A method, apparatus and system for single-stage joint frequency offset and two-stage DC calibration

By employing a single-stage joint frequency offset and two-stage DC calibration method, utilizing a shared training preamble and zero-mean sequence, and combining coarse and fine delay autocorrelation, the problem of mutual influence between DC offset and carrier frequency offset is solved, thereby improving the calibration accuracy and frequency offset estimation capability of the wireless communication system.

CN123123091APending Publication Date: 2026-11-17BEIJING BLUE WHALE INTELLIGENT TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611267932.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In wireless digital communication systems, how can we reduce the mutual influence between DC bias and carrier frequency offset, and reduce the impact of the residual non-zero mean after frequency offset compensation on subsequent channel estimation and frequency domain equalization?

Method used

A single-stage joint frequency offset and two-stage DC calibration method is adopted. By acquiring the I/Q baseband data of the shared training preamble, DC calibration is performed using the zero-mean sequence. The carrier frequency offset phase increment is calculated by combining the coarse and fine delay autocorrelation, and digital downconversion compensation is performed. Finally, DC calibration is performed.

Benefits of technology

It reduces the pull of DC bias on the relevant phase, improves calibration accuracy, and is suitable for communication scenarios with short data frames and limited burst reception delay, while taking into account the frequency offset acquisition range and estimation accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123091A_ABST
    Figure CN123123091A_ABST
Patent Text Reader

Abstract

This invention relates to the field of wireless communication technology and provides a method, apparatus, and system for single-shot joint frequency offset and two-stage DC calibration. The method includes: acquiring I / Q baseband data containing a shared training preamble; performing DC calibration based on I-channel and Q-channel DC estimates to obtain first-stage calibration data; obtaining coarse-level and fine-level correlation results; obtaining the phase increment corresponding to the carrier frequency offset; obtaining frequency offset calibration data; and performing DC calibration on the frequency offset calibration data to obtain second-stage calibration data. The advantage of this application is that it can simultaneously consider both frequency offset acquisition range and estimation accuracy; the two-stage DC calibration separately processes the main DC offset before correlation calculation and the residual mean after frequency offset correction, reducing the pull of DC offset on the correlation phase and avoiding directly retaining the first-stage finite window estimation error in the output data, thus improving the overall calibration accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of wireless communication technology, and particularly relates to a method, apparatus and system for single-stage joint frequency offset and two-stage DC calibration. Background Technology

[0002] In wireless digital communication systems, single-carrier frequency domain equalization is a common digital communication reception method. The transmitter sends data in the form of a single carrier, and the receiver processes the received data in data blocks, converts the time-domain data to the frequency domain using a fast Fourier transform, and calculates the equalization coefficients using the channel estimation results obtained from the training sequence, thereby compensating for signal distortion caused by multipath propagation.

[0003] In a single-carrier frequency-domain equalized receiver, after frame synchronization, the receiver typically needs to sequentially perform DC bias calibration and carrier frequency offset estimation before proceeding to channel estimation and frequency domain equalization. Current mainstream solutions usually include a repetitive training preamble in the data frame. After frame synchronization, the receiver extracts training data from the training preamble, averages the I-channel and Q-channel data respectively to obtain a DC bias estimate, and subtracts this estimate from the received data. Then, a single-delay autocorrelation is performed on the training data, and the carrier frequency offset estimate is obtained based on the phase of the correlation result. Frequency offset compensation is then performed on the received data using digital down-conversion. Some existing technologies also utilize different training fields to perform coarse and fine frequency offset estimations in stages, or correlate the received training data with a local pseudo-random sequence to estimate the frequency offset. After DC bias calibration and frequency offset compensation, channel estimation, frequency domain equalization, and demodulation are then performed on the received data.

[0004] Therefore, how to reduce the mutual influence between DC bias and carrier frequency offset, and reduce the impact of the residual non-zero mean after frequency offset compensation on subsequent channel estimation and frequency domain equalization, remains a technical problem that needs to be solved. Summary of the Invention

[0005] The purpose of this application is to provide a single-stage joint frequency offset and two-stage DC calibration method, which aims to solve the problem of reducing the impact of the residual non-zero mean after frequency offset compensation on subsequent channel estimation and frequency domain equalization while minimizing the mutual influence between DC offset and carrier frequency offset.

[0006] This application provides a method for single-stage joint frequency offset and two-stage DC calibration, the method comprising:

[0007] Acquire I / Q baseband data containing a shared training preamble, wherein the effective estimation segment of the shared training preamble is composed of multiple identical complex training blocks, and the I-path training sequence and Q-path training sequence of each complex training block are constructed as zero-mean sequences at the transmitter.

[0008] Based on the mean values ​​of the I-channel training data and the Q-channel training data within the effective estimation segment, the DC estimation values ​​of the I-channel and Q-channel are obtained respectively. Based on the DC estimation values ​​of the I-channel and Q-channel, the I / Q baseband data is DC calibrated to obtain the first-level calibration data.

[0009] Calculate the coarse-scale delay autocorrelation and fine-scale delay autocorrelation of the first-level calibration data to obtain the coarse-scale correlation result and the fine-scale correlation result. The delay length of the fine-scale delay autocorrelation is greater than the delay length of the coarse-scale delay autocorrelation.

[0010] Based on the ratio of the delay length of the fine-length delay autocorrelation to that of the coarse-length delay autocorrelation, the phase of the coarse-length correlation result is converted, and the phase of the fine-length correlation result is de-fuzzified using the converted phase as a reference to obtain the phase increment corresponding to the carrier frequency offset.

[0011] Based on the phase increment, digital down-conversion compensation is performed on the first-level calibration data to obtain frequency offset calibration data;

[0012] Based on the I-channel mean and Q-channel mean of the training data corresponding to the effective estimation segment in the frequency offset calibration data, DC calibration is performed on the frequency offset calibration data to obtain the second-level calibration data.

[0013] Another objective of this application is to provide a single-stage combined frequency offset and two-stage DC calibration device, the device comprising:

[0014] The baseband data acquisition module is used to acquire I / Q baseband data containing a shared training preamble. The effective estimation segment of the shared training preamble is composed of multiple identical complex training blocks. The I-path training sequence and Q-path training sequence of each complex training block are constructed as zero-mean sequences at the transmitter.

[0015] The first-level calibration module is used to obtain the DC estimation values ​​of the I-channel and Q-channel training data according to the mean values ​​of the I-channel training data and Q-channel training data in the effective estimation segment, respectively, and to perform DC calibration on the I / Q baseband data based on the DC estimation values ​​of the I-channel and Q-channel training data to obtain the first-level calibration data;

[0016] The delay autocorrelation module is used to calculate the coarse delay autocorrelation and fine delay autocorrelation of the first-level calibration data, and obtain the coarse correlation result and the fine correlation result. The delay length of the fine delay autocorrelation is greater than the delay length of the coarse delay autocorrelation.

[0017] The phase increment acquisition module is used to convert the phase of the coarse-level correlation result based on the ratio of the delay length of the fine-level delay autocorrelation to the delay length of the coarse-level delay autocorrelation, and to de-blur the phase of the fine-level correlation result using the converted phase as a reference to obtain the phase increment corresponding to the carrier frequency offset.

[0018] The frequency offset calibration module is used to perform digital down-conversion compensation on the first-level calibration data based on the phase increment to obtain frequency offset calibration data.

[0019] The secondary calibration module is used to perform DC calibration on the frequency offset calibration data based on the I-channel mean and Q-channel mean of the training data corresponding to the effective estimation segment in the frequency offset calibration data, so as to obtain the second-level calibration data.

[0020] Another objective of this application is to provide a communication system, the system including a transmitter and a receiver;

[0021] The sending end is used to send a data frame containing training data to the receiving end;

[0022] The receiving end is used to receive the data frame and perform the single joint frequency offset and two-stage DC calibration method as described above.

[0023] The single-shot joint frequency offset and two-stage DC calibration method provided in this application has the following advantages: it reduces the transmission resources occupied by the training field by utilizing a shared training preamble, making it suitable for communication scenarios with short data frames and limited burst reception delay; the coarse delay provides a larger unambiguous frequency offset range, while the fine delay improves the ability to distinguish small frequency offset differences; after recovering the integer phase of the fine delay using the coarse delay result, both the frequency offset capture range and estimation accuracy can be taken into account; the two-stage DC calibration processes the main DC offset before correlation operation and the residual mean after frequency offset correction, respectively, reducing the pull of the DC offset on the correlation phase and avoiding the direct retention of the first-stage finite window estimation error in the output data, thus improving the overall calibration accuracy. Attached Figure Description

[0024] Figure 1 A flowchart of a single-stage joint frequency offset and two-stage DC calibration method provided in this application embodiment;

[0025] Figure 2 A schematic diagram of a shared training preamble structure provided in an embodiment of this application;

[0026] Figure 3 A timing diagram for delayed alignment DC calibration provided in an embodiment of this application;

[0027] Figure 4 This is a structural block diagram of a single-stage combined frequency offset and two-stage DC calibration device provided in an embodiment of this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0029] It is understood that the terms "first," "second," etc., used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are used only to distinguish the first unit or module from another unit or module. For example, without departing from the scope of this application, the first unit may be referred to as the second unit, and similarly, the second module may be referred to as the first module.

[0030] like Figure 1 As shown, in one embodiment, a single-stage joint frequency offset and two-stage DC calibration method is proposed, which may specifically include the following steps:

[0031] Step S10: Obtain I / Q baseband data containing a shared training preamble. The effective estimation segment of the shared training preamble is composed of multiple identical complex training blocks. The I-path training sequence and Q-path training sequence of each complex training block are constructed as zero-mean sequences at the transmitter.

[0032] In this embodiment, the shared training preamble refers to setting a training preamble in the same frame of I / Q baseband data that is used jointly by DC bias estimation and carrier frequency offset estimation. The receiver extracts training data from the effective estimation segment of the shared training preamble. First, it performs a first-level DC bias estimation based on the training data. After completing the first-level DC calibration, it uses the training data corresponding to the effective estimation segment in the first-level calibration data to perform coarse-scale delay autocorrelation and fine-scale delay autocorrelation to obtain the phase increment corresponding to the carrier frequency offset. After completing the frequency offset calibration, it uses the training data corresponding to the effective estimation segment in the frequency offset calibration data to perform a second-level residual DC bias estimation.

[0033] In this embodiment, the zero-mean sequence refers to the ideal training sequence constructed and generated at the transmitting end. The transmitting and receiving ends pre-agree on information such as the content, repetition count, and position of the training block in the data frame, and the receiving end extracts the effective estimation segment accordingly. The effective estimation segment is a continuous data interval allowed for parameter estimation in the shared training preamble, and its start and end positions can be determined by the frame synchronization result, the training field start flag, and the preset training field length. This interval is set to avoid unstable samples at the receiving gain switching, filter establishment process, and frame boundaries, so as to prevent the transient response from entering the mean calculation and delay autocorrelation calculation in subsequent steps. The complex training block is a complex domain training block, and the zero-mean characteristic makes the I-channel mean and Q-channel mean of the ideal received training data both zero. When there is a DC bias in the received data caused by analog-to-digital converter offset, zero-IF mixer leakage, or analog front-end imbalance, the mean of the training data deviates from zero accordingly, and this deviation can be used to estimate the DC bias. Multiple training blocks have the same symbol arrangement. The coarse and fine delay lengths can be chosen as integer multiples of the training block length, ensuring that the training symbols before and after the delay are in the same position within the block. When calculating the delay autocorrelation, the modulation phase of the repeated training symbols is canceled out, and the remaining correlation phase mainly reflects the phase accumulation caused by the carrier frequency offset within the corresponding delay interval.

[0034] Step S20: Based on the mean values ​​of the I-channel training data and the Q-channel training data within the effective estimation segment, obtain the I-channel DC estimation value and the Q-channel DC estimation value respectively. Perform DC calibration on the I / Q baseband data based on the I-channel DC estimation value and the Q-channel DC estimation value to obtain the first-level calibration data.

[0035] In this embodiment, the mean of the training data within the effective estimation segment refers to the mean of the training data collected by the receiver after being affected by DC bias, noise, carrier frequency offset, etc. The first-stage DC calibration is performed before the delay autocorrelation calculation. A large DC bias will generate a correlation term of the DC bias itself and a cross term between the DC bias and the training signal in the correlation result. The correlation term of the DC bias itself is usually concentrated in the real part of the correlation result. When this term accounts for a high proportion, the phase of the correlation result will be biased towards zero, thus affecting the frequency offset estimation. The first-stage calibration first suppresses this interference, so that the subsequent correlation phase is mainly determined by the carrier frequency offset. Affected by carrier frequency offset, finite training length, quantization error, and noise, the DC bias estimate obtained at this time may still have residuals. The first-stage calibration undertakes the main DC suppression before the correlation operation, and the remaining error is estimated again by the second-stage DC calibration after frequency offset correction.

[0036] Step S30: Calculate the coarse-scale delay autocorrelation and fine-scale delay autocorrelation of the first-level calibration data to obtain the coarse-scale correlation result and the fine-scale correlation result. The delay length of the fine-scale delay autocorrelation is greater than the delay length of the coarse-scale delay autocorrelation.

[0037] In this embodiment, "coarse" and "fine" are relative terms and do not imply the need for two training pre-processing steps or the sequential execution of two frequency offset estimation stages. Coarse-stage delay autocorrelation uses a shorter delay length for autocorrelation, primarily to obtain a larger frequency offset capture range; fine-stage delay autocorrelation uses a longer delay length for autocorrelation, primarily to improve the resolution of frequency offset estimation. The coarse and fine stages are divided into two sets of frequency offset estimation parameters based on the delay length of the delay autocorrelation. Both process the same first-level calibration data, obtaining a complex correlation result by multiplying and accumulating the conjugates of the baseband samples before and after the delay. The difference lies in the time interval between the two baseband samples involved in the conjugate multiplication. A shorter delay interval results in a smaller cumulative phase generated by the carrier frequency offset, making it less likely to exceed the system's allowed carrier frequency offset range. and The phase boundary is such that it can maintain a large unambiguous frequency offset estimation range.

[0038] Since the correlated phase ultimately needs to be divided by a relatively short delay length, the frequency offset estimation error corresponding to the same phase measurement error is relatively large. Therefore, the coarse-grained result is mainly used to determine the frequency offset range and assist in the unambiguity resolution of the fine-grained correlated phase. A longer delay interval can cause a larger cumulative phase difference for the same carrier frequency offset. When the frequency offset values ​​of two carriers are close, the cumulative phase difference formed by them within the fine-grained delay interval will also amplify with the delay length. After dividing the fine-grained correlated phase by a longer delay length, the impact of the phase measurement error on the phase increment of each modulation symbol is smaller. The longer the fine-grained delay, the more the principal phase value spans a complete 2π / 2. The higher the probability of a cycle, the more necessary it is to use coarse-grained correlation phases to determine the number of integer cycles that should be recovered.

[0039] In this embodiment, coarse and fine correlation profiles can be used to obtain observation results of the same carrier frequency offset at two different delay scales. When the two profiles are used together, the coarse correlation result is first used to determine how many complete cycles the fine cumulative phase has lost. Then, the fine correlation result is deblurred, and finally, the phase increment per symbol is calculated using the deblurred fine cumulative phase. The receiver first uses the coarse correlation phase to determine the integer cycle interval of the fine correlation phase, then recovers the fine cumulative phase, and calculates the final phase increment using the recovered fine cumulative phase. The coarse correlation result does not directly replace the fine correlation result; the final phase increment is determined based on the recovered fine cumulative phase, which can better balance the frequency offset acquisition range and the frequency offset estimation accuracy.

[0040] Step S40: Based on the ratio of the delay length of the fine-scale delay autocorrelation to the delay length of the coarse-scale delay autocorrelation, the phase of the coarse-scale correlation result is converted, and the phase of the fine-scale correlation result is de-fuzzified using the converted phase as a reference to obtain the phase increment corresponding to the carrier frequency offset.

[0041] In this embodiment, defuzzification refers to determining the missing phases of the fine-grained correlation outside the principal value range. Number of integer cycles. Coarse results provide integer cycle positions, while fine results retain the higher resolution of long-delay measurements.

[0042] For example, coarse autocorrelation compares the current training sample with the training samples from the previous 128 symbols, while fine autocorrelation compares the current training sample with the training samples from the previous 512 symbols; the ratio of the delay lengths is 4. With the carrier frequency offset remaining constant, the cumulative phase caused by the frequency offset is proportional to the delay length. Therefore, the actual cumulative phase within the fine delay is theoretically four times the cumulative phase within the coarse delay. Multiplying the coarse phase by this ratio allows prediction of the cumulative phase that the carrier frequency offset should form during the fine delay period.

[0043] Step S50: Perform digital down-conversion compensation on the first-level calibration data based on the phase increment to obtain frequency offset calibration data.

[0044] In this embodiment, the first-level calibration data is written to a delay buffer, enabling coarse and fine-scale correlation operations, phase calculations, and integer cycle determination to be completed before the buffered data is output. The delay buffer only changes the output time of the data, without altering the order of baseband samples within the data frame. Digital down-conversion compensation refers to sample-by-sample phase rotation performed by a phase accumulator, a sine / cosine generation unit, and a complex multiplication unit. The phase accumulator updates the compensation phase once for each baseband sample received, ensuring that each sample receives the rotation amount corresponding to its timing position. This compensation can suppress persistent phase drift between training symbols and payload symbols caused by carrier frequency offset.

[0045] Step S60: Based on the I-channel mean and Q-channel mean of the training data corresponding to the effective estimation segment in the frequency offset calibration data, DC calibration is performed on the frequency offset calibration data to obtain the second-level calibration data.

[0046] In this embodiment, the training data corresponding to the effective estimation segment refers to the training samples that, after the first-level DC calibration, delay buffering, and digital down-conversion compensation, still have the same intra-frame position as the effective estimation segment selected in step S10. During data processing, the training segment flag and intra-frame sample number can be transmitted synchronously to relocate these samples in the frequency offset calibration data. Frequency offset correction realigns the phase relationship between repeated training blocks, reducing the impact of carrier frequency offset on the mean of the training data. The I-path mean and Q-path mean are recalculated, yielding the residual DC offset still present in the frequency offset calibration data after the first-level calibration. The second-level calibration processes the finite window error, frequency offset effect, fixed-point quantization error, and the residual mean offset at the output of the frequency offset correction module. After processing, the centers of the training data and payload data are closer to the origin, reducing the impact of DC residuals on synchronization, channel estimation, and symbol decision.

[0047] The method provided in this embodiment has the advantage of reducing the transmission resources occupied by the training field by sharing the training preamble, making it suitable for communication scenarios with short data frames and limited burst reception latency. The coarse-level delay provides a larger unambiguous frequency offset range, while the fine-level delay improves the ability to distinguish small frequency offset differences. After recovering the integer phase of the fine-level result using the coarse-level result, both the frequency offset capture range and estimation accuracy can be considered simultaneously. Two-stage DC calibration processes the main DC bias before correlation operations and the residual mean after frequency offset correction, respectively. This sequence reduces the pull of the DC bias on the correlation phase, avoiding the direct retention of the first-stage finite window estimation error in the output data. The delay buffer allows the estimated parameters to be aligned with the baseband samples of the same data frame without changing the original timing of the samples in the data frame.

[0048] As an embodiment of this application, this method can be used in a zero-IF burst receiver. The transmitter sets a shared training preamble consisting of eight identical complex training blocks at the beginning of the data frame. Each complex training block contains 128 modulation symbols, and the I-channel and Q-channel sequences of each training block satisfy the zero-mean condition. The receiver uses 128 modulation symbols as the coarse-length delay and 512 modulation symbols as the fine-length delay, with a ratio of 4. Each modulation symbol corresponds to four baseband samples.

[0049] Assuming the actual phase increment per modulation symbol caused by carrier frequency offset is 0.012 rad, the cumulative phase within the coarse delay interval is:

[0050] 128 × 0.012 = 1.536 rad

[0051] The actual cumulative phase within the fine-scale delay interval is:

[0052] 512 × 0.012 = 6.144 rad

[0053] Since 6.144 rad exceeds the principal phase range, the phase directly output by the fine-file correlation result is approximately:

[0054] 6.144−2π=−0.139rad,

[0055] After converting the coarse-related phase according to the ratio of delay length, we get:

[0056] 4 × 1.536 = 6.144 rad

[0057] Using this value as a reference, it can be determined that the phase of the fine-section needs to be compensated for a complete 2π cycle, and the recovered cumulative phase of the fine-section is 6.144 rad. Therefore:

[0058] Δθsym = 6.144 / 512 = 0.012 rad / modulation symbol

[0059] Δθsamp = 0.012 / 4 = 0.003 rad / baseband sample point

[0060] The digital downconversion module accumulates a compensation phase of -0.003 rad point-by-point according to the baseband sample number, rotating the I / Q baseband samples output from the delay buffer. If the frequency offset calibration data after the first-stage DC calibration still has an I-channel mean of 0.0015 and a Q-channel mean of -0.0012, the second-stage DC calibration uses these two means as residual DC offset estimates, subtracting them from the corresponding I-channel and Q-channel components. This further brings the center of the output data closer to the origin, providing frequency offset and DC offset corrected baseband data for subsequent synchronization, channel estimation, and demodulation.

[0061] In a preferred embodiment, the shared training preamble includes a guard segment and the effective estimation segment, the guard segment preceding the effective estimation segment; the complex training block is constructed to satisfy:

[0062]

[0063] The I-path training sequence and the Q-path training sequence each consist of different lengths. The pseudo-random sequence is obtained by mapping positive and negative 1, where bit 1 is mapped to +1 and bit 0 is mapped to -1, and the following conditions are met:

[0064]

[0065]

[0066] Repeat the retraining block continuously The resulting effective estimation segment is as follows:

[0067]

[0068]

[0069] Frame synchronization is performed on the I / Q baseband data. Based on the shared training preamble start position obtained from frame synchronization, the guard segment is skipped, and the effective estimation segment is truncated as a training window. The training window is used for both DC estimation and carrier frequency offset estimation. Indicates the first training block A complex training symbol, Indicates the first I-path training symbols for a complex training symbol, Indicates the first Q-path training symbols for a complex training symbol, Represents the imaginary unit. Indicates the symbol index within the complex training block. The symbol number representing the training sequence of path I. Indicates the symbolic index of the Q-path training sequence. Indicates the length of the complex training block. Indicates the number of repetitions of the training block. Indicates the length of the effective estimated segment. Indicates the length of the protected section. Indicates the length of the shared training leader.

[0070] In the embodiments of this application, such as Figure 2The diagram illustrates a shared training preamble structure, with the transmission or reception time sequence of each component arranged horizontally from left to right. The shared training preamble includes a guard segment and an effective estimation segment arranged sequentially. The guard segment is 64 modulation symbols long; the receiver skips this guard segment after determining the start position of the shared training preamble to reduce the impact of frame synchronization position deviation, receiver filter transient response, and automatic gain control state switching on subsequent parameter estimation. The effective estimation segment is 1536 modulation symbols long, formed by the continuous repetition of 24 complex training blocks of 64 modulation symbols each. The total length of the shared training preamble is 1600 modulation symbols. Each complex training block uses the same complex training sequence, with the I-path and Q-path components of the complex training sequence generated from different pseudo-random sequences. Bit 1 in the pseudo-random sequence is mapped to +1, bit 0 to -1, and the symbol sums of the I-path and Q-path training sequences within each complex training block are both zero. Repeated arrangement of identical complex training blocks ensures that training symbols separated by one or more complex training block lengths have the same sequence position, allowing the computation of autocorrelation results at different delay lengths. Autocorrelation results corresponding to shorter delays provide a larger frequency offset estimation range, while those corresponding to longer delays improve the resolution of the frequency offset estimation. Setting the zero mean of the I-channel and Q-channel training sequences reduces the impact of the DC component of the training sequence itself on the DC offset estimation, making the mean within the training window more reflective of the DC offset in the received data.

[0071] In this embodiment, when the training window's valid signal is active, the I-channel and Q-channel training samples within the training window are accumulated. After the training window ends, the I-channel DC estimate and Q-channel DC estimate are calculated based on the two accumulation results, and the two DC estimates are latched when the DC estimate valid signal is active. The original I-channel and Q-channel data are written to a delay buffer during training data accumulation. The length of the delay buffer covers the time required for training window accumulation, mean calculation, DC estimate latching, and control signal transmission. After the two DC estimates are latched, the delayed I-channel and Q-channel data are read in the original sample order, and the latched I-channel and Q-channel DC estimates are subtracted from them respectively. The compensation in the figure begins with the DC bias subtraction starting from the first delayed-aligned baseband sample. The entire frame delay alignment length represents the fixed delay experienced by the corresponding original baseband sample from input to entering the DC calibration position. Through the above timing arrangement, the DC estimate calculated by the training window can be used to calibrate baseband data belonging to the same data frame as the training window. The output data maintains the same sample order as the original data, and the DC estimate remains unchanged during the processing of the corresponding data frame, thus avoiding the loss of compensation opportunities for data arriving before the end of the training window.

[0072] In this embodiment, the guard segment can be used to accommodate transitional samples caused by frame detection errors, the receiver filter setup process, automatic gain control switching, and channel delay spread. The guard segment can be a cyclic extension of the end of the complex training block or a guard symbol pre-agreed upon by the transmitter and receiver. This segment does not participate in DC bias estimation and delay autocorrelation calculation to avoid sample interference with parameter estimation before the front-end has stabilized. Different I-path and Q-path pseudo-random sequences can reduce the correlation between the two training data streams, distributing the complex training symbols in different quadrants of the complex plane and preventing training energy from being concentrated in a single direction for a long time. The zero-mean condition ensures that the accumulated I-path and Q-path values ​​of a complete training block are both zero when there is no DC bias, providing a known benchmark for DC bias estimation. After consecutive repetition of complex training blocks, symbols separated by one or more complete training blocks have the same training content. This structure allows the receiver to select different training block intervals for coarse and fine delay autocorrelation without needing to set different training sequences for the two delays.

[0073] In this embodiment, the same training window is used for both stages of DC bias estimation, thereby reducing the training overhead in the data frame. For example, =128、 =8、 When the value is 32, the effective estimation segment contains 1024 modulation symbols, and the shared training preamble contains 1056 modulation symbols. If =4, the training window truncated after skipping the guard segment contains 4096 baseband samples, which can be used for the first-level DC estimation; the 1024 complex training symbols extracted according to the symbol interval can be used for coarse and fine delay autocorrelation.

[0074] In a preferred embodiment, the method for obtaining the first-level calibration data is as follows:

[0075] Within the training window corresponding to the effective estimation segment, the I-path training data and Q-path training data are accumulated respectively. After the training window ends, the I-path DC estimate and Q-path DC estimate are calculated.

[0076]

[0077]

[0078] During the accumulation process, the I / Q baseband data is delayed and buffered. Each sample point, cache This is the equivalent delay from the frame synchronization completion point to the end point of the training window;

[0079] After the training window ends, the I-channel DC estimates and Q-channel DC estimates are latched. The I / Q baseband data corresponding to the I-channel DC estimates and Q-channel DC estimates are read from the delay buffer. DC calibration is performed according to the following formula to obtain the I-channel data in the first-level calibration data. And Q-road data :

[0080]

[0081]

[0082] in, This represents the estimated value of DC path I. This represents the estimated DC value for the Q-channel. This represents the total number of training samples within the training window that participate in the first-level DC estimation. This indicates the training sample number used in the first-stage DC estimation. and They represent the first I-path data and Q-path data of each training sample point This indicates the length of the delay buffer used in the first-stage DC calibration. This indicates the baseband sample number output by the first-level delay buffer. and These represent the I-path data and Q-path data after delay alignment, respectively.

[0083] In this embodiment, the receiving end obtains and latches the DC estimates for the I and Q channels after the training window ends. When performing the first-level DC calibration on subsequent data of the same data frame, these two estimates are always used, and no more sample-by-sample updates are performed. After the current frame is processed, the estimates are recalculated for the next frame to ensure that a consistent DC calibration compensation amount is used within the same frame, avoiding changes in the compensation value within the frame.

[0084] In this embodiment, This buffer is used to store the I / Q baseband samples arriving in the same frame before the DC estimate is generated. Its length must at least cover the total processing delay caused by training window accumulation, mean calculation, result latching, and control flag transmission. The buffered data should be delayed synchronously with the frame flag, training window flag, and sample sequence number, ensuring that the DC estimate applies to the same data frame that generated it. The DC bias in the input signal will generate DC bias autocorrelation terms and cross-terms between the DC bias and training data in the correlation results, potentially causing the phase of the correlation results to deviate from the cumulative phase caused by the carrier frequency offset. Subtracting the training window mean first can reduce the impact of the above interference on the coarse and fine correlation results. The receiver obtains and latches the I and Q DC estimates after the training window ends; when processing baseband samples belonging to the same data frame as the training window and after delay alignment, these two estimates are always used, and no further sample-by-sample updates are performed. After processing the data frame, a new DC estimate is generated at the end of the training window of the next data frame. This avoids baseband data mean jumps caused by changes in compensation within the same data frame.

[0085] In a preferred embodiment, the method for calculating the coarse-scale delay autocorrelation and fine-scale delay autocorrelation of the first-level calibration data to obtain the coarse-scale correlation results and fine-scale correlation results is as follows:

[0086] Training data corresponding to the effective estimation segment is extracted from the first-level calibration data, and the training data is represented as follows:

[0087]

[0088] Based on the repetitive structure of the training block, autocorrelation accumulation is performed on training samples in the training data that are separated by corresponding delay lengths, according to both coarse-level and fine-level delay lengths:

[0089]

[0090]

[0091] The coarse-file delay length and the fine-file delay length are both integer multiples of the length of the complex training block, satisfying:

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] in, Indicates the first level of calibration. A complex training symbol, and These represent the I-channel and Q-channel data of the complex training symbol, respectively. Indicates the sequence number of the training symbols. This indicates the training symbol number used in the coarse autocorrelation operation. Indicates the training symbol number used in the fine-file autocorrelation operation. Represents the imaginary unit. This indicates the results related to the rough document. This indicates the detailed results. Indicates the coarse delay length. Indicates the fine-scale delay length. This indicates the number of complex training blocks included in the coarse-grained delay length. This indicates the number of complex training blocks contained within the fine-file delay length. Indicates the length of the complex training block. This indicates the ratio of the fine-scale delay length to the coarse-scale delay length. Indicates the length of the effective estimated segment. This indicates taking the conjugate of a complex number. This indicates that the relevant products within the corresponding sequence number range are accumulated. This indicates the modulo operation.

[0099] In this embodiment, coarse and fine are two sets of frequency offset estimation parameters divided according to the delay length used in the delay autocorrelation. The coarse-level delay autocorrelation uses a shorter delay length and is mainly used to determine the unambiguous range of the carrier frequency offset; the fine-level delay autocorrelation uses a longer delay length to improve the resolution of the carrier frequency offset estimation. Both process the same set of first-level calibration data. "Coarse" and "fine" correspond to different frequency offset resolution capabilities; it is understood that the naming does not indicate the quality level of the correlation data. The coarse-level result is responsible for providing the integer interval of the fine-level correlation phase, and the recovered fine-level cumulative phase is used to calculate the final phase increment.

[0100] In a preferred embodiment, the phase of the coarse-level correlation result is converted based on the ratio of the delay length of the fine-level delay autocorrelation to that of the coarse-level delay autocorrelation. The phase of the fine-level correlation result is then de-ambiguously de-defined using the converted phase as a reference to obtain the phase increment corresponding to the carrier frequency offset.

[0101] Take the phase of the coarse-file correlation results and the fine-file correlation results respectively:

[0102]

[0103]

[0104] Based on the proportional relationship between the coarse-length delay and the fine-length delay, the fine-length predicted phase is obtained:

[0105]

[0106]

[0107] Obtain the phase difference between the predicted phase and the phase of the correlation result in the fine-file section:

[0108]

[0109] The phase ambiguity order is obtained based on the phase difference:

[0110]

[0111] Recover the detailed cumulative phase based on the stated phase ambiguity order:

[0112]

[0113] The recovered detailed cumulative phase is converted into a phase increment per modulation symbol. The phase increment corresponding to the carrier frequency offset is obtained:

[0114]

[0115] in, This indicates the phase of the coarse correlation result within a preset phase period. This indicates the phase of the detailed correlation results within a preset phase period. This indicates obtaining the phase of a complex number. This indicates the results related to the rough document. This indicates the detailed results. Indicates the coarse delay length. Indicates the fine-scale delay length. This indicates the ratio of the fine-scale delay length to the coarse-scale delay length. This represents the detailed predicted phase calculated from the coarse-grained correlation results. This represents the phase difference between the predicted phase and the phase of the correlation result in the fine-file section. Indicates the phase ambiguity order. This indicates taking the nearest integer. Represents pi (π). This represents the cumulative phase of the fine-grained data after deblurring.

[0116] In this embodiment, the approximate range of carrier frequency offset is determined using the coarse-level correlated phase. Then, the coarse-level correlated phase is converted to the fine-level delay scale according to the ratio of the coarse and fine-level delay lengths. By comparing the conversion result with the fine-level correlated phase, the complete phase period missed by the fine-level correlated phase is determined, the actual cumulative phase within the fine-level delay interval is recovered, and thus the phase increment per modulation symbol is obtained. This method uses the coarse-level result to solve the problem of integer cycle foldback that easily occurs with long-delay correlated phases, while retaining the high estimation accuracy brought by the long delay of the fine-level phase, thus balancing the carrier frequency offset acquisition range and frequency offset resolution capability.

[0117] In a preferred embodiment, the method for obtaining frequency offset calibration data by performing digital down-conversion compensation on the first-level calibration data based on the phase increment is as follows:

[0118] Write the first-level calibration data into the delay buffer;

[0119] Based on the number of baseband samples contained in each modulation symbol, the phase increment per symbol is converted into the phase increment per sample:

[0120]

[0121] Read the first-level calibration data from the delay buffer, and accumulate the phase increment of each sample according to the baseband sample number to obtain the compensated phase of each baseband sample:

[0122]

[0123] The first-level calibration data is represented as:

[0124]

[0125] Based on the compensation phase, the first-level calibration data is subjected to a complex phase rotation in the opposite direction:

[0126]

[0127] Set the phase-rotated data as the frequency offset calibration data;

[0128] in, This represents the phase increment per symbol. This indicates the number of baseband samples contained in each symbol. This represents the phase increment at each sample point. This indicates the baseband sample number used in the digital down-conversion compensation. Indicates the first Compensation phase for each baseband sample point This indicates the first level of calibration data. A complex number of sample points, and These represent the I-channel data and Q-channel data of the complex sample points, respectively. Indicates the first in the frequency offset calibration data A complex number of sample points, Represents the imaginary unit. This indicates complex exponentiation.

[0129] In this embodiment, the phase increment of each modulation symbol is converted into the phase increment of each baseband sample, and then the compensation phase is accumulated point by point according to the baseband sample number. After reading the first-level calibration data in the delay buffer, a complex phase rotation opposite to the carrier frequency offset direction is performed on each baseband sample to obtain frequency offset calibration data. This process applies the frequency offset estimation result at the symbol scale to the actual baseband sampling data, so that each baseband sample obtains the compensation amount corresponding to its time position. The delay buffer reserves processing time for frequency offset estimation and also allows the estimation result to be used in the same data frame, making it suitable for continuous pipelined implementation.

[0130] In a preferred embodiment, the method for obtaining the second-level calibration data is as follows:

[0131] A training window corresponding to the effective estimation segment is extracted from the frequency offset calibration data. Within the training window, the I-channel training data and the Q-channel training data are accumulated respectively, and the residual I-channel DC estimate is obtained according to the following formula. and residual Q-path DC estimate :

[0132]

[0133]

[0134] During the accumulation process, the frequency offset calibration data is written to a delay buffer. After the training window ends, the residual I-channel DC estimate and the residual Q-channel DC estimate are latched. The frequency offset calibration data corresponding to the residual I-channel DC estimate and the residual Q-channel DC estimate are read from the delay buffer, and DC calibration is performed according to the following formula to obtain the second-level calibration data:

[0135]

[0136]

[0137] in, This represents the total number of training samples within the training window that participate in the second-level DC estimation. This indicates the I-path training sample number used in the second-level DC estimation. This indicates the Q-path training sample number used in the second-stage DC estimation. and These represent the I-channel and Q-channel data of the corresponding frequency offset calibration training samples, respectively. This indicates the sequence number of the data sample output from the second-level delay buffer. and These represent the I-channel and Q-channel data of the frequency offset calibration data after delay alignment, respectively. and These represent the I-channel data and Q-channel data of the second-level calibration data, respectively.

[0138] In this embodiment, for each baseband sample that belongs to the same data frame as the training data used to determine the residual I-channel DC bias estimate and the residual Q-channel DC bias estimate, after delay alignment, the residual I-channel DC bias estimate and the residual Q-channel DC bias estimate are kept unchanged, and the corresponding DC bias estimate is subtracted from the I-channel component and Q-channel component of each baseband sample to obtain the second-level calibration data.

[0139] In this embodiment, after frequency offset compensation is completed, training data corresponding to the effective estimation segment is re-extracted from the frequency offset calibration data. Two residual DC estimates are determined based on the I-path and Q-path mean values ​​of this training data. After the frequency offset calibration data is delayed and buffered, the corresponding residual DC estimates are subtracted from the I-path and Q-path components, respectively. Frequency offset compensation reduces the impact of continuous rotation of training symbols on mean calculation. The second mean estimation can further address the finite training window error, fixed-point calculation bias, and residual mean offset after frequency offset compensation in the first-level DC estimation. After completing the second-level calibration, the constellation center of the baseband data is closer to the origin, which is beneficial for subsequent channel estimation and symbol decision.

[0140] In a preferred embodiment, the method for calculating the phase of the coarse correlation result and the fine correlation result further includes: inputting the real and imaginary parts of the coarse correlation result into a coordinate rotation digital calculation module to calculate the coarse correlation phase; inputting the real and imaginary parts of the fine correlation result into the coordinate rotation digital calculation module to calculate the fine correlation phase; when the real and imaginary parts of the coarse correlation result are not simultaneously zero, and the real and imaginary parts of the fine correlation result are not simultaneously zero, setting the phase validity flag to a valid state and setting the error flag to an invalid state; when the real and imaginary parts of the coarse correlation result are simultaneously zero, or when the real and imaginary parts of the fine correlation result are simultaneously zero, setting the phase validity flag to an invalid state and setting the error flag to a valid state.

[0141] In this embodiment, the coordinate rotation digital calculation module calculates the corresponding phase based on the real and imaginary parts of the correlation results. When both correlation results can determine the phase, a valid state is output; when either correlation result is a zero complex number, an error state is output, preventing that phase result from entering subsequent integer recovery and frequency offset compensation. The coordinate rotation digital calculation module can perform complex phase calculations through shift and addition / subtraction operations, and can run in an FPGA or dedicated digital circuit. Valid and error flags can intercept undefined phase results, preventing erroneous phase increments from causing continuous erroneous rotation of the entire frame of data.

[0142] In one embodiment, such as Figure 3 The diagram illustrates a timing diagram for delayed alignment DC calibration. The horizontal axis represents time from left to right. After the training window valid signal becomes active, the receiver begins accumulating the I-channel and Q-channel training samples within the training window. When the training window valid signal returns to inactive, it indicates that all training samples used for this DC estimation have been received, corresponding to the training end time in the diagram. Subsequently, the receiver calculates the I-channel and Q-channel DC estimates based on the I-channel and Q-channel accumulation results and the number of training samples involved in the accumulation. After the DC estimation calculation is complete, the DC estimation valid signal becomes active. This signal can maintain one or more processing clock cycles to indicate that the two DC estimates have been generated and latching is enabled. "DC latching" in the diagram indicates writing the I-channel and Q-channel DC estimates calculated in this training window into the corresponding registers. After latching, the two DC estimates remain unchanged during the processing of baseband data belonging to the same data frame as this training window.

[0143] Upon reaching the compensation start time shown in the diagram, the receiver reads the in-phase / quadrature data from the delay buffer according to the original sample point order. It subtracts the latched I-channel DC estimate from each I-channel baseband sample point and subtracts the latched Q-channel DC estimate from each Q-channel baseband sample point to obtain the output in-phase / quadrature data. Output = Delayed Data - DC Estimate, meaning the output data is obtained by subtracting the delayed-aligned baseband data from the corresponding DC estimate.

[0144] In a preferred embodiment of this application, the length of the training block can be 32, 64, or 128 modulation symbols, and the coarse delay length can be 64 or 128 modulation symbols. The specific combination of values ​​should satisfy... for Integer multiples of. Fine-grained delay length is determined by... It is determined that the delay length ratio can be 2, 4, or 8, and must satisfy... < An effective estimated segment length should also satisfy the condition that it can be... and Divisible, and within the effective estimation segment, there exist sign pairs available for use with both coarse-scale and fine-scale delayed autocorrelation. Increase This improves the sensitivity of the fine-file accumulated phase to carrier frequency offset, resulting in higher resolution for the phase increment per modulation symbol obtained from the recovered fine-file accumulated phase. Under the same effective estimation segment length, Increasing the value also reduces the number of sign pairs included in the fine-file correlation accumulation and increases the likelihood that the fine-file correlation phase will span multiple full phase cycles, thus requiring the coarse-file correlation phase to have sufficient accuracy. The value of can be determined based on the maximum carrier frequency offset, training window length, and minimum operating signal-to-noise ratio. Zero-mean I-path and Q-path training sequences can be generated using balanced pseudo-random sequences, Walsh codes satisfying the zero-mean condition, or positive complementary sequences. When using m-sequences, since the standard m-sequence usually has a difference in the number of symbols after positive-negative-one mapping, this can be achieved by sequence concatenation, complementary combination, or adding balancing symbols to make the number of +1 and -1 in the final training block equal. When using Walsh codes, non-zero-mean codewords with equal numbers of +1 and -1 should be selected.

[0145] As a preferred embodiment of this application, the integer recovery of the fine-file correlated phase can also be jointly judged using three or more delay lengths, and the candidate integer number can be voted on based on multiple delay correlation results; a lookup table between the coarse-file correlated phase, the fine-file correlated phase and the phase ambiguity order can also be pre-established; under low signal-to-noise ratio conditions, the correlation error corresponding to each candidate phase ambiguity order can also be calculated, and the candidate value with the smallest error or the largest likelihood value can be selected.

[0146] In a preferred embodiment of this application, the second-level DC bias estimation can use the one-time mean of the entire training window, or it can use a moving average or block update. When using a moving average, the update process can be limited to the training field or other known zero-mean data intervals; when using block update, new residual DC estimates can be latched at the boundaries of each data block, avoiding changes to the DC calibration compensation amount point by point within the data block.

[0147] In a preferred embodiment of this application, the total length of the shared training preamble can be set to 1152, 1600, or 1664 modulation symbols, depending on the frequency offset acquisition range, the correlation accumulation length, and the receiver setup time. When adjusting the guard segment length, the number of repetitions of the retraining block, or the effective estimation segment length accordingly, the effective estimation segment should be composed of a complete retraining block, and both the coarse-level delay and the fine-level delay should be aligned with the repetition position of the retraining block. This allows for adaptation to different data frame lengths, frequency offset ranges, and hardware buffer capacities without altering the coarse- and fine-level frequency offset estimation principles.

[0148] In a preferred embodiment, such as Figure 4As shown, a single-stage combined frequency offset and two-stage DC calibration device is provided, the device comprising:

[0149] The baseband data acquisition module 510 is used to acquire I / Q baseband data containing a shared training preamble. The effective estimation segment of the shared training preamble is composed of multiple identical complex training blocks. The I-path training sequence and Q-path training sequence of each complex training block are constructed as zero-mean sequences at the transmitter.

[0150] The first-level calibration module 520 is used to obtain the DC estimation value of the I-channel and the DC estimation value of the Q-channel based on the mean of the I-channel training data and the Q-channel training data in the effective estimation segment, respectively, and to perform DC calibration on the I / Q baseband data based on the DC estimation value of the I-channel and the DC estimation value of the Q-channel to obtain the first-level calibration data;

[0151] The delay autocorrelation module 530 is used to calculate the coarse delay autocorrelation and fine delay autocorrelation of the first-level calibration data to obtain the coarse correlation result and the fine correlation result. The delay length of the fine delay autocorrelation is greater than the delay length of the coarse delay autocorrelation.

[0152] The phase increment acquisition module 540 is used to convert the phase of the coarse-level correlation result based on the ratio of the delay length of the fine-level delay autocorrelation to the delay length of the coarse-level delay autocorrelation, and to de-blur the phase of the fine-level correlation result using the converted phase as a reference to obtain the phase increment corresponding to the carrier frequency offset.

[0153] The frequency offset calibration module 550 is used to perform digital down-conversion compensation on the first-level calibration data based on the phase increment to obtain frequency offset calibration data.

[0154] The secondary calibration module 560 is used to perform DC calibration on the frequency offset calibration data based on the I-channel mean and Q-channel mean of the training data corresponding to the effective estimation segment in the frequency offset calibration data, so as to obtain the second-level calibration data.

[0155] In the embodiments of this application, the aforementioned device can be a computer program software, a data processing module, or an integrated circuit hardware module composed of several component units. This integrated circuit, when running, can realize the functions of the corresponding module. Explanations of the modules are provided in the corresponding steps and will not be repeated here. For example, this device is implemented using a digital signal processor in conjunction with an application-specific integrated circuit (ASIC). Each module can continuously process baseband data at a fixed pace. Coarse-to-fine correlation operations and two-level averaging calculations can be implemented in parallel or reused computing resources depending on the processing speed, balancing processing latency and hardware usage.

[0156] In one embodiment, a communication system is provided, the system including a transmitter and a receiver;

[0157] The sending end is used to send a data frame containing training data to the receiving end;

[0158] The receiving end receives the data frame and performs the single-stage joint frequency offset and two-stage DC calibration method as described above.

[0159] In this embodiment, the transmitting end sends a data frame containing a shared training preamble; the receiving end performs first-level DC estimation, coarse-to-fine carrier frequency offset estimation, sample-by-sample frequency offset compensation, and second-level residual DC estimation based on the shared training preamble. The same training preamble simultaneously serves DC calibration and carrier frequency offset calibration, reducing the transmission overhead caused by setting multiple training fields separately. This communication system is suitable for burst communication, short frame transmission, and rapid link establishment scenarios, helping to shorten the time required for the receiving end to complete parameter estimation.

[0160] It should be understood that although the steps in the flowcharts of the various embodiments of this application are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0161] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Furthermore, any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory.

[0162] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A single-shot joint frequency offset and two-stage DC calibration method, applied at the receiving end, characterized in that, The method includes: Acquire I / Q baseband data containing a shared training preamble, wherein the effective estimation segment of the shared training preamble is composed of multiple identical complex training blocks, and the I-path training sequence and Q-path training sequence of each complex training block are constructed as zero-mean sequences at the transmitter. Based on the mean values ​​of the I-channel training data and the Q-channel training data within the effective estimation segment, the DC estimation values ​​of the I-channel and Q-channel are obtained respectively. Based on the DC estimation values ​​of the I-channel and Q-channel, the I / Q baseband data is DC calibrated to obtain the first-level calibration data. Calculate the coarse-scale delay autocorrelation and fine-scale delay autocorrelation of the first-level calibration data to obtain the coarse-scale correlation result and the fine-scale correlation result. The delay length of the fine-scale delay autocorrelation is greater than the delay length of the coarse-scale delay autocorrelation. Based on the ratio of the delay length of the fine-length delay autocorrelation to that of the coarse-length delay autocorrelation, the phase of the coarse-length correlation result is converted, and the phase of the fine-length correlation result is de-fuzzified using the converted phase as a reference to obtain the phase increment corresponding to the carrier frequency offset. Based on the phase increment, digital down-conversion compensation is performed on the first-level calibration data to obtain frequency offset calibration data; Based on the I-channel mean and Q-channel mean of the training data corresponding to the effective estimation segment in the frequency offset calibration data, DC calibration is performed on the frequency offset calibration data to obtain the second-level calibration data.

2. The method for single-stage joint frequency offset and two-stage DC calibration according to claim 1, characterized in that, The shared training preamble includes a guard segment and the effective estimation segment, wherein the guard segment precedes the effective estimation segment; the complex training block is constructed to satisfy: The I-path training sequence and the Q-path training sequence each consist of different lengths. The pseudo-random sequence is obtained by mapping positive and negative 1, where bit 1 is mapped to +1 and bit 0 is mapped to -1, and the following conditions are met: Repeat the retraining block continuously The resulting effective estimation segment is as follows: The I / Q baseband data is frame synchronized. Based on the shared training preamble start position obtained from the frame synchronization, the guard segment is skipped, and the effective estimation segment is truncated as a training window. The training window is used for both DC estimation and carrier frequency offset estimation. in, Indicates the first training block A complex training symbol, Indicates the first I-path training symbols for a complex training symbol, Indicates the first Q-path training symbols for a complex training symbol, Represents the imaginary unit. Indicates the symbol index within the complex training block. The symbol number representing the training sequence of path I. Indicates the symbolic index of the Q-path training sequence. Indicates the length of the complex training block. Indicates the number of repetitions of the training block. Indicates the length of the effective estimated segment. Indicates the length of the protected section. Indicates the length of the shared training leader.

3. The method for single-stage joint frequency offset and two-stage DC calibration according to claim 1, characterized in that, The method for obtaining the first-level calibration data is as follows: Within the training window corresponding to the effective estimation segment, the I-path training data and Q-path training data are accumulated respectively. After the training window ends, the I-path DC estimate and Q-path DC estimate are calculated. During the accumulation process, the I / Q baseband data is delayed and buffered. Each sample point, cache This is the equivalent delay from the frame synchronization completion point to the end point of the training window; After the training window ends, the I-channel DC estimates and Q-channel DC estimates are latched. The I / Q baseband data corresponding to the I-channel DC estimates and Q-channel DC estimates are read from the delay buffer. DC calibration is performed according to the following formula to obtain the I-channel data in the first-level calibration data. And Q-road data : in, This represents the estimated value of DC path I. This represents the estimated DC value for the Q-channel. This represents the total number of training samples within the training window that participate in the first-level DC estimation. This indicates the sequence number of the I-path training samples used in the first-stage DC estimation. This indicates the sequence number of the Q-path training samples used in the first-stage DC estimation. and These represent the I-path and Q-path data for the corresponding training samples, respectively. This indicates the length of the delay buffer used in the first-stage DC calibration. This indicates the sequence number of the data sample output from the first-level delay buffer. and These represent the delayed and aligned I-channel and Q-channel data, respectively. and These represent the I-channel data and Q-channel data of the first-level calibration data, respectively.

4. The method for single-stage joint frequency offset and two-stage DC calibration according to claim 1, characterized in that, The method for calculating the coarse-scale delay autocorrelation and fine-scale delay autocorrelation of the first-level calibration data, and obtaining the coarse-scale correlation results and fine-scale correlation results, is as follows: Training data corresponding to the effective estimation segment is extracted from the first-level calibration data, and the training data is represented as follows: Based on the repetitive structure of the training block, autocorrelation accumulation is performed on training samples in the training data that are separated by corresponding delay lengths, according to both coarse-level and fine-level delay lengths: The coarse-file delay length and the fine-file delay length are both integer multiples of the length of the complex training block, satisfying: in, Indicates the first level of calibration. A complex training symbol, and These represent the I-channel and Q-channel data of the complex training symbol, respectively. Indicates the sequence number of the training symbols. This indicates the training symbol number used in the coarse autocorrelation operation. Indicates the training symbol number used in the fine-file autocorrelation operation. Represents the imaginary unit. This indicates the results related to the rough document. This indicates the detailed results. Indicates the coarse delay length. Indicates the fine-scale delay length. This indicates the number of complex training blocks included in the coarse-grained delay length. This indicates the number of complex training blocks contained within the fine-file delay length. Indicates the length of the complex training block. This indicates the ratio of the fine-scale delay length to the coarse-scale delay length. Indicates the length of the effective estimated segment. This indicates taking the conjugate of a complex number. This indicates that the relevant products within the corresponding sequence number range are accumulated. This indicates the modulo operation.

5. The method for single-stage joint frequency offset and two-stage DC calibration according to claim 1, characterized in that, The method for converting the phase of the coarse-level correlation result based on the ratio of the delay length of the fine-level delay autocorrelation to that of the coarse-level delay autocorrelation, and then using the converted phase as a reference to de-ambiguate the phase of the fine-level correlation result to obtain the phase increment corresponding to the carrier frequency offset is as follows: Take the phase of the coarse-file correlation results and the fine-file correlation results respectively: Based on the proportional relationship between the coarse-length delay and the fine-length delay, the fine-length predicted phase is obtained: Obtain the phase difference between the predicted phase and the phase of the correlation result in the fine-file section: The phase ambiguity order is obtained based on the phase difference: Recover the detailed cumulative phase based on the stated phase ambiguity order: The recovered detailed cumulative phase is converted into a phase increment per modulation symbol. The phase increment corresponding to the carrier frequency offset is obtained: in, This indicates the phase of the coarse correlation result within a preset phase period. This indicates the phase of the detailed correlation results within a preset phase period. This indicates obtaining the phase of a complex number. This indicates the results related to the rough document. This indicates the detailed results. Indicates the coarse delay length. Indicates the fine-scale delay length. This indicates the ratio of the fine-scale delay length to the coarse-scale delay length. This represents the detailed predicted phase calculated from the coarse-grained correlation results. This represents the phase difference between the predicted phase and the phase of the correlation result in the fine-file section. Indicates the phase ambiguity order. This indicates taking the nearest integer. Represents pi (π). This represents the cumulative phase of the fine-grained data after deblurring.

6. The method for single-stage joint frequency offset and two-stage DC calibration according to claim 5, characterized in that, The method for obtaining frequency offset calibration data by performing digital down-conversion compensation on the first-level calibration data based on the phase increment is as follows: Write the first-level calibration data into the delay buffer; Based on the number of baseband samples contained in each modulation symbol, the phase increment per symbol is converted into the phase increment per sample: Read the first-level calibration data from the delay buffer, and accumulate the phase increment of each sample according to the baseband sample number to obtain the compensated phase of each baseband sample: The first-level calibration data is represented as: Based on the compensation phase, the first-level calibration data is subjected to a complex phase rotation in the opposite direction: Set the phase-rotated data as the frequency offset calibration data; in, This represents the phase increment per symbol. This indicates the number of baseband samples contained in each symbol. This represents the phase increment at each sample point. This indicates the baseband sample number used in the digital down-conversion compensation. Indicates the first Compensation phase for each baseband sample point This indicates the first level of calibration data. A complex number of sample points, and These represent the I-channel data and Q-channel data of the complex sample points, respectively. Indicates the first in the frequency offset calibration data A complex number of sample points, Represents the imaginary unit. This indicates complex exponentiation.

7. The method for single-stage joint frequency offset and two-stage DC calibration according to claim 1, characterized in that, The method for obtaining the second-level calibration data is as follows: A training window corresponding to the effective estimation segment is extracted from the frequency offset calibration data. Within the training window, the I-channel training data and the Q-channel training data are accumulated respectively, and the residual I-channel DC estimate is obtained according to the following formula. and residual Q-path DC estimate : During the accumulation process, the frequency offset calibration data is written to the delay buffer; After the training window ends, the residual I-channel DC estimates and residual Q-channel DC estimates are latched. The frequency offset calibration data corresponding to the residual I-channel DC estimates and residual Q-channel DC estimates are read from the delay buffer. DC calibration is performed according to the following formula to obtain the second-level calibration data: in, This represents the total number of training samples within the training window that participate in the second-level DC estimation. This indicates the I-path training sample number used in the second-level DC estimation. This indicates the Q-path training sample number used in the second-stage DC estimation. and These represent the I-channel and Q-channel data of the corresponding frequency offset calibration training samples, respectively. This indicates the sequence number of the data sample output from the second-level delay buffer. and These represent the I-channel and Q-channel data of the frequency offset calibration data after delay alignment, respectively. and These represent the I-channel data and Q-channel data of the second-level calibration data, respectively.

8. The method for single-stage joint frequency offset and two-stage DC calibration according to claim 6, characterized in that, The method for calculating the phase of the coarse-file correlation result and the fine-file correlation result further includes: The real and imaginary parts of the coarse correlation result are input into the coordinate rotation digital calculation module to obtain the coarse correlation phase; The real and imaginary parts of the detailed correlation results are input into the coordinate rotation digital calculation module to obtain the detailed correlation phase; When the real and imaginary parts of the coarse correlation result are not both zero, and the real and imaginary parts of the fine correlation result are not both zero, the phase valid flag is set to valid and the error flag is set to invalid. When the real and imaginary parts of the coarse correlation result are both zero, or the real and imaginary parts of the fine correlation result are both zero, the phase valid flag is set to invalid and the error flag is set to valid.

9. A single-stage combined frequency offset and two-stage DC calibration device, characterized in that, The device includes: The baseband data acquisition module is used to acquire I / Q baseband data containing a shared training preamble. The effective estimation segment of the shared training preamble is composed of multiple identical complex training blocks. The I-path training sequence and Q-path training sequence of each complex training block are constructed as zero-mean sequences at the transmitter. The first-level calibration module is used to obtain the DC estimation values ​​of the I-channel and Q-channel training data according to the mean values ​​of the I-channel training data and Q-channel training data in the effective estimation segment, respectively, and to perform DC calibration on the I / Q baseband data based on the DC estimation values ​​of the I-channel and Q-channel training data to obtain the first-level calibration data; The delay autocorrelation module is used to calculate the coarse delay autocorrelation and fine delay autocorrelation of the first-level calibration data, and obtain the coarse correlation result and the fine correlation result. The delay length of the fine delay autocorrelation is greater than the delay length of the coarse delay autocorrelation. The phase increment acquisition module is used to convert the phase of the coarse-level correlation result based on the ratio of the delay length of the fine-level delay autocorrelation to the delay length of the coarse-level delay autocorrelation, and to de-blur the phase of the fine-level correlation result using the converted phase as a reference to obtain the phase increment corresponding to the carrier frequency offset. The frequency offset calibration module is used to perform digital down-conversion compensation on the first-level calibration data based on the phase increment to obtain frequency offset calibration data. The secondary calibration module is used to perform DC calibration on the frequency offset calibration data based on the I-channel mean and Q-channel mean of the training data corresponding to the effective estimation segment in the frequency offset calibration data, so as to obtain the second-level calibration data.

10. A communication system, characterized in that, The system includes a transmitter and a receiver; The sending end is used to send a data frame containing training data to the receiving end; The receiving end is used to receive the data frame and execute the single joint frequency offset and two-stage DC calibration method according to any one of claims 1 to 8.