JP2855717B2 - Intersymbol interference canceller - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an intersymbol interference canceller, and more particularly to an intersymbol interference canceller to which a Viterbi algorithm is applied.

[Conventional technology]

When high-density magnetic recording is performed on a magnetic recording medium, intersymbol interference inevitably occurs. Hereinafter, the intersymbol interference will be described.

For example, when a digital signal is transmitted through a linear transmission path, the input data sequence is [a _i ] (a _i = 0, 1), the output signal is y (t), and the impulse response of the transmission path is g.
Assuming that (t), the output signal c (t) when the output signal e (t) [e (t) = 1, where 0 ≦ t ≦ T] is transmitted is expressed by the following equation.

When the bit interval is T and c (t) is used, the output signal y (t) is expressed by the following equation.

Here, the input data sequence is expressed as [a _i ] = [101100111001
0], if c (nT) = 0 (n ≠ 0), the ninth
As shown in the figure, since the output value at the adjacent bit is not affected by c (t), the output signal y (n
T) becomes 0 or 12, making detection easy.

Therefore, usually, an equalizer (equalizer) is required and c
Pulse shaping is performed so that (nT) = 0.

However, if the equalization is not sufficient and an equalization error remains and c (nT) ≠ 0, adjacent bits are affected as shown in FIG. 10, and the output signal y at the time of detection is affected.
(NT) does not become 0 or 1.

 This is called intersymbol interference.

[Problems to be solved by the invention]

The above-mentioned intersymbol interference includes linear distortion and non-linear distortion.

Conventionally, intersymbol interference has been removed by equalizing linear distortion with a linear equalizer (a combination of an integrator and a low-pass filter). However,
In the case of this linear equalizer, noise is also emphasized, so that there is a problem that S / N is deteriorated.

Therefore, it has been proposed to remove intersymbol interference by path feedback Viterbi decoding [“Path feedback Viterbi decoding method for detecting without opening the eye” Ikeya, Ide, Yamamitsu, IEICE Tech.
87-38]. When the inter-symbol interference is removed by the path feedback Viterbi decoding, only the inter-symbol interference due to the already decoded data can be removed, and the intersymbol interference from the data that has not been decoded cannot be removed.

On the other hand, as an example of the nonlinear distortion, there is a magnetic recording of a digital signal which has been performed at a high density in recent years.

This magnetic recording process is a non-linear process, but assuming that the magnetization pattern in the medium is equal to the linear addition of the isolated magnetization reversal, the linear reproduction path has the same isolated reproduction waveform as shown in FIG. A reproduced waveform can be obtained by convolution. In the drawing, AR1 denotes a transition area, and 100 denotes a magnetic tape.

When recording a digital signal, as shown in FIG. 12, if the recording wavelength T is relatively long, the above assumption is satisfied, so that a reproduced waveform can be obtained by convolution of an isolated reproduced waveform. However, as shown in FIG. 13, as the recording wavelength T becomes shorter due to high-density recording, the distance between adjacent bits becomes shorter, and as shown by the arrow Y in FIG. 13, the transition areas AR1 overlap. Become like
Then, the residual magnetization of the already recorded recording signal magnetic field is disturbed by the new recording signal magnetic field.

For this reason, it is difficult to establish the above-described assumption of the linear addition of the isolated magnetization reversal, and the actual reproduced waveform does not match the waveform obtained by convolution of the isolated reproduced waveform. As a result, there is a problem that a non-linear distortion occurs as shown by the arrow X in FIG.

Up to now, little has been done about the removal of the nonlinear distortion as described above, and it is desired to remove the intersymbol interference due to the nonlinear distortion.

By the way, it is conceivable to apply an equalizer to remove the non-linear distortion, but there is a problem that the conventional linear equalization is insufficient and cannot cope with the non-linear distortion. In addition, it is conceivable to apply the above-mentioned path feedback Viterbi decoding.However, in this case, since the influence of subsequent data is not removed, accurate decoding is difficult and nonlinear distortion cannot be dealt with. Was.

Further, there is a problem that the error rate is not improved due to the intersymbol interference.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide an intersymbol interference removing apparatus capable of removing intersymbol interference such as non-linear distortion and linear distortion in data recorded magnetically at high density.

[Means for solving the problem]

According to the present invention, a state is set from a bit being reproduced and a plurality of bits in the vicinity of the reproduced bit. Viterbi which selects by comparing the value of a metric obtained by arithmetic processing including integration from a calculated value obtained by a predetermined calculation based on a reproduced signal corresponding to the reproduced signal, and generates a decoded value based on the selected sequence of state transitions It is configured to apply an algorithm.

[Action]

According to the present invention, a state is set from a bit being reproduced and a plurality of bits near the reproduced bit, and a Viterbi algorithm is applied to this state. That is, a metric is obtained based on a signal detected from a bit being reproduced and a predetermined expected sample value, a surviving path is determined based on a comparison of the metric, and a sequence of decoded values with the highest likelihood is determined. Desired.

Thus, the data recorded on the magnetic recording medium is reproduced.

〔Example〕

Hereinafter, an embodiment of the present invention will be described with reference to FIGS.

In the embodiment of the present invention, the removal of the nonlinear distortion is described in one embodiment, and then the removal of both the linear distortion and the nonlinear distortion is described in another embodiment.

1 to 5 show one embodiment of the present invention.

FIG. 1 shows a state setting of one embodiment, and FIG. 2 shows a block diagram of an intersymbol interference removing apparatus.

First, in order to facilitate understanding of the embodiment, the application of the Viterbi algorithm will be described, and then the intersymbol interference removing apparatus will be described.

As described above, the non-linear distortion may be caused by the recording magnetic field of the subsequent bit disturbing the already formed recording magnetic field (residual magnetization) or the recording demagnetization. It depends on the following data (data to be reproduced). Therefore, in the present invention, the state of the detection bit _aK and the bit pattern of the subsequent data is set as the state, and the Viterbi algorithm is applied.

The number of the above conditions is defined by either assumed to influence subsequent data after many bits from the detected bits a _K. For example, as shown in FIG. ₁ , a case will be described in which the influence of a total of three bits of a detection bit a _K and two bits a _{K + 1} and a _{K + 2} as subsequent data is assumed. As shown in FIG.
There are 2 ³ = 8 cases.

In the expected sample values Ys are considered in the detection bit a _K corresponding to each case. If there is no non-linear distortion and no equalization error, the expected sample value Ys is 0 for Y1 to Y4 and 1 for Y5 to Y8. By applying non-linear distortion according to (S = S1 to S8), eight values of Ys = Y1 to Y8 can be considered. Note that the expected sample value Ys is a value obtained experimentally. The expected sample value Ys can be determined adaptively, for example, by digitizing the reproduced waveform and actually measuring the measured value, or measuring the error rate. This 8
The different cases are expressed using state transitions. Consider a 2-bit sequence (a _K , a _{K + 1} ) as a state, and change state 1 to (0,
0), state 2 is (0,1), state 3 is (1,0), and state 4 is (1,1). The above eight cases can be represented by transitions between these four states. For example, case 1 (0,0,0) is state S
This is the sequence when the state transits from 1 (0, 0) to state S1 (0, 0) at the next time. Case 2 (0,0,1) is the state S1 (0,0,1)
0) to state S2 (0, 1). Hereinafter, similarly, case 3 (0, 1, 0) changes from state S2 (0, 1) to state S3 (1,
0), case 4 (0,1,1) is a transition from state S2 (0,1) to state S4 (1,1), and case 5 (1,0,0) is a state S3 (1,1).
0) to state S1 (0,0), case 6 (1,0,1) is a transition from state S3 (1,0) to state S2 (0,1), case 7 (1,1,
0) is a transition from state S4 (1,1) to state S3 (1,0), case 8 (1,1,1) is a transition from state S4 (1,1) to state S4 (1,1) Can be expressed as In general, when assuming effects over N bits, 2 ^{N -1} states are possible.

FIG. 3 shows state transition diagrams for eight cases.
Expected sample values corresponding to state transitions are also shown. FIG. 4 shows a trellis diagram.

The actual reproduced signal is Zk (Zk = Ys + nk: nk is assumed to be Gaussian noise), and the Viterbi algorithm is applied by setting the “length” of the transition to a certain state S to (− (Zk + Ys) ² ). . That is, a time point (k-1) to at metrics for each state, when the respective _{^{L k-1 1, L k}} -1 2, L k-1 3, L k-1 4, for each transition length With the maximum L _k-1 ¹ and the maximum L _k-1
^2, ..., transition as L _k-1 ⁴ is obtained, are respectively selected. For example, when calculating L _k ¹ , since S 1 or S 3 leads to the state S 1 from the state transition diagram of FIG.
_k ¹ is ^{_{[L k-1 1 - (Zk}} -Y1) 2 ], ^{_{[L k-1 3 - (Zk}} -Y5)
² ] and the larger value is selected. Thereby, the most probable state transition, that is, the most probable detection sequence is obtained.

In this way, the most probable detection sequence can be selected while predicting the bit pattern of the subsequent data that has not been decoded yet.

 Hereinafter, the algorithm will be described with reference to FIG.

In the configuration of FIG. 2, a signal Zk reproduced from the magnetic tape 1 by the magnetic head 2 at time t (k) is:
PLL4 via equalizer amplifier 3 and terminal of switch 5
Supplied to 5a.

In the PLL 4, a clock signal CLK is extracted from the reproduced signal Zk. This clock signal CLK is taken out from the terminal 6 and supplied to each circuit inside the device.

The switch 5 receives the clock signal CL supplied from the terminal 7
Opening and closing is controlled by K. When the switch 5 is closed, that is, when the terminals 5a and 5b are connected, the signal Zk is
~ 15 respectively.

The adders 8 to 15 add the predetermined sample values Y1 to Y8 supplied from the terminals 16 to 23 to the signal Zk. Outputs from the adders 8 to 15 are supplied to squaring circuits 24 to 31.

In the squaring circuits 24 to 31, the values of the outputs (Zk-Y1) to (Zk-Y8) supplied from the adders 8 to 15 are squared and assigned negative signs. The output from the squaring circuits 24 to 31 (− (Zk−Y
1) Each of ² ) to (− (Zk−Y8) ² ) is an adder 32 to 39
Supplied to

In the adder 32, the output (− (Zk−Y1) ² ) described above and the metric L at the time t (k−1) supplied from the comparator 40 are output.
adding the _k-1 ¹ is made, it is supplied to the memory 44.

In the memory 44, L _k− ¹¹ − (Zk−Y 1) ² is fetched based on the clock signal section CLK supplied via the terminal 52 and supplied to the comparator 40.

In the adder 33, the output (− (Zk−Y2) ² ) described above and the metric L at the time t (k−1) supplied from the comparator 40 are output.
adding the _k-1 ¹ is made, it is supplied to the memory 45.

In the memory 45, L _k− ¹¹ − (Zk−Y1) ² is fetched based on the clock signal CLK supplied via the terminal 52,
The signal is supplied to the comparator 41.

The adder 34 outputs the above output (− (Zk−Y3) ² ) and the metric L at the time t (k−1) supplied from the comparator 41.
_k-1 ² from ^{_{L k-1 2 - (Zk}} -Y3) 2 are obtained and supplied to the memory 46. In the memory 46, based on the clock signal ^{_{CLK, L k-1 2 -}} (Zk-Y3) 2 is taken, it is supplied to the comparator 42.

In the adder 35, the output (− (Zk−Y4) ² ) described above and the metric L at time t (k−1) supplied from the comparator 41 are output.
Based on the _{^{k-1 2, L k-}} 1 2 - (Zk-Y4) 2 are obtained and supplied to the memory 47.

In the memory 47, L _k−1 ² −
(Zk-Y4) ² is taken in and supplied to the comparator 43.

In the adder 36, the output (− (Zk−Y5) ² ) described above and the metric L at the time t (k−1) supplied from the comparator 42 are output.
Based on the _{^{k-1 3, L k-}} 1 3 - (Zk-Y5) 2 are obtained and supplied to the memory 48.

In the memory 48, based on the clock signal CLK, L _k-1 ³ −
(Zk−Y5) ² is taken in and supplied to the comparator 40.

In the adder 37, the output (− (Zk−Y6) ² ) described above and the metric L at the time t (k−1) supplied from the comparator 42 are output.
Based on the _{^{k-1 3, L k-}} 1 3 - (Zk-Y6) 2 are obtained and supplied to the memory 49.

In the memory 49, based on the clock signal CLK, L _k−1 ³ −
(Zk−Y5) ² is fetched and supplied to the comparator 41.

The adder 38 outputs the output (− (Zk−Y7) ² ) and the metric L at the time t (k−1) supplied from the comparator 43.
Based on the _{^{k-1 4, L k-}} 1 4 - (Zk-Y7) 2 are obtained and supplied to the memory 50.

In the memory 50, based on the clock signal CLK, L _k−1 ⁴ −
(Zk−Y7) ² is taken in and supplied to the comparator 42.

In the adder 39, the above output (− (Zk−Y8) ² ) and the metric L at the time t (k−1) supplied from the comparator 43 are output.
Based on the _{^{k-1 4, L k-}} 1 4 - (Zk-Y8) 2 are obtained and supplied to the memory 51.

In the memory 51, based on the clock signal CLK, L _k-1 ⁴ −
(Zk-Y8) ² is taken in and supplied to the comparator 43.

In the comparators 40 to 43, according to the trellis diagram of FIG.
A magnitude comparison of each memory output is performed, a surviving path is determined from the comparison result, and based on a rule described later,
The state transition is determined. And “1” or “0”
Are supplied to the shift registers 53 to 56. At the same time, the comparators 40 to 43 output the metric L _k ^{1 at} time t (k).
~L _k ⁸ is fed back to the adder 32-39.

The shift registers 53 to 56 are connected to each other according to a trellis diagram, and perform serial loading or parallel loading of decoded values according to the selection of the surviving path described above. The serial loading or parallel loading of the decoded values in the shift registers 53 to 56 is performed as shown in FIG. 5 in synchronization with the clock signal CLK supplied from the terminal 52. This corresponds to merging of paths on the trellis diagram. The decoded values are supplied to the determination circuit 57 from the shift registers 53 to 56.

Although the four decoded values supplied to the discriminating circuit 57 should originally match, it is fully conceivable that they do not actually match due to noise or other effects. Therefore, in the discriminating circuit 57, the decoded value is selected based on a criterion such as majority decision or metric size, and is taken out from the terminal 58.

According to this embodiment, the detection bit a _K and the detection bit a _K
Since the state is set for 2 bits a _{K + 1} and a _{K + 2} as the subsequent data following, and the data is reproduced and decoded by applying the Viterbi algorithm, the data that is magnetically recorded at high density In this case, the nonlinear distortion caused by the two bits a _{K + 1} and a _{K + 2 of} the continuous data can be removed, and the intersymbol interference can be removed.

In addition, in removing the intersymbol interference, a linear equalizer is not used, so that it is possible to prevent the S / N from deteriorating.Because it is not based on path feedback Viterbi decoding, data that has not been decoded yet Can be eliminated.

In this embodiment, in order to remove the intersymbol interference, the state is set with respect to the detection bit a _K and the three bits of the subsequent data bits a _{K + 1} and a _{K + 2} . The present invention is not limited to this, and the number of target bits can be set arbitrarily.

6 to 8 show another embodiment of the present invention.

The other embodiment is different from the above-described embodiment in that bits a _K−1 and a _{K + 1} before and after the detection bit a _K are considered. That is, in the above-described embodiment, an example is described in which the non-linear distortion which is the effect of the subsequent bits a _{K + 1} and a _{K + 2 on} the detection bit is removed. That is, the linear distortion, which is the effect of the bit a _K−1 of FIG.

In this other embodiment, as shown in FIG. 6, eight cases can be considered from the values of each bit a _K−1 and a _{K + 1} before and after the detection bit a _K. Then, a decoded sequence with the highest likelihood is obtained based on the Viterbi algorithm.

 First, the intersymbol interference in each case will be described.

Assuming that there is no noise, the value of the signal at the detection time t = KT is uniquely determined. If the equalization error C _K is 0, a _K = 0
If y _K = 0, then if a _K = 1, then y _K = 1. When there is an equalization error as shown in FIG.
y _K is represented by the following equation.

Generally, when the equalization error C _K remains over n points, y _K takes 2 ⁿ values. For example, as shown in FIG. 8, the equalization error C _-1 over three points,
If the C _0, C ₁ remains, the value of y _K are considered are 2 ³ = 8.

_{(A K-1, a K} , a K + 1) = (000) y K = 0 (= y S1) if _{S1 (a K-1, a} K, a K + 1) = (001) y K = When C ₋₁ (= y _S2 ), S2 (a _K−1 , a _K , a _{K + 1} ) = (010) y _K = 1 (= y _S3 ), and S3 (a _K−1 , a _K , a _{K) +1) = (011) y K} = 1 + C -1 (= y S4) when _{S4 (a K-1, a} K, a K + 1) = (100) y K = C 1 (= y S5) If S5 _{(a K-1, a K} , a K + 1) = (101) y K = C -1 + C 1 (= y S6) If _{S6 (a K-1, a} K, a K + 1) = (110 _{) y K = 1 + C 1} (= y S7) when _{S7 (a K-1, a} K, a K + 1) = (111) y K = 1 + C -1 + C 1 (= y S8) If S8 Furthermore, non-linear When distortion is considered, the nonlinear distortion α _i corresponding to each pattern may be obtained and added by any method. Since the nonlinear distortion basically depends on the pattern, detection in this way can be expected with sufficient consideration of the influence of the nonlinear characteristic.

(A _K−1 , a _K , a _{K + 1} ) = (000) y _k = 0 + α ₁ (= y _S1 ) S1 (a _K−1 , a _K , a _{K + 1} ) = (001) y _K = C _-1 + α ₂ (= y _S2 ) S2 (a _{K -1} , a _K , a _{K +1} ) = (010) y _K = 1 + α ₃ (= y _S3 ) S3 (a _{K -1} , a _{K, a K + 1) =} (011) y K = 1 + C -1 + α 4 (= y S4) when _{S4 (a K-1, a} K, a K + 1) = (100) y K = C 1 + α ₅ (= y _S5) If _{S5 (a K-1, a} K, a K + 1) = (101) y K = C -1 + C 1 + α 6 (= y S6) If S6 (a _K-1, a _{K, a K + 1) =} (110) y K = 1 + C 1 + α 7 (= y S7) when _{S7 (a K-1, a} K, a K + 1) = (111) y K = 1 + C -1 + C ₁ + α ₈ (= y _S8 ) S8 At this time, (a _K−1 , a _K , a _{K + 1} ) is considered as a state transition and a trellis diagram is considered, and the Viterbi algorithm is applied. still,
The contents of the application of the Viterbi algorithm, the circuit, the transition diagram, the trellis diagram, and the like are the same as those in the above-described embodiment.
Common parts are denoted by the same reference numerals, and redundant description is omitted.

According to the other embodiment, in addition to the effects of the above-described embodiment, the detection bit _aK and the one bit before and after the detection bit _aK-1 and _{aK + 1} are added.
The data is decoded by applying the Viterbi algorithm to the target, so that in the data that is magnetically recorded at a high density, the linear distortion caused by each of the preceding and succeeding one bit a _K-1 and a _{K + 1} , and , The bit of the subsequent data following the detection bit a _K
a Both nonlinear distortion caused by _{K + 1} can be removed, and intersymbol interference can be removed.

〔The invention's effect〕

According to the intersymbol interference removing apparatus according to the present invention, since the data is decoded by applying the Viterbi algorithm to the detected bit and the bits before and after the detected bit, the data in the magnetically recorded data at high density can be obtained. This has the effect of eliminating nonlinear distortion and intersymbol interference. In addition, there is an effect that linear distortion in data that is magnetically recorded at high density can be removed and intersymbol interference can be removed.

Thereby, there is an effect that the error rate can be improved.

Since the linear equalizer is not used for removing the intersymbol interference, noise is not emphasized and deterioration of S / N can be prevented. Further, since the path feedback Viterbi decoding is not used, there is an effect that intersymbol interference from data that has not been decoded can be removed.

[Brief description of the drawings]

FIG. 1 is an explanatory view showing a state setting in one embodiment of the present invention, FIG. 2 is a block diagram showing one embodiment of the present invention, FIG. 3 is a state transition diagram, and FIG. Figure, fifth
FIG. 6 is a schematic diagram showing the operation of the shift register, FIG. 6 is an explanatory diagram showing a state setting in another embodiment of the present invention, FIG. 7 is an explanatory diagram showing intersymbol interference due to an equalization error, FIG. 8 is an explanatory diagram showing an example of an equalization error, FIGS. 9 and 10 are diagrams showing levels of output signals according to the presence or absence of an equalization error, respectively.
11 shows the magnetization state of the isolated reproduction waveform, FIG. 12 shows the magnetization state when the recording signal has a long wavelength, and FIG. 13 shows the magnetization state and nonlinear distortion when the recording signal has a short wavelength. FIG. Description of the key code in the drawing, 100: magnetic tape, a _K: detection _{bit, a K-1, a K} + 1, a K + 1, a K + 2: bit, S, S1 to S8: Status.

Claims (1)

(57) [Claims] 1. A state is set from a bit being reproduced and a plurality of bits in the vicinity of the reproduced bit, and a maximum likelihood sequence of transitions between the states is determined by a reproduction bit including the reproduced bit. Selection is made by comparing the value of a metric obtained by arithmetic processing including integration from a calculated value obtained by a predetermined calculation based on a reproduced signal corresponding to the sequence, and a decoded value is generated based on the selected sequence of state transitions. An inter-symbol interference removing apparatus to which a Viterbi algorithm is applied.