JP5631220B2 - Symbol estimation circuit and demodulation circuit - Google Patents

Description

  The present invention relates to a symbol estimation circuit capable of estimating a symbol rate when demodulating a reception signal modulated at an unknown symbol rate, and a demodulation circuit mounted with the symbol estimation circuit. .

With the rapid development of software defined radio (SDR) using programmable signal processing devices such as DSP (Digital Signal Processor) and FPGA (Field-Programmable Gate Array) in recent years, the same H / W platform has been developed. Multi-mode communication devices capable of supporting a plurality of communication systems have been developed.
In addition, automatic selection of the optimal communication system according to changes in the communication environment, environment-adaptive communication that dynamically optimizes the modulation type, and adaptive modems that realize higher transmission quality and throughput according to the transmission path conditions Or a system for monitoring incoming radio waves has been developed.

In the above-described multimode communication device and system, attention is focused on a modulation type identification circuit that estimates a modulation type of a received signal that is an incoming wave.
However, when the modulation type identification circuit estimates the modulation type of the received signal, it is necessary to know the specifications such as the center frequency and symbol rate of the received signal in advance. If not known, the modulation type of the received signal cannot be estimated and the received signal cannot be demodulated.
The following Patent Documents 1 and 2 disclose a technique for estimating a modulation type of a received signal, but it is assumed that specifications such as a center frequency are known.

JP 2002-064577 A JP 2004-248219 A

  Since the conventional modulation type identification circuit is configured as described above, if the specifications such as the center frequency of the received signal and the symbol rate are not known in advance, the modulation type of the received signal cannot be estimated. There was a problem that the received signal could not be demodulated.

The present invention has been made to solve the above problems, and an object of the present invention is to provide a symbol estimation circuit capable of estimating a symbol rate when demodulating a received signal modulated at an unknown symbol rate. And
Another object of the present invention is to obtain a demodulation circuit capable of demodulating a received signal whose modulation type is unknown using the symbol rate estimated by the symbol estimation circuit.

  The symbol estimation circuit according to the present invention converts an IQ signal orthogonally detected from a received signal modulated at an unknown symbol rate into a frequency domain signal, and obtains a frequency value at which peak power is obtained in the frequency domain signal. A symbol coarse estimator for coarsely estimating the symbol rate from the frequency value, and detecting a Nyquist phase in each symbol from the IQ signal, calculating a frequency deviation from the Nyquist phase in each symbol, A symbol precision estimator for adding the deviation to the symbol speed roughly estimated by the symbol coarse estimator is provided.

  According to the present invention, an IQ signal orthogonally detected from a reception signal modulated at an unknown symbol rate is converted into a frequency domain signal, and a frequency value at which peak power is obtained in the frequency domain signal is specified. A symbol coarse estimator that roughly estimates a symbol rate from the frequency value, a Nyquist phase in each symbol detected from the IQ signal, a frequency deviation is calculated from the Nyquist phase in each symbol, and the frequency deviation is symbolized Since a symbol precision estimator for adding to the symbol rate roughly estimated by the coarse estimator is provided, the symbol rate can be estimated when demodulating a received signal modulated at an unknown symbol rate. effective.

It is a block diagram which shows the demodulation circuit which mounts the symbol estimation circuit by Embodiment 1 of this invention. It is a block diagram which shows the symbol rough estimator of the symbol estimation circuit by Embodiment 1 of this invention. It is a block diagram which shows the symbol precision estimator of the symbol estimation circuit by Embodiment 1 of this invention. It is explanatory drawing which shows symbol frequency (omega) _PT1 after the interpolation process which is a detection result of the symbol information detector 10, and peak electric power MAX (PAVE1 ((omega))). Is an explanatory diagram showing an example of a symbol rate deviation is the Nyquist phase θ _{N (n)} and the frequency deviation Delta] f _R in each symbol that is detected by the BTR detector circuits 61 and 62.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a demodulating circuit equipped with a symbol estimating circuit according to the first embodiment of the present invention.
In FIG. 1, a symbol estimation circuit 1 is a device that estimates a symbol rate when demodulating a received signal modulated at an unknown symbol rate.
The symbol coarse estimator 2 is implemented with a band limiting filter 2a that receives an IQ signal orthogonally detected from a received signal modulated at an unknown symbol rate and removes noise (Gaussian noise) superimposed on the IQ signal. ing.
The coarse symbol estimator 2 converts the IQ signal after noise removal by the band limiting filter 2a into a frequency domain signal, specifies a frequency value at which peak power can be obtained from the frequency domain signal, and generates a symbol from the frequency value. A process for roughly estimating the speed is performed.

  The band limiting filter 2a is, for example, an occupied bandwidth estimation value Fw output from an estimation circuit (not shown) that estimates the occupied bandwidth by implementing a 99% method or a 3dB method, or the known information by a spectrum analyzer or the like. (Measured occupied bandwidth value) is input, and the IQ signal is passed at a cutoff frequency that is α times the estimated occupied bandwidth value Fw (for example, about 1.3 times = 1.3 × Fw). . As a result, Gaussian noise is removed to such an extent that the IQ signal necessary for symbol band estimation does not drop due to filtering.

The symbol precision estimator 3 detects the Nyquist phase in each symbol from the IQ signal after noise removal by the band limiting filter 2a, calculates the frequency deviation from the Nyquist phase in each symbol, and calculates the frequency deviation as a symbol coarse estimator. 2 is added to the roughly estimated symbol speed.
The demodulator 4 demodulates the received signal using the symbol rate to which the frequency deviation is added by the symbol precision estimator 3 of the symbol estimation circuit 1.

FIG. 2 is a block diagram showing a rough symbol estimator 2 of the symbol estimation circuit 1 according to the first embodiment of the present invention.
In FIG. 2, the symbol information detector 10 performs a nonlinear process and an FFT (Fast Fourier Transform) process on the IQ signal, thereby obtaining a frequency value at which a peak power MAX (PAVE1 (ω)) is obtained in the frequency domain signal. A process of specifying ω _PT1 is performed. The symbol information detector 10 constitutes a first symbol information detector.

The symbol information detector 20 performs a process of specifying a frequency value ω _PT2 at which a peak power MAX (PAVE2 (ω)) is obtained in a frequency domain signal by performing a zero cross point detection process and an FFT process of the IQ signal. To implement. The symbol information detector 20 constitutes a second symbol information detector.
The symbol information detector 30 performs a frequency variation detection process, a nonlinear process, and an FFT process on the IQ signal to obtain a frequency value ω _PT3 from which a peak power MAX (PAVE3 (ω)) is obtained in the frequency domain signal. Perform the specified process. The symbol information detector 30 constitutes a third symbol information detector.

The LPF unit 11 of the symbol information detector 10 performs a filtering process on the orthogonally detected IQ signal S _R (n). This filtering process uses, for example, a root Nyquist filter (for example, cutoff frequency fc = 3 dB: occupied bandwidth estimated value Fw × 1.3), and corresponds to the filtering process of the band limiting filter 2a.
The nonlinear unit 12 performs nonlinear processing (square processing) on the IQ signal S ′ _R (n) after filtering by the LPF unit 11 and outputs the IQ signal f (n) after nonlinear processing.

The FFT unit 13 multiplies the IQ signal f (n) after nonlinear processing by the nonlinear unit 12 by the Hanning window function coefficient W (n), and then performs complex FFT processing of N points (for example, 1024 points).
The FFT time domain averaging unit 14 performs time addition processing a plurality of times on the plurality of FFT results F (ω) by the FFT unit 13, so that the frequency domain signal in the frequency value ω range from 0 to 1023 is obtained. The power PAVE1 (ω) is calculated.

The peak power detection unit 15 detects the highest peak power MAX (PAVE1 (ω)) among the power PAVE1 (ω) of the frequency domain signal calculated by the FFT time domain averaging unit 14, and the peak power MAX ( A process of outputting a frequency value ω _{PAVE1 from} which PAVE1 (ω)) is obtained as a symbol frequency is performed.
The Lagrangian interpolation unit 16 _performs an interpolation process (for example, a Lagrange secondary interpolation process) on the symbol frequency ω _PAVE1 output from the peak power detection unit 15, and _performs the symbol frequency ω _PT1 and the peak power MAX (PAVE1) after the interpolation process. (Ω)) is output to the frequency value selector 40.

Frequency deviation detecting unit 21 of the symbol information detector 20 from a quadrature detection IQ signal S _{R (n),} for example, by implementing such multiplication method to estimate the center frequency Δfc the IQ signal S _{R (n)} . Here, the frequency deviation detector 21 estimates the center frequency Δfc, but the center frequency Δfc may be given in advance as known information.
The frequency rotator 22 performs processing for rotating the frequency of the IQ signal S _R (n) by the center frequency Δfc estimated by the frequency deviation detector 21.

The LPF unit 23 performs a filtering process on the IQ signal S _ROT (n) after the frequency rotation process by the frequency rotation unit 22. This filtering process uses, for example, a root Nyquist filter (for example, cutoff frequency fc = 3 dB: occupied bandwidth estimated value Fw × 1.3), and corresponds to the filtering process of the band limiting filter 2a.
The zero cross detection unit 24 performs a zero cross point detection process between the IQ signal S ′ _ROT (n) after the filtering process by the LPF unit 23 and the IQ signal S ′ _ROT (n−1) one symbol before, A pulse wave IQ signal f (n) is generated.

The FFT unit 25 multiplies the IQ signal f (n) of the pulse wave generated by the zero cross detection unit 24 by the Hanning window function coefficient W (n), and then performs complex FFT processing of N points (for example, 1024 points). carry out.
The FFT time domain averaging unit 26 performs a plurality of times of time addition processing on the plurality of FFT results F (ω) by the FFT unit 25, so that the frequency domain signal in the frequency value ω range from 0 to 1023 is obtained. The power PAVE2 (ω) is calculated.

The peak power detector 27 detects the highest peak power MAX (PAVE2 (ω)) among the power PAVE2 (ω) of the frequency domain signal calculated by the FFT time domain averaging unit 26, and the peak power MAX ( A process of outputting a frequency value ω _{PAVE2 from} which PAVE2 (ω)) is obtained as a symbol frequency is performed.
The Lagrangian interpolation unit 28 _performs an interpolation process (for example, a Lagrange secondary interpolation process) on the symbol frequency ω _PAVE2 output from the peak power detection unit 27, and the symbol frequency ω _PT2 and the peak power MAX (PAVE2) after the interpolation process. (Ω)) is output to the frequency value selector 40.

The LPF unit 31 of the symbol information detector 30 performs a filtering process on the orthogonally detected IQ signal S _R (n). This filtering process uses, for example, a root Nyquist filter (for example, cutoff frequency fc = 3 dB: occupied bandwidth estimated value Fw × 1.3), and corresponds to the filtering process of the band limiting filter 2a.
The frequency discriminator unit 32 differentiates the phase information θ (n) of the IQ signal S ′ _R (n) after the filtering process by the LPF unit 31 to obtain the instantaneous frequency fluctuation of the IQ signal S ′ _R (n). To detect.
The non-linear unit 33 performs non-linear processing (square processing) on the detection result of the frequency discriminator unit 32 and outputs a signal f (n) after the non-linear processing.

The FFT unit 34 multiplies the signal f (n) after nonlinear processing by the nonlinear unit 33 by a Hanning window function coefficient W (n), and then performs complex FFT processing of N points (eg, 1024 points).
The FFT time domain averaging unit 35 performs a plurality of times of time addition processing on the plurality of FFT results F (ω) by the FFT unit 34, so that the frequency domain signal in the range of frequency values ω from 0 to 1023 is obtained. The power PAVE3 (ω) is calculated.

The peak power detection unit 36 detects the highest peak power MAX (PAVE3 (ω)) among the power PAVE3 (ω) of the frequency domain signal calculated by the FFT time domain averaging unit 35, and the peak power MAX ( A process of outputting a frequency value ω _{PAVE3 from} which PAVE3 (ω)) is obtained as a symbol frequency is performed.
The Lagrangian interpolation unit 37 _performs an interpolation process (for example, a Lagrange secondary interpolation process) on the symbol frequency ω _PAVE3 output from the peak power detection unit 36, and the symbol frequency ω _PT3 and the peak power MAX (PAVE3) after the interpolation process. (Ω)) is output to the frequency value selector 40.

The frequency value selection section 40 is a symbol frequency (frequency value) at which the highest peak power is obtained from the interpolated symbol frequencies ω _PT1 , ω _PT2 , ω _PT3 output from the symbol information detectors 10, 20, 30. A process of selecting ω _PT is performed.
The symbol rate calculation unit 50 performs a process of calculating the symbol rate f _R from the symbol frequency ω _PT selected by the frequency value selection unit 40.

FIG. 3 is a block diagram showing the symbol precision estimator 3 of the symbol estimation circuit 1 according to the first embodiment of the present invention.
In FIG. 3, BTR by non-linear processing (Bit Timing Recovery) detection circuit 61 symbol frequency omega _PT selected by the frequency value selection unit 40, the symbol information detector symbol frequency after the output interpolation process from 10 omega _PT1 or In the case of the symbol frequency ω _PT3 after the interpolation process output from the symbol information detector 30, each non-linear process is performed on the IQ signal S ′ _R (n) after the noise removal by the band limiting filter 2a. A process for detecting the Nyquist phase θ _N (n) in the symbol is performed.

When the symbol frequency ω _PT selected by the frequency value selection unit 40 is the symbol frequency ω _PT2 after interpolation processing output from the symbol information detector 20, the zero cross BTR detection circuit 62 removes noise by the band limiting filter 2a. A process of detecting the Nyquist phase θ _N (n) in each symbol is performed by performing a process of detecting the zero cross point of the subsequent IQ signal S ′ _R (n).
The symbol deviation detection circuit 63 calculates a frequency deviation Δf _R from the Nyquist phase θ _N (n) in each symbol detected by the BTR detection circuit 61 or the BTR detection circuit 62, and uses the frequency deviation Δf _R as a symbol rough estimator. It was added to the symbol rate f _R calculated by the second symbol rate calculation unit 50, and carries out a process of outputting to the demodulator 4 and the addition result f _{R +} Delta] f _R as fine estimation result of the symbol rate.

Next, the operation will be described.
The rough symbol estimator 2 receives the IQ signal S _R (n) orthogonally detected from the received signal modulated at an unknown symbol rate, and uses the band limiting filter 2a to generate the IQ signal S _R (n). Remove superimposed noise (Gaussian noise).
The symbol rough estimator 2 converts the IQ signal after noise removal by the band limiting filter 2a into a frequency domain signal, and specifies a symbol frequency ω _PT (frequency value) at which peak power is obtained in the frequency domain signal. At the same time, the symbol speed f _R is roughly estimated from the symbol frequency ω _PT .
The rough symbol estimator 2 performs rough symbol accuracy estimation processing based on FFT point accuracy (for example, when 1024-point FFT processing is performed, a deviation accuracy of about 1000 ppm).
The processing contents of the symbol rough estimator 2 will be specifically described below.

When the IQ signal S _R (n) orthogonally detected from the received signal modulated at an unknown symbol rate is input, the symbol information detector 10 performs nonlinear processing and FFT processing on the IQ signal S _R (n). Thus, the frequency value ω _PT1 at which the peak power MAX (PAVE1 (ω)) is obtained among the signals in the frequency domain is specified.
In other words, the LPF unit 11 of the symbol information detector 10 uses, for example, a root Nyquist filter (for example, a cutoff frequency fc = 3 dB: an occupied bandwidth estimation value Fw × 1.3) to perform orthogonal detection on the IQ signal S. Filtering processing is performed on _R (n), and the filtered IQ signal S ′ _R (n) is output to the nonlinear unit 12.
S ′ _R (n) = S ′ _RI (n) + jS ′ _RQ (n) (1)
However, j is a complex number.

When the LPF unit 11 performs the filtering process on the IQ signal S _R (n), the nonlinear unit 12 performs a nonlinear process on the IQ signal S ′ _R (n) after the filtering process as shown in the following equation (2). The square process is performed, and the IQ signal f (n) after the nonlinear process is output to the FFT unit 13.
f (n) = S ′ _R (n) × S ′ _R (n) ^*
= _S'RI (n) * _S'RI (n) + _S'RQ (n) * _S'RQ (n)
(2)
However, * is a symbol indicating conjugation.

When receiving the IQ signal f (n) after nonlinear processing from the nonlinear unit 12, the FFT unit 13 performs a Hanning window on the IQ signal f (n) after nonlinear processing as shown in the following equation (3). After multiplying the function coefficient W (n), a complex FFT process of N points (for example, 1024 points) is performed to output a plurality of FFT results F (ω) to the FFT time domain averaging unit 14.

The FFT time domain averaging unit 14 performs a time addition process a plurality of times on a plurality (= m) of FFT results F (ω) by the FFT unit 13 to obtain a frequency value ω range of 0 to 1023. Thus, the power PAVE1 (ω) of the frequency domain signal is calculated.
That is, the FFT time domain averaging unit 14 slides the n value of the IQ signal f (n) at the time of deriving F (ω), for example, by 256, as shown in the following equation (4), (4m−1) ) Times of addition processing.

The peak power detection unit 15 detects the highest peak power MAX (PAVE1 (ω)) among the power PAVE1 (ω) of the frequency domain signal calculated by the FFT time domain averaging unit 14.
When the peak power MAX (PAVE1 (ω)) is detected, the peak power detection unit 15 outputs the frequency value ω _{PAVE1 from} which the peak power MAX (PAVE1 (ω)) is obtained to the Lagrange interpolation unit 16 as a symbol frequency.
However, when detecting the peak power, the frequency value range ω may be limited to a range where the peak power is expected to be actually detected.

When the Lagrangian interpolation unit 16 receives the symbol frequency ω _PAVE1 from the peak power detection unit 15, the Lagrangian interpolation unit 16 _performs an interpolation process on the symbol frequency ω _{PAVE1 and performs} the symbol frequency ω _PT1 and the peak power MAX (PAVE1 (ω1) after the interpolation process. )) Is output to the frequency value selector 40.
For example, a Lagrangian secondary interpolation process as shown in the following equation (5) is performed.

FIG. 4 is an explanatory diagram showing the symbol frequency ω _PT1 after interpolation processing and the peak power MAX (PAVE1 (ω)), which are detection results of the symbol information detector 10.
In the example of FIG. 4, the frequency with the FFT point number “256” is the symbol frequency ω _PT1 , and power of about −12 dB is obtained.

Symbol information detector 20 inputs the IQ signal is orthogonally detected from a received signal modulated by the unknown symbol rate S _{R (n),} the detection process and the FFT process of the zero-crossing point of the IQ signal S _{R (n)} Is performed to identify the frequency value ω _PT2 from which the peak power MAX (PAVE2 (ω)) is obtained in the frequency domain signal.
That is, the frequency deviation detecting unit 21 of the symbol information detector 20, the orthogonally detected IQ signal _S R (n), by carrying out, for example, multiplication method, the center frequency Δfc the IQ signal _S R (n) Estimate
Here, the frequency deviation detector 21 estimates the center frequency Δfc, but the center frequency Δfc may be given in advance as known information.

When the frequency deviation detector 21 estimates the center frequency Δfc, the frequency rotation unit 22 rotates the frequency of the IQ signal S _R (n) by the center frequency Δfc as shown in the following equation (6).
S _ROT (n) = S _R (n) × exp (−2π (Δfc / Fs) × n)
(6)
Here, Fs is an A / D (Analog to Digital Converter) sampling frequency obtained by sampling a digital IQ signal, and n is a time sample number (n = 0,..., 1023).

The LPF unit 23 uses, for example, a root Nyquist filter (for example, a cut-off frequency fc = 3 dB: an occupied bandwidth estimation value Fw × 1.3), and the IQ signal S _ROT ( The filtering process is performed on n), and the filtered IQ signal S ′ _ROT (n) is output to the zero-cross detection unit 24.
S ′ _ROT (n) = S ′ _ROTI (n) + jS ′ _ROTQ (n) (7)
However, j is a complex number.

Zero-cross detector 24, 'receives the _ROT (n), the IQ signal S' IQ signal after the filtering processing from the LPF section 23 S _ROT (n) and the preceding symbol of the IQ signal S _'ROT (n-1) The zero cross point detection process between the two is performed to generate a pulse wave IQ signal f (n).
f (n) = doutI (n) + j × doutQ (n) (8)
However, j is a complex number.

That is, the zero cross detection unit 24 generates the pulse wave IQ signal f (n) by performing the following calculation on the Ich signal and the Qch signal.
However, in the following calculation:
InZeroDet = 1 indicates that there is a zero cross, and InZeroDet = 0 indicates that there is no zero cross.

When the zero cross detection unit 24 generates the pulse wave IQ signal f (n), the FFT unit 25 generates a Hanning window function for the pulse wave IQ signal f (n) as shown in the following equation (9). After multiplying the coefficient W (n), a complex FFT process of N points (for example, 1024 points) is performed to output a plurality of FFT results F (ω) to the FFT time domain averaging unit 26.

The FFT time domain averaging unit 26 performs a plurality of time addition processes on a plurality (= m) of FFT results F (ω) by the FFT unit 25, so that the frequency value ω ranges from 0 to 1023. Thus, the power PAVE2 (ω) of the frequency domain signal is calculated.
That is, the FFT time domain averaging unit 26 slides the n value of the IQ signal f (n) at the time of deriving F (ω), for example, by 256, as shown in the following equation (10), (4m−1 ) Times of addition processing.

The peak power detection unit 27 detects the highest peak power MAX (PAVE2 (ω)) among the power PAVE2 (ω) of the frequency domain signal calculated by the FFT time domain averaging unit 26.
When the peak power MAX (PAVE2 (ω)) is detected, the peak power detection unit 27 outputs the frequency value ω _{PAVE2 from} which the peak power MAX (PAVE2 (ω)) is obtained to the Lagrange interpolation unit 28 as a symbol frequency.
However, when detecting the peak power, the frequency value range ω may be limited to a range where the peak power is expected to be actually detected.

When the Lagrangian interpolation unit 28 receives the symbol frequency ω _PAVE2 from the peak power detection unit 27, the Lagrangian interpolation unit 28 _performs an interpolation process on the symbol frequency ω _{PAVE2 and performs} the symbol frequency ω _PT2 and the peak power MAX (PAVE2 (ω )) Is output to the frequency value selector 40.
For example, a Lagrangian secondary interpolation process as shown in the following equation (11) is performed.

Symbol information detector 30 inputs the IQ signal is orthogonally detected from a received signal modulated by the unknown symbol rate S _{R (n),} the detection processing of the frequency variation of the IQ signal S _{R (n),} the nonlinear processing In addition, by performing the FFT process, the frequency value ω _PT3 at which the peak power MAX (PAVE3 (ω)) is obtained in the frequency domain signal is specified.
In other words, the LPF unit 31 of the symbol information detector 30 uses, for example, a root Nyquist filter (for example, a cutoff frequency fc = 3 dB: an occupied bandwidth estimation value Fw × 1.3) to perform orthogonal detection on the IQ signal S. the filtering process carried out with respect to _R (n), and outputs the IQ signal after the filtering processing S _'R (n) to the frequency discriminator 32.
S ′ _R (n) = S ′ _RI (n) + jS ′ _RQ (n) (12)
However, j is a complex number.

Frequency discriminator 32 'receives the _R (n), as shown in the following equation (13), the IQ signal S' IQ signal after the filtering processing from the LPF unit 31 S phase of _R (n) By differentiating the information θ (n), the instantaneous frequency fluctuation of the IQ signal S ′ _R (n) is detected.

The non-linear unit 33 performs non-linear processing (square processing) on the detection result f (n) of the frequency discriminator unit 32, and outputs the non-linear processed signal f (n) to the FFT unit 34.
When receiving the signal f (n) after nonlinear processing from the nonlinear unit 33, the FFT unit 34 applies a Hanning window function coefficient to the signal f (n) after nonlinear processing as shown in the following equation (14). After multiplying by W (n), a complex FFT process of N points (for example, 1024 points) is performed to output a plurality of FFT results F (ω) to the FFT time domain averaging unit 35.

The FFT time domain averaging unit 35 performs a time addition process a plurality of times on a plurality (= m) of FFT results F (ω) by the FFT unit 34 to obtain a frequency value ω range of 0 to 1023. Thus, the power PAVE3 (ω) of the frequency domain signal is calculated.
That is, the FFT time domain averaging unit 35 slides the n value of the signal f (n) at the time of deriving F (ω), for example, by 256, as shown in the following equation (15), (4m−1) Addition process is performed once.

The peak power detecting unit 36 detects the highest peak power MAX (PAVE3 (ω)) among the power PAVE3 (ω) of the frequency domain signal calculated by the FFT time domain averaging unit 35.
When the peak power MAX (PAVE3 (ω)) is detected, the peak power detector 36 outputs the frequency value ω _{PAVE3 from} which the peak power MAX (PAVE3 (ω)) is obtained to the Lagrange interpolation unit 37 as a symbol frequency.
However, when detecting the peak power, the frequency value range ω may be limited to a range where the peak power is expected to be actually detected.

When the Lagrangian interpolation unit 37 receives the symbol frequency ω _PAVE3 from the peak power detection unit 36, the Lagrange interpolation unit 37 _performs an interpolation process on the symbol frequency ω _{PAVE3, and performs} the symbol frequency ω _PT3 and the peak power MAX (PAVE3 (ω )) Is output to the frequency value selector 40.
For example, a Lagrange secondary interpolation process as shown in the following equation (16) is performed.

The frequency value selection unit 40 receives the symbol frequencies ω _PT1 , ω _PT2 , ω _PT3 after the interpolation processing from the symbol information detectors 10, 20, 30, and the peak power MAX (PAVE 1 (ω)), MAX (PAVE 2 (ω)). , MAX (PAVE3 (ω)), the peak powers MAX (PAVE1 (ω)), MAX (PAVE2 (ω)), MAX (PAVE3 (ω)) are compared, and the symbol frequency ω _PT1 after interpolation processing is compared. , Ω _PT2 , ω _PT3, the symbol frequency ω _PT that provides the highest peak power is selected.
For example, if the MAX (PAVE2 (ω))> MAX (PAVE3 (ω))> MAX (PAVE1 (ω)), and select the symbol frequency omega _PT2 after interpolation processing, MAX (PAVE3 (ω)) > MAX (PAVE1 (ω))> if the MAX (PAVE2 (ω)), to select the symbol frequency ω _PT3 after the interpolation processing.

When the frequency value selection unit 40 selects the symbol frequency ω _PT , the symbol rate calculation unit 50 calculates the symbol rate f _R from the symbol frequency ω _PT as shown in the following equation (17), and the symbol rate f _R is output to the symbol precision estimator 3 as a rough estimation result. Further, the symbol frequency ω _PT selected by the frequency value selection unit 40 is output to the symbol precision estimator 3.
f _R = ω _PT / N × Fs (17)
Here, N is the number of FFT points, and Fs is the A / D sampling frequency.

When the symbol precision estimator 3 receives the symbol rate f _R and the symbol frequency ω _PT from the symbol coarse estimator 2, the Nyquist phase in each symbol is derived from the IQ signal S ′ _R (n) after noise removal by the band limiting filter 2 a. detects the θ _{n (n),} the Nyquist phase theta _n to calculate the frequency deviation Delta] f _R from _(n), its frequency deviation Delta] f _R a is rough estimated by the symbol crude estimator 2 symbol rate f _R in each symbol Add to.

That is, the BTR detection circuit 61 based on nonlinear processing of the symbol estimator 3 uses the symbol frequency ω _PT output from the symbol rough estimator 2 as the symbol frequency ω _PT1 after interpolation processing output from the symbol information detector 10. In some cases, or when the symbol frequency ω _PT3 after interpolation processing output from the symbol information detector 30 is applied, nonlinear processing is performed on the IQ signal S ′ _R (n) after noise removal by the band limiting filter 2a. Thus, the Nyquist phase θ _N (n) in each symbol is detected.
Since the BTR detection circuit 61 based on nonlinear processing detects the Nyquist phase θ _N (n) itself is a known technique, detailed description thereof is omitted.

When the symbol frequency ω _PT output from the symbol rough estimator 2 is the symbol frequency ω _PT2 after the interpolation processing output from the symbol information detector 20, the zero cross BTR detection circuit 62 performs noise caused by the band limiting filter 2a. The Nyquist phase θ _N (n) in each symbol is detected by performing the zero cross point detection process of the IQ signal S ′ _R (n) after the removal.
Since the process of detecting the Nyquist phase θ _N (n) by the zero cross BTR detection circuit 62 is a known technique, detailed description thereof is omitted.
Here, FIG. 5 is an explanatory diagram showing an example of the symbol speed deviation which is the Nyquist phase θ _N (n) and the frequency deviation Δf _R in each symbol detected by the BTR detection circuits 61 and 62.

When the BTR detection circuit 61 or the BTR detection circuit 62 detects the Nyquist phase θ _N (n) in each symbol, the symbol deviation detection circuit 63 receives the Nyquist phase θ _N (n ) To calculate the frequency deviation Δf _R.
Δf _R = ( (θ _N (n _B ) −θ _N (n _A )) / 360 ) × (f _R / (n _B −n _A ))
(18)
In the example of FIG. 5, since _{n A-th} symbol of the Nyquist phase theta _{N (n A)} is 140, _{n B} th symbol of the Nyquist phase θ _{N (n B)} is 230, the frequency deviation Delta] f _R is as follows become.
Δf _R = ( (θ _N (n _B ) −θ _N (n _A )) / 360 ) × (f _R / (n _B −n _A ))
= ((230-140) / 360) × (f R / (n B -n A)) (Hz)

After calculating the frequency deviation Δf _R from the Nyquist phase θ _N (n), the symbol deviation detection circuit 63 adds the frequency deviation Δf _R to the symbol speed f _R calculated by the symbol speed calculation unit 50 of the symbol coarse estimator 2. Then, the addition result f _R + Δf _R is output to the demodulator 4 as a precise estimation result of the symbol rate.
Thereby, the coarse estimation result of the symbol coarse estimator 2 has an accuracy of FFT 1024 pt (about 1000 ppm), but the fine estimation result of the symbol fine estimator 3 improves to a deviation of about 50 ppm or less.

The demodulator 4 demodulates the received signal using the symbol rate f _R + Δf _R that is the result of the fine estimation of the symbol precision estimator 3 of the symbol estimation circuit 1.
Since the demodulation process itself of the demodulator 4 is a known technique, a detailed description thereof will be omitted.

As apparent from the above, according to the first embodiment, a band limiting filter that removes noise superimposed on the IQ signal S _R (n) orthogonally detected from the received signal modulated at an unknown symbol rate. 2a is implemented, the IQ signal S ′ _R (n) after noise removal by the band limiting filter 2a is converted into a frequency domain signal, and a symbol frequency ω _PT from which peak power is obtained in the frequency domain signal is specified. A symbol coarse estimator 2 that roughly estimates the symbol rate f _R from the symbol frequency ω _PT and a Nyquist phase θ _N (in each symbol from the IQ signal S ′ _R (n) after noise removal by the band limiting filter 2a. detecting the n), calculates a frequency deviation Delta] f _R from the Nyquist phase θ _{n (n)} in each symbol, crude its frequency deviation Delta] f _R by the symbol crude estimator 2 Since it is configured so as to provide a symbol fine estimator 3 for adding a constant symbol rate f _R, when demodulating the received signal modulated by the unknown symbol rate, it is possible to estimate the symbol rate with high precision There is an effect.

  In the present invention, any constituent element of the embodiment can be modified or any constituent element of the embodiment can be omitted within the scope of the invention.

  1 symbol estimation circuit, 2 symbol coarse estimator, 2a band limiting filter, 3 symbol fine estimator, 4 demodulator, 10 symbol information detector (first symbol information detector), 11 LPF unit, 12 nonlinear unit, 13 FFT unit, 14 FFT time domain averaging unit, 15 peak power detection unit, 16 Lagrange interpolation unit, 20 symbol information detector (second symbol information detector), 21 frequency deviation detection unit, 22 frequency rotation unit, 23 LPF Unit, 24 zero cross detection unit, 25 FFT unit, 26 FFT time domain averaging unit, 27 peak power detection unit, 28 Lagrange interpolation unit, 30 symbol information detector (third symbol information detector), 31 LPF unit, 32 Frequency discriminator section, 33 nonlinear section, 34 FFT section, 35 FFT time domain averaging section, 36 Over click power detection unit, 37 Lagrange interpolation unit, 40 a frequency value selection unit, 50 symbol rate calculator, BTR detector circuit according to 61 non-linear processing, BTR detector by 62 zero cross, 63 symbols deviation detecting circuit.

Claims (5)

  An IQ signal orthogonally detected from a received signal modulated at an unknown symbol rate is converted into a frequency domain signal, and a frequency value at which peak power is obtained in the frequency domain signal is specified. A symbol coarse estimator that roughly estimates the velocity, and detects a Nyquist phase in each symbol from the IQ signal, calculates a frequency deviation from the Nyquist phase in each symbol, and roughly calculates the frequency deviation by the symbol coarse estimator. A symbol estimation circuit comprising: a symbol precision estimator for adding to an estimated symbol rate. The symbol rough estimator has a band limiting filter for removing noise superimposed on the IQ signal, and converts the IQ signal after noise removal by the band limiting filter into a signal in a frequency domain. Item 2. The symbol estimation circuit according to Item 1. The symbol crude estimator may implement a non-linear process on IQ signal after noise removal by the band-limiting filter, and converts the IQ signal after the nonlinear processing on the signal in the frequency domain, the peak power in the signal in the frequency domain The first symbol information detector that identifies the frequency value from which is obtained, and the zero cross point detection processing of the IQ signal after noise removal by the band limiting filter, is performed, and the detection result is converted into a frequency domain signal, A second symbol information detector that identifies a frequency value at which peak power is obtained in the frequency domain signal, and a frequency fluctuation detection process of the IQ signal after noise removal by the band limiting filter, and a detection result thereof The third thin film is converted into a frequency domain signal by performing non-linear processing on the frequency domain, and a frequency value at which peak power is obtained in the frequency domain signal is specified. A frequency information selection unit, a frequency value selection unit that selects a frequency value from which the highest peak power is obtained from the frequency values specified by the first, second, and third symbol information detectors, and the frequency 3. The symbol estimation circuit according to claim 2, further comprising: a symbol rate calculation unit that calculates a symbol rate from the frequency value selected by the value selection unit. The symbol fine estimator, a BTR (Bit Timing Recovery) detection circuit for detecting the Nyquist phase in each symbol from the IQ signal after noise removal by the band limiting filter, is detected by the BTR (Bit Timing Recovery) detection circuit A symbol deviation detection circuit that calculates a frequency deviation from the Nyquist phase in each symbol, adds the frequency deviation to the symbol speed calculated by the symbol speed calculation unit, and outputs the addition result as a precise estimation result of the symbol speed The symbol estimation circuit according to claim 3, comprising:   An IQ signal orthogonally detected from a received signal modulated at an unknown symbol rate is converted into a frequency domain signal, and a frequency value at which peak power is obtained in the frequency domain signal is specified. A symbol coarse estimator that roughly estimates the velocity, and detects a Nyquist phase in each symbol from the IQ signal, calculates a frequency deviation from the Nyquist phase in each symbol, and roughly calculates the frequency deviation by the symbol coarse estimator. A demodulation circuit comprising: a symbol precision estimator that adds to an estimated symbol rate; and a demodulator that demodulates the received signal using a symbol rate obtained by adding a frequency deviation by the symbol precision estimator.

Images