JPH11177471A - Adaptive equalizer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an adaptive equalizer that avoids the occurrence of a frame erasure resulting from non-termination of a sequential series estimate algorithm, even when a pattern which is incapable of being undecoded uniquely in an impulse response of a channel appears, or resulting from occurrence of a buffer overflow. SOLUTION: A sampler 2 samples each element of an array antenna 1 to obtain a snapshot vector 3, an impulse response estimate value 5 of a channel of each antenna element is obtained by the snapshot vector, and an update circuit 8 uses the estimated value of each channel, a fano metric value 9 before updating an object series and an object series length 10 in order to update the fano metric value of the object series. A circuit 12 receives the vector 3 and estimates the series sequentially, based on the likelihood comparison using the fano metric through a sequential decoding algorithm, and outputs the result as a transmission series.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is applied to, for example, an adaptive equalizer that obtains a transmission sequence from a reception output of an array antenna in mobile communication. The present invention relates to an adaptive equalizer that can perform the processing efficiently.

[0002]

2. Description of the Related Art A feature of mobile communication is that a radio propagation environment is a multipath propagation path. Considering the upstream (mobile station transmission, base station reception) communication path, a bundle of transmission element waves affected by scattering, diffraction, and reflection around the mobile station is directly or after being reflected at a further distance, Arrives at the base station. Therefore, the base station receives the transmission signal on a path divided into a plurality of components having different arrival angles. These components corresponding to each path are composed of the bundle of the above-mentioned transmission element waves, and the process of creating this bundle (scattering, diffraction, reflection, etc.).
Is different for each path, so that each path is subject to independent fading.

[0003] When communication is performed in such a mobile communication environment, different phenomena appear in the received signal depending on the signal bandwidth. If the bit rate of the transmitted signal is low and its bandwidth is sufficiently smaller than the coherence bandwidth of the channel, the propagation delay difference of the signal received via each of the above paths is the symbol time length of the signal (usually the symbol rate Smaller than the reciprocal). In this case, on the receiving side, the same information symbol is received on each path, and no waveform distortion occurs in the received signal due to intersymbol interference.

As the bit rate of the transmitted signal becomes progressively higher and the bandwidth of the signal becomes approximately equal to the coherence bandwidth of the channel, different information symbols are received on each path. In this case, waveform distortion occurs in the received signal due to intersymbol interference of several symbols before and after. As described above, since each path is subjected to independent fading, the intersymbol interference of several symbols before and after this becomes time-varying intersymbol interference that changes with time. Therefore, the equalizer for removing such intersymbol interference has a function of estimating the impulse response of a channel (equal to the complex amplitude of each path arranged at its arrival time) and estimating a transmission sequence. Is required. Maximum likelihood sequence estimation (MLS
E: An algorithm based on the Maximum Likelihood Sequence Estimation theory can be applied. If the intersymbol interference is several symbols before and after, signal processing combining this channel estimation and MLSE can be realized with a realistic processing amount. Detail is,
For example, references: Fukawa and Suzuki, "Sequential Least Square Form Maximum Likelihood Sequence Estimation (RLS-MLSE)-Application of Maximum Likelihood Estimation Theory to Mobile Radio-", IEICE (B-II), J76-B-II, No.
4, pp. 202-214, Apr. 1993.

[0005] As the bit rate of the transmitted signal is further increased, the received signal is subject to time-varying intersymbol interference of more symbols. Theoretically, the above-described combined signal processing of channel estimation and MLSE can be applied to the equalization of the intersymbol interference. However, the number of states of the Viterbi algorithm used in MLSE increases exponentially with the length of intersymbol interference (equal to the length of storage of the channel). For example, when binary phase modulation (BPSK) is used as the modulation method, the number of states (corresponding to the 12-path propagation path) becomes 2048 when 11 symbols are stored in the channel, which exceeds the actual processing amount. Would.

On the other hand, in high-speed transmission, fading fluctuation normalized by the symbol length becomes relatively slow. For example, when transmitting a signal of 10 Msymbol / sec (= 1 / T, T = symbol length), the maximum Doppler frequency of fading f
_D is 100Hz (about 150k at 1GHz carrier frequency)
m / sec), the maximum Doppler frequency f _D T standardized by the symbol length is 10 ^−5.
Becomes Therefore, even if a frame is composed of several hundred symbols, it can be considered that there is no fading fluctuation within one frame. Therefore, it is possible to perform channel estimation at the head of the frame (for this reason, a known training sequence is transmitted at the head of the frame in both transmission and reception), and thereafter, the estimated value can be continuously used (no updating is performed). As a result, the processing amount is reduced.

As can be seen from the above considerations, it can be said that the technical problem in configuring an adaptive equalizer in high-speed transmission on the order of megabits lies in the sequence estimation unit. In order to avoid this problem, an adaptive equalizer that uses successive sequence estimation instead of the maximum likelihood sequence estimation MLSE is described in Matsumoto, "Sequential decoding equalization in high-speed mobile communication", 1997 IEICE Society Conference B -5-149. Sequential sequence estimation is a method originally developed as a decoding algorithm for convolutional codes with a long storage length.
There has been a history of studies on codes having a storage length of 40 to 60 for application and purpose to deep space communication. Since the process in which intersymbol interference occurs is nothing less than convolution on a complex number field, it is clear that there is affinity for adaptive equalization. The above literature indicates that successive sequences are estimated for BPSK in a 12-path channel, and as a result, good characteristics can be obtained with a realistic processing amount. However, at the same time, in the sequential sequence estimation, a phenomenon that the equalization processing of one frame does not end within a predetermined time (or a phenomenon that a buffer prepared for the processing overflows) rarely occurs. At this time, the equalization result can be output. It is also shown that the disappearance (referred to as frame erasure) greatly affects the overall characteristics.

[0008] This phenomenon mainly appears when the impulse response of the channel is a pattern that cannot be uniquely decoded or a pattern close thereto. For example, consider a case where the complex amplitudes of the preceding wave and the delayed wave are each 1 in a two-path propagation path. At this time, if the information symbols of the preceding wave and the delayed wave are each 1, the received signal is 2, and the receiving side knows that the information symbol is 1 for both. On the other hand, the information symbols of the preceding wave and the delayed wave are-
If it is 1, the received signal is -2. In this case as well, it can be seen that both sides have the information symbol of -1 on the receiving side.
However, the information symbols of the preceding wave and the delayed wave are each 1
If -1 and -1, or vice versa, the received signal is 0 in both cases, and the receiving side cannot determine which pattern has been transmitted. Such an impulse response is called a pattern that cannot be uniquely decoded. In the sequential sequence estimation, when a pattern that cannot be uniquely decoded appears in the impulse response, the probability of many sequence candidates becomes equal or close, and the equalization process is continued for all of them. For this reason, the above-described equalization processing does not end, or a phenomenon that the buffer overflows occurs.

[0009]

SUMMARY OF THE INVENTION An object of the present invention is to provide a frame erasure system in which the sequential sequence estimation algorithm does not end or a buffer overflow occurs even if a pattern that cannot be uniquely decoded appears in the impulse response of the channel. An object of the present invention is to provide an adaptive equalizer capable of avoiding the following phenomenon.

[0010]

According to the present invention, by receiving a signal with an array antenna, the impulse response of a channel observed by each antenna element is prevented from being a pattern that cannot be uniquely decoded. The principle will be described below. Each path has a high probability that not only the delay time is different but also the arrival direction is different. The phase difference of the received signal at each element of the array antenna depends on the propagation distance difference between the antenna elements, which is uniquely determined given the spatial arrangement of the antenna elements and the angle of incidence of the signal. If the arrival directions of the respective paths are different, the phase difference between the elements of the antenna of the signal passing through the respective paths should be different for each path. For example, if the phase of the preceding wave received by the four-element array antenna changes in the order of π / 4 → π / 2 → 3π / 4 → π in the order of the elements, the delayed wave arrives from the same direction as the preceding wave. Otherwise, the phase will not change in the same way. This means that even if the impulse response of the channel observed by a certain antenna element is a pattern that cannot be uniquely decoded, there is a high probability that the impulse response will not be obtained by another element.

The present invention is based on this principle. The output of each element of the array antenna composed of a plurality of antenna elements is sampled by the sampling means and a snapshot vector is output. From the snapshot vector, the impulse response of the channel in each antenna element is calculated by the impulse response estimation means by the least square method. Using the impulse response estimation values of these channels and the snapshot vector, obtain a fanometric by a sequential sequence estimating means, and sequentially perform a sequence estimation based on a likelihood comparison using the fanometric. , And output as a transmission sequence.

[0012]

FIG. 1A shows an example of the configuration of an adaptive equalizer according to the present invention. The array antenna 1 is composed of M elements. The output is sampled by the sampler 2. The sampler output 3 is called a snapshot vector. The snapshot vector is input to the impulse response estimator 4, and the impulse response is estimated. The method of least squares can be used to estimate the impulse response (for example, Simon Haykin,
“Adaptive Filter Theory”, Prentice-Hall, pp. 381-4
50,1986). Impulse response estimated value output 5
Is a matrix of M × N (the number of array elements × the number of paths) when estimating the impulse response of the N path. As described above, in high-speed transmission, channel fluctuations can be ignored within one frame, so that channel estimation is performed only at the beginning of a frame, and the result can be used over the entire frame. For this purpose, a training sequence known in both transmission and reception is transmitted at the beginning of the frame.

FIG. 1A shows a configuration example of the sequential sequence estimator 6 used in the present invention. The sequential sequence estimator 6 receives the snapshot vector 3 and the impulse response estimated value output 5 as inputs. In the process of successive sequence estimation, it is necessary to compare the likelihood of candidate sequences having different lengths. Using a phanometric as a value representing this certainty (for example, the theoretical background and the calculation method of the phantom are described in Imai, Iwatari, Miyagawa, "Coding Theory" Shokodo, pp.355-367,
1973). The fanometric update circuit 8 updates the fanometric of the candidate sequence using the sampler output 3 and the impulse response estimated value output 5, the fanometric value 9 before updating the candidate sequence, and the length information 10 of the candidate sequence. The update output 11 is input to a sequential decoding algorithm execution circuit 12. As described above, the sequential sequence estimation has been developed as a decoding algorithm for a convolutional code having a long storage length. From the viewpoint of decoding the convolutional code, the sequential sequence estimator 6 is a “decoder”, and in the present invention, the “decoding” and the algorithm are used as they are. Therefore, the sequential decoding algorithm is called a sequential "decoding" algorithm. As the sequential decoding algorithm, a Fano algorithm, a stack algorithm, and the like are known. These require updating the metric of the candidate sequence during the execution of these algorithms. This processing is performed by the fanometric update circuit 8. The pre-updated phanometric value 9 and the length information 10 of the candidate sequence required for this
Is sequentially stored in the memory, and these are read out and output to the phantom updating circuit 8 when necessary.

[0014]

As described above, according to the present invention, the fanometrics used in the sequential sequence estimation are obtained from the estimated values of the channel impulse responses of all M antenna elements. Even if the impulse response of the channel becomes a pattern that cannot be uniquely decoded, the probability that this will not be the case with other elements can be increased. As a result, it is possible to avoid a phenomenon that the sequential sequence estimation algorithm does not end within a predetermined time or a buffer overflow occurs.

[Brief description of the drawings]

FIG. 1A is a block diagram illustrating a functional configuration example of an adaptive equalizer according to the present invention, and FIG. 1B is a block diagram illustrating a functional configuration example of a sequential sequence estimator 6;

Claims (1)

[Claims] 1. A means for obtaining a snapshot vector by sampling the output of each element of an array antenna composed of a plurality of antenna elements, and using the snapshot vector as an input, the impulse of the channel in each element of the array antenna is obtained by the least square method. Means for estimating a response, means for obtaining a fanometric using the impulse response estimation value of the channel in each element of the array antenna and the snapshot vector, and sequence estimation based on likelihood comparison using the fanometric And a means for performing successive sequence estimation as an estimated output of the transmission sequence.