Hybrid reference signal with low PAPR for DFT-S-OFDM

By inserting reference samples and cyclic prefix samples in the DFT-Extended-OFDM signal, combined with the Zadoff-Chu sequence, the problems of low PAPR and power efficiency in DFT-Extended-OFDM modulation are solved, and low power fluctuations and high power in satellite transmission are achieved Spectral efficiency radio signals.

CN114731321BActive Publication Date: 2024-11-19MITSUBISHI ELECTRIC CORP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202080076155.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2019-11-05
Filing Date
2020-10-20
Publication Date
2024-11-19
Estimated Expiration
2040-10-20

AI Technical Summary

Technical Problem

DFT-Extension-OFDM modulation in applications in radio signals makes it difficult to maintain low power envelope fluctuations and low PAPR characteristics, resulting in poor power efficiency in environments where near-saturation power amplifiers are required, especially in satellite transmission reference Too many symbols are inserted to reduce spectral efficiency.

Method used

By inserting S reference samples and L cycle prefix samples into the DFT-Expand-OFDM signal, the signal structure is adjusted to generate mixed symbols, ensuring the ratio matching of data and reference samples, and using OFDM pilots in the channel estimation to keep them low PAPR features while inserting Zadoff-Chu sequences into the signal to achieve low power fluctuations.

Benefits of technology

Achieves adjusting the reference sample to total sample ratio without changing the signal structure, providing radio signals with low PAPR and low power fluctuations, suitable for environments where power savings are required, such as satellite transmission, improving channel estimation efficiency and robustness sex.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114731321B_ABST
    Figure CN114731321B_ABST
Patent Text Reader

Abstract

The invention relates to a method for transmitting at least one set of Kd data samples in a radio signal to be transmitted by a communication system, the radio signal being intended to be transmitted by a transmitter comprising at least one transmitting antenna, the radio signal being provided by: determining a first position J1 and a second position J2 in a block of K samples; applying a DFT-spread-OFDM-type scheme to the block of K samples A=(A0; ...; AK1) and obtaining an output signal representing a symbol comprising N complex samples Sn in the radio signal, where n=0 to N-1; determining a first position T1 and a second position T2 in the output signal such that: J 1=floor(T1×K / N); J2=ceil(T2×K / N); T2-T1=R, R is a strictly positive integer; inserting S reference samples and L cyclic prefix samples between a first position T1 and a second position T2 of the updated output signal, where L is a positive integer or 0, S is a strictly positive integer, and R=S+L, to obtain a subsequent signal representing a mixed symbol including a data segment and a reference segment; appending a cyclic prefix of at least part of the mixed symbol to the mixed symbol to obtain a signal representing the mixed symbol appended with the cyclic prefix; and transmitting a radio signal corresponding to the signal representing the mixed symbol appended with the cyclic prefix.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates generally to the field of telecommunications and more particularly to the incorporation of symbols similar to reference signals in the context of communications using OFDM (Orthogonal Frequency Division Multiplexing) based transmission schemes.

[0002] More specifically, DFT-spread-OFDM modulation is used to implement uplink transmissions in 3GPP / LTE networks under the acronym SC-FDMA (Single Carrier Frequency Division Multiple Access). It is also part of the New Radio (NR) standard for 5G systems under the 3GPP standardization body. This modulation is also standardized to implement the satellite component of the hybrid distribution of the DVB-NGH system under the acronym SC-OFDM (Single Carrier Orthogonal Frequency Division Multiplexing). Background Art

[0003] Two main technologies are used to achieve wireless transmission:

[0004] - Time Division Multiplexing (TDM) also known as Single Carrier (SC) modulation, and

[0005] - Orthogonal Frequency Division Multiplexing (OFDM) modulation.

[0006] In the first case, the information is sent as an M-ary symbol stream at the so-called baud rate (1 / T Hz), shaped with a time waveform to reduce the spectral occupation. These modulations are usually linear modulations of amplitude and phase. It is also possible to modulate the signal in frequency, but this approach is not described here. To simplify the receiver, the duration of the waveform can be limited to the symbol period T of the signal. Unfortunately, this will lead to too large a spectral occupation. There is nothing to prevent relying on waveforms with longer time support, but this usually requires the use of more complex receivers. In practice, as long as the waveform regularly goes to zero every period T, it is possible to limit the spectral occupation while maintaining the ability to sample the received signal at the baud rate. This is the case, for example, with the so-called square root raised cosine (SRRC) filter. The main limitation is to achieve very accurate time synchronization (sampling when the eye is closed). This description applies to the baseband. Obviously, the signal can be transposed in any frequency band by zero-IF or superheterodyne modulation. The SC signal can be described as follows:

[0007]

[0008] where f0 is the carrier frequency, T is the symbol period, g(t) is a signal waveform with finite energy - typically a square root raised cosine (SRRC) filter, and {x k} is a sequence of M discrete symbols carrying the information bits. Using a constellation alphabet with no or low amplitude variations (usually a phase-shift keying (PSK) alphabet) results in a modulated signal with low-power envelope fluctuations, which is directly related to the sharpness of the filtering in the frequency domain (the roll-off of the SRRC filter, from 0 to 100%). The power fluctuations are usually described by the peak-to-average power ratio (PAPR), which is a metric used to quantify the envelope excursion of the signal v(t) within a time interval τ:

[0009]

[0010] PAPR represents the ratio of the maximum instantaneous peak power of a signal to the average power within an observation period τ, usually expressed in dB. Due to the low PAPR, SC transmitters can drive their power amplifiers with low input back-off (IBO), which directly improves power efficiency. In contrast, high-speed filtering of symbols makes it difficult to compensate for the frequency selectivity of the transmission channel. Once the channel power delay profile significantly exceeds the symbol period duration, demodulation of SC signals requires fairly complex inverse filters with a large number of taps to eliminate inter-symbol interference (ISI). SC signals are widely used in satellite communications because the transmission channel is in most cases only flat in frequency and it is strongly recommended to operate the onboard high power amplifier (HPA) close to the saturation point (i.e., low IBO) to save power.

[0011] The second and by far the most popular modulation technique is nowadays Orthogonal Frequency Division Multiplexing (OFDM) modulation. The principle is quite simple: instead of modulating a single carrier at a high rate (the occupied bandwidth is defined by the baud rate), it is possible to modulate multiple adjacent subcarriers at a much lower rate, hereafter referred to as K adjacent subcarriers. To do so, even waveforms of finite length (and therefore with large support in the frequency domain) can be used, such as rectangular windows of duration T. As long as the baud rate of each individual subcarrier is much longer than the channel delay spread, the channel in the frequency domain looks just flat on each subcarrier. In other words, the effects of the channel can be easily compensated for each subcarrier by simple complex amplitude scaling. What makes OFDM modulation so attractive is the possibility to interleave the subcarriers every 1 / T Hz, since the contributions of all adjacent subcarriers cancel to zero at the output of a rectangular matched filter of length T seconds (AWGN channel assumption).

[0012] The second reason for the widespread use of OFDM modulation is its ease of implementation. It can be easily shown that a sampled version of the true analog OFDM signal can be obtained by simply computing the inverse discrete Fourier transform (IDFT) of the K samples to be sent after adding P empty subcarriers at both edges of the subcarrier multiplexing (here N = K + 2 × P). The IDFT generates only an oversampled version of the ideal analog OFDM signal, i.e., with aliasing. The more empty subcarriers are added, the better the quality of the sampling. In practice, to reduce the computational burden, the number of empty subcarriers is limited to allow only the first oversampling filtering to be applied before the digital-to-analog conversion without introducing too much in-band ripple. The spectral occupancy of OFDM modulation is indeed an inherent weakness of the modulation, and the occupied bandwidth is usually reduced using sharp analog filters. The OFDM signal is defined in baseband as:

[0013]

[0014] where Π(t) is a rectangular window that is equal to 1 in the time interval [0,T] and zero elsewhere. It can be seen that detection can be achieved independently on a symbol basis in the time domain (OFDM symbol level) and in the frequency domain (subcarrier level). The so-called subcarrier orthogonality brings another benefit: even for waveforms of finite length (such as rectangular waveforms), OFDM modulation achieves fairly good spectral efficiency, in fact similar to the one of SC modulation with zero roll-off waveform. In practice, the symbol period cannot be arbitrarily long, because for time-varying channels, it will become impossible to separate the subcarriers due to phase noise or Doppler spread. Instead, symbols are usually sent with a guard interval (GI) fed as a cyclic prefix (CP) to "swallow" the inter-symbol interference (ISI) between consecutive symbols. Due to the cyclic prefix, the linear convolution of the received signal with the channel impulse response appears as a circular convolution, which is the basic assumption for performing filtering in the DFT domain. It is as if the channel was applied in the DFT domain (assuming that the signal is periodic). Thus, a discrete channel can be accurately estimated in the frequency domain by simply dividing the received subcarriers by a reference subcarrier known in advance at the receiver (i.e., satisfying the duality - time convolution, frequency division). In the same way, the channel impact can also be retrieved in the frequency domain by simply dividing the received subcarriers by the estimated channel coefficients (one-tap equalization). The CP can also be used to mitigate the effects of jitter on time synchronization by simply positioning the demodulation FFT window to a few samples back into the CP.

[0015] As the sum of many independent signals, OFDM signals quickly appear as random Gaussian noise with considerable envelope fluctuations. To avoid any adverse effects on the receiver performance, a back-off of about 15dB to 17dB relative to the average power should be applied when saturating the amplitude of the OFDM signal. In this regard, OFDM is clearly less suitable than SC modulation when saving power by applying a low-power amplifier IBO. This is where DFT-spread-OFDM modulation builds a bridge between single-carrier (SC) and multi-carrier (MC) solutions. In DFT-spread-OFDM modulation, the constellation samples are first spread in frequency by DFT. After adding null subcarriers at the two band edges, the spread symbols are OFDM modulated to obtain a signal with the expected spectral shape. It is obvious that combining the DFT for spreading and the IDFT for modulation, the resulting signal is more or less a copy of the original samples. In fact, the resulting signal is just a sampled version of the original samples, which is filtered out using a Dirichlet waveform, a discrete time Fourier transform (DTFT) of a discrete rectangular window. This is in fact the principle of the well-known FFT or Fourier oversampling algorithm. From the above description, it can be seen that DFT-spread-OFDM modulation can be interpreted as an alternative to basic time-domain filtering to achieve the generation of SC signals. However, there is one major difference between the two methods: due to the cyclic nature of the DFT convolution, the first and last samples on each OFDM symbol are correlated over a number of N / K samples (the first significant lobe of the Dirichlet waveform).

[0016] For any OFDM signal, demodulation of DFT-spread-OFDM modulation can essentially be achieved in the frequency domain by a low-complexity one-tap equalizer. As can be seen, the receiver architecture is only the same as an OFDM receiver, but an additional IDFT is significantly required to recover the transmitted samples.

[0017] Reference symbols need to be defined to allow channel estimation for DFT-spread OFDM modulation.

[0018] As mentioned before, OFDM modulation is particularly well suited for frequency selective channels, since no matter what the selectivity of the channel is (as long as it remains constant for the symbol duration), the effects of the channel can be readily restored with the aid of a single one-tap equalizer on each subcarrier. A basic step is to first estimate the channel coefficients on all data subcarriers. The channel is typically estimated using reference symbols (RS), also known as pilots known at the receiver. In OFDM, reference symbols are typically generated by inserting samples known at the receiver on some or all of the modulated subcarriers. The basic method for estimating the channel in OFDM involves performing:

[0019] - Channel estimates of pilot locations,

[0020] - frequency domain smoothing, and

[0021] - Temporal interpolation.

[0022] In OFDM, the ratio data / pilot can be adjusted according to the channel properties in the frequency and time domain. For example, if the channel coherence bandwidth is very high but the channel changes rapidly, only a few pilots can be inserted in the frequency domain (1 pilot per γ subcarriers), but regularly spaced in time. It is even possible to change the position of the so-called scattered pilots from time to time, such as the PPx symbols in DVB-T2 systems. This is clearly one of the main advantages of OFDM modulation over SC modulation.

[0023] If the DFT-spread-OFDM modulation benefits from its OFDM spectrum for equalization, there are fewer cases where the pilot is inserted. In fact, the low-power fluctuations of its envelope are caused by the OFDM modulation of the DFT-spread symbols. Any arbitrary change to the extended subcarriers, such as by directly inserting the reference subcarrier, will destroy the PAPR characteristics of the signal (this is also the case in the case of insertion after puncturing of the extended samples). For this reason, the 3GPP / LTE uplink system specifies a full pilot, i.e., all subcarriers of the OFDM symbol carry reference pilots, which are directly inserted into the frequency domain as Zadoff-Chu sequences. The Zadoff-Chu sequence has a constant amplitude and remains the Zadoff-Chu sequence for certain parameter sets after the DFT. Therefore, the transmitted pilot maintains the good PAPR characteristics of the SC signal.

[0024] In order to reduce pilot overhead, the DVB-NGH system specifies a pilot called PP9, which combines data and reference information (hereinafter referred to as a mixed pilot symbol). The data is obtained by applying an extended DFT on a data block with a length equal to half the size of the conventional DFT. The extended data is then interleaved with the Zadoff-Chu sequence at one frequency per two subcarriers. For each component of the reference symbol (data and pilot), the resulting signal is simply an oversampled version of two consecutive copies of the original half-length sequence. As the sum of two SC signals, the resulting signal is no longer a pure SC signal. In DVB-NGH, the two signals are added with a time shift of half a sample duration to reduce envelope fluctuations. This results in a reference symbol with fairly good PAPR characteristics. Since the ratio of data and pilot samples is equal, the PP9 RS of DVB-NGH is particularly suitable for channels with small coherent bandwidths (i.e., severe frequency distortion).

[0025] Figure 1The principle of another pilot scheme called time division multiplexing pilot (TDMPi) is described. In this scheme, the reference symbol is modulated like any other DFT-spread-OFDM symbol. The sample vector to be sent is divided into two parts: the first part contains a given number of reference samples known at the receiver, and the second part contains the information data. The number of reference samples is usually selected according to the coherence bandwidth of the channel. The principle here is to generate symbols containing a predefined pattern for the receiver to use as short OFDM symbols. However, due to the circular convolution applied by DFT-spread-OFDM modulation, the modulated samples at the front of the reference symbol depend not only on the reference sample, but also on the adjacent samples on the right hand side of the reference sample and the cyclicity of the last sample in the vector.

[0026] To reduce data interference with the pilot, the TDMPi scheme modulates the pilot and information data separately, and the information data samples contribute zero to the pilot portion before adding the two modulation vectors (DFT-spread-OFDM modulation is linear). A key advantage of this approach is that the zeroing operation does not change the envelope of the generated signal. Therefore, regardless of the length of the front portion, the reference symbol maintains the good PAPR characteristics of DFT-spread-OFDM modulation. Another key feature of the TDMPi scheme is the insertion of a cyclic prefix that is independent of the entire symbol but is related to the front portion. In doing so, the receiver is able to estimate a channel from the front portion of the reference symbol that is very similar to a regular OFDM symbol including its CP.

[0027] This scheme is specifically designed for satellite transmission environments, where the channel is typically only flat and varies slowly over time. The basic principle is therefore to be able to make accurate channel estimates from a few reference symbols at the expense of a slightly degraded demodulation performance of the data carried by the reference symbols. The demodulation of these data suffers from two adverse effects: the destruction of the orthogonality of the OFDM modulation caused by the zeroing operation, and the lack of circularity due to the replacement of the global CP by a pilot CP (PCP) covering the previous sample as a small OFDM reference symbol. In fact, the degradation is particularly significant at the two edges of the data segment.

[0028] Another limitation of the TDMPi pilot scheme is related to the lack of cyclicity of the whole symbol (PCP instead of CP). If the demodulation FFT window is located a few samples later in the CP, the signal does not contain information for the last sample of the demodulated data segment. This is quite common in environments other than satellite transmissions to apply an offset in the CP of up to half its length to avoid synchronization problems and reduce some imperfections in the channel estimation due to the lack of knowledge of its exact power delay profile.

[0029] There is a need to define hybrid reference symbols (also called pilots) with low PAPR to allow channel estimation in DFT-spread-OFDM modulation. This modulation is mainly used in situations where power needs to be saved by driving power amplifiers close to saturation, such as satellite transmission. Therefore, it is necessary to rely on signals that show low power fluctuations, including reference symbols. The construction of reference symbols with low PAPR is more complex than for OFDM, because any change of the signal will significantly reduce its power envelope. In order to benefit from the low PAPR, existing solutions either exhibit a fixed and high ratio of pilot samples to data, or require arbitration between channel estimation and embedded data demodulation quality.

[0030] As mentioned earlier, TDMPi reference symbols are particularly suitable for environments such as satellite transmission, where the key requirement is to reduce the insertion of reference symbols to maximize power savings at the satellite while maintaining low PAPR. This is the main reason for replacing the regular CP with the pilot CP instead of building specific samples of a dedicated CP for the pilot pattern. A side effect is that if the FFT window is located a few samples behind in the CP, the receiver may lack information for the last sample in the demodulated symbol. In addition, unlike the global CP, there is no amplitude continuity at the connection between the end of the PCP and the beginning of the DFT-spread-OFDM symbol. This is a source of out-of-band (OOB) radiation, which can be reduced by filtering, but at the expense of reduced cyclicity of small OFDM symbols. One solution is to add reference samples at the end of the symbol, but at the expense of reduced spectral efficiency. Summary of the invention

[0031] The present invention aims to improve this situation.

[0032] To this end, the present invention relates to a method for transmitting at least one set of Kd data samples in a radio signal to be transmitted through a communication system, the radio signal being intended to be transmitted by a transmitter comprising at least one transmission module, the at least one transmission module being configured for transmitting on at least K different subcarriers, K and Kd being strictly positive integers, Kd being strictly smaller than K, the radio signal being provided by:

[0033] -Determine K samples (A0; ...; A K-1 ) in a first position J1 and a second position J2 in a block of K samples, such that a block of K samples comprises:

[0034] οKd data samples to be sent;

[0035] ο and K0 consecutive samples (A J1 ;...;A J2 );

[0036] οwhere J2-J1+1=K0 and Kd+K0=K;

[0037] - for a block of K samples A = (A0; ...; A K-1 ) applies a DFT-spread-OFDM type scheme and obtains N complex samples S n An output signal comprising symbols in a radio signal, wherein n=0 to N-1;

[0038] - Determine a first position T1 and a second position T2 in the output signal based on J1, J2, K and N, such that:

[0039] οT2-T1=R, R is a strictly positive integer;

[0040] - In the output signal, R complex samples (S T1 ;…;S T2 ) is set to zero to obtain the updated output signal;

[0041] - inserting S reference samples between the first position T1 and the second position T2 of the updated output signal, S is a strictly positive integer, and R=S+L, where L is a positive integer or 0, to obtain a subsequent signal representing a mixed symbol, comprising:

[0042] For each of n = [0; T1-1] ∪ [T2+1; N-1], the composite sample S in the radio signal n , the composite samples form a data segment of a mixed symbol;

[0043] For each of n=[T1+L; T2], the reference sample RS in the wireless signal n , the reference samples form a reference segment of the hybrid symbol;

[0044] - appending to the mixed symbol G samples forming at least part of a cyclic prefix of the mixed symbol to obtain a signal representing the mixed symbol appended with the cyclic prefix; considering transmitting a radio signal corresponding to the signal representing the mixed symbol appended with the cyclic prefix.

[0045] The communication system is understood to be a wireless communication system or a wired communication system. The transmission module is understood to be a transmission antenna when transmitting radio signals via a wireless communication system, or a wired transmission interface when transmitting radio signals via a wired communication system.

[0046] The data samples to be transmitted are understood as information samples to be included in the radio signal to be transmitted. A mixed symbol is understood as a symbol comprising data samples and reference samples. A reference sample is understood as a sample known at the receiver for demodulation purposes such as channel estimation and synchronization.

[0047] A DFT-spread-OFDM-type scheme is understood as a DFT-spread-OFDM type signal modulation. A DFT-spread-OFDM type scheme is applied to signal modulation according to any existing or future standard or to systems that rely on DFT-spread-OFDM modulation (DVB, 3GPP, etc.). Applying a DFT-spread-OFDM scheme can be interpreted as oversampling the original samples of a block of K samples, with a ratio of N / K. In other words, the modulated signal includes N / K times the original number of samples Kd. If N / K is an integer, the original Kd samples are found directly in the radio signal to be transmitted.

[0048] "G samples form a cyclic prefix of at least part of a mixed symbol" is generally understood to mean a substantially identical copy of the last part of the mixed symbol. Depending on the values ​​of G, T1 and T2, the cyclic prefix of the mixed symbol may include at least part of the data segment of the mixed symbol and / or at least part of the reference segment of the mixed symbol.

[0049] The original samples in the DFT-spread-OFDM data symbols are replaced by:

[0050] - zeroing the corresponding part of the modulated signal; and

[0051] - vector to insert reference samples,

[0052] This is equivalent to overwriting the contribution of the original samples. The indexes T1 and T2 identify the positions of these original samples.

[0053] Therefore, the NR data samples are arranged in a comb structure by applying a DFT-spread-OFDM type scheme, and the low power fluctuation of the single carrier modulation is retained even if the NR data samples are multiplexed with S reference samples to form a hybrid symbol as long as the reference samples are also generated to show low PAPR.

[0054] In one embodiment, inserting S reference samples between the first position T1 and the second position T2 of the updated output signal to obtain the subsequent signal further includes: inserting L cyclic prefix samples between the first position T1 and the second position T2, so that for each of n=[T1; T1+L-1], the subsequent signal will have cyclic prefix samples CPRS n Included in the radio signal, the cyclic prefix samples form a cyclic prefix of a reference segment of the mixed symbol.

[0055] As an example, the inserted S+L samples forming the reference segment of a hybrid symbol may correspond to a real OFDM pilot protected by its own cyclic prefix, in which case the S reference samples of such a hybrid symbol allow to benefit from the efficiency and robustness of the OFDM modulation for estimating the channel.

[0056] Furthermore, the structure of the mixed symbols in the radio signal to be transmitted allows the ratio of S reference samples to N total samples (or equivalently, the ratio of Kd data samples and K0 reference samples to the total number of samples K) to be adjusted after applying a DFT-spread-OFDM type scheme without changing the structure of the transmitted signal. It also allows the definition of specific positions where the S reference samples are inserted. In practice, depending on the requirements of the channel in terms of coherence bandwidth, it may be desirable to provide a radio signal with a predefined reference-data ratio. It may also be desirable that the reference sub-portion is located at a position known to the receiver, such as a default position, so that the receiver can, for example, demodulate only the reference sub-portion of the mixed symbol to perform channel estimation without demodulating the entire mixed symbol.

[0057] In order to provide a radio signal that matches these additional requirements, a first position J1 in the block of K samples may be determined such that a position T1 at which the S reference samples are inserted matches a target position, and a second position J2 in the block of K samples may be determined such that a ratio of the S reference samples to the K0 data samples matches a target ratio.

[0058] It should be noted that by applying a DFT-spread-OFDM type scheme to a block of K samples, it is necessary to generate S reference samples with similar spectral characteristics to the signal obtained from modulating the Kd data samples. In order to facilitate the estimation of the channel, it is recommended to generate it as an OFDM signal. With regard to the goal of generating reference symbols with low power envelope fluctuations, it is also preferred to generate S reference samples in order to avoid reducing the PAPR as much as possible. As mentioned before, OFDM modulation is not particularly suitable for generating low-PAPR signals. However, especially for the reference signal, the power fluctuations can be kept low by carefully selecting the samples to be modulated.

[0059] A typical but not exclusive example is the Zadoff-Chu sequence. Any other type of sequence and modulation scheme may be selected as long as it is able to estimate the channel and show low PAPR.

[0060] The advantages of inserting a true OFDM pilot with low PAPR and protected by its own cyclic prefix are:

[0061] - the radio signal to be transmitted exhibits low power fluctuations for both data samples and reference samples, which is particularly desirable in situations such as satellite transmission where it is necessary to save power by driving a power amplifier close to saturation,

[0062] - demodulation of the OFDM pilot at the receiver can benefit from an accurate estimate of the channel used to transmit the radio signal using any of several available OFDM channel estimation techniques,

[0063] -The hybrid symbol constructed in this way can include any proportion of reference samples while maintaining the same number of samples as any other symbol carried in the signal to be transmitted, which allows, for example, one or more hybrid symbols constructed in this way to be sent together with data symbols according to the LTE radio frame structure consisting of subframes composed of OFDM symbols (each OFDM symbol has the same length) in the time domain.

[0064] In one embodiment, predetermined reference samples are generated by applying OFDM modulation to a CAZAC sequence, which may more specifically be a Zadoff-Chu sequence. The result of an OFDM type scheme being applied to a CAZAC sequence (respectively a Zadoff-Chu sequence) with some specific parameters is an oversampled version of another CAZAC sequence or another Zadoff-Chu sequence. In addition to good correlation properties, such sequences in radio signals also have low envelope variations, i.e., low PAPR. Reference samples to be modulated may be constructed by applying an OFDM type scheme to these sequences, which are then inserted as reference sub-parts of mixed symbols with low computational complexity.

[0065] In one embodiment, the block of K samples includes a front portion of K1 consecutive samples with K1=J1 and a rear portion of K2 consecutive samples with K2=K-J2 and K1+K2=Kd. In other words, J1 and J2 are defined such that J1 is different from 0 and J2 is different from K-1. As a result, in the mixed symbol to be sent in the radio signal, the front portion and the rear portion are non-empty, that is, the first and last samples of the mixed symbol are data samples. Alternatively, K1 or K2 can be equal to 0, resulting in the first sample or the last sample of the mixed symbol being a reference sample.

[0066] Several embodiments are disclosed below. These embodiments differ in particular in the values ​​held by K0 consecutive samples between a first position J1 and a second position J2 in a block of K samples.

[0067] In one embodiment, the method further comprises setting the values ​​of K0 consecutive samples between the first position J1 and the second position J2 in the block of K samples to zero before applying a DFT-spread-OFDM type scheme thereto and obtaining an output signal.

[0068] Setting the values ​​of K0 consecutive samples to zero is a simple and cost-free way to obtain a suitable output signal in which the reference symbol will be introduced. However, after applying a DFT-spread-OFDM type scheme to a block of K samples including K0 consecutive samples set to zero in the middle (between the first position J1 and the second position J2), the resulting output signal includes non-zero symbols in the middle (between the moments corresponding to the first position J1 and corresponding to the second position J2) due to the inter-sample interference caused by the filtering effect of the modulation by applying the DFT-spread-OFDM type scheme. Setting these symbols to zero results in a modified signal that no longer has the characteristics of an OFDM type signal and, in particular, reduces the quality of the demodulation.

[0069] In one embodiment, at least a portion of K0 consecutive samples in a block of K samples is obtained by the following steps before applying a DFT-spread-OFDM type scheme to the block of K samples and obtaining an output signal:

[0070] (a) Get K samples (A'0; ...; A' K-1 ), each temporary block of K samples includes:

[0071] - Kd data samples to send, and

[0072] -K0 temporary continuous ghost samples (A' J1 ;...;A' J2 );

[0073] (b) for each temporary block of K samples, determining R complex samples S′ for each of n=[T1; T2] obtained between a first position T1 and a second position T2 as a result of applying a DFT-spread-OFDM type scheme to the temporary block of K samples n at least part of the value of ;

[0074] (c) for each temporary block of K samples, determining a criterion associated with the temporary block of K samples indicating a similarity between:

[0075] - the R composite samples S' n at least part of the value of

[0076] - Null value; and

[0077] (d) determining the K0 consecutive samples (A) of the block of K samples based on the value of each of the temporary blocks of K samples and each criterion determined in (c) J1 ;...;A J2) such that the R complex samples S for each of n=[T1; T2] obtained between a first position T1 and a second position T2 as a result of applying a DFT-spread-OFDM type scheme to a block of K samples n At least some of the values ​​of exhibit an optimized level of similarity to the null value.

[0078] It should be noted that, according to this embodiment, the K0 temporary consecutive samples (A') selected to obtain the K0 consecutive samples of the block of K samples are different from the Kd data samples carrying the useful data to be transmitted. J1 ;...;A' J2 ) are ghost samples, i.e., samples that are involved in generating the signal to be sent but are not themselves sent as samples (i.e., as samples that carry information).

[0079] Due to the linearity of DFT-spread-OFDM modulation, the criterion indicating the similarity level can be written as a linear system. In general, the system is overdetermined. Therefore, the least squares method can be applied to calculate the solution.

[0080] Throughout this document, a criterion may be any value associated with an algebraic or statistical method for evaluating the similarity between a first data set and a second data set. Here, the first data set corresponds to R composite samples S' n The standard may include, for example, L1 norm, Euclidean distance (L2 norm), etc. for algebraic methods, and may include, for statistical methods, the average value, standard deviation, weighted average value, etc. of R composite samples.

[0081] The sequence of (a), (b), (c), (d) above may be part of an algebraic or statistical method for determining the optimal value of K0 temporary consecutive samples. Examples of relevant statistical methods that can be used for this purpose include least squares, weighted least squares, χ 2 Law, etc...

[0082] Throughout this document, determine K0 consecutive samples of a block of K samples (A J1 ;...;A J2 ) may include solving a system of equations, such as a linear system of equations. For each potential solution (here K0 temporary consecutive ghost samples (A' J1 ;...;A' J2)), such a system of linear equations generally includes a linear equation, which is a relationship between the value associated with each sample of the potential solution and the value of a criterion associated with the potential solution. Such a system of linear equations generally also includes a linear equation, which is a relationship between the value of the solution to be determined and the value of the criterion associated with the solution to be determined, which is set to an optimized value (here corresponding to the value of perfect similarity of the solution determined with a null value).

[0083] Solving such a system of equations does not necessarily correspond to finding an exact solution, here K0 consecutive samples of a block of K samples (A J1 ;...;A J2 ) is an exact value of at least part of it.

[0084] In practice, the linear system is usually overdetermined. In this case, solving the linear system results in converting the K0 consecutive samples (A) of a block of K samples into J1 ;...;A J2 ) is at least partially determined as a result of an algebraic transformation.

[0085] If necessary, K0 consecutive samples can be determined by continuous iterations so that after modulation by applying a DFT-spread-OFDM type scheme to the block of K samples, the contribution of the K0 consecutive samples tends to offset the contribution of the Kd data samples, so that the S'n symbols for each of n=[T1;T2] are close to zero.

[0086] In fact, the contribution of the K0 samples thus determined at the position [T1, T2] is nearly opposite to the contribution of the data samples to the same part of the signal.

[0087] The main advantage of this approach is that setting the modulated data contribution to zero for all or part of the sample positions corresponding to the index of the RSn reference sample for each of n=[T1;T2] generates less noise, which may be detrimental to demodulation performance.

[0088] In one embodiment, determining at least the portion of K0 provisional consecutive samples comprises applying a least squares method.

[0089] Applying the least squares method allows setting a value for the K0 temporary consecutive samples that minimizes the Euclidean distance between the contribution of the K0 consecutive samples and the contribution of the modulated Kd data samples.

[0090] In one embodiment, K0 consecutive samples (A) in a block of K consecutive samples J1 ;...;A J2) are ghost samples of the ghost pilot type. Ghost samples of the ghost pilot type or ghost pilots are samples used to generate a signal, which is a portion between [T1, T2], simulating samples inserted after zeroing after modulation by applying a DFT-spread-OFDM type scheme thereto. The term ghost is related to the fact that their direct contribution to the pilot portion is smeared out. In some cases, it can be useful to know the ghost pilots at the receiver to compensate for their contribution to the data samples after demodulation. This is especially true when the ghost pilots are generated only from the samples S to be inserted and are not affected by the data samples.

[0091] In one embodiment, K0 consecutive samples (A) in a block of K consecutive samples J1 ;...;A J2 ) is determined by the following steps before applying a DFT-spread-OFDM type scheme to a block of K samples and obtaining an output signal:

[0092] - obtaining a reference signal comprising S reference samples to be inserted appended with L cyclic prefix samples to be inserted,

[0093] - applying an inverse DFT-spread-OFDM type scheme to the reference signal to obtain a demodulation reference signal comprising a block of K demodulation samples, the block of K demodulation samples comprising K0 consecutive demodulation samples,

[0094] - and inserting at least part of K0 consecutive demodulated samples into the block of K samples as ghost samples of a ghost pilot type forming at least part of the K0 consecutive samples.

[0095] The advantage of this approach is that after modulation, the Sn symbols for each of n=[T1;T2] simulate the S reference samples and L cyclic prefix samples to be inserted. The closer the Sn samples are to the S reference samples, the less degradation the insertion will apply.

[0096] The reference signal may be obtained by setting the value of any sample adjacent to the S reference samples to be inserted appended with the L cyclic prefix samples to be inserted to zero. In this case, the reference signal includes only the reference samples.

[0097] Alternatively, the reference signal may also include additional samples, adjacent to the S reference samples to be inserted with L cyclic prefix samples appended thereto, and having values ​​different from zero. An example of obtaining such a reference signal is:

[0098] - applying a DFT-spread-OFDM-type scheme to a block of K samples comprising Kd data samples to be transmitted and K0 consecutive samples having a value set to zero, and obtaining a signal representing a radio signal comprising N complex samples S n An output signal of the sign of, wherein n=0 to N-1;

[0099] - In the output signal thus obtained, R complex samples (S T1 ;…;S T2 ) is set to zero to obtain the updated output signal;

[0100] - inserting S reference samples and L cyclic prefix samples between the first position T1 and the second position T2 of the updated output signal thus obtained, to obtain a reference signal.

[0101] It should be noted that after modulation, the Sn symbol of each of n=[T1; T2] is slightly different from the S reference samples and L cyclic prefix samples to be inserted after modulation to form a mixed symbol.

[0102] In fact, by applying an inverse DFT-spread-OFDM type scheme to the reference signal, K demodulated samples can be obtained. These K samples contain all the information originally present in the reference signal. Among these K demodulated samples, only K0 consecutive demodulated samples are retained, resulting in information loss. When modulating a block of K samples containing Kd data samples and K0 consecutive demodulated samples, it is impossible to accurately obtain the Sn symbol originally present in the reference signal for each of n=[T1; T2]. However, setting the value of the Sn symbol for each of n=[T1; T2] to zero and then inserting S reference samples and L cyclic prefix samples results in a subsequent signal similar to the output signal and maintains similar characteristics in terms of demodulation performance and emission of out-of-band radiation. It should be noted that, in general, replacing the modulation symbol with the inserted OFDM symbol will result in amplitude discontinuity, thereby resulting in the emission of out-of-band radiation. In the above embodiment, since the sample to be modulated is determined so that the modulation symbol to be replaced is similar to the OFDM symbol to be inserted, the amplitude jump is minimized.

[0103] In an alternative embodiment, K0 consecutive samples (A) in a block of K consecutive samples J1 ;...;A J2 ) is determined before applying a DFT-spread-OFDM-type scheme to a block of K samples and obtaining an output signal by:

[0104] (a) Obtain K samples (A”0; ...; A” K-1), each temporary mixing block of K samples includes:

[0105] - Kd data samples to send, and

[0106] - K0 temporary consecutive ghost samples of ghost pilot type (A') J1 ;...;A” J2 );

[0107] (b) for each temporary block of K samples, determining R complex samples S" for each of n = [T1; T2] obtained between a first position T1 and a second position T2 as a result of applying a DFT-spread-OFDM-type scheme to the temporary mixed block of K samples n at least part of the value of ;

[0108] (c) for each temporary mixed block of K samples, determining a criterion associated with said temporary mixed block of said K samples indicating a level of similarity between:

[0109] - the R composite samples S" n at least part of the value of

[0110] - values ​​of at least part of the S reference samples to be inserted appended with the L cyclic prefix samples to be inserted; and

[0111] (d) determining the K0 consecutive samples (A) of the block of K samples based on the temporary mixed block of K samples and each criterion determined in (c) J1 ;...;A J2 ) such that the R complex samples S for each of n=[T1; T2] obtained between a first position T1 and a second position T2 as a result of applying a DFT-spread-OFDM type scheme to a block of K samples n At least part of the values ​​of exhibit an optimized level of similarity to at least part of the values ​​of the S reference samples to be inserted appended with the L cyclic prefix samples to be inserted.

[0112] Here, each temporary mixed block of K samples in (a) comprises K0 temporary consecutive ghost samples (A) of the ghost pilot type, according to any algebraic or statistical method such as that proposed above. J1 ;...;A” J2 ) as a candidate for the optimization problem solved in (d). The solution to the optimization problem is to form K0 consecutive samples (A) in a block of K samples. J1 ;...;A J2 ) of at least a portion of the ghost samples of the ghost pilot type.

[0113] Therefore, after modulation, the S reference samples and L cyclic prefix samples to be inserted are also simulated for each Sn symbol in n=[T1; T2].

[0114] In one embodiment, the method further comprises:

[0115] (e) Obtain K samples (A'0; ...; A' K-1 ), each temporary block of K samples includes:

[0116] - Kd data samples to send, and

[0117] -K0 temporary consecutive ghost samples (A' J1 ;...;A' J2 );

[0118] (f) for each temporary block of K samples, determining R complex samples S′ for each of n=[T1; T2] obtained between a first position T1 and a second position T2 as a result of applying a DFT-spread-OFDM-type scheme to the temporary block of K samples n at least part of the value of ;

[0119] (g) for each temporary block of K samples, determining a criterion associated with said temporary block of K samples indicating a level of similarity between:

[0120] - the R composite samples S' n at least part of the value of

[0121] - Null value; and

[0122] (h) determining at least a portion of K0 consecutive samples based on the value of each of said temporary blocks of K samples and each criterion determined in (g) such that R complex samples S for each of n=[T1; T2] obtained between a first position T1 and a second position T2 as a result of applying a DFT-spread-OFDM-type scheme to the block of K samples n At least some of the values ​​of exhibit an optimized level of similarity to the null value.

[0123] (i) updating K0 consecutive samples in the block of K samples (A) based on at least the portion of the K0 consecutive samples determined in (h); J1 ;...;A J2 ) at least part of.

[0124] Update K0 consecutive samples in a block of K samples (A J1 ;...;A J2) may correspond, for example, to K0 consecutive samples of a block of K samples (A J1 ;...;A J2 ) is the same as the K0 temporary consecutive ghost samples (A' J1 ;...;A' J2 ) are added together.

[0125] Thus, after modulation by applying a DFT-spread-OFDM type scheme to a block of K samples, the contribution of the Kd data samples is compensated by the contribution of the K0 consecutive samples thus modified. Thus, the L cyclic prefix samples and the S reference samples to be inserted are more accurately simulated for each Sn symbol in n=[T1; T2].

[0126] In one embodiment, updating the value of at least part of K0 consecutive samples of the block of K samples comprises:

[0127] - Determine a third position J1+Pa1 and a fourth position J2-Pa2 in the block of K samples such that K0 consecutive samples include:

[0128] οPa the front of 1 consecutive samples (A J1 ;...;A J1 +P a1 );

[0129] ο Middle (A J1+Pa1+1 ;...;A J2-Pa2-1 );

[0130] οPaThe rear part of 2 consecutive samples (A J2-Pa2 ;...;A J2 ),

[0131] Pa1 and Pa2 are positive integers, Pa1+Pa2 is strictly inferior to K0,

[0132] - and updating the sample values ​​of the front and back of K0 consecutive samples.

[0133] Therefore, only the front and rear of K0 consecutive samples are updated. This update is sufficient to reduce the amplitude discontinuity between the first and last samples of the pilot portion of the data segment and the mixed symbol. In particular, with respect to the signal characteristics of the transmission and demodulation performance of the out-of-band radiation that have been solved above, these signal characteristics can be optimized to a large range by updating only the front and rear of K0 consecutive samples. In fact, the effect of updating the middle of K0 consecutive samples on the characteristics is relatively small. For example, the transmission of out-of-band radiation is related to the amplitude discontinuity, which is related to replacing the samples of the front and rear of K0 consecutive samples with inserted samples (such as, OFDM samples).

[0134] In one embodiment, Pa1 = Pa2 results in a symmetric effect on the front and back of the K0 sample.

[0135] Data zeroing in the middle of the symbol after modulation and insertion of an OFDM pilot in the middle of the symbol to form the pilot portion of the mixed symbol may both result in degradation at the receiver. This degradation may be minimized by optimizing the block of K samples as a whole as described above, or - when the ghost samples are not determined based on the values ​​of the data samples - by precoding the Kd data samples individually according to the following embodiment. In this embodiment, K0 consecutive samples between the first position J1 and the second position J2 are ghost samples, and the values ​​of at least part of the Kd data samples are determined to compensate for the contribution of the K0 ghost samples to the data segment of the mixed symbol that occurs when demodulating the mixed symbol.

[0136] In another embodiment, a data segment of a mixed symbol obtained by applying a DFT-spread-OFDM-type scheme to a block of K samples including Kd precoded data samples and K0 consecutive ghost samples is obtained by applying the same DFT-spread-OFDM-type scheme to a block of K samples including Kd non-precoded data samples (A0; ...; A J1-1 ; A J2+1 ;...;A K-1 ) and K0 consecutive samples (A J1 ;...;A J2 ) values ​​set to zero are essentially the same.

[0137] Therefore, before modulation, the Kd data samples are precoded to compensate for the contribution to the K0 samples after modulation, so that in the mixed symbol, the data segment corresponds to the Kd data samples to be sent, and the precoding offsets the degradation introduced by applying a DFT-spread-OFDM type scheme.

[0138] In one embodiment, J1, J2, T1 and T2 are defined relative to each other, such as:

[0139]

[0140] and

[0141]

[0142] Wherein, each of w1 and w2 is a non-negative integer.

[0143] Thus, it is ensured that the Kd data samples to be transmitted are not erased after modulation during the setting of the value of the Sn symbol for each of n=[T1;T2] to zero. Furthermore, each positive increment of one of the values ​​of w1 or w2 provides an additional degree of freedom in determining the ghost pilots.

[0144] Another aspect of the invention relates to a method for determining a channel estimate of a propagation channel used for sending a radio signal transmitted in accordance with the method as described above, transmitted through the propagation channel and received by a receiver configured to be effective on at least K different subcarriers, K being a strictly positive integer, the method comprising the following steps:

[0145] - based on the received radio signal, obtaining a signal comprising N samples forming a mixed symbol appended with G samples, the G samples forming at least part of a cyclic prefix of the mixed symbol,

[0146] - obtain the first position T1 and the second position T2 in the signal, T2-T1=R, R is a strictly positive integer, R=S+L, L is a positive integer or 0, S is a strictly positive integer,

[0147] So that:

[0148] For each of n=[0; T1-1]∪[T2+1; N-1], the signal comprises a composite sample S n , the composite samples form a data segment of a mixed symbol;

[0149] For each of n=[T1+L; T2], the signal includes reference samples RS n , the reference samples form the reference segment of the mixed symbol;

[0150] - and determining a channel estimate of a propagation channel used for transmitting the received radio signal based at least on a reference segment of the mixed symbol.

[0151] The received signal is the observation of the transmitted signal after it has been sent over the propagation channel. After transmission, the received signal is subject to different degradations such as fading, Doppler shift, phase noise, additive noise, and inter-symbol interference (ISI) due to multipath delay spread.

[0152] A reference segment of mixed symbols in the received signal enables determination of a channel estimate, which itself can be used to mitigate the effects of ISI, for example using an equalizer.

[0153] The channel estimate thus determined can be used for demodulating the received radio signal, in particular for demodulating data segments of mixed symbols, and other data symbols only.

[0154] Determining the channel estimate based at least on the reference segment of the mixed symbols may be performed by any channel estimation method.

[0155] A simple channel estimation method comprises applying an inverse OFDM type scheme to mixed symbols to obtain an inverse OFDM type demodulated signal indicative of a channel estimate of the propagation channel used to transmit the received radio signal, smoothing and interpolating over the frequency of the demodulated signal to reduce noise indicative of multipath delay spread using, for example, a minimum mean square estimator (MMSE) using knowledge of transmitted reference samples, and determining a channel estimate based on the smoothed signal.

[0156] In an alternative embodiment, determining the channel estimate based at least on a reference segment of the mixed symbol comprises:

[0157] - The composite sample S n The values ​​of and the samples of the cyclic prefix forming at least part of the mixed symbol are set to zero to obtain a reference sample RS comprising S n The updated signal of

[0158] - applying an inverse DFT-spread-OFDM type scheme to the updated signal to obtain an inverse DFT-spread-OFDM type demodulated signal, a DFT-spread-OFDM type demodulated signal comprising a block of consecutive reference samples indicative of a channel estimate of a propagation channel used to transmit the received radio signal; and

[0159] -Determining a channel estimate based on the demodulated signal.

[0160] Another aspect of the invention relates to a telecommunication device comprising means for transmitting at least one set of Kd data samples in a radio signal to be transmitted by a communication system, the means comprising a transmitter configured to transmit on at least K different subcarriers, K and Kd being strictly positive integers, Kd being strictly less than K, the telecommunication device further comprising a processing circuit for generating said radio signal by:

[0161] -Determine K samples (A0; ...; A K-1 ) in a first position J1 and a second position J2 in a block of K samples, such that a block of K samples comprises:

[0162] ο Kd data samples to send;

[0163] ο and K0 consecutive samples (A J1 ;...;A J2 );

[0164] οwhere J2-J1+1=K0 and Kd+K0=K;

[0165] - for a block of K samples A = (A0; ...; A K-1 ) applies a DFT-spread-OFDM-type scheme and obtains a representation of the N complex samples S nAn output signal comprising symbols in a radio signal, wherein n=0 to N-1;

[0166] - Determine a first position T1 and a second position T2 in the output signal based on J1, J2, K and N, such that:

[0167] οT2-T1=R, R is a strictly positive integer;

[0168] - In the output signal, R complex samples (S T1 ;...;S T2 ) is set to zero to obtain the updated output signal;

[0169] - inserting S reference samples between the first position T1 and the second position T2 of the updated output signal, S is a strictly positive integer, and R=S+L, where L is a positive integer or 0, to obtain a subsequent signal representing a mixed symbol, comprising:

[0170] For each of n = [0; T1-1] ∪ [T2+1; N-1], the composite sample S in the radio signal n , the composite samples form a data segment of a mixed symbol;

[0171] For each of n=[T1+L; T2], the reference sample RS in the radio signal n , the reference samples form a reference segment of the hybrid symbol;

[0172] - appending at least part of the cyclic prefix of the mixed symbol to the mixed symbol to obtain a signal representing the mixed symbol to which the cyclic prefix is ​​appended; and

[0173] - generating a radio signal corresponding to a signal representing a mixed symbol to which a cyclic prefix is ​​appended, with a view to transmitting said radio signal.

[0174] Another aspect of the invention relates to a telecommunication device comprising a module for receiving at least one set of Kd data samples in a radio signal to be transmitted by a communication system, the module being configured to operate on at least K different subcarriers, K and Kd being strictly positive integers, Kd being strictly less than K, the telecommunication device further comprising a processing circuit for determining a channel estimate of a propagation channel for transmitting the radio signal to the telecommunication device, wherein it is received as a received radio signal by:

[0175] - based on the received radio signal, obtaining a signal comprising N samples forming a mixed symbol appended with G samples, the G samples forming at least part of a cyclic prefix of the mixed symbol, wherein N and G are strictly positive integers,

[0176] - obtaining a first position T1 and a second position T2 in the signal, T2-T1=R, R is a strictly positive integer, and R=S+L, L is a positive integer or 0, and S is a strictly positive integer,

[0177] So that:

[0178] For each of n=[0; T1-1]∪[T2+1; N-1], the signal comprises a composite sample S n , the composite samples form a data segment of a mixed symbol;

[0179] For each of n=[T1+L; T2], the signal includes reference samples RS n , the reference samples form the reference segment of the mixed symbol;

[0180] - The composite sample S n The values ​​of and the samples of the cyclic prefix forming at least part of the mixed symbol are set to zero to obtain a reference sample RS comprising S n The updated signal of

[0181] - applying an inverse OFDM-type scheme to the updated signal to obtain a demodulated signal comprising a block of consecutive reference samples indicative of a channel estimate of the propagation channel used for transmitting the radio signal.

[0182] In a non-limiting embodiment, such a transmitting device is Figure 3 Shown in.

[0183] refer to Figure 3 , a transmitter 3.1 is shown transmitting a radio signal to a receiver 3.2. The receiver 3.2 is within the coverage of the transmitter 3.1. In the context of an OFDM-based system, this transmission is a DFT-spread-OFDM type transmission. In this example, the transmitter 3.1 is a fixed station and the receiver 3.2 is a mobile terminal, which in the context of a 3GPP cellular system would be named a base station and a user equipment. The transmitter 3.1 could also be a mobile terminal and the receiver 3.2 could also be a fixed station.

[0184] The transmitter 3.1 comprises a communication module (COM_trans) 3.3, a processing module (PROC_trans) 3.4 and a storage unit (MEMO_trans) 3.5. MEMO_trans 3.5 comprises a non-volatile unit for retrieving a computer program and a volatile unit for retrieving symbol merging parameters. PROC_trans 3.4 is configured to transmit mixed symbols according to the invention. COM_trans 3.3 is configured to transmit a radio signal to a receiver 3.2. As previously mentioned, the communication module 3.3, the processing module 3.4 and the storage unit 3.5 may constitute a device for transmitting mixed symbols.

[0185] The receiver 3.2 comprises a communication module (COM_recei) 3.6, a processing module (PROC_recei) 3.7 and a storage unit (MEMO_recei) 3.8. MEMO_recei 3.8 comprises a non-volatile unit for retrieving a computer program. PROC_recei 3.7 is configured to retrieve symbols from a radio signal. COM_recei 3.6 is configured to receive a radio signal from the transmitter 3.1.

[0186] Another aspect of the invention relates to a computer program product comprising code instructions, which, when executed by a processor, perform the method as described above.

[0187] In a non-limiting embodiment, the general algorithm of such a computer program product is Figure 2 As described in detail below.

[0188] The present invention is illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings in which like references refer to similar elements. BRIEF DESCRIPTION OF THE DRAWINGS

[0189] [ Figure 1 ]

[0190] Figure 1 The generation of TDMPi reference symbols is shown.

[0191] [ Figure 2 ]

[0192] Figure 2 A block diagram indicating a general algorithm of a procedure for executing a method according to an embodiment of the present invention.

[0193] [ Figure 3 ]

[0194] Figure 3 A telecommunication device according to an embodiment of the invention is shown.

[0195] [ Figure 4 ]

[0196] Figure 4 A flowchart illustrating generation of a hybrid symbol according to a first embodiment of the present invention is indicated.

[0197] [ Figure 5 ]

[0198] Figure 5 A flow chart illustrating generation of a hybrid symbol according to a second embodiment of the present invention is indicated. DETAILED DESCRIPTION

[0199] refer to Figure 2, we get K samples (A0, ...A K-1 ) block 2.1. A first position J1 and a second position J2 in the sample block are also obtained, so that a block of K samples includes Kd data samples to be sent and K0 consecutive samples not to be sent (A J1 ;...;A J2 ).

[0200] A K-size Discrete Fourier Transform (DFT) module 2.2 is applied to blocks of K samples to obtain K complex samples in the frequency domain. That is, one complex sample is obtained for each l-th subcarrier of the allocated K subcarriers.

[0201] These complex samples are mapped to K of the N inputs of the N-size IDFT module 2.4 by the subcarrier mapping module (KxN) 2.3 in the frequency domain. With regard to subcarrier mapping, the vector of complex samples The subcarrier mapping module 2.3 maps the K allocated subcarriers of the N existing subcarriers. The subcarrier mapping may be, for example, local, i.e., the K elements of the vector S are mapped to K consecutive subcarriers of the N existing subcarriers. For example, the subcarrier mapping may be distributed, i.e., the K elements of the vector S are mapped equidistantly over the entire bandwidth, with zeros occupying unused subcarriers.

[0202] The N-dimensional inverse DFT module 2.4 is then applied to the resulting vector of subcarrier mapping module 2.3 This generates the DFT-spread OFDM-symbols that are transmitted via the transmit antenna 2.0. More precisely, at the output of the IDFT module 2.4, the signal The signal occupies K allocated subcarriers out of the N existing subcarriers during the time interval corresponding to the DFT-spread-OFDM-symbol. is a time domain signal whose frequency domain representation is a composite sample S for each lth occupied subcarrier in the time interval l , and l = 0 to M-1. The time domain signal corresponds to a DFT-spread OFDM-symbol. Therefore, the signal The samples in refer to the samples in a DFT-spread-OFDM symbol.

[0203] At the output of IDFT module 2.4, an output signal is obtained, which represents the N complex samples S n A DFT-spread OFDM symbol is included in a radio signal, where n=0 to N-1.

[0204] refer to Figure 3, the proposed hybrid symbol is obtained by inserting OFDM reference symbols within the DFT-spread-OFDM symbols of the output signal.

[0205] The advantage of inserting a true OFDM reference symbol including its own CP benefits from an accurate estimate of the channel using any of a number of available OFDM channel estimation techniques. It is assumed here that the DFT-spread-OFDM symbol carries K time domain samples, which are oversampled by a factor of N / K, adding P empty subcarriers (N=2P+K) on both edges of the multiplexing. The size of the OFDM pilot symbols inserted in the DFT-spread-OFDM-reference symbol is defined according to the requirements of the channel in terms of coherence bandwidth. The OFDM pilot can be inserted into any position within the DFT-spread-OFDM-reference symbol. The OFDM pilot is inserted by first zeroing the corresponding samples in the previously generated DFT-spread-OFDM reference symbol.

[0206] To this end, the output signal is sent to a data zeroing module 2.5 which has access to the values ​​of a first position T1 in the output signal and a second position T2 in the output signal, the values ​​of T1 and T2 being determined based on J1, J2, K and N. The output signal thus comprises N complex samples forming:

[0207] - the front of N1 composite samples (S0; ...; S T1-1 ),

[0208] -The middle of R composite samples (S T1 ;...;S T2 ),

[0209] - and the rear of N2 composite samples (S T2+1 ;...;S N-1 ).

[0210] At the output of the data zeroing module 2.5, the values ​​of the R complex samples in the middle of the output signal are set to zero, thereby generating an updated output signal.

[0211] The updated output signal is sent to the pilot merging module 2.6, which merges the OFDM pilots appended with their own cyclic prefix. The pilot merging module 2.6 has access to the first position T1 and the second position T2, so that the OFDM pilot and the cyclic prefix are inserted from the first position T1 to the second position T2. ​​At the output of the pilot merging module 2.6, a subsequent signal representing the mixed symbol is obtained.

[0212] The cyclic prefix appending module 2.7 appends the cyclic prefix of the mixed symbol to the mixed symbol, and then the signal can be converted using the digital-to-analog converter module (DAC / RF) 2.8 to generate a radio signal to be transmitted by the antenna 2.0. In particular, due to the propagation conditions of electromagnetic signals and current telecommunication regulations, the digitally processed signal cannot be transmitted in the air. The DAC / RF module overcomes this disadvantage by performing frequency conversion on the signal.

[0213] The generation of the proposed hybrid symbol will be further described in detail in the following embodiments.

[0214] As mentioned earlier, TDMPi reference symbols are particularly suitable for environments such as satellite transmission, where the key requirement is to minimize the insertion of reference symbols to maximize power savings at the satellite while maintaining a low PAPR. This is the main reason for replacing the regular CP with a pilot CP instead of investing in specific samples to build a dedicated CP for the pilot pattern. A side effect is that if the FFT demodulation window is located a few samples behind in the CP, the receiver may lack information for the last few samples in the demodulated symbol. In addition, unlike the global CP, there is no amplitude continuity at the connection between the end of the PCP and the beginning of the DFT-spread-OFDM symbol. This is a source of out-of-band (OOB) radiation, which can be reduced by filtering, but at the expense of degradation of the cyclicity of small OFDM symbols. One solution is to add reference samples at the end of the symbol, but at the expense of additional performance degradation.

[0215] The proposed hybrid symbol compensates for the limitations of the TDMPi scheme. In one embodiment, the basic principle of the proposed hybrid symbol is to insert a real OFDM symbol within a DFT-expanded OFDM-symbol. The advantage of inserting a real OFDM reference symbol including its own CP benefits from the accurate estimation of the channel using any one of a plurality of available OFDM channel estimation techniques. It is assumed here that the DFT-expanded OFDM symbol carries K time domain samples, which are oversampled by a factor of N / K, and P empty subcarriers are added at the two edges of the multiplexing. The size of the OFDM pilot symbol inserted in the DFT-expanded OFDM-reference symbol is defined according to the requirements of the channel in terms of coherent bandwidth. The OFDM pilot can be inserted into any position within the DFT-expanded OFDM-reference symbol. The OFDM pilot is implicitly inserted by first zeroing the corresponding samples in the previously generated DFT-expanded OFDM reference symbol.

[0216] Obviously, the erasure of original samples carrying information data degrades their demodulation.

[0217] refer to Figure 4In one embodiment, a solution is to generate the DFT-spread-OFDM-reference symbols with a vector containing zeros at the index corresponding to the position of the OFDM pilots. After modulation, due to the circular convolution with the Dirichlet kernel, the part of the signal corresponding to the position of the previously modulated zeros will contain a contribution from the data adjacent to the zeros. This contribution is first zeroed out to be able to insert the samples of the OFDM pilot symbols.

[0218] As mentioned before, the portion of the signal corresponding to the location of the previously modulated zero contains contributions from the data adjacent to the zero. Applying the zeroing operation to these samples naturally destroys the orthogonality of the underlying OFDM signal, resulting in some degradation of the demodulated data, especially for those close to zero samples.

[0219] refer to Figure 5 In one embodiment, it is proposed to insert pilot samples before the modulation calculation to minimize the contribution of the modulated data to the middle part. Since the contribution of these pilot samples will be mostly offset by the zeroing operation, they are described in this article as so-called ghost pilot type ghost samples or ghost pilots or ghost pilot samples.

[0220] The OFDM pilot consists of S samples extended with a CP of length L, resulting in a total number of samples R = S + L. The OFDM pilot is assumed to be inserted at index T1 of a DFT-spread-OFDM reference symbol of length N, assumed to be numbered [0, N-1]. The last sample of the OFDM pilot is therefore inserted at index T2 = T1 + R-1. The indices of the first and last samples carrying pilot samples before modulation are defined as:

[0221]

[0222] and

[0223]

[0224] Where floor() is a function that takes a real number x as input and gives as output the largest integer less than or equal to x, ceil() is a function that gives as output the smallest integer greater than or equal to x, and w1 and w2 are positive integers, usually small (from 0 to 1 or 2) to limit the loss of spectral efficiency. Note that in cases where the first or last sample of the inserted OFDM pilot is very close to the data location, it is best to set wi to at least 1. As a result, the original vector is divided into three parts:

[0225] - Front: K1 = J1 information data, index [0, J1-1]

[0226] - Middle: K0 = J2-J1+1 pilot samples, index [J1, J2]

[0227] - and the back part: K2 = K-K1-K0 information data, index [J2+1, K-1]

[0228] The total number of data of the front and rear executions in the original vector is represented as Kd=K1+K2.

[0229] The samples carried in the front and back parts are represented as information or user or useful data. After DFT-spread-OFDM modulation, the middle part depends mainly on the inserted pilot samples, but also contains the contribution of data due to the circular convolution of the front and back samples with the Dirichlet kernel. Before inserting the OFDM pilot, the signal is decomposed into three parts:

[0230] - Front: N1=T1 data samples remain unchanged, index [0, T1-1]

[0231] - Middle: N0 = R samples forced to zero, index [T1, T2]

[0232] - and the latter part: N2=N-N1-N0 data samples remain unchanged, indexed [T2+1,N-1].

[0233] Finally, an OFDM pilot of length R, generated according to any method that guarantees low PAPR, is inserted into the positions occupied by the R pilots. A typical but non-limiting solution is the so-called Zadoff-Chu sequence used in the uplink of 3GPP / LTE cellular systems. The total number of data performed in the front and rear parts of the modulated signal is denoted as Nd=N1+N2.

[0234] The proposed algorithm is best presented using a matrix description of DFT-spread-OFDM modulation. In the first step, both modulation and demodulation are described as matrix transforms to serve as a reference for the rest of the description. k represents a data (information) sample that will be parsed into a data block x of size K. Therefore, the i-th data block x (i) can be written as:

[0235] The contribution of the data to the overall signal is detailed below. Data block x (i) First, the K-point normalized direct form discrete Fourier transform (DFT) is expanded in frequency. The discrete Fourier transform (DFT) can be used: A = P c F K can be described as a matrix product, where P c is a (K, K) permutation matrix that swaps half of each column of the matrix to which it is applied, where F Kis the element at row k and column n K-point normalized discrete Fourier transform (DFT) in the form of a (K, K) matrix, where k, n = 0...K-1, ω K = exp(j2π / K) is the primitive root of unity. After expansion and adding null subcarriers, use the matrix The samples are modulated by OFDM, where P r is an (N,N) permutation matrix to which the swap is applied to half of each row of the matrix, and where, is the N-point normalized inverse discrete Fourier transform (IDFT). In this matrix, only the contribution of the extended data is retained, and the null subcarriers are ignored:

[0236] Bm=B(:;ModSubCarrierMask),

[0237] Among them, ModSubCarrierMask corresponds to the index vector of length N of the K modulated subcarriers in the multiplexing.

[0238] Finally, the generated signal can be calculated using the (N, K) matrix: C = Bm A.

[0239] The generated signal is given by: (i) =Cx (i) .

[0240] For simplicity, references to the rank (i) of the symbol are ignored in the rest of the document. At the receiver, the matrix Perform OFDM demodulation on the signal. In this matrix, only the contribution of the modulated subcarrier is retained:

[0241] Dm=D(ModSubCarrierMask,:).

[0242] Assume perfect channel estimation, with ideal zero-forcing equalization and no additive noise. (Extension to more correlated equalization is straightforward.) After equalization using the results of the channel estimate, the demodulated samples are given by the matrix Despread.

[0243] In general, the samples are demodulated by a (K, N) matrix: F = E Dm.

[0244] The contribution of the zeroing operation to the demodulated useful data will be described later. Again, perfect channel estimation is assumed, with ideal zero-forcing equalization and no added noise. The useful data contribution to the front and back of the generated signal is calculated. Considering the useful data samples and ghost pilots, the spreading operation calculated using the matrix A as described above is first considered. After spreading and adding null subcarriers, the samples are OFDM modulated using the matrix Bm, where only the front and back samples are retained:

[0245] Bdm=B(DataPosMaskAfter, ModSubCarrierMask),

[0246] where DataPosMaskAfter corresponds to an index vector of length N of the Nd useful front and back parts of the generated signal. In general, the contributions of the original vectors to the front and back parts of the generated signal are obtained from the (Nd, K) matrix Cd=Bdm A.

[0247] At the receiver, the signal is first OFDM-demodulated using the matrix Dm, where only the front and rear contributions are retained (the middle is zeroed):

[0248] Dmd=D(ModSubCarrierMask, DataPosMaskAfter).

[0249] After equalization using the results of the channel estimation, the demodulated samples are despread using the matrix E, where only the results of the rear and front parts of the demodulated vector are retained:

[0250] Ed=E(DataPosMaskBefore,:).

[0251] Demodulation is therefore performed using the following matrix: Fz = Ed Dmd.

[0252] In summary, the result of demodulating the useful data using the zeroing application is therefore calculated using the (Kd, K) matrix Gz=Fz Cd.

[0253] In one embodiment, the generation of ghost pilots for reducing the data contribution to the middle portion is detailed hereinafter. The goal is to calculate a vector of pilot samples that cancels or at least reduces the data contribution to the middle portion.

[0254] s k Specifies the pilot samples of the short OFDM symbol inserted at the position of the zero samples after modulation. The short OFDM pilot is described as the following vector with length R: s = [s0, ..., s R-1 ] T .

[0255] p kSpecify the ghost pilot samples that appear in the middle of the vector x before modulation. The ghost pilot is described as the following vector of length K0: p = [p0, ..., p K0-1 ] T .

[0256] Finally, for convenience, the information dataset is denoted as d k and grouped into a single vector of length Kd: d = [d0, ..., d Kd-1 ] T .

[0257] The contribution of the useful data to the middle part of the generated signal will be described later. The extended operation calculated using the matrix A takes into account only the useful data:

[0258] Ad=A(:,DataPosMaskBefore),

[0259] Among them, DataPosMaskBefore corresponds to the index vector of length K of Kd useful data in data block x. After extension and adding null subcarriers, the samples are OFDM modulated using matrix Bm, where only the samples in the middle are retained:

[0260] Bpm=B(PilotPosMaskAfter, ModSubCarrierMask),

[0261] Where PilotPosMaskAfter corresponds to the index vector of the middle R samples in the generated signal. In general, the contributions of the front and back of the original vector to the middle of the generated signal are obtained from the following (R, Kd) matrix: Cpd = Bpm Ad.

[0262] Similarly, the contribution of the ghost pilots to the middle of the generated signal is calculated. The extended operation calculated using matrix A only considers the ghost pilots appearing in the middle:

[0263] Ap=A(:, PilotPosMaskBefore),

[0264] where PilotPosMaskBefore corresponds to a vector of length K0 indices of the ghost pilot samples within the data block x. After extension and addition of null subcarriers, the samples are OFDM modulated using the matrix Bpm. In general, the contribution of the ghost pilots of the original vector to the middle of the generated signal is obtained from the (R, K0) matrix: Cpp = BpmAp. The set of ghost pilot samples is obtained as the solution to the following linear system: Cpp p = -Cpd d.

[0265] Since the (N0, K0) matrix C ppis full rank and has more rows than columns, so the corresponding system has no exact solution. In this case, the least squares solution is usually computed:

[0266] Experiments show that the error is usually small, so the vector obtained reduces the impact of the zeroing operation on the decoded data. It is also possible to reduce the number of points where the data contribution is forced to be cancelled, as well as the number of ghost pilots used to achieve this goal. Without loss of generality, it is assumed that the ghost pilots are located on both sides of the middle with the same number of samples Kb. Similarly, the samples to be cancelled are located on both sides of the middle with an equal number of samples Ka. Two situations can be met:

[0267] -Ka ≥ Kb: Again, the system is multi-factor dependent and only a least squares solution can be computed. The advantage of this approach is the reduced computational cost. In addition to the residuals, the demodulation of the data may be degraded due to the application of zeroing to the remaining data contributions in the middle (and no ghost samples are used to reduce this). This effect is usually small because the data contribution amplitude decreases as one moves away from the preceding and following data segments.

[0268] -Ka≤Kb: In this case, the system is underdetermined and multiple solutions exist. The solution with the least squares magnitude is usually chosen. In this case, zeroing can be strictly performed. However, since the system may have multiple degrees of freedom, the system may result in solutions where the ghost pilot samples located in the middle of the middle show very large magnitudes. This will lead to PAPR degradation. Then, an alternative approach is to solve the system by constraining the magnitude of the pilots, but at the expense of the residual.

[0269] In order to obtain a significant reduction in data contribution, it is preferred to benefit from at least one pilot on each side of the middle before modulation, without falling in the middle of the generated signal. To this end, the indices of the first and last samples carrying pilot samples are defined as follows:

[0270]

[0271]

[0272] should be chosen so that and is at least equal to 1. Better results can be obtained by using 2 pilots on each side, i.e. and is at least equal to 2. However, the solution found may show high amplitudes specifically for these samples. Again, an alternative is to solve the system while limiting the amplitude of the pilot.

[0273] The application of ghost pilots to eliminate the data contribution to all or part of the middle of the generated signal can only come from a trade-off. In fact, when the data contribution is reduced by inserting pilots in the middle, those pilots can easily extend outside the middle, that is, to the data segment. This is not a problem for demodulation as long as the overall contribution of the pilots remains unchanged. But in this case, the zeroing operation breaks the orthogonality of the ghost pilot components. As a result, the demodulated samples may be degraded by the pilot contribution. This is especially true when the system is uncertain, resulting in ghost samples with large amplitudes. It should also be noted that the demodulated data is also degraded by the insertion of short OFDM pilots.

[0274] refer to Figure 5 In one embodiment, it is proposed to insert ghost pilot samples in order to generate a signal very similar to a short OFDM symbol inserted after zeroing.

[0275] The two main advantages of this approach are:

[0276] -Since the signal generated by using ghost pilots is very similar to OFDM pilot symbols, the disturbance caused by zeroing fully or partially compensates for the error caused by inserting OFDM pilots after zeroing;

[0277] - and the introduction of ghost pilots to obtain a signal very similar to the inserted OFDM signal significantly reduces the amplitude discontinuity after zeroing and insertion of the OFDM pilot, thereby reducing OOB radiation.

[0278] The set of phantom pilot samples simulating OFDM pilot symbols is obtained as the solution to the following linear system: Cpp p = s - Cpd d.

[0279] It can be noticed in the linear system above that the calculated ghost pilots are similar to the inserted samples after retrieving the data contribution to the same samples. This is indeed the overall result of modulating the data and pilot samples, which builds the shape of the samples that will be replaced by the short OFDM pilots.

[0280] Since the (N0, K0) matrix C pp is full rank and has more rows than columns, so the corresponding system has no exact solution. In this case, the least squares solution is usually computed:

[0281] As mentioned before, it is important to benefit from at least one pilot on each side of the middle before modulation, rather than falling in the middle of the generated signal, in order to perform mathematical operations significantly on the inserted samples. However, it may appear that the solution found can show high amplitudes specifically for these two samples. Then, an alternative is to solve the system by limiting the amplitude of the pilots.

[0282] In one embodiment, ghost samples are simply generated as a result of demodulating inserted OFDM symbols or in combination with data samples.

[0283] In an alternative embodiment, the initial value of the ghost pilot samples is set to a vector calculated according to the solution mentioned in the previous embodiment. Then the samples on both sides of the pilot are updated to further reduce OOB.

[0284] The zeroing operation introduces performance degradation at the receiver due to the change in OFDM orthogonality. In another embodiment of the proposed solution, the set of original data samples is updated before modulation in order to minimize the degradation at the receiver. This updating is called data precoding.

[0285] In the first case, it is assumed that the pilot samples are simply set to zero. Before actually generating the DFT-spread-OFDM symbol, the transmitter updates the data vector in order to compensate for the effect of zeroing and the contribution of the pilot samples at the receiver. Thereafter the contribution of the pilot samples to the useful data is calculated. At the receiver, the signal is first OFDM-demodulated using the matrix Dm, where only the contribution of the middle part is retained.

[0286] Dmp=D(ModSubCarrierMask, PilotPoSMaSkAfter)

[0287] Where PilotPosMaskAfter corresponds to an index vector of length N of the R pilot samples in the generated signal. After equalization, the demodulated samples are despread using the matrix Ed, where only the results of the rear and front parts of the demodulated vector are retained. Therefore, the demodulation results of the OFDM pilot pairs of the useful data are calculated using the (Kd, R) matrix Fp = EdDmp. The precoded data set is obtained as the solution of the following linear system: Gz d updated =d original -Fps where d updated , where d updated is the information dataset after applying precoding, d original It is the original information dataset.

[0288] In the above linear system, a new data vector is calculated to compensate for the degradation caused by the zeroing operation and the degradation caused by the pilot insertion. This component can also be retrieved at the receiver. Since the matrix Gz is full rank, any suitable algorithm can be used to calculate the updated sample vector.

[0289] In another embodiment, the size of the vector of updated data may be reduced to reduce complexity at the expense of a small performance degradation.

[0290] The data vector can be updated to compensate for the effect of the ghost pilots and the insertion of the OFDM pilots. In this case, the system is almost similar to the previous one, except that the contribution of the ghost pilots to the decoded data calculated using the matrix Gzp=Fz Cdp needs to be added to the component to be compensated, where Cdp=Bdm Ap.

[0291] The new data set is obtained as the solution of the following linear system: Gz d updated =d original -Fp s-Gzp p.

[0292] In one embodiment, the calculation of the phantom pilots and precoded data are implemented jointly for better performance, but at the expense of increased complexity.

[0293] Assume that Pa ghost samples are inserted on both sides of the middle before modulation (a total of 2×Pa samples). The ghost samples are calculated to achieve equality with the inserted OFDM pilot for Pb samples on both sides of the middle after modulation (a total of 2×Pb samples). This operation can be performed alone or in combination with compensation for zeroing and pilot insertion. Here, the algorithm is described in this more general case.

[0294] First, consider the expansion operation calculated using the (K, Kd+2×Pa) matrix extracted from A, taking into account the useful data and ghost samples:

[0295] Ade=A(:, ExtendedDataPosMaskBefore),

[0296] Where ExtendedDataPosMaskBefore corresponds to a length K index vector of Kd useful data enhanced with 2×Pa ghost samples. After extension and adding null subcarriers, the samples are OFDM modulated using the (Nd, K) matrix extracted from Bdm. In general, the contributions of the front and back parts of the ghost sample enhancement of the original vector to the front and back of the generated signal are obtained from the (Nd, Kd+2×Pa) matrix Cdde=Bdm Ade.

[0297] At the receiver, the signal is first OFDM demodulated using the matrix Dmd. After equalization using the result of the channel estimation (assuming perfect channel estimation, with ideal zero-forcing equalization and no added noise), the demodulated samples are despread using the (K, K) matrix Ed. Demodulation is therefore performed using the (K, Nd)Fz matrix.

[0298] In general, the (Kd, Kd+2×Pa) matrix Gze=Fz Cdde is used to calculate the result of demodulating the useful data enhanced with ghost samples taking into account the use of zeroing application.

[0299] To ensure equal amplitudes, we first need to calculate the contribution of the useful data plus the ghost samples to the Pb samples on both sides of the modulated middle. The starting point is the matrix Ade, which performs the expansion operation taking into account the useful data enhanced with ghost samples. After expansion and adding null subcarriers, the samples are OFDM modulated using the matrix Bm, where only the samples on the two edges of the middle are kept:

[0300] Bpm=B(GhostPosMaskAfter, ModSubCarrierMask)

[0301] Among them, GhostPosMaskAfter corresponds to the index vector of length N of 2×Pb samples selected on the middle edge after modulation, which is guaranteed to be equal to the corresponding samples of the inserted OFDM pilot.

[0302] In general, the contributions of the front and rear parts of the original vector enhanced with 2×Pa ghost samples to the modulated middle edge are obtained from the (2×Pb, Kd+2×Pa) matrix Cedp=BpmAde.

[0303] s - Denotes a vector of pilot samples that we wish to obtain that are equal to the corresponding samples obtained after modulating the updated data and 2×Pa ghost samples:

[0304] s - =[s0,...,s Pb-1 ,s R-Pb-1 , ..., s R-1 ] T

[0305] x + A vector representing useful data augmented with 2×Pa ghost samples:

[0306] d + =[d0, ..., d J1-1 , p0, ..., p Pa-1 , p Pa , ..., p K0-Pa+1 , d J1 , ..., d Ka-1 ] T .

[0307] The updated data and ghost samples are obtained together as a solution to the following (Kd+2×Pb, Kd+2×Pa) system:

[0308]

[0309] To ensure that there is at least one solution, it is required to select Pb ≤ Pa. Just like the data update operation, the amount of updated data can be reduced to reduce the computational cost, but at the cost of performance degradation.

Claims

1. A method for transmitting at least one set of Kd data samples in a radio signal to be transmitted through a communication system, the radio signal being intended to be transmitted by a transmitter comprising at least one transmission module, the at least one transmission module being configured to transmit on at least K different subcarriers, K and Kd being strictly positive integers, Kd being strictly less than K, the radio signal being provided by the following steps: -Determine K samples (A0; ...; A K-1 ) in a first position J1 and a second position J2 in a block of K samples, such that a block of K samples comprises: ο Kd data samples to send; ο and K0 consecutive samples (A J1 ;... ; A J2 ); ο and J2-J1+1=K0, and Kd+K0=K; - for a block of K samples A = (A0; ...; A K-1 ) applies a DFT-spread-OFDM-type scheme and obtains a representation comprising N complex samples S of the radio signal n An output signal of the sign of, and n = 0 to N-1; - Determine a first position T1 and a second position T2 in the output signal based on J1, J2, K and N, such that: οT2-T1=R, and R is a strictly positive integer; - In the output signal, R complex samples (S T1 ;...; S T2 ) is set to zero to obtain the updated output signal; The method is characterized in that: - inserting S reference samples between the first position T1 and the second position T2 of the updated output signal, S is a strictly positive integer, and R=S+L, where L is a positive integer or 0, to obtain a subsequent signal representing a mixed symbol, comprising: - a composite sample S of the radio signal n , wherein n=[0:T1-1]∪[T2+1:N-1], the composite samples form a data segment of the mixed symbol; o Reference samples RS in the radio signal n , wherein n=[T1+L; T2], the reference samples form a reference segment of the mixed symbol; - in view of transmitting a radio signal corresponding to a signal representing said mixed symbol appended with a cyclic prefix, appending to said mixed symbol G samples, said G samples forming at least part of a cyclic prefix of said mixed symbol.

2. The method according to claim 1, wherein: Inserting S reference samples between the first position T1 and the second position T2 of the updated output signal to obtain a subsequent signal further comprises: inserting L cyclic prefix samples between the first position T1 and the second position T2, so that the subsequent signal comprises: o Cyclic prefix samples CPRS in the radio signal n , wherein n=[T1+L; T2], the cyclic prefix samples form a cyclic prefix of the reference segment of the mixed symbol.

3. The method according to claim 1, further comprising the steps of: Before applying the DFT-spread-OFDM-type scheme to the block of K samples and obtaining the output signal, the values ​​of K0 consecutive samples between the first position J1 and the second position J2 in the block of K samples are set to zero.

4. The method according to claim 1, wherein: At least part of the K0 consecutive samples in the block of K samples is obtained by the following steps before applying the DFT-spreading-OFDM-type scheme to the block of K samples and obtaining the output signal: (a) Get K samples (A'0; ...; A' K-1 ), each temporary block of the K samples comprises: - Kd data samples to send, and -K0 temporary consecutive ghost samples (A' J1 ;...;A' J2 ); (b) determining, for each temporary block of K samples, said R complex samples S′ for each of n=[T1; T2] obtained between said first position T1 and said second position T2 as a result of applying said DFT-spread-OFDM-type scheme to said temporary block of K samples n at least part of the value of ; (c) for each temporary block of K samples, determining a criterion associated with said temporary block of K samples indicating a level of similarity between: - the R composite samples S' n the value of at least part of - Null value; and (d) determining the K0 consecutive samples (A) of the block of K samples based on the value of each of the temporary blocks of K samples and each criterion determined in (c). J1 ;...;A J2 ) such that a first value exhibits an optimized level of similarity to a second value, wherein the first value is such that the R complex samples S of each of n=[T1; T2] obtained between the first position T1 and the second position T2 as a result of applying the DFT-spread-OFDM-type scheme to the block of K samples n At least part of the values ​​in the above table, the second value is a null value.

5. The method according to claim 4, wherein: Determine the K0 consecutive samples (A) of the block of K samples J1 ;... ; A J2 ) at least part of the steps include: applying the least squares method.

6. The method according to claim 1, wherein: The K0 consecutive samples (A J1 ;...;A J2 ) are at least part of the phantom samples of the phantom pilot type.

7. The method according to claim 5, wherein: Before applying the DFT-spread-OFDM-type scheme to the block of K samples and obtaining the output signal, the K0 consecutive samples (A) in the block of K samples are determined by the following steps: J1 ;...;A J2 ) at least part of: - obtaining a reference signal comprising S reference samples to be inserted appended with L cyclic prefix samples to be inserted, - applying an inverse DFT-spreading-OFDM-type scheme to said reference signal to obtain a demodulation reference signal comprising a block of K demodulation samples, said block of K demodulation samples comprising K0 consecutive demodulation samples, - and inserting at least part of said K0 consecutive demodulated samples into said block of K samples as ghost samples of a ghost pilot type forming at least part of said K0 consecutive samples.

8. The method according to claim 6, wherein: Before applying the DFT-spread-OFDM-type scheme to the block of K samples and obtaining the output signal, the K0 consecutive samples (A) in the block of K samples are determined by the following steps: J1 ;...;A J2 ) at least part of: (a) Get K samples (A'0; ...; A' K-1 ), each temporary mixed block of the K samples comprises: - the Kd data samples to be sent, and - K0 temporary consecutive ghost samples of ghost pilot type (A' J1 ;...;A' J2 ); (b) determining, for each temporary block of K samples, said R complex samples S" for each of n=[T1; T2] obtained between said first position T1 and said second position T2 as a result of applying said DFT-spread-OFDM-type scheme to said temporary mixed block of K samples n at least part of the value of ; (c) for each temporary mixed block of said K samples, determining a criterion associated with said temporary mixed block of said K samples indicating a level of similarity between: - the R composite samples S" n the value of at least part of - values ​​of at least part of the S reference samples to be inserted appended with the L cyclic prefix samples to be inserted; and (d) determining the K0 consecutive samples (A′) of the block of K samples based on the temporary mixed block of K samples and each criterion determined in (c) J1 ;...;A' J2 ), so that the first value exhibits an optimized level of similarity to the second value, wherein the first value is such that the R complex samples S for each of n=[T1; T2] obtained between the first position T1 and the second position T2 as a result of applying the DFT-spread-OFDM type scheme to the block of K samples n The second value is at least a part of the value of the S reference samples to be inserted, and is appended with L cyclic prefix samples to be inserted.

9. The method according to claim 8, further comprising the steps of: (e) Get K samples (A'0; ... ; A' K-1 ), each temporary block of the K samples comprises: - Kd data samples to send, and -K0 temporary consecutive ghost samples (A' J1 ;...;A' J2 ); (f) determining, for each temporary block of K samples, said R complex samples S′ for each of n=[T1; T2] obtained between said first position T1 and said second position T2 as a result of applying said DFT-spread-OFDM-type scheme to said temporary block of K samples n at least part of the value of ; (g) for each temporary block of K samples, determining a criterion associated with said temporary block of K samples indicating a level of similarity between: - the R composite samples S' n the value of at least part of - Null value; and (h) determining at least a portion of K0 consecutive samples based on the value of each of the temporary blocks of K samples and each criterion determined in (g) such that a first value exhibits an optimized level of similarity to a second value, wherein the first value is such that, as a result of applying the DFT-spread-OFDM-type scheme to the block of K samples, each of the R complex samples S of n=[T1; T2] obtained between the first position T1 and the second position T2 n , the second value is a null value; (i) updating the K0 consecutive samples of the block of K samples based on the at least part of the K0 consecutive samples determined in (h); J1 ;...;A J2 ) at least part of.

10. The method according to any one of claims 4, 5 and 9, wherein: Updating the values ​​of at least part of the K0 consecutive samples in the block of K samples comprises: - determining a third position J1+Pa1 and a fourth position J2-Pa2 in the block of K samples, such that the K0 consecutive samples include: οPa the front of 1 consecutive samples (A J1 ;...;A J1+Pa1 ); ο middle (A J1+Pa1+1 ;...; A J2-Pa2-1 ); οPaThe rear part of 2 consecutive samples (A J2-Pa2 ;...;A J2 ), Pa1 and Pa2 are positive integers, and Pa1+Pa2 is strictly less than K0. - and updating the values ​​of the front and rear samples of the K0 consecutive samples.

11. The method according to any one of claims 1 to 9, wherein: The reference samples are generated by applying an OFDM type scheme to a CAZAC sequence.

12. The method according to claim 10, wherein: The reference samples are generated by applying an OFDM type scheme to a CAZAC sequence.

13. The method according to claim 11, wherein: The reference samples are generated by applying an OFDM type scheme to a Zadoff-Chu sequence, wherein the Zadoff-Chu sequence is a CAZAC sequence.

14. The method according to claim 12, wherein: The reference samples are generated by applying an OFDM type scheme to a Zadoff-Chu sequence, wherein the Zadoff-Chu sequence is a CAZAC sequence.

15. The method according to claim 1, wherein: The K0 consecutive samples between the first position J1 and the second position J2 are ghost samples, and values ​​of at least part of the Kd data samples are determined to compensate for contributions of the K0 ghost samples to the data segment of the mixed symbol.

16. The method according to claim 1, wherein: J1, J2, T1, T2 are defined with respect to each other as follows: and Wherein, each of w1 and w2 is a non-negative integer.

17. A method for determining a channel estimate of a propagation channel used for transmitting a radio signal transmitted according to the method of claim 1, transmitted through the propagation channel and received by a receiver configured to be active on at least K different subcarriers, K being a strictly positive integer, the method comprising the steps of: - based on the received radio signal, obtaining a signal comprising N samples forming a mixed symbol appended with G samples, said G samples forming at least part of a cyclic prefix of said mixed symbol, - obtaining a first position T1 and a second position T2 in the signal, and T2-T1=R, R is a strictly positive integer, R=S+L, L is a positive integer or 0, and S is a strictly positive integer, So that: o The signal comprises a composite sample S n , wherein n=[0; T1-1]∪[T2+1; N-1], the composite samples form a data segment of the mixed symbol; o The signal includes a reference sample RS n , where n=[T1+L; T2], the reference samples form a reference segment of the mixed symbol; - The composite sample S n The values ​​of and the samples of the cyclic prefix forming at least part of the mixed symbol are set to zero to obtain a reference sample RS comprising S n The updated signal of - applying an inverse OFDM type scheme to the updated signal to obtain a demodulated signal comprising a block of consecutive reference samples indicative of said channel estimate of said propagation channel used for transmitting said radio signal, - and determining said channel estimate of said propagation channel used for transmitting a received radio signal based at least on said reference segment of said hybrid symbol.

18. A computer program product comprising code instructions, which, when executed by a processor, perform the method according to claim 1.

19. A telecommunications device, comprising means for transmitting at least one set of Kd data samples in a radio signal to be transmitted over a communication system, the means comprising a transmitter configured to transmit on at least K different subcarriers, K and Kd being strictly positive integers, Kd being strictly less than K, the telecommunications device further comprising processing circuitry to generate the radio signal by: -Determine K samples (A0; ...; A K-1 ) in a first position J1 and a second position J2 in a block of K samples, such that the block of K samples comprises: ο Kd data samples to send; ο and K0 consecutive samples (A J1 ;... ; A J2 ); οwhere J2-J1+1=K0 and Kd+K0=K; - for a block of K samples A = (A0; ...; A K-1 ) applies a DFT-spread-OFDM-type scheme and obtains a representation of the N complex samples S n an output signal comprising symbols in said radio signal, wherein n=0 to N-1; - Determine a first position T1 and a second position T2 in the output signal based on J1, J2, K and N, such that: οT2-T1=R, and R is a strictly positive integer; - in the output signal, the R complex samples (S) between the first position T1 and the second position T2 are T1 ;...;S T2 ) is set to zero to obtain the updated output signal; The telecommunication device is characterized in that the processing circuit further generates the radio signal by: - inserting S reference samples between the first position T1 and the second position T2 of the updated output signal, where S is a strictly positive integer, R=S+L, where L is a positive integer or 0, to obtain a subsequent signal representing a mixed symbol, comprising: o The composite sample S in the radio signal n , wherein n=[0; T1-1]∪[T2+1; N-1], the composite samples form a data segment of the mixed symbol; o Reference samples RS in the radio signal n , where n=[T1+L; T2], the reference samples form a reference segment of the mixed symbol; - appending a cyclic prefix of at least part of the mixed symbol to the mixed symbol to obtain a signal representing the mixed symbol appended with the cyclic prefix; and - generating a radio signal corresponding to a signal representing said mixed symbol to which said cyclic prefix is ​​appended, in view of transmitting said radio signal.

20. A telecommunication device comprising means for receiving at least one set of Kd data samples in a radio signal transmitted via a communication system, the means being configured to operate on at least K different subcarriers, K and Kd being strictly positive integers, Kd being strictly less than K, The telecommunication device further comprises a processing circuit which determines a channel estimate of a propagation channel for transmitting the radio signal to the telecommunication device in the following manner, wherein The radio signal is received as a received radio signal: - based on the received radio signal, obtaining a signal comprising N samples forming a mixed symbol, the mixed symbol being appended with G samples, the G samples forming at least part of a cyclic prefix of the mixed symbol, wherein N and G are strictly positive integers, - obtaining a first position T1 and a second position T2 in the signal, wherein T2-T1=R, R is a strictly positive integer, R=S+L, L is a positive integer or 0, and S is a strictly positive integer, So that: o The signal comprises a composite sample S n , wherein n=[0; T1-1]∪[T2+1; N-1], the composite samples form a data segment of the mixed symbol; o The signal includes a reference sample RS n , wherein n=[T1+L; T2], the reference samples form a reference segment of the mixed symbol; - The composite sample S n The values ​​of and the samples of the cyclic prefix forming at least part of the mixed symbol are set to zero to obtain a reference sample RS comprising S n The updated signal of - applying an inverse OFDM-type scheme to said updated signal to obtain a demodulated signal comprising a block of consecutive reference samples indicative of said channel estimate of said propagation channel used for transmitting said radio signal.