Improved method for a channel equalization system with cyclic decision feedback
The channel equalization system using cyclic decision feedback dynamically updates the channel gain coefficient of the equalizer by combining pilot zero-padding and cyclic feedback with pre-decision, thus solving the channel estimation error problem of frequency domain channel equalization technology when multipath delay spread and channel time-varying speed, and improving signal demodulation performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU UNIV
- Filing Date
- 2026-02-03
- Publication Date
- 2026-05-15
AI Technical Summary
Existing frequency domain channel equalization techniques are prone to increasing channel estimation errors when multipath delay spread is large and channel time-varying is fast. It is difficult to achieve a balance between estimation accuracy and computational complexity. Furthermore, they are not adaptable to non-stationary channels, resulting in residual inter-symbol interference and affecting signal demodulation performance.
The channel equalization system employing cyclic decision feedback dynamically updates the equalizer's channel gain coefficient by using a frame reassembly module, a cyclic equalization module, and a decision feedback module, and by employing zero-padding processing of known and received pilots. Combined with cyclic feedback and pre-decision methods, it reduces the impact of inter-symbol interference.
It improves the accuracy of channel estimation and the adaptability of equalization algorithms, reduces the influence of colored noise in digital demodulation, and improves the accuracy of signal decision results, bringing them closer to the real data of the transmitted signal.
Smart Images
Figure CN121644288B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of channel equalization, and particularly relates to a channel equalization system and method with cyclic decision feedback, as well as improvements to the method. Background Technology
[0002] In wireless communication systems, due to the multipath propagation effect of the channel, transmitted signals experience delays, attenuation, and phase shifts along different paths, resulting in inter-symbol interference (ISI). In broadband communication systems based on OFDM (Orthogonal Frequency Division Multiplexing), multipath interference can also induce inter-carrier interference (ICI), further threatening transmission reliability. The presence of ISI severely degrades the quality of received signals, leading to increased bit error rates and decreased communication reliability. Channel equalization technology, as a core means of combating ISI, can recover interfered transmission signals by estimating and compensating for channel characteristics. It is a key supporting technology for ensuring high-reliability, high-speed wireless communication in scenarios such as 5G / 6G, IoT, and vehicular networks.
[0003] Currently, frequency domain channel equalization has become a key component in combating multipath interference in communication systems such as OFDM and SC-FDE due to its ability to reduce computational complexity through efficient algorithms such as the Fast Fourier Transform. However, existing frequency domain channel equalization techniques still have many shortcomings. In situations with large multipath delay spread and rapid channel time variation (such as in high-speed mobile scenarios), the rapid changes in channel characteristics can easily lead to increased channel estimation errors. Furthermore, traditional frequency domain equalization algorithms struggle to balance estimation accuracy and computational complexity, and they lack adaptability to non-stationary channels, easily leaving residual equalization interference, which further affects signal demodulation performance. Therefore, how to solve residual equalization interference and improve the adaptability of equalization algorithms has become an urgent technical problem to be solved in this field. Summary of the Invention
[0004] To address the aforementioned shortcomings in the prior art, this invention provides a cyclic decision feedback channel equalization system and method, as well as improvements to the method, which solve the problem of residual inter-symbol interference caused by equalizers in existing OFDM, SC-FDE and other communication systems.
[0005] To achieve the above-mentioned objectives, the technical solution adopted by this invention is: a channel equalization system with cyclic decision feedback, comprising:
[0006] 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.
[0007] 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.
[0008] 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.
[0009] Furthermore, the known pilot signals are a sequence pre-set by both communicating parties.
[0010] Furthermore, the data frame to be equalized includes a received pilot and a data segment to be equalized.
[0011] The beneficial effects of this invention are as follows: Through a special frame format design, it can dynamically reflect the differences in interference caused by the channel to different frame data, and quickly update the channel gain coefficient of the equalizer. By using time-domain zero-padding, the number of points in the channel estimation is increased, resulting in a more accurate estimation of the channel's true frequency response compared to traditional channel estimation methods. The combination of cyclic feedback and pre-decision allows for dynamic adjustment of the estimated channel frequency response and the equalizer's gain coefficient, further reducing the impact of inter-symbol interference and minimizing colored noise in digital demodulation. The signal decision result after equalization is more accurate, and compared to traditional frequency-domain channel equalization methods, the demodulated data is closer to the actual data of the transmitted signal.
[0012] A channel equalization method, comprising:
[0013] S1. Obtain the data to be equalized and determine the known pilot frequencies;
[0014] S2. Extract the data frames to be balanced sequentially from the data to be balanced;
[0015] 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.
[0016] 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.
[0017] 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;
[0018] S6. Calculate the gain coefficient of the equalizer based on the channel response at each frequency point;
[0019] 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;
[0020] S8. Perform FFT transformation on the reconstructed frame data to obtain the frequency domain signal of the reconstructed frame;
[0021] S9. Multiply the reconstructed frame frequency domain signal by the gain coefficient of the equalizer to obtain the equalized reconstructed frame frequency domain signal.
[0022] 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.
[0023] S11. Demodulate the nth data of the recombined frame equalization signal, and modulate the demodulation result to obtain modulated data;
[0024] S12. Replace the nth data of the known pilot after zero-padding with the modulation data to obtain the updated known pilot;
[0025] 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.
[0026] S14. Return to step S2 to perform equalization processing on the next data frame to be equalized.
[0027] Furthermore, the data to be balanced includes several data frames to be balanced; there is a guard interval between each data frame to be balanced.
[0028] Furthermore, 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.
[0029] The beneficial effects of this invention are as follows: Through a special frame format design, it can dynamically reflect the differences in interference caused by the channel to different frame data, and quickly update the channel gain coefficient of the equalizer. By using time-domain zero-padding, the number of points in the channel estimation is increased, resulting in a more accurate estimation of the channel's true frequency response compared to traditional channel estimation methods. The combination of cyclic feedback and pre-decision allows for dynamic adjustment of the estimated channel frequency response and the equalizer's gain coefficient, further reducing the impact of inter-symbol interference and minimizing colored noise in digital demodulation. The signal decision result after equalization is more accurate, and compared to traditional frequency-domain channel equalization methods, the demodulated data is closer to the actual data of the transmitted signal.
[0030] An improvement to a channel equalization method includes the following steps:
[0031] A1. Obtain the data frame to be balanced;
[0032] 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.
[0033] 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.
[0034] 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.
[0035] A5. Demodulate the nth data point of the mean result, and then modulate the demodulated result to obtain the modulated data;
[0036] A6. Update the known pilot signals using the modulated data;
[0037] A7. Update the gain coefficient of the equalizer based on the updated received pilot and the updated known pilot;
[0038] 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.
[0039] Furthermore, updating the receiving pilot in A3 specifically involves: keeping the length of the receiving pilot unchanged, and placing the current data to be equalized at the beginning of the receiving pilot.
[0040] Furthermore, updating the known pilot in A6 specifically involves keeping the length of the known pilot unchanged and placing the modulated data at the beginning and end of the known pilot.
[0041] Furthermore, A7 specifically refers to:
[0042] 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.
[0043] 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.
[0044] A703. The received pilot sequence and the known pilot sequence are used as the first frequency point values of the received pilot and the known pilot, respectively, to calculate the single-point channel response and obtain the single-point channel response result.
[0045] 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.
[0046] A705. Recalculate the gain coefficient corresponding to the first frequency point of the equalizer based on the updated channel response sequence.
[0047] The beneficial effects of this invention are as follows: Through a special frame format design, it dynamically reflects the differences in interference caused by the channel to different data, quickly updates the channel gain coefficient of the equalizer, and adopts a combination of cyclic feedback and pre-decision to dynamically adjust the estimated value of the channel frequency response and the gain coefficient of the equalizer, further reducing the impact of inter-symbol interference and the influence of colored noise in digital demodulation. Furthermore, when only a small amount of data changes, the improved method, through the butterfly computational structure of the FFT algorithm, avoids recalculating the overall frequency domain information of the data. This greatly simplifies the channel response and equalizer coefficient calculation process while effectively improving the performance of the equalization algorithm, significantly saving limited computing resources in practical engineering deployments. Attached Figure Description
[0048] Figure 1 This is a schematic diagram of a channel equalization system architecture with cyclic decision feedback proposed in this invention.
[0049] Figure 2 This is a schematic diagram of a channel equalization method with cyclic decision feedback in an embodiment.
[0050] Figure 3 This is a schematic flowchart illustrating an improved channel equalization method with cyclic decision feedback in an embodiment.
[0051] Figure 4 This is a schematic diagram of the frame structure in the embodiment.
[0052] Figure 5 This is a schematic diagram of the structure for quickly solving the equilibrium result in the embodiment. Detailed Implementation
[0053] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.
[0054] like Figure 1 As shown, in one embodiment of the present invention, a channel equalization system with cyclic decision feedback includes:
[0055] 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.
[0056] 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.
[0057] 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.
[0058] The known pilot signals are a sequence pre-set by both communicating parties.
[0059] The data frame to be equalized includes a received pilot and a data segment to be equalized.
[0060] In this embodiment, Figure 1 In This is the zero-padded receiving pilot; The known pilot signal after zero-padding; For reassembling frame data; For the reassembled frame data nOne 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.
[0061] Example 2
[0062] like Figure 2 As shown, a channel equalization method includes:
[0063] S1. Obtain the data to be equalized and determine the known pilot frequencies;
[0064] S2. Extract the data frames to be balanced sequentially from the data to be balanced;
[0065] 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.
[0066] 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.
[0067] 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;
[0068] S6. Calculate the gain coefficient of the equalizer based on the channel response at each frequency point;
[0069] 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;
[0070] S8. Perform FFT transformation on the reconstructed frame data to obtain the frequency domain signal of the reconstructed frame;
[0071] S9. Multiply the reconstructed frame frequency domain signal by the gain coefficient of the equalizer to obtain the equalized reconstructed frame frequency domain signal.
[0072] 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.
[0073] S11. Demodulate the nth data of the recombined frame equalization signal, and modulate the demodulation result to obtain modulated data;
[0074] S12. Replace the nth data of the known pilot after zero-padding with the modulation data to obtain the updated known pilot;
[0075] 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.
[0076] S14. Return to step S2 to perform equalization processing on the next data frame to be equalized.
[0077] 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.
[0078] 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.
[0079] 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.
[0080] 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.
[0081] 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.
[0082] 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.
[0083] 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.
[0084] 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.
[0085] 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.
[0086] 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.
[0087] 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.
[0088] 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.
[0089] This explanation uses a pilot p1 length of 128 and a data segment y1 length of 512 as an example:
[0090] B1. Obtain 1 frame of data (assuming it is the first frame): "Pilot p1 + Data segment y1 + Guard interval GI", without guard interval GI;
[0091] 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.
[0092] B3. Calculate the coefficients of the equalizer (LS, MMSE equalization method) using the channel response H of this 512 frequency point;
[0093] 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;
[0094] 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.
[0095] B6. Replace the zero-padded data at the corresponding positions in the known pilot sequence Tx_p with the remodulated data;
[0096] 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).
[0097] B8. Repeat steps B4-B7 until data segment y1 is balanced.
[0098] 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.
[0099] Example 3
[0100] like Figure 3 As shown, an improvement to the channel equalization method includes the following steps:
[0101] A1. Obtain the data frame to be balanced;
[0102] 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.
[0103] 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.
[0104] 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.
[0105] A5. Demodulate the nth data point of the mean result, and then modulate the demodulated result to obtain the modulated data;
[0106] A6. Update the known pilot signals using the modulated data;
[0107] A7. Update the gain coefficient of the equalizer based on the updated received pilot and the updated known pilot;
[0108] 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.
[0109] 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.
[0110] 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.
[0111] 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.
[0112] 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.
[0113] Specifically, A7 refers to:
[0114] 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.
[0115] 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.
[0116] A703. The received pilot sequence and the known pilot sequence are used as the first frequency point values of the received pilot and the known pilot, respectively, to calculate the single-point channel response and obtain the single-point channel response result.
[0117] 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.
[0118] A705. Recalculate the gain coefficient corresponding to the first frequency point of the equalizer based on the updated channel response sequence.
[0119] 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.
[0120] 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.
[0121] In this embodiment, the description is similarly based on pilot p1 having a length of 128 and data segment y1 having a length of 512:
[0122] 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;
[0123] C2. Perform a 128-point FFT transformation 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.
[0124] 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.
[0125] 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).
[0126] 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).
[0127] 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.
[0128] C71. Sum all the data of the updated known pilot sequence Tx_p in C6, and denote it as Sum_tx_p_1;
[0129] C72. Sum all the data in C4 after shifting in the first data y1_1, and record the result as Sum_y1_1;
[0130] 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).
[0131] 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.
[0132] 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).
[0133] C8. Repeat steps C4-C7 until the data segment y1 is balanced.
[0134] 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 channel equalization system with cyclic decision feedback, characterized in that, include: The reassembled frame module is used to extract the data frame to be equalized, pad the length of the known pilot and the received pilot in the data frame to be equalized with zeros until it is the same as the data segment to be equalized in the data frame, and replace the nth data of the zero-padding received pilot with the nth data of the data segment to be equalized to obtain the reassembled frame data; using the zero-padding known pilot and the zero-padding received pilot, the channel response in the initial state is calculated, and the gain coefficient of the equalizer is updated using the channel response in the initial state; The nth data of the known pilot after zero-padding is replaced with the modulation data to obtain 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.
2. The channel equalization system with cyclic decision feedback according to claim 1, characterized in that, The known pilot signals are a sequence pre-set by both communicating parties.
3. The channel equalization system with cyclic decision feedback according to claim 1, characterized in that, The data frame to be equalized includes a received pilot and a data segment to be equalized.
4. A channel equalization method for a channel equalization system based on the cyclic decision feedback of any one of claims 1-3, characterized in that, include: 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.
5. The channel equalization method according to claim 4, characterized in that, The data to be balanced includes several data frames to be balanced; there is a guard interval between each data frame to be balanced.
6. The channel equalization method according to claim 4, characterized in that, 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 channel equalization method, characterized in that, 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 point of the mean result is the equilibrium data corresponding to the current data to be balanced. 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.
8. The channel equalization method according to claim 7, characterized in that, 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.
9. The channel equalization method according to claim 7, characterized in that, 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.
10. The channel equalization method according to claim 7, characterized in that, 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. The received pilot sequence and the known pilot sequence are used as the first frequency point values of the received pilot and the known pilot, respectively, to calculate the single-point channel response and 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.