JP3436647B2 - Diversity receiver and diversity receiving method - Google Patents

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention is used for wireless communication. INDUSTRIAL APPLICABILITY The present invention is suitable for use in digital mobile communication in which complicated multiple paths occur in a wireless section.

[0002]

2. Description of the Related Art In a radio transmission path, a phenomenon called fading in which a received signal level fluctuates drastically occurs and transmission quality may be significantly deteriorated. Diversity receiver is one of the compensation techniques for fading.

Signals received from a plurality of antennas spatially sufficiently separated fluctuate almost independently. Therefore, by combining or selecting these to obtain a received signal, deterioration of reception quality due to a drop in reception level is significantly reduced. be able to.

Further, in a transmission line with a large difference in propagation delay time, intersymbol interference occurs due to interference with a delayed wave, and reception quality deteriorates. In such a case, the reception quality can be improved by combining the diversity receiver with an equalizer.

As an example in which the equalizer is applied to a diversity receiver, a training interval is a recursive least square (Recursive L).
east squares (RLS) algorithm and operates as an equalizer with the same number of diversity branches, and uses the Least Mean Square (LMS) algorithm in the data section, and operates as a single combined diversity receiver. The method of doing is proposed. The iterative least squares algorithm has a fast convergence property, but the amount of computation and the memory capacity required for computation increase in proportion to the square of the number of taps, and as a method of implementing a diversity receiver,
It is known that the properties of compound molding are superior to those of selective molding. For this reason, in this method, it is possible to reduce the amount of calculation and realize good reception quality while maintaining the high-speed initial pull-in characteristic.

As a conventional technique, the operation of the diversity receiver for switching the equalizer configuration and the algorithm between the training section and the data section will be described. FIG. 5 shows a block configuration of a diversity receiver showing an example of a conventional two-branch configuration when a decision feedback equalizer is used. Note that FIG.
The switches 40, 41, 42, 43, and 44 are in the state during the training section operation.

During training, the training signal is supplied from the training signal memory 30 to the feedback filters 80 and 81 and the tap coefficient estimators 32 and 34 based on the recursive least squares algorithm.

The received signals input to the input terminals 61 and 62 pass through the feedforward filters 10 and 20 and are added to the outputs of the feedback filters 80 and 81, respectively, and then input to the tap coefficient estimators 32 and 34. . The tap coefficient estimators 32 and 34 use the training signal input by the recursive least squares algorithm and the outputs of the feedforward filters 10 and 20, and taps 15 to 1 so that the root mean square value of the differences is minimized.
Determine a coefficient of 9, 25-29.

When entering the data section, the switches 40, 41,
42, 43, 44 is switched. At this time, the outputs of the tap coefficient estimators 32 and 34 based on the successive least squares algorithm are weighted by the initialization unit 31, and are input as initial values to the tap coefficient estimator 33 based on the least mean square algorithm. The received signals input to the input terminals 61 and 62 pass through the feedforward filters 10 and 20 and the adder 72.
Is added in. Further, after being added by the output of the feedback filter 80 by the adder 73, it is identified by the code identifier 36 and output from the output terminal 50.

The output of the code discriminator 36 is input to the feedback filter 80 and used for processing the next symbol. In the data section, the coefficients of the taps 15 to 19 and 25 to 28 of the feedforward filters 10 and 20 and the feedback filter 80 are determined by the tap coefficient estimator 33 using the least mean square algorithm by regarding them as one filter.

When the processing shifts from the training section to the data section and the equalizer configuration and the algorithm are switched, the integral value σ of the equalization squared error defined by the equation (1).
_{Based on 1} ² and σ ₂ ² , the tap coefficients in the training interval are weighted using equation (2), and this is used as the initial value, and the entire tap coefficient is used as a single equalizer thereafter. Update. Equalization error power is considered to be inversely proportional to the carrier-to-noise ratio of the received signal, and the maximum ratio combining is performed by weighting the tap coefficient for the equalizer in each branch of the diversity receiver based on the equalization error power. Can be realized. However, i is the branch number (i =
1, 2), h _ij is the training section, h _ij ′ is the tap coefficient of the data section, N _cyc is the number of equalization squared error integrations, e
_ik is the equalization error of the kth symbol.

[0012]

[Equation 1] FIG. 6 is a time chart for explaining the order of operations in the conventional example. The horizontal axis shows time, and the vertical axis shows main parameters. According to this, in the training section, the equalizer 1
0 and 20 are independent, and the signal type arriving at the input terminals 61 and 62 is a training signal. In the data section, data signals arrive at the input terminals 61 and 62, and the outputs of the feedforward filters 10 and 20 are combined by the adder 72. The algorithm is a recursive least squares algorithm (RLS) in the training interval and a least mean squares algorithm (LMS) in the data interval.
The initialization by the initialization unit 31 is performed at the end of the training section when the tap coefficient estimators 32 and 34 take over control to the tap coefficient estimator 33.

[0013]

It is desirable that the number of training symbols is small from the viewpoint of transmission efficiency and the amount of calculation. However, if the number of training symbols is small, the configuration and algorithm of the equalizer will be switched before convergence is sufficient. Therefore, it is necessary to pull in the tap coefficient again after the switching.

If the tap coefficients are not sufficiently converged during training, the equalization error power is not necessarily inversely proportional to the carrier-to-noise ratio, so that maximum ratio combining is not achieved. Due to these causes, the reception quality is deteriorated.
In solving these problems, since the merit of reducing the number of training symbols decreases when the amount of calculation increases, the amount of calculation needs to be about the same as in the conventional technique with the same number of training symbols.

The present invention has been made against such a background, and when the number of training symbols of the diversity receiver to which the equalizer is applied is reduced, the reception quality is improved without greatly increasing the calculation amount. It is an object of the present invention to provide a diversity receiver and a diversity receiving method capable of achieving the above. It is an object of the present invention to provide a diversity receiver and a diversity receiving method capable of performing diversity reception with high reception quality.
An object of the present invention is to provide a diversity receiver and a diversity receiving method capable of improving reception quality without increasing the number of training symbols.

[0016]

In order to solve these problems, according to the present invention, after the training by the successive least squares algorithm is completed, the training signal is re-trained M times (M ≧ 2 ) by using the received signal in the training section. The main feature is to do. As a result, the equalizer is sufficiently retracted. Data section processing is performed after retraining.

At this time, the weighting coefficient for weighting the tap coefficient of the equalizer of each branch of the diversity receiver is used for the received signal in the training section by using the tap coefficient which has been sufficiently converged after the training. It is desirable to calculate. This improves the accuracy of tap coefficient weighting.

Further, in order to realize a faster pull-in characteristic during retraining or in the data section,
It is desirable to use the Normalized LMS (NLMS) algorithm. The normalized least mean square algorithm is an algorithm that normalizes the step size that defines the amount of change when updating the tap coefficient in the least mean square algorithm with the average power of the received signal, and has a faster pull-in characteristic than the least mean square algorithm. Have.
In order to use the normalized least mean square algorithm, it is necessary to calculate the average power of the received signal, but in the present invention, the average value of the diagonal components of the correlation matrix in the recursive least squares algorithm used in the training section is equivalent to this. Therefore, it is not necessary to calculate again, and the value at the end of the training section may be held for one slot.

That is, a first aspect of the present invention is a diversity receiver, which is provided with an equalizer that switches the configuration and algorithm between the training section and the data section of the received signal. A feature of the present invention is that the received signal is stored in a training section of the received signal by using the received signal storage means for accumulating the received signal and making it possible to read out a plurality of times and the first equalizer configuration and the first algorithm. The first means for updating the tap coefficient and the second equalizer configuration and the second algorithm for switching the algorithm to the second equalizer configuration and the second algorithm when the processing by the first means completes one cycle. Means and the received signals of the training section accumulated in the received signal storage means,
And third means for updating process again tap coefficients over multiple times again using the second equalizer structure and said second algorithm, said second equalized after the end the third means And a fourth means for updating the tap coefficient in the data section by using the second configuration and the second configuration.

Further, the present invention is a diversity receiver, in which N delay line filters with taps corresponding to inputs from N demodulators and the tap coefficients are regarded as N individual filters. Output of the first tap coefficient estimator for setting, a second tap coefficient estimator for setting the tap coefficient by regarding the N delay line filters with taps as one filter To the output of the second tap coefficient estimator, and tap coefficient switching means for providing the switched output to the N tapped delay line filters, and the output of the first tap coefficient estimator at the time of switching. And initializing means for converting and setting as the initial value of the second tap coefficient estimator, the initializing means weighting the tap coefficients of the N tapped delay line filters. A diversity receiver including a.

The feature of the present invention is that N
Received signals from the demodulators of the stage are accumulated and the output is
A reception signal memory connected to the tapped delay line filters, a training signal memory in which the signals received in the training section are stored in advance, and an adder for adding the outputs of the N tapped delay line filters, A code discriminator that discriminates the output of the adder to generate a demodulated signal sequence, the training signal memory output and the output of the code discriminator are input, and one of the first and second taps is selected. Training signal switching means to be supplied to the coefficient estimator, and each output of the N tapped delay line filters with respect to the received signal of the training section accumulated in the received signal memory as initial training Input to the tap coefficient estimator, and the training signal switching means outputs the output of the training signal memory to the first The output of the adder is input to the second tap coefficient estimator and supplied to the tap coefficient estimator, and subsequently, as retraining, with respect to the received signal of the training section accumulated in the received signal memory, the training is performed. The signal switching means supplies the training signal memory output to the second tap coefficient estimator, and as the processing of the data section, the output of the adder is the output of the adder with respect to the reception signal of the data section accumulated in the reception signal memory. The training signal switching means is supplied to the second tap coefficient estimator, and the training signal switching means supplies the code discriminator output to the second tap coefficient estimator for processing.

The initialization means uses a weighting coefficient calculated from the tap coefficient immediately before switching from the first to the second tap coefficient estimator and the received signal in the training section accumulated in the received signal memory, It is desirable to include means for weighting the tap coefficients of the N tapped delay line filters.

Preferably, the first tap coefficient estimator operates using a recursive least squares algorithm and the second tap coefficient estimator operates using a least mean square algorithm.

The least-squares-means algorithm is preferably a normalized least-squares-means algorithm in which the step size in updating the tap coefficient is normalized by the average power of the received signal.

A second aspect of the present invention is a diversity receiving method, which is provided with an equalizer that switches the configuration and algorithm between the training section and the data section of the received signal. A feature of the present invention is that the received signal is stored in the received signal memory, and the tap coefficient in the training section of the received signal is updated using the first equalizer configuration and the first algorithm,
At the time when this update process has completed one cycle, the configuration and algorithm are switched to the second equalizer configuration and the second algorithm, and the second equalizer is again applied to the received signal in the training section accumulated in the received signal memory. performs update processing of the tap coefficients over multiple times with equalizer structure and the second algorithm, after the update processing ends, the second equalizer structure and the second data section by using the algorithm It is in the process of updating the tap coefficient of.

[0026]

DETAILED DESCRIPTION OF THE INVENTION

[0027]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The configuration of an embodiment of the present invention will be described with reference to FIG. FIG. 1 is a block diagram of a diversity receiver showing an example of a two-branch configuration when the decision feedback equalizer of the embodiment of the present invention is used.

The present invention is a diversity receiver, comprising:
A feedforward filter 10 as a delay line filter with two taps corresponding to inputs from a single demodulator,
(20) and feedback filter 80, (81)
The first tap coefficient estimator 32, which sets the tap coefficients by regarding these two filters (10 and 80 form a set, 20 and 81 form a set) as individual filters. 34), the second tap coefficient estimator 35, and the outputs of the tap coefficient estimators 32 and 34 are switched to the outputs of the tap coefficient estimator 35, and the switched output is fed-forward filter 10, (20) and feedback. Switches 40 and 41 as tap coefficient switching means provided to filters 80 and (81), switches 42 and 43 as switches for switching the configuration of the equalizer, and outputs of tap coefficient estimators 32 and 34 at the time of switching. And an initialization unit 31 as an initialization means for converting the values to set as an initial value of the tap coefficient estimator 35. Forward filter 10, (20)
Also, the diversity receiver weights the tap coefficients of the feedback filters 80 and (81).

The features of the present invention are as follows.
Received signal memory 4 for accumulating received signals from demodulators
5, (46). Further, a training signal memory 30 in which signals received in the training section are stored in advance, an adder 72 for adding the outputs of the feedforward filters 10 and 20 via a switch 42, an output of the adder 72 and a feedback filter. Adder 7 for adding 80 and the output of (81)
3, (74), the code discriminator 36 for discriminating the output of the adder 73 to generate a demodulated signal sequence, the output of the training signal memory 30 and the output of the code discriminator 36, and selecting one of them. And a switch 44 as a training signal switching means to be supplied to the tap coefficient estimators 32, 34, and 35. The output of the adder 73 is branched and given to the tap coefficient estimators 32 and 35, and the output of the adder 74 is given to the tap coefficient estimator 34.

The operation of the embodiment of the present invention will be described. The received signals input to the input terminals 61 and 62 are stored in the received signal memories 45 and 46, respectively. The training signal is a known pattern such as UW (unique word). The training signal is supplied from the training signal memory 30 to the feedback filters 80 and 81 and the tap coefficient estimators 32 and 34 based on the recursive least squares algorithm. The received signals in the training section accumulated in the received signal memories 45 and 46 are sequentially read out, passed through the feedforward filters 10 and 20, respectively, and output from the feedback filters 80 and 81 and the adder 7 respectively.
After being added by 3 and 74 respectively, they are input to the tap coefficient estimators 32 and 34. Tap coefficient estimator 32,
In 34, the feedforward filter 10, (20) and the feedback filter 80, (81) are used as individual filters, and the training signal input by the successive least squares algorithm and the adders 73, 74 are input.
Output, the coefficients of taps 15 to 19 and (25 to 29) are determined so that the mean square value of the difference is minimized.

Initialization will be described. Initialization unit 31
From the tap estimator 32, 34 to the tap coefficient estimator 35
The tap coefficient immediately before switching to the
Using the weighting factors calculated from the received signals of the training section accumulated in 46, the feedforward filter 1
The tap coefficients of 0 and 20 and the feedback filters 80 and 81 are weighted.

After the above initialization, retraining is started. The operation during retraining will be described below. Tap coefficient estimators 32, 34 based on the recursive least squares algorithm
The output of is weighted by the initialization unit 31, and the result is input to the tap coefficient estimator 35 by the normalized least mean square algorithm as an initial value. Switch 40, 4 here
1, 42, 43 are switched, and the reception signal memory 45,
The received signals of the training sections accumulated in 46 are input and processed. Received signals in the training section are sequentially read out, and each is fed-forward filter 1
After passing through 0 and 20, after being added by the adder 72, the output of the feedback filter 80 and the adder 73 are added and input to the tap coefficient estimator 35. The training signal from the training signal memory 30 is supplied to the feedback filter 80 and the tap coefficient estimator 35. The tap coefficient estimator 35 determines the coefficients of the taps 15 to 19 and 25 to 28 from the input training signal and the output of the adder 73 by the normalized least mean square algorithm. This retraining process is repeated M times (M ≧ 2 ).

After the retraining process is completed, the switch 44 is changed over, and the received signal in the data section is processed thereafter. The output of the discriminator 36 is used in this data section. At this time, the tap coefficient estimator 35 continuously processes the signals in the data section thereafter. The reception signals in the data sections accumulated in the reception signal memories 45 and 46 are sequentially read out, and added in the adder 72 via the feedforward filters 10 and 20, respectively.
The output of the adder 72 is added to the output of the feedback filter 80 in the adder 73, identified by the code identifying unit 36, and output from the output terminal 50. The output of the code discriminator 36 is input to the feedback filter 80, fed back to the adder 73 for use in processing the next symbol, and combined with the output of the adder 72. This combined output is the tap coefficient estimator 35.
Entered in. The tap coefficient estimator 35 continues to update the tap coefficient by the normalized least mean square algorithm.

FIG. 2 is a time chart for explaining the sequence of operations of the above embodiment of the present invention. The horizontal axis shows time, and the vertical axis shows main parameters.

The internal structure of the reception signal memories 45 and 46 is shown in FIG. FIG. 3 shows the internal structure of the received signal memory and the switch status in the retraining section. In the training section, both switches SW1 and SW2 are closed,
Write the training signal in the memory M. In the retraining section, the switch SW1 is opened and the input terminal 61,
The input from 62 is cut off, and the training signal stored in the memory M is read. In the data section, the switch SW1 is closed, the switch SW2 is opened, and the data signal is received. (Summary of Examples) For a diversity receiver having 6 feedforwards and 2 taps feedback, the number of times of retraining is twice, and N _{cyc in the} formula (1) is 8, the training 12 symbols, and the data 160 symbols as compared with the conventional system. A comparison was made regarding the amount of calculation and the reception quality.
The calculation amount is increased by 6% to 9% as compared with the conventional technique.

FIG. 4 shows the result of measurement of the bit error rate characteristics under the two-wave independent Rayleigh fading with the maximum Doppler frequency of 10 Hz and the delay time difference of 2 μS. FIG. 4 is a diagram showing the effect of the embodiment of the present invention, in which the carrier-to-noise power ratio is plotted on the horizontal axis and the bit error rate is plotted on the vertical axis. It was found that the carrier-to-noise power ratio C / N was improved by about 1 dB in any case when compared with the reception quality of the prior art at 10 ⁻⁴ points.

After the training by the successive least squares algorithm is completed, the amount of computation remains the same as that of the conventional technique, the algorithm and the configuration are switched, retraining is performed M times (M ≧ 2 ), and then the data section is processed. Thus, even with a small number of training symbols, the equalizer can be sufficiently pulled in and the reception quality is improved.

[0038]

As described above, according to the present invention,
When the number of training symbols of the diversity receiver to which the equalizer is applied is reduced, it is possible to improve the reception quality without significantly increasing the calculation amount. By this means, diversity reception with high reception quality can be performed, and reception quality can be improved without increasing the number of training symbols.

[Brief description of drawings]

FIG. 1 is a block diagram of a diversity receiver showing an example of a two-branch configuration when a decision feedback equalizer according to an embodiment of the present invention is used.

FIG. 2 is a time chart illustrating the order of operations according to the embodiment of the present invention.

FIG. 3 is a diagram showing an internal configuration of a reception signal memory and a switch state in a retraining section.

FIG. 4 is a diagram showing the effect of the embodiment of the present invention.

FIG. 5 shows the prior art 2 when a decision feedback equalizer is used.
The block block diagram of the diversity receiver which shows the example of a branch structure.

FIG. 6 is a time chart illustrating the order of operations in a conventional example.

[Explanation of symbols]

10, 20 Feed forward filter 11-14, 21-24 Delay circuit 15-19, 25-29 taps 30 training signal memory 31 initialization unit 32-35 tap coefficient estimator 40-44, SW1, SW2 switches 45,46 Received signal memory 50 output terminals 61,62 Input terminal 70-74 adder 80, 81 Feedback filter M memory

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04B ⁷ /02-7/12 H04L 1/02-1/06 H04B 7/005

Claims (6)

(57) [Claims] 1. A training section and a data section of a received signal
Equipped with an equalizer that switches the configuration and algorithm between and
e, Received signal that accumulates received signals and enables multiple readings
No. storage means, Receiving using the first equalizer configuration and the first algorithm
Update process of tap coefficient in training section of signal
The first means of doing When the processing by the first means completes, the equalizer structure
The second equalizer configuration and the second
A second means of switching to an algorithm, Receiving the training section stored in the received signal storage means
The second equalizer configuration and the second
Using the algorithmReUpdate the tap coefficient
Third means, This third meansOf processing byAfter completion, the second equalizer
Configuration and data interval using the second algorithm
And a fourth means for updating the tap coefficient ofDa
In the diversity receiver, The third means is for the received signal in the toning section.
Repeat tap coefficient update process multiple times That
Characteristic diversity receiver. 2. N pieces corresponding to inputs from N demodulators
Delay line filter with tap, Considering this as N individual filters, tap coefficients are set.
A first tap coefficient estimator that determines The N tapped delay line filters are combined into one filter.
Second tap coefficient estimation to set the tap coefficient
A vessel, Estimate the output of the first tap coefficient estimator to the second tap coefficient estimator
To the output of the vessel, and the switched output is
Switching coefficient of tap coefficient for tapped delay line filter
Dan, Converts the output of the first tap coefficient estimator when switching
ShiThe tap coefficient of the delay line filter with N taps
Do weightingInitial value of the second tap coefficient estimator
Initialization means to set asWhen, Received signals from N demodulators are accumulated and the output is
Received signal connected to N tapped delay line filter
Memory and Pre-store the signals received in the training section
Training signal memory, Add the outputs of the N tapped delay line filters
An adder, A code that identifies the output of this adder and generates a demodulated signal sequence.
No. discriminator, Output of the training signal memory and output of the code discriminator
Force and input, select one of the above and the first and second
The training signal supplied to the tap coefficient estimator of
And means, As an initial training, it is stored in the received signal memory
For the received signal in the training section,
Each output of the delay line filter with a clock
Input to the number estimator, The training signal switching means is the training signal.
Supply the output of the No. memory to the first tap coefficient estimator
Then The received signal memory is subsequently retrained.
The received signal of the training section accumulated in
The output of the adder is input to the second tap coefficient estimator.
And The training signal switching means is the training signal.
No. memory output to the second tap coefficient estimator, As the processing of the data section, In the received signal of the data section accumulated in the received signal memory
In contrast, the output of the adder is the second tap coefficient estimate
Input into the vessel, The training signal switching means outputs the code discriminator.
Force to the second tap coefficient estimator for processingDa
In the diversity receiver, Repeat the retraining multiple times Special
Diversity receiver to collect. 3. The initialization means uses a weighting coefficient calculated from a tap coefficient immediately before switching from the first to second tap coefficient estimator and a received signal in a training section accumulated in the received signal memory. 3. The diversity receiver according to claim 2, further comprising means for weighting the tap coefficients of the N tapped delay line filters. 4. The diversity receiver of claim 2, wherein the first tap coefficient estimator operates using a recursive least squares algorithm and the second tap coefficient estimator operates using a least mean squares algorithm. . 5. The diversity receiver according to claim 2, wherein the least-squares-means algorithm is a normalized least-squares-means algorithm in which a step size in updating a tap coefficient is normalized by an average power of a received signal. 6. A training section and a data section of a received signal
Equipped with an equalizer that switches the configuration and algorithm between and
In the diversity reception method, the received signal is
The first equalizer configuration and the first algorithm stored in memory
The rhythm is used to
UP coefficient update processing (hereinafter referred to as training processing)
When the training process is completed,
And the algorithm to the second equalizer configuration and the second
Switch to the algorithm and store the received signal in the memory.
For the received signal in the training section, the second equal
And the second algorithm is used.TeUp
Coefficient update processing (hereinafter referred to as retraining processing)
After completing this retraining process,
Data using the rectifier configuration and the second algorithm
Update the tap coefficient of the sectionHow to receive diversity
In law, Repeat the retraining multiple times Special
Diversity reception method to collect.