JP2540931B2 - PSK signal demodulation method - Google Patents

Description

TECHNICAL FIELD The present invention relates to a synchronous detection demodulation method for PSK modulated signals.

(Prior Art) Demodulation methods of PSK modulated signals are roughly classified into a synchronous detection method and a differential detection method. Among them, the synchronous detection method has a more complex demodulator configuration than the differential detection method, but has an excellent demodulation bit error rate characteristic. Therefore, PSK
As a demodulation method of a modulation signal, a synchronous detection method has been widely used in recent years.

In the synchronous detection method, it is necessary for the receiving side to regenerate the reference carrier that is phase-synchronized with the received signal carrier, but as the carrier regenerating method, the multiplication / division method, PLL (P
hase Locked Loop) method has been proposed.

(Problems to be solved by the invention) According to these conventional carrier recovery methods, when the line condition is poor, for example, when the received signal is weak, that is, the carrier power to noise power ratio (C / N) of the received signal is When it is low, it takes a long time to establish phase synchronization between the received signal carrier and the regenerated carrier, and it is easy to lose synchronization due to fluctuations in the receiving frequency and C / N. There was a problem that it took a long time to do.

As described above, in the conventional PSK signal synchronous detection method, carrier recovery is possible, but the C / N of the received signal, the frequency deviation of the steady received signal, or the magnitude of the frequency fluctuation is limited. However, there is a problem that the carrier wave cannot be accurately reproduced when the line condition is poor, and the application of the communication method using the PSK signal is limited.

The present invention has been made to solve the above-mentioned drawbacks of the prior art, and it is possible to accurately demodulate even when the C / N of the received signal is very low and the received signal has a large frequency deviation. It is an object to provide a PSK signal demodulation method that can be performed.

(Means for Solving the Problems) A feature of the present invention is that after phase-modulated PSK reception signals are quasi-coherently detected, sampling is performed at a predetermined cycle to convert PSK signals converted into digital values by digital signal processing. In the PSK signal demodulation method for demodulating, one of the branched PSK signals by branching the digitized PSK signal is converted into a frequency domain signal from a time domain received signal by an FFT to obtain its power spectrum, Estimating the carrier frequency of the received signal from the obtained power spectrum,
After correcting the frequency of the PSK signal in the other branched time domain using the estimated carrier frequency, a noise component added on the transmission line by a bandpass filter having the estimated carrier frequency at the center After performing the clock regeneration of one of the branched ones, the signal conversion component of the other branched one is removed to form an unmodulated signal, and a sliding DFT that performs a sequential process adapted to the input frequency fluctuation.
Converted to the frequency domain by the method, after correcting the frequency deviation deviation and phase of the received signal clock-regenerated by reducing the noise component using the unmodulated signal converted to the frequency domain, demodulation of the PSK signal There is something to do.

(Principle of the Invention) The present invention is to perform quasi-coherent detection of a PSK signal on the receiving side once with a reference carrier having a frequency close to the carrier frequency of the received signal, and obtain the received signal and reference signal by quasi-synchronous detection A PSK modulated signal with a difference in carrier frequency is converted into digital data by A / D conversion. The power spectrum of the received signal in the time domain, which is a digital value, is obtained by the FFT method, and the carrier frequency of the received PSK signal is estimated in the frequency domain. The received signal in the time domain is noise-reduced by a band-pass filter having the estimated frequency as the center frequency or by correcting the estimated signal at the estimated frequency and a low-pass filter. The noise-reduced received signal in the time domain is, for example, multiplied by M to become a non-modulated signal in order to remove a modulation component. Here, M represents the number of phases of the PSK signal, and for example, in the case of a two-phase PSK (BPSK) signal, M =
It becomes 2. The time domain unmodulated signal then slides
The power spectrum is obtained by the DFT method (hereinafter referred to as “S-DFT”), and the carrier frequency and phase are estimated.
Then, if the quasi-coherent detection signal in the time domain is corrected using these estimated frequency and phase values, it means that the PSK signal is coherently detected.

Here, the S-DFT method does not collectively convert an input consisting of a certain N points into a signal in the frequency domain block by block as in a normal DFT or FFT method, but is described in, for example, the literature (LR Rabiner et al., “Theory and Application of D
igital Signal Processing ", Prentice-Hall 1975), the DFT operation is sequentially performed on the sample values in each time domain of the input. Generally, in the DFT or FFT method, the input is observed. The longer the time, the more accurate the frequency spectrum can be obtained, but the more points, the longer the calculation time.On the other hand, if the S-DFT method is used, conversion to the frequency domain without delay is performed. Is possible with high accuracy.

Therefore, in the present invention, by performing carrier recovery in the frequency domain by the S-DFT method using a non-modulated signal in the time domain, it is possible to obtain an accurate PSK signal with low reception C / N and even when the frequency deviation is large. This enables demodulation.

(Example) The present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram of a demodulation device using the PSK signal demodulation method according to the present invention. In the following, input PSK
It is assumed that the modulation signal is an M-phase PSK signal.

In the figure, 100 is an input terminal of the received PSK signal, 1 is a band pass filter (hereinafter referred to as "BPF"), 2'is a multiplier, 3 is a π / 2 phase shifter, 4 is a fixed frequency oscillator, 5,
5'is a low-pass filter (hereinafter referred to as "LPF"), 6 is A
/ D (analog / digital) converter, 7 is a digital signal processing type demodulator which is a characteristic part of the present invention, 111 is a demodulated data output terminal, and 101 to 110 are respective signals.

FIG. 6 is a flow chart for schematically explaining the flow of the PSK signal demodulation method according to the present invention, and the present embodiment will be described below also using this figure.

The M-phase PSK signal represented by the following equation is input to the input terminal 100.

In equation (1), noise added on the transmission path is not considered. A indicates the amplitude level, ω _c ′ indicates the carrier angular frequency, θ ₀ indicates the initial phase, and θ _k is the information phase corresponding to the kth information bit. For example, two-phase PSK (M
= 2), 0 or π, and 4-phase PSK (M = 4), 0, π /
Takes 2, π, or 3π / 2. Here, the angular frequency ω _c ′ of the carrier wave of the received signal expressed by the equation (1) is determined by the stability of the transmission carrier wave and the stability of the frequency converter, fixed oscillator, etc. used in the line. Frequency ω _c
Has a different value than. That is, on the receiving side (ω _c ′ −ω
It will have a frequency deviation of _c ). Therefore, the bandwidth of the bandpass filter 1 must be determined in consideration of the maximum value of the frequency deviation that can occur in the transmission line. The out-of-band signal and the signal from which noise has been removed by the band-pass filter 1 are quasi-coherently detected by the reference carrier obtained from the fixed oscillator 4 (Step 1 in FIG. 6).

The reference carrier wave 104 (R _c ) generated by the fixed oscillator 4 and the reference carrier wave 103 (R _s ) output from the π / 2 phase shifter 3 are expressed as follows.

In equations (2) and (3), ω _c ″ and θ ₁ respectively indicate the angular frequency and the initial phase value of the fixed oscillator 4. Here, ω
_c ″ has a value close to the transmitted angular frequency ω _c , and in the following description it is assumed that ω _c = ω _c ″.

The received signals 101 and 102 of the equation (1) have the following equations (2) and (3).
The reference carriers 103 and 104 indicated by are multiplied to obtain signals 105 and 106 as in the following equation.

_{e 1 (t) = S k} (t) · R c (t) = Acos {(ω c '-ω c) t + θ k + (θ 0 -θ 1)} + Acos
{(Ω _c ′ + ω _c ) t + θ _k + θ _o + θ ₁ } ... (4) e ₂ (t) = S _k (t) · R _s (t) = −Asin {(ω _c ′ −ω _c ) T + θ _k + (θ ₀ −θ ₁ )} + Asin {(ω
_c ′ + ω _c ) t + θ _k + θ _o + θ ₁ } (5) The signals represented by the equations (4) and (5) have their harmonic components removed by low-pass filters (LPFs) 5 and 5 ′, respectively. As a result, the signals 107 and 108 are expressed by the following equations.

e ₁ ′ (t) = Acos (Δωt + θ _k + Δθ) (6) e ₂ ′ (t) = − A sin (Δωt + θ _k + Δθ)
(7) Here, Δω and Δθ represent the frequency deviation amount (ω _c ′ −ω _c ) between the received signal and the reference carrier and the phase difference (θ ₀ −θ ₁ ) between the received signal and the reference carrier, respectively. .

The signals represented by the equations (6) and (7) are converted into digital signals 109 and 110 by the A / D converter 6 as in the following equations (Step 2 in FIG. 6).

x _i = e ₁ ′ (t _i ) = A cos (Δωt _i + θ _k + Δθ) (8) y _i = e ₂ ′ (t _i ) = − A sin (Δωt _i + θ _k + Δθ) (9) ) However, t _i represents the i-th sample time.

These digital data signals are sequentially input to the digital signal processing demodulation circuit 7, which is a characteristic part of the present invention.

These signals are signals that have passed through the band pass filter (BPF) 1, and assuming that the frequency deviation of the received signal is large, the bandwidth of BPF1 must be wide as described above. In this case, these signals x _i , y _i will include noise corresponding to the bandwidth of BPF1, and the larger the reception frequency deviation, the worse the C / N of the BPF1 output signal, and the carrier recovery. It will be difficult. Therefore, in the present invention, in order to demodulate the received signal in such a bad channel condition, the signal converted into the frequency domain using digital signal processing is processed to reduce the noise level. There is.

FIG. 2 is a concrete block diagram of the digital signal processing type demodulation circuit 7 according to the present invention. Signals x _i and y _i represented by equations (8) and (9) are input to the input terminals indicated by 109 and 110 in FIG. These signals are converted into frequency domain signals by the FFT circuit 8 (Step in FIG. 6).
3). The output of the FFT circuit 8 is a real value in the frequency domain,
An imaginary value amplitude spectrum is obtained.

The frequency estimator 9 estimates the frequency after obtaining the power spectrum of the input signal from these amplitude spectra (Step 4 in FIG. 6). In the frequency domain, the white noise component has a constant power level, whereas the modulated signal component has a large power level only in the band in which the signal exists. Therefore, a noise component irrelevant to the modulation signal having a frequency correlation can be easily identified in the frequency domain despite the low received C / N. In this way, the center frequency of the received signal can be estimated relatively accurately using the FFT method even when the line condition is poor.

The input signals x _i and y _i are frequency-corrected in the time domain by the frequencies estimated by the frequency estimator 9 in the multipliers 10 and 10 ′, whereby the signals whose center frequency is 0 are obtained.
The noise component is removed from the signal, which is almost in the base band, by the LPF 11 having a pass band width similar to the modulation band width (Step 5 in FIG. 6). By the above operation, the C / N of the received signal is improved by the bandwidth of LPF11 / the bandwidth of BPF1, and carrier recovery described later becomes easy. Do here
Frequency estimation using the FFT method shall be performed as long as resynchronization is necessary at the time of initial synchronization of continuous mode signals, or when synchronization is lost or occurs once synchronization is established.

The output signal of the LPF11 is multiplied by M by the multiplier 12 to obtain the expression (1
An unmodulated signal S ₂ such as 0) is obtained (Step 6 in FIG. 6).

This unmodulated signal represents the received carrier component itself. As the output of the multiplier 12, the noise component is actually included in the signal of Expression (10).

The unmodulated signal is then transmitted to the sliding DFT.
The signal is input to the circuit (hereinafter referred to as "S-DFT") 13 and the frequency spectrum is calculated (Step 7 in FIG. 6). S-D
The FT13 is for converting a signal, which is sequentially input, for each time sample into a frequency domain. That is, the frequency component that changes with the passage of time can be obtained for each time sample. As the output of the S-DFT 13, a maximum of N frequency components can be obtained for N signals in the time domain from the first sample to the Nth sample, as in the case of a normal DFT. For each subsequently input sample, frequency components corresponding to the N time signals input up to that point are output.

FIG. 3 is a specific block diagram of the S-DFT13 circuit used in the present invention. Here, the signal is a complex signal. In Figure 3, the 30 _a to 30 _n in the delay circuit (Z ^-N), N is the number of delay samples, the 31 _a to 31 _n are adders, the 32 _a to 32 _m represent multipliers, respectively. Further, the input terminal 120 _a to 120 _m, the weighting factor determined from the frequency and phase estimator 14 of FIG. 2 is given. Thus, up to N frequency values can be sequentially obtained for the i-th input of a certain time sample.

FIG. 4 is a block diagram showing a specific configuration of the frequency / phase estimator 14, and FIG. 5 is a frequency component diagram of the input signal (output of the S-DFT 13) of the frequency / phase estimator 14.

5, Y ₀ , Y ₁ , ..., Y _m are S-DFT 13
The output signal of FIG. 4 shows that the unmodulated signal has a large level frequency component in the frequency domain. Therefore, in the maximum value level determination unit 41 shown in FIG. 4, the input signals Y ₀ , Y ₁ , ...
..., the frequency component of the highest level among Y _m is determined, and the frequency / phase calculation unit 42 uses the maximum level frequency component determined by the maximum value level determination unit 41 or the adjacent frequency component as a carrier wave. The frequency and phase are interpolated (estimated) and output (Step 8 in FIG. 6). Next, the correction signal generator 43 generates a correction signal for correcting the received signal based on the obtained carrier frequency and phase.

On the other hand, in order to calculate m frequency components centering on the frequency f _i from the carrier frequency estimated by the frequency / phase calculating unit 42, the weighting factor determining unit 44 uses the weighting factors W (0) to W (W).
(M) is determined and output. If the received carrier frequency is
f _{i +} when shifted to _1, this time the frequency f _{i + 1} to calculate the m-number of frequency components around the a new weight coefficient W (0)
It is sufficient to give ~ W (m).

That is, as shown in FIG. 4, from the frequency estimation result of the frequency / phase estimator 14, weighting factors (0) to W are calculated so that only m frequency components centering on the received carrier frequency are calculated.
Even if the frequency of the non-modulated signal is deviated, the frequency component less than N centered on the frequency may be always obtained by giving (m).

As described above, in the present invention, the circuit scale can be suppressed by calculating the frequency components only in the vicinity of the frequency of interest. In addition, even when the carrier frequency changes rapidly,
It is possible to follow. Furthermore, in the S-DFT method for frequency estimation, the carrier power to noise power ratio in the frequency domain is improved by making the observation time of the input signal sufficiently long, and the S-DFT method using the estimation result of a small number of frequency components is used. -It is also possible to reduce the circuit scale of the DFT13. In this way, the frequency pull-in range is not limited unlike the conventional PLL carrier recovery method, and the pull-in range can be made very wide with respect to slow frequency fluctuations.

When obtaining the frequency component, the frequency spectrum of the non-modulated signal obtained here has a large power as a line spectrum at the frequency MΔω. On the other hand, since the noise component is present at all frequencies in the band on average, it can be said that the method performed in the frequency domain is excellent for estimating the frequency and phase of the unmodulated signal.

In the S-DFT 13, as in the ordinary DFT or FFT method, the longer the observation time of the input time signal, that is, the more the number of points, the higher the resolution of the obtained frequency spectrum, and therefore the accuracy is good. Frequency estimation is possible. Since the input unmodulated signal is a line spectrum, a high frequency resolution means that the power of the noise component is small because the frequency step size is small, and a relatively high level line spectrum is obtained. Therefore, even when the line condition is poor and the received C / N is low, the unmodulated signal, which is the received carrier component, can be easily identified in the frequency domain. However, the scale of the circuit increases in proportion to the number of points N, and the operation time will increase with the usual DFT and FFT methods. However, here, it suffices to know only the frequency having the maximum line spectrum in the band.
Unlike the usual DFT and FFT methods, T13 does not need to calculate all N components as described above, so if the frequency fluctuation is continuous in time, the result of frequency estimation by the frequency / phase estimator 14 is used. Then, only the frequency component near the carrier frequency needs to be calculated, and the circuit scale can be reduced.

Carrier frequency estimation performed in the frequency domain is low reception C / N.
In addition to being able to perform with high accuracy, there is an advantage that the estimation is easy even if there is a large frequency deviation in the carrier frequency. This is the conventional multiplication / division method or PLL
This is a characteristic point of the present invention that cannot be obtained by the carrier recovery method by the method.

However, the estimated value obtained in the frequency domain is still S-
There is only the accuracy of the resolution Δf of the DFT 13, and there are also errors due to the S-DFT 13. Therefore, even if the frequency and phase are corrected using the estimation result, a very slight frequency error may remain. In such a case, the frequency domain data obtained with a small number of samples can be used to obtain a more accurate frequency and phase by using an interpolation formula, or a simple configuration with good followability can be used instead of a narrow pull-in frequency range. It is also possible to realize highly accurate correction by the PLL.

The frequency deviation and the phase value estimated by the frequency / phase estimator 14 are input to the frequency / phase corrector 16. Here, the frequency of the signal whose clock component has been reproduced by the clock reproducing circuit 15 (Step 9 in FIG. 6) and the frequency deviation and the phase are corrected by the phase corrector 16, and the carrier is reproduced here. (Step 10 in FIG. 6).

The signal reproduced from the carrier wave is judged as information bit data by the judging device 17, and demodulated data is output to the output terminal 111 (Step 11 in FIG. 6).

(Effects of the Invention) As described in detail above, according to the present invention, a weighting coefficient is adaptively obtained by S-DFT and frequency / phase estimation to convert it into a frequency domain signal for PSK demodulation. , The deviation amount of the received carrier frequency is large and low C /
Even in the N state, it is possible to accurately demodulate in a short time.

Further, in the present invention, by using the S-DFT method, it is possible to demodulate successively input time-continuous signals without delay time.

[Brief description of drawings]

FIG. 1 is a block diagram of a demodulation device using a PSK signal demodulation method according to the present invention, FIG. 2 is a block diagram of a digital signal processing type demodulation circuit according to the present invention, and FIG. 3 is an outline of an S-DFT used in the present invention. FIG. 4 is a concrete block diagram of the frequency / phase estimator according to the present invention, and FIG. 5 is an input signal (S-DFT13) of the frequency / phase estimator 14.
Output), and FIG. 6 is a flow chart for schematically explaining the flow of the PSK signal demodulation method according to the present invention. 1 ... band pass filter (BPF), 2, 2 '... multiplier, 3 ... π / 2 phase shifter, 4 ... fixed frequency oscillator, 5, 5' ... low pass filter (LPF), 6 ... A / D (analog / digital) converter, 7 ... digital signal processing demodulator, 8 ... FFT circuit, 9 ... frequency estimator, 10, 10 '... multiplier, 11 ... LPF , 12 …… Multiplier, 13 …… S-DFT, 14 …… Frequency / phase estimator, 15 …… Clock regenerator, 16 …… Frequency / phase corrector, 17 …… Judger, 30 _a to 30 _n ...... Delay circuit (Z ^-N ), 31 _{a to} 31 _n ...... Adder, 32 _{a to} 32 _m ...... Multiplier, 41 ...... Maximum value level judgment unit, 42 ...... Frequency / phase calculation unit, 43 ... ... correction signal generation unit, 44 ...... weighting factor generator, an input terminal 100 ...... received PSK signal, 101 to 110 ...... signal, 111 ...... demodulated data output terminal, 120 _a to 120 _m ...... input terminal

Claims (1)

(57) [Claims] 1. A PSK signal demodulating method for demodulating a PSK signal, which has been phase-modulated by quasi-coherent detection and then sampled at a predetermined cycle and converted into a digital value by digital signal processing, in a continuous mode. At the time of initial synchronization of the signal, or when resynchronization is necessary after the synchronization is once established, the digitized PSK signal is branched and one of the branched PSK signals is subjected to FFT to convert the time domain signal to the frequency domain. To obtain the power spectrum of the received signal, estimate the carrier frequency of the received signal from the obtained power spectrum, and use the estimated carrier frequency to split the other time domain PSK.
After correcting the frequency of the signal, for the frequency-corrected PSK signal, a noise component added on the transmission path is reduced by a bandpass filter having the estimated carrier frequency at the center, and the PSK after passing the bandpass filter. The signal is branched, and clock recovery is performed for one of the branched PSK signals, and the modulation component of the time domain signal is removed for the other of the branched PSK signals to obtain a non-modulated signal, and then input frequency fluctuation or The unmodulated signal is converted into a frequency domain signal using a sliding DFT method that performs highly accurate domain conversion by performing sequential processing adapted to the deviation, and the unmodulated signal converted into the frequency domain is used to A PSK signal demodulation method characterized in that the PSK signal is demodulated after correcting the frequency deviation and phase of the clock-regenerated PSK signal.