JP2006115261A - Waveform equalizer for radio communication apparatus and receiving method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To correct the amplitude and phase of the factor of a filter by using a pilot symbol, to eliminate a discontinuity between slots, and to enable the training of the filter factor extensively over the slots by using a convergence algorithm at a low speed such as an LMS. <P>SOLUTION: A waveform equalizer for a radio communication apparatus has a waveform equalizing means having the filter, an error arithmetic processor and a filter-factor training means. The waveform equalizer further has steps of: computing the filter factor by utilizing a preamble in the waveform equalizing means; computing the correction factor of the filter factor on the basis of an amplitude ratio to the symbol in the case of the transmission of a pilot symbol and a phase difference by arithmetically operating an output from the equalizer at a time when the reception signal of the pilot symbol is input; and temporarily storing the correction factor. The waveform equalizer conducts an initial training in a channel for a control when the radio communication apparatus conducts a reception, and conducts the training while using the stored correction factor as an initial value after a transfer to the channel for a communication. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a waveform equalizer and a reception method for compensating for propagation path distortion of a radio communication device, and more particularly to a waveform equalizer and a reception method having a function of compensating for a discontinuity between slots of a received signal.

In digital wireless communication, there is a problem that a received signal is distorted due to frequency selective fading caused by a large number of reflected waves with delay, and it is necessary to compensate for this distortion by a waveform equalizer.
Among the waveform equalizer methods, the transversal equalizer is easy to realize because of its relatively simple circuit configuration, and is widely used. This waveform equalizer generally uses an LMS (Least Mean Square) algorithm or an RLS (Recursive Least Square) algorithm for training of its own filter coefficients.
The LMS algorithm is relatively easy to process, but has a slow convergence rate and requires many known symbols (approximately several tens of symbols) for training the filter coefficients. The RLS algorithm, on the other hand, has a fast convergence speed (roughly converges at about 10 symbols), but requires the matrix operation to be performed by floating-point arithmetic, which increases the circuit scale for the operation compared to the LMS algorithm. There is. If the fluctuation of the propagation path characteristic is sufficiently slow compared with the convergence speed of the LMS algorithm and the length of the known symbol included in the received signal is sufficiently longer than the convergence time of the LMS algorithm, the LMS algorithm can be used.

  In fixed wireless communication in which a transmitter and a receiver are installed in a fixed place, the propagation path characteristic fluctuation is slow, so that the LMS algorithm can be used if the length of a known symbol used for filter coefficient training is sufficient. For example, the system of Non-Patent Document 1 described below is used for fixed communication, and since there are 64 known symbols in a transmission slot, an LMS algorithm can be used.

In the following, a receiver incorporating a transversal equalizer using the LMS algorithm will be described taking the municipal digital broadcast communication system (ARIB STD-T86), which is a digital fixed wireless communication system, as an example.
FIG. 9 is a diagram showing a frame configuration of this system. In this system, the signal is divided into frames (1 frame = 80 ms), and then one frame is divided into 6 slots (1 slot = 80 ms / 6 = 13.33 ms). To slot 0, slot 1,..., Slot 5. FIG. 3 is a diagram showing the data format of one slot of this system, and there are three types of signals: a control physical channel, a communication physical channel, and a synchronization burst.
In this system, slots 0 to 2 are basically downlinks (transmitted by the master station and received by the slave stations), slots 3 to 5 are uplinked (transmitted by the slave stations and received by the master station), and slots 0 and 3 are A control physical channel (slot 0 is downlink and slot 3 is uplink) is assigned for communication control.

FIG. 10 is a diagram showing an example of slot assignment in this system. FIG. 10 (1) shows the slot assignment in the case of a contact call, and FIG. 10 (2) shows the slot assignment in the case of loud sound broadcasting and deemed voice FAX.
In the case of a communication call, the downlink control physical channel in slot 0 is basically transmitted from the master station at all times during a call, but the uplink control physical channel in slot 3 is transmitted only during call connection. The physical channel for communication is allocated in order from the slot 1 for the downlink and in order from the slot 4 for the uplink.
When making a loudspeak report from the master station to the slave station or sending an assumed voice FAX, slots 1, 2, 4, and 5 are assigned as communication physical channels (downlinks), and the downlink control physical in slot 0 The channel is basically transmitted from the master station at all times during communication.
The control physical channel and the synchronization burst have a preamble (the AGC preamble AP of the control physical channel and the fixed pattern FP of the synchronization burst are collectively referred to simply as a preamble), and a long pattern of '20A800080A' There are 64 symbols, and the preamble is used for initial training of the filter coefficients of the waveform equalizer. On the other hand, the known pattern arranged in the communication physical channel is only the synchronization word SW (also arranged in the communication physical channel and synchronization burst) and has a length of 10 symbols. Convergence by the LMS algorithm requires at least 20 to 30 known symbols, and therefore initial training cannot be performed on the communication physical channel. Therefore, training must be performed each time.

FIG. 5 is a block diagram showing an example of a receiver incorporating a waveform equalizer.
An antenna (not shown) is connected to the terminal A, and a received signal is input to the high frequency circuit 101. The high-frequency circuit 101 converts the frequency of the radio band to an intermediate frequency band that can be sampled by an A / D (analog / digital) converter 102 connected at a subsequent stage, amplifies the power, and converts the frequency into an A / D converter. Input to 102. In the following description, a signal output from the high frequency circuit 101 is referred to as an IF signal.
The A / D converter 102 samples and quantizes the IF signal into a digital signal, and inputs the digital signal to the orthogonal demodulation unit 103. The quadrature demodulation unit 103 performs frequency conversion of the digitally converted IF signal into a baseband band, and inputs the in-phase component and the quadrature component to the reception filter unit 104. In the figure, two signals of an in-phase component and a quadrature component are input from the quadrature demodulation unit 103 to the reception filter unit 104, but are represented by a single line for the purpose of simplifying the drawing. In the following description, a set of in-phase component and quadrature component signals is represented by a single line on the drawing, represented by a complex number in the equation, an in-phase component is represented by a real part, and a quadrature component is represented by an imaginary part.
The reception filter unit 104 performs removal of unnecessary frequency components and waveform shaping, and inputs the result to an AFC (Automatic Frequency Control) unit 105. The AFC unit 105 corrects the phase rotation of the baseband signal generated by the frequency error of the local oscillator included in the high frequency unit circuit 101, and inputs the correction to the frame synchronization unit. The frame synchronization unit searches the frame head timing and symbol timing of the received signal, and stores the baseband signal for one slot in the buffer 107 in synchronization with these timings. The buffer 107 is connected to the amplitude / phase correction unit 501, and the baseband signal accumulated in the buffer 107 is read from the amplitude / phase correction unit 501.
The amplitude / phase correction unit 501 has a buffer for storing the corrected baseband signal therein, and uses a pilot symbol P of the received baseband signal (two symbols are arranged apart from each other by 128 symbols in the slot). The amplitude and phase are corrected, and the corrected signal is stored in the internal buffer. Since the signal to be received is a burst signal, discontinuity occurs in the amplitude and phase of the received signal if there is a non-transmission slot in between. Since the filter equalizer training of the waveform equalizer starts with the coefficient trained in the slot received immediately before as an initial value, the filter coefficient diverges when there is discontinuity. In order to prevent this, amplitude and phase are corrected by using pilot symbols for each slot to remove discontinuities.
The equalizer 108 reads the baseband signal accumulated in the internal buffer of the amplitude / phase correction unit 501, repeats waveform equalization and demodulation for the number of symbols of the accumulated signal, and outputs the demodulated bit data to the terminal Output via B.

FIG. 2 is a block diagram showing an example of a transversal equalizer. When the number of taps is N _tap , a shift register 201, N _tap complex multipliers 201-1, 201-2,. Ntap, complex adder 203, symbol determination unit 204, switch 205, complex subtractor 206, LMS algorithm calculation unit 207, filter coefficient storage buffer 208, symbol number is n, and n = 0, 1,. It is processed repeatedly.
There are filter coefficient intervals of one symbol and narrower ones (called fractional interval equalizers), but more effectively compensate for delay distortion and suppress deterioration due to symbol timing errors. For this purpose, in FIG. 2, the coefficient interval is set to 1/2 symbol. Therefore, the signal accumulated in the buffer 107 is oversampled twice as much as the signal of the symbol period.
The shift register 201 stores N _tap complex numbers in the storage areas u ₀ , u ₁ ,..., U _Ntap-1 , and inputs signals from the terminals A1 and A2 to u ₀ , u _{1 ,.} The contents of ₃ are shifted by 2 samples to u ₂ , u ₃ ,..., U _Ntap-1 , and the input signal from terminal A1 is stored in u ₀ and the input signal from A2 is stored in the storage area of u ₁ , respectively. If the number of the symbol to be processed by the equalizer is n (n = 0, 1, ..., 149), the signal x _2n of the 2n sample is input to the center tap. N _tap -1) / 2 sample signal x _{2n + (Ntap-1) / 2} is input, and terminal A2 receives 2n + (N _tap -1) / 2-1 sample signal x _{2n + (Ntap-1) / 2-1} is input respectively.
The complex multipliers 201-1, 201-2,..., 202-Ntap include complex conjugate operations from the contents u ₀ , u ₁ ,..., U _Ntap-1 of the shift register 201 and the filter coefficient buffer 208. through the vessel 209, h of the filter coefficients _{0, h 1, ..., h} Ntap-1 complex conjugate ^{_{h * 0, h * 1,}} ..., h * Ntap-1 is inputted, the complex multiplication result u ₀ h ^* ₀ , u ₁ h ^* ₁ ,..., U _Ntap-1 h ^* N _tap-1 are input to the complex adder 203. Complex adder 203 calculates the sum of the input values from complex multipliers 201-1, 201-2,..., 202-Ntap, and outputs the calculation result to symbol determination unit 206 and the “−” side terminal of the subtractor. input. The calculation result is expressed by Expression (1), where the tap input vector is U (n) and the coefficient vector is H (n).
Here, “*” in h ^* ₀ or H ^* (n) is a symbol representing a complex conjugate, and means that the sign of the imaginary part is inverted. Also, "T" in U ^T (n) is a transpose symbol, meaning that the rows and columns of the matrix are interchanged. Since U (n) is a 1 × N _tap column vector, U ^T (n) is N _tap × 1 row vector.
The symbol determination unit 204 performs symbol determination of the filter output value y _n input from the complex adder 203, inputs the determination symbol D (y _n ) to the terminal b of the switch 205, and inputs the determination bit b _n to the terminal D Output to. In this system, 16QAM symbol determination is performed.
The switch 205 selects the determination symbol D (y _n ) and the reference symbol r _n (symbol at the time of preamble transmission) input from the terminal C, and uses the “+” side terminal of the subtractor 206 as a desired response d _n. To enter. When having received a control physical channel or synchronization bursts, by connecting the terminals ac at 4 ≦ n ≦ 68, and select the reference symbol r _n as desired responses d _n, connects the terminals ab otherwise , D _n is selected as a determination symbol D (yn). When having received a communication physical channel always connects the terminals ab, selects a determination symbol D (yn) as d _n. That is, r _n is selected in the preamble, and D (y _n ) is selected in the other cases.
Complex subtractor 206 calculates an error e _n of the filter output value y _n for the response d _n desired, and inputs the calculation result e _n to LMS algorithm operating unit 207.
The LMS algorithm calculation unit 207 receives the tap input vector U (n) from the shift register 201, the error e _n from the complex subtractor 206, the constant μ from the terminal B, and the coefficient vector H (n) from the filter coefficient storage buffer 208, The coefficient vector H (n + 1) at the next symbol is calculated according to equation (4), and the contents of the filter coefficient storage buffer 208 are updated to H (n + 1).
Here, the constant μ is called a step size parameter, and 0 <μ <1.
Equation (4), in the preamble, the (transmission time of a symbol of the preamble) r _n reference symbols in response d _n desired, so as to approximate the output y _n of the equalizer in the reference symbol r _n (error e _n The coefficient is controlled (to be close to 0), and, except for the preamble, the symbol D (y _n ) determined on the receiving side is the desired response d _n and the output y _n of the equalizer is set to the determination symbol D (y _n ) controls the (error e _n as close to 0) coefficient closer.

FIG. 8 is a flowchart showing a processing procedure of the amplitude / phase correction unit 501 and the equalizer 108 after the baseband signal to be processed is accumulated in the buffer 107. When the baseband signal to be processed is accumulated in the buffer 107, the amplitude / phase correction processing by the amplitude / phase correction unit 501 described above is performed (step 801). When the processing in step 801 is completed, it is determined in step 802 whether a control physical channel or synchronization burst has been received. If a control physical channel or synchronization burst has been received, the filter coefficient H is reset to H = 0. (Step 803). Otherwise, or after the end of step 803, the equalization process by the equalizer 108 described above is performed for each symbol for n = 0, 1,..., 149 (steps 804, 805, and 806).
When a control physical channel or synchronization burst is received, the filter coefficient H of the equalizer 108 is reset to 0, initial training is performed with the preamble as a known symbol, and training is continued with a signal following the preamble. When receiving a communication physical channel, training is performed with the filter coefficient trained in the slot received immediately before as an initial value.

For example, Non-Patent Document 1 is a technique related to a conventional wireless communication system.
Municipal digital broadcast communication system ARIB STD-T86 The Japan Radio Industry Association

In the conventional receiver described above, the amplitude / phase correction unit 501 corrects the amplitude / phase of the received signal by using the pilot symbol in order to compensate for the amplitude / phase discontinuity between the slots of the received signal. . When the received signal is distorted, the distortion causes an error in the correction of amplitude and phase, and the correction varies from slot to slot. Therefore, there is a problem that the discontinuity remains between the slots and the convergence characteristic is deteriorated.
In addition, when it is necessary to use a slow convergence algorithm such as LMS, it is necessary to improve the deterioration of the convergence characteristic.
Further, when a non-transmission slot is received by a burst signal having a non-transmission slot between slots and the filter coefficient diverges, it is necessary to restore the filter coefficient to a state before divergence.

  The object of the present invention has been made in view of the above, and the amplitude and phase of the coefficient of the filter unit are corrected using pilot symbols, discontinuities between slots are removed, and the like in the LMS algorithm. It is an object of the present invention to provide a waveform equalizer and a receiving method that enable training of filter coefficients across slots using a slow convergence algorithm.

In order to solve the above problems, the filter coefficient correction means of the present invention is an equalizer when the received signal of the pilot symbol part is input to a transversal equalizer whose filter part is a FIR (Finite Impulse Response) filter. calculating the output y _P of the symbols during transmission of the pilot symbols p, the amplitude ratio of the y _p for the _{p a = | y p | /} | p |, the phase difference between the p and the y _p theta = θ _yp −θ _p (unit is radians, θ _yp is the phase of y _p , θ _p is the phase of p), and filter coefficient H = [h ₀ , h ₁ , ..., h _N-1 ] (N is a tap The coefficient obtained by dividing a by the amplitude of (number) and rotating the phase by θ H ′ = He ^j θ / a = [h ₀ e ^j θ / a, h ₁ e ^j θ / a,..., H _N−1 e ^j θ / a] (e is the base of natural logarithm, j is an imaginary unit, and e ^j θ is a rotation factor of angle θ (radian) with e ^j θ = cos θ + jsin θ).
Alternatively, said calculation value g = e ^j θ / a multiplying coefficient H, wherein y value _p divided by the _{p k = y p / p =} y p p * / | p | computes ^2, the k is divided by the magnitude | k | of k, and the division result g ′ = k / | k | ² is set as the value g.

  Further, the waveform equalizer of the present invention is a transversal equalizer in which a filter unit is a FIR (Finite Impulse Response) filter, and the H ′ is calculated by the filter coefficient correction unit, and the H ′ is used as a filter coefficient. Train as an initial value.

In order to solve the above problem, the waveform equalizer of the present invention uses the center tap coefficient of the filter coefficient H = [h ₀ , h ₁ ,..., H _N−1 ] (N is the number of taps). h _Ncenter = 1 (N _center is the center tap number), otherwise h ₀ = h ₁ =… = h _Ncenter-1 = 0, h _{Ncenter + 1} = h _{Ncenter + 2} =… h _N-1 = 0 the _{H 0 = [0, ...,} 0, 1, 0, ..., 0] as an initial value of H, the filter by coefficient correcting means 'calculates the coefficient H of the calculated' year H a filter coefficient initial Train as a value.

Further, in order to solve the above problem, the waveform equalizer of the present invention has a memory for saving the filter coefficient, and the average value in the slot of the determination error power | e _n | ² (n is the symbol number in the slot) < | e _n | ² > and compare <| e _n | ² > with a preset constant e _th (e _th > 0), and <| e _n | ² ><e _th (or <| e _n | ² > ≦ e _th ), the filter coefficient H is saved in the memory, and if <| e _n | ² > ≧ e _th (or <| e _n | ² >> e _th ), the memory The coefficient saved in is restored as H and used as the initial value of the filter coefficient at the next slot processing.

In order to solve the above-described problem, the wireless communication device reception method of the present invention is a wireless communication device reception method including a waveform equalization unit having a filter unit, an error calculation unit, and a filter coefficient training unit. A step of calculating a filter coefficient using a preamble as a waveform equalization means, and an output of an equalizer when a received signal of a pilot symbol part is inputted to calculate an amplitude ratio of the pilot symbol part to a symbol at the time of transmission And a step of calculating a correction coefficient of the filter coefficient based on the phase difference and a step of temporarily storing the correction coefficient, and when the wireless communication apparatus performs reception, the control channel performs initial training. After the communication channel shift, training is performed using the stored correction coefficient as an initial value.

According to the present invention, since the amplitude and phase of the coefficient of the filter unit are corrected using pilot symbols, discontinuity between slots is removed, and a slow convergence algorithm such as an LMS algorithm is used to span between slots. Training of filter coefficients is possible.
Since the correction of the amplitude and phase of the filter coefficient is performed by comparing the equalizer output signal and the pilot symbol at the time of transmission, it can be performed without being affected by the distortion of the propagation path, and the convergence characteristics are not deteriorated.
Further, a correction coefficient H ′ is calculated using H ₀ = [0,..., 0, 1, 0,..., 0] with only the center tap as 1, and training is performed with the H ′ as an initial value. As a result, the time required for training (convergence time) is shortened as compared with the case of training with the initial value of the filter coefficient set to zero.
Furthermore, the average value of judgment errors in the slot <| e _n | ² > is compared with a threshold value. If the average power of judgment errors is greater than the threshold value, the filter coefficients saved in the memory are restored. In the next slot, training is performed with the restored filter coefficient as an initial value. Therefore, even when a non-transmission slot is received by a burst signal having a non-transmission slot between slots and the filter coefficient diverges, A burst signal can be received to train with the trained filter coefficients as initial values.

(Example 1)
FIG. 1 is a block diagram showing an embodiment of a receiver incorporating an equalizer according to the present invention. In FIG. 1, the terminal A, the high frequency circuit 101, the A / D converter 102, the quadrature demodulation unit 103, the reception filter unit 104, the AFC unit 105, the frame synchronization unit 106, and the buffer 107 are the same as those in FIG. Omitted.
An equalizer 108 and a coefficient correction unit 109 are connected to the buffer 107, and the equalizer 108 is a transversal equalizer shown in FIG. 2, which reads out the baseband signal accumulated in the buffer 107, performs waveform equalization and Demodulation is repeated for the number of symbols of the accumulated signal, and demodulated bit data is output via terminal B. Coefficient correction section 109 reads a baseband signal around the pilot section from buffer 107, and corrects the amplitude and phase of the filter coefficient of equalizer 108 using the read signal. The processing by the coefficient correction unit 109 is performed before the processing of the equalizer 108.

FIG. 4 is a block diagram showing the coefficient correction unit 109 in detail. The coefficient correction unit 109 includes N _tap +2 complex multipliers 401-0, 401-1, ..., 401-Ntap-1, 403, 404, a complex adder 402, a complex vector multiplier 405, and an amplitude square operation 406, reciprocal calculator 407, and complex conjugate calculator 408.
Terminals A0, A1,..., ANtap-1 receive from the buffer 107 the received baseband signal centered on the pilot symbol portion x _{2Np + (Ntap-1) / 2} , x _{2Np + (Ntap-1) / 2-1} , ..., x _{2Np- (Ntap-1) / 2} is read and input to the complex multipliers 401-0, 401-1, ..., 401-Ntap-1. Here, N _p is the symbol number of the pilot symbol, and N _p = 10 from the data format of FIG.
The filter coefficient H = [h ₀ h ₁ ... h _Ntap-1 ] ^{T of} the equalizer trained up to the immediately preceding slot is input to the terminal B and input to the complex vector multiplier 405 and the complex conjugate calculator 408 Is done. The complex conjugate calculator 408 calculates the complex conjugate H ^* = [h ^* ₀ h ^* ₁ … h ^* _Ntap-1 ] ^T of the filter coefficient H, and each element h ^* ₀ , h ^* ₁ , ..., h ^* _{Ntap -1} is input to the other input terminal of the complex multiplier 401-0, 401-1, ..., 401-Ntap-1.
Complex multipliers 401-0, 401-1,..., 401-Ntap-1 and complex adder 402 calculate the equalizer output y _Np at the time of receiving the pilot symbol portion received signal, using Equation (5).
The complex multiplier 403 sets the symbol at the time of transmission of the pilot symbol to p, and p '= p / | p | ² 's complex conjugate (p') * obtained by dividing p by the square of the magnitude of p to y _Np Multiplication is performed, and the multiplication result k is calculated by Expression (6), and the calculation result k is input to the complex multiplier 404 and the amplitude square calculator 406.
Amplitude square calculator 406 and reciprocal calculator 407 calculate the reciprocal 1 / | k | ² of the magnitude of k, and complex multiplier 404 multiplies k by 1 / | k | ² . The calculation result g is input to the complex vector multiplier 405.
The complex vector multiplier 405 receives the elements h ₀ , h ₁ ,..., H _{Ntap− of the} filter coefficients H = [h ₀ h ₁ … h _Ntap−1 ] ^T input from the equalizer 108 via the terminal B. ₁ is multiplied by g inputted from the complex multiplier 404, and the multiplication result H ′ = [h ′ ₀ h ′ ₁ ... H ′ _Ntap−1 ] ^T is _sent to the equalizer 108 via the terminal C. input.

FIG. 6 is a flowchart showing a processing procedure in the present embodiment. In step 601, it is determined whether a control physical channel or synchronization burst is received. If a control physical channel or synchronization burst is received, the filter coefficient is reset to H = 0 (step 602). Otherwise, the coefficient correction unit 109 calculates the correction coefficient H ′ and sets H ′ as the initial value of H (step 603). When the processing of step 602 or step 603 is completed, equalization processing by the equalizer 108 is performed for each symbol for n = 0, 1,..., 149 (steps 604, 605, and 606).
When a control physical channel or synchronization burst is received, the filter coefficient H of the equalizer 108 is reset to 0, initial training is performed with the preamble as a known symbol, and training is continued with a signal following the preamble. When receiving a physical channel for communication, filter coefficient correction processing is performed before equalization processing, and training is performed using the corrected filter coefficients as initial values.

In the above description, if the filter coefficient trained in the slot received immediately before is H = [h ₀ h ₁ ... h _Ntap-1 ] ^T and there is no change in amplitude and phase between the slots, the filter output of the nth symbol y _n substantially matches the symbol x _n at the time of transmission, but if the amplitude of the received signal is a times and the phase changes θ between slots,
Thus, the amplitude and phase of the equalizer output change the same as the received signal. In this state, when the received signal of the pilot symbol part is input to the equalizer,
And substituting this into equation (6),
And substituting equation (11) into equation (7),
It becomes. The new coefficient is from equation (8)
Thus, a coefficient for rotating the amplitude by 1 / a and rotating the phase by θ is calculated.
Since the calculation of the equalizer output is convolution of the tap input vector and the complex conjugate of the coefficient vector from equation (1), the phase of the filter output is corrected to the phase opposite to the coefficient phase. Therefore, in the updated H ′, the amplitude is corrected to 1 / a and the phase is rotated by −θ.
In the coefficient before update, since the equalizer output is a times larger in amplitude and the phase is rotated by θ from equation (9), the amplitude and phase are changed by replacing the updated coefficient H ′ with the coefficient of the equalizer. It is corrected.
In this embodiment, since the amplitude and phase of the filter coefficient are matched to the equalizer output y _Np at the time of pilot symbol input, the coefficient correction processing is hardly affected by propagation path distortion, and discontinuity is unlikely to occur between slots. Therefore, deterioration of convergence characteristics can be prevented.

(Example 2)
In the first embodiment, the baseband signal input to the equalizer 108 is not subjected to amplitude / phase correction, and at the start of processing, the amplitude / phase of the baseband signal is unknown on the receiver side. The initial value of the filter coefficient was set to 0 when performing initial training on a channel or a synchronous burst. If the present invention is applied, the initial value of the filter coefficient is set to impulse H ₀ = [0 ... 0 1 0 ... 0] ^T (only the coefficient of the center tap is set to 1) having no frequency characteristics, The coefficient correction unit 109 corrects the amplitude and phase of the coefficient in accordance with the received baseband signal, and initial training can be performed using the corrected coefficient H ′ as the initial value of the filter coefficient.

FIG. 7 is a flowchart showing a processing procedure in the present embodiment. In step 701, it is determined whether a control physical channel or synchronization burst is received. If a control physical channel or synchronization burst is received, the filter coefficient is set to H = H ₀ = [0… 0 1 0… 0] ^T (Step 702). Otherwise, or after completion of step 702, the coefficient correction unit 109 calculates the correction coefficient H ′ and sets H ′ as an initial value of H (step 703). When the processing of step 703 is completed, equalization processing by the equalizer 108 is performed for each symbol for n = 0, 1,..., 149 (steps 704, 705, and 706).
When receiving a control physical channel or synchronization burst, the filter coefficient H of the equalizer 108 is reset to H ₀ = [0… 0 1 0… 0] ^T, and the amplitude and phase of H ₀ are received baseband signals. In accordance with the above, the corrected coefficient H ′ is set as the initial value of the coefficient, the initial training is performed with the preamble as a known symbol, and the training is continued with the signal following the preamble. When receiving a physical channel for communication, filter coefficient correction processing is performed before equalization processing, and training is performed using the corrected filter coefficients as initial values.

  In the first embodiment, the initial value of the filter coefficient of the equalizer is set to 0, and initial training is performed using a preamble. Therefore, in addition to adapting the filter coefficient to the frequency characteristic (distortion characteristic) of the propagation path, Since the phase is also adapted, the training time (convergence time) becomes long. On the other hand, in this embodiment, since the initial value of the filter coefficient of the equalizer is the initial value of the coefficient obtained by combining the amplitude and phase of the received signal, the equalizer has a frequency characteristic (distortion characteristic) of the propagation path. It is only necessary to adapt the filter coefficient to, so that the training time (convergence time) is shortened.

(Example 3)
In the first and second embodiments, the average value <| e _n | ² > in the slot of the determination error power | e _n | ² (n is the symbol number in the slot) is obtained. After the equalization processing, <| e _n | ² > is compared with a preset constant e _th (e _th > 0), and <| e _n | ² ><e _th (or <| e _n | ² > ≦ e _th ), the filter coefficient H is saved in the memory. If <| e _n | ² > ≧ e _th (or <| e _n | ² >> e _th ), the coefficient saved in the memory is restored as H and used as the initial coefficient value at the time of the next slot processing.
The judgment error average power <| e _n | ² > is not greater than −10 dB when a signal according to the data format is received and filter coefficient training is performed normally. When a burst signal with a transmission slot is received and a non-transmission slot is received, it becomes larger than -10 dB.
In this embodiment, when a non-transmission slot is received, <| e _n | ² > is larger than e _th, the coefficient saved in the memory is restored as H, and the initial value of the filter coefficient when receiving the next slot is Become. For this reason, while receiving the non-transmission slot, the saved filter coefficient is restored, and when the burst signal is received, the restored trained filter coefficient becomes the initial value, and the burst signal may be received. it can.

(Example 4)
In the first to third embodiments, assuming that the frequency tracking accuracy in the AFC unit 105 is sufficient, it is assumed that there is almost no phase rotation in the slot. Since this is only continuous and is removed by the coefficient correction unit 109, the amplitude / phase correction unit 501 described with reference to FIG. 5 of the prior art is unnecessary.
However, if the frequency tracking accuracy in the AFC unit 105 is insufficient and the phase rotation in the slot remains, the LMS algorithm cannot follow this phase rotation, and reception characteristics deteriorate or filter coefficients diverge. End up. In such a case, correction means such as the amplitude / phase correction unit 501 is incorporated in the receiver, and the baseband signal corrected in amplitude / phase is accumulated in the buffer 107, and the equalizer 108 and the coefficient correction unit 109 Process.

According to the embodiment of the present invention, for example, the uplink reception during the communication call is initially trained by using the preamble of the control physical channel in the call connection, and is received immediately after the transition to the communication physical channel. Training is performed with the filter coefficient drawn in the slot as an initial value. Downlink reception during contact communication, loudspeaking notification, and deemed voice FAX, initial training is performed with the preamble of the control physical channel in slot 0, and the communication channel is trained with the filter coefficient drawn in the slot received immediately before as the initial value. It is also possible to perform an operation such as
In particular, it is possible to construct a system that is effective in the case of a disaster prevention administrative radio system equipped with fixed receivers in various places in municipalities.

The block diagram of the receiver which implemented the waveform equalizer by this invention. The block diagram which shows one Example of a transversal equalizer. The figure which shows the data format of the slot by ARIB STD-T86. The detailed block diagram of the coefficient correction means of this invention. The block diagram of the conventional receiver. 3 is a flowchart illustrating a processing procedure according to the first embodiment. 10 is a flowchart illustrating a processing procedure according to the second embodiment. The flowchart which shows the process sequence of the conventional receiver. The figure which shows the frame structure by ARIB STD-T86. The figure which shows an example of slot allocation.

Explanation of symbols

101: high-frequency circuit, 102: A / D (analog / digital) converter, 103: quadrature demodulation unit, 104: reception filter unit, 105: AFC (automatic frequency control) unit, 106: frame synchronization unit, 107: buffer, 108: equalizer, 109: coefficient correction unit,
201: Shift register, 202-1 to 202-Ntap: Complex multiplier, 203: Complex adder, 204: Symbol decision unit, 205: Switch, 206: Complex subtractor, 207: LMS algorithm operation unit, 208: Filter coefficient Storage buffer,
401-0 to 401-Ntap-1, 403, 404: Complex multiplier, 402: Complex adder, 405: Complex vector multiplier, 406: Amplitude square calculator, 407: Reciprocal calculator, 408: Complex conjugate calculation vessel,
501: Amplitude / phase correction unit.

Claims (6)

In the filter coefficient correction means used in the waveform equalizer having the filter section, the error calculation section, the filter coefficient training means, and the filter section is a FIR (Finite Impulse Response) filter,
The output y _P of the equalizer when the received signal of the pilot symbol part is input to the equalizer is calculated, the symbol at the time of transmitting the pilot symbol is p, and the amplitude ratio of the y _{p to} the _p is a = | y _p | / | p |, and the phase difference between _p and y _p is θ = θ _yp −θ _p (the unit is radian, θ _yp is the phase of y _p , θ _p is the phase of p), and the filter coefficient H = [h ₀ , h ₁ ,..., h _N−1 ] (N is the number of taps) The coefficient obtained by dividing the a and rotating the phase by θ H ′ = He ^j θ / a = [h ₀ e ^j θ / a, h ₁ e ^j θ / a,…, h _N-1 e ^j θ / a] (e is the base of natural logarithm, j is imaginary unit, e ^j θ is e ^j θ = cos θ + and a coefficient correction means for a waveform equalizer, wherein jsinθ is a rotation factor of angle θ (radian). 2. The coefficient correction means according to claim 1, wherein an operation of a value g = e ^j θ / a multiplied by the coefficient H is a value obtained by dividing the y _p by the p k = y _p / p = y _p p ^* / | p | ² is calculated, the k is divided by the magnitude | k | of the k, and the division result g ′ = k / | k | ² is set to the value g. Coefficient correction means. 3. The filter coefficient corrected by the coefficient correction unit according to claim 1 or 2, in a waveform equalizer having a filter unit, an error calculation unit, a filter coefficient training unit, and the filter unit is an FIR (Finite Impulse Response) filter A waveform equalizer that calculates H ′ and trains H ′ as an initial value of a filter coefficient. In an equalizer having a filter unit, an error calculation unit, filter coefficient training means, and the filter unit is a FIR (Finite Impulse Response) filter, filter coefficients H = [h ₀ , h ₁ ,..., H _N-1 ] ( _Where N is the number of taps) and the center tap coefficient is h _Ncenter = 1 (N _center is the center tap number), otherwise h ₀ = h ₁ =… = h _Ncenter-1 = 0, h _{Ncenter + 1} = h _{Ncenter +2 = ... h N-1 =} 0 and the _{H 0 = [0, ...,} 0, 1, 0, ..., 0] as an initial value of H, the coefficient correction means according to claim 1 or claim 2, wherein A waveform equalizer that calculates and trains the calculated coefficient H ′ as an initial value of a filter coefficient. 5. The waveform equalizer according to claim 3, further comprising a memory for saving filter coefficients, and an in-slot average value <| e _n | of determination error power | e _n | ² (n is a symbol number in the slot). ² >, and the <| e _n | ² > is compared with a preset constant e _th (e _th > 0), and <| e _n | ² ><e _th (or <| e _n | ² > ≦ e _th ), the filter coefficient H is saved in the memory. If <| e _n | ² > ≧ e _th (or <| e _n | ² >>> e _th ), the saved coefficient is stored in the memory. Is restored as H, and the initial value of the filter coefficient during the next slot processing is used as a waveform equalizer. In the reception method of the wireless communication device including the waveform equalization unit having the filter unit, the error calculation unit, and the filter coefficient training unit,
Calculating a filter coefficient using a preamble for the waveform equalization means;
The output of the equalizer when the received signal of the pilot symbol part is input is calculated, and the correction coefficient of the filter coefficient is calculated based on the amplitude ratio and the phase difference of the pilot symbol part with respect to the symbol at the time of transmission. Steps,
Temporarily storing the correction coefficient,
When the wireless communication device performs reception, initial training is performed in the control channel, and training is performed using the stored correction coefficient as an initial value after switching to the communication channel. .