Channel equalization system and method for cyclic decision feedback and improvement of method

Abstract

The invention discloses a channel equalization system and method for cyclic decision feedback and improvement of the method, and belongs to the field of channel equalization, the system comprises a recombination frame module used for replacing nth data of a zero-filled receiving pilot frequency with nth data of a data segment to be equalized to obtain recombination frame data; updating the equalizer coefficient by using the channel response in the initial state; updating the known pilot frequency; the cyclic equalization module is used for completing channel compensation of the recombined frame data on a frequency domain and converting an equalized recombined frame frequency domain signal back to a time domain through IFFT (Inverse Fast Fourier Transform) to obtain a recombined frame equalization signal; updating an equalizer coefficient according to the recombined frame frequency domain signal and the updated known pilot frequency domain signal; and the decision feedback module is used for demodulating the nth data of the recombined frame equalization signal and modulating a demodulation result to obtain modulated data. According to the invention, the problem of residual inter-symbol interference caused by an equalizer in existing communication systems such as OFDM, SC-FDE and the like is solved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of channel equalization, and particularly relates to a cyclic decision feedback channel equalization system and method and improvement of the method. BACKGROUND

[0002] In a wireless communication system, due to the multipath propagation effect of the channel, the transmitted signal will pass through different paths with delay, attenuation and phase shift, thereby generating inter-symbol interference (ISI). In a wideband communication system taking OFDM (Orthogonal Frequency Division Multiplexing) as the core, multipath interference will also cause inter-carrier interference (ICI), further threatening the transmission reliability. The existence of inter-symbol interference will seriously degrade the quality of the received signal, leading to an increase in the bit error rate and a decrease in communication reliability. Channel equalization technology, as a core means to combat inter-symbol interference, can restore the disturbed transmission signal by estimating and compensating the channel characteristics, and is a key supporting technology for guaranteeing high-reliability and high-speed wireless communication in 5G / 6G, Internet of Things, Internet of Vehicles and other scenarios.

[0003] Currently, frequency domain channel equalization has become a key link to combat multipath interference in OFDM, SC-FDE and other communication systems because it can reduce computational complexity by means of fast Fourier transform and other efficient algorithms. However, the existing frequency domain channel equalization technology still has many deficiencies. When the multipath time delay is large and the channel time variation is fast (such as in high-speed mobile scenarios), the rapid change of channel characteristics easily leads to an increase in channel estimation error, and the traditional frequency domain equalization algorithm is difficult to balance between estimation accuracy and computational complexity, and has insufficient adaptability to non-stationary channels, which easily leaves residual equalization interference, further affecting the signal demodulation performance. Therefore, how to solve the residual equalization interference and improve the adaptability of the equalization algorithm has become a technical problem to be solved in the field. SUMMARY

[0004] In view of the above deficiencies in the prior art, the present application provides a cyclic decision feedback channel equalization system and method and improvement of the method, which solves the problem of residual inter-symbol interference caused by the equalizer in the existing OFDM, SC-FDE and other communication systems.

[0005] In order to achieve the above-mentioned application purposes, the technical scheme adopted by the present application is as follows: a cyclic decision feedback channel equalization system, comprising: The recombining frame module is used for extracting the data frame to be equalized, zero-padding the length of the known pilot and the received pilot in the data frame to be equalized to be the same as that of the data segment to be equalized, replacing the nth data of the received pilot after zero-padding with the nth data of the data segment to be equalized to obtain recombined frame data; the channel response in the initial state is calculated by using the known pilot after zero-padding and the received pilot after zero-padding, and the gain coefficient of the equalizer is updated by using the channel response in the initial state; the nth data of the known pilot after zero-padding is replaced with modulation data to obtain updated known pilot; The cyclic equalization module is used for transforming the recombined frame data to the frequency domain to obtain recombined frame frequency domain signals, completing the channel compensation of the recombined frame data in the frequency domain according to the recombined frame frequency domain signals, obtaining equalized recombined frame frequency domain signals, and transforming the equalized recombined frame frequency domain signals back to the time domain by IFFT to obtain recombined frame equalization signals; the gain coefficient of the equalizer is updated according to the recombined frame frequency domain signals and the updated known pilot frequency domain signals; The decision feedback module is used for demodulating the nth data of the recombined frame equalization signals and modulating the demodulated result to obtain modulation data.

[0006] Further, the known pilot is a sequence preset by both parties of communication.

[0007] Further, the data frame to be equalized includes the received pilot and the data segment to be equalized.

[0008] The application has the following beneficial effects: through special frame format design, the interference difference caused by different frame data of the channel can be dynamically reacted, and the channel gain coefficient of the equalizer can be quickly updated; through the time domain zero-padding method, the number of channel estimation points is improved, and the real frequency response of the channel can be more accurately estimated compared with the traditional channel estimation; the combination of cyclic feedback and pre-decision can dynamically adjust the estimated value of the channel frequency response and the gain coefficient of the equalizer, further reduce the influence of inter-symbol interference, and reduce the influence of colored noise of digital demodulation. The equalized signal decision result is more accurate, and compared with the traditional frequency domain channel equalization method, the demodulated data is closer to the real data of the signal at the transmitting end.

[0009] A channel equalization method, comprising: S1, acquiring data to be equalized and determining a known pilot; S2, sequentially extracting data frames to be equalized from the data to be equalized; S3, respectively performing tail zero-padding on the received pilot and the known pilot until the lengths of the pilots are the same as those of the data segment to be equalized; S4, respectively performing FFT transformation on the received pilot after zero-padding and the known pilot after zero-padding to obtain received pilot frequency domain signals and known pilot frequency domain signals; S5, performing channel estimation according to the received pilot frequency domain signal and the known pilot frequency domain signal to obtain channel responses of each frequency point; S6, calculating gain coefficients of the equalizer according to the channel responses of each frequency point; S7, replacing the nth data of the received pilot after zero padding with the nth data of the data segment to be equalized to obtain reorganized frame data; S8, performing FFT transformation on the reorganized frame data to obtain reorganized frame frequency domain signals; S9, performing point multiplication on the reorganized frame frequency domain signals and the gain coefficients of the equalizer to obtain equalized reorganized frame frequency domain signals; S10, transforming the equalized reorganized frame frequency domain signals back to the time domain through IFFT transformation to obtain reorganized frame equalized signals; the nth data of the reorganized frame equalized signals is the equalized data corresponding to the nth data of the data segment to be equalized; S11, demodulating the nth data of the reorganized frame equalized signals and modulating the demodulated results to obtain modulation data; S12, replacing the nth data of the known pilot after zero padding with the modulation data to obtain updated known pilots; S13, performing FFT transformation on the updated known pilots to obtain updated known pilot frequency domain signals, and updating the gain coefficients of the equalizer according to the reorganized frame frequency domain signals and the updated known pilot frequency domain signals; returning to step S7 to perform equalization processing on the next data of the data segment to be equalized until the equalization processing of the data segment to be equalized is completed, and entering step S14; S14, returning to step S2 to perform equalization processing on the next data frame to be equalized.

[0010] Further, the data to be equalized includes a plurality of data frames to be equalized; and there is a guard interval segment between each data frame to be equalized.

[0011] Further, the number of frequency points of the received pilot frequency domain signal and the known pilot frequency domain signal is the same as the length value of the data segment to be equalized.

[0012] The present application has the following advantages: through special frame format design, the interference difference caused by different frame data of the channel can be dynamically reacted, and the channel gain coefficient of the equalizer can be quickly updated. Through the time domain zero padding method, the number of channel estimation points is improved, and the real frequency response of the channel can be more accurately estimated compared with the traditional channel estimation; the estimation value of the channel frequency response and the gain coefficient of the equalizer can be dynamically adjusted by using the combination of cyclic feedback and pre-decision, further reducing the influence of inter-symbol interference, and reducing the influence of colored noise of digital demodulation. The equalized signal decision result is more accurate, and compared with the traditional frequency domain channel equalization method, the demodulated data is closer to the real data of the transmitting end signal.

[0013] An improvement of a channel equalization method, comprising the following steps: A1, obtaining a data frame to be equalized; A2, performing channel estimation by receiving a pilot and a known pilot sequence, obtaining a channel response sequence, and updating a gain coefficient of an equalizer according to the channel response sequence; A3, extracting an nth data from a data segment to be equalized in the data frame to be equalized as current data to be equalized, and updating the pilot by using the current data to be equalized; A4, performing dot multiplication on the current data to be equalized and the gain coefficient of the equalizer and summing the results, and dividing the sum by the number of the gain coefficients of the equalizer to obtain a mean result; the nth data of the mean result is the equalized data corresponding to the current data to be equalized; A5, demodulating the nth data of the mean result, and modulating the demodulated result to obtain modulated data; A6, updating the known pilot by using the modulated data; A7, updating the gain coefficient of the equalizer according to the updated pilot and the updated known pilot; A8, returning to step A3 to perform equalization processing on the next data to be equalized in the data segment to be equalized until the equalization of the data segment to be equalized is completed.

[0014] Further, the updating of the pilot in A3 is specifically: keeping the length of the pilot unchanged, and placing the current data to be equalized at the first position of the pilot.

[0015] Further, the updating of the known pilot in A6 is specifically: keeping the length of the known pilot unchanged, and placing the modulated data at the first position of the known pilot.

[0016] Further, A7 is specifically: A701, determining a pilot sequence sum, when n=1, the pilot sequence sum is the sum of all data of the updated pilot, and when n>1, the pilot sequence sum is the sum of the current data to be equalized and the pilot of the last round; A702, determining a known pilot sequence sum, when n=1, the known pilot sequence sum is the sum of all data of the updated known pilot, and when n>1, the known pilot sequence sum is the sum of the modulated data and the known pilot sequence sum of the last round; A703, performing single-point channel response calculation on the pilot sequence sum and the known pilot sequence sum as the first frequency point values of the pilot and the known pilot respectively to obtain a single-point channel response result; A704, keeping the length of the channel response sequence unchanged, placing the single-point channel response result at the first position of the channel response sequence to obtain an updated channel response sequence; A705、According to the updated channel response sequence, the gain coefficient corresponding to the first frequency point of the equalizer is recalculated.

[0017] The application has the following advantages: through special frame format design, dynamic response to the interference difference caused by different data to the channel, fast update of the channel gain coefficient of the equalizer, using the combination of cyclic feedback and pre-decision, the estimated value of the channel frequency response and the gain coefficient of the equalizer can be dynamically adjusted, the influence of inter-symbol interference is further reduced, the colored noise influence of digital demodulation is reduced. In the case of only a small amount of data change, the improved method avoids the recalculation of the overall frequency domain information of the data through the butterfly calculation structure characteristics of the FFT algorithm, greatly simplifies the channel response and equalizer coefficient calculation process, effectively improves the performance of the equalization algorithm, and greatly saves the limited computing resources in the actual deployment of the project. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 A schematic diagram of a channel equalization system architecture of the application is shown.

[0019] Figure 2 A flowchart of a channel equalization method of the application is shown.

[0020] Figure 3 A flowchart of an improved channel equalization method of the application is shown.

[0021] Figure 4 A schematic diagram of a frame structure of the application is shown.

[0022] Figure 5 A schematic diagram of a structure for quickly solving the equalization result of the application is shown. DETAILED DESCRIPTION

[0023] The specific embodiments of the application are described below to facilitate understanding of the application by those skilled in the art, but it should be clear that the application is not limited to the scope of the specific embodiments, and for those skilled in the art, any changes within the spirit and scope of the application defined and determined by the appended claims are obvious, and all applications utilizing the concept of the application are within the scope of protection.

[0024] As shown in Figure 1 In one embodiment of the application, a channel equalization system with cyclic decision feedback includes: The reassembled frame module is used to extract the data frame to be equalized. It pads the length of the known pilot and the received pilot in the data frame to be equalized with zeros until it matches the length of the data segment to be equalized in the data frame. It then replaces the nth data point of the zero-padding received pilot with the nth data point of the data segment to be equalized, obtaining the reassembled frame data. Using the zero-padding known pilot and the zero-padding received pilot, it calculates the channel response in the initial state and updates the equalizer's gain coefficient using the initial channel response. Finally, it replaces the nth data point of the zero-padding known pilot with the modulation data, obtaining the updated known pilot. The cyclic equalization module transforms the recombined frame data to the frequency domain to obtain the recombined frame frequency domain signal. Based on the recombined frame frequency domain signal, it performs channel compensation on the recombined frame data in the frequency domain to obtain the equalized recombined frame frequency domain signal. The equalized recombined frame frequency domain signal is then transformed back to the time domain using IFFT to obtain the recombined frame equalization signal. Based on the recombined frame frequency domain signal and the updated known pilot frequency domain signal, the gain coefficient of the equalizer is updated. The decision feedback module is used to demodulate the nth data of the recombined frame equalization signal and modulate the demodulated result to obtain modulated data.

[0025] The known pilot signals are a sequence pre-set by both communicating parties.

[0026] The data frame to be equalized includes a received pilot and a data segment to be equalized.

[0027] In this embodiment, Figure 1 In This is the receiving pilot after zero padding; The known pilot signal after zero-padding; For reassembling frame data; For the reassembled frame data n One data point; FFT stands for Fast Fourier Transform, used to convert the reconstructed frame data to the frequency domain; The frequency domain signal is reconstructed; the equalizer is used to perform channel compensation. The reconstructed frame frequency domain signal after equalization; The result is the equilibrium data.

[0028] Example 2 like Figure 2 As shown, a channel equalization method includes: S1. Obtain the data to be equalized and determine the known pilot frequencies; S2. Extract the data frames to be balanced sequentially from the data to be balanced; S3. Pad the received pilot and the known pilot with zeros at the end until the pilot length is the same as the data segment to be equalized. S4. Perform FFT transformation on the zero-padded received pilot and the zero-padded known pilot respectively to obtain the received pilot frequency domain signal and the known pilot frequency domain signal. S5. Perform channel estimation based on the received pilot frequency domain signal and the known pilot frequency domain signal to obtain the channel response at each frequency point; S6. Calculate the gain coefficient of the equalizer based on the channel response at each frequency point; S7. Replace the nth data of the zero-padded received pilot with the nth data of the data segment to be equalized to obtain the recombined frame data; S8. Perform FFT transformation on the reconstructed frame data to obtain the frequency domain signal of the reconstructed frame; S9. Multiply the reconstructed frame frequency domain signal by the gain coefficient of the equalizer to obtain the equalized reconstructed frame frequency domain signal. S10. The equalized recombined frame frequency domain signal is transformed back to the time domain by IFFT to obtain the recombined frame equalization signal; the nth data of the recombined frame equalization signal is the equalization data corresponding to the nth data of the data segment to be equalized. S11. Demodulate the nth data of the recombined frame equalization signal, and modulate the demodulation result to obtain modulated data; S12. Replace the nth data of the known pilot after zero-padding with the modulation data to obtain the updated known pilot; S13. Perform FFT transformation on the updated known pilot to obtain the updated known pilot frequency domain signal, and update the gain coefficient of the equalizer according to the reconstructed frame frequency domain signal and the updated known pilot frequency domain signal; return to step S7 to perform equalization processing on the next data of the data segment to be equalized until the equalization processing of the data segment to be equalized is completed, and proceed to step S14. S14. Return to step S2 to perform equalization processing on the next data frame to be equalized.

[0029] like Figure 4 As shown, the data to be balanced includes several data frames to be balanced; there is a guard interval between each data frame to be balanced.

[0030] In this embodiment, the data to be equalized is a time-domain serial signal. According to a pre-defined special transmission frame structure, the data to be equalized is divided into frames, as shown in the following frame structure: Figure 4 As shown, each frame of signal is divided into a pilot segment, an effective data segment, and a guard interval segment. The sequence length of the pilot segment and the guard interval segment is shorter than that of the effective data segment.

[0031] In this embodiment, the pilot sequence can include, but is not limited to, unique word (UW) sequences and constant amplitude zero autocorrelation (CAZAC) sequences, and the guard interval uses the same data as the pilot. By using different pilot data for different frames, the dynamic changes of the channel can be estimated, and the guard interval can act as a cyclic prefix, converting the linear convolution of the channel into a cyclic convolution.

[0032] Zeros are added to the end of the extracted pilot data sequence. The number of pilot data in each frame is L, the number of valid data is M, and the number of pilot zeros is ML. The data in the guard interval is not processed and is directly discarded.

[0033] In this embodiment, the purpose of zero-padding the pilot in the time domain is as follows: zero-padding in the time domain is equivalent to interpolation in the frequency domain. After zero-padding the pilot in the time domain, the frequency response of the channel can be estimated more accurately, improving the integrity and continuity of the channel response in the entire frequency domain, and making the subsequent channel equalization effect better.

[0034] In this embodiment, the channel frequency response is estimated by receiving pilot signals and known pilot sequences. The channel estimation method can be based on least squares (LS) and minimum mean square error (MMSE) or other methods. The equalizer gain coefficient calculation method can be selected from, but is not limited to, least squares (LS), minimum mean square error (MMSE), linear minimum mean square error (LMMSE), etc.

[0035] In this embodiment, by reframing the data to be equalized with a known sequence, the known channel frequency response characteristics are utilized to use an equalizer to eliminate inter-symbol interference from other unknown symbols in the data to be equalized.

[0036] In this embodiment, the transmitting end can use digital modulation methods such as BPSK and QPSK. Since the pilot sequence modulation method is the same as the effective data, the receiving end uses the corresponding demodulation method.

[0037] In this embodiment, to reflect the channel changes in the data to be equalized, the channel frequency response needs to be re-estimated. Therefore, the equalized data is first demodulated and then digitally modulated. The data at the corresponding positions of the known pilot sequence is replaced to update the known pilots. Then, the channel is re-estimated to complete the update of the channel frequency response. This allows the updated channel estimate to dynamically reflect the differences in the channel's impact on different data segments.

[0038] In this embodiment, because the channel estimate and the data to be equalized have changed, the gain coefficient of the equalizer needs to be recalculated to improve the equalization effect of the equalizer on the next data.

[0039] The number of frequency points of the received pilot frequency domain signal and the known pilot frequency domain signal is the same as the length of the data segment to be equalized.

[0040] In this embodiment, taking the data frame "pilot p1 + data segment y1" as an example, pilot p1 is called the received pilot Rx_p1, which is the sequence after the known pilot sequence Tx_p passes through the multipath channel. The Rx_p of each frame of data is the result of Tx_p passing through the multipath channel. However, due to the randomness of the channel, Rx_p may be different, so it is named p1, p2... The received pilot described refers to the pilot Rx_pn of the current frame, and the transmitted pilot refers to Tx_p.

[0041] This explanation uses a pilot p1 length of 128 and a data segment y1 length of 512 as an example: B1. Obtain 1 frame of data (assuming it is the first frame): "Pilot p1 + Data segment y1 + Guard interval GI", without guard interval GI; B2. Pad the end of pilot p1 with zeros to the length of data segment y1 (512 points). Given that the pilot sequence Tx_p is also padded with zeros to 512, perform a 512 FFT transform on the two pilot segments to estimate the channel response H with 512 frequency points. B3. Calculate the coefficients of the equalizer (LS, MMSE equalization method) using the channel response H of this 512 frequency point; B4. Replace the first data y_1 of data segment y1 with the first data p_1 of the zero-padded received pilot p (the recombined frame mentioned), and perform FFT transformation; B5. Multiply the equalizer coefficients obtained in B3 by a dot to complete the frequency domain channel equalization. Transform back to the time domain through IFFT, extract the first point of the recombined frame data after equalization (that is, y_1 after equalization), demodulate (i.e., pre-decision, QPSK demodulation is used if QPSK modulation is used at the transmitting end), and then remodulate. B6. Replace the zero-padded data at the corresponding positions in the known pilot sequence Tx_p with the remodulated data; B7. Perform an FFT transform on the known pilot sequence Tx_p updated in B6, and recompile the reconstructed frame data after the FFT transform in B4. Then, recalculate the 512-point channel response H (update the channel response H) and recalculate the coefficients of the equalizer (LS or MMSE equalization method) (update the equalizer coefficients). B8. Repeat steps B4-B7 until data segment y1 is balanced.

[0042] B9. After the first frame of data ("pilot p1 + data segment y1") is processed, "pilot p2 + data segment y2 + guard interval GI" is extracted. Similarly, the guard interval GI is discarded. Repeat steps B1 to B8 to complete the equalization operation of the second frame of data.

[0043] Example 3 likeFigure 3 As shown, an improvement to the channel equalization method includes the following steps: A1. Obtain the data frame to be balanced; A2. Channel estimation is performed by receiving pilot signals and known pilot sequences to obtain the channel response sequence, and the gain coefficient of the equalizer is updated according to the channel response sequence. A3. Extract the nth data from the data segment to be equalized in the data frame to be equalized as the current data to be equalized, and update the receiving pilot using the current data to be equalized. A4. Multiply the current data to be equalized by the gain coefficient of the equalizer and sum them. Divide the sum by the number of gain coefficients of the equalizer to obtain the mean result. The nth data in the mean result is the equalized data corresponding to the current data to be equalized. A5. Demodulate the nth data point of the mean result, and then modulate the demodulated result to obtain the modulated data; A6. Update the known pilot signals using the modulated data; A7. Update the gain coefficient of the equalizer based on the updated received pilot and the updated known pilot; A8. Return to step A3 to perform the balancing process for the next data segment to be balanced in the data segment to be balanced, until the balancing of the data segment to be balanced is completed.

[0044] In this embodiment, the equilibrium results of the valid data to be balanced can be applied. Figure 5 The special structure in the middle is quickly calculated. The data to be equalized is shifted into the receiving pilot and the data at the end of the receiving pilot is shifted out. The same shift operation is performed on the channel response sequence. The output result is the weighted sum of the effective data to be equalized and the equalizer coefficients. Figure 5 In this context, p(n-1) is the (n-1)th element of the time-domain sequence of the received pilot; y(1) is the first element of the time-domain sequence y(n); y(n) is the time-domain sequence of the data to be equalized; Y(1) is the frequency value corresponding to y(1) after FFT transformation of the first element of the time-domain sequence y(n); Z(n-1) are the frequency values ​​corresponding to the other elements of the time-domain sequence after FFT transformation; W(n) is the frequency domain coefficient of the equalizer; and n is the index of the discrete sequence.

[0045] The specific steps for updating the receiving pilot in A3 are as follows: keep the length of the receiving pilot unchanged, and place the current data to be equalized at the beginning of the receiving pilot.

[0046] The specific steps for updating the known pilot in A6 are as follows: keep the length of the known pilot unchanged, and place the modulated data at the beginning and end of the known pilot.

[0047] In this embodiment, in order to reflect the real-time changes of the channel, a cyclic shift operation is performed on the pilot sequence. The data to be equalized is shifted into the first position of the receiving pilot, the remodulated decision data is shifted into the first position of the transmitting pilot, and the shifted-out last data is discarded, thus completing the update process of the two pilot time-domain sequences.

[0048] Specifically, A7 refers to: A701. Determine the sum of the received pilot sequence. When n=1, the sum of the received pilot sequence is the sum of all data of the updated received pilot. When n>1, the sum of the received pilot sequence is the sum of the received pilot from the previous round and the current data to be equalized. A702. Determine the sum of known pilot sequences. When n=1, the sum of known pilot sequences is the sum of all data of the updated known pilot sequences. When n>1, the sum of known pilot sequences is the sum of the sum of known pilot sequences from the previous round and the modulated data. A703. Using the received pilot sequence and the known pilot sequence as the first frequency point values ​​of the received pilot and the known pilot respectively, perform single-point channel response calculation to obtain the single-point channel response result; A704. Keeping the channel response sequence length unchanged, insert the single-point channel response result into the beginning of the channel response sequence to obtain the updated channel response sequence. A705. Recalculate the gain coefficient corresponding to the first frequency point of the equalizer based on the updated channel response sequence.

[0049] In this embodiment, during the loop processing, the update of the first frequency domain data of the received pilot sequence can be quickly obtained by time-domain summation; similarly, the first frequency domain data of the known pilot sequence can be obtained. Using the updated pilot sequence, the first frequency domain data of the equalizer gain coefficient is recalculated.

[0050] In this embodiment, the two pilot sequences are summed separately, and the summation result is used as the difference between the transmit and receive ends at the first frequency point in the frequency domain. The channel frequency response at this point is calculated as the channel state update, and the gain coefficient of the equalizer at this point is updated at the same time.

[0051] In this embodiment, the pilot p1 length is 128 and the data segment y1 length is 512, as described above: C1. Obtain 1 frame of data (assuming it is the first frame) "pilot p1 + data segment y1 + guard interval GI", and discard the guard interval GI as well; C2. Perform a 128-point FFT transform on pilot p1 and the known pilot sequence Tx_p, and then perform channel estimation to obtain the channel response H1 at 128 frequency points. C3. Calculate the coefficients (128 tap coefficients) of the equalizer (including but not limited to LS and MMSE equalization methods) using the channel response H1 of these 128 frequency points. C4. Shift the first data y1_1 of data segment y1 into the first position of the receiving pilot p1 (in the loop processing, it is also shifted into the first index position). Shift the other data indices of pilot p1 one position to the right. Discard the 128th data bit of pilot p1, keeping the overall length of 128 points unchanged (the first data shifted in, y1_1, is the data to be equalized in the improved algorithm; the data shifted in for the nth time in the loop is y1_n). C5. Multiply the data to be equalized in C4 by the equalizer coefficients obtained in C3 and sum them. Divide the sum by the number of equalizer coefficients (equalization process). Take out the first point after equalization (that is, y1_1 after equalization). Demodulate (i.e., pre-decision, here only for the equalized y1_1, QPSK demodulation is used for QPSK modulation used by the transmitter). Assume that the demodulation result of this data is correct, and then re-modulate (this is the same as the pre-decision and modulation process in the unimproved algorithm, except that it processes a single data). C6. Move the remodulated data into the first index position of the known pilot sequence Tx_p (in the loop processing, it is also moved into the first index position). Shift the other data indices of the known pilot sequence Tx_p to the right by 1 bit. Discard the 128th bit of the known pilot sequence Tx_p. C71. Sum all the data of the updated known pilot sequence Tx_p in C6, and denote it as Sum_tx_p_1; C72. Sum all the data in C4 after shifting in the first data y1_1, and record the result as Sum_y1_1; C73. Use Sum_tx_p_1 and Sum_y1_1 as the first frequency point values ​​in the frequency domain of the two sequences respectively (here, no FFT transformation is performed, but the two values ​​are directly used as frequency domain values. During the loop, Sum_tx_p_1 and Sum_y1_1 are accumulated with their respective updated values. The summation of the entire sequence is no longer calculated. The data is cleared after all the data in the current frame is equalized). C74. Calculate the channel response at a single point, denoted as H1_1. Shift the data into the channel response sequence H1 from step C3 (in the loop, this is also shifted into the first index position), and shift all other data indices one position to the right. Discard the 128th bit of the channel response sequence H1. C75. Recalculate the equalizer coefficients at this point and replace the equalizer coefficients at the first point (unlike the channel response and pilot sequence, here we replace the data at the first point of the equalizer coefficients, rather than shifting them). C8. Repeat steps C4-C7 until the data segment y1 is balanced.

[0052] C9. After the first frame of data ("pilot p1 + data segment y1") is processed, "pilot p2 + data segment y2 + guard interval GI" is extracted. Similarly, the guard interval GI is discarded. Repeat steps C1 to C8 to complete the equalization operation of the second frame of data.

Claims

1. A system for channel equalization with cyclic decision feedback, comprising: The application relates to a method for equalizing data, which comprises the following steps: a recombining frame module is used for extracting a data frame to be equalized, zero-padding the length of a known pilot and a received pilot in the data frame to be equalized to be the same as that of a data segment to be equalized, replacing the nth data of the zero-padded received pilot with the nth data of the data segment to be equalized to obtain recombined frame data; the initial state channel response is calculated by using the zero-padded known pilot and the zero-padded received pilot, and the gain coefficient of the equalizer is updated by using the initial state channel response; the nth data of the zero-padded known pilot is replaced with modulation data to obtain updated known pilot; a cyclic equalization module is used for transforming the recombined frame data into a frequency domain to obtain recombined frame frequency domain signals, and completing channel compensation of the recombined frame data in the frequency domain according to the recombined frame frequency domain signals to obtain equalized recombined frame frequency domain signals; the equalized recombined frame frequency domain signals are transformed back into a time domain through IFFT to obtain recombined frame equalization signals; the gain coefficient of the equalizer is updated according to the recombined frame frequency domain signals and the updated known pilot frequency domain signals; a decision feedback module is used for demodulating the nth data of the recombined frame equalization signals and modulating the demodulated results to obtain modulation data.

2. The system for channel equalization with cyclic decision feedback according to claim 1, wherein, The known pilot is a sequence preset by both sides of communication.

3. The system of claim 1, wherein the decision feedback is cyclic. The data frame to be equalized comprises a received pilot and a data segment to be equalized.

4. A channel equalization method for a channel equalization system based on the cycle- by-cyclic decision feedback of any one of claims 1 to 3, characterized by The application further relates to a method for equalizing data, which comprises the following steps: S1, acquiring data to be equalized and determining a known pilot; S2, extracting data frames to be equalized from the data to be equalized in sequence; S3, zero-padding the tails of the received pilot and the known pilot respectively until the lengths of the pilots are the same as that of the data segment to be equalized; S4, performing FFT transformation on the zero-padded received pilot and the zero-padded known pilot respectively to obtain received pilot frequency domain signals and known pilot frequency domain signals; S5, performing channel estimation according to the received pilot frequency domain signals and the known pilot frequency domain signals to obtain channel responses of each frequency point; S6, calculating the gain coefficient of the equalizer according to the channel responses of each frequency point; S7, replacing the nth data of the zero-padded received pilot with the nth data of the data segment to be equalized to obtain recombined frame data; S8, performing FFT transformation on the recombined frame data to obtain recombined frame frequency domain signals; S9, point-multiplying the recombined frame frequency domain signals with the gain coefficient of the equalizer to obtain equalized recombined frame frequency domain signals; S10, transforming the equalized recombined frame frequency domain signals back into a time domain through IFFT to obtain recombined frame equalization signals; the nth data of the recombined frame equalization signals is equalization data corresponding to the nth data of the data segment to be equalized; S11, demodulating the nth data of the recombined frame equalization signals and modulating the demodulated results to obtain modulation data; S12, replacing the nth data of the zero-padded known pilot with the modulation data to obtain updated known pilot; S13, performing FFT transformation on the updated known pilot to obtain updated known pilot frequency domain signals, and updating the gain coefficient of the equalizer according to the recombined frame frequency domain signals and the updated known pilot frequency domain signals; returning to step S7 to perform equalization processing on the next data of the data segment to be equalized until the data segment to be equalized is equalized, and entering step S14; S14, return to step S2 to perform equalization processing on the next data frame to be equalized.

5. The channel equalization method of claim 4, wherein, The data to be equalized includes a plurality of data frames to be equalized; and there is a guard interval between each data frame to be equalized.

6. The channel equalization method of claim 4, wherein, The number of frequency points of the received pilot frequency domain signal and the known pilot frequency domain signal is the same as the length of the data segment to be equalized.

7. A method of equalizing a channel as defined in any one of claims 4-6, characterized by The method comprises the following steps: A1, obtaining a data frame to be equalized; A2, performing channel estimation by receiving a pilot and a known pilot sequence to obtain a channel response sequence, and updating the gain coefficient of the equalizer according to the channel response sequence; A3, extracting the nth data from the data segment to be equalized in the data frame to be equalized as the current data to be equalized, and updating the received pilot using the current data to be equalized; A4, point-multiplying the current data to be equalized with the gain coefficient of the equalizer and summing, and dividing the sum by the number of gain coefficients of the equalizer to obtain a mean result; The nth data of the mean result is the equalized data corresponding to the current data to be equalized; A5, demodulating the nth data of the mean result, and modulating the demodulated result to obtain modulated data; A6, updating the known pilot with the modulated data; A7, updating the gain coefficient of the equalizer according to the updated received pilot and the updated known pilot; A8, returning to step A3 to perform equalization processing on the next data to be equalized in the data segment to be equalized until the equalization of the data segment to be equalized is completed.

8. The improvement of the channel equalization method of claim 7 wherein, The updating of the received pilot in A3 is specifically: keeping the length of the received pilot unchanged, and placing the current data to be equalized at the beginning of the received pilot.

9. The improvement of the channel equalization method of claim 7 wherein, The updating of the known pilot in A6 is specifically: keeping the length of the known pilot unchanged, and placing the modulated data at the beginning of the known pilot.

10. The improvement of the channel equalization method of claim 7 wherein, The A7 is specifically: A701, determining a received pilot sequence sum, when n=1, the received pilot sequence sum is the sum of all data of the updated received pilot, and when n>1, the received pilot sequence sum is the sum of the current data to be equalized and the received pilot of the last round; A702, determining a known pilot sequence sum, when n=1, the known pilot sequence sum is the sum of all data of the updated known pilot, and when n>1, the known pilot sequence sum is the sum of the modulated data and the known pilot sequence sum of the last round; A703, performing single-point channel response calculation on the first frequency point value of the received pilot and the known pilot respectively using the received pilot sequence sum and the known pilot sequence sum to obtain a single-point channel response result; A704, keeping the length of the channel response sequence unchanged, and placing the single-point channel response result at the beginning of the channel response sequence to obtain an updated channel response sequence; A705, recalculating the gain coefficient corresponding to the first frequency point of the equalizer according to the updated channel response sequence.

Images