KR100474821B1 - Apparatus and method for reproducing signal with nonlinear characteristics - Google Patents

Abstract

The present invention relates to a reproduction signal processing apparatus and method having a non-linear characteristic, the reproduction signal processing apparatus having a non-linear characteristic includes a pre-processing filter for processing the reproduction signal as a standard response signal having a causal characteristic to output; A first adder for subtracting the intersymbol interference signal estimated from the previous signal from the signal processed by the preprocessing filter; A slicer which compares the value added by the first adder with a predetermined reference threshold and converts the value into a digital value according to the result; A first delayer for delaying the output of the slicer for a predetermined time; And an intersymbol interference estimator for estimating the intersymbol interference signal according to the transition polarity of the previous output signal from the value delayed in the first delay unit.

According to the invention. In processing the signal reproduced from the data storage device, the distortion due to the intersymbol interference signal is estimated by removing the distortion characteristic through a predetermined feedback filter according to the transition polarity of the signal, and the threshold value of the slicer is adjusted. Nonlinear distortion due to asymmetry and saturation caused by the difference in response can be effectively eliminated. In addition, it is possible to approximate the channel coefficient more quickly and accurately by using the parameter estimation method when setting the initial coefficients of the preprocessing filter and the feedback filter.

Description

Apparatus and method for reproducing signal having nonlinear characteristics

The present invention relates to an apparatus and method for processing a reproduced signal, and more particularly, to an apparatus and method for processing a signal having a nonlinear characteristic reproduced from a data storage device.

Signals reproduced from the data storage device interfere with each other, making signal detection difficult. In particular, when a magnetic resistance head is used for high-density recording, the signal is distorted due to the asymmetry between the positive and negative pulses of the reproduction signal and the saturation of the magnetic flux. Even more difficult. In addition, if the DC bias position of the MR head is not set correctly, an asymmetry phenomenon occurs in the reproduction signal, and the response characteristic is different depending on the sign of the input transition. Figure 1 shows the response characteristics of an MR head according to an applied magnetic field. According to this response curve in the form of a quadratic curve, for example, if the input signal is located in the center of the response curve, the output signal The frequency of is doubled as the input signal frequency and output. That is, according to the position of the DC bias field H _b , the asymmetric effect of the shape of the positive peak and the negative peak becomes incorrect. Therefore, the correct H _b value must be applied so that the MR head operates in quasi-linear mode so that the positive and negative pulses have the same shape. Although the SAL (Soft Adjacent Layer) bias method and the coupled-element method are used to accurately make the above-described H _b , this method improves the symmetry of the waveform but has a problem that the signal-to-noise ratio is deteriorated.

One way to model asymmetry and saturation is by Proakis. The above-described Pro-Akis method can model non-linear characteristics by changing the power according to the polarity of the signal using a single PW50 model. However, since only a single PW50 is used, inter-symbol interference depends on the polarity of the pulse. It is difficult to model the signal). Here, pw50 is the pulse width measured at two points corresponding to 50% of the peak regeneration pulse voltage.

Viterbi algorithm for equalizing the signal with ISI and nonlinear distortion from the reproduced signal, adding the square of the difference between the reproduced signal sequence and all possible signal sequences that have been recorded for each signal sequence, and then selecting the signal sequence whose value is minimum There is also a way to use (Viterbi Algorithm). Differential PR4 VA is applied to the real channel, which has the advantage of simplicity of implementation and fast recovery of signal sequence, but it does not fit with the PR4 channel as the recording density of the storage device increases. As the ISI of the portion increases, system performance deteriorates during high density recording. Decision-Feedback Equalizer (hereinafter referred to as DFE) may be used as a method for solving the above problems. However, if the feedback filter is linear, performance may be degraded by asymmetric distortion, which is a nonlinear characteristic. There is nothing else.

Another method is to change the function of the analog-to-digital converter to reduce the asymmetry effect in the detector. In other words, the asymmetry existing in the analog stage is different from the digital stage by adjusting the number of quantization bits in the positive part and the number of quantization bits in the negative part. However, this method does not consider the ISI but simply considers the asymmetry difference of the peak, and thus is not a solution when the threshold value is changed by the ISI or when a nonlinear effect due to saturation occurs.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and provides a nonlinear characteristic for supplying a reproduction signal having accurate and symmetrical by suppressing inter-symbol interference signal from a reproduction signal having a nonlinear characteristic and mitigating distortion of the signal due to saturation. It is an object of the present invention to provide an apparatus and method for processing a reproduction signal having the same.

According to the present invention for achieving the above object, the apparatus for processing a reproduction signal having a nonlinear characteristic comprises a pre-processing filter for processing the reproduction signal as a standard response signal having a causal characteristic and outputs; A first adder for subtracting the intersymbol interference signal estimated from the previous signal from the signal processed by the preprocessing filter; A slicer which compares the value added by the first adder with a predetermined reference threshold and converts the value into a digital value according to the result; A first delayer for delaying the output of the slicer for a predetermined time; And an intersymbol interference estimator for estimating the intersymbol interference signal according to the transition polarity of the previous output signal from the value delayed in the first delay unit.

According to another aspect of the present invention, there is provided a method of processing a reproduction signal having a nonlinear characteristic, comprising: a first step of preprocessing an input signal expressed in symbol units using a predetermined filter; A second step of subtracting the intersymbol interference signal from the result of the preprocessing step in the first step; A third step of detecting a symbol based on a predetermined threshold value from the signal from which the interference signal is removed in the subtraction step; A fourth step of delaying the signal detected in the symbol detection step by one period when a predetermined time unit is one period, and then examining transition polarity of the delayed signal; And a fifth step of estimating the inter-symbol interference signal to be removed from the next signal using a predetermined feedback filter which inputs predetermined values output according to the transition polarity irradiated in the fourth step. Do.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. 2 is a block diagram of an apparatus for processing a reproduction signal having a nonlinear characteristic according to the present invention. Referring to FIG. 2, an apparatus for processing a reproduction signal having a nonlinear characteristic includes a linear preprocessor filter (LPF, 200), Adder 210, Variable Threshold Slicer (VTS, 220), Delay 230, Nonlinear Volterra Signal Debinder (NVSD, 240), First Feedback Filter 250, and First It consists of two feedback filters 260.

First, the LPF 200 is composed of a Linear Transversal Filter to output an input signal as a canonical system response having a causal characteristic, and the adder 210 is an LPF 200. The ISI which is the output of the first and second feedback filters is subtracted from the output signal of?. The VTS 220 determines the output of the adder 210 as 0 or 1 based on the variable threshold. The delay unit 230 delays the output of the VTS 220 by one cycle, and the NVSD 240 divides the delayed signal from the delay unit 230 into a nonlinear combination according to polarity. The first and second feedback filters 250 and 260 are general filters formed in the form of a linear delay line filter to remove the ISI by filtering the signal according to the ISI pattern according to the polarity of the output signal of the NVSD.

The VTS 220 described above is composed of a selector 310 and a biased slicer 320 as shown in FIG. 3. The selector 310 outputs one of r ₊ and r _- which adjusts the threshold upward or downward according to the input signal, and the biased slicer 320 is shifted by the output value of the selector 310 according to the threshold of FIG. 2. The signal output from the adder 210 is determined. That is, the output signal of the adder 210 (see FIG. 2) is determined according to the threshold biased by one polarity according to the degree of asymmetry, and the binary signal b [k], which is one of {0,1}, is output, thereby saturating magnetic flux. It is possible to alleviate the distortion of the signal due to.

4 is a block diagram of an apparatus for adapting r ₊ , r ₋ of FIG. 3, wherein the apparatus for adapting r ₊ , r ₋ includes a selector 310, a delayer 410, an adder 420, and a switch 440. ) And a multiplier 430. First, the selector 310 is the same as the selector of FIG. 3, the delayer 410 delays the output value of the selector 310 by one period, and the adder 420 is the output of the delayer 410 and the multiplier 430. Add the output of. The switch 440 outputs one of r ₊ and r ₋ according to the input signal and adds the corresponding value to r ₊ and r ₋ of the selector 310. Multiplier 430 multiplies the biased threshold of 1, the error value e [k], which is the difference between the input signal and the determined signal, and the step size μ.

The above-described NVSD 240 is shown in block diagram in FIG. 5A, and FIG. 5B is a diagram in which FIG. 5A is configured by a simple logical operator. First, the delay unit 510 of FIG. 5A delays the input signal by one cycle and inputs the non-linear voltage converter 520. The nonlinear Volterra converter 520 outputs different values according to the transition polarity by comparing the signal delayed by the delay unit 510 with the input signal. FIG. 5B is a simple logic gate of FIG. 5A and includes inverters 530 and 540 and logic sum gates 550 and 560.

FIG. 6 is a block diagram of an apparatus for adapting the above-described first and second feedback filters, wherein the first adder 600, the first and second multipliers 612 and 622, and the second and third adders 614, 624, first and second delayers 616 and 626, and first and second feedback filters 250 and 260.

First, referring only to the left side of FIG. 6, the first adder 610 obtains a difference between the output of the adder 210 of FIG. 2 and the signal or target signal determined by the VTS 220 (see FIG. 2), and the first multiplier 612. ) Outputs a result of multiplying the result of the first adder 610 by the step size μ and a signal string having a positive polarity among the outputs of the nonlinear Volterra signal divider 240 of FIG. 2. The second adder 614 outputs the result of adding the output of the first multiplier 612 and the output of the first feedback filter 250 the previous number of times, that is, the output delayed by one period through the first delay 616. Then, the coefficient of the first feedback filter 250 is adjusted to this value. The right side of FIG. 6 has the same structure as the left side, except that one of the inputs of the second multiplier is a signal sequence having negative polarity among the outputs of the nonlinear Volterra signal divider 240 of FIG. 2, and the adaptation target is the second feedback. The filter 260 is different. [K-1: k-K] in the figure means a signal sequence from the (k-1) to the K-th of each signal. K is the number of taps in the filter.

Meanwhile, the operation principle of the present invention will be described with reference to the flowchart of FIG. 7. 7 is a flowchart illustrating a method of processing a reproduction signal having a nonlinear characteristic according to the present invention. First, the reproduction signal a [k] read from the storage device is transformed into a causal standard system response x [k] through the LPF (step 710), and the system response x [k] at this time is as follows. .

[Equation 1]

Where K is the number of taps in the filter

g _k : filter coefficient

D ^k : Delay of the filter.

Since LPF is a linear delay type filter, it is output in the form of standard system response, but it is difficult to remove the influence of ISI which is different depending on the polarity of the signal. That is, the output of this filter is a signal x [k], which is a combination of two ISIs depending on the polarity.

In the above-described x [k], the ISI estimated by the first and second feedback filters 250 and 260 is subtracted through the adder 210 of FIG. 2 (step 720). The subtracted signal z [k] in step 720 is detected as 1 when the threshold is greater than the threshold value or 0 when it is smaller through the VTS 220 (see FIG. 2) (step 730). When the detected signal of step 730 is b [k], the threshold of the VTS 220 (see FIG. 2) is changed according to the state of the previous output value, that is, b [k-1]. If b [k-1] is known, the polarity of b [k] can be predicted so that the threshold value is changed using the state of b [k-1]. The polarity of the signal according to b [k-1] is shown in the following table. The threshold term zero is a reference threshold when not varied.

TABLE 1

The variable threshold will be described with reference to FIG. 3. First, in the selector 310, r ₊ is output when b [k-1] is 1, and r ₋ is output when b [k-1] is 0. The threshold of the slicer is moved by one of r ₊ and r _- according to the value output from the selector 310 at the 0 level.

[k]만큼 가산되어

[k]가 된다. 이 때,

[k]는 소정의 스텝사이즈, 바이어스 문턱값 1 그리고 b[k]-z[k]인 e[k]가 승산기(430)에 의해 곱해진 값이다. 상술한

[k]는 스위치(440)로 입력되어 b[k-1]가 1이면 r₊에, b[k-1]가 0이면 r_-에 더해진다.At this time, r ₊ and r _- are adapted according to the value of b [k-1]. The adaptation method will be described with reference to FIG. If the r value, r [k], for the k-th signal is determined by the selector 310, r [k-1] delayed by one period by the delayer 410 is increased by a predetermined increment.

is added by [k]

becomes [k]. At this time,

[k] is a value obtained by multiplying multiplier 430 by a predetermined step size, bias threshold 1 and e [k] of b [k] -z [k]. Above

[k] is input to the switch 440 and added to r ₊ if b [k-1] is 1 and r ₋ if b [k-1] is 0.

The signal b [k] passing through the VTS 220 is delayed by one cycle through the delayer 230, and then in accordance with the transition polarity from b [k-2] to b [k-1] via the NVSD 240. the positive transition (positive transition) surface _- and the _{(b + [k-1]} , b [k-1]) is (1,0), is converted into a negative transition (negative transiton) surface (0,1). 5A and 5B, the conversion method is as follows. If b [k-2] with a delay of b [k-1] and b [k-1] is input to the nonlinear Volterra converter, the temporary signal c [k] is converted to c [k-1] = b [k- can be obtained _{[k-1] - 1]} -b [k-2] to La, and by the food _{b + [k-1],} b.

[Equation 2]

Output at positive transition polarity: b ₊ [k-1] = (c [k-1] + l) c [k-1] / 2

Output when the negative polarity _{transition: b - [k-1]} = (c [k-1] -1) c [k-1] / 2

When the above equation 2 is converted into a logical operation, it is shown in the following table and can be configured as a logical operator as shown in FIG. 5B. In the following table, b [k-1] is A1, b [k-2] is A2, the output of b ₊ [k-1] is _{B +, b - [k-} 1] is B _- a.

TABLE 2

Rule: B ₊ = A1 AND-A2

B _- = _- A1 AND A2

The output of the _{NVSD (240) b + [k} -1], b - [k-1] is to adjust the coefficients of each of the feedback filter as shown in Figure 6 is supplied to the input of the first and the second feedback filter 250, respectively According to the polarity of the signal, the ISI, which is combined into the next signal sequence, is estimated from the previous signal sequence (step 740).

[k-1:K]가 된다.

[k-1:K]는 다시 제1귀환필터(250)로부터 나오는 한 주기 지연된 출력 b+[k-1:K]과 제2승산기(622)를 통해 가산된 다음, 제1귀환필터(250)의 i번째 횟수의 적응계수에 대해 i+1번째 횟수의 k번째 탭계수 W_i+1[k]를 조정한다. 조정규칙은 다음과 같다.An adaptive method of each feedback filter for ISI approximation will be described with reference to FIG. 6. First, the input signal z [k] becomes the error signal e [k] through the determined signal b [k] or the predetermined target signal d [k] and the first adder 610. e [k] is multiplied by the adaptive constant μ and the positive signal sequence b ₊ [k-1; K] through the first multiplier 612

[k-1: K].

[k-1: K] is added to the output b + [k-1: K] delayed by the first feedback filter 250 through the second multiplier 622, and then the first feedback filter 250 is added. The k-th tap coefficient W _{i + 1} [k] of the _{i + 1} th time is adjusted with respect to the i th coefficient of the adaptive number of times. The adjustment rules are as follows.

[Equation 3]

In other words,

Here, 2 may or may not be added as a constant.

[k-1:K]가

[k-1:K]로 되는 것이 다르다. 도 2의 LPF(200)도 입력신호를 등화하기위해 필터의 계수들을 적응시키는데, 이들의 적응방법도 상술한 귀환필터의 적응방법과 같다. 다만, b₊[k-1:K]가 a[k]로 바뀔 뿐이다.The same as the Fig adaptation of the feedback filter 260 described above, but _{b + [k-1: K} ] is _{b - [k-1: K} ] replaced with,

[k-1: K]

[k-1: K] is different. The LPF 200 of FIG. 2 also adapts the coefficients of the filters to equalize the input signal, and their adaptation method is the same as that of the feedback filter described above. It just changes b ₊ [k-1: K] to a [k].

As described above, the ISI from the past signal sequence is estimated according to the polarity of the signal, and then the estimated ISI is removed from the output of the LPF 200 (see FIG. 2).

Here, a method of determining the initial coefficients of the LPF 200 and the first and second feedback filters 250 and 260 (see FIG. 1), which are components of the present invention, will be described.

In general, the initialization of the filter coefficients uses the least mean square method or the recursive least square method, which requires a great deal of training to find the optimal coefficient. However, the above-described methods have a low convergence speed, fall into a local minimum, and obtain a large value as a solution.

Therefore, the present invention avoids the above-described methods, and instead uses a parameter estimating method (weakly referred to as PE) that can be easily realized in hardware using a fast lookup table and a look-up table. We obtain the initial coefficient for the channel. Using the characteristic that the linear response of the MR head is similar to the Lorentz distribution curve, PE finds the channel parameters pw50 +, pw50-, and gain, without finding all the tap coefficients of the channel, and then approximates the Lorentz distribution curve. As a method of obtaining channel characteristics, the convergence speed and efficiency are high.

The obtained channel characteristic is converted into the tap coefficient of the filter described above using a known Levinson-Trench algorithm. Where pw50 + is the pulse width measured at 50% of the pulse voltage peak with positive polarity and pw50 + is the pulse width measured at 50% position of the pulse voltage peak with negative polarity. Gain refers to channel gain, and refers to peak value when the channel characteristics are approximated by the Lorentz distribution curve described above. The Lorentz distribution curve can be expressed by the following equation.

[Equation 4]

To find the channel parameters pw50 +, pw50-, and gain, a known Stochastic gradient-based algorithm is used for a predetermined number of times.

[Equation 5]

Where C: channel estimate

β, r: step size

b _j : a _j -a _j-1 ; a _j is the playback signal

e _j : Actual channel output-channel estimate

The step sizes β and r can be found by using the formula that the Lorentz distribution curve can be represented by a closed form equation. Since pw50 + and pw50- are independent of each other, only one is considered to find the correlation with the gain, and then β and r values are set to zero, which is the cost function of initial values p ₀ and g ₀ . The cost function J (p, g) for pw50 and the gain is obtained using MAPLE, a symbolic mathematic tool.

[Equation 6]

Where p: pw50

p ₀ : Initial value of pw50

g: gain

g ₀ : initial value of gain

FIG. 8 illustrates a trajectory of a cost function for pw50 and a gain when p ₀ = 3.0 and g ₀ = 2.0 using Equation 6 described above, and FIG. 9 illustrates values of β and r found using FIG. 8. As a result of finding the parameter by applying the above equation (5), it can be seen that convergence in about 15 iterations.

According to the invention. In processing the signal reproduced from the data storage device, the distortion due to the intersymbol interference signal is estimated by removing the distortion characteristic through a predetermined feedback filter according to the transition polarity of the signal, and the threshold value of the slicer is adjusted. Nonlinear distortion due to asymmetry and saturation caused by the difference in response can be effectively eliminated. In addition, it is possible to approximate the channel coefficient more quickly and accurately by using the parameter estimation method when setting the initial coefficients of the preprocessing filter and the feedback filter.

1 is a response characteristic of an MR head according to an applied magnetic field.

2 is a block diagram of a reproduction signal processing apparatus having a nonlinear characteristic according to the present invention.

3 is a block diagram of the variable threshold slicer of FIG. 2.

4 is a block diagram of an apparatus for adapting r ₊ , r ₋ of FIG. 3.

FIG. 5A is a block diagram of the nonlinear Volterra signal divider of FIG. 2.

FIG. 5B is a diagram illustrating a simple logical operator of FIG. 5A.

6 is a block diagram of an apparatus for adapting the first and second feedback filters of FIG.

7 is a flowchart illustrating a reproduction signal processing method having a nonlinear characteristic according to the present invention.

8 shows the trajectory of pw50 and gain.

FIG. 9 illustrates a convergence speed when finding a parameter using the trajectory of FIG. 8.

Claims (13)

A preprocessing filter for processing the reproduction signal into a standard response signal having causal characteristics and outputting the processed signal; A first adder for subtracting the intersymbol interference signal estimated from the previous signal from the signal processed by the preprocessing filter; A slicer which compares the value added by the first adder with a predetermined reference threshold and converts the value into a digital value according to the result; A first delayer for delaying the output of the slicer for a predetermined time; And Compare the value delayed by the first delayer with the value output from the first delayer in the previous time to obtain a signal sequence according to the transition polarity, and estimate the inter-symbol interference signal from the obtained signal sequence and output the symbol to the first adder. And an interference estimator. The method of claim 1, wherein the slicer A selector configured to select and output the second or third input according to the first input by using the output signal of the first delay as a first input and setting two predetermined values as second and third inputs; And And a biased slicer which adds the value selected by the selector to a predetermined threshold and converts the value added by the first adder to a digital value according to the result value, and outputs the digital signal. Device. The method of claim 2, wherein the selector A second delayer for delaying the output of the selector for a predetermined time; A multiplier that multiplies the bias reference value, a predetermined step size and the subtracted value of the output of the first adder from the slicer output; A second adder for adding the output of the second delayer and the output of the multiplier; And And a switch for transferring the output of the second adder to the second or third input of the selector in accordance with the output of the first delay unit. The method of claim 1, wherein the inter-symbol interference estimator A signal string separator having two output terminals and outputting predetermined values to the two output terminals according to the transition polarity of the signal delayed from the first delay unit; A first feedback filter connected to one output terminal of the signal splitter and modeling an intersymbol interference signal according to a case in which the transition polarity is positive, and then outputting an interference signal modeled by the adder; And And a second feedback filter connected to another output terminal of the signal splitter and configured to model the inter-symbol interference signal due to the case where the transition polarity is denied, and then output the interference signal modeled by the adder. Apparatus for processing reproduction signal having characteristics. The method of claim 4, wherein the signal splitter A third delayer for delaying the output of the first delayer for a predetermined time; And The output of the first delayer and the output of the third delayer are input, and the outputs of the third delayer and the output of the first delayer are compared with each other according to the transition polarity. Apparatus for processing a reproduction signal having a non-linear characteristics, characterized in that it comprises a non-linear converter for outputting. The method of claim 5, wherein the nonlinear transducer is A first inverter for inverting the output of the first delay unit; A second inverter inverting the output of the third delay unit; A first logical operator configured to perform an AND operation on one input of the output of the first delay and the other of the output of the second inverter; And A second logical product operator configured to perform an AND operation with one input as an output of the third delay and an output of the first inverter as an input. Device. A first step of preprocessing an input signal expressed in symbol units using a predetermined filter; A second step of subtracting the intersymbol interference signal estimated at a previous time from the result of the preprocessing step in the first step; And A third step of detecting a symbol based on a predetermined threshold value from the signal from which the interference signal is removed in the subtraction step, The intersymbol interference signal is generated at the previous time. Delaying the signal detected by the third step by a predetermined time and generating a signal sequence according to the transition polarity of the delayed signal; And And estimating the inter-symbol interference signal using a predetermined feedback filter as the input of the signal sequence. The method of claim 7, wherein the first step Setting an initial value of a predetermined channel parameter; Repeating the adaptive algorithm for a predetermined number of times to adapt the channel parameter set to the initial value; Obtaining a channel coefficient by approximating the channel characteristic to a predetermined statistical distribution curve using the adapted channel parameter; And And loading the obtained channel coefficient into the coefficients of the preprocessing filter using a predetermined conversion algorithm, and then filtering the input signal. The method of claim 8, wherein the adaptive algorithm A method of processing a reproduction signal having a nonlinear characteristic, characterized by using a stochastic gradient-based algorithm. The method of claim 9, Obtaining a predetermined cost function that defines the correlation between the channel parameters, and then determining the first step size of the stochastic gradient-based algorithm to approach the cost function as zero; . 8. The method of claim 7, wherein the coefficients of the preprocessing filter or the feedback filter are adapted each time an input is received. When the i-th input comes in, the k-th coefficient of each filter coefficient is W _i [k], and the predetermined second step size is μ and e [k] (the detected symbol value-actual output value). When the input value of these fields is called x [k], the kth coefficient when the i + 1st input is inputted is [Equation 3] W _{i + 1} [k] = W _i [k] +2 μ e [k] x [k] The processing method of the reproduction signal having a non-linear characteristic further comprising. The method of claim 7, wherein the third step And adjusting the threshold value by adding or subtracting a predetermined value to the threshold value according to the output of the signal before one cycle. The method of claim 7, wherein The predetermined values B1 and B2 output according to the transition polarity in the estimating step are inputted as a signal A2 which is further delayed by the delayed signals A1 and A1. B1 = A1 AND-A2 B2 = -A1 AND A2 Provided that AND is an AND and N is not A method of processing a reproduction signal having a nonlinear characteristic, characterized by performing logical operation as described above.

Images