JP2000091928A - Receiver - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a receiver at a low level bit error rate BER by realizing provision of an adaptive equalizer where the divergence of an equalizer is avoided, based on information of maximum likelihood decoding (Viterbi decoding). SOLUTION: The receiver having a discrimination-decoder 114 that decides a discrimination point in response to a state of a convolution coder at a transmitter side, applies maximum likelihood decoding to a received signal by using an error signal at a receiving point and the discrimination point to decode the received signal, is provided with a survival path monitor section 120 that traces back a trellis used for maximum likelihood decoding of the received signal by a cut-off path length to monitor whether or not the survival paths is concentrated onto one path and with an adaptive equalizer control section 122 that controls an adaptive equalizer 110 correcting distortion in the received signal due to a frequency of a transmission line by periodical storage of a tap gain or reading of the stored tap gain, based on the result of monitor by the survival path monitor section 120.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is used in a facsimile apparatus and the like to demodulate a modulated signal and perform maximum likelihood decoding when the demodulated data sequence is a pre-convolutionally encoded data sequence. The present invention relates to an adaptive automatic equalizer that prevents divergence of an equalizer from information of maximum likelihood decoding (Viterbi decoding), and relates to a receiver that reduces BER (bit error rate).

[0002]

2. Description of the Related Art Conventionally, in a facsimile machine or the like,
In the case where transmission is performed within a limited band of 0.3 to 3.4 kHz and good transmission quality is obtained at a transmission rate exceeding 9.6 kbps, modems (modulators / demodulators) have recently used convolutional coding. As a decoding method, an error correction technique of Viterbi decoding has been introduced.

[0003] This is because, in recent years, it has become essential to allocate a large number of bits to one symbol in accordance with a demand for high-speed modems. For example, a reference technical document “2.1 Application of DPS to a modem” by the author “Yawata”, Information Processing Vol. 30 No. 1989 11
Month, pp. 1315 to 1323, it is common to use a trellis coded modulation method and a Viterbi decoding method.

FIG. 2 is a block diagram showing a schematic configuration of a modem to which a Viterbi decoding method is applied as a convolutional coding method and its decoding method. Referring to FIG. 2, a transmitting modem 1 and a demodulator 5 provided with a scrambler 8, a trellis encoder 2, and a modulator 3, an equalizer 6, a decision / decoder 7, The receiving modem 4 is provided with a descrambler 9.

Here, the scrambler 8 of the transmitting modem 1
Has a function of randomizing transmission data in order to facilitate bit synchronization extraction and automatic equalization in the receiving modem 4.

The trellis encoder 2 of the transmitting modem 1 has, for example, a configuration as shown in FIG. That is, by the contents of three registers F2, F1, and F0 of the trellis encoder (convolutional encoder) shown in FIG.
The trellis encoder 2 has eight states S0 to S7, and two inputs Y of the trellis encoder 2.
There are four input states in combination of 2n and Y1n.

FIG. 4 shows a state transition diagram of the trellis encoder 2 depending on the input state. This state transition diagram is called a trellis diagram. In this trellis diagram, the trellis encoder 2
Of the three bits output from, two bits are Y2n, Y1
n, Y0n is coded to “0” if it is in states S0 to S3 in the trellis diagram, and S4 to S
If it is at 7, it is coded to "1". As is clear from this trellis diagram, there are two paths that exit the same state and enter the same state between two separate times, and at three separate times, the overlap of the intermediate paths occurs. I chose four so as not to have one.

[0008] This trellis coding is based on V. This is a type of error correction code using a convolutional code, such as a 17-modem, which performs transmission more efficiently than a band, and adds redundant bits to transmission data. Therefore, for example, In the transmission of 14,400 b / s by the 17 modem, it should be possible to transmit at 2,400 baud at 2 ⁶ = 64 values, but when trellis coding is used, 12
By transmitting in eight values, the number of signal points increases.

As described above, by using trellis coding, the number of signal points increases, but the transition between signals is restricted. By utilizing this and performing Viterbi decoding on the receiving side as described later, the SN ratio versus error rate characteristics can be improved.

The modulator 3 modulates two carriers whose phases differ by 90 ° obtained by mapping the output from the trellis coding 2 to each signal point on a phase plane, and modulates the two carriers. Combining two signals and PSK
(Phase shift keying) or QAM (quadrature amplitude)
It generates a modulation waveform of “tude modulation (quadrature amplitude modulation)”.

The demodulator 5 of the receiving modem 4 is configured to demodulate a transmission signal transmitted from the transmitting modem 1 via the transmission line 10. In this case, the high-speed modem first has a fixed frequency oscillator close to the carrier frequency in the receiving modem 4, and demodulates the received signal at this frequency. This is called quasi-synchronous detection.

In the method used for PSK and QAM, two demodulated signals are obtained because demodulation is performed with two reproduced carriers (or fixed frequency signals) whose phases are shifted by 90 °.

The determination / decoder 7 of the receiving modem 4
Uses the trellis encoder 2 as the encoder of the transmitting modem 1, and performs Viterbi decoding as the maximum likelihood decoding for this. In Viterbi decoding, features of a trellis diagram as shown in FIG. 4 are used.
That is, for the received signal sequence R, each likelihood P (R | Si) when the codeword sequence is Si is obtained, and the sequence Sk having the maximum likelihood is subjected to maximum likelihood estimation decoding as a transmitted signal. Things.

The code word sequence is one-to-one of the paths of the trellis diagram.
, It corresponds to finding the path with the highest likelihood. Therefore, a path having a high likelihood can be sequentially found using the trellis diagram. That is, when there are a plurality of paths starting from the same state and entering the same state, the likelihood of each path is calculated, and the path having the highest likelihood is obtained. Then, the decoded word by this path is set as the transmission codeword corresponding to that part, and the other paths are discarded. In this way, the number of surviving paths (remaining paths) is always limited to eight. As described above, when unnecessary paths are discarded and decoding is advanced while leaving only surviving paths, all surviving paths (remaining paths) start from one point. then,
The code before the one point is subjected to maximum likelihood decoding.

The above is the principle of the Viterbi decoding method. In a modem, one assumption is made for obtaining the likelihood of a path. In other words, this assumption is affected by inter-symbol interference (residual inter-symbol interference after equalization and noise) before the signal point on the transmitting side is sent and reaches the decision circuit on the receiving side. However, this effect (the difference between the transmission signal point and the reception point) follows a Gaussian distribution, and this assumption is a valid assumption.

FIG. Some of the signal points of 17 modems are shown. If a signal point whose lower 3 bits are (a2, a1, a0) is transmitted when the signal of the reception point R is received, a subpoint of the signal point whose lower 3 bits are (a2, a1, a0) • The closest signal point S (R: a
2, a1, a0) is calculated, and a distance r (a2, a1, a0) from R is calculated as a candidate for a transmission signal point.

In the example shown in FIG. 5, (a2, a1, a0)
It can be seen that the signal point S for the sub-set of = (0,0,0) is (0011000), and similarly the signal point S for the sub-set of (0,1,0) is (0101010). The respective signal points S for the eight sub-sets are obtained, and the distance r (≡ri) from the signal points S
Ask for. Since r (≡ri) follows a Gaussian distribution,
This probability density function is given by the following equation (1).

P (r) ∝exp ((− r ² ) / (2 × σ ² )) (1)

Therefore, r (≡ri) is used as the metric of the path between the states, and the evaluation value of the likelihood of a certain partial path can be viewed from the total of the metrics of the path. Of course, the lower the value, the higher the likelihood. Thereby, decoding by the maximum likelihood estimation can be advanced.

FIG. 6 is a block diagram showing the configuration of a general Viterbi decoder that decodes a convolutionally coded received signal sequence using the Viterbi algorithm. In general, a Viterbi decoder adds a stored path metric and a branch metric obtained from an input data sequence in some combinations, compares the added results, selects a new path metric, and selects a stored path metric. The most probable data is decoded by repeating the process of updating the contents of the path metric being used to the contents of the selected new path metric.

Note that a branch refers to a decoding path from one state to the next state, and a path refers to a series of decoding paths formed by a series of branches. Further, the branch metric and the path metric are data obtained from the above-mentioned branch and path by a certain calculation formula, respectively.

In FIG. 6, reference numeral 11 denotes input information input to a later-described branch metric calculator of the Viterbi decoder;
Reference numeral 12 denotes a branch metric calculation unit for calculating a branch metric from the input information 11, 13 denotes a branch metric calculated by the branch metric calculation unit 12, 15
a is the stored path metric described above, and 15b is the new path metric described above.

An addition / comparison / selection unit (ACS) 16 selects a surviving path and a new path metric 15b.
), 14 are the surviving path and the new path metric 15
b, a path metric memory for storing b, remaining path information 17 indicating the remaining path selected by the addition / comparison / selection section 16, a path memory 18 for storing the remaining path information 17, 1
Reference numeral 9 denotes a plurality of remaining path information 1 stored in the path memory 18.
7 is a path selection unit that selects the most likely state signal (most likely path) which is the most likely data.

FIG. 7 is a flowchart showing the processing operation of the Viterbi decoder of FIG. Referring to FIG. 7, in the Viterbi decoder, in the branch metric calculation unit 12,
First, calculate the received signal (input information) 11 and the receiver square of the distance between the signal points of baseband signal within the constellation demodulating unit (branch metric) r ² (S701). That is,
A branch metric is obtained for all branches from each state at a certain point in the input information 11 to each state at the next point (each state in trellis coding).

When the branch metric is calculated in step S701 in this manner, the addition / comparison / selection unit (ACS
In section 16, all the branch metrics 1
3 and the path metric 15a in each state stored in the path metric memory 14 are added in various combinations (S702). Subsequently, an addition / comparison / selection unit (AC
(S unit) 16 compares the magnitude relation of the path metrics for each state, which is the result of the addition, and detects the combination of the branch metric and the path metric giving the maximum value as the surviving path information (S703).

The addition / comparison / selection unit (ACS unit) 16 stores the remaining path information 17 for each state (each state in trellis coding) detected as described above in the path memory 18 and also stores the detected remaining path information. Of the various path metrics 15a stored in the path metric memory 14 based on the
Is selected for each state, and the selected new path metric 1
The contents of the path metric memory 14 are updated by 5a (S704).

After repeating such a loop process several times, the path selecting unit 19 sets the path memory 1
The remaining path information 17 giving the maximum value is selected as the maximum likelihood state signal (maximum likelihood path) from the remaining path information 17 corresponding to the number of states of the input information 11 stored in 8 and output (S70).
5). That is, in the path selection unit 19, the path memory 18
The maximum likelihood path is selected from the paths stored in the path, and the first branch of the path, that is, the signal point is output as a decision point.

This Viterbi algorithm assumes the above-described transmission path model as a theoretical background. That is, when the transmitter 1 transmits the signal point S,
Is assumed to superimpose Gaussian white noise.
In this case, the probability P (R | S) that the receiving point is R is represented by the following equation (2).

P (R | S) = A × exp ((− r ² ) / (2 × σ ² )) (2)

Here, r (≡ri) is the distance r = RS between the signal point and the receiving point, σ ² is the noise variance, and A is a constant for normalization.

On the other hand, the probability that the transmission signal point is S when R is received is expressed by the following equation (3) according to Bayes' theorem.

P (R | S) = (P (S) / P (R)) × P (R / S) (3)

Here, it is assumed that the occurrence probabilities of the signals are uniform (P (Si) = P (Sj) for the signals Si and Sj.
Is always satisfied), and the following equation (4) is obtained by substituting equation (2), which is a model equation of the transmission path, into equation (3).

P (R | S) ∝exp ((− r ² ) / (2 × σ ² )) (4)

From the above equation (4) (that is, from the above equation (1)), a signal having a small r ² is a signal having the maximum posterior probability, that is, the signal S having the maximum likelihood. As aforementioned,
In a normal Viterbi algorithm, a signal sequence (path) is estimated using the above equation (4) (equation (1)). When the received signal sequence (R0, R1,..., Rn) is received, the probability that it is a signal sequence (S0, S1,..., Sn) is represented by the following equation (5).

[0036] P (S0, S1, ··, Sn | R0, R1, ··, Rn) = Πn, i P (Si | Ri) αexp (Σn, i ((- r 2) / (2 × σ ² )) ・ ・ ・ (5)

In the Viterbi algorithm, a path having the shortest total distance between the receiving point and the signal point is selected. This path is a path that maximizes the posterior probability of the above equation (5).

The equalizer 6 removes intersymbol interference caused by transmission waveform distortion, and has a basic configuration of a transversal filter having a variable tap gain. In PSK and QAM, two systems of demodulated signals are obtained, and intersymbol interference occurs between the two systems, so that the effect is eliminated.

The demodulator 5 plays a role of converting a modulated signal coming from a line into a complex baseband signal. Subsequently, the complex baseband signals are sequentially input to the equalizer 6 forming a transversal filter.

[0040]

However, in the conventional receiving apparatus as described above, tap gain updating is performed when tap interruption or sudden level fluctuation occurs on the line. If the equalization process is adaptively continued according to the algorithm, the tap gain will be inadvertently greatly changed because the equalization error takes an extremely large value.

As a result, the equalizing capability of the adaptive automatic equalizer rapidly deteriorates, and if the determination error occurs frequently or the degree of the line failure is bad, the equalizer diverges.
Degree and equalization ability may not recover.

The present invention has been made in view of the above, and provides an adaptive equalizer capable of avoiding divergence of an equalizer from information of maximum likelihood decoding (Viterbi decoding). ,
An object of the present invention is to provide a low-level BER (bit error rate) receiving apparatus.

[0043]

In order to achieve the above object, in a receiving apparatus according to the present invention, a decision point according to a state of a convolutional encoder on a transmitting apparatus side is determined, and a receiving point is determined. In a receiving apparatus having maximum likelihood decoding means for performing maximum likelihood decoding using an error signal between a point and a determination point and decoding a received signal, a trellis used for the maximum likelihood decoding of the received signal is traced back by a path truncation length, Surviving path monitoring means for monitoring whether the number of surviving paths is reduced to one, and an adaptive equalizer for correcting distortion received by the received signal at a transmission line frequency based on the monitoring result of the surviving path monitoring means. Adaptive equalizer control means for controlling by periodically storing tap gains or reading stored tap gain values;
It is provided with.

According to a second aspect of the present invention, in the first aspect, the maximum likelihood decoding means uses Viterbi decoding.

[0045]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a receiver according to the present invention will be described in detail with reference to the accompanying drawings.

The present invention realizes an adaptive automatic equalizer that can prevent the divergence of the equalizer from the information of the maximum likelihood decoding (Viterbi decoding) even when the line condition is poor. Error rate (error rate) can be reduced.

That is, as described above, in Viterbi decoding, even if the trellis is traced back along the surviving path (remaining path) for a certain period (path cutoff length), the number of surviving paths is not reduced to one. Therefore, using this information, when the number of surviving paths is narrowed down to one, it is determined that tap gain updating is performed correctly, and the value is periodically saved, and the number of surviving paths is reduced to one. When the situation is not narrowed down, the tap gain is prevented from being carelessly updated, and the adaptive equalizer is executed by using the previously stored tap gain value. Hereinafter, the configuration and operation will be described with a specific example.

(Configuration of Receiving Apparatus (Modem)) FIG. 1 is a block diagram showing a configuration of a receiving apparatus (modem) according to an embodiment of the present invention. In the figure, reference numeral 100 denotes an A / D converter for converting a received signal received from a transmission path (not shown) into a digital signal, and reference numeral 102 denotes a complex for subjecting a signal output from the A / D converter 100 to complex processing. Processing unit, 1
Reference numeral 04 denotes a baseband filter (BBF) that removes unnecessary high-frequency components and the like included in a received signal received from the transmission path and removes intersymbol interference.

Reference numeral 106 denotes a first-order IIR (infinite impulse response) for correcting the loss of the subscriber line.
(esponse: infinite impulse response) A fixed equalizer composed of a filter, 108 is an automatic gain control unit (AGC) for automatically adjusting the signal level, and 110 is an adaptive circuit for compensating for the distortion of the received signal due to the frequency of the transmission line. The equalizer 112 is a phase synchronization unit that compensates for a frequency shift and a phase fluctuation included in a signal output from the adaptive equalizer 110.

A convolutional encoder 114 (for example, the above-described convolutional encoder on the transmitting apparatus side) uses reception points on a signal plane (signal point arrangement) of a quadratic plane having two axes of a real number axis and an imaginary number axis. A decision point according to the state of the trellis encoder 2) is determined, Viterbi decoding (maximum likelihood decoding) is performed using an error (error signal e) between the reception point and the decision point, and the signal is decoded (maximum likelihood decoding). ) Is a decision / decoder (maximum likelihood decoding unit) as maximum likelihood decoding means.

Reference numeral 116 denotes a descrambler for returning a signal randomized by a scrambler (for example, the scrambler 8 similar to that shown in FIG. 2) to a signal SGN.
18 is an error calculator for calculating an equalizer error signal e, 120
Is determined by the decision / decoder (maximum likelihood decoding unit) 114 by traversing the trellis for a certain period (path truncation length) and the number of surviving paths is 1
A surviving path monitoring unit as surviving path monitoring means for monitoring whether or not the book is focused on is a surviving path monitoring unit 1.
An adaptive equalizer control unit serving as an adaptive equalizer control unit that controls the adaptive equalizer 110 by periodically storing or reading tap gains based on the monitoring result of 20.

Here, the adaptive equalizer 110 is provided with a transversal type FIR (Finite Impulse RE).
(sponse: finite impulse response) filter is used, and its filter tap coefficient Ci is calculated by an error calculation unit 118.
Is updated as shown in the following equation (6) based on the error signal e from

Ci = Ci−μ · e · xi ^* (6)

In the above equation (6), xi ^* is a conjugate complex number of the signal xi input to the adaptive equalizer 110, and μ is a coefficient (convergence coefficient) for adjusting the convergence speed.

(Operation of Receiving Apparatus (Modem)) Next, the characteristic operation of the receiving apparatus (modem) configured as described above will be described. The surviving path monitoring unit 120 is adapted to respond to an instantaneous interruption or a sudden level change in the line.
In Viterbi decoding, even if the trellis is traced back along the surviving path (surviving path) for a certain period (path truncation length),
Monitor the situation where the number of surviving paths is not limited to one.

The adaptive equalizer control unit 122 determines that the state reduced to one is a normal state based on the monitoring result by the surviving path monitoring unit 120, and periodically saves the tap gain value at that time. I do. In addition, it is determined that the case where the number is not limited to one is not normal, the update of the tap gain is stopped, the stored tap gain value is re-read,
Perform an adaptive equalizer.

The surviving path monitoring unit 120 performs the recovery from the momentary interruption and the AGC following the level fluctuation,
It monitors the state where the surviving paths are reduced to one. Then, the adaptive equalizer control unit 122 outputs the surviving path monitoring unit 1
Based on the monitoring result by the control unit 20, control is performed to restart the update of the adaptive equalizer tap gain.

Therefore, as described above, even if an instantaneous interruption or a sudden level change occurs in the line, the surviving path cannot be reduced to one in the above-described receiving apparatus. Upon detecting this, the equalizer keeps the frozen state, and normal equalization is performed only when the number of surviving paths is reduced to one by recovery from instantaneous interruption or following AGC for level fluctuation. Since the processing is executed, the divergence of the equalizer can be avoided.

[0059]

As described above, according to the receiving apparatus according to the present invention (claims 1 and 2), the state of the surviving path which is traced back by the trellis length used for maximum likelihood decoding (Viterbi decoding) of data is determined. When the tap gain in the state where the data decoding is correctly performed is periodically saved, the error occurs frequently in the data decoding and the tap gain of the adaptive equalizer is not correctly updated. , Execute the loading of the stored tap gain. As described above, since the normal equalization processing is executed only when the number of surviving paths is reduced to one, the adaptive automatic equalizer that can prevent the divergence of the equalizer can be prevented. Offer is realized,
BER (bit error rate) can be reduced.

[Brief description of the drawings]

FIG. 1 is a receiving apparatus (modem) according to an embodiment of the present invention.
FIG. 3 is a block diagram showing the configuration of FIG.

FIG. 2 is a block diagram showing a schematic configuration of a conventional modem to which a Viterbi decoding method is applied as a conventional convolutional coding and its decoding method.

FIG. 3 is a block diagram illustrating a configuration example of a trellis encoder of a transmission-side modem in FIG. 2;

FIG. 4 is a trellis diagram showing a state transition of the trellis encoder in FIG. 3;

FIG. It is explanatory drawing which shows some signal points of 17 modems.

FIG. 6 is a block diagram showing a configuration of a general Viterbi decoder that decodes a convolutionally encoded received signal sequence using a Viterbi algorithm.

FIG. 7 is a flowchart showing a processing operation of a Viterbi decoder in FIG. 6;

[Explanation of symbols]

 Reference Signs List 100 A / D converter 102 Complexization processing unit 104 Baseband filter (BBF) 106 Fixed equalizer 108 Automatic gain control unit (AGC) 110 Adaptive equalizer 112 Phase synchronization unit 114 Judgment / demultiplexer 116 Descrambler 118 Error calculation unit 120 Surviving path monitoring unit 122 Adaptive equalizer control unit

Claims (2)

[Claims] 1. A maximum likelihood decoding for determining a decision point according to a state of a convolutional encoder on a transmitting apparatus side, performing maximum likelihood decoding using an error signal between a reception point and a decision point, and decoding a reception signal. A surviving path monitoring means for monitoring whether a trellis used for the maximum likelihood decoding of the received signal is traced back to a path truncation length and narrowing down a surviving path to one, and monitoring of the surviving path monitoring means. An adaptive equalizer that controls an adaptive equalizer that corrects distortion of the received signal at a frequency of a transmission line based on a result by periodically storing tap gains or reading a stored tap gain value. A receiving device comprising: a control unit. 2. The receiving apparatus according to claim 1, wherein the maximum likelihood decoding means uses Viterbi decoding.