JP6975760B2 - Autocorrelator and receiver - Google Patents

Description

The present invention relates to an autocorrelator and a receiver, and more particularly to an autocorrelator for generating symbol timing in single carrier block transmission, and a receiver using the autocorrelator.

In wideband digital power line transport using a high voltage transmission line, a high multi-level modulation method of 64QAM (Quadrature Amplitude Modulation) is used for high-speed transmission in a limited frequency band (150 kHz to 450 kHz). .. In order to obtain good BER (Bit Error Ratio) characteristics, it is necessary to match the symbol timing of the Nyquist interval generated on the transmitting side with the symbol timing generated on the receiving side. However, normally, the transmission / reception symbol timing is asynchronous, and the reception side symbol timing is not synchronized with the transmission side symbol timing.

Therefore, it is necessary to match the transmission / reception symbol timings, and it is necessary for the receiving side to estimate the transmission symbol timing with high accuracy and generate the reception symbol timing synchronized with the transmission symbol timing. As a result, interference between symbols can be reduced at the output of the roll-off filter on the receiving side, and good BER characteristics can be obtained. In particular, the higher the multi-value modulation method, the higher the accuracy is required. Therefore, in the wideband digital power line transfer method, the development of a DPL (Digital Phase Lock Loop) method that can generate highly accurate reception symbol timing has been developed. It becomes important.

For example, Patent Document 1 describes an autocorrelation calculation unit that performs an autocorrelation calculation on an input time signal, a cross-correlation calculation unit that performs a cross-correlation calculation on an input time signal, and an autocorrelation calculation unit. And the one that detects the timing synchronization position by using the autocorrelation calculation result and the cross-correlation calculation result from the cross-correlation calculation unit is disclosed. The apparatus disclosed in Patent Document 1 states that the timing for demodulating the received data can be accurately generated with high accuracy even under various reception conditions.

Japanese Unexamined Patent Publication No. 2006-2383667

As disclosed in Patent Document 1, the autocorrelation calculation uses the repetition pattern of the signal itself to obtain the correlation only from the received signal, and in general, the autocorrelation is strong against reflection and fading. There is an advantage that it can be realized on a small scale in terms of circuit configuration. However, there is a characteristic drawback that even data and noise with periodicity other than preamble show correlation. On the other hand, the cross-correlation calculation holds the data string of the signal itself known in advance on the receiving side and correlates the received signal with the expected value, and the cross-correlation is for noise and irrelevant data. There is an advantage that highly accurate timing synchronization can be realized even in a Gaussian noise transmission line with low C / N (carrier signal / noise) with almost no correlation detected. However, when the reception waveform changes due to a large deviation in the reception frequency, reflection, fading, etc., each peak becomes smaller and the synchronization characteristics deteriorate.

Since the symbol rate used in the conventional digital power line carrier for transmission lines is about several tens of kilobps, there is no large deviation in the correlation peak position from the correlator even if a delayed wave is present in the received signal. In addition, even if the correlation peak position is deviated and the timing position is deviated, the deviation width is acceptable by the equalizer, so that the BER characteristic is not significantly affected. However, when the method is wideband transmission and the symbol rate is about several hundred kilobps, the influence of the delayed wave, which has not affected the characteristics of DPLL until now, becomes noticeable, and depending on the phase characteristics of the delayed wave, the correlator The correlation peak position from is greatly deviated. Therefore, the equalizer cannot handle it, and the BER characteristics are greatly deteriorated. Since the time width between symbols is shortened in order to estimate the transmission symbol timing with high accuracy, the number of oversamplings at a higher rate is required to extract the symbol timing with high accuracy.

For example, it is necessary to interpolate the output signal of the receiving roll-off filter at a symbol rate of about 128 times oversampling. For this reason, a complex number correlator that operates with 128 times oversampling is required, but when a correlator with a large number of taps is used at a high sampling rate, the scale of the FPGA (Field Programmable Gate Array) becomes large. Is a concern.

For example, when 128 times oversampling is used, at 64 taps, the FPGA has a multiplier of 256 and the DSP (Digital Signal Processing) block requires 64 modules, and at 32 taps, the multiplier is 128 and the DSP. 32 blocks are required. Further, with 16 taps, it can be seen that the multiplication number is 64 and the DSP block requires 16 modules, and when the number of taps is doubled, the circuit scale is proportionally increased. This is a problem when applying a correlator during oversampling, and since the correlator occupies most of the number of modules allocated to FPGA (for example, 274 modules), it can be performed by other processing such as FFT (Fast Fourier Transform) processing. The number of modules is tight. From this, it is desirable that the correlator is composed of the correlation with the smallest possible number of taps. However, when the number of taps of the correlator is reduced, there arises a problem that the correlation peak power value and the correlation characteristic cannot appear at the required timing.

The present invention has been made in view of these circumstances, and as a correlator for supporting a wideband digital power line transfer method for transmission lines, a required correlation peak is required even when the scale of the number of taps of the correlator is reduced. It is an object of the present invention to provide an autocorrelator capable of ensuring a power value and a receiver having this autocorrelator.

In order to solve the above problems, the first technical means of the present invention is an autocorrelator that generates symbol timing in single carrier block transmission, from received data having a training sequence in which a Zadoff-Chu sequence is arranged. , An autocorrelation device that calculates the autocorrelation value output using the even symmetry of the Zadoff-Chu series. The training sequence is arranged at the beginning and the back of the data block, and the autocorrelation value output is at least. , (N-1) The first autocorrelation value output obtained by using the training sequence after the th data block or the training sequence before the nth data block, and the first autocorrelation value output. A second autocorrelation value output that is separated by a predetermined number of samples and is obtained by using the training sequence after the (n-1) th data block and the training sequence before the nth data block following it. It is characterized in that the total value of the two autocorrelation value outputs of the autocorrelation value output of is output.

で表されることを特徴とするものである。 The second technical means is that in the first technical means, the autocorrelation value output Ca (t) of the autocorrelator has k as the number of oversamplings of the training signal, M as the number of multipliers (number of taps), and a. Is the number of symbols with the minimum distance that is evenly symmetric in the Zadoff-Chu series, rd (t) is the value of the received data (complex number) at the time index t, and rd (t) * is the received data rd (t). When the value of the complex conjugate number is used, the following equation

It is characterized by being represented by.

The third technical means is characterized in that, in the first or second technical means, the number M of the multiplier is 16 to 48.

The fourth technical means is that in any one of the first to third technical means, the predetermined number of samples is the number of samples corresponding to the number of symbols in the training sequence, or one half of the number of symbols. It is characterized by having a corresponding number of samples.

The fifth technical means is a receiver, characterized in that it has an autocorrelator of any one of the first to fourth technical means.

According to the present invention, as a correlator for supporting a wideband digital power line carrier system for transmission lines, it is possible to obtain a correlator that can secure a required correlation peak power value even when the scale of the number of taps of the correlator is reduced. Can be done.

It is a figure which shows a part of the structure of the receiver side which uses the autocorrelator which concerns on this invention. It is a figure which shows the structural example of the block transmission data for wideband digital power line carrier used in this invention. It is a figure which shows the example of the autocorrelator which concerns on this invention. It is a figure for demonstrating the relationship of the training sequence autocorrelation measurement in which the Zadoff-Chu sequence is arranged in the autocorrelation device which concerns on this invention. It is a figure which shows the example of a cross-correlator. It is a figure which shows the time characteristic of the autocorrelator (tap = 16) which concerns on this invention. It is a figure which shows the time characteristic of the autocorrelator (tap = 20) which concerns on this invention. It is a figure which shows the time characteristic of the autocorrelator (tap = 24) which concerns on this invention. It is a figure which shows the time characteristic of the cross-correlator (tap = 16) as a comparative example. It is a figure which shows the time characteristic of the cross-correlator (tap = 32) as a comparative example. It is a figure which shows the time characteristic of the cross-correlator (tap = 64) as a comparative example. It is a figure which shows the correlation peak response characteristic of each correlator of the autocorrelator which concerns on this invention and the cross-correlator as a comparative example.

Hereinafter, preferred embodiments of the autocorrelator and receiver of the present invention will be described with reference to the drawings. In the following description, the configurations with the same reference numerals may be omitted because the configurations with the same reference numerals are the same in different drawings. It should be noted that the present invention is not limited to the examples in these embodiments, but includes all modifications within the scope of the matters described in the claims and within the scope of equality. Further, as long as a combination of a plurality of embodiments is possible, the present invention includes a combination of arbitrary embodiments.

FIG. 1 is a diagram showing a part of the configuration of a receiver in which the autocorrelator according to the present invention is used.
The RF signal received from the transmission line is demodulated by the demodulator 11, and the demodulated signal is decimmed to 1 / d by the 1 / d decimation unit 12 so as to have a required sampling speed. The decimation signal is band-limited by the Nyquist filter (roll-off filter) unit 13. Then, the echo signal wrapping around from the transmission signal is removed by the echo canceling unit 14, and only the element of the received signal is extracted.

Next, in order to generate the symbol timing of the received signal from the training signal transmitted from the transmitting side, the k-fold interpolation unit 15 performs interpolation with k-fold oversampling. The k-fold oversampled signal is output to the DPLL 20 and the symbol data generation unit 16. In the DPLL 20, the correlation measurement value of the received signal is output by the autocorrelator 21, the correlation peak value is searched by the correlation peak search unit 22, and the reception symbol timing is calculated from the correlation peak position searched by the correlation peak position calculation unit 23. ..

The correction value control unit 24 for the symbol timing position corrects the reception symbol timing so that the transmission symbol timing and the reception symbol timing are synchronized, and the symbol timing generation unit 25 generates a symbol timing signal at the corrected symbol timing. Then, the symbol data generation unit 16 punches out the received data interpolated based on the generated symbol timing signal, and extracts the symbol data corresponding to the symbol rate. Since the output symbol data has a frequency offset, the frequency offset is corrected by the frequency offset estimation / correction unit 17 and input to the equalizer 18. In the equalizer 18, the element of the delayed wave is removed from the symbol data, and a signal that estimates the transmitted symbol data is output.

FIG. 2 is a diagram showing a configuration example of block transmission data for carrying a wideband digital power line. In wideband digital power line carrier for transmission lines, in order to apply single carrier block transmission, training sequences (hereinafter, also referred to as “TS”) are arranged at the beginning and the end of the data symbol as shown in FIG. In this embodiment, the Zadoff-Chu sequence represented by the equation (1) is used as the TS in order to generate the received symbol timing with high accuracy.

In the equation (1), N _ts is the number of symbols of TS and is given as the series length (integer) of the Zadoff-Chu series. m is an arbitrary integer. The constant m is selected to be relatively prime with the _{number of symbols N ts} equal to the sequence length in order to reduce the PAPR (Peak to Average Power Ratio).

FIG. 4 is a diagram for explaining the relationship between the training sequence in which the Zadoff-Chu sequence is arranged and the autocorrelation measurement in the autocorrelation device according to the present invention. Each value shown in FIG. 4 shows the value u (t) of the Zadoff-Chu series when the _{number of symbols N ts of TS is 64 and the constant m is 61.} As can be seen from FIG. 4, the Zadoff-Chu series is an even symmetric series. Incidentally, the symbol number _{N d} data symbols shown in FIG. 2, for example, 960, number of symbols in the block is 1088. The TS and data symbols are DFT (Discrete Fourier Transform) blocks.

FIG. 3 is a diagram showing an example of an autocorrelator according to the present invention. Autocorrelator 21 shown in FIG. 3 (in the present embodiment, M = 16) M-number multipliers _W 1 to _W-M of, 2M-1 single delay units _{D 1 ~D 2 (M-1} ) +1 It is equipped with two adders A1 and A2, and is configured to measure the autocorrelation of the Zadoff-Chu series with even symmetry. In FIG. 3, k is the oversampling number in the k times interpolation section 15 shown in FIG. 1, delay units _D 1 _{to D M-1} and delayer _{D M + 1 ~D 2 (M} -1) +1 is It has a delay amount of k times (128 times) the number of samplings. Further, when the number of samples between symbols measuring the autocorrelation and 128 oversampling delay ak of the delay device D _M is the 256 sampling the sampling number between symbols of the even symmetric having the minimum distance (2 symbols) Used.

And one value of the delayed sequence from delay device D ₁ in D _M-1, and the complex conjugate value of the other values of the delayed sequence with D _{2 (M-1) +1} from delay unit D _M is the multiplication is multiplied by the vessel W ₁ to _W-M, multiplication results are added for each 1 sampling by summer a _1, total output from the adder a _1. The autocorrelation value output Ca (t), which is the output value of the adder A _{1, is given by the following equation (2).} As described above, in equation (2), k is the number of oversamplings of the training signal, M is the number of multipliers, a is the number of symbols with the minimum distance that is even symmetric in the Zadoff-Chu series, and rd (t) is. , The value of the received data (complex number) at the time index t, rd (t) ^* is the value of the complex conjugate number of the received data rd (t).

It is possible to obtain the correlation peak and obtain the symbol timing signal based only on the autocorrelation value output Ca (t) of the autocorrelation device 21, but the TS of the Zadoff-Chu series is in front of the _{Nd symbol of the data block.} On the receiving side, it is possible to obtain a large correlation peak by using the autocorrelation value output Ca (t) at a plurality of points by using the even symmetry of the Zadoff-Chu series.

As shown in FIG. 4, when the number of symbols N _ts of the Zadoff-Chu series constituting the TS is an even number, the two TSs are _{half the value of the number of symbols N ts} of each TS (time index t). In addition to being even-symmetrical before and after (value at = 32), even-numbered symmetry occurs before and after the value at the beginning of the TS in front of the rear block (value at t = 0), and peaks occur in the correlation values.

ここで、Ｃad(t)は現在の自己相関値出力Ｃa(t)と、所定サンプルとして、３２シンボルに相当するサンプル数（３２×１２８=４０９６サンプリング）遅延した自己相関値出力Ｃa(t-Ｎ)の和である。このように、前方のＴＳ系列と後方のＴＳ系列の２つのＴＳを利用することにより、２個所ピーク値（３２シンボル間隔）の和をとることによって、よりノイズに強い相関ピーク値を得ることができる。 Therefore, as shown in equation (3), by adding in an adder A ₂ autocorrelation value output Ca (t) and N samples (N = 32) delayed autocorrelation value output Ca (t-N) Since the current correlation peak value and the correlation peak value delayed by N samples are added, it can be obtained as a correlation function of a large correlation peak value.

Here, Cad (t) is the current autocorrelation value output Ca (t) and the autocorrelation value output Ca (t-N) delayed by the number of samples corresponding to 32 symbols (32 × 128 = 4096 samplings) as a predetermined sample. ) Is the sum. In this way, by using two TSs, the front TS series and the rear TS series, it is possible to obtain a correlation peak value that is more resistant to noise by summing the peak values at two locations (32 symbol intervals). can.

Explaining the autocorrelation value output CAD (t) with reference to FIG. 4, the TS behind the front block and the TS in front of the rear block are continuously received. Then, for example, the autocorrelation value output Ca (t) is obtained from the TS behind the front (n-1) th block, and the TS behind the front (n-1) th block and the nth rear. Using the two TSs of the TS in front of the block, the autocorrelation value output Ca (t-N) delayed by the N symbol (N = 32) is obtained, and both are added. Alternatively, the autocorrelation value output Ca (t) and the N symbol (N = 32) delayed autocorrelation value output Ca (t-N) are the TS behind the front (n-1) th block and the n behind. The autocorrelation value output Ca (t) is obtained using the two TSs of the TS in front of the second block, and the autocorrelation delayed by N symbols (N = 32) from the TS in front of the nth block behind. The value output Ca (t-N) may be obtained. By the way, the output value of each multiplier based on the two values of TS for measuring the autocorrelation value shown by the arrow in FIG. 4 is 1, and the values of CAD (t) and Ca (t-N) are 16 respectively. It becomes a peak value having a value.

Added by the adder A ₁ autocorrelation value output Cad (t) is squared by that the absolute value | Cad | as autocorrelation value output of autocorrelator 21 as ^{2 (t),} the correlation peak searcher 22 Is output to. In the equation (3), the autocorrelation value outputs at two places separated by 32 symbols are added, but it is also possible to add the autocorrelation value outputs at three places delayed by 32 symbols.

Next, the time characteristics of the autocorrelator and the conventional cross-correlator according to the present invention will be described. FIG. 5 is a diagram showing an example of a cross-correlator. When the cross-correlator 31 is used, the autocorrelator 21 of the DPLL 20 in FIG. 1 is replaced with the cross-correlator 31. In the cross-correlator 31 used for DPLL, when the number of taps of the cross-correlator 31 _{is equal to the number of symbols N ts of the TS of the Zadoff-Chu series, the delay devices D 1 to} D have the same delay time as the oversampling number k. _{N-1 is arranged by the number of symbols N ts-} 1 of the TS of the Zadoff-Chu series, and the cross-correlation signal is input from the k-fold cross-correlation unit 15.

A multiplier W _{1 to} W _{Nts having N ts} of symbols of the TS of the Zadoff-Chu series is arranged at the output of the delay device D, and is multiplied by the complex conjugate Zadoff-Chu series signal serving as a reference signal. The multiplied N _ts signals are added by an adder for each sampling, and the total value C (t) is output. Then, by multiplying the total value C (t) and the complex conjugate signal C (t) ^{* of the} total value C (t), the absolute value correlation value | C (t) | ² is output. When reducing the number of taps of the cross-correlator 31, the number of delay devices D can be reduced together with the multiplier W.

FIG. 6A is a diagram showing the time characteristics of the autocorrelator (tap = 16) according to the present invention, and FIG. 6B is a diagram showing the time characteristics of the autocorrelator (tap = 20) according to the present invention. FIG. 6C is a diagram showing the time characteristics of the autocorrelator (tap = 24) according to the present invention. Further, FIG. 7A is a diagram showing the time characteristics of the cross-correlator (tap = 16) as a comparative example, and FIG. 7B is a diagram showing the time characteristics of the cross-correlator (tap = 32) as a comparative example. Yes, FIG. 7C is a diagram showing the time characteristics of a cross-correlator (tap = 64) as a comparative example. Further, FIG. 8 is a diagram showing the correlation peak response characteristics of each correlator of the autocorrelator according to the present invention and the cross-correlator as a comparative example. 7A, 7B, and 7C show the time characteristics according to the circuit configuration of the conventional cross-correlator, and the circuit method is different from the time characteristics due to the circuit configuration of the autocorrelator according to the present invention. It is a thing.

In the autocorrelator according to the present invention, as shown in FIGS. 6A, 6B, and 6C, when the number of taps is 16, 20, or 24, the two peak points have a constant sampling interval (N symbol × k). It has the characteristics that appear in the oversampling). If the number of taps is increased, a larger peak value can be obtained, so it is practical until the number of taps is about 48. However, increasing the number of taps increases the circuit scale by multiplying the number of oversamplings. It is not practical to increase it to 64. Therefore, 16 to 48 taps are realistic at the time of oversampling from the viewpoint of ease of mounting on FPGA and cost. This becomes more pronounced as the number of oversamplings increases, such as 128 times.

Further, FIG. 8 shows the relative sampling number when the sample point at the peak position is 0 on the horizontal axis, and the correlation value is shown on the vertical axis. As shown in FIG. 8, when the correlation functions before and after the maximum correlation value is compared between the autocorrelator 21 having 16 taps and the partial cross-correlator 31 having 16 taps, the partial cross-correlation with 16 taps is compared. The correlator 31 has not been able to obtain steep peak characteristics. Therefore, when extracted from these as symbol timing, it is presumed that the algorithm and applied circuit for estimating the timing position will be very large. This point is the same as when the number of taps of the cross-correlator 31 is 32 and 64, and when comparing the correlation functions before and after the maximum correlation value, the 16-tap autocorrelator is the steepest. A characteristic with a peak is obtained.

Next, in the correlation function of the cross-correlator 31 with 32 taps, as shown in FIG. 7B, five correlation peak positions appear, and the intervals between the peak positions become aperiodic. There is. The distance between the two peaks at which the highest peak value is obtained is constant (64 symbols = 8192 samples). It is possible to extract the symbol timing using these two peak points, but since the level difference from other peak values has not been obtained, the threshold value for peak position search is taken into consideration when considering the effects of noise and the like. It is thought that it will be difficult to set. This point is the same for the time characteristic of the cross-correlator 31 having 16 taps shown in FIG. 7A.

On the other hand, in the correlation function of the cross-correlator 31 with 64 taps, as shown in FIG. 7C, high correlation peak values appear at two locations, and the peak position interval is a constant sampling interval (64 symbols = 8192 samples). It has become. Therefore, the cross-correlator having 64 taps is the most appropriate as the characteristic of the correlation peak value, but it is difficult to apply from the viewpoint of the circuit scale because the scale of the number of DSP modules of the FPGA increases as described above.

From the above, the correlation peak characteristic of the autocorrelator 21 with 16 taps is inferior to the cross-correlator 31 with 64 taps, but the correlation function at the peak time for extracting the symbol timing is sufficient power. It is the value and the steepest peak characteristic. From the viewpoint that the circuit scale of the FPGA can be minimized, the autocorrelator 21 having 16 taps is suitable as a correlator applied to the DPLL method of wideband digital power line carrier.

As described above, the autocorrelator of the present embodiment does not require the Zadoff-Chu sequence as a reference signal on the receiving side, and the deterioration of the correlation characteristic is small even in the symbol string having the frequency offset as compared with the cross-correlator. .. Further, when calculating the sum of the correlation value outputs separated by a predetermined number of samples, the cross-correlator is affected by the error due to the frequency offset for the delay time of the TS before and after, but the autocorrelator is affected by such an error. Never receive. Therefore, in the autocorrelator of the present embodiment, when the influence of the delayed wave or noise is small, the correlation peak position can be applied as the symbol timing.

11 ... demodulator, 12 ... 1 / d decimation section, 13 ... Nyquist filter (roll-off filter) section, 14 ... echo cancel section, 15 ... k times cross-correlation section, 16 ... symbol data generation section, 17 ... correction section , 18 ... Equalizer, 20 ... DPLL, 21 ... Autocorrelator, 22 ... Correlation peak search unit, 23 ... Correlation peak position calculation unit, 24 ... Correction value control unit, 25 ... Symbol timing generation unit, 31 ... Cross-correlation Instrument, A1 ... adder, A2 ... adder, D ... delayer, W ... multiplier.

Claims (5)

An autocorrelator that generates symbol timing in single-carrier block transmission, and calculates the autocorrelation value output using the even symmetry of the Zadoff-Chu series from received data that has a training sequence in which the Zadoff-Chu series is placed. Is an autocorrelator
The training sequence is arranged at the beginning and the back of the data block, and as the autocorrelation value output, at least the training sequence after the (n-1) th data block or the training before the nth data block. The first autocorrelation value output obtained by using the sequence and the autocorrelation value output separated from the first autocorrelation value output by a predetermined number of samples, which are behind the (n-1) th data block. The autocorrelation is characterized by outputting the sum of the two autocorrelation value outputs of the second autocorrelation value output obtained by using the training sequence and the training sequence in front of the nth data block consecutive to the training sequence. Correlator. In the autocorrelation value output Ca (t) of the autocorrelator, k is the number of oversamplings of the training signal, M is the number of multipliers (number of taps), and a is the symbol of the minimum distance that is even symmetric in the Zadoff-Chu series. When the number and rd (t) are the value of the received data (complex number) at the time index t and the rd (t) * is the value of the complex conjugate number of the received data rd (t), the following equation is used.

The autocorrelator according to claim 1, wherein the autocorrelator is represented by. The autocorrelator according to claim 1 or 2, wherein the number M of the multipliers is 16 to 48. One of claims 1 to 3, wherein the predetermined number of samples is a number of samples corresponding to the number of symbols of the training sequence or a number of samples corresponding to one half of the number of symbols. The autocorrelator described in. The receiver having the autocorrelator according to any one of claims 1 to 4.

Images