KR20010086792A - Data-aided frequency estimation method for PSK signaling in frequency-selective fading - Google Patents

Abstract

PURPOSE: A method for estimating a carrier frequency offset for a phase shift keying signal in a frequency selective fading channel is provided to improve the estimating performance of a radio communication system by estimating a carrier frequency offset with the aid of a training signal when there occurs a frequency error between carrier frequencies due to the Doppler or oscillator deflection between a transmitter and a receiver in the radio communication system. CONSTITUTION: A reference frequency estimator is provided(S10), wherein the reference frequency estimator is obtained through the extension of an estimator based on a maximal likelihood under the assumption that a channel impulse response is known. The estimator based on the maximal likelihood is modified to the reference frequency estimator proposed which does not need channel information by means of a training signal(S20).

Description

Data-aided frequency estimation method for PSK signaling in frequency-selective fading}

In the present invention, the limited oscillator accuracy or the Doppler effect during the synchronous demodulation of a PSK (Phase Shift Keying) series signal may cause a carrier frequency offset, which may cause a significant performance degradation of the demodulator. It relates to a method of estimating error.

Conventional data-assisted frequency estimators are based on the maximum likelihood (ML) theory, all of which assume white noise AWGN (Additve White Gaussian Noise) or flat fading channels.

As the prior art, US Patent Publication No. US05943606 and IEEE Trans. Commun., Vol. 43, pp 1 169-1178 and IEEE Trans. Commun., Vol. 46, pp553-560 .

The carrier frequency recovery and carrier frequency error estimation schemes disclosed in the above-mentioned US Patent Publication No. US05943606 have been proposed assuming channels that do not cause mutual interference signals, such as AWGN channels, and can only be used in AWGN and flat fading channels.

IEEE Trans. Commun., Vol. 43, pp1169-1178, relates to a carrier frequency error estimation scheme for an AWGN channel in a communication system given a training signal.

IEEE Trans. Commun., Vol. The technique disclosed in 46, pp553-560 also relates to a carrier frequency error estimation method for a frequency selective fading channel in a communication system in which a training signal is given.

The proposed method for the conventional frequency selective fading channel requires a correlation function of the channel, and does not consider the channel characteristics that change with time, resulting in poor performance.

SUMMARY OF THE INVENTION The present invention has been made to solve the problems of the prior art, and an object of the present invention is to provide a new frequency estimation method in a frequency selective fading channel under the assumption that the channel impulse response is known by extending the estimator for the existing AWGN. have.

As a technical idea for achieving the above object, a process of suggesting a base frequency estimator under the assumption that a channel impulse response is known in the estimation of a carrier frequency error during synchronous demodulation of a PSK (Phase Shift Keying) signal and the base frequency estimator The present invention is characterized by comprising a process of transforming to a frequency estimator that does not require channel information using a white type training signal.

1 is a block diagram showing a baseband system model.

2 is a flow chart according to the present invention.

3 is a block diagram illustrating a base frequency estimator.

4 is a configuration diagram illustrating an m th block in FIG. 3.

5 is a block diagram showing a frequency estimator according to the present invention.

6 is a diagram showing a fourth order moment for a training signal.

7 is a diagram illustrating a model used for simulation.

8 is a chart showing frequency error estimation performance at τ = T / 2 in a frequency selective fading channel.

9 is a chart showing frequency error estimation performance at τ = 3T / 4 in a frequency selective fading channel.

<Explanation of symbols for the main parts of the drawings>

1: transmission signal 2: Rayleigh fading channel

3: first multiplier 4: adder

5: second multiplier 6: reception filter

7: carrier offset estimator 8: switch

9: output signal 10: estimator input signal

11, 12, 13: autocorrelation function estimation unit

14: adder 15: arg calculation unit

16 multiplier 17 estimator output signal

Hereinafter, with reference to the accompanying drawings with respect to the configuration and operation of the present invention will be described in detail.

1 is a block diagram showing a baseband system model.

As shown in Fig. 1, a transmission signal 1, a Rayleigh fading channel 2 modeling a mobile communication channel, and a carrier frequency error Using a multiplier (3) to multiply, an adder (4) to add noise η (t), and an estimated carrier frequency error A multiplier (5) multiplying by the receiver, a receiver filter (6) connected to the multiplier, a switch (7) for intermitting a signal passing through the receiver filter, and a carrier frequency connected by feedback from the signal passing through the switch Error estimator 8.

In the above configuration, the transmission signal 1 is the sum of the product of the transmission symbol d (j) and the baseband waveform shaping function h (t), n (t) is a noise signal, θ is an initial random phase, and Δf is a carrier wave. Indicates frequency error.

Assuming perfect timing recovery, the output of the receive filter at t = kT is given by

Here, g _k (l) represents the response by the impulse 1 hour before the time k, and represents the equivalent channel impulse response in the baseband. g _k (l) represents both h (t) and the Rayleigh fading channel block in FIG. 1 in the discrete time domain and is L + 1 in length.

η (k) is the variance Assume that is a white Gaussian noise. L = 0 for flat fading and L≥1 for frequency selective fading channel.

2 is a flow chart according to the present invention.

As shown in Fig. 2, the estimator proposed in the present invention is derived through the following two steps.

First, based on the assumption that the channel impulse response is known, a process (S10) in which a base frequency estimator obtained through direct expansion of a conventional ML-based estimator is proposed,

Secondly, a process of transforming the proposed frequency estimator from the base frequency estimator to the proposed frequency estimator which does not require channel information is performed (S20).

Hereinafter, the above configuration will be described in more detail with reference to the following equation.

Training signal {d (k) | k = 1, ..., K} and {g _k (l)} is fixed during the training signal interval, i.e. g _k (l) = g (l), k = 1, ..., K Assuming, the received signal can be rewritten as the following equation.

If the impulse response of the channel, {g (l)}, is known, then the base estimator is proposed as follows.

Here, R _{l, l} (m) is an estimate of the autocorrelation function of γ _l (k), which is defined by the following equation.

In the estimator of Equation 3, if g (i) g ^* (l) included in R _{l, l} (m) is changed to E [g (i) g ^* (l)], the IEEE Trans. Commun., Vol. An estimator is obtained as in 46, pp553-560. In this case, the equation (5) is substituted into the equation (4).

Where H _l (m) is equal to the following equation.

Therefore, the estimator of Equation 3 can be rewritten as in the following equation.

3 is a block diagram of a frequency estimator called a base estimator.

As shown in FIG. 3, an adder 12 that adds a received signal r (k) 10, an estimate calculation unit 11, 12, 13 of an autocorrelation function for the received signal, and the result of the estimate calculation unit 12. ), An arg calculator 14 connected to the result of the adder, and the result of the calculation. It consists of a multiplier 15 to multiply by.

4 is 3 is a block diagram showing the m-th block in FIG.

The estimator to be proposed is derived from the base estimator. The frequency range that the base estimator can estimate is equal to | ΔfT | ≤0.5 / N.

Here, the problem of applying the proposed base estimator to the real system is that the problem of applying the proposed base estimator to the real system is that the channel parameters | g (l) | are not known in the frequency selective fading channel. The carrier frequency error must be corrected to estimate the values, which makes it difficult to apply this frequency estimator to the actual system.

Therefore, we propose a practical frequency estimator through the modification of the base estimator.

Substituting Equation 3 into Equation 7 approximates the following Equation.

Furthermore, we make the following assumptions.

Where 0 ≦ i ≦ L, 0 ≦ l ≦ L, 0 ≦ m ≦ L and 1 ≦ m ≦ N.

This assumption can be justified from the following equation if modeled as random signals with an average of 0 and a variance of 1, where d (k) is not correlated.

In general, Equation 10 is satisfied if d (k) is an output of white noise having an average of zero having a fourth order moment such as an impulse. This d (k) is called a white type signal. For this training signal, H _l (m) of Equation 8 is expressed by the following equation.

According to Equation 12 Equation 9 is modified as follows.

5 is a block diagram illustrating a proposed frequency estimator according to Equation (13).

As shown in FIG. 5, since the received signal r (k) through the frequency selective fading channel is present in the proposed estimator, there is a mutual interference signal. Data modulation is removed from each branch and the carrier frequency is estimated by differential decoding. Each branch is weighted with the absolute value of H_ {0} (m), which corresponds to the amplitude response of the channel if the training signal is white.

This estimator, shown in Figure 5, does not need any channel information. Therefore, if a white type training signal is provided, the proposed estimator may be useful in a real system.

However, this white type training signal should be used is the biggest problem in using the proposed estimator.

In general, these training signals should be searched for the full possible training signal pattern and then determined to be the signal pattern most similar to the whitest type.

In mobile communication systems such as GSM and IS-136, a given training signal is close to the white type.

Training signals for these systems are shown in Table 1.

6 shows the fourth order moments defined in Equation 11 for each training signal. Since the fourth moments are nearly impulse, these signals are white, so the proposed frequency estimator can be applied to GSM or IS-136 systems.

Gaussian minimum shift keying (GMSK), which is a modulation scheme defined by GSM, can be regarded as a linear modulation scheme, that is, offset quadrature phase-shift keying (OQPSK) through an appropriate basic waveform shaping function.

GSM 0,0,1,0,0,1,0,1,1,1,0,0,0,0,1,0,0,0,1,0,0,1,0,1,1, One IS-136

The proposed estimator and the existing estimator have the following differences. Firstly, the estimator designed assuming basically AWGN channel does not take into account the mutual interference signal generated by the channel at all, and therefore the performance may be poor in the frequency selective fading channel. Next, the ML-based estimator is a proposed structure that takes into account mutual interference signals, but does not consider a channel that changes with time, but fixes E [g (i) g ^* (l)] as a channel parameter as described above. By using these values, the frequency estimation performance may deteriorate significantly if the values of g (i) g ^* (l) and E [g (i) g ^* (l)] by the actual channel are different. This will be explained through the following computer simulation.

7 is a block diagram showing a system model used in the simulation.

The proposed estimator is compared with the proposed estimator in the frequency selective Rayleigh fading channel through computer simulation.

As the system model in this simulation, the structure shown in FIG. 5 was used.

The IS-136 training signal (K = 14) was used as the transmission signal and the Rayleigh fading channel with two rays was considered. This channel is modeled as

Here, α ₀ (t) and α ₁ (t) are complex Gaussian random processes that are independent of each other and have an average of zero. τ represents the time delay between the two rays.

It is assumed here that the powers of the two ray are the same as in the following equation.

In Equation 14, two parameters, T / 2 and 3T / 4, were used and were fixed at L = 2. The carrier frequency is assumed to be 900MHz and the speed of the moving object is 100km / h. The normalized carrier frequency error ΔfT varied from 0 to 0.5.

In addition, the IEEE Trans. Commun., Vol. In designing the estimator proposed in 46, pp553-560, the exponential channel correlation function is assumed as follows.

8 and 9 are diagrams showing simulation results, and root mean square (RMS) frequency estimation error according to a carrier frequency error, that is, Indicates.

Increasing the N value for all estimators reduces the frequency estimation range, but yields a more precise estimate. In all cases the proposed estimator outperformed the conventional methods.

IEEE Trans. Commun., Vol. The estimator proposed in 43, pp1169-1178 is the worst performing because it is proposed in the AWGN channel. See also IEEE Trans. Commun., Vol. The estimator proposed in 46, pp553-560 also shows a larger RMS error value than the proposed method because it does not take into account the characteristics of the channel that change over time.

As can be seen from the above description, the present invention is directed to a training signal when a frequency error occurs between carrier frequencies due to a deviation of a Doppler or oscillator between a transmitter and a receiver in a wireless communication system in a frequency selective fading channel environment. By estimating the carrier frequency error, it is possible to improve the estimation performance and to more accurately compensate the carrier frequency error while mitigating the oscillator constraint that is within accuracy of 0.3-5ppm if it is applied to mobile communication system such as GSM or IS-136. Can be.

Claims (8)

In Estimating Carrier Frequency Error During Synchronous Demodulation of Phase Shift Keying (PSK) Series Signal, The process of suggesting a base frequency estimator on the assumption that the channel impulse response is known, And a step of transforming the proposed frequency estimator from the base frequency estimator into a proposed frequency estimator which does not need channel information using the white type training signal. The method of claim 1, wherein the base frequency estimator is a following equation. Where g (l) is the impulse response of the channel 3. The method of claim 2, wherein H ₁ (m) is the following equation. 4. Where d (k): training signal r (k): Received signal 4. The method of claim 3, wherein r (k) is the following equation. Where η (k): noise Δf: carrier frequency error θ: initial random phase The method of claim 1, wherein the proposed frequency estimator is a following equation. 6. The method of claim 5, wherein H _l (m) is the following equation. Where g (l) is the impulse response of the channel Δf: carrier frequency error 7. The method of claim 6, wherein H _l (m) is made by assuming the following equation 22 in the following equation 23. Where 0≤i≤L, 0≤l≤L, 0≤m≤L, 1≤m≤L The method of claim 7, wherein Equation 21 may be justified from the following equation if d (k) is modeled as random signals having an uncorrelated average of 0 and a variance of 1. A method for estimating data-assisted carrier frequency error for a PSK signal in a frequency selective fading channel, characterized in that d (k) satisfies if the output is a white noise with an average of zero having a fourth order moment such as an impulse.