posteriori data, there will be positive feedback causing the data to diverge and be corrupted [1].

[0136] Data Flow Example

[0137] exist Figure 9The diagram illustrates an example of how three different types of LLRs flow between decoder 912 and soft demapper circuit 907, and is also explained with reference to Figures 35(A) to 35(B). Assume that a new code, code A, is received at receiver chain 901. As in step 4409 of Figures 35(A) to 35(B), demodulator 118 has generated a set of N / 4 × 16 correlator output amplitudes 109 for code A, and in step 4410 of Figures 35(A) to 35(B), the same number of OS LLRs are output from demapper circuit 113, which is arranged to convert the correlator outputs into OS symbol LLRs. Bit-to-LLR conversion circuit 906 receives these OS symbol LLRs 908 as prior data and converts them (as shown in step 4413 of Figures 35(A) to 35(B)) into N-bit LLRs 909. The bit-to-LLR circuit 906 also stores copies of the 4411 OSLLRs in its internal memory for reuse when applying iterative decoding 4421 (step 4419 in Figures 35(A) to 35(B)). At this point (as shown by step 4412 in Figures 35(A) to 35(B)), the feedback bit LLR 910 does not contain an LLR (or equivalently carries N zero-value LLRs) because this is the first time the data corresponding to code A is processed in the soft demapper circuit 907. Typically, in some examples of the invention, the bit-to-LLR circuit 906 generates its output based on the received N / 4 × 16 OS symbol LLRs 908 and the N feedback bits LLR 910 received from the feedback line. Assuming the feedback bits LLR 910 represent pre-existing data, even if they take zero values ​​in the first iteration (step 4412 in Figures 35(A) to 35(B)), the output of the bit-to-LLR circuit 906 will be an N-bit LLR 909 of a posteriori type.

[0138] Because the decoder 912 provides feedback bits LLR 910 and there is a closed-loop data flow between the decoder 912 and the soft demapper circuit 907, the decoder 912 expects the extrinsic type input LLR as part of its processing, which is the opposite of the a posteriori LLR, which will create a positive feedback loop [1]. Therefore, the soft demapper circuit 907, which is set in the closed loop together with the decoder 912, subtracts the N a posteriori bits LLR 910 from the N a posteriori bits LLR 909 output of the bit LLR circuit 906 to generate N extrinsic bits LLR 904, as in step 4414 of Figures 35(A) to 35(B). As described above, the first iteration of the conversion-to-bit LLR circuit 906 for code A (step 4412 in Figures 35(A) to 35(B)) will have a zero-valued a priori bit LLR for the feedback bit LLR 910. Therefore, the extrinsic bit LLR 904 output from the soft demapper circuit 907 will be equal to the output of the conversion-to-bit LLR circuit 906 in the first iteration. The extrinsic bit LLR 904 of the demapper output from the soft demapper circuit 907 is deinterleaved (π) in the decoder 912. -1 This is followed by N prior bits LLR 916, which are then fed into the decoder core 915 as the prior type of decoder 912. After this, decoder 912 can use these prior LLR 916 to generate the decoded bit vector 917, which is then fed into... Figure 9 The CRC decoder 911 is shown. If the CRC decoder 911 passes check 4422, the transmission process is successfully completed.

[0139] In Figures 35(A) and 35(B), if decoder 912 fails 4421 due to CRC decoding 911 of code A and decides to apply iterative decoding, decoder 912 can generate a sequence 910 of N feedback bits LLR. As a result, the LLR feedback output in the form of, for example, the a posteriori LLR 914 of decoder core 915 will be a posteriori, and for the same reason explained earlier, to avoid positive feedback in the closed loop with soft demapper circuit 907, decoder 912 subtracts the input a priori bits LLR 916 from the a posteriori LLR feedback output 914 of decoder core 915. The resulting sequence of N extrinsic LLR 918 is interleaved (π) and fed back to soft demapper circuit 907 as a priori feedback bit LLR 910 through the feedback line (step 4418 in Figures 35(A) and 35(B)), the feedback line this time possibly having a non-zero value. Then, when the bit-to-LLR circuit 906 also uses (step 4419 of Figures 35(A) to 35(B)) the OS LLR of code A already stored in its memory from the previous iteration 4411, it uses the feedback bit LLR 910 to perform a second OS symbol to bit LLR conversion on code A (step 4413 of Figures 35(A) to 35(B)). The bit-to-LLR circuit 906 will still keep the same OS LLR in its memory because the CRC decoder 911 may fail a second time 4421 and may need a third OS symbol to bit LLR conversion (step 4413 of Figures 35(A) to 35(B)). The decoder 912 determines the end of the iterative decoding, which may occur due to a CRC pass (leading to a successful transmission process) or reaching the maximum number of decoding iterations (so the transmission process is abandoned and considered unsuccessful) 4422. After processing code A, the next code block (code 'B') may have already been received in receiver chain 901 4402, and the above process will be repeated for code B.

[0140] Calculate bit LLR

[0141] In the previous description of the bit LLR conversion of the OS scheme, it was explained how to generate the OS symbol LLR 908 from the correlator output value. Among them, it was explained how to calculate bit LLR 909 in the iterative decoding method based on the prior OS symbol LLR 908 and the feedback prior bit LLR 910 (step 4413 in Figures 35(A) to 35(B)).

[0142] No feedback from the decoder

[0143] First, assuming there is no feedback bit LLR 910 from the decoder, or equivalently, the a priori feedback bit LLR 910 to the soft demapper circuit 907 has a zero value, there exists a 4-bit message m for which k = 4 bits of LLR need to be found; each message bit has one bit of LLR. In some examples, the 4-bit message can have any value from the set of 4-bit binary combinations {0000, 0001, ..., 1111} 3701, such as... Figure 28 As shown in the figure, this diagram illustrates the 4-bit message combination and the corresponding OS symbol LLR parameters. The following parameters can also be used to represent the bit LLR of a message: LLR 3 bit,out , LLR 2 bit,out , LLR 1 bit,out , LLR 0 bit,out They correspond in the same order from the most significant bit (MSB) to the least significant bit (LSB) of the message. For the input to the bit-to-LLR circuit 906, there are M=16 OS symbol LLR values, where each value corresponds to one of the M=16 message combinations 3702. Using parameters... LLR 0 sym,in , LLR 15 sym,in , ..., are used to represent the OS symbol LLR 3703, such as Figure 28 As listed in the table. Therefore, on one side, there are M=16 OS symbol LLR values ​​corresponding to M=16 message combinations, while on the other side, M=16 symbol LLRs will be used to obtain k=4 bit LLRs for 4-bit messages.

[0144] calculate

[0145] Now for reference Figure 29 According to some exemplary embodiments of the present invention, terms in the calculation of the bit LLR are shown when there is no feedback from the decoder. Here, examples are used to better understand the calculation of the bit LLR (step 4413 in Figures 35(A) to 35(B)). If for all i in the range [0, M-1] or [0, 15], it is assumed that... LLR i sym = LLR i sym,in If the LSB of the transmitted message is "1", such as message "0001" or "1011", then it can be expected that, under good channel conditions, demodulator 118 will represent the set probability of all message combinations whose LSB is "1". Figure 29(Column 8 in the table) The probability of this set is relatively higher than the probability of the set of messages whose LSB is "0" ( Figure 29 (See column 7 in the table). Similarly, if bit 1 of a message has the value "1", such as "0010" or "1111", it can be expected that demodulator 118 will represent the set probability of all message combinations whose bit 1 is equal to "1" (see column 7 in the table). Figure 29 (Column 6 in the table) The probability of this set is higher than the probability of the set whose bit 1 is equal to "0" (see column 6 in the table). Figure 29 (Column 5 in the table). This can also be said for the case where the message bit is "0". The decoder does not know the actual value of the message bits, but attempts to determine them by considering the combination probabilities (denoted by LLR), thus obtaining the final message combination that is most likely to be any actual message. Figure 29 The left column 3801 shows different message combinations of a 4-bit message (k=4). The right column 3802 shows how each of the M=16 symbols LLR is related in the bit LLR calculation (step 4413 of Figures 35(A) to 35(B)) for each bit of the 4-bit message.

[0146] To calculate (step 4413 in Figures 35(A) to 35(B)) the bit LLR value of a bit, according to Figure 29 First, the M=16 symbol LLRs are divided into two groups of 8 LLRs each. For each bit of a 4-bit message (with an index in the range [0, k-1] or [0, 3]), there are M / 2 = 8 message combinations where the same bit index has a value "1", and M / 2 = 8 combinations where the same bit index has a value "0". For example, for bit 3 (MSB) of a 4-bit message, the MSB of message combination {0000, 0001, ..., 0111} is equal to "0", and the MSB of message combination {1000, 1001, ..., 1111} is equal to "1". Therefore, the LLR values ​​corresponding to the first and second groups of 8 message combinations with the MSB of this message are { LLR 0 sym , LLR 1 sym , … , LLR 7 sym}and{ LLR 8 sym , LLR 9 sym , … , LLR 15 sym}, and respectively in Figure 29 As shown in columns 1 and 2. It is well known that the log-likelihood ratio is defined as:

[0147]

[0148] And multiplication and division in the probability domain correspond to addition and subtraction in the LLR domain, respectively. Therefore, for example, for bit 3, the bit LLR value will be the LLR probability when bit 3 is "0" minus the LLR probability when bit 3 is "1", that is, LLR 3 bit,out = L 3 bit,0 - L 3 bit,1 ,like Figure 29 As shown in 3803.

[0149] In order to calculate Figure 29 The parameter in the bottom line 3804 L 3 bit,0 , L 3 bit,1 , L 2 bit,0 Each of these needs to be combined with the corresponding LLR value. As an instance, this will be... L 3 bit,0 The LLR parameter in column 1. To obtain the probability of an event occurring for any message combination in a given set of combinations, the probabilities of all combinations in the given set must be added together. This is because message combinations are mutually exclusive—only one message combination is the actual value of a 4-bit message. For example, the probability that a message is one of the values ​​{"0000", "0101", "1010"} is Pr["0000"]. "0101" "1010"] = Pr["0000"] + Pr["0101"] + Pr["1010"].

[0150] The sum of probabilities can be approximated using the Jacobian logarithm in the LLR field. The Jacobian logarithm is:

[0151]

[0152] And assumptions and If the probabilities are mutually exclusive, then it can be solved by writing... They are combined using their LLR values. This operation is called the 28axstar operation, and is used with... It is represented by symbols. The addition of probabilities can be represented using the 28axstar operator in the LLR field. For example, L 3 bit,0 Defined as L 3 bit,0 LLR 0 sym LLR 1 sym … LLR 7 sym It contains Figure 29 The items in column 1. In the same way, the other parameters... L 3 bit,1 , L 2 bit,0 ...can be seen from Figure 29 The results are calculated from their corresponding columns.

[0153] There is feedback from the decoder.

[0154] As mentioned earlier, for each 4-bit message (k=4), there are k=4 prior feedback bits LLR 910 fed back from the decoder 912 to the bit conversion LLR circuit 906, such as... Figure 9 As shown. The following parameters are used to represent the prior soft bit 910 (feedback bit LLR). LLR 3 bit,in , LLR 2 bit,in , LLR 1 bit,in , LLR 0 bit,in These correspond to the MSB to LSB of the message, respectively.

[0155] Suppose that the soft demapper circuit 907 is now receiving multiple inputs, which can be listed according to the following formula:

[0156] The M output amplitudes of the correlator are referred to as the first input set of the soft demapper circuit 907. They are used to generate the first a priori soft signal set including M OS symbols LLR 908;

[0157] The second input set of the soft demapper circuit 907 can be defined as a second a priori soft signal set including k feedback bits LLR 910 (or k soft bits).

[0158] The current problem is how to improve the k=4 feedback bits LLR909, which is the output of the LLR 906 converted to a bit LLR circuit, by using these new k=4 values ​​of the soft bits or feedback bits LLR 910, together with the previous M=16 OS symbols LLR 908. Figure 28 The diagram shows that each of the M=16 OS symbols in LLR 908 corresponds to... Figure 29The left sidebar 3801 shows the probability of one of the M=16 message combinations. However, in this case, what is the correlation of the k=4 prior feedback bits LLR 910 with respect to each message combination? In some examples, a positive LLR is known to mean that the bit is more likely to be "0", while a negative LLR means that it is more likely to be "1". Now, if the actual message is "0001", the prior LLR corresponding to bits 1-3 can be expected to be positive, and the LLR value corresponding to bit 0 can be negative. Therefore, for the message "0001", the expected value is... LLR 3 bit,in + LLR 2 bit,in + LLR 1 bit,in – LLR 3 bit,in It is a positive value. Similarly, we can say, for example, that if the message is "1010", the following values ​​can be expected to be positive: - LLR 3 bit,in + LLR 2 bit,in – LLR 1 bit,in + LLR 0 bit,in This interpretation applies to all M=16 message combinations: if a message combination is an actual received message, its corresponding prior bit LLR combination is expected to be positive. Figure 30 A 4-bit message combination is shown, which has a corresponding OS symbol LLR parameter and a priori bit LLR combination. A priori bit LLR combination 3901 is listed, and calculating the output bit LLR (step 4413 in Figures 35(A) to 35(B)) requires dividing by 2 in each combination.

[0159] The above explanation can be further understood as follows: if the message combination is the actual message, its corresponding a priori bit LLR combination is expected to have the maximum value among all other bit LLR combinations 3901. Therefore, by calculating... Figure 29 and Figure 30 The left column 3801 has M=16 prior bits of LLR combinations, and the largest combination will point the decoder to the message combination that is most likely to be the actual message according to k=4 bits of LLR.

[0160] So far, in some examples, for Figure 29 and Figure 30 The left column 3801 contains M=16 message combinations, and there are two sets of M=16 soft signals indicating the likelihood that the message combination represents an actual 4-bit message: symbol LLR 3703 and a priori bit LLR combination 3901, such as... Figure 30As shown. Since both sets are LLR representations, they can be added together to form a single LLR set corresponding to the message combination. Therefore, if Figure 30 The symbolic LLR parameter is defined as for all i in the range [0, M-1] or [0, 15]. LLR i sym = LLR i sym,in + LLR i sym,bit Then you can use the targeting Figure 29 The method of interpretation is to calculate the bit LLR generated by the bit conversion LLR circuit 906. The resulting bit LLR 909 (as shown in step 4413 of Figures 35(A) to 35(B)) is the a posteriori LLR because it represents all the information, including the newly received OS symbol LLR 908 from upstream and the old a priori feedback bit LLR 910 received from decoder 912.

[0161] Summary

[0162] In summary, the demapper circuit 113 and the bit-to-LLR circuit 906 have been described, which are arranged to convert the correlator output into OS symbol LLRs and are used to generate (step 4413 of Figures 35(A) to 35(B)) the a priori soft bits LLR 909, including a second set of k soft bits 910 provided as the second set of inputs to the circuit. Furthermore, the circuit combines all the a priori soft signals to generate 2 k A set of posterior soft signals, and where 2 k The set of posterior soft signals is combined to obtain a posterior soft bit LLR set 909. Finally, the posterior soft bit LLR 909 is combined with the soft bits or feedback bits LLR 910 of the second prior soft signal set to obtain an external bit LLR set 904 including k soft bits, as shown in steps 4414 of Figures 35(A) to 35(B).

[0163] As shown in the flowcharts of Figures 35(A) to 35(B), the process 4400 of the soft demapper in the OS scheme can be summarized as follows. Initially at 4401, the demapper receives the signal set 4402 as the output of the demodulator, which corresponds to the transmitted frame. Then, in some examples, it is decided which of three different methods to use for estimating the correlator output distribution. If the offline method 4423 is selected, channel characteristics, such as signal-to-noise ratio and channel type (e.g., AWGN), are identified in the offline channel estimation method 4405. The results of the identification are then used to address a pre-calculated table of distribution parameters of the correlator output amplitude, such as... Figure 1The pre-calculated table 112. Conversely, if an online distribution estimation method is selected at 4424, the selection will be either using a synchronization sequence in the first online channel estimation method 4407 or applying the maximum amplitude method of the second online channel estimation method 4408. Regardless of the estimation method used, the estimated parameters of the amplitude distribution and the correlator output amplitude collected in step 4409 of Figures 35(A) to 35(B) are used to calculate the OS soft signal in step 4410 of Figures 35(A) to 35(B), which can be in the form of an LLR, as previously described in the section on calculating the OS LLR.

[0164] Subsequently, the calculated OS soft signal 4410 is stored 4411 in internal memory in case a demapping decoding round requiring iterative decoding later. In the first iteration of a possible iterative decoding process, there is no value on the feedback path from the decoder. Therefore, the feedback will be all zero values, as shown in steps 4412 of Figures 35(A) to 35(B). Next, the a posteriori soft bits in the form of LLR (OS bits LLR) are calculated, as shown in steps 4413 of Figures 35(A) to 35(B). This is followed by converting the a posteriori soft bits (as shown in steps 4414 of Figures 35(A) to 35(B)) into extrinsic soft bits of the type required by the decoder and sending them (4415) to the decoder. After the decoder has completed the process of decoding the received extrinsic information, the demapping unit receives (4416) the CRC status and the decoded soft bits from the decoder. After checking the CRC status at 4417, if the check fails at 4421, Turbo decoding can be applied in some examples of the invention, and the decoded soft bits received at 4416 can be considered as prior information fed back from the downstream decoder (as shown in step 4419 of Figures 35(A) to 35(B)). Furthermore, the OS soft signal previously calculated at 4410 is loaded from memory as another set of prior information for soft bit calculation (as shown in step 4418 of Figures 35(A) to 35(B)). Symbol-to-bit LLR conversion uses all prior information to perform a second calculation of the soft bits as bit LLR at 4413. The demapping-decoding iteration of Turbo decoding ends at 4422 by either passing the CRC check or by abandoning further decoding operations on the current frame, which may result in, for example, Hybrid Automatic Repeat Request (HARQ) retransmissions. At this point, the demapping of the current frame ends at 4420.

[0165] Test platform

[0166] A method has previously been described that converts the OS symbol LLR 908 into a 4413-bit LLR 909 using an iterative decoding principle, providing the decoder 912 with the external bits LLR 904 of the received message, for example, in soft bits. Here, the evaluation of the conversion to the external bits LLR 904 is described. Different methods can be used to evaluate the result of the conversion to the bit LLR circuit 906, which is also the output of the external bits LLR 904 of the soft demapper circuit 907. One method is to observe the bit error rate (BER) when using soft decision demapping and compare it with the BER of the same transmission, but using a hard demapping of hard bits to hard bits instead of LLRs using the correlator output value. Other methods characterize the external bits LLR 904 generated from the soft demapper circuit 907, such as a measured mutual information (MI) histogram and a measured mutual information average, which will be described later.

[0167] It should be noted that because some of the methods described above rely on a comparison of the decoded bit 919 with the transmitted bit 920, they are only applicable to laboratory testbeds where the transmitted bit 920 is accessible, rather than to real-world systems where only the 4402 data 921 is received. Examples of these laboratory-only evaluation methods include BER and MI histogram measurements. However, although these methods are only applicable to laboratory environments, they provide good measurements for evaluating the quality of the external bit LLR904 output of the soft demapper circuit 907.

[0168] Advanced computing

[0169] A test platform was developed in Matlab to measure the quality of the external bit LLR 904 output from the soft demapper circuit when the soft demapper circuit 907 uses the bit LLR conversion described above. Figure 10 A block diagram of a test platform 1000 for a 16OS scheme (orthogonal signaling of M=16, and a dataset taken from [5]) is shown. As shown, there is no encoder or decoder: instead, random bits from a uniformly distributed random number source 1002 are used to simulate coded bits 1001. Furthermore, data intended to be fed into the decoder in the actual system is generated from a soft demapper circuit 907 and subjected to several measurements (such as BER 1005 and MI calculations 1006, 1007). The measurement results are used by a final processing function 1008, which plots an EXIT graph. Figure 10In this set, blocks and procedures in sets 113, 114, 906, 907, 1001, 1004, 1013, 1020, 1023, 1024, 1026, 1031, 1032, and 1035 correspond to portions of the test platform shared with the actual transmitter-receiver communication system 900. The remaining blocks and procedures are specific only to the laboratory test platform and are generally not present in the actual system.

[0170] Each execution of the test platform is based on the selection of the real channel and the estimated channel, generating an EXIT chart 1011. In a real system, receiver chain 901 may not perfectly estimate channel characteristics. Figure 10 In the test bench, this is represented by simulating transmission using a first channel model (referred to as the real channel model 1009) and providing channel characteristics obtained by a (potentially imperfect) channel estimator using a second (potentially different) model 1010. In practice, in some applications, there may be no channel estimation at all in the receiver chain 901, and a fixed set of (worst-case) channel characteristics can be assumed regardless of the actual channel characteristics. The user of the test bench will select two channel models as inputs 1009 and 1010, such as... Figure 10 As shown on the right, and the expected output EXIT chart 1011 specific to the two selected channels is shown on the left side of the block diagram. The test platform 1000 is applied to several real and estimated channel combinations, where their EXIT charts are as follows... Figures 12 to 1 Figures 4(A) through 14(B) are shown. In these figures, each EXIT chart has the same true and estimated channel types, but sometimes the SNR differs. The true and estimated channels in each EXIT chart are based on either an additive white Gaussian noise (AWGN) model or a multipath (MP) model. Although Figure 12 The actual channel and the estimated channel have different SNRs, but Figures 13(A) to 13(B) and Figures 14(A) to 14(B) have the same SNR. Figure 11 This is an example EXIT chart when both the real channel and the estimated channel are based on the AWGN_-7.5dB dataset.

[0171] Mutual Information

[0172] The EXIT graph illustrates the characteristics of the unit that generates the EXIT graph; this unit could be a demapper or a decoder. In the example described in this paper, which seeks to evaluate a soft demapped model, Figures 11 to 1All EXIT charts 1100, 1200, 1300, and 1400 shown in Figures 4(A) through 14(B) belong to the soft demapper circuit 907. In these charts, the X-axis is the mutual information of the prior bit LLR 1013: it is a measure of the feedback quality (which would come from the channel decoder in a real system, but in test platform 1000, it comes from the random LLR generation circuit 1014) and is fed to the bit-converted LLR circuit 906. The mutual information between two random variables quantifies the amount of information in one variable that can be obtained by observing the other, and the MI value is in the range [0, 1]. An MI value of MI=0 means there is no feedback from the decoder, and the feedback line will consist of a sequence of N zero-value bit LLRs. Any MI>0 corresponds to the presence of feedback from the decoder, and a larger MI value means better quality feedback. Figures 11 to 1 The Y-axis in the EXIT charts 1100, 1200, 1300, and 1400 shown in Figures 4(A) to 14(B) is also a quality measurement between 0 and 1, and corresponds to the output value generated by the soft demapper. In each of these EXIT charts 1100, 1200, 1300, and 1400 in this document, there are three curves 1101, 1102, and 1103 and a point represented by a cross 1104, as well as some numbers 1105 regarding the maximum coding rate achievable by the soft decision method.

[0173] Test platform process

[0174] To generate EXIT chart data, the test platform is cyclically tested across multiple test indices 1016, where each 1017 corresponds in the EXIT chart 1011 to a point in all charts that has the same MI value 1018 for the prior bit LLR 1013. For each test index 1017 of the test indexes 1016, a prior MI value 1018 is selected. The system is simulated using this MI value 1018, measuring the demapper output (such as BER 1005) of the N extrinsic bit LLR 1004 quality parameters, and ultimately identifying some points in the EXIT chart 1011. For example, the test platform where MI=0.3 is executed as follows: Figure 11 Each of the curves 1101, 1102, and 1103 in the EXIT chart 1100 generates a point.

[0175] transmitter

[0176] In each simulation with test index 1017, coded bits 1001 are generated from a uniformly distributed random number source 1002 and fed to a bit-to-symbol circuit 114, which converts the bits into symbols based on modulation with 16 symbols (M=16). Therefore, for N coded bits 1001, there will be N / 4 symbols (k=4) 1020 to be modulated.

[0177] Dataset

[0178] The datasets included in the previous correlator distribution survey and the OS scheme to symbol LLR conversion represent the models used in this test platform for different modulation schemes and channel types, and an additional dataset is added. These datasets are provided by [5] and are represented using the following notations: AWGN_100dB, AWGN_0dB, AWGN_-3dB, AWGN_-7.5dB, MP_0dB, MP_-3dB, MP_-7.5dB. Each of these datasets is a large matrix of 8100 rows, where each row contains (a) symbol values ​​3502 in the range of [0, M-1] or [0, 15], and (b) M=16 correlator output amplitudes 3503, which are generated for the symbol values ​​according to the mathematical models of the corresponding modulation schemes and channels. Figure 10 The memory 1012 in the memory stores and copies all of the above datasets, of which there are a total of seven datasets in this example.

[0179] Models of the real channel and the estimated channel

[0180] In this example, selecting the real channel 1009 means choosing one of the seven available datasets and treating it as the modulation-channel simulation model 1021 of the test platform 1000. During the test platform simulation, for the data from... Figure 1 The bits of each symbol value generated by the symbol circuit 114 are derived from... Figure 10 In the selected modulation and channel model 1022, rows with the same sign value are randomly selected, and the corresponding 16 correlator values ​​in the same row are output in 1023. This preserves the correlation between the correlator outputs. However, it should be noted that the storage in the channel between consecutive transmissions is not modeled in this way. However, this is reasonable because the proposed soft-decision demodulator does not utilize this storage, and because the spreading sequence used in [5] is designed to mitigate the dispersion that leads to this storage. The user also needs to select the estimated channel. In this example, Figure 10The data analyzer circuit 1024 fits the Rice distribution to all rows of the user-selected estimation dataset 1025 (1010) and returns four parameters 1026 corresponding to the distributions of the correct and incorrect correlator data sets, as defined and explained in the previous correlator distribution survey and the conversion to the notation LLR of the OS scheme. While the channel estimation task of the data analyzer circuit 1024 is performed in parallel with the simulation in this test platform 1000, in a real system, it can be performed online by analyzing the received signal, or offline using a very large set of channel data if a fixed set of channel characteristics is to be assumed. Performing channel estimation offline avoids significant online complexity overhead.

[0181] Hard judgment process

[0182] In the test platform 1000, a hard-decision process 1027, including receiver 1028, is used to calculate its BER 1029 and compare it with the BER performance 1030 of a soft-decision bit LLR model. The hard-decision process of symbol circuit 1031 receives M=16 correlator values ​​from the channel model and performs hard decision by using the index 1034 of the largest correlator value as a 4-bit message value (k=4), then passes the index to bit conversion circuit 1032 to generate equivalent bits 1035. For example, if the largest correlator value in M=16 has an index "9", the result of the hard demapper circuit 1033 for that symbol will be "1001". As the hard demapper circuit 1033 demaps all symbols into bits, the BER measurement circuit 1036 compares these bits with the true encoded bits 1001 from transmitter chain 1037 and calculates the BER value using the formula BER = number of erroneous bits divided by the total number of bits. The BER value 1029 will be a fraction in the range [0, 1]. While a smaller BER indicates better performance, plotting the 1-BER value (using the cross "X") on the EXIT plot for comparison with other quality measures shows that a larger value also indicates better performance, and similarly, values ​​between 0 and 1 are used. The hard-decision process 1027 is independent of the soft-decision process 1038 and its iterative feedback, and for each dataset, there will be only one hard BER value 1029 1104. Furthermore, since the hard-decision process 1027 does not use feedback, its BER 1029 can be compared to the case where there is no feedback in the soft-decision process. Thus, as... Figure 11 As shown, in the EXIT chart, 1104 hard decision BER points are plotted at the prior MI=0.

[0183] LLR quality

[0184] The soft decision process 1038 uses the received prior bit LLR 1013 as feedback to the soft demapper circuit 907, such as... Figure 10 As shown. Instead of using a decoder model in the test bench, it is sufficient for the test bench to generate prior bits LLR 1013 to the soft demapper circuit 907, which is different from the decoder. In this case, the decoder is generated by the random LLR generation circuit 1014. In order to simulate a real receiver, the generated prior bits LLR 1013 must be similar to the coded bits 1001 with a certain quality. Therefore, the random LLR generation circuit 1014 takes the real coded bits 1001 as input to create LLRs with this similarity. These random LLRs are generated according to a Gaussian distribution, as shown in [1]. Figure 4 As described in [the text].

[0185] Feedback Prior Bits (LLR): The MI function in this test platform

[0186] Represented by MI value as mutual information Figure 10 The quality of the feedback prior bit LLR 1013 in the test platform 1000. MI is a statistical measure that can be used to represent how much shared information the demodulated extrinsic LLR (or feedback prior bit LLR 1013) has relative to the real encoded bit 1001. Therefore, as a statistical measure, it is expected that the MI value is usually calculated to find the correlation between the encoded and demodulated bits, for example, in a laboratory environment test platform where a decoder model is present. However, in this example test platform 1000, since the feedback prior bit LLR 1013 is not generated from the decoder model but is a function of the real encoded bit 1001, the MI of the encoded and demodulated bits can be defined for different test indices. The MI value 1018 will become the input to the random LLR generation circuit 1014 to generate analog decoded bits. The random LLR generation circuit 1014 contains a mathematical function that produces an LLR whose mutual information with respect to a given input bit is equal to the input MI value. The MI value generation circuit 1039 uses test index 1017 to generate the MI value 1018 in the range [0, 1].

[0187] Soft BER calculation: Why posterior data?

[0188] In soft-decision procedure 1038, calculating the bit error rate requires posterior information from the soft demapper-decoder interface. This is because the output of the soft decoder is typically posterior data (as explained previously in the section describing the EXIT graph evaluation of the bit LLR conversion and OS scheme), therefore the posterior information will be in the form closest to the decoder output. Figure 10The diagram illustrates the acquisition of posterior data from the decoder-demapper interface, where the summation 1040 of N extrinsic bits LLR 1004 and N prior bits LLR 1013 originates from and proceeds to the soft demapper circuit 907. The test platform 1000 then makes a hard decision 1041 on the posterior bits LLR 1042 to provide N bits 1043 from the N posterior bits LLR 1042. Here, a hard decision of binary "0" is made for positive posterior LLRs, and a hard decision of "1" is made for negative LLRs. The resulting N bits 1043 are compared with the actual N encoded bits 1001 from the transmitter chain 1037 according to BER 1005 to provide BER measurements 1030 for the executed test indices 1017 and MI 1018.

[0189] Comparison of soft BER and hard BER

[0190] When MI=0, the comparison between the soft BER value and the hard BER 1029 is valid because both values ​​are calculated based on bits obtained from the same type of (extrinsic) data through demapping from their symbols to bits. This is obvious for the hard decision process 1027, since there is no feedback in this process, and therefore the data from the hard demapping circuit 1033 is extrinsic. Furthermore, since the soft decision stream 1038 has a zero-valued feedback a priori bit LLR 1013 when MI=0, in this case, the posterior bit LLR 1042 is equal to the N extrinsic bits LLR 1004 generated from the soft demapping circuit 907, as shown below. Figure 10 As shown. Therefore, the decrease in BER value from hard-decision procedure 1027 to soft-decision procedure 1038 means that the use of LLR has been able to improve receiver quality by reducing the bit error rate. In the EXIT graph, if the crossover point 1104 of the hard BER values ​​is exactly below the curve 1101 of the soft-decision BER values, the BER performance will be improved because the curve values ​​are in the form of 1-BER. Generating soft-decision BER values ​​also shows the impact of improving the quality of feedback LLR on BER performance. For example, Figure 11 This demonstrates that using the example soft demapping method, the quality of the LLR to the decoder improves as the quality of the feedback LLR increases. It can be observed that... Figures 12 to 1 In all cases shown in Figures 4(A) to 14(B), the soft (decision) demapper circuit 907 produces a BER equal to or better than that of the hard (decision) demapper circuit 1033.

[0191] MI: Histogram and Averaging Methods

[0192] Two other quality measures characterize the output of the soft demapper circuit: the mutual information histogram 1102 and the mutual information average 1103, as shown in the EXIT chart. Both measures provide mutual information and both look at the N extrinsic bits LLR 1004 of the output of the soft demapper circuit 907 to calculate their MI, which is the opposite of the BER measurement that requires the a posteriori bits LLR 1042. The histogram method (Equation (2) in [1]) provides the mutual information 1044 between the N extrinsic bits LLR 1004 and the true encoded bits 1001 from the transmitter chain 1037. This known method does not assume that “the LLR value cannot be wrong”—that is, this known method does not trust the LLR—and checks those LLRs against the true encoded bits 1001. In contrast, the average method (Equation (2) in [1] immediately follows Figure 4 The equation below only considers N extrinsic bits (LLR 1004) and provides a quality measurement (1045) without an external reference. Therefore, it provides a trust-based quality measurement of the LLR. For this reason, the averaging method performs best when the decoding algorithm is optimal.

[0193] The upper limit of achievable coding rate

[0194] When using soft demapper circuits characterized by MI histograms, the MI histogram method can also be used to provide an upper limit on the achievable coding rate. Generally, the achievable coding rate between the decoder and demodulator (including the demapper) depends primarily on two aspects. The first aspect is how well the decoder and demapper match (see the discussion on “area gap” in [1]). Like the demapper, the behavior of the decoder can also be described using an EXIT chart. The EXIT chart of the demapper can be compared with the EXIT chart of the decoder to determine how well they match. The second aspect of identifying the achievable coding rate is the block length applied when using the system: shorter block lengths reduce the coding rate required in practice (see [1] for more details).

[0195] Regardless of whether the decoder's EXIT graph exists or whether a code length is selected, the demapper's EXIT graph contains a value defining a theoretical upper limit to the coding rate, which supports reliable low BER operations using the demapper. Figure 11 The upper limit differs between the case with iterative feedback in 1107 and the case without iterative feedback in 1106. For example, according to Figure 11The EXIT chart shows that when the feedback quality has a MI of zero, the prior MI histogram value is 0.87 (rounded to zero). This means that, as shown in 1106, for an AWGN 7.5 dB channel, the maximum achievable coding rate using the example soft-decision demapping method without iterative feedback is 0.87. In the presence of iterative feedback (when the MI of the feedback is >0), the area under the MI histogram will be the maximum achievable coding rate [1]. Figure 11 In the example, the area under the MI histogram is 0.92 (rounded to zero). Therefore, as shown in 1107, the maximum achievable coding rate in the example soft-decision demapping method is 0.92 when there is iterative feedback for the AWGN_7.5dB channel.

[0196] Result: EXIT chart

[0197] High SNR chart

[0198] Figures 12 to 1 The EXIT charts in Figures 4(A) to 14(B) show that for SNR ≥ -3dB in all channels (a total of 5 charts), there are no bit errors: all hard BER values ​​and soft BER values ​​are zero. Furthermore, for all MI values ​​of feedback quality, the MI quality measurement (histogram and averaging method) of the external LLR of the demapper has a value of 100%. This verifies the soft-decision demapping model for normal high SNR conditions according to an exemplary embodiment of the invention: the model successfully provides the decoder with bit LLR where the real bits transmitted from the transmitter have 100% mutual information and a bit error rate of zero.

[0199] Low SNR chart

[0200] No iterative feedback: BER

[0201] The EXIT charts for the small SNR of -7.5 dB (e.g., two charts, 1301 and 1401, in Figures 13(A) to 13(B) and 14(A) to 14(B)) also demonstrate that the soft-decision demapping model according to the exemplary embodiment of the invention operates as expected. The two EXIT charts for the AWGN 1301 and MP 1401 channels in Figures 13(A) to 13(B) and 14(A) to 14(B), each with an SNR of -7.5 dB, have a 1-BER value of 0.97 (rounded to infinity) for their hard decisions without iterative feedback, and a slightly better value for their soft decisions. This means that the soft demapping model according to the exemplary embodiment of the invention manages to maintain the bit error rate when switching from a hard model to a soft model without feedback.

[0202] Iterative feedback: The impact of feedback MI on BER and the two MI values

[0203] The two low SNR EXIT charts 1301 and 1401 in Figures 13(A) to 13(B) and 14(A) to 14(B) also show that as the quality of the feedback LLR increases (when MI>0), the quality of the output LLR of the demapper circuit and the bit error rate increase. When the feedback quality MI reaches a value of 1, the 1-BER number approaches 100%, which is expected for cases where the feedback is fully informative. Furthermore, for feedback MI=0, the output quality MI values ​​of the histogram and averaging methods of the demapper circuit start from values ​​between 85% and 90%, and become approximately equal to 97% when the feedback MI is “1”. For all feedback MI values, the two MI values ​​also have values ​​close to each other, which means that the two MI measurements reliably represent the quality of the feedback LLR and show that the consistency condition of [1] is met.

[0204] Maximum achievable coding rate

[0205] Depending on the decoder and the selected block length, the receiver may be able to converge to the portion of the EXIT chart of its demapper circuit with the highest quality value. In the illustrated EXIT chart, the fact that both the highest BER performance and MI quality value are close to 97%-100% when the feedback MI equals 1 means that the demapper circuit method proposed according to the exemplary embodiment of the invention can provide the highest quality LLR. The low SNR EXIT charts 1301 and 1401 of Figures 13(A) to 13(B) and Figures 14(A) to 14(B) also show upper limits for the coding rate of 0.88 and 0.93 (rounded to the nearest second decimal place) for the absence of iterative feedback and the presence of iterative feedback, respectively.

[0206] OS-PSK software demapping and EXIT chart evaluation

[0207] In this section, according to some examples of the invention, a proposed soft-decision method is described that demaps the correlator output of a demodulator from an orthogonal signaling-phase-shift keying (OS-PSK) modulation scheme to an external soft bit in the form of a log-likelihood ratio (LLR), also referred to as a bit LLR or soft bit. The operation of an example of the proposed soft demapper is illustrated in the flowcharts of Figures 36(A) to 36(B).

[0208] The previous section introduced the soft demapping method for OS correlator output only. The operation of some circuits in Figures 15(A) to 15(B) has already been described in the preceding figures and will not be repeated to avoid confusion with the concepts described here. Figures 15(A) to 15(B) introduce the following additional circuits: the PSK symbol-to-LLR circuit 1501 and the symbol-to-symbol LLR circuit 1502. This section also explains how, in the transmission system shown in Figures 15(A) to 15(B), transmitter chain 1503 applies Direct Sequence Spread Spectrum (DSSS) technology (using spread spectrum sequence generator circuitry) and modulator circuit 1504 to convert information bits into OS-PSK modulated data.

[0209] Modulation model

[0210] transmitter chain

[0211] As shown in Figures 15(A) to 15(B), according to the operation performed in [7], the OS-PSK modulation scheme transforms N coded bits 1505 into multiple OS symbols 1506 and PSK symbols 1507. Taking an OS scheme with a modulation parameter (called the OS base) of M=16 (also known as the 16OS scheme) as an example, this corresponds to k=log2M or 4 bits per OS symbol. Similarly, consider that the PSK modulation parameter (called the number of bits per PSK symbol) is Q. m =2, which corresponds to 2 Qm = 4 is the PSK base number (also known as the modulation order). Where Q m A PSK scheme with a value of 2 is also called orthogonal PSK (QPSK). In encoding the message, every k + Q... m = 4 + 2 = 6 bits. In this example, 4 bits are converted to OS notation and 2 bits are converted to PSK notation, both of which have complex values. Assuming N is an integer multiple of 6, the total number of bits is N / (k+Q). m () or N / 6 OS symbols 1506 and the same number of PSK symbols 1507. In transmitter chain 1503, the k bits of the OS mapping are called the first bit set 1508, and the Q bits of the PSK encoding are... m The bit is called the second bit set 1509.

[0212] It should be noted that in this section, according to some examples of the invention, a fixed OS radix of M=16 is used, but it has a variable PSK radix. This approach allows the inventors to explore the effects of changing the PSK radix. However, it is envisioned that in other exemplary embodiments, the soft demapping methods and concepts described herein can be applied to any number of OS bits 'k'.

[0213] In practice, in extreme cases, a skilled practitioner can devise a scheme where the OS cardinality is M=1, and k=0 bits are transmitted by selecting an extension sequence, and all N=Q bits are transmitted by the phase of the extension sequence. m In this scheme, the transmitter will repeatedly transmit the same extended sequence, each time with a transmission Q. m Phase rotation of bits.

[0214] Direct sequence spread spectrum

[0215] To improve the signal-to-noise ratio (SNR) of receiver chain 1510, DSSS technology is used in transmitter chain 1503. Transmitter chain 1503 multiplies OS symbol 1506 by a pseudo-random base sequence of the signal represented by a complex number 1517. The base sequence is called the spreading sequence, and each signal value in the spreading sequence 1511 is called a chip. Compared to coded bits 1505, chips have a shorter duration and therefore a larger bandwidth. The conversion of information bits to chips has the effect of scrambling information bits and widening their spectrum, thus reducing overall signal interference when modulating the signal. Receiver chain 1510 knows the spreading sequence and uses it to despread the received signal.

[0216] modulation

[0217] In the example above, of the 6 bits of encoded bit 1505, k=4 bits 1508 are converted into the OS symbol 1506, and then into the complex-valued extended sequence 1511, while the other Q bits... m Two PSK coded bits 1509 are mapped 1515 to one of the four possible complex PSK symbols 1507 in the orthogonal PSK scheme. The four possible QPSK mapping values ​​have different phases but the same amplitude and are equidistant in the complex plane. m The PSK symbol 1507, consisting of 2 PSK coded bits 1509, is multiplied by each value in the extended sequence 1511 in the modulator circuit 1504. The result is a complex-valued sequence 1518, which is a rotated version of the extended sequence 1511 corresponding to k=4 OS coded bits 1508, where the rotation amount is based on other Q values. m = 2 PSK encoded bits, 1509. The final sequence of this complex number, 1518, contains all k+Q. m =6 encoded bits of 1505 information, and transmitted through the antenna.

[0218] data

[0219] The receiver chain 1510 will demodulate the signal based on its understanding of the applied scheme, such as the PSK modulation order (from Q). mThe values ​​of ) and extended sequence 1511. [6] provides the datasets used in this run to evaluate the example soft demapping methods, which are based on the 16OS-QPSK scheme (that is, M=16, Q m =2), and generated based on an extended sequence with 50 chips. The following section proposes a technique that allows these 16OS-QPSK datasets to be generalized and used to model any PSK cardinality. These datasets belong to the same set of channel and SNR values ​​previously used: AWGN_100dB, AWGN_0dB, AWGN_-3dB, AWGN_-7.5dB, MP_0dB, MP_-3dB, MP_-7.5dB. AWGN and MP terms refer to additive white Gaussian noise and multipath channels, respectively. For example, Figures 31(A) to 31(B) show the AWGN_0db dataset of the 16OS-QPSK scheme [6]: the table shows the first 20 rows out of 8100 rows. Each dataset consists of 8100 rows, where each row includes a 6-bit value 4001 of the message and the corresponding M=16 pairs of in-phase (I) and quadrature-phase (Q) correlator values ​​for 4002.

[0220] amplitude of the correlator IQ pair

[0221] In each IQ pair, the I and Q values ​​represent the real and imaginary parts of the complex correlator output, respectively. For example, in row 1 of Figures 31(A) to 31(B), the first correlator output has a complex value of -0.9-6. i , where i is the imaginary part. The magnitude of the IQ pair is analogous to the probability of the index value of the IQ pair relative to the other pairs in a row of 16 pairs (and thus a cross-correlation value, similar to the output of a correlator), to represent 4 bits of the original transmitted 6-bit message. Therefore, in a manner similar to the example described earlier, it is generally expected that the pair with the largest magnitude among M=16 pairs (under favorable channel conditions) has an index value that is (k+Q) m In a 6-bit message, k=4 least significant (LS) bits are identical. Here, PSK bits are used to provide Q. mThe convention is that the most significant bit (LSB) and the OS bit provide k least significant bits. It is conceivable that other conventions, such as the PSK bit providing k least significant bits, could also be used in the inventive concept described herein. In Figures 31(A) and 31(B), the IQ pairs with the largest amplitude in the rows of IQ pairs are underlined. For example, row 9, 4003 in Figures 31(A) and 31(B), has a value of 7 (“0111”) for its k=4 least significant bits of message, and the index of the largest amplitude IQ pair in that row also has a value of 7. As previously stated, in a row of the dataset, the IQ pair whose index equals the value of the k=4 least significant bits of the message can be called the “correct” correlator output, while the remaining M-1=15 IQ pairs are called the “incorrect” correlator output. More specifically, the “correct” correlator output is provided by a correlator that: for this correlator, 2 k One of the possible signals was selected as the transmission signal.

[0222] Phase of the correlator IQ pair

[0223] When the amplitude of the correlator IQ pair contains (k+Q) m In a 6-bit message, k=4 bits of information, the phase value of the IQ pair represents the remaining Q. m = 2 bits of information. The dataset performs its QPSK modulation on the two most significant (MS) bits of the message. For example, the two most significant bits of the message in row 1, 4004 of Figures 31(A) to 31(B) are '01'. In this example, this means that the message is in the range [0, 3] (in general, it is [0, 2). Qm The phase index with index 1 in [-1] is used for QPSK modulation. According to QPSK coding, the phase index is mapped to a specific set of Gray code phase values. For example, using the phase mapping from [π / 4, 3π / 4, -π / 4, -3π / 4] to [00, 01, 10, 11], the phase index "01" is mapped to the phase value 3π / 4, so the correct correlator output IQ pair in row 1 (Figures 31(A) to 31(B)) is expected to have a phase value of 3π / 4.

[0224] Subsequently, a technique is described that allows these 16OS-QPSK datasets to be generalized and used to simulate any PSK cardinality. For simplicity, in the examples shown, a phase mapping starting with phase 0 will be used, such as [0, -π] for binary PSK (BPSK), or [0, π / 2, -π / 2, -π] for QPSK, and so on.

[0225] Correlator output distribution

[0226] Correlator output amplitude: OS modulation

[0227] Similar to the investigation of data from the OS-only model in the previous section, the distributions of correct and incorrect correlator output amplitudes in the OS-PSK scheme can be aggregated into two distributions, serving as the first and second aggregated correlator output amplitude distributions, respectively. The parameter 1520 used to represent each of the two aggregated distributions can also be referred to as the first distribution parameter set. As previously mentioned, the Rice distribution can be identified as the single distribution that best fits each of the sets of correct and incorrect correlator output amplitudes. As previously mentioned, for the AWGN and MP channel datasets, the distributions of correct and incorrect correlator output amplitudes are shown in Figures 21(A) through 21(C) and Figures 23(A) through 23(C), and... Figure 16 An example for the MP-7.5dB dataset is provided. A comparison of these distributions with those in Figures 5(A) through 5(C) and Figures 6(A) through 6(C) previously discussed for the OS-only scheme shows that the data distributions are very similar between the OS-PSK (this section) and OS-only (previous sections) models. This is demonstrated by the non-centrality parameter(s) and scaling parameter(s) of the Rice distribution. This will be used to prove that for all SNR values ​​and the first set of distribution parameters between the two models, they have the same... Figure 3 The parameter 303 has a similar value.

[0228] In summary, as previously described, a method for estimating the parameters of the output amplitude distributions of the first and second aggregation correlators has been described, in which at least one of the following is used:

[0229] (i) Estimate the first set of distribution parameters for the first aggregate correlator output amplitude distribution by fitting a first probability distribution to the amplitude of the correlator output obtained by correlating the received synchronization signal with a known synchronization signal (as in step 4507 of Figures 36(A) to 36(B)); (ii) Estimate the first set of distribution parameters for the second aggregate correlator output amplitude distribution by fitting a second probability distribution to the amplitude of the correlator output obtained by correlating the received synchronization signal with a signal different from the known synchronization signal (as in step 4507 of Figures 36(A) to 36(B)).

[0230] (iii) By fitting the third probability distribution to the set of 2 k The two correlators obtained 2 k The amplitude of the correlator output with the largest amplitude in the set of correlator outputs is used to estimate the first set of distribution parameters of the first aggregate correlator output amplitude distribution (as in step 4508 of Figures 36(A) to 36(B)).

[0231] (iv) By fitting the fourth probability distribution to the set of 2k The two correlators obtained 2 k The set of correlator outputs does not contain the maximum amplitude of 2. k -1 The amplitude of the correlator output is used to estimate the first set of distribution parameters of the second aggregate correlator output amplitude distribution (as in step 4508 of Figures 36(A) to 36(B));

[0232] (v) Select, for example, a first set of distribution parameters for the first aggregate correlator output amplitude distribution and the second aggregate correlator output amplitude distribution from lookup table 1527 (as in step 4505 of Figures 36(A) to 36(B)).

[0233] In other examples, it can be demonstrated that the Rice distribution can be used to represent the first probability distribution, the second probability distribution, the third probability distribution, or the fourth probability distribution.

[0234] Correlator output phase: PSK modulation

[0235] Phase error value

[0236] For AWGN 2200 and Multipath 2400 channels, and for sets of correct and incorrect IQ pairs, the aggregated distribution of the correlator output phase values ​​is shown in Figures 22(A) to 22(C) and Figures 24(A) to 24(C). The data in these figures are represented by phase error: for each row in the dataset (as shown in Figures 31(A) to 31(B)), the “true” phase of the transmitted message (determined by the Q of the message) is subtracted from the phase of each correlator IQ pair. m =Defined by the 2 most significant bits) as the "received" phase, thus generating the received phase error. Next, in some examples, the phase error of the "correct" correlator output is extracted from each row of the dataset to obtain the set of "correct" phase errors, while the remaining M-1=15 phase errors from each row are collected into the set of "incorrect" phase errors. Figure 17 Histograms and fitted distributions for the correct phase error 1701 and incorrect phase error 1702 are provided for the MP_-7.5dB dataset.

[0237] Fit the distribution to the phase error

[0238] The parameters of the aggregated correct correlator output phase error distribution can be referred to as the second set of distribution parameters. The correct phase errors 1701 and 1703 are best-fitted to a normal distribution 1704, which is constrained to have zero values. This constraint is achieved by fitting the absolute values ​​of the correct phase errors to a half-normal distribution, such that the distribution is determined by only a single extended parameter. 相位1705 is used to characterize this. This is shown in Figures 15(A) and 15(B), where the correct phase error distribution is... 相位 Parameter 1723 is sent to the PSK symbol LLR circuit 1501. As shown in Figures 22(A) through 22(C) and Figures 24(A) through 24(C), the incorrect phase errors 1702 all have a uniform distribution. This means that the incorrect phase does not contain any information and does not need to be characterized by any distribution parameters.

[0239] Based on the above, only the correct phase error 1703 will convey information from the correlated data, not the incorrect phase error 1702. Therefore, the distribution of all correct correlator output phase error values ​​can be called the aggregate correlator output phase distribution. In this name, the term "correct" is explicitly excluded, even though the distribution is constructed from the set of correct phase error values. This maintains the general definition of the aggregate correlator output phase distribution. As mentioned before, the correct phase error is provided by the correlator, for which 2 k One of the possible signals is selected as the transmitted signal.

[0240] In summary, a method has been proposed in which the output phase distribution of multiple aggregation correlators is 1, and when 2 k When one of the possible signals is selected as the transmitted signal, the phase distribution of the aggregate correlator output is approximately 2. k The output phase of each correlator is 2 k Aggregation of distributions. A method is also proposed where the phase distribution of the aggregate correlator output is represented by a second set of distribution parameters, and where the second set of distribution parameters includes extended parameters. 相位 .

[0241] Estimation methods

[0242] The above analysis assumes a test platform environment with prior knowledge of the true values ​​of transmitted bit 1508 and PSK encoded bit 1509. However, in a real receiver, such knowledge is unavailable, and the distribution parameters must be estimated in the absence of this knowledge. 相位1523. To obtain the distribution of the correlator output phase, online (step 4524 in Figures 36(A) to 36(B)) or offline (step 4523 in Figures 36(A) to 36(B)) methods are proposed. In an example of the proposed online method, the data analyzer circuit 1524 is responsible for analyzing the correlator output 1525 received during runtime from the demodulator 1519, finding the optimal distribution to fit the data, and calculating the parameters 1526 of the distribution, which benefit from real-time tuning of the distribution parameters. This is shown in Figures 15(A) to 15(B) with the data analyzer circuit 1524 and is detailed below. However, some offline methods are discussed first, which use a library lookup table 1527 of the correlator distribution shown in Figures 15(A) to 15(B).

[0243] Offline estimation methods

[0244] As previously described for the OS-only scheme, in the offline method, correlator distributions can be computed offline to establish a library lookup table 1527 for the distributions of all applicable modulation channel combinations. For example, the previously mentioned dataset can be used as a model for different channel conditions and modulation schemes. As shown in Figures 15(A) and 15(B), the configuration time switch 1528 can be configured to select (step 4505 of Figures 36(A) and 36(B)) the correct dataset based on the system's modulation scheme and channel type. This can be performed as an extension of any other channel estimation task performed by the demodulator 1519. For example, the outputs of these channel estimation tasks can be used to index (step 4505 of Figures 36(A) and 36(B)) the lookup table 1527 for the pre-computed correlator output distributions. More specifically, offline estimation of the correlator distributions can be used to record the correlator outputs for several channel conditions and to establish lookup tables for these distributions. Then, during actual data transmission, the channel estimation task identifies (step 4505 of Figures 36(A) to 36(B)) the current state of the channel and selects the corresponding distribution from lookup table 1527, as shown in Figures 36(A) to 36(B).

[0245] Another offline approach option is to use a single set of correlator distributions in all cases, regardless of varying channel conditions. This single set of correlator distributions can be recorded under worst-case conditions, perhaps at the lowest SNR where reliable synchronization is achievable. When channel conditions match this worst-case scenario, using the appropriate correlator distribution will ensure the best possible performance. When channel conditions are better than this worst-case scenario, even if the assumed correlator distribution is pessimistic compared to the true distribution, the chance of successful decoding can be expected to increase.

[0246] Online estimation methods

[0247] In addition to offline distribution estimation using a dataset, two online methods 4507 and 4508 are proposed, which compute distribution parameters in the early stages of real data transmission, such as when a channel estimation task is underway. The first online method, step 4507 in Figures 36(A) to 36(B), involves associating the received synchronization signal with a known synchronization signal. Using this first online method of step 4507, since the receiver chain 1510 knows the true symbol phase, it can associate each set of M=16 correlator outputs 1525 transmitted with its correct phase. This first online method of step 4507 can be referred to as correlator distribution estimation using the synchronization sequence. After distribution aggregation in the data analyzer circuit 1524, the parameters of the aggregated distribution can be estimated (as shown in step 4507 of Figures 36(A) to 36(B)) using a distribution fitting method (like Matlab's histfit function). 相位 1523. Correlator distribution estimation using synchronization sequences has the advantage of online operation, but it also has its own drawbacks. One drawback is that synchronization sequences are typically short unless embedded in time-consuming channel estimations. Therefore, they do not always provide enough samples to obtain an accurate distribution, which may degrade the quality of the symbol LLR 1529 calculated using those distribution parameters. To achieve sufficient accuracy, longer sequences are required, which increases the time overhead of channel estimation phase and also increases power and bandwidth consumption.

[0248] Another method for estimating the distribution parameters is the online method of step 4508 in Figures 36(A) to 36(B), which is the method previously described for the OS-only scheme, considering the maximum amplitude correlator output. The online method of step 4508 divides the correlator outputs 1525 into two groups: a) maximum amplitude correlator outputs (each maximum amplitude will be in M ​​outputs of the same received message), and b) non-maximum amplitude correlator outputs (M-1 outputs have non-maximum amplitudes of the same received message and are used for several transmissions / messages). In the online method of step 4508, it is assumed that any maximum amplitude correlator output in the M correlator outputs 1525 corresponds to its transmission sequence 1518 from transmitter chain 1503. Based on this assumption, the phase of the maximum amplitude correlator output should also correspond to the phase value of the transmission sequence 1518, so each phase should represent Q in the message. m1509 PSK encoded bits. Therefore, similar to the offline estimation method in step 4523, where the incorrect phase error value 1702 of the incorrect correlator output follows a uniform distribution, the phase error value of the non-maximum amplitude correlator output is also expected to have a uniform distribution. Therefore, the valuable distribution in the maximum amplitude of step 4508 will be the phase error value of the maximum amplitude correlator output. As previously described, these values ​​form the aggregated correlator output phase distribution, but this time it is for the maximum amplitude of step 4508. The calculation of the phase error in the maximum amplitude of step 4508 is the same as described above, except that the "true" phase value (known in other methods) is assumed to be the phase in the phase map that is closest to the phase of the maximum amplitude correlator output. More precisely, the phase error can be estimated as the difference between the phase of the maximum amplitude correlator output and the phase map closest to it.

[0249] It should be noted that the maximum amplitude correlator distribution estimation method in step 4508 can be further refined through the following steps: completing the parameters 相位 The first estimate of 1523 is used to calculate the LLR before it is provided to the channel decoder 912, as described later regarding the soft demapper circuitry. The channel decoder 912 can then attempt to remove any errors in the LLR sequence and use the Cyclic Redundancy Check (CRC) decoder 911 to determine if it has succeeded. If unsuccessful, the channel decoder can provide feedback LLRs to the data analyzer circuitry 1524 of Figures 15(A) through 15(B). These LLRs can then be considered, and they can cause the classification of correlator outputs between the maximum amplitude group and the non-maximum amplitude group to be overridden. More specifically, when the maximum amplitude group is formed, if the feedback LLR provides a sufficiently strong indication that the amplitude does not reflect the correct transmission, the correlator output with the maximum amplitude for a particular transmission can be swapped for another correlator output without the maximum amplitude.

[0250] The definition of the output phase distribution of the aggregate correlator is now provided, which applies to all estimation methods described above. When one of the M possible signals is selected as the transmitted signal, the output phase distribution of the aggregate correlator approximates the aggregation of the M distributions of the output phases of the M correlators. The output phase distribution of the aggregate correlator is represented by a second set of distribution parameters.

[0251] Summary

[0252] In summary, methods have been proposed for estimating the parameters of the phase distribution of the aggregate correlator output, which utilize at least one of the following:

[0253] The second set of distribution parameters of the aggregated correlator output phase distribution is estimated by fitting the fifth probability distribution to the phase error of the correlator output obtained by correlating the received synchronization signal with a known synchronization signal (step 4507 in Figures 36(A) to 36(B)).

[0254] By fitting the sixth probability distribution to the set of 2 k The two correlators obtained 2 k The phase error of the correlator output with the largest amplitude in the set of correlator outputs is used to estimate the second set of distribution parameters of the aggregate correlator output phase distribution (as shown in step 4508 of Figures 36(A) to 36(B)).

[0255] Select the second set of distribution parameters for the output phase distribution of the aggregate correlator from lookup table 1527 (step 4505 in Figures 36(A) to 36(B)).

[0256] It was also proven that the fifth or sixth probability distribution can be represented by a Gaussian distribution.

[0257] In the example embodiment below, the soft demapping circuit 1522 uses the phase 1530 of M=16 correlator IQ pairs to obtain the log-likelihood ratio (LLR) of possible phase values.

[0258] Software demapper

[0259] Similar to Figure 9 and Figure 10 The previously proposed soft demapper circuit 907, for an OS-only scheme, requires an LLR value for each message bit to be fed to the soft decision decoder 912. This means that for (k+Q) m A 10-bit message needs to be generated by the soft demapping circuit 1522 (k+Q). m Therefore, for a total of N encoded bits 1505 transmitted, there will be N external bits LLR 904 output by the soft demapping circuit 1522, as shown in Figures 15(A) to 15(B). Since the information of k=4 OS bits (from (k+Q) is transmitted through the correlator output amplitude 1531 (as previously described), mThe amplitudes of these bits (bit messages) are fed as a first set of inputs to the soft demapper circuit. The conversion to OS LLR circuit 1513 in Figures 15(A) and 15(B) will provide (as shown in step 4510 of Figures 36(A) and 36(B)) M=16 OS symbols LLR 908 as a first set of a priori soft signals corresponding to M=16 correlator outputs 1525. Furthermore, as in the previously discussed OS-only scheme, the OS a priori feedback bit LLR set 910 is fed from the decoder 912 to the soft demapper circuit 1522 as a second set of inputs to the soft demapper circuit 1522, where those feedback bit LLR 910 can be referred to as the second set of a priori soft signals.

[0260] In this example embodiment, a soft demapping circuit 1522 is described, which can use any number of PSK bits Q per symbol. m This operates under the general conditions of M-OS-PSK. However, in the following discussion, Q will be used frequently. m =2 QPSK as a specific example, in this case (k+Q m =6.

[0261] Convert to bit LLR

[0262] The generation of the external bit LLR 904 is performed using a bit-to-bit LLR circuit 1521 (Figures 15(A) to 15(B)). In an example embodiment without feedback from the decoder, as previously described, a bit-to-bit LLR circuit 906 is introduced, which converts the M=16 OS symbols LLR 908 of the 4-bit message into a k=4-bit LLR. The bit-to-bit LLR circuit 1521 in the current section needs to generate k+Q for a 6-bit message. m =6 external bits LLR 904, including LLR corresponding to OS modulation and LLR corresponding to PSK scheme, as previously described for the modulation model. The conversion to bit LLR circuit 1521 in this section uses the same method as the similar circuits discussed previously for OS-only schemes, but this time it has information about bits representing the phase of the demodulator correlator output 1525 as input. This new information is provided by the PSK symbol LLR 1532 by the conversion to PSK symbol LLR circuit 1501, as shown in Figures 15(A) to 15(B), and in some examples, it can be referred to as the third a priori soft signal set.

[0263] The symbol-to-bit LLR circuit 1533 of Figures 15(A) to 15(B) is described below. It is responsible for generating the a posteriori bit LLR 909. The symbol-to-bit LLR circuit 1533 takes input from the symbol-to-symbol LLR circuit 1502 of Figures 15(A) to 15(B), which will be further described below and is responsible for combining the LLR information from three sources (i.e., LLRs 908, 1532, and 1534). These three sources are the bit-to-symbol LLR circuit 1535 (described below, which handles the feedback a posteriori bit LLR 910), the "convert to OS LLR" circuit 1513 (which handles the amplitude 1531 of the correlator output), and the PSK-to-symbol LLR circuit 1501 (which handles the phase 1530 of the correlator output).

[0264] Symbol-bit LLR circuit

[0265] Similar to the OS-only scheme, it can be assumed that each message combination in the OS-PSK scheme has a corresponding symbol LLR. For a 6-bit message, such as in the 16OS-QPSK scheme, this results in a total of 2 LLRs for a single message. 6 = 64-symbol LLR: LLR 63 sym , LLR 62 sym , …, LLR 0 sym These values ​​are shown by annotated lines in Figures 15(A) and 15(B), which represent the total number of symbols LLR 1529 with N-bit coded bits 1505 as N / (k+Q). m ) x M x 2 Qm or N / 6 x 16 x 2 2 The symbol LLR 1529 is converted to bit LLR 909 using the symbol-to-bit LLR circuit 1533 in Figures 15(A) to 15(B) (step 4513 in Figures 36(A) to 36(B)). Figures 32(A) to 32(B) are extensions of Figures 31(A) to 31(B) and are applicable to 4-bit messages. Figures 32(A) to 32(B) illustrate the symbol-to-bit LLR conversion for 6-bit messages.

[0266] In Figures 32(A) to 32(B), the LLR term L5 bit,0 L5 bit,1 L4 bit,0 L4 bit,1 , … , L0 bit,0 L0 bit,1The value of each item in the column is defined by the item above it in the same column. For example, in column 4... L 4 bit,0 Defined as L 4 bit ,0 LLR 0 sym LLR 1 sym … LLR 15 sym LLR 32 sym LLR 33 sym … LLR 47 sym .symbol The maxstar operation, defined in the first part of the OS-only scheme, provides a detailed explanation of how the bit LLR is calculated from the symbol LLR. The bottom row 3804 in Figures 32(A) and 32(B) defines the value of the a posteriori output bit LLR, which includes two sets of soft bits in this OS-PSK scheme: OS bit LLR (MS Q). m LLR: LLR 5 bit,out LLR 4 bit,out ) and PSK bit LLR (LS k LLR: LLR 3 bit,out down to LLR 0 bit,out The output OS bit LLR can be referred to as the first a posteriori soft bit set including k soft bits, and the output PSK bit LLR can be referred to as the set including Q... m The second posterior soft bit set of 1 soft bit.

[0267] As shown in Figures 15(A) and 15(B), to obtain the extrinsic bit LLR 904 (step 4514 in Figures 36(A) and 36(B)), the feedback prior bit LLR 910 is subtracted from the posterior bit LLR 909. This relationship between the bits corresponding to the mutually separate OS and PSK schemes can be described as follows: The soft bits of the first prior soft bit set are combined with the soft bits of the second prior soft signal set to obtain a first extrinsic soft bit set comprising k soft bits. The soft bits of the second prior soft bit set are combined with the soft bits of the fourth prior soft signal set to obtain a set comprising Q...m The second set of external soft bits of each soft bit.

[0268] Symbol-to-symbol LLR circuit

[0269] The current problem is how to use the LLR from upstream 908, 1532 and downstream feedback bit LLR 910 to calculate symbol LLR 1529 (step 4528 in Figures 36(A) to 36(B)). In the OS-only scheme, OS symbol LLR 908 is a function of correlator output amplitude 109 (as the first input set of the soft demapper circuit 1522) and feedback OS a priori feedback bit LLR 910 (as the second input set of the soft demapper circuit). Here, symbol LLR is also a function of correlator output phase 1530 as the third input set of the soft demapper circuit 1522 and feedback PSK a priori bit LLR as the fourth input set. In some examples, feedback PSK bit LLR can be equivalently referred to as the fourth a priori soft signal set.

[0270] The symbol-to-symbol LLR circuit 1502 takes the OS symbol LLR 908, PSK symbol LLR 1532, and symbol LLR 1534, which are converted from the downstream bit LLR (step 4527 in Figures 36(A) to 36(B)), as input to generate symbol LLR 1529 (in step 4528 in Figures 36(A) to 36(B)). The subsequent symbol LLR 1534 can be referred to as the feedback symbol LLR (as shown in Figures 15(A) to 15(B)), which includes the OS and PSK prior symbol LLRs as the second and fourth prior soft signal sets, respectively. Each of the M=16 OS symbol LLRs corresponds to one of the M=16 combinations of k=4 LS bits of a 6-bit message, and 6 bits can represent 64 combinations. In some examples, it is assumed that the Q of the message... m The two most significant bits themselves can also have four combinations ("00", "01", "10", "11"). Figure 28 According to Figure 33 The extension is used for this part of the operation. Figure 33 The 6-bit message combination 4201 and the corresponding LLR parameters 3703, 4202, and 4203 are shown.

[0271] exist Figure 33 In this context, the OS symbol LLR 3703 has the same value as the symbol LLR 3703 in the previously discussed OS-only scheme: for all i in the range [0, M-1] or [0, 15], LLR i OS,in OS-PSK = LLR isym,in 仅OS This is because the operation discussed above is based on an OS-only scheme, where the cross-correlation data is only the amplitude, not the phase. For example... Figure 33 As shown, after every M=16 combinations, the OS symbol LLR value 3703 is repeated for M=16 times. Furthermore, it is assumed that each message combination (from 2...) k+Qm =64) has its own PSK symbol LLR, which will explain why this assumption holds and how the PSK symbol LLR4202 is calculated in step 4525. Each message combination will also have a corresponding feedback symbol LLR 4203, such as Figure 33 As shown in the rightmost column. The symbolic LLR value for each message combination will be the sum of all LLR terms in the corresponding row of Figures 31(A) to 31(B). For example, LLR 31 sym = LLR 15 OS,in + LLR 31 PSK,in + LLR 31 sym,bit In other words, all prior soft signals 908, 1532, and 1534 are combined in the symbol-to-symbol LLR circuit 1502 to generate 2 in step 4528. k+Qm A set of posterior soft signals or symbols LLR 1529, which are combined 1533 in step 4514 to obtain a first set of posterior soft bits LLR 909 and a second set of posterior soft bits.

[0272] Bit-symbol LLR circuit

[0273] The feedback bit LLR 910 in Figures 15(A) and 15(B) may contain the LLR value fed back from the decoder 912 to the soft demapping circuit 1522. If the decoder 912 does not feed back any LLR, the feedback bit LLR 910 will have a bit LLR value of zero (as in step 4512 of Figures 36(A) and 36(B)), which has no effect on the calculation of symbol LLR 1529 (as in steps 4527 and 4528). Otherwise, in the iterative decoding operation of step 4521, the feedback bit LLR value affects (as in steps 4519 and 4527) the calculation of step 4528 within the bit LLR circuit 1521. Figure 34 The transformation from the a priori bit LLR to the symbolic LLR field is shown, which is determined by the feedback symbolic LLR parameter. LLR 63 sym,bit , LLR 62 sym,bit, …, LLR 0 sym,bit 4203 ( Figure 33 The rightmost column represents this. An explanation of an example method for calculating the feedback symbol LLR 4301 in step 4527 has already been described.

[0274] It should be noted that in some examples, () can be used. LLR 5 bit,in + LLR 4 bit,in + LLR 3 bit,in + LLR 2 bit,in + LLR 1 bit,in + LLR 0 bit,in Add the constant value of ) / 2 to Figure 34 All 64 lines are included without affecting the extrinsic LLR ultimately generated by the entire soft demapping circuit 1522. The advantage of adding this constant is that, after some simplification, it eliminates... Figure 34 The calculation includes all divisions by 2 and reduces the number of additions.

[0275] Convert to PSK notation LLR

[0276] exist Figure 33 In the middle, assuming 2 k+Qm Each of the message combinations 4201 has a corresponding PSK LLR. The PSK LLR is generated by the PSK symbol conversion LLR circuit 1501 shown in Figures 15(A) to 15(B) (step 4525 in Figures 36(A) to 36(B)). For a 16OS-QPSK scheme, this conversion uses (step 4509 in Figures 36(A) to 36(B)) M = 16 correlator outputs 1525 as (k+Q) m ) 6-bit message input, and generate 2 k+Qm =64 PSK LLR 1532 as output. As can be seen from the previous section, the M=16 complex correlator outputting a phase of 1525 means that the Q value contains the 6 bits of the transmitted message. m= 2 most significant bits of information. For example, consider the case where the truth bit of the 6-bit message x is -π / 2. Now, it can be observed that under normal channel conditions, one correlator output (most likely the correct IQ pair) has a phase of -π / 2, and the phase values ​​of the remaining M-1=15 correlator outputs (most likely the incorrect IQ pairs) will be uniformly distributed from the applicable range of [-π, π]. Therefore, if the true phase of -π / 2 is subtracted from all M=16 correlator output phases, at least one phase error is expected to be close to zero, and the remaining phases are expected to have random values ​​within the aforementioned applicable range.

[0277] It should be noted that while the QPSK phases [π / 4, 3π / 3, -π / 4, -3π / 4] were used in the datasets previously described for OS-only schemes, an example technique will be introduced that allows generalization of these 16OS-QPSK datasets and can be used to model any PSK cardinality in the previous testbed discussion. For simplicity, in this example embodiment, a phase mapping starting with phase 0 will be used, such as [0, π / 2, -π / 2, -π] for QPSK.

[0278] On the other hand, the distributions of correct and incorrect phase error values ​​can be considered. The phase error calculated above (the result of subtracting the "hypothetical" true phase from the M=16 received phases) can be applied to the probability density function (PDF) of both distributions to obtain the probability that each correlator output is a) a correct correlator output or b) an incorrect correlator output. Using these conditional probabilities, the probability ratio (i.e., PSK LLR) for each of the M=16 correlator outputs can be calculated according to the following formula:

[0279]

[0280] Applying Bayes' theorem, we give

[0281]

[0282] Here, the conditional probability is characterized by the normal distribution and the uniform distribution, and we can assume that Pr( correct ) = 1 / M and Pr (not correct ) = (M-1) / M. Performing these substitutions, we get...

[0283]

[0284] Here, 0.5 / π is a uniformly distributed value within the range of -π to +π. It should be noted that the "phase error" here refers to the phase of the corresponding correlator output and 2π / π. k+Qm The difference between the phases of the corresponding messages in a message combination.

[0285] In the example above, the true phase is assumed to be -π / 2; that is, the two most significant bits of the message are assumed to be fixed values. However, this assumption is made without knowing the true phase, therefore, the above method must be applied to all possibilities of PSK phases (i.e., the four phases in the QPSK scheme). Therefore, the steps for converting 1501 (step 4525 in Figures 36(A) to 36(B)) to PSK symbol LLR 1532 can be summarized as follows: assume a possible PSK phase value (e.g., for QPSK, a value in [0, π / 2, -π / 2, -π]), subtract it from the M correlator phases 1530, apply the resulting phase error value to the correct and incorrect PDFs, calculate the PSK LLR for each correlator value from the PDF output, and repeat these steps starting from the new PSK phase value. In the 16OS-QPSK scheme, the process of step 4525 produces M.2 Qm Or 16 × 4 = 64 PSK symbols LLR 1532, which correspond to Figure 33 The middle is displayed as LLR 63 PSK,in , LLR 62 PSK,in , …, LLR 0 PSK,in The input PSK LLR parameter is 4202. In summary, it can be said that the output phases 1530 of the M correlators (1525) and the second distribution parameter set of the aggregate correlator output phase distribution (...) 相位 ) 1523 combined, so as to obtain in step 4525 including 2 k+Qm The third prior soft signal set of a soft phase (PSK symbol LLR 1532).

[0286] Summary

[0287] In summary, a soft demapping circuit 1522 has been proposed for performing soft decision demodulation in the receiver chain 1510 of a transmission system 1500 comprising a transmitter chain 1503, a channel 1536, and a receiver chain 1510. The transmitter chain 1503 transmits data from 2 k bits based on a k-bit value... k The extended sequence 1511 selected from the set of possible signals is used to signal the first bit set 1508, which includes k bits, and the transmitter uses the Q signal in the modulator circuit 1504. m The value of the bit is from 2 Qm The rotation is selected from a set of possible rotations to rotate the phase of the extended sequence 1511 to signal the Q signal. m The second set of bits, 1509, is used by receiver chain 1510. kA correlator is used to detect the transmission of each possible signal, and the output of the correlator is 1525. k Amplitude 1531 is provided as the first input set to the soft demapping circuit 1522, and a second a priori soft signal set or feedback bit LLR 910, including k soft bits, is provided as the second input set to the soft demapping circuit 1522. Furthermore, the correlator output 1525 is... k Phase 1530 is provided as the third input set to the soft demapping circuit 1522, and will include Q m The fourth prior soft signal set or feedback bit LLR 910 of k soft bits is provided as the fourth input set to the soft demapping circuit 1522. The soft demapping circuit 1522 calculates a first posterior soft bit set comprising k soft bits based on statistics and parameters 1520 derived from the amplitude distributions of the outputs of multiple aggregate correlators, wherein the soft demapping circuit 1522 is based on statistics (parameters) derived from the phase distributions of the outputs of multiple aggregate correlators. 相位 1523) to calculate including Q m The second posterior soft bit set of 1 soft bit. Furthermore, the soft demapping circuit 1522 combines (step 4528 of Figures 36(A) to 36(B)) all prior soft signals 908, 1532, and 1534 to generate 2 soft bits. k+Qm A set of 1529 posterior soft signals or symbols LLRs, and combinations thereof 2 k+Qm A posterior soft signal or symbol LLR 1529 (as shown in step 4513 of Figures 36(A) to 36(B)) is used to obtain a first posterior soft bit LLR set 909 and a second posterior soft bit set. Subsequently, the first prior soft bit set is combined with the soft bits of the second prior soft signal set to obtain a first extrinsic soft bit set. Finally, the second prior soft bit set is combined with the soft bits of a fourth prior soft signal set to obtain a set including Q. m The second set of external soft bits of each soft bit.

[0288] As shown in the flowcharts of Figures 36(A) to 36(B), the operation 4500 of the example soft demapper for the OS-PSK scheme can be summarized as follows. Initially at 4501, the demapper receives a set of signals 4502 as the output of the demodulator, which corresponds to the transmitted frame. Then, it is decided which of three different methods to use for estimating the amplitude and phase distribution of the correlator output. If the offline method, such as step 4523, is selected, then in step 4505, characteristics of the channel, such as signal-to-noise ratio and channel type (e.g., AWGN), are identified. The results of the identification are then used to address a pre-computed lookup table 1527 for the parameters of the correlator output amplitude and phase distribution. Conversely, if the online method of distribution estimation in step 4524 is selected, the choice will be to use the synchronization sequence (as shown in step 4507 of Figures 36(A) to 36(B)) or apply the maximum amplitude in step 4508. Regardless of the estimation method used, the estimated parameters of the amplitude distribution and the correlator output amplitude collected in step 4509 of Figures 36(A) to 36(B) are used to calculate (step 4510 of Figures 36(A) to 36(B)) the OS soft signal, which, as previously described, can be in the form of an LLR. Similarly, the calculation of the PSK soft signal in step 4525 is performed using the estimated parameters of the correlator output phase value and the phase error distribution, as described in the example embodiment for conversion to PSK notation LLR.

[0289] Subsequently, the calculated OS (step 4510 in Figures 36(A) to 36(B)) and PSK (step 4525) soft signals are stored (as in steps 4511 and 4526) in internal memory to prevent demapping-decoding rounds that require iterative decoding later. Up to this point, because only the first iteration has begun in the possible iterative decoding process, there are no values ​​on the feedback path from the decoder. Therefore, the feedback will be entirely zero, as in step 4512 of Figures 36(A) to 36(B). Although currently zero, the feedback path is in the LLR domain (soft bits), and is therefore converted to the symbol domain in step 4527 to provide the feedback a priori soft signal (as previously described regarding the bit-to-symbol LLR circuitry of the soft demapper). The OS and PSK soft signals previously calculated in steps 4510 and 4525 are also of a priori type and are combined with the feedback a priori soft signal, which is used to calculate the OS-PSK soft signal in step 4528 (as previously described regarding the symbol-to-symbol LLR circuitry of the soft demapper). These signals are prior information because they are functions of the feedback path, although they are zero values ​​so far.

[0290] Next, in step 4513, the posterior OS-PSK symbols are converted to the LLR domain to give the posterior soft bits (as described above regarding the symbol-bit LLR circuit). Then, in step 4514, the posterior soft bits are converted to extrinsic soft bits of the type required by the decoder, and sent to the decoder in step 4515. After the decoder completes the decoding of its received extrinsic information, in step 4516, the demapper receives the CRC status and the decoded soft bits from the decoder. After checking the CRC status in step 4517, if there is a failure in step 4521, Turbo decoding is applied, and the decoded soft bits received in step 4516 are treated as prior information fed back from the downstream decoder in step 4519. Furthermore, in steps 4518 and 4529, the OS and PSK soft signals previously calculated in steps 4510 and 4525 are loaded from memory as another set of prior information for the second demapping iteration. In step 4527, the feedback prior soft bits are converted to the soft signal domain, and the soft demapper uses all prior information to perform a second calculation of the prior soft signal (step 4528 in Figures 36(A) to 36(B)). In step 4522, the Turbo decoding demapping-decoding iteration ends by either passing the CRC check or abandoning further decoding of the current frame, which may result in, for example, HARQ retransmission. At this point, the demapping of the current frame ends in step 4520.

[0291] Test platform

[0292] Here, the test platform evaluates the soft demapping circuit 1522 used for the OS-PSK scheme, and the test platform is shown in the block diagrams of Figures 18(A) and 18(B). Figures 18(A) and 18(B) show the 16OS-PSK test platform using the 16OS-QPSK dataset. For a given dataset, the test platform 1800 generates three external information transfer (EXIT) diagrams 1801 for different parts of the message bit / LLR: namely, the OS bit / LLR of OS stream 1802, the PSK bit / LLR of PSK stream 1803, and the combined OS-PSK bit / LLR of combined stream 1804 (i.e., the entire encoded / decoded message). In Figures 18(A) to 18(B), sets 113, 114, 904, 909, 1001, 1501, 1506, 1507, 1508, 1509, 1511, 1515, 1517, 1521, 1522, 1524, 1526, 1530, 1531, 1532, 1808, 1811, 1814, and 1815 represent a portion of the test platform 1800 existing in the real transmission system 1500, and the remaining portion is a test platform-specific flow 1806. The test platform 1800 in Figures 18(A) to 18(B) has the same characteristics as... Figure 10 The test platform shown in the image operates the same way as 1000.

[0293] Since the dataset used in the currently running test platform 1800 is based on the OS-PSK scheme, which differs from the previously discussed OS-only model and dataset 1012, the functionality of dataset 1807 and analog channel circuit 1900 in Figures 18(A) to 18(B) differs from the corresponding circuits 1012 and 1021 in the OS-only scheme. The example OS-PSK dataset 1807 has been previously described and introduced, and the analog channel circuit 1900 is described below. The portions listed above in Figures 18(A) to 18(B) share the same circuitry as the block diagrams in Figures 15(A) to 15(B), except that a hard demapper circuit 1808 is added to this test platform 1800 (as it is used in the OS-only scheme), allowing the bit error rate (BER) 1809 of the hard-decision demapping process 1812 to be compared with the BER 1810 of the soft-decision demapping process 1813.

[0294] Channel simulation

[0295] In example test platform 1800, a given channel is simulated in analog channel circuit 1900 using its corresponding dataset, which is derived from the previously presented dataset 1807, all of which belong to the 16OS-QPSK scheme. In some examples, test platform 1800 can be configured to simulate other PSK schemes; such as BPSK, 8PSK, or 16PSK. For this, it is necessary to remove the dependency on the QPSK scheme from dataset 1807. This can be done using the method described for pure OS schemes: for each pair of “1 16OS-QPSK symbol - 16 correlator IQ pairs”, the true phase (as the received QPSK phase) embedded in the QPSK portion of the 16OS-QPSK symbol value is subtracted from the phase of the complex value of each correlator. This is done in... Figure 19 The figure illustrates a 16OS-nPSK channel simulation using a 16OS-QPSK dataset. To find the true phase value, the 16OS-QPSK symbol value is divided by 1901 M = 16 (1902 rounded to zero), and the result is used to select the phase in the QPSK phase mapping circuit 1903. The test platform 1800 is universal for OS modulation orders and can have M=2. k Any value of .

[0296] Assume the 16OS-QPSK symbol value 4001 is in the range [0, 2]. k+Qm Within [-1] or [0, 63] (Figures 31(A) to 31(B)), dividing by 1901 M=16 (and then rounding to 1902) will give the range [0, 2]. QmThe value within [-1] or [0, 3] is equal to the binary value of Q in the 16OS-QPSK symbol. m = 2 most significant bits. Phase subtraction is performed by dividing the complex number 1904, which has the received phase, by a complex number with an amplitude of 11905 and the true phase value, as follows: Figure 19 The "÷" operator is shown in the diagram. Through this calculation, the 16OS-QPSK dataset is transformed into a new dataset 1906: 16OS-PSK data with 16OS signs and PSK phase errors. This transformation was performed offline in 1909, as shown in the diagram. Figure 19 As shown by the dashed lines in the image. Figure 19 The solid line in the diagram illustrates the process during the simulation of 1910.

[0297] Typically, the new 16OS-PSK data 1906 can be used to simulate 16OS-PSK schemes with any PSK modulation; such as BSPK, 8PSK, or nPSK (n=2Qm). From Figure 19 The input to the circuit, k=4 bits, defines the 16OS symbol value, and a row with the same input 16OS symbol value is randomly selected from the 16OS-PSK data 1906. The corresponding row of data contains M=16 complex numbers IQ pairs multiplied by a complex number (Q) with an amplitude of 1 (defined by another input to the analog channel circuit 1900) of 1907. m bits (where n=2) Qm The multiplication 1907 reverses the effect of the previously performed offline division 1904 to calculate the phase error: multiplication 1907 adds the phase from the nPSK scheme to the phase error. The result is M=16 correlator IQ pairs becoming the output 1811 of the analog channel circuit 1900.

[0298] It should be noted that, based on the dataset described earlier in this section, in Figure 19 The QPSK phase-mapped circuit in the 1903 always uses phase mapping [π / 4, 3π / 4, -π / 4, -3π / 4]. Conversely... Figure 19The nPSK phase mapping circuit 1908 uses different phases depending on the PSK modulation order being simulated. For simplicity, a phase mapping starting with phase "0" is used. In the case of QPSK, a phase mapping of [0, π / 2, -π / 2, -π] is used here, which preserves the Gray mapping but is rotated by π / 4 radians relative to the dataset 1807 above. For BPSK, [0, -π] is used; for 8PSK, [0 1 3 2 -1 -2 -4 -3]*π / 4 is used; for 16PSK, [0 1 3 2 7 6 4 5 -1 -2 -4 -3 -8 -7 -5 -6]*π / 8 is used; and for 32PSK, [0 1 3 2 7 6 4 5 15 14 12 13 8 9 11 10 -1 -2 -4 -3 -8 -7 -5 -6 -16 -15 -13 -14 -9 -10 -12 -11]*π / 16 is used. All of these implement Gray mapping.

[0299] PSK Hardware Decoding Mapping

[0300] As mentioned earlier, the hard decision demapping process 1812 is used in the test platform 1800 to compare the hard BER 1809 and soft BER 1810 in the EXIT chart 1801. The k=4 bits of the message modulated by the OS scheme are demapped by the hard demapping circuit 1808 in the symbol hard decision circuit 1814 by taking the index of the maximum amplitude IQ complex pair and converting it to binary in the bit conversion circuit 1815. This has been described in detail above for the OS-only scheme.

[0301] To find the Q of messages modulated by the PSK scheme mThe bit, hard demapper circuit 1808 (Figures 18(A) to 18(B)) applies the following method. The symbol hard decision circuit 1814 first calculates the difference between the phase of the maximum amplitude correlator IQ pair (in the received M=16 correlator outputs) and each phase in the phase mapping vector in the applied PSK scheme (e.g., the vector [0, π / 2, -π / 2, -π] in QPSK). Then, the symbol hard decision circuit 1814 of the hard demapper circuit 1808 takes the PSK phase with the smallest difference from the phase of the maximum amplitude IQ complex value as the phase of the PSK symbol, and converts the index of this phase into binary according to the PSK phase mapping vector in the bit conversion circuit 1815. For example, it can be assumed that QPSK modulation with mapping vector [0, π / 2, -π / 2, -π] is applied, and the phase of the maximum amplitude IQ pair is 5π / 3. Through some calculations, it can be observed that the angle 5π / 3 in the range [0, 2π] is closest to the angle -π / 2 in the range [-π, π] (phase mapping). The value -π / 2 has an index of 2 in the QPSK mapping [0, π / 2, -π / 2, -π]. Therefore, in this example, the Q of the 16OS-QPSK modulation is... m The binary value of the 1808 hard demapper circuit with 2 most significant bits will be "10".

[0302] Quality measurement of each scheme

[0303] Using the 16OS-PSK scheme simulated with the example test platform 1800, it is possible to measure not only the quality parameters representing the result of all message bits in the combined stream 1804, but also the quality parameters of the corresponding message portions in each of the OS stream 1802 and PSK stream 1803. For example, in a 16OS-QPSK scheme with a total of N bits in a code block, the quality parameters of each k+Q bit of the transmitter's coded bits 1001 are measured. m =6 bits and the output bits 1035 of the hard demapper will include k=4 16OS bits, Q m =2 QPSK bits and a total of 6 16OS-QPSK bits — each of these three bits is called a bit group. The same applies to the bit LLRs generated from the soft demapper circuit 1522, which includes extrinsic bit LLR 904 and a posteriori bit LLR 1042, in the case of LLR groups. When processed simultaneously, the bit and LLR groups are indicated by three cascaded arrows, pointing to the OS stream 1802, PSK stream 1803, and combined stream 1804 in Figures 18(A) to 18(B).

[0304] Figures 20(A) to 20(B) show three EXIT graphs for three different 16OS-QPSK modulation schemes. The middle curve 2001 in Figures 20(A) to 20(B) corresponds to the Q value of the first 6-bit transmission during the processing of N bit sets. m = 2 QPSK bits / bit LLR, and corresponding to the Q in the second transmission m = 2 QPSK bits / bit LLR, and so on, until the Q of the last transmission during the processing of N bits. m = 2 QPSK bits / bit LLR. The same applies to the right curve 2002 in Figures 20(A) to 20(B): the curve data corresponds to the set of all k = 4 16OS bits / bit LLRs, where for N bits, each 16OS bit / bit LLR appears in different transmissions in all transmissions.

[0305] Measuring the hard BER 1809 of the 1036 16OS-PSK, 16OS and PSK streams requires the coded bit group 1001 from the transmitter and the output bits 1035 of the hard demapper, as shown in Figures 18(A) to 18(B). The three soft BER values ​​1810 are calculated from the coded bit group 1001 of the transmitter chain (as a reference) and the a posteriori bit LLR group 1042 that is converted 1041 into hard bits 1816 (as soft-to-hard bits). The mutual information histogram (MI) parameters of the bit LLRs of the three modulation schemes in the 1006 message can be measured by using the coded bit group 1001 from the transmitter chain and the external bit LLR group 904 from the soft demapper circuit 1522. [1] These external bit LLR groups 904 are also used to calculate the MI average parameters 1007, as shown in Figures 18(A) to 18(B).

[0306] result

[0307] Figures 25(A) through 25(F) show the EXIT charts for the MP -7.5dB dataset with different PSK modulation orders for QPSK 2501, 8PSK 2502, 16PSK 2503, and 32PSK 2504. Only the worst-case SNR value at -7.5dB in EXIT chart 2500 was selected to be included because the higher SNR causes all quality amplitudes (including MI values ​​and 1-BER figures) to approach 1. Furthermore, there was no significant difference between the results for AWGN and MP channels, so only the multipath channel chart 2500 was included. Details regarding which information is included in the EXIT charts were described above for the OS-only scheme.

[0308] EXIT chart

[0309] OS Chart – PSK Chart

[0310] As can be observed from EXIT chart 2500, apart from minor variations, increasing the PSK modulation order has little impact on the quality of the 16OS LLR characterized in EXIT chart 2505. This means that the models below dataset 1807 can largely maintain the independence of the modulation schemes from each other. In contrast, increasing the PSK modulation order does have a significant impact on the quality of the PSK bit LLR characterized in EXIT chart 2506: the quality measurement decreases with increasing PSK modulation order. This is expected, as increasing the number of modulation phases increases the chance of (i) generating accurate phase in the transmitted signal and (ii) identifying the correct phase from more error-prone data.

[0311] It is worth noting that, in certain cases of 16OS-QPSK, the OS bit LLR and PSK bit LLR have similar qualities. In higher-order PSK schemes, the PSK bit LLR has a lower quality than the OS bit LLR.

[0312] OS-PSK charts

[0313] The EXIT chart 2507 for the OS-PSK symbol is based on measurements of all bit LLRs of the combined stream 1804 (including the OS bit LLRs of OS stream 1802 and the PSK bit LLRs of PSK stream 1803), thus showing the average behavior between the corresponding OSEXIT chart 2505 and PSK EXIT chart 2506 for each PSK modulation order. For example, for the 16OS-32PSK scheme with MI=0, although the histogram MI measurements for the 16OS bit LLR and the 32PSK bit LLR are 0.87 and 0.44 respectively, the histogram MI value for the 16OS-32PSK bit LLR is 0.63, which is between the other two measurements. Here, in the case of 16OS-32PSK, the average is a weighted average, with the OS bit LLR carrying only 4 / 5 of the weight of the PSK bit LLR.

[0314] In the discussion of OS-only schemes, 1000 EXIT charts 1300-1400 are generated from a single scheme (16OS), and the upper limit of the coding rate is represented by the area under the MI histogram. In this section, schemes with different PSK modulation orders exist, and therefore, different total numbers of transmitted bits exist for these schemes. For example, the 16OS-QPSK scheme in EXIT chart 2501 contains 4 + 2 = 6 bits per transmission, and the 16OS-8PSK scheme in EXIT chart 2502 is based on 4 + 3 = 7 bits per transmission. However, although the number of coded bits per transmission increases with the PSK modulation order, the reliability of the transmission decreases, and a lower channel coding rate is required. It is anticipated that an "optimal point" can be found where the benefits of increasing the number of coded bits per transmission maximize outweigh the cost of requiring a lower coding rate. Assuming the coding rate is a function of the total number of bits transmitted, in each 16OS-nPSK EXIT chart 2500, the area under the histogram MI plot is multiplied by the total number of bits to provide a common measure of comparable information transmission across different nPSK schemes, where n is different. For example, when moving from 16OS-QPSK EXIT chart 2501 to 16OS-8PSK EXIT chart 2502, although the value of the MI plot decreases, it can be observed that the upper limit 2508 of the achievable coding rate increases (for both with and without iterative feedback). This is because, although the area under the histogram MI plot decreases in this movement, the increase in the number of bits transmitted from 6 to 7 results in a considerable jump in the maximum coding rate, for example, from 5.71 to 6.26 in the iterative case.

[0315] Generally speaking, it can be said that using a higher-order PSK allows for more coded bits per symbol, but results in a lower channel coding rate for reliable decoding—these two effects are opposite, but to varying degrees at different PSK modulation orders. Results show that the 16OS-16PSK scheme in EXIT chart 2503 has maximum achievable coding rates of 6.31 and 5.79 with and without iterative feedback, respectively, conveying the most information. However, for such a higher PSK modulation order, phase tracking is expected to become more challenging, which may be beneficial in practice for reducing the value. Furthermore, 4 PSK bits are more error-prone than 4 OS bits, potentially making channel decoder optimization more difficult. In contrast, it can be observed that in the 16OS-QPSK scheme in EXIT chart 2501, 2 PSK bits and 4 OS bits are equally error-prone (more or less) and support reliable phase tracking, which is expected to reduce implementation challenges.

[0316] A sample application

[0317] Now for reference Figure 37 This illustration demonstrates a typical computing system 4600 that can be used to implement software demapping according to some exemplary embodiments of the present invention. This type of computing system can be used in wireless communication units. Those skilled in the art will also recognize how to implement the invention using other computer systems or architectures. The computing system 4600 can represent, for example, a desktop computer, laptop or notebook computer, handheld computing device (PDA, mobile phone, PDA, etc.), mainframe, server, client, or any other type of dedicated or general-purpose computing device that may be ideal or suitable for a given application or environment. The computing system 4600 may include at least one processor, such as processor 4604. Processor 4604 may be implemented using a general-purpose or dedicated processing engine (e.g., a microprocessor, controller, or other control logic). In this example, processor 4604 is connected to bus 4602 or other communication medium. In some examples, the computing system 4600 may be a non-transitory tangible computer program product, including executable code stored therein for implementing software demapping.

[0318] The computing system 4600 may also include a main memory 4608, such as random access memory (RAM) or other dynamic memory, for storing information and instructions to be executed by the processor 4604. The main memory 4608 may also be used to store temporary variables or other intermediate information during the execution of instructions to be executed by the processor 4604. The computing system 4600 may also include read-only memory (“ROM”) or other static storage devices coupled to the bus 4602 to store static information and instructions for the processor 4604.

[0319] The computing system 4600 may also include an information storage system 4610, which may include, for example, a media drive 4612 and a removable storage interface 4620. The media drive 4612 may include a drive or other mechanism that supports fixed or removable storage media, such as a hard disk drive, floppy disk drive, magnetic tape drive, optical disc drive, compact disc (CD) or digital video drive (DVD) read or write drive (R or RW), or other removable or fixed media drive. Storage medium 4618 may include, for example, a hard disk, floppy disk, magnetic tape, optical disc, CD or DVD, or other fixed or removable media read and written by the media drive 4612. As these examples illustrate, storage medium 4618 may include a computer-readable storage medium in which specific computer software or data is stored.

[0320] In an alternative embodiment, information storage mechanism 4610 may include other similar components for allowing computer programs or other instructions or data to be loaded into computing system 4600. Such components may include, for example, removable storage unit 4622 and interface 4620, such as program box and box interface, removable memory (e.g., flash memory or other removable memory modules) and memory slots, as well as other removable storage units 4622 and interfaces 4620 that allow software and data to be transferred from removable storage unit 4618 to computing system 4600.

[0321] The computing system 4600 may also include a communication interface 4624. The communication interface 4624 can be used to allow software and data to be transferred between the computing system 4600 and external devices. Examples of the communication interface 4624 may include a modem, a network interface (such as Ethernet or other NIC cards), a communication port (such as a Universal Serial Bus (USB) port), a PCMCIA slot, and cards. The software and data transmitted via the communication interface 4624 are in the form of signals, which may be electronic, electromagnetic, and optical signals, or other signals that can be received by the communication interface 4624. These signals are provided to the communication interface 4624 via a channel 4628. The channel 4628 can carry signals and can be implemented using wireless media, wires or cables, optical fibers, or other communication media. Some examples of channels include telephone lines, cellular telephone links, RF links, network interfaces, local area networks (LANs) or wide area networks (WANs), and other communication channels.

[0322] In this document, the terms "computer program product," "computer-readable medium," etc., can generally be used to refer to media such as memory 4608, storage device 4618, or storage unit 4622. These and other forms of computer-readable media can store at least one instruction for use by processor 4604 to cause the processor to perform a specified operation. Such instructions, commonly referred to as "computer program code" (which may be grouped as a computer program or other groupings), when executed, enable computing system 4600 to perform the functions of embodiments of the present invention. It should be noted that code may directly cause the processor to perform a specified operation, be compiled to perform this operation, and / or be combined with other software, hardware, and / or firmware components (e.g., libraries for performing standard functions) to perform this operation.

[0323] In embodiments using software-implemented elements, the software may be stored on a computer-readable medium and loaded into the computing system 4600 using, for example, a removable storage drive 4622, a drive 4612, or a communication interface 4624. When executed by the processor 4604, control logic (in this example, software instructions or computer program code) causes the processor 4604 to perform the functions of the invention as described herein.

[0324] In the foregoing description, the invention has been described with reference to specific examples of embodiments thereof. However, it will be apparent that various modifications and alterations may be made therein without departing from the scope of the invention as set forth in the appended claims, and the claims are not limited to the specific examples described above.

[0325] The connections discussed herein can be any type of connection suitable for transmitting signals from or to a corresponding node, unit, or device, for example, via an intermediate device. Therefore, unless otherwise implied or stated, a connection can be, for example, a direct connection or an indirect connection. Connections can be illustrated or described with reference to a single connection, multiple connections, unidirectional connections, or bidirectional connections. However, different embodiments can change the implementation of the connection. For example, a single unidirectional connection can be used instead of a bidirectional connection, and vice versa. Furthermore, multiple connections can be replaced by a single connection that transmits multiple signals serially or in a time-multiplexed manner. Similarly, a single connection carrying multiple signals can be separated into various different connections carrying subsets of those signals. Therefore, there are many options available for transmitting signals.

[0326] Those skilled in the art will recognize that the architecture described herein is merely exemplary, and that many other architectures that achieve the same functionality can actually be implemented.

[0327] Any arrangement of components that achieve the same function is effectively “associated” to enable the desired functionality. Therefore, any two components combined in this paper to achieve a specific function can be considered “associated” with each other to achieve the desired functionality, regardless of the architecture or intermediate components. Similarly, any two such associated components can also be considered “operationally connected” or “operationally linked” to each other to achieve the desired functionality.

[0328] Furthermore, those skilled in the art will recognize that the boundaries between the operations described above are merely illustrative. Multiple operations can be combined into a single operation, a single operation can be assigned to additional operations, and operations can be performed with at least partial overlap in time. Additionally, alternative embodiments may include multiple instances of a specific operation, and the order of operations may vary in various other embodiments.

[0329] This invention is described herein with reference to integrated circuit devices, such as microprocessors configured to perform software demapping functions. However, it should be understood that the invention is not limited to such integrated circuit devices and can be equally applied to integrated circuit devices including any alternative types of operational functions. Examples of such integrated circuit devices including alternative types of operational functions may include application-specific integrated circuit (ASIC) devices, field-programmable gate array (FPGA) devices, or those integrated with other components, etc. Furthermore, since the illustrated embodiments of the invention can be largely implemented using electronic components and circuits known to those skilled in the art, details are not explained beyond any greater scope deemed necessary for understanding and comprehending the basic concepts of the invention, and so as not to obscure or distract from the teachings of the invention. Alternatively, examples of circuits and / or components may be implemented as any number of discrete integrated circuits or discrete devices interconnected to each other in a suitable manner.

[0330] For example, an example or a portion thereof may be implemented as a soft or coded representation of a physical circuit or a logical representation that can be converted into a physical circuit, such as in any suitable type of hardware description language.

[0331] Furthermore, embodiments of the present invention are not limited to physical devices or units implemented in non-programmable hardware, but can also be applied to programmable devices or units capable of performing desired software demapping by operating according to appropriate program code, such as minicomputers, personal computers, notepads, personal digital assistants, video games, automobiles and other embedded systems, mobile phones and various other wireless devices, generally referred to as "computer systems" in this application.

[0332] However, other modifications, variations, and alternatives are also possible. Therefore, the specification and drawings are to be regarded as illustrative rather than restrictive.

[0333] In the claims, any reference signs placed between parentheses shall not be construed as limiting the claims. The word “comprising” does not exclude the presence of other elements or steps besides those features or steps listed in the claims. Furthermore, the term “a / an” as used herein is defined as at least one. Additionally, the use of introductory phrases such as “at least one” and “at least one” in the claims should not be construed as implying that another claim element introduced by the indefinite article “a (a)” or “an” limits any specific claim that includes such introduced claim element to an invention that includes only one such element, even when the same claim includes the introductory phrase “at least one” or “at least one” and indefinite articles such as “a (a)” or “an”. The same applies to the use of definite articles. Unless otherwise noted, terms such as “first” and “second” are used to make arbitrary distinctions between the elements described by such terms. Thus, such terms are not necessarily intended to indicate the time or other priority of such elements. The mere fact that certain measures are referenced in mutually different claims does not indicate that a combination of such measures cannot be used advantageously. The term “subset” refers to the selection of elements from a set, where the selection may include one, some, or all of the elements in the set.

[0334] References

[0335] [1] J. Hagenauer, "The exit chart - introduction to extrinsicinformation transfer in iterative processing," 200412thEuropeanSignalProcessi ngConference , Vienna, 2004, pp. 1541-1548.

[0336] [2] 3GPP TS 38.212, "NR; Multiplexing and channel coding", v16.1.0, March 2020.

[0337] [3] ZB Kaykac Egilmez, L. Xiang, RG Maunder and L. Hanzo, "The Development, Operation and Performance of the 5G Polar Codes," in IEEE Communications Surveys & Tutorials, vol. 22, no. 1, pp. 96-122, Q1 2020.

[0338] [4] S. B. Wicker and V. K. Bhargava, eds. "Reed-Solomon codes and their applications," John Wiley & Sons, 1999。

[0339] [5] J. Neasham, “Simulated cross-correlation dataset for the 16-ary Orthogonal Signalling scheme,” Newcastle University, 10 / 01 / 2020。

[0340] [6] J. Neasham, “Simulated cross-correlation dataset for the 16-ary Orthogonal Signalling – Phase Shift Keying scheme,” Newcastle University, 10 / 01 / 2020。

[0341] [7] “Modulation using the 16-ary Orthogonal Signalling scheme, and the 16-ary Orthogonal Signalling – Phase Shift Keying scheme,” Newcastle University and Sonardyne International Ltd, 2020。

Claims

1. A communication unit for performing soft-decision demodulation, the communication unit comprising a receiver, wherein the receiver is arranged to receive a transmitted signal having a first bit set comprising k bits, the transmitted signal being determined from 2... k The transmitted signal is selected from a set of possible signals, and the transmitted signal includes a second set of bits, the second set of bits including those based on 2... Qm The phase rotation Q of the transmitted signal is selected from a set of possible rotations. m bits, wherein the receiver includes: Demodulator, which includes a set of 2 k One correlator, and configured as follows: Detect the transmission of each possible transmitted signal, and The output of the correlator is 2 k Each phase is used as the third input set; Demapper circuitry, which is connected to the demodulator and configured as follows: Receive the third input set; as well as Determine the statistics derived from the phase distributions of the multiple aggregate correlator outputs of the third input set, and calculate, based on the statistics, Q... m The second a posteriori soft bit set of 1 soft bit is output.

2. The communication unit according to claim 1, wherein the number of phase distributions output by the aggregation correlator is 1, and when the 2 k When one of the possible signals is selected as the transmitted signal, the phase distribution of the aggregate correlator output is approximately equal to that of the 2... k The output phase of the correlator is 2 k Aggregation of distributions.

3. The communication unit according to claim 2, wherein the phase distribution of the aggregate correlator output is represented by a second set of distribution parameters.

4. The communication unit according to claim 3, wherein the second distributed parameter set includes extended parameters. 相位 .

5. The communication unit of claim 3, wherein the demodulator is further configured to, based on the detected possible transmission signals, be connected by the set of 2 k The correlator outputs 2 k Each amplitude is used as the first input set.

6. The communication unit of claim 5, wherein the demapping circuitry is configured to perform at least one of the following: The second set of distribution parameters of the aggregated correlator output phase distribution is estimated by fitting the fifth probability distribution to the phase error of the correlator output obtained by correlating the received synchronization signal with a correlator corresponding to a known synchronization signal. By fitting the sixth probability distribution to the set of 2 k The two correlators obtained 2 k The second set of distribution parameters for estimating the aggregated correlator output phase distribution is determined by finding the phase error of the correlator output with the largest amplitude among the set of correlator outputs. The second set of distribution parameters for the output phase distribution of the aggregate correlator is selected from a lookup table connected to the demapping circuit.

7. The communication unit according to claim 6, wherein at least one of the fifth probability distribution or the sixth probability distribution is a Gaussian distribution.

8. The communication unit according to claim 3, wherein the 2 k The output phase of each correlator is combined with the second set of distribution parameters of the aggregate correlator output phase distribution to obtain a set including 2 k+Qm The third prior soft signal set of each soft phase.

9. The communication unit of claim 3, wherein the demapping circuit is configured to receive Q... m The fourth set of soft prior signals of 1 soft bits is used as the fourth input set.

10. The communication unit of claim 9, wherein the soft bits of the second posterior soft bit set are combined with the soft bits of the fourth prior soft signal set to obtain a combination including Q. m The second set of external soft bits of each soft bit.

11. The communication unit of claim 8, wherein the communication unit combines all prior soft signals to generate 2 k +Qm A set of posterior soft signals, and wherein the 2 k+Qm The set of posterior soft signals is combined to obtain the second posterior soft bit set.

12. A method for performing soft-decision demodulation, comprising a communication unit having a receiver having a demapper circuit coupled to a demodulator, the demodulator including a set of 2 k A correlator, the method at the receiver comprising: Receive a transmitted signal having a first bit set and a second bit set comprising k bits, the transmitted signal being determined according to the values ​​of the k bits from 2... k The second bit set is selected from a set of possible signals, including those based on 2... Qm The phase rotation Q of the transmitted signal is selected from a set of possible rotations. m bits, Detect the transmission of each possible transmitted signal, and The output of the correlator is 2 k Each phase is used as the third input set; Receive the third input set; as well as Determine the statistics derived from the phase distribution of the outputs of multiple aggregate correlators in the third input set; as well as Calculate Q based on the aforementioned statistics. m The second a posteriori soft bit set of 1 soft bit is output.

Images