阅读说明：本技术 一种gfsk接收机中的频偏估计方法 (Frequency offset estimation method in GFSK receiver ) 是由 梁晓峰 李啟彬 陈智德 于 2021-10-27 设计创作，主要内容包括：本发明公开了一种GFSK接收机中的频偏估计方法。本发明方法首先在解调后的调制信号频率中找出连续bit时的极大值和极小值,以及单bit时的极大值和极小值；然后对两种极大值和极小值分别进行滤波处理；最后将连续bit时的极大值的滤波结果与极小值的滤波结果进行平均,将单bit时的极大值的滤波结果与极小值的滤波结果进行平均,将两个平均结果再做一次平均,最终得到的结果即为当前接收信号的频率偏移值。本发明方法可应用于接收全过程,不受数据内容影响,具有较高的通用性,可同时兼容低中频接收机与零中频接收机。本发明方法完全采用实测结果来进行频偏估计,精确度更高。(The invention discloses a frequency offset estimation method in a GFSK receiver. The method firstly finds out a maximum value and a minimum value when the modulation signal frequency is demodulated, and a maximum value and a minimum value when the modulation signal frequency is single bit; then, respectively carrying out filtering processing on the two maximum values and the two minimum values; and finally, averaging the filtering result of the maximum value and the filtering result of the minimum value when the signals are continuously bit, averaging the filtering result of the maximum value and the filtering result of the minimum value when the signals are singly bit, averaging the two averaging results again, and finally obtaining a result which is the frequency deviation value of the current receiving signal. The method can be applied to the whole receiving process, is not influenced by data contents, has higher universality and can be compatible with a low intermediate frequency receiver and a zero intermediate frequency receiver at the same time. The method of the invention completely adopts the actual measurement result to carry out frequency offset estimation, and has higher accuracy.)

1. A frequency offset estimation method in a GFSK receiver is characterized in that the method specifically comprises the following steps:

finding out a maximum value and a minimum value when the modulation signal frequency is continuously bit and a maximum value and a minimum value when the modulation signal frequency is single bit;

step (2) filtering the two maximum values and the two minimum values respectively;

and (3) averaging the filtering result of the maximum value and the filtering result of the minimum value when the signals are continuously bit, averaging the filtering result of the maximum value and the filtering result of the minimum value when the signals are singly bit, averaging the two averaging results again, and finally obtaining a result which is the frequency deviation value of the current received signal.

2. A method of estimating frequency offset in a GFSK receiver as claimed in claim 1, wherein step (1) is specifically:

in the demodulated modulation signal frequency, every three continuous sampling points are used as a group of decision data, if the amplitude of the middle sampling point is larger than that of the other two sampling points, the sampling point is used as the maximum value of the group of decision data, and if the amplitude of the middle sampling point is smaller than that of the other two sampling points, the sampling point is used as the minimum value of the group of decision data; then judging whether the sampling point belongs to continuous bit or single bit;

amplitude A of demodulation signal corresponding to modulation frequency offset set by current GFSK system_devThree decision thresholds are set, THR1 < THR2 < THR3, THR1 ═ alpha.A_dev，THR2＝β·A_dev，THR3＝γ·A_devAlpha, beta and gamma are set parameters; in the demodulated modulation signal frequency, the belonged situation of bit conversion is judged according to the absolute value delta of the difference between adjacent extreme values: if THR1 is less than or equal to delta < THR2, the case one is true; if THR2 is less than or equal to delta < THR3, the case is two or three; if Δ ≧ THR3, then case four; the maximum value or the minimum value of a single bit belongs to the case one or three, and the maximum value or the minimum value of a continuous bit belongs to the case two or four; the first case is the conversion from single bit to single bit, the second case is the conversion from single bit to continuous bit, and the third case is the conversion from continuous bit to single bitAnd C, converting, wherein the fourth condition is conversion from continuous bit to continuous bit.

3. The frequency offset estimation method in a GFSK receiver according to claim 2, wherein in step (1), only the time interval between the two acquired extrema is kept as the set extremum, and the two maxima and minima are found as follows: the conversion from single bit to single bit is carried out, and the time interval is set to be 0.8 Ts-1.2 Ts; the time interval is set to be greater than 1.2Ts for the conversion from single bit to continuous bit or from continuous bit to single bit.

4. A method of frequency offset estimation in a GFSK receiver as claimed in claim 2, wherein: alpha is more than or equal to 0.7 and less than or equal to 1.2, beta is more than 1.2 and less than or equal to 1.6, and gamma is more than 1.6 and less than or equal to 2.

5. The frequency offset estimation method in a GFSK receiver according to claim 1, wherein in step (2) the maximum value at continuous bit, the minimum value at continuous bit, the maximum value at single bit and the minimum value at single bit are filtered separately; the filtering mode adopts moving average or Kalman filtering.

6. The frequency offset estimation method in a GFSK receiver according to claim 1, wherein in step (2) the maximum value at continuous bit, the minimum value at continuous bit, the maximum value at single bit and the minimum value at single bit are filtered separately; firstly, feeding back a frequency deviation rough calculation result to a digital mixer in a moving average filtering mode to remove large frequency deviation, and then continuously estimating and correcting the statistical characteristic of noise in a Kalman filtering mode.

Technical Field

The invention belongs to the technical field of wireless communication, and relates to a frequency offset estimation method in a GFSK receiver.

Background

GFSK modulation is a modulation method in which a carrier frequency is modulated by a baseband signal, and therefore the baseband signal itself is a signal of frequency dimension, and the receiver demodulates the output signal, i.e., the signal of frequency dimension. Because the local oscillation signal frequency of the receiver and the local oscillation signal frequency of the transmitting end cannot be completely consistent, slight deviation exists inevitably, and the direct influence is that the demodulated waveform has certain direct current offset. The decision circuit behind the demodulation module is processed based on the set threshold to restore the original data, and the direct current offset of the demodulated data greatly influences the signal decision. The worst possibility is that the demodulated data is completely larger or smaller than the demodulation threshold, which results in the case that the decision is long 1 or long 0, and any useful data cannot be output. Therefore, before sending the frequency signal to the decision module, the frequency offset estimation must be performed on the demodulated output frequency signal, and the frequency offset is calibrated according to the estimation result, and then the decision stage can be entered.

How well the compensation is depends on an accurate estimation of the frequency offset. For the estimation of the frequency offset, it is a common practice to accumulate and average the frequency signals after receiving and demodulating, which is simple to implement, but the estimation result of the frequency offset is greatly affected by the form of actual data and the duration of the accumulated signal. In the conventional technology, as shown in the four patents listed in the patent comparison, a preamble (0101 alternate data) is added before valid data, and the method of direct sampling and then accumulation averaging is required for the receiver to perform calibration. This method has limitations on the data format, and if the data format is changed, a large calculation error may be caused, and the number of accumulated data also affects the calculation accuracy, and if the number of accumulated data is too small, the interference in the received signal may be greatly affected, and if the number of accumulated data is too large, the frequency offset calculation time is too long and the consumption of hardware resources is large.

The invention patent No. 201811200773.9 discloses a frequency offset estimator for use in a GFSK demodulator and a method thereof. In the frequency deviation estimator, a signal obtained by a digital differential phase discriminator is firstly transmitted to a feedback frequency deviation estimator with the characteristic of fast convergence of frequency deviation estimation, an average value calculator is used for calculating the initial frequency deviation of the signal and feeding the initial frequency deviation back to a digital down-conversion module at the front end to complete the recovery of the initial frequency deviation, a median selector and a Kalman filter are used for periodically estimating a frequency deviation value, and when a frequency deviation accumulated value reaches an adjustment threshold value, the estimated value is fed forward to an adjustable threshold value decider to change the decision threshold value. The method selects the intermediate sampling value of two symbols (single-bit data) with the code element sequence of 10 or 01 for estimation, cannot distinguish continuous bits or the change from the continuous bits to the single bits, and has limitation in application. It can only be used for data format with preamble fixed to 010101 and cannot perform continuous frequency drift tracking during the acceptance process.

The invention patent with patent number 201811452250.3 discloses a method and a system for estimating and compensating sampling frequency offset, wherein the method comprises the following steps: determining the windowing position of the current symbol, and acquiring the sampling point of the current symbol from the data cache according to the windowing position; after the absolute value of the sampling point of the current symbol is taken, five sampling values at equal intervals in each symbol period are calculated; utilizing five equal-interval sampling values to iteratively calculate a first decision expression and a second decision expression; and when the sampling frequency offset compensation is determined to be required according to the first decision expression, determining a windowing adjusting parameter by using a second decision expression so as to adjust the windowing position. In the method, frequency offset estimation is calculated by five equally-spaced sampling points in a symbol, each symbol used for calculation must have a consistent shape to obtain an accurate result, namely a fixed data form 010101, and if the data has certain randomness, calculation cannot be performed. If the data has certain interference, the estimation result is also affected.

The invention patent with the patent number 201210144850.X discloses a frequency offset estimation method and a system for coherent demodulation frequency shift keying modulation signals. The prefix signal of the GFSK/FSK signal is designed into a string of 0 and 1 alternating codes with fixed length, and the use of the prefix code enables a receiver to completely remove c (i) from x (i) without copying an ideal received signal c (i), so that the problem that the ideal received signal c (i) cannot be accurately copied at a receiving end due to the fact that the frequency modulation index of the GFSK/FSK signal cannot be accurately obtained and changes along with time and temperature is avoided. The method designs the lead code signal into a 010101 sequence with fixed length, calculates frequency offset through phase change and then averages the frequency offset, firstly has a certain requirement on a data format, and can average the frequency offset value only by storing a large amount of hardware resources.

The invention patent No. 201410686680.7 discloses a method for fast carrier frequency offset estimation and correction of FSK signals. The method comprises the steps of inputting two paths of low-intermediate frequency I/Q signals generated by a radio frequency module, carrying out correlation operation and filtering processing on the signals generated by a DDS (direct digital synthesizer), carrying out mean value operation, carrying out corresponding query in a pre-stored mean value-frequency offset lookup table to estimate an estimated value of frequency deviation, and feeding the estimated value back to the previous DDS to carry out frequency deviation elimination to complete a frequency deviation correction process. The method generates a string of 1010 or 0101 lead codes, estimates the direct current component through an average filtering method after obtaining the frequency, and then converts the direct current into a frequency offset value through an average-frequency offset lookup table, thereby realizing direct current component compensation.

Disclosure of Invention

The invention aims to provide a frequency offset estimation method in a GFSK receiver.

The received signal of GFSK is expressed as:t represents the time of receiving the signal, i represents the number of accumulated phases, i ≧ 0, f₀(t) represents a carrier frequency at time t, ω (t) represents a modulation signal frequency before demodulation at time t, and η (t) represents an interference signal at time t;

the demodulated modulated signal frequency is represented as:η "(t) represents the system noise at time t.

It can be seen that the frequency of the demodulated modulated signal has a certain deviation from the actual frequency value, including f₀(t) and η "(t). The frequency value needs to be determined to restore the information sequence, so the frequency offset estimation and frequency offset calibration must be completed before the determination.

The waveform of the GFSK signal received by the receiver after demodulation is represented by frequency change, and if the carrier of the receiving and transmitting system is absolutely same in frequency and has no noise influence, the signal output after demodulation is changed by taking 0 as the center. If the frequency deviation exists between the received signal and the carrier wave, the fact that the direct current offset exists in the demodulation signal can be reflected. Therefore, the frequency offset estimation of the GFSK receiver can be simply understood as calculating the DC offset of the demodulated signal.

The common method for calculating the dc offset in a digital manner is accumulation averaging, but this requires that the number of 0 s and 1 s in the calculated data segment is equivalent and the distribution is uniform, so that the dc offset can be calculated more accurately. However, if the numbers of 0 and 1 in the transmission data segment are obviously different and the distribution is not uniform, the demodulation data is unlikely to be uniformly distributed above and below the 0 point, and may be more biased to positive values or more biased to negative values, resulting in a larger deviation between the result of accumulation and averaging and the true value.

The method specifically comprises the following steps:

finding out a maximum value and a minimum value when the modulation signal frequency is continuously bit and a maximum value and a minimum value when the modulation signal frequency is single bit after demodulation in the step (1):

in the demodulated modulation signal frequency, every three continuous sampling points are used as a group of decision data, if the amplitude of the middle sampling point is larger than that of the other two sampling points, the sampling point is used as the maximum value of the group of decision data, and if the amplitude of the middle sampling point is smaller than that of the other two sampling points, the sampling point is used as the minimum value of the group of decision data; and then whether the sample point belongs to a continuous bit or a single bit is judged.

Amplitude A of demodulation signal corresponding to modulation frequency offset set by current GFSK system_devThree decision thresholds are set, THR1 < THR2 < THR3, THR1 ═ alpha.A_dev，THR2＝β·A_dev，THR3＝γ·A_devAlpha, beta and gamma are set parameters; in the demodulated modulation signal frequency, the belonged situation of bit conversion is judged according to the absolute value delta of the difference between adjacent extreme values: if THR1 is less than or equal to delta < THR2, the case one is true; if THR2 is less than or equal to delta < THR3, the case is two or three; if Δ ≧ THR3, then case four; the maximum value or the minimum value of a single bit belongs to the case one or three, and the maximum value or the minimum value of a continuous bit belongs to the case two or four; the first condition is the conversion from single bit to single bit, and the first condition is the conversion from single bit to single bitThe second is the conversion from single bit to continuous bit, the third is the conversion from continuous bit to single bit, and the fourth is the conversion from continuous bit to continuous bit.

Step (2) filtering the two maximum values and the two minimum values respectively; the filtering mode adopts moving average and/or Kalman filtering.

And (3) averaging the filtering result of the maximum value and the filtering result of the minimum value when the signals are continuously bit, averaging the filtering result of the maximum value and the filtering result of the minimum value when the signals are singly bit, averaging the two averaging results again, and finally obtaining a result which is the frequency deviation value of the current received signal.

The frequency offset estimation method in the GFSK receiver provided by the invention records four waveform extreme values of demodulated effective data, calculates the receiving frequency offset, can be applied to the whole receiving process, is not influenced by data contents, has higher universality, and can be compatible with a low intermediate frequency receiver and a zero intermediate frequency receiver at the same time. For a low-IF receiver, the frequency offset estimate may be fed back to the digital mixer, and the frequency offset calibration may be performed simultaneously with the removal of the IF. For zero-if receivers, a digital mixer is added to calibrate the frequency offset. The invention can effectively combine the advantages of operation accuracy, calibration real-time performance, lower hardware resources and the like.

Drawings

Fig. 1 is a schematic diagram of four extreme values of an ideal GFSK modulated signal;

fig. 2 is a diagram of a pseudo-extremum of a GFSK modulated signal with interference.

Detailed Description

A frequency deviation estimation method in a GFSK receiver adopts a method of finding extremum and averaging, and the deviation of the integral modulation waveform and a 0 value is calculated by finding an effective extremum in a demodulation waveform. This method will be described below.

Finding out a maximum value and a minimum value when the modulation signal frequency is continuously bit and a maximum value and a minimum value when the modulation signal frequency is single bit after demodulation in the step (1):

as can be known from the time domain waveform of the GFSK, the amplitude of the modulated signal is not a fixed value, and a single "1" or a single "0" of the baseband signal has substantially the same absolute value of the corresponding amplitude in the modulated signal; while the amplitude of consecutive "1" s or consecutive "0" s is higher than a single "1" or a single "0".

In the demodulated modulation signal frequency, every three continuous sampling points are used as a group of decision data, if the amplitude of the middle sampling point is larger than that of the other two sampling points, the sampling point is used as the maximum value of the group of decision data, and if the amplitude of the middle sampling point is smaller than that of the other two sampling points, the sampling point is used as the minimum value of the group of decision data; and then whether the sample point belongs to a continuous bit or a single bit is judged.

The continuous bit maximum value and the continuous bit minimum value, and the single bit maximum value and the single bit minimum value are always symmetrical by taking a direct current point as a center, so that the direct current value of the demodulated signal, namely the frequency offset of the received signal, can be accurately obtained as long as the four waveform extreme values can be accurately found, and the frequency offset can be calibrated by feeding back the frequency offset to the digital mixer. Therefore, the preamble of 0101 at the beginning of data transmission in a specific communication mode is not required to be relied on, and the method can be applied to data transmission to perform frequency offset tracking, so that frequency drift generated in the effective data transmission process can be effectively detected and eliminated, and the receiving quality in the whole data transmission process is ensured.

According to the shape of the GFSK modulated signal, the amplitude variation of the modulated signal frequency is divided into four cases:

the first condition is as follows: single bit to single bit conversion, such as "0101" or "1010";

case two: conversion of a single bit to a continuous bit, such as "01000" or "10111";

case three: continuous bit to single bit transitions, such as "00010" or "11101";

case four: continuous bit to continuous bit transitions, such as "000111" or "111000".

Under ideal conditions, the amplitude variation value of the modulation signal frequency in case one is minimum, the amplitude variation values in case two and case three are basically consistent and are greater than the amplitude variation value in case one and the amplitude variation value of the modulation signal frequency in case four is maximum. Amplitude A of demodulation signal corresponding to modulation frequency offset set by current GFSK system_devThree decision thresholds are set, THR1 < THR2 < THR3, THR1 ═ alpha.A_dev，THR2＝β·A_dev，THR3＝γ·A_devAlpha, beta and gamma are set parameters, alpha is more than or equal to 0.7 and less than or equal to 1.2, beta is more than 1.2 and less than or equal to 1.6, and gamma is more than 1.6 and less than or equal to 2.

In the demodulated modulation signal frequency, the belonged situation of bit conversion is judged according to the absolute value delta of the difference between adjacent extreme values: if THR1 is less than or equal to delta < THR2, the case one is true; if THR2 is less than or equal to delta < THR3, the case is two or three; case four is true if Δ ≧ THR 3. The maximum value or the minimum value of a single bit belongs to the case one or three, and the maximum value or the minimum value of a continuous bit belongs to the case two or four. As shown in fig. 1, the a portion is a maximum value when the number of bits is continuous, the B portion is a minimum value when the number of bits is continuous, the C portion is a maximum value when the number of bits is single, and the D portion is a minimum value when the number of bits is single.

Considering that the received data is not ideal data, the demodulated data may have interference, so that false extreme values (E is a false maximum value, and F is a false minimum value) as shown in fig. 2 occur, and a judgment only from the amplitude change may cause a false judgment, so that the judgment condition needs to be strengthened, and the time interval between two collected extreme values is used as another judgment basis.

The conversion from single bit to single bit, under ideal conditions, the time interval of the amplitude change is 1Ts duration, and considering interference factors, the time interval can be set to 0.8 Ts-1.2 Ts;

the conversion from single bit to continuous bit or from continuous bit to single bit, because of Guassian filter, the time interval of amplitude change is longer than the jump from single bit to single bit, the time interval can be set to be larger than 1.2 Ts;

the conversion from continuous bit to continuous bit takes into account that the amplitude variation is maximum, and the threshold value can cover the influence of interference, so that the time interval does not need to be considered.

And only the extreme value which is set according with the time interval is reserved for filtering.

And (2) respectively carrying out filtering processing on the two maximum values and the minimum values:

and respectively filtering the maximum value when the bit is continuously held, the minimum value when the bit is continuously held, the maximum value when the bit is single held and the minimum value when the bit is single held. By configuring two filtering implementation methods of switching sliding average and/or Kalman filtering, the method is respectively suitable for application scenes with quick response and application scenes with long-term tracking and insensitivity to response time.

The moving average filtering mode is applied to rough estimation of frequency offset, and a frequency offset rough calculation result is fed back to the digital mixer to remove larger frequency offset. The filtering mode of the moving average has a common effect, but has the advantages that the response speed is higher, the method is suitable for data used for calibration, such as a lead code and a synchronous code, of a received signal just after entering a receiving mode, and the frequency offset calibration does not influence the receiving of effective data.

The Kalman filtering mode is applied to fine adjustment of frequency offset, a small amount of frequency offset residues are still reserved after rough estimation, filtering is performed through Kalman filtering, the statistical characteristics of noise can be continuously estimated and corrected, the response time of a Kalman filter is long, a certain time is required for reaching a stable state, the Kalman filtering mode is not suitable for an application scene of rapid calculation, but the precision is high, and the Kalman filtering mode is suitable for processing residual frequency offset caused by frequency drift or insufficient rough correction precision in the receiving process.

And (3) averaging the filtering result of the maximum value and the filtering result of the minimum value when the signals are continuously bit, averaging the filtering result of the maximum value and the filtering result of the minimum value when the signals are singly bit, averaging the two averaging results again, and finally obtaining a result which is the frequency deviation value of the current received signal.

Averaging the continuous bit maximum value filtering result and the continuous bit minimum value filtering result, and storing the result as MULTI _ AVG; averaging the maximum value filtering result and the minimum value filtering result of a SINGLE bit to store the result as SINGLE _ AVG, averaging the two averaging results again to obtain a final result which is the frequency deviation value of the current received signal, and performing frequency deviation estimation by completely adopting an actual measurement result, so that the accuracy is higher.