Method for symbol time recovery by a receiving device

Abstract

The present invention relates to a method (100) for recovering symbol times by a receiving device (31) to decode a sequence of symbols (40) transmitted by a transmitting device (20) when the symbol times of the transmitting device are biased with respect to the symbol times of the receiving device. When a transition between two consecutive symbols is detected (102), an absolute error at the current symbol instance is measured (103), and a statistical model of bias is updated (104). A correction (105) can then be applied to subsequent symbol instants according to the measured absolute error and / or the bias estimated from the statistical model. During periods without transitions between symbols, the absolute error cannot be measured, but it is still possible to apply a correction (107) to subsequent symbol instants according to the relative error extrapolated from the statistical model.

Description

Technical Field

[0001] The present invention belongs to the field of digital communication. More specifically, the present invention relates to a method for recovering symbol time by a receiving device in order to decode a sequence of symbols transmitted by a transmitting device when the symbol time of the transmitting device can be modeled by a stationary process having a non-negligible bias with respect to the symbol time of the receiving device.

Background Art

[0002] In a digital communication system, when a transmitter transmits a message in the form of a sequence of symbols, a coherent receiver needs to recognize the synchronization of the symbols transmitted by the transmitter so as to be able to decode the message.

[0003] The symbol time in a transmitting device is determined from a clock belonging to the transmitting device. The symbol time in a receiving device is determined from a clock belonging to the receiving device. However, if the two clocks have a non-negligible frequency difference from each other, a bias may exist between the clock of the transmitting device and the clock of the receiving device. Such a situation results in a drift of the symbol time of the receiving device with respect to the symbol time of the transmitting device, which leads to an increase in the error in the determination of the instance of the symbol during sampling by the receiving device. This may cause an error in symbol decoding and the possibility of being unable to decode the received message.

[0004] There are various methods to recover the symbol time so that the ideal sampling point of the symbol can be estimated. Methods for tracking the synchronization of digital signals are usually based on successive measurement and correction steps. Each method includes specific constraints on channel coding, modulation pulse shaping, symbol generation, signal sampling, etc. For example, methods based on the Gardner algorithm allow measuring and correcting the error in the symbol time at each transition between two consecutive symbols.

[0005] However, when the bias between the symbol time of the transmitting device and the symbol time of the receiving device is large, especially when the modulation used does not necessarily cause a state transition between two consecutive symbols, the existing methods are not always effective enough. For example, it is certainly possible to use a modulated pulse with a shape that includes a state transition at each symbol, such as return-to-zero (RZ) coding. With such coding, the signal has a state transition at each symbol, even if two identical symbols are consecutive. However, return-to-zero coding consumes twice the bandwidth to achieve the same throughput compared to non-return-to-zero (NRZ) formats.

[0006] In Internet of Things (IoT) type wireless communication systems, the transmitting devices are generally low-cost devices and the technical specification of the symbol transmission frequency ("baud rate") should not be too restrictive. For such systems, it is necessary to find a solution that can effectively synchronize the symbol times at the receiving devices, especially when the modulation used does not prioritize transitions between symbols. Summary of the Invention

[0007] The present invention aims to overcome all or some of the drawbacks of the prior art, in particular those mentioned above.

[0008] For this purpose, and according to a first aspect, the present invention proposes a method for recovering the symbol time by a receiving device in order to decode a sequence of symbols transmitted by a transmitting device when the symbol time of the transmitting device has a bias with respect to the symbol time of the receiving device. This method comprises the following steps, namely, - Sampling of a baseband signal representing a sequence of symbols, each symbol being associated with a sampling "symbol instant", the sampling and - Detection of a transition between a current symbol, called "current symbol", and a symbol preceding the current symbol, said transition being detected when the current symbol and the preceding symbol have different states, the detection and - If a transition exists, ○ Measurement of the absolute error at the symbol instant of the current symbol and ○ Update of a statistical model of the bias between the symbol time of the transmitting device and the symbol time of the receiving device according to the measured absolute error and ○ Correction of the symbol instants of subsequent symbols according to the measured absolute error and / or according to the bias estimated from the statistical model and - If no transition exists, ○ Extrapolation from the statistical model of the relative error at the symbol instant of the current symbol and ○ Correction of the symbol instants of subsequent symbols according to the extrapolated relative error, and includes.

[0009] The absolute error measured or the relative error extrapolated at the symbol instant of the current symbol represents a delay or an advance of the estimated symbol instant with respect to the exact symbol instant of the current symbol.

[0010] The bias estimated from the statistical model represents the drift of the symbol time of the transmitting device with respect to the symbol time of the receiving device. This drift causes additional delay or advance for each new symbol. Therefore, the estimated bias represents the time error per symbol.

[0011] In such a configuration, even when an absolute error cannot be measured during a period without transitions between symbols, it is possible to apply a correction to the subsequent symbol instances according to the bias estimated using the statistical model.

[0012] In a particular implementation, the present invention may further include one or more of the following features, either alone or in all technically possible combinations.

[0013] In a particular implementation, the correction of the symbol instance of the subsequent symbol is performed only when a predetermined criterion is verified during the verification step.

[0014] In a particular implementation, the verification of the predetermined criterion includes a comparison between the value of the correction to be performed and a predetermined threshold.

[0015] Such a configuration can avoid inadvertent corrections caused by, for example, temporal variations in symbol time that affect only locally a particular symbol. In particular, this can prevent the error after correction from worsening compared to before correction.

[0016] In a particular implementation, the verification of the predetermined criterion includes a comparison between some elements included in the statistical model and a predetermined threshold.

[0017] In a particular implementation, the verification of the predetermined criterion includes verifying that the value of the correction to be performed is within the confidence interval targeted by a given statistical test output.

[0018] In fact, the greater the number of elements included in the statistical model, the higher the accuracy of the bias estimation. Therefore, with such a configuration, when the reliability given to this correction is not sufficient (that is, when the probability that the correction is actually relevant is not sufficiently high), the application of the correction can be avoided.

[0019] In a particular implementation, the baseband signal is obtained by a receiving device from a signal that is a phase or frequency modulated by a sequence of symbols by a transmitting device.

[0020] In a particular implementation, the signal is modulated by a transmitting device by BPSK, DBPSK, GFSK, or DGFSK modulation.

[0021] In a particular implementation, this method includes a preliminary step of inserting a synchronization pattern at the beginning of the symbol sequence by a transmitting device before transmitting the symbol sequence, and the synchronization pattern includes a sequence of symbols known to both the transmitting device and the receiving device.

[0022] In a particular implementation, the sequence of symbols of the synchronization pattern has a ratio of the number of transitions between symbols to the number of symbols constituting the synchronization pattern equal to at least a predetermined threshold. For example, the number of transitions between symbols is at least equal to half the number of symbols constituting the synchronization pattern.

[0023] According to a second aspect, the present invention relates to a computer program product including a set of program code instructions that configure the processor to implement a method for recovering symbol time according to any one of the foregoing implementations when executed by the processor.

[0024] According to a third aspect, the present invention relates to a receiving device of a communication system including means configured to implement a method for recovering symbol time according to any one of the foregoing implementations.

[0025] According to a fourth aspect, the present invention relates to a communication system including such a receiving device.

[0026] The present invention is given by way of non-limiting examples and will be better understood by reading the following description made with reference to FIGS. 1 to 6, which represent the following.

Brief Description of the Drawings

[0027]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Modes for Carrying Out the Invention

[0028] In these figures, the same reference from one figure to another indicates the same or similar elements. For clarity, unless otherwise specified, the elements shown are not necessarily to be at the same scale.

[0029] As described above, the present invention finds particularly advantageous applications in, but is not limited to, IoT-type wireless communication systems. In the remainder of the description, the case of such a system is considered as a non-limiting example.

[0030] FIG. 1 schematically represents an IoT-type wireless communication system 10 including one or more terminals 20 and an access network 30. The access network 30 includes several base stations 31 and a server 32 connected to the base stations 31.

[0031] In such a wireless communication system 10, data exchange is essentially unidirectional, in this case, from the terminal 20 to the access network 30 of the wireless communication system 10 by uplink. In the remainder of the description, as a non-limiting example, the terminal 20 is considered to be the transmitting device 20. However, nothing prevents the terminal from also functioning as a receiving device for receiving messages transmitted from the access network 30. When the terminal functions as a receiving device, a method for recovering the symbol time according to the present invention can also be implemented.

[0032] To minimize the risk of losing messages transmitted by the transmitting device 20, the access network planning is often carried out such that a given geographical area is covered simultaneously by several base stations 31, and the messages transmitted by the transmitting device 20 will be received by several base stations 31.

[0033] Each base station 31 is adapted to receive messages from the transmitting devices 20 within its range. Each message received in this way is transmitted, for example, to the server 32 of the access network 10, perhaps accompanied by other information such as the identifier of the base station 31 that received this message, a value representing the quality of the radio signal transmitting the message, the center frequency at which the message was received, the date on which the message was received, etc. The server 32 processes, for example, all the messages received from different base stations 31. The communication link between the base station 31 and the server 32 can be supported by an optical fiber or an electrical cable, but it can also be a wireless communication link.

[0034] In the remainder of the description, as a non-limiting example, base station 31 is considered to be a receiving device 31 that implements a method for recovering symbol time in accordance with the present invention. However, nothing prevents the base station from also functioning as a transmitting device for transmitting messages to terminal 20. In a variant, the decoding of the message can be performed at server 32 (and not at base station 31). In such a case, it is server 32 that functions as a receiving device that implements a method for recovering symbol time in accordance with the present invention.

[0035] Communication system 10 is, for example, a wireless low-power wide-area network known by the term LPWAN. Such a wireless communication system is a long-range access network (over 1 kilometer or even over dozens of kilometers), with low energy consumption (for example, the energy consumption during message transmission and reception is less than 100 mW, less than 50 mW, or even less than 25 mW), and its throughput is usually less than 1 Mbits / s. Such a wireless communication system is particularly suitable for applications including connected IoT-type objects.

[0036] In a particular implementation, communication system 10 can be an ultra-narrowband communication system. The term "ultra-narrowband" (or UNB) means that the instantaneous frequency spectrum of the radio signal emitted by transmitting device 20 has a frequency width of less than 2 kilohertz, or even less than 1 kilohertz. Such a system can significantly limit the power consumption of transmitting device 20 when communicating with the access network.

[0037] The transmitting device 20 is configured to transmit a message to the receiving device 31. For this purpose, the transmitting device 20 includes a processing circuit including a memory, one or more processors, and a communication module. The communication module can implement different steps of a digital transmission chain (source coding, channel coding, modulation, frequency translation, wireless transmission, etc.) in a conventional manner. For this purpose, the communication module includes a set of hardware and / or software means (encoder, local oscillator, mixer, filter, digital / analog converter, amplifier, antenna, etc.) that are considered known to those skilled in the art.

[0038] In the remainder of the description, by way of example and without limitation, the case where a message transmitted by the transmitting device 20 directed to the receiving device 31 is encoded by a sequence of binary symbols is considered. However, the present invention can also be applied to modulations using symbols that can take several different values greater than 2.

[0039] Figure 2 schematically represents a sequence 40 of binary symbols that encodes a message to be transmitted by the transmitting device 20 to the receiving device 31. Each binary symbol takes the value "1" or the value "0". Each symbol is transmitted at a symbol transmission frequency f’ S =1 / T’ S where T’ S is the duration of the transmission of the binary symbol.

[0040] Figure 2 also shows a signal 41 representing a sequence 40 of symbols to be transmitted. Signal 41 takes a "high" state for each symbol with a value of "1" and a "low" state for each symbol with a value of "0". Signal 41 is used, for example, to modulate a carrier wave in the form of a high-frequency sine wave signal such as a frequency in the ISM ("Industrial, Scientific and Medical") band. In the example considered, the carrier wave has a frequency of 868 MHz and the symbols are transmitted at a throughput of 100 baud (100 symbols per second). The modulation used is phase modulation, frequency modulation, or amplitude modulation. More specifically, phase modulation of the BPSK ("Binary Phase Shift Keying") type, or the DBPSK ("Differential Binary Phase Shift Keying") type can be used. According to other examples, frequency modulation of the GFSK ("Gaussian Frequency Shift Keying") or DGFSK ("Differential Gaussian Frequency Shift Keying") type can also be used. Such modulation is relatively easy to implement and is particularly suitable for low-cost IoT-type transmission devices. However, these modulations do not prioritize transitions between symbols. It should be noted that other modulations can be used and the choice of a particular modulation is only one variation of the present invention.

[0041] Next, the signal for transmitting the message transmitted by the transmitting device 20 corresponds to a carrier wave modulated by a signal 41 representing a sequence 40 of binary symbols encoding the message.

[0042] The receiving device 31 is configured to receive the message transmitted from the transmitting device 20. For this purpose, the receiving device 31 includes a processing circuit including a memory, one or more processors, and a communication module.

[0043] The communication module can, in a conventional manner, perform different steps of a digital reception chain (such as wireless reception, frequency conversion, demodulation, channel decoding, source decoding, etc.). For this purpose, the communication module includes a set of hardware and / or software means (antenna, amplifier, local oscillator, mixer, analog / digital converter, filter, decoder, etc.) that are considered known to those skilled in the art.

[0044] A computer program stored in the memory of the receiving device, when executed by a processor, includes a set of program code instructions that configure the processor to implement a method 100 for recovering symbol time in accordance with the present invention.

[0045] Alternatively, or in addition thereto, the receiving device 31 includes one or more programmable logic circuits (such as FPGA, PLD, etc.), and / or one or more application-specific integrated circuits (ASIC), and / or a set of discrete electronic components, etc., that are adapted to perform all or part of the steps of a method for recovering symbol time 100 in accordance with the present invention. In other words, the receiving device 31 includes means that are software (a specific computer program product) and / or hardware (FPGA, PLD, ASIC, discrete electronic components, etc.) configured to perform the steps of method 100 in accordance with the present invention.

[0046] FIG. 3 schematically represents a baseband signal 51 constructed in a conventional manner by a receiving device 31 from a transmitted and modulated carrier wave for transmitting a message encoded by a sequence 40 of binary symbols represented in FIG. 2. The baseband signal 51 represents the sequence 40 of binary symbols and assumes a "high" state when the symbol takes the value "1" and a "low" state when the symbol takes the value "0". Conventionally, the signal 51 is sampled by the receiving device 31 and each symbol is associated with a sampling "symbol instant". In FIG. 3, the value of each sample corresponding to a symbol instant is represented by a dot. The time difference between two symbol instants is T S as indicated.

[0047] The sampling frequency is at least equal to the symbol reception frequency f S = 1 / T S The sampling frequency is usually a multiple of the symbol reception frequency f S In FIG. 3, each cross represents a sample taken at the center of the interval separating two symbol instants.

[0048] Each binary symbol is associated with an index i. In FIG. 2, the samples with indices i = 0, 2, 3, 4, 6, 9, 10, 11 and 12 correspond to symbols with the value "1". This is because in these samples, the signal 51 assumes a "high" state corresponding to a sample value greater than a predetermined high threshold. The samples with indices i = 1, 5, 7, 8, 13 and 14 correspond to symbols with the value "0". This is because in these samples, the signal 51 assumes a "low" state corresponding to a sample value less than a predetermined low threshold.

[0049] As shown in Fig. 5, the value of the sample corresponding to the symbol with index i is represented by x(i). The value of the sample of the signal 51 located at the center of the interval between two samples with indices (i - 1) and i is represented by x(i - 1 / 2). In the example considered and illustrated in Fig. 5, the sampling frequency is 8 times the symbol reception frequency f S and the samples x(i - 7 / 8), x(i - 3 / 4), x(i - 5 / 8), x(i - 1 / 2), x(i - 3 / 8), x(i - 1 / 4) and x(i - 1 / 8) are found between the sample with index (i - 1) and the sample with index i.

[0050] To maintain perfect symbol synchronization, the frequency f’ S used by the transmitting device to transmit the symbol and the frequency f S used by the receiving device to receive the symbol must be the same. However, there may be a drift between the clock used by the transmitting device for symbol transmission and the clock used by the receiving device for symbol reception. In other words, there may be a bias between the symbol time of the transmitting device and the symbol time of the receiving device (the values T’ S and T S are not exactly the same). This bias is particularly important for low-cost transmitting devices where the reliability of the clock is not always high.

[0051] Fig. 4 shows the main steps of a method for recovering the symbol time 100 implemented by the receiving device 31 in order to efficiently decode the sequence 40 of binary symbols transmitted by the transmitting device 20, even when there is a non-negligible bias between the symbol time of the transmitting device and the symbol time of the receiving device.

[0052] The first step corresponds to the sampling 101 of the baseband signal 51. This step has already been described with reference to Figs. 2 and 5.

[0053] The second step corresponds to the detection 102 of the transition between the current symbol, which has index i, and the preceding symbol, which has index (i - 1). A transition is detected when the states of the current symbol with index i and the preceding symbol with index (i - 1) are different. In the example illustrated in FIG. 2, the symbols with indices i = 1, 2, 5, 6, 7, 9, and 13 correspond to the symbols for which a transition has been detected with respect to the preceding symbol. However, in the case of the symbols with indices i = 3, 4, 8, 10, 11, 12, and 14, no transition is detected with respect to the preceding symbol. Different methods can be used to detect the transition between the current symbol with index i and the preceding symbol with index (i - 1). For example, a transition is detected when the absolute value of the difference between sample x(i) and sample x(i - 1) is greater than a predetermined threshold.

[0054] When a transition is detected, the method for recovering the symbol time 100 includes the following steps, namely, - measuring 103 the absolute error at the symbol instant of the current symbol with index i, and - updating 104 the estimated average bias between the symbol time of the transmitting device and the symbol time of the receiving device according to the measured absolute error, and - correcting 105 the symbol instant of the subsequent symbol with index (i + 1) according to the measured absolute error and / or according to the estimated average bias. It should be noted that the order of steps 104 and 105 is not necessarily fixed. Thus, the correction 105 of the symbol instant of the subsequent symbol is probably performed according to the previous estimated average bias before the average bias is updated.

[0055] When no transition is detected, the method for recovering the symbol time 100 includes the following steps, namely, - The extrapolation 106 from the estimated average bias of the relative error at the symbol instant of the current symbol which is index i, and - The correction 107 of the symbol instant of the subsequent symbol which is index (i + 1) according to the extrapolated relative error.

[0056] The estimated average bias is stored by the receiving device 31. The estimated average bias is updated, for example, each time a transition is detected between two consecutive symbols. If there is no transition between two consecutive symbols, the estimated average bias is not updated, but it can still be used to correct the symbol time. Thus, according to the present invention, the method 100 can correct the symbol time even when there is no transition between symbols. This is particularly advantageous when many identical symbols follow each other without a transition, because symbol synchronization can be maintained even when the absolute error of the current symbol cannot be measured.

[0057] FIG. 5 schematically shows an example of the implementation of step 103 for measuring the absolute error at the symbol instant of the current symbol which is index i. The absolute error is E abs and is represented, for example,

[0058]

Equation

[0059] calculated in the form of.

[0060] The sign of the value of the term [x(i) - x(i - 1)] indicates whether the transition corresponds to a transition from a high state to a low state (the value of this term becomes negative), or a transition from a low state to a high state (the value of this term becomes positive).

[0061] When the transition corresponds to a transition from a high state to a low state, the term x(i - 1 / 2) is negative when the estimated symbol instant of the current symbol is delayed with respect to the exact instant of the symbol, and is positive when the estimated symbol instant of the current symbol precedes the exact instant of the symbol. Conversely, when the transition corresponds to a transition from a low state to a high state, the term x(i - 1 / 2) is positive when the estimated symbol instant of the current symbol is delayed compared to the exact instant of the symbol, and is negative when the estimated symbol instant of the current symbol precedes the exact instant of the symbol.

[0062] Therefore, the sign of the absolute error E abs indicates whether the measured error corresponds to a delay (the absolute error E abs is positive) or a lead (the absolute error E abs is negative) with respect to the exact symbol instant of the current symbol.

[0063] The absolute value of the term x(i - 1 / 2) specifically represents the importance of the error (corresponding to a delay or a lead) at the symbol instant of the current symbol. The larger this value, the greater the delay or lead with respect to the exact instant of the symbol.

[0064] K is a positive constant, and from the amplitude measurement performed on the current symbol, the absolute error E abs can be obtained in the time domain.

[0065] In the example illustrated in FIG. 5, the measured absolute error E abs represents the delay D of the estimated symbol instant of the symbol with index i with respect to the exact symbol instant of the symbol with index i. In the absence of a delay (and a lead), the sample x(i - 1 / 2) at the center of the time interval separating the sample x(i - 1) and the sample x(i) takes on a zero value.

[0066] Using other methods, the absolute error E of the symbol instance of the current symbol can be measured. The selection of a specific method for measuring this absolute error is merely one variation of the present invention. abs The value of the estimated mean bias is represented, for example, by β. The mean bias β is initialized to zero, for example, and when a transition is detected for the current symbol with index i > 0, step 104 for updating the mean bias β is executed, for example, as follows.

[0067]

[0068]

Equation

[0069] The estimated mean bias β then corresponds to a delay (when β > 0) or a lead (when β < 0) and is corrected for each newly considered current symbol. Other calculation methods can be used to estimate the mean bias β (for example, by linear regression or other statistical models, and in some cases, by removing specific measurements with excessive variations). The selection of a specific method for estimating the mean bias β is merely one variation of the present invention.

[0070] Next, the correction 105 can be applied to the symbol time. The value of the correction to be applied can be calculated according to the measured absolute error E abs and / or according to the estimated mean bias β.

[0071] For example, consider the scenario where the sampling frequency f E is N times the symbol reception frequency f S (f E = N × f S ). The correction 105 can consist of shifting the symbol instance of the subsequent symbol by a duration corresponding to a part of the duration T S . For example, the sample corresponding to the subsequent symbol with index (i + 1) becomes the following sample.

[0072] [Number]

[0073] Here, |A| is the integer part of A. In this formula, the average bias β used may be the average bias estimated before and after the update executed from the measured absolute error E abs from the estimated average bias before and after the update executed from the measured absolute error E

[0074] According to another example, when the correction 105 is calculated directly from the measured absolute error, the sample corresponding to the subsequent symbol with index (i + 1) becomes the sample.

[0075] [Number]

[0076] In the example illustrated in FIG. 5, the value of N is 8 (N = 8), and the correction 105 can be configured by advancing the subsequent symbol instant by the duration corresponding to T S / 8. Therefore, the symbol instant of the subsequent symbol with index (i + 1) is corrected to be the instant of the sample x(i + 7 / 8).

[0077] When there is no transition between the preceding symbol with index (i - 1) and the current symbol with index i, the absolute error at the estimated instant of the current symbol cannot be measured. However, the relative error at the estimated instant of the current symbol can be estimated from the estimated average bias β (in step 106). Therefore, this enables the symbol instant of the subsequent symbol with index (i + 1) to be corrected (in step 107) according to the extrapolated relative error. For example, the extrapolated relative error is equal to the estimated average bias β, and the correction applied is similar to that represented by the above (Equation 3).

[0078] However, it should be noted that corrections are not always applied for each new current symbol. For example, corrections are only applied when a predetermined criterion is verified. In this case, the cumulative error E acc needs to be maintained. The cumulative error E acc is updated for each new current symbol and reset each time a symbol time correction is performed.

[0079] Applying corrections only when a specific criterion is verified can, on the one hand, limit the number of corrections to be performed and, on the other hand, avoid inadvertent corrections due to temporal variations (jitter) of the symbol time that affect only locally, for example, a specific symbol. The bias is estimated according to a statistical model, and as the series of measurements increases, the weight of the measurements decreases. The application of the criterion can be linked to the resolution of the correction that can be set by the oversampling value. By using a specific criterion, it is possible, in particular, to prevent the error after correction from worsening compared to before correction.

[0080] FIG. 6 schematically represents the specific implementation steps of a method for recovering the symbol time 100 in which the correction 109 is applied only when a specific criterion is verified during the verification step 108.

[0081] Steps 101 to 106 correspond to steps 101 to 106 described above with reference to FIG. 4.

[0082] The correction 109 for the instance of the subsequent symbol which is the index (i + 1) corresponds to either the correction 105 described with reference to FIG. 4 when the transition of the current symbol which is the index i is detected, or the correction 107 described with reference to FIG. 4 when the transition of the current symbol which is the index i is not detected. However, the correction 109 is only performed when a specific criterion is verified during the verification step 108.

[0083] According to the first example, the cumulative error E accis calculated for each current symbol, and the correction 109 of the symbol instance of the subsequent symbol is performed only when the cumulative error E acc is greater than a predetermined threshold value. When the correction is performed, the cumulative error is reset to zero. According to another example, the correction 109 is performed only when the number of symbols elapsed since the last correction is greater than a specific threshold value.

[0084] In the case of the embodiment described with reference to FIG. 6, the estimated mean bias β is updated in step 104 according to the following formula, for example, when a transition is detected for the current symbol (if no transition is detected, the mean bias is not updated).

[0085]

Equation

[0086] The absolute error E abs corresponds to the temporal error measured for the instance of the current symbol. The absolute error E abs is specifically composed of the sum of the errors accumulated and not corrected between the preceding symbols.

[0087] The cumulative error E acc initially takes the value zero. When the current symbol with index i is processed, the cumulative error E acc is updated according to the following formula, for example. E acc = E acc + β (Equation 6)

[0088] According to another example, when the measurement of the absolute error E abs is performed for the current symbol, it is also possible to assign the value of the absolute error E abs to the cumulative error E acc (E acc = E abs ).

[0089] If the criterion is verified in step 108, the correction 109 is applied in a manner similar to that described with reference to the above formula (formula 3), and the cumulative error is reset (E acc = 0).

[0090] According to another example, instead of resetting the cumulative error E acc to the value zero, it is possible to subtract the value of the applied correction from the cumulative error E acc .

[0091] The above description has been made considering the calculation of the average bias between the symbol time of the transmitting device and the symbol time of the receiving device. The calculation of the average value of the bias can simplify implementation because it is not necessary to store the values of the continuously measured errors for each current symbol having a transition to the preceding symbol. In fact, it is sufficient to store the cumulative error to calculate the average value of the bias. However, when estimating the bias from the average value, the accuracy of the bias estimation may be insufficient. As described above, the bias can be estimated using a statistical model. The use of a statistical model necessarily means collecting and storing a large number of values corresponding to the continuously measured errors for symbols having transitions. In return, the accuracy of the bias estimation is significantly improved.

[0092] The statistical model can be, in particular, linear regression. However, other statistical models (non - linear regression, machine learning algorithms, etc.) can be considered.

[0093] Referring back to FIG. 4, when a statistical model is used, update step 104 corresponds to providing a statistical model with an absolute error measured with respect to the current symbol instant. Thereafter, step 105 of correcting the symbol instant of subsequent symbols can be performed in accordance with the measured absolute error and / or in accordance with a bias β estimated from the statistical model. The statistical model can provide information regarding, in particular, the estimated absolute value of the bias at the current instant (the estimated absolute error between the estimated symbol instant and the exact symbol instant of the current symbol), and / or the estimated value of the bias for each time symbol (the relative error introduced for each new symbol). In the absence of a transition, step 106 of extrapolating the relative error at the symbol instant of the current symbol is performed using the statistical model.

[0094] Note that the evolution of the existing bias between the symbol time of the transmitting device and the symbol time of the receiving device can change over time. Advantageously, the statistical model can model this evolution and accurately estimate future bias values even during periods where there is no transition between symbols and thus the error at the current symbol instant cannot be measured.

[0095] Repeating, the correction is not necessarily applied for each new current symbol. In fact, it is possible to perform the correction of the symbol instant only if a predetermined criterion is verified.

[0096] Referring to FIG. 6, when a statistical model is used, step 108 of verifying a predetermined criterion can include, in particular, comparing the value of the correction to be made with a predetermined threshold. Such a configuration can avoid inadvertent corrections, for example, due to significant temporal variations in the symbol time that only locally affect a particular symbol. In particular, this can prevent the error after correction from worsening compared to before correction.

[0097] Step 108 of verifying the predetermined criteria may also include comparing the number of elements included in the statistical model with a predetermined threshold. The number of elements included in the statistical model corresponds to the number of measurements 103 of absolute error that have been performed over time for the symbol for which the transition has been detected. Each absolute error measured in this way is actually stored and used to update the statistical model (step 104). The greater the number of elements included in the statistical model, the higher the accuracy of the bias estimation. Thus, with such a configuration, if the reliability given to this correction is not sufficient (or, in other words, if the correction does not have a sufficiently high probability of actually being relevant), the application of the correction can be avoided.

[0098] According to yet another example, the verification of the predetermined criteria includes verifying that the value of the correction to be performed is within the confidence interval for which the given statistical test output is targeted. The statistical test output of the test is the probability of wisely rejecting a hypothesis (null hypothesis) considered to be true a priori (because it is false). The statistical test output is the value (1 - b), where b is the probability of not rejecting the null hypothesis when the null hypothesis is false (b is the "type II risk"). The statistical test output is generally defined according to the number of elements in the statistical model, the spread of the elements, and the threshold of the test (critical probability or "p-value"). The confidence interval provides a range of possible values. The confidence interval is generally defined according to the number of elements in the statistical model, their variance, and the threshold of the test. The test threshold sets the confidence level of this interval. The confidence of the test is the probability of not rejecting the null hypothesis when the null hypothesis is true. The confidence is the value (1 - a), where a is the probability of rejecting the null hypothesis when the null hypothesis is true (a is the "type I risk").

[0099] As described above, using a statistical model can also exclude specific measurements with too much model variation. This can optimize the accuracy of bias estimation from the statistical model. Statistically, the accuracy of bias estimation improves over time. Advantageously, to optimize the convergence of the estimated bias to the actual value of the bias existing between the symbol time of the transmitting device and the symbol time of the receiving device, it is possible to insert a synchronization pattern including a sequence of symbols known to both the transmitting device and the receiving device at the beginning of each message. Preferably, the synchronization pattern is selected such that the symbols constituting the synchronization pattern have a number of transitions equal to at least half of the number of symbols constituting the synchronization pattern, i.e., a number of state transitions.

[0100] The above description clearly shows that the present invention achieves the set objectives through its different features and their advantages. In particular, the present invention can correct the symbol time of the receiving device even during a period when there is no transition between symbols. Such a configuration can maintain symbol synchronization between the transmitting device and the receiving device even when there is a significant bias between the symbol time of the transmitting device and the symbol time of the receiving device.

[0101] Measurements were carried out to demonstrate the effectiveness of the method according to the present invention. For a theoretical frequency of 100 baud symbols (100 symbols per second) and a signal-to-noise ratio (SNR) of 20 dB, a receiving device implementing the method according to the present invention can decode messages with a success rate close to 100% if the actual transmission frequency of the symbols of the transmitting device is composed of 92 to 108 baud (i.e., an error of + / - 8% of the theoretical symbol frequency). However, a receiving device not implementing the method according to the present invention can only decode messages with a success rate close to 100% when the actual transmission frequency of the symbols of the transmitting device is composed of 99.5 to 100.5 baud (i.e., an error of + / - 0.5% of the theoretical symbol frequency).

[0102] More generally, the implementations and embodiments considered above are illustrated by non-limiting examples, and thus it should be noted that other variations are possible. In particular, for detecting transitions between two consecutive symbols, for measuring the error at the instant of the current symbol having a transition with respect to the preceding symbol, and / or for estimating a bias from several measurements performed on symbols with transitions, the choice of a particular method simply corresponds to a variation of the present invention.

[0103] The present invention has been described with respect to binary modulation. However, nothing prevents the present invention from being applied to modulations in which symbols can take several discrete values greater than 2. Using modulation with symbols having three or more states can affect not only the method for detecting state transitions between two symbols, but also the method for measuring the absolute error regarding symbol instants. However, the core of the present invention, which is in the modeling of the bias for applying corrections to the symbol time during periods when there are no transitions between symbols, remains applicable.

[0104] The present invention has been described considering an IoT type wireless communication system 10. However, considering other digital communication systems including wire communication systems is not excluded.

Claims

1. A method (100) for symbol time recovery, the method being performed by a receiving device (31) to decode a sequence (40) of symbols received from a transmitting device (20), the symbol time of the transmitting device having a bias with respect to the symbol time of the receiving device, the method (100) comprising: - Sampling (101) to obtain samples of a baseband signal (51) representing the sequence of symbols, each symbol of the sequence of symbols being associated with a symbol instant corresponding to one sampling instant of the samples, the sampling (101); - Detection (102) of a transition between the current symbol and a preceding symbol immediately preceding the current symbol in the sequence of symbols, the transition being detected when the current symbol and the preceding symbol have different states, the detection (102); - If a transition exists, 〇 Measurement (103) of an absolute error at the symbol instant associated with the current symbol; 〇 Update (104) of a statistical model of the bias between the symbol time of the transmitting device and the symbol time of the receiving device according to the measured absolute error; 〇 Correction (105) of the symbol instant associated with a subsequent symbol immediately following the current symbol, the correction (105) including changing the sampling instant corresponding to the symbol instant associated with the subsequent symbol based on at least one of the measured absolute error and the bias estimated from the statistical model, the correction (105); - If no transition exists, 〇 Extrapolation (106) from the statistical model of a relative error at the symbol instant associated with the current symbol; 〇 Correction (107) of the symbol instant associated with the subsequent symbol, the correction (107) including changing the sampling instant corresponding to the symbol instant associated with the subsequent symbol based on the extrapolated relative error, the correction (107), the method (100).

2. The method (100) according to claim 1, wherein the statistical model is linear regression.

3. The method (100) according to claim 1, wherein the correction of the symbol instance of the subsequent symbol is performed only when a verification of a predetermined criterion is satisfied.

4. The method (100) according to claim 3, wherein the verification (108) of the predetermined criterion includes a comparison between a value of the correction to be performed and a predetermined threshold value.

5. The method (100) according to claim 3, wherein the verification (108) of the predetermined criterion includes verification that a value of the correction to be performed is within a confidence interval that is the subject of a given statistical test output.

6. The method (100) according to claim 5, wherein the baseband signal (51) is obtained by the receiving device (31) from a signal that is phase or frequency modulated by the sequence (40) of symbols by the transmitting device (20).

7. The method (100) according to claim 6, wherein the signal is modulated by the transmitting device (20) by BPSK, DBPSK, GFSK, or DGFSK modulation.

8. The method (100) according to claim 1, including a preliminary step of inserting a synchronization pattern at the beginning of the sequence (40) of symbols before transmission of the sequence (40) of symbols by the transmitting device (20), the synchronization pattern including a sequence of symbols known to both the transmitting device (20) and the receiving device (31).

9. The method (100) according to claim 8, wherein the sequence of symbols of the synchronization pattern is selected to indicate a ratio between the number of transitions between symbols and the number of symbols constituting the synchronization pattern equal to at least a predetermined threshold value.

10. A computer-readable storage medium including a set of program code instructions for configuring the processor to implement the method (100) for recovering the symbol time of the receiving device according to any one of claims 1 to 7 when executed by a processor of the receiving device.

11. A receiving device (20) of a communication system (10) including a processor configured to implement the method (100) for recovering the symbol time of the receiving device according to any one of claims 1 to 7.

12. A communication system (10) including the receiving device (20) according to claim 11.

Images