A pre-equalization compensation method for terminals with known transmitter IQI and nonlinear parameters in a TDD-OFDM direct conversion system.

Abstract

This invention discloses a pre-equalization compensation method for a terminal in a TDD-OFDM direct conversion system with known transmitter IQI and nonlinear parameters. The method includes: the base station cyclically sending P different training sequences to the terminal; the terminal using a sliding window least squares-based channel state information tracking algorithm to track and estimate the channel state information, where δ is a threshold representing the correlation of the channel within adjacent symbol periods; and pre-equalization compensation performed directly at the terminal. Using the pre-equalization compensation algorithm, a modified data sequence can be obtained and sent to the base station. This invention considers three factors in wireless communication systems: I-path and Q-path imbalance caused by the non-ideal nature of analog components in RF transceivers, power amplifier nonlinearity, and multipath channels. It enables distributed processing within the cell, effectively reducing the demodulation burden on the base station and has wider applicability than traditional compensation methods.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of wireless communication, and relates to a pre-equalization compensation method for terminal known transmitter IQI (In-phase Quadrature Imbalance) and non-linear parameters in a TDD-OFDM (Time Division Duplexing-Orthogonal Frequency Division Multiplexing) direct frequency conversion system. BACKGROUND

[0002] In a wireless device, due to the influence of non-linear components in the transmitter radio frequency front-end circuit, the signal emitted by the device is distorted; in addition, the differences in components and factory production cause these distortions to have individual differences, which have long-term stability and uniqueness, mainly including IQI and non-linear effects of front-end power amplifiers. If these characteristics are not compensated in the communication system, the transmission performance of the communication system will be reduced.

[0003] The traditional transmitter compensation method for OFDM signals mainly compensates for individual IQI or individual non-linear effects, usually adopts linear (least squares, minimum mean square error, adaptive equalization) or non-linear (maximum likelihood) algorithms, and performs equalization compensation at the receiving end. However, the above prior art has the following disadvantages: it does not simultaneously consider IQI, non-linear effects of front-end power amplifiers, and multipath coefficients, and there is no pre-equalization method for TDD systems, and there is a lack of distributed processing method in this case. SUMMARY

[0004] The present application aims to overcome the defects of the prior art and provide a pre-equalization compensation method for terminal known transmitter IQI and non-linearity in a TDD-OFDM direct frequency conversion system, which can, in the case that the terminal knows the transmitter IQI and non-linear parameters in the TDD-OFDM system, through pre-equalization compensation operation, enable the receiving end to directly obtain the real transmitted data sequence without demodulation, thereby effectively reducing the demodulation burden of the base station.

[0005] To solve the above technical problems, the present application adopts the following technical solutions.

[0006] A pre-equalization compensation method for terminal known transmitter IQI and non-linear parameters in a TDD-OFDM direct frequency conversion system adopts two core algorithms: a channel state information tracking algorithm based on a sliding window least squares for estimating a channel, and a pre-equalization compensation algorithm for pre-equalization compensation of transmitted data;

[0007] In the TDD-OFDM direct frequency conversion system, let dB (n) represents the baseband training sequence sent from the base station to the terminal, and the signal received by the terminal is...

[0008]

[0009] In equation (1), h(n) is the multipath channel coefficient, and w dwon_r (n) is additive white Gaussian noise; and It is the IQI coefficient of the terminal receiver, and g represents the amplitude deviation of the terminal receiver, and θ′ represents the phase deviation of the terminal receiver; adding the received signal and its conjugate gives...

[0010]

[0011] Due to the conjugate symmetry of the Fourier transform, DFT{x * (n)}=X * (Nk), where DFT represents the Discrete Fourier Transform, which can be calculated using FFT; applying the DFT to both sides of equation (2) and using conjugate symmetry, we obtain

[0012]

[0013] Where D B (k), H(k), W down_r (k) and Y down_r (k) represent d respectively B (n), h(n), w dwon_r (n) and y down_r The discrete Fourier transform of (n), where Re{·} denotes taking the real part; it can be seen from equation (3) that when sending P (P≥2) different sequences, we can obtain There are P independent linear equations, each of which can estimate two channel frequency domain variables; considering the time-varying nature of the wireless channel, the length of the P training sequences must not exceed the channel coherence time length T. c

[0014]

[0015] Among them T+T cp This represents the length of an OFDM symbol, where T is the time length of the OFDM symbol itself. cp It is the duration of the cyclic prefix; f m It is a Doppler frequency shift; to better track performance, P consecutive but different training sequences can be sent cyclically within the channel coherence time, and the channel coefficients can be estimated using the sliding window least squares algorithm; by introducing the symbol period index i, equation (3) can be rewritten in matrix form to obtain

[0016] In formula (5)

[0017]

[0018]

[0019] in, and express The first and second columns; x mod y means x divided by y and taking the remainder; Im{·} means taking the imaginary part;

[0020] Obtain the channel frequency domain coefficients for the i-th symbol period.

[0021]

[0022] Because swapping rows or columns of a matrix does not change its singular values, It is a fixed value;

[0023] make:

[0024] so

[0025]

[0026] A k (i) represents A k The i-th row; to avoid large estimation errors, A is needed. k The condition number is relatively small;

[0027] The method includes the following steps:

[0028] Step 1: The base station sends P different training sequences to the terminal in a loop;

[0029] Step 2: At the terminal, use a channel state information tracking algorithm based on sliding window least squares to track and estimate channel state information.

[0030] Step 3, when the conditions are met By performing pre-equalization compensation directly at the terminal and utilizing the pre-equalization compensation algorithm, a modified data sequence can be obtained. And send it to the base station; δ is a threshold that represents the correlation of the channel within adjacent symbol periods.

[0031] Specifically, in step two, the channel state information is tracked and estimated at the terminal using a channel state information tracking algorithm based on sliding window least squares. The process includes:

[0032] Initialization: i = P-1, given training sequence d B(n), the baseband y after direct downconversion. down_r,i (t)

[0033] The baseband signal y after passing through the analog-to-digital converter down_r,i (n) Remove the CP (Cyclic Prefix), then perform a serial-to-parallel transformation, and calculate.

[0034] Offline computing and

[0035] For i≥P

[0036] calculate

[0037] renew

[0038] i = i + 1;

[0039] End

[0040] because It is a constant value that can be calculated offline without participating in the iteration process; and the iteration process only needs to update Υ. k The computational complexity is very small, so the entire channel state information tracking algorithm based on sliding window least squares is highly efficient; furthermore, in the coherence time T... c It has a certain channel tracking capability.

[0041] Specifically, in step three, if the terminal transmits a baseband data sequence of length N, d = [d(0), d(1), d(2)...d(N-1)] T After passing through the nonlinear effects of the transmitter IQI and the RF front-end power amplifier, the baseband signal received by the receiver is...

[0042] r=(Sb)*h+n (9)

[0043] Among them, [S] i,m =|s(i)| 2(m-1) s(i)=ud(i)+vd * (i); b = [b0,...,b M-1 ] T Here, n is the nonlinear coefficient of the terminal transmitter, h is the additive noise vector, and v and u are the IQI coefficients of the terminal transmitter, defined by the following formula:

[0044] u = cos(θ / 2) + jαsin(θ / 2)

[0045] v=αcos(θ / 2)+jsin(θ / 2) (10)

[0046] Where α represents the amplitude deviation of the terminal transmitter, and θ represents the phase deviation of the terminal transmitter.

[0047] Furthermore, once the terminal knows its transmitter IQI and nonlinear coefficients, and estimates the channel using a channel state information tracking algorithm based on sliding window least squares, a pre-equalization compensation algorithm can be used to modify the actual data sequence to be sent to the base station. The baseband signal received by the base station is the actual data.

[0048] Specifically, in step three, the pre-equalization compensation algorithm is used to obtain a modified data sequence d, which is then sent to the base station. The process includes:

[0049] Given: the actual data sequence d to be transmitted, length N; nonlinear coefficients b, length M; and multipath frequency domain coefficients estimated by a channel state information tracking algorithm based on sliding window least squares.

[0050] Removing the effects of multipath channels: Where F is the Fourier matrix. This represents the element-wise division of the X and Y vectors, and Fd can be calculated using FFT; in this case,

[0051]

[0052] In equation (11),

[0053] Equation (11) can be written in terms of d i =N independent equations of d(i),

[0054] or

[0055]

[0056] in

[0057]

[0058] Φx m =bx m (u x d xi -u y d yi +v x d xi +v y d yi )-by m (u x d yi +u y d xi -vx d yi +v y d xi )

[0059] Φy m =by m (u x d xi -u y d yi +v x d xi +v y d yi )+bx m (u x d yi +u y d xi -v x d yi +v y d xi )

[0060]

[0061] Equation (12) is N terms relating to d i The real part d xi and the imaginary part d yi The system of two nonlinear equations can be solved using the Newton-Raphson method; the subscript i is omitted, and each baseband data to be transmitted is written in vector form d. Δ =[d x ,d y ] T

[0062] d Δ k+1 =d Δ k -J(d Δ k ) -1 F(d Δ k (14)

[0063] In equation (14), It is a Jacobi matrix, where

[0064]

[0065] and

[0066]

[0067] Therefore, the baseband signal received by the base station from the terminal is the actual data sequence d.

[0068] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0069] 1. This invention discloses a method in a TDD-OFDM direct frequency conversion system where, given the terminal's transmitter IQI and nonlinearity, the downlink channel coefficients are estimated and the uplink data is pre-equalized at the terminal, so that the baseband signal received by the base station is the actual transmitted data. This enables distributed processing of terminals within the cell, reducing the demodulation burden on the base station and demonstrating high application value.

[0070] 2. This invention comprises two core algorithms: a channel state information tracking algorithm based on sliding window least squares and a pre-equalization compensation algorithm. During channel estimation, multiple different sequences are transmitted cyclically, resulting in high channel estimation accuracy, low computational cost of the least squares solution, and a certain tracking capability within the coherence time. The pre-equalization compensation algorithm offers the advantages of fast parallel computation.

[0071] 3. This invention considers four factors in wireless communication systems caused by the non-ideal nature of the analog devices at the front end of the RF transmitter: transmitter IQI, transmitter power amplifier nonlinearity, terminal receiver IQI, and multipath channel. It has a wider range of applicability than traditional compensation methods. Attached Figure Description

[0072] Figure 1 This is a flowchart of a pre-equalization compensation method for a TDD-OFDM direct frequency conversion system where the transmitter IQI and nonlinear parameters are known at the terminal.

[0073] Figure 2 This is a block diagram of a terminal receiver with IQI according to an embodiment of the present invention.

[0074] Figure 3 This is a block diagram of a terminal transmitter with IQI and a nonlinear power amplifier according to an embodiment of the present invention.

[0075] Figures 4(a)-4(d)This is a performance diagram of a channel state information tracking algorithm based on sliding window least squares according to an embodiment of the present invention. It shows a comparison of channel tracking diagrams for the amplitude coefficients of five channels under different channel jitter (stable channel) and signal-to-noise ratios when the number of training sequences P = 7. In Figure 4(a), the root mean square (RMS) of the channel is 0.02, and the SNR is 5 dB; in Figure 4(b), the RMS is 0.02, and the SNR is 20 dB; in Figure 4(c), the RMS is 0.005, and the SNR is 5 dB; in Figure 4(d), the RMS is 0.005, and the SNR is 20 dB. In the figures, h1_true, h2_true, h3_true, h4_true, and h5_true represent the true values ​​of the first to fifth channel tap coefficients, respectively; h1_SWLS, h2_SWLS, h3_SWLS, h4_SWLS, and h5_SWLS represent the estimated values ​​of the first to fifth channel tap coefficients obtained through the channel state information tracking algorithm based on sliding window least squares, respectively. Figures 4(a)-4(d) As can be seen, the difference in signal-to-noise ratio did not significantly change the tracking performance. This performance is mainly affected by the jitter of the channel itself. This is because using the P training sequence to track the channel effectively suppresses additive noise, while the time-varying nature of the P training sequence itself has a significant cumulative impact on the tracking accuracy of the current symbol period.

[0076] Figure 5 This is a schematic diagram illustrating the performance of the pre-equalization mean square error (mean square error between the baseband data received by the terminal after pre-equalization and the real baseband data) under different signal-to-noise ratios according to an embodiment of the present invention. A selection was made... Figures 4(a)-4(d) The method transmits four baseband data points. The nonlinear coefficients are b = [1, 0.3, 0.2 + j0.3, -j0.1], and the IQI coefficients are v(5°, 0.2) and u(5°, 0.2). Experimental results show that after pre-equalization, the error between the baseband data received by the base station and the actual data is small, especially when the signal-to-noise ratio is greater than 5 dB, the accuracy is even higher, indicating that this method has reliable demodulation performance. Detailed Implementation

[0077] This invention considers a TDD-OFDM direct conversion system where the terminal knows its transmitter IQI and nonlinearity. It estimates channel coefficients via downlink and performs pre-equalization compensation at the terminal, achieving distributed processing within the cell and reducing the demodulation burden on the base station. This invention has high practical value. It includes two core algorithms: a channel state information tracking algorithm based on sliding window least squares and a pre-equalization compensation algorithm. The former is used to estimate the channel, and the latter is used to pre-equalize and compensate transmitted data.

[0078] In the aforementioned TDD-OFDM direct frequency conversion system, let d B (n) represents the baseband training sequence sent from the base station to the terminal, and the signal received by the terminal is...

[0079]

[0080] In equation (1), h(n) is the multipath channel coefficient, and w dwon_r (n) is additive white Gaussian noise. and It is the IQI coefficient of the terminal receiver, and has Where g represents the amplitude deviation of the terminal receiver, and θ′ represents the phase deviation of the terminal receiver, such as Figure 2 As shown. Adding the terminal's received signal to its conjugate yields...

[0081]

[0082] Due to the conjugate symmetry of the Fourier transform, DFT{x * (n)}=X * (Nk), where DFT represents the Discrete Fourier Transform, which can be calculated using FFT (Fast Fourier Transform). Applying conjugate symmetry to the DFT of both sides of equation (2) yields...

[0083]

[0084] Where D B (k), H(k), W down_r (k) and Y down_r (k) represent d respectively B (n), h(n), w dwon_r (n) and y down_r The discrete Fourier transform of (n) is given by Re{·}, where Re represents taking the real part. From equation (3), it can be seen that when sending P (P≥2) different sequences, the following can be obtained: There are P independent linear equations, each of which can estimate two channel frequency domain variables. Considering the time-varying nature of the wireless channel, the length of the P training sequences must not exceed the channel coherence time length T. c

[0085]

[0086] Among them T+T cp This represents the length of an OFDM symbol (T is the time length of the OFDM symbol itself, T...). cp (This is the duration of the cyclic prefix), f m It is the Doppler frequency shift. To improve tracking (estimation) performance, P consecutive but different training sequences can be sent cyclically within the channel coherence time, and the channel coefficients can be estimated using the sliding window least squares algorithm. Introducing the symbol period index i, equation (3) can be rewritten in matrix form to obtain

[0087] In formula (5)

[0088]

[0089] in, and express The first and second columns; x mod y means x divided by y and taking the remainder; Im{·} means taking the imaginary part.

[0090] Obtain the channel frequency domain coefficients for the i-th symbol period.

[0091]

[0092] Because swapping rows or columns of a matrix does not change its singular values, It is a fixed value. Let

[0093]

[0094] so

[0095]

[0096] A k (i) represents A k The i-th row. To avoid large estimation errors, A is needed. k The condition number is relatively small.

[0097] The steps of the channel state information tracking algorithm based on sliding window least squares are as follows:

[0098] Initialization: i = P-1, given training sequence d B (n), the baseband y after direct downconversion. down_r,i (t)

[0099] The baseband signal y after passing through the analog-to-digital converter down_r,i (n) Remove the CP (Cyclic Prefix), then perform a serial-to-parallel transformation, and calculate.

[0100] Offline computing and

[0101] For i≥P

[0102] calculate

[0103] renew

[0104] i = i + 1;

[0105] End

[0106] because It is a constant value that can be calculated offline without participating in the iteration process; and the iteration process only needs to update Υ. k The computational complexity is very small, so the entire channel state information tracking algorithm based on sliding window least squares is highly efficient; furthermore, in the coherence time T... c It has a certain channel tracking capability, such as Figures 4(a)-4(d) As shown.

[0107] If the terminal transmits a baseband data sequence of length N, d = [d(0), d(1), d(2)...d(N-1)] T After passing through the nonlinear effects of the transmitter IQI and the RF front-end power amplifier, the baseband signal received by the receiver is...

[0108] r=(Sb)*h+n (9)

[0109] Among them, [S] i,m =|s(i)| 2(m-1) s(i)=ud(i)+vd * (i). b = [b0,...,b M-1 ] T is the nonlinear coefficient of the terminal transmitter, n is the additive noise vector, and h is the multipath channel coefficient vector. v and u are the IQI coefficients of the terminal transmitter, defined by the following formula:

[0110] u = cos(θ / 2) + jαsin(θ / 2)

[0111] v=αcos(θ / 2)+jsin(θ / 2) (10)

[0112] Where α represents the amplitude deviation of the terminal transmitter, and θ represents the phase deviation of the terminal transmitter, such as... Figure 3 As shown. After the terminal knows its transmitter IQI and nonlinear coefficients, and estimates the channel using a channel state information tracking algorithm based on sliding window least squares, a pre-equalization compensation algorithm can be used to modify the actual data sequence to be sent to the base station. The baseband signal received by the base station is the actual data.

[0113] In conclusion, the method proposed in this invention includes the following steps:

[0114] The first step is for the base station to send P different training sequences to the terminal in a loop;

[0115] The second step involves using a channel state information tracking algorithm based on sliding window least squares at the terminal to track and estimate the channel state information.

[0116] Third step, when the conditions are met By performing pre-equalization compensation directly at the terminal and utilizing the pre-equalization compensation algorithm, a modified data sequence can be obtained. And send it to the base station; δ is a threshold that represents the correlation of the channel within adjacent symbol periods.

[0117] In this way, the base station receives the baseband signal sent by the terminal, which is the actual data sequence d.

[0118] The pre-equalization compensation algorithm steps are as follows:

[0119] Given: the actual data sequence d to be transmitted, length N; nonlinear coefficients b, length M; and multipath frequency domain coefficients estimated by a channel state information tracking algorithm based on sliding window least squares.

[0120] Removing the effects of multipath channels: Where F is the Fourier matrix. This represents the element-wise division of the X and Y vectors; here, Fd can be calculated using FFT. At this point,

[0121]

[0122] In equation (11),

[0123] Rewrite equation (11) in terms of d i =N independent equations of d(i),

[0124] or

[0125]

[0126] in

[0127]

[0128] Φx m =bx m (u x d xi -u y d yi +v x d xi +v y d yi )-by m (u x d yi +u y d xi -v x d yi +v y d xi )

[0129] Φy m =by m (u x d xi -u y d yi +v x d xi +v y d yi )+bx m (u x d yi +u y d xi -v x d yi +v y d xi )

[0130]

[0131] Equation (12) is N terms relating to d i The real part d xi and the imaginary part d yi The system of two nonlinear equations can be solved using the Newton-Raphson method. The subscript i is omitted, and each baseband data to be transmitted is written in vector form d. Δ =[d x ,d y ] T

[0132] d Δ k+1 =d Δ k -J(d Δ k ) -1 F(d Δ k (14)

[0133] In equation (14), It is a Jacobi matrix, where

[0134]

[0135] and

[0136]

[0137] This pre-equilibrium compensation algorithm solves N independent systems of two nonlinear equations and can perform parallel computation, thus its operation speed is very fast.

[0138] The present invention will now be described in further detail with reference to the accompanying drawings.

[0139] like Figure 1 The diagram shown is a flowchart of a pre-equalization compensation method for a TDD-OFDM direct conversion system with known transmitter IQI and nonlinear parameters at the terminal, comprising the following steps:

[0140] Step 1: The base station sends P different training sequences to the terminal in a loop;

[0141] Step 2: At the terminal, use a channel state information tracking algorithm based on sliding window least squares to track and estimate channel state information.

[0142] Step 3, when the conditions are met By performing pre-equalization compensation directly at the terminal and utilizing the pre-equalization compensation algorithm, a modified data sequence can be obtained. And send it to the base station; δ is a threshold that represents the correlation of the channel within adjacent symbol periods.

[0143] In step two, the channel state information is tracked and estimated at the terminal using a channel state information tracking algorithm based on sliding window least squares. The process includes:

[0144] Initialization: i = P-1, given training sequence d B (n), the baseband y after direct downconversion. down_r,i (t)

[0145] The baseband signal y after passing through the analog-to-digital converter down_r,i (n) Remove CP, then perform serial-to-parallel transformation, and calculate.

[0146] Offline computing and

[0147] For i≥P

[0148] calculate

[0149] renew

[0150] i = i + 1;

[0151] End

[0152] In step three, the pre-equalization compensation algorithm can be used to obtain a modified data sequence. And send it to the base station, the process of which includes:

[0153] Given: the actual data sequence d to be transmitted, length N; nonlinear coefficients b, length M; and multipath frequency domain coefficients estimated by a channel state information tracking algorithm based on sliding window least squares.

[0154] Removing the effects of multipath channels: Where F is the Fourier matrix. This represents the element-wise division of the X and Y vectors; here, Fd can be calculated using FFT. At this point,

[0155]

[0156] In equation (11),

[0157] Rewrite equation (11) in terms of d i =N independent equations of d(i),

[0158] or

[0159]

[0160] in

[0161]

[0162] Φx m =bx m (u x d xi -u y d yi +v x d xi +v y d yi )-by m (u x d yi +u y d xi -v x d yi +v y d xi )

[0163] Φy m =by m (u x d xi -u y d yi +v x d xi +v y d yi )+bx m (u x d yi +uy d xi -v x d yi +v y d xi )

[0164]

[0165] Equation (12) is N terms relating to d i The real part d xi and the imaginary part d yi The system of two nonlinear equations can be solved using the Newton-Raphson method. The subscript i is omitted, and each baseband data to be transmitted is written in vector form d. Δ =[d x ,d y ] T

[0166] In equation (14), It is a Jacobi matrix, where

[0167]

[0168]

[0169] and

[0170]

[0171] Therefore, the baseband signal received by the base station from the terminal is the actual data sequence d.

[0172] In summary, this invention discloses a method for distributed processing within a cell in a TDD-OFDM direct conversion system. This method involves estimating channel coefficients via downlink and performing pre-equalization compensation at the terminal, given the terminal's known transmitter IQI and nonlinear coefficients. This reduces the demodulation burden on the base station and has high practical value. First, the base station cyclically sends multiple different training sequences to the terminal. The terminal estimates (tracks) the frequency domain channel information using a sliding window least squares-based channel state information tracking algorithm. Then, the terminal performs pre-equalization compensation on the uplink data before sending it back to the base station. This invention comprises two core algorithms: a sliding window least squares-based channel state information tracking algorithm and a pre-equalization compensation algorithm. During channel estimation, multiple different sequences are cyclically sent, effectively tracking the time-varying frequency domain channel within the coherence time. Subsequent pre-equalization processing can be performed in parallel, resulting in high efficiency. This method considers four factors in wireless communication systems: transmitter IQI caused by the non-ideal nature of the RF transmitter front-end analog devices, transmitter power amplifier nonlinearity, terminal receiver IQI, and multipath channels. This method has broader applicability than traditional compensation methods.

Claims

1. A pre-equalization compensation method for terminal known transmitter IQI and non-linear parameters in a TDD-OFDM direct conversion system, characterized in that, Two core algorithms are adopted: a channel state information tracking algorithm based on sliding window least square for estimating the channel and a pre-equalization compensation algorithm for pre-equalization compensation of the transmitted data; In the TDD-OFDM direct conversion system, let d B (n) represents the baseband training sequence sent by the base station to the terminal, and the signal received by the terminal is In Equation (1), h(n) is a multipath channel coefficient, w down_r (n) is additive white Gaussian noise; and is an IQ imbalance coefficient of the terminal receiver, and g represents an amplitude deviation of the terminal receiver, and θ' represents a phase deviation of the terminal receiver; and the terminal received signal is added to its conjugate Due to the conjugate symmetry of the Fourier transform DFT{x * (n)} = X * (N-k), where DFT denotes the discrete Fourier transform, which can be computed using the FFT; taking the DFT of both sides of equation (2) and using the conjugate symmetry gives Where D B (k), H(k), W down_r (k) and Y down_r (k) represent d respectively B (n), h(n), w down_r (n) and y down_r The discrete Fourier transform of (n), where Re{·} denotes taking the real part; from equation (3), it can be seen that when sending P different sequences, P≥2, we can obtain There are P independent linear equations, each of which can estimate two channel frequency domain variables; considering the time-varying nature of the wireless channel, the length of the P training sequences must not exceed the channel coherence time length T. c where T + T cp denotes an OFDM symbol length, T is the length of an OFDM symbol itself, T cp is the length of a cyclic prefix; f m is the Doppler shift; for better tracking performance, P consecutive but different training sequences can be transmitted in a cycle within the channel coherence time, and the channel coefficients can be estimated by using a sliding window least square algorithm; introducing the index i of the symbol period, equation (3) is rewritten in matrix form as in formula (5) wherein and denote the 1st and 2nd columns; x mod y denotes x divided by y with remainder; Im{•} denotes taking imaginary part; The channel frequency domain coefficient of the ith symbol period is obtained Because the rows or columns of an interchange matrix do not change its singular values, so is a constant value; Let Thus A k (i) denotes the i-th row of A k To avoid large estimation errors, the condition number of A k should be small. The method comprises the following steps: Step one, the base station cyclically transmits P different training sequences to the terminal; Step two, the terminal utilizes a channel state information tracking algorithm based on sliding window least square to track the estimated channel state information Step three, when meeting By directly pre-equalization compensation in the terminal, using pre-equalization compensation algorithm, a modified data sequence And send to the base station; δ is a threshold, indicating the correlation size of adjacent symbol period channel; wherein: [S] i,m = |s(i)| 2(m-1) , s(i) = ud(i) + vd * (i); b = [b0,..., b M-1 ] T is the terminal transmitter non-linearity coefficient, n is the additive noise vector, h is the multipath channel coefficient vector; v and u are the terminal transmitter IQI coefficients defined by: u = cos(θ / 2) + jαsin(θ / 2) v = αcos(θ / 2) + jsin(θ / 2) (10) Wherein, α represents the amplitude deviation of the terminal transmitter, and θ represents the phase deviation of the terminal transmitter.

2. The pre-equalization compensation method of terminal known transmitter IQI and non-linear parameters in a TDD-OFDM direct conversion system according to claim 1, characterized in that, In the third step, if the terminal transmits a baseband data sequence d = [d(0), d(1), d(2)...d(N-1)] of length N T The baseband signal received at the receiving end is affected by the non-linear effects of the transmitter IQI and the radio frequency front-end power amplifier r = (Sb) * h + n (9).

3. The pre-equalization compensation method of terminal known transmitter IQI and non-linear parameters in a TDD-OFDM direct conversion system according to claim 1, characterized in that, After the terminal estimates the channel by the channel state information tracking algorithm based on sliding window least square and knows the IQI and the non-linear coefficient of the transmitter, the pre-equalization compensation algorithm is adopted to modify the real data sequence to be transmitted and sent to the base station, and the baseband signal received by the base station is the real data.

4. The pre-equalization compensation method of terminal known transmitter IQI and non-linear parameters in a TDD-OFDM direct conversion system according to claim 1, characterized in that, In step three, the pre-equalization compensation algorithm is used to obtain a modified data sequence and sends to the base station, the process includes: Given: real data sequence d to be transmitted, length N; nonlinear coefficient b, length M; multipath frequency-domain coefficients estimated by a channel state information tracking algorithm based on sliding window least squares Removing the effect of multipath channels: where F is the Fourier matrix, denotes element-wise division of the elements of the X vector and the Y vector, and Fd can be computed using an FFT; in this case, in formula (11), Write equation (11) as N independent equations in d i = d(i) or Wherein Q = |ud i +vd i * | 2 Φx m = bx m (u x d xi - u y d yi + v x d xi + v y d yi ) - by m (u x d yi + u y d xi - v x d yi + v y d xi ) Φy m = by m (u x d xi -u y d yi +v x d xi +v y d yi )+bx m (u x d yi +u y d xi -v x d yi +v y d xi ​ Equation (12) is a system of N binary nonlinear equations in the real part d i and the imaginary part d xi of d yi , which can be solved by the Newton-Raphson method; omitting the subscript i, and writing each baseband data to be transmitted as a vector d Δ = [d x , d y ] T d Δ k+1 = d Δ k - J(d Δ k ) -1 F(d Δ k ) (14) In formula (14), is a Jacobian matrix, wherein And Thus, the baseband signal sent by the terminal and received by the base station is the real data sequence d.

Images