JP3003711B2 - Demodulator of magnetic recording / reproducing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a demodulator of a magnetic recording / reproducing apparatus for sequentially demodulating a reproduction signal from a recording medium, and more particularly, to a reproduction signal from a recording medium having a certain regularity in the reproduction signal, for example, a magnetic signal. The present invention relates to a demodulator of a magnetic recording / reproducing apparatus that demodulates a reproduction signal of a disk device by using a neural net.

With the speeding up of computer systems, higher speeds and larger capacities are also required for magnetic disk devices as external storage devices. For this reason, the frequency of the signal handled by the demodulation circuit of the magnetic disk device is increased, the recording density (BPI) on the medium is increased, and the signal quality is degraded. Therefore, not only the improvement of the head-medium system,
There is a need for measures to improve signal quality in a demodulation circuit system for a reproduction signal from a magnetic disk drive.

[0003]

2. Description of the Related Art FIG. 5 shows the configuration of a demodulation circuit in a conventional magnetic disk drive. The signal reproduced by the magnetic head 70 is amplified by a preamplifier 71 arranged near the magnetic head 70, and then amplified by an automatic gain control circuit (AG
C) 72, waveform equalization circuit 73, and low-pass filter
After passing through the (LPF) 74, the amplitude reaches a point P with sufficient amplitude for performing demodulation processing.

At point P, this signal is split into two
One to the differentiation circuit 75 and the other to the level slice circuit 77
Led to. The signal differentiated by the differentiating circuit 75 is input to a zero-cross detecting circuit 76 via a point Q, and a zero-cross pulse corresponding to the zero-cross point of the differentiated waveform is generated. The pulse from the zero-cross detection circuit 76 is input to the gate circuit 78 through the R point. On the other hand, the output at the point P input to the level slice circuit 77 is compared with a predetermined slice level voltage, and when the signal waveform exceeds the level slice voltage, the level slice gate is opened and the signal is level sliced. The signal is input from the circuit 77 to the gate circuit 78 via the point S. The gate circuit 78 removes a pulse irrelevant to the peak of the reproduction signal using the zero-cross pulse and the gate signal, and generates pulsed data corresponding to the peak point of the analog signal input from the preamplifier 71.

FIGS. 6A to 6E show waveforms at points P to T shown in the conventional demodulation circuit of FIG. Fig. 6 (a)
FIG. 6 shows a reproduced waveform signal of the head 70 at the point P.
(b) is a waveform at point Q, which is a differential waveform of the waveform at point P. Therefore, this waveform includes a pulse irrelevant to the peak of the reproduced waveform. FIG. 6C shows the output waveform of the zero-crossing detection circuit 76 at the point R, which is a signal that goes high when the waveform at the point Q passes through the zero point. FIG. 6D shows the output waveform of the level slice circuit 77 at the point S. The slice level voltages VH and VL (shown in FIG. 6A) set in the level slice circuit 77 are reproduced at the point P. It goes to high level "H" when the signal exceeds. FIG. 6E shows the gate circuit 78 at the point T.
5 shows demodulated data from which pulses irrelevant to the peak of the reproduced waveform have been removed.

As described above, according to the conventional circuit, pulsed data corresponding to the peak of the reproduced signal can be obtained.

[0007]

However, in the conventional demodulation circuit configured as described above, the recording data "
The peak corresponding to 1 "is detected by the level slice and the zero-cross point detection of the differentiated waveform. Therefore, when the waveform is observed, such as the central peak in a 3-bit pattern, it is expected to be a peak from the front-back relationship. However, if the data is below the slice level, the signal may not be recognized as a peak and a demodulation error may occur.
There is a problem that even if a portion corresponding to the above exceeds the slice level, it is recognized as a peak.
As described above, in the conventional demodulation method, if the level of the portion corresponding to "0" of the recording data exceeds the level of the peak corresponding to "1" of the recording data, demodulation cannot be performed. Therefore, when performing higher density recording,
The quality of the signal to be handled is increasingly degraded, causing a reduction in S / N and a decrease in resolution.

The present invention solves the problem in the method of demodulating a regular reproduction signal in the reproduction signal as in the above-mentioned conventional magnetic recording / reproducing apparatus, and makes the reproduction signal "1" lower than the conventional slice level. Crossing the signal or slice level "
Even if the "0" signal is included, the "1" signal is regarded as "1" and "
An object of the present invention is to provide a highly reliable demodulator of a magnetic recording / reproducing apparatus which can demodulate a "0" signal as "0" using a neural net.

[0009]

FIG. 1 shows the principle configuration of a demodulator of a magnetic recording / reproducing apparatus according to the present invention which achieves the above object. As shown in FIG. 1, the demodulator of the magnetic recording / reproducing apparatus of the present invention is a demodulator of a magnetic recording / reproducing apparatus for demodulating recorded data from a reproduced signal from the magnetic recording medium 1, and is connected in series. comprising a plurality number of delay elements with data bit period the same as or shorter delay amount Td,, it offsets the input reproduced signal every time interval of the delay amount Td, the slow
A delay means 2 for outputting N types of signals , where N is a number greater than the number of delay elements by one, and a sampling means 3 for sampling these N types of signals in parallel at a sampling period Ts N times the delay amount Td. , An input element to which the N sample values are input, and output elements corresponding to a plurality of M types of data patterns that can be taken by the sample values, and only one of the output elements according to the input data pattern. Neural network means 4 learned to output signal
And a reproduction signal output means 5 for demodulating a data pattern in accordance with the M types of output signals from the neural network means 4 and outputting it as a reproduction signal.

[0010] Neural net means 4 described above, the N input elements and L plural hidden elements, M number of output elements,
And a perceptron consisting of a two-stage neuron layer connecting these, and writing the optimum weight of the neuron layer to the storage means by learning before shipment of the magnetic recording / reproducing apparatus, and transferring the weight data from the storage means during operation to the neuron layer. To perform demodulation.

The demodulator of the magnetic recording / reproducing apparatus performs learning a plurality of times by using a plurality of learning patterns at the time of operation of the apparatus or periodically at intervals of time. The weight data of the neuron layer stored in the storage means may be rewritten with the weight data. Further, the demodulator of the magnetic recording / reproducing device is
A period Ts ′ for sampling N kinds of signals in parallel is represented by T
When s ′ = Td * (NC), N samples are input to the perceptron while overlapping by a predetermined number of samples C , and when the data pattern of the previous and current overlapping portions matches, the previous demodulation was performed. A configuration may be adopted in which a data pattern of a portion where reproduction output does not overlap is output as a correct reproduction signal.

[0012]

According to the demodulator of the magnetic recording and reproducing apparatus of the present invention,
Reproduction signal from the magnetic recording and reproducing apparatus, a plurality number of delay elements having the same data bit period or a short delay Td,, shifted by a time interval of the delay amount Td, the delay element
The signals are output as N types of signals that are one more than the number , sampled in parallel with a sample period Ts N times the delay amount Td, and input to the neural network. In the neural network, the state of the input N sample values is determined according to a plurality of M types of data patterns that the sample values can take, and one of the output elements corresponding to the determined data pattern is determined. Only a signal is output, and a data pattern is demodulated in accordance with M types of output signals from the neural network and output as a reproduced signal.

[0013]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 2 shows the configuration of an embodiment of the demodulator of the magnetic recording / reproducing apparatus of the present invention. The present invention is a demodulator of a magnetic recording / reproducing apparatus which demodulates recording data from a reproducing signal from a magnetic recording medium. Hereinafter, a reproducing signal of a magnetic disk apparatus in which positive and negative peaks always appear alternately in the reproducing signal will be described. bet it will be described as an example. Here, a case where demodulation is performed based on N consecutive sample values is shown, which is obtained from the characteristics of magnetic recording and conditions such as codes used.
A case where the total number of states that N consecutive sample values can take is M is shown.

In FIG. 2 , reference numeral 21 denotes a delay element sequence,
N pluralities with delay time τ = Td (Td is a data period)
Of delay elements. Reproduction signal from the magnetic disk device (not shown) to be input to the delay element row 21, the N signals with shifted by a time delay Td from each other
And output from the delay element sequence 21. The total N of the reproduced signal and the N signals shifted from each other by the time Td.
+1 signal as N + 1 samples, and N + 1 inputs element 22, and L plural hidden elements 23, M + 1
It is input to a perceptron 20 including a plurality of output elements 24 and two stages of neuron layers 26 and 27 connecting these. Number M + 1 of output elements 24 of perceptron 20
Is the number of patterns that can be taken by N + 1 sample values of the reproduction signal from the magnetic recording medium. The perceptron learns each neuron layer 2 so that when a certain data pattern is input to the input element 22, an output signal is output to the corresponding output element 24 by learning described below.
Weights 6 and 27 are determined.

Each neuron layer 2 of the perceptron 20
6 _and 27 have weights represented by Vij _and Wij, respectively, and the values Yp Zq of the hidden element 23 and the output element 24 are expressed by the following equations. Yp = σ (ΣWkp · Xk) Zq = σ (ΣVpq · Yp) where Σ is the sum of k = 0 to N, and 0 ≦ p ≦ L _, 0
It is assumed that ≦ q ≦ L. Then, σ (S) is defined by the following equation.

Σ (S) = 1 / (1 + exp (−S)) Now, let X be a set of learning patterns, h (X) be the output of the perceptron, and d (X) be the target function. It is expressed as follows. h (X) = (h0 (X), h1 (X), h2 (X),... hq (X),... hM (X))... d (X) = (d0 (X), d1 (X), d2 (X),... dq (X),... dM (X)) where hq (X) = Zq and the target function dq (X) is a learning pattern of X Of which
"1" is output for the learning pattern of "1", and "0" is output for other learning patterns.

Dq (X) = 1 (when learning pattern of q)
dq (X) = 0 (for other learning patterns)
'The state of the hidden element 23 when the learning pattern X is given is represented by f (X), and is represented by the following equation. f (X) = (f ₀ (X), f ₁ (X), f ₂ (X),... f _q (X),... f _L (X)). -d (X)} as ² is minimized, the procedure of determining a hidden element 23 weight Vpq neuron layer 27 between the output element _24, and the weight Wkp neuron layer 26 between the input element 22 and the hidden element 23 Is shown in FIG.

In step 301, the weight Vpq of the neuron layer 27 and the weight Wkp of the neuron layer 26 are arbitrarily determined.
Then, in step 302, p = 0, 1, 2,..., L, q = 0, 1, 2, 2, using the equation: Vpq = Vpq−ΔΣδ ₂ q (X) · fp (X)
, M, the weight V between the hidden element 23 and the output element 24
.., N, p = 0, 1, 2,..., using the equation: Wkp = Wkp−ΔΣδ ₁ p (X) · X
, L, the weight W between the hidden element 23 and the input element 22
Update kp. Where δ ₁ p (X) and δ ₂ q
(X) is represented by the following formula, and Σ is a sum of q = 0 to M.

Δ ₁ p (X) = ｛Σδ ₂ q (X) · Vpq｝ · fp (X) · (1−fp (X)) δ ₂ q (X) = ｛hq (X) −dq ( X)｝ hq (X) · (1−hq (X)) In the following step 304, it is determined whether or not the error {h (X) −d (X)} ² is sufficiently small. Returns to step 302 and repeats steps 302 and 303. When it is sufficiently small (YES), this routine ends in step 305.

Thus, in the above embodiment, N + 1
The optimum weights Wkp and Vpq of the neuron layer 26 and the neuron layer 27 are learned so that when a pattern having a number of sample values is input to the perceptron 20, an output is output to the output element 24 corresponding to the pattern one by one. Ask by. The optimum weights Wkp and Vpq of the neuron layers 26 and 27 are determined by learning before shipment of the magnetic recording / reproducing apparatus, and are determined by ROM.
The weight data is supplied from the ROM to the neuron layers 26 and 27 during operation.

As described above, since an output corresponding to the input data pattern is output from the output element 24 of the perceptron 20 on a one-to-one basis, the demodulated data output circuit 25 connected to the output element 24 corresponds to the location of the output element 24. The demodulated data thus prepared is prepared, and a data pattern corresponding to the output signal can be output as demodulated data (reproduced signal). The determination by learning the weights Wkp and Vpq of the neuron layers 26 and 27 of the perceptron 20 is performed not only before shipment of the magnetic recording / reproducing apparatus, but also during operation of the apparatus or periodically at time intervals. Learning may be performed a plurality of times according to the learning pattern described above, and the weight data of the neuron layer stored in the RAM or the like may be rewritten with the weight data of the neuron layer obtained by this learning routine.

FIG. 4 shows the configuration of a demodulator of a magnetic recording / reproducing apparatus according to another embodiment of the present invention. The same components as those of the demodulator of FIG. Omitted. The demodulator of FIG. 4 differs from the demodulator of FIG. 2 in that, in order to increase the reliability, demodulated data determined at a certain sampling clock fs having a period shorter than a sample period Ts which is N times the data period Td; The point is that it is determined whether or not the data sequence of the overlapping portion of the demodulated data determined at the next sampling clock fs matches.

Therefore, in the apparatus of the embodiment shown in FIG. 4, N + 1 sample values are input to the input element 22 of the perceptron 20 for each sampling clock fs. One of the demodulated data determined by the sampling clock fs is delayed by the delay circuit 41 by a predetermined time U of the data period Td, and
2 and the other is directly input to the comparison circuit 42.
As a result, demodulated data different by the time U × Td is input to the comparison circuit 42, and the number of samples is set to N +
If it is set to 1 , data of (N + 1) -U among the two demodulated data will be inputted in duplicate. Comparison circuit 42
Determines whether or not the overlapped data matches, and outputs only non-overlapping data of the demodulated data input through the delay circuit 41 as correct demodulated data only when they match, and outputs an error when they do not match. Outputs signals and the like.
Note that the number U of overlapping samples can be adjusted.

As a result, the demodulator of FIG. 4 always outputs the demodulated data while checking the value, thereby increasing the reliability of the data value.

[0025]

As described above, according to the present invention, even if the resolution of a reproduced signal from a magnetic recording medium is reduced by high-density recording or the signal is degraded by noise,
Even if the level of the peak value in the reproduced signal fluctuates, the "1" signal is regarded as "1" and the "0" signal is regarded as "0", so that the signal can be demodulated with high reliability.

[Brief description of the drawings]

FIG. 1 is a principle configuration diagram showing a configuration of a demodulator of a magnetic recording / reproducing apparatus of the present invention.

FIG. 2 is a configuration diagram showing a configuration of an embodiment of a demodulator of the magnetic recording / reproducing apparatus of the present invention.

FIG. 3 is a flowchart showing an operation of a demodulator of the magnetic recording / reproducing apparatus of FIG.

FIG. 4 is a configuration diagram of another embodiment of the demodulator of the magnetic recording / reproducing apparatus of the present invention.

FIG. 5 is a demodulation circuit diagram of a conventional magnetic disk drive.

FIG. 6 is a waveform diagram showing operation waveforms of various parts of the conventional circuit of FIG.

[Explanation of symbols]

 Reference Signs List 20 perceptron 21 delay element sequence 22 input element 23 hidden element 24 output element 25 demodulation data output circuit 26 neuron layer 27 neuron layer 41 delay circuit 42 comparison circuit

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Takenori Oshima 1015 Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture Inside Fujitsu Limited (56) References JP-A-4-216343 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G11B 20/10 321

Claims (4)

(57) [Claims] 1. A demodulator for a magnetic recording / reproducing apparatus for demodulating recorded data from a reproduced signal from a magnetic recording medium (1), wherein the demodulator has a delay Td equal to or shorter than a data bit cycle connected in series. comprising a plurality of delay elements, it offsets the input reproduced signal every time interval of the delay amount Td, the delay requirements
A delay means (2) for outputting N types of signals , where N is a number one greater than the prime number, and a sample period T of N times the amount of delay Td
sampling means (3) for sampling in parallel at s; an input element to which the N sample values are input; and output elements corresponding to a plurality of M types of data patterns that can be taken by the sample values. A neural network means (4) trained to output an output signal to only one of the output elements in accordance with the pattern; and demodulating and reproducing a data pattern in accordance with M kinds of output signals from the neural network means. A demodulator for a magnetic recording / reproducing apparatus, comprising: reproduction signal output means (5) for outputting as a signal. 2. The neural network means (4) comprises a perceptron comprising N input elements, L plural hidden elements, M output elements, and a two-stage neuron layer connecting these elements. 2. The method according to claim 1, wherein the optimum weight of the neuron layer is written into the storage means by learning before shipment of the magnetic recording / reproducing apparatus, and the demodulation is performed by giving the weight data from the storage means to the neuron layer during operation. 3. A demodulator for a magnetic recording and reproducing apparatus according to claim 1. 3. Learning is performed a plurality of times by a plurality of learning patterns at the time of operation of the apparatus or periodically at intervals of time, and the weighting data of the neuron layer obtained by the learning routine is stored in the storage means. 3. The demodulator according to claim 2, wherein the weight data of the neuron layer is rewritten. 4. A period Ts 'for sampling N types of signals in parallel is defined as Ts' = Td * (N-C), and the N samples are overlapped by a predetermined number C of samples. 4. The apparatus according to claim 1, wherein a data pattern of a non-overlapping portion of a previously demodulated reproduction output is output as a correct reproduction signal when the data pattern is inputted to a perceptron and the data pattern of the previous and current overlapping portions coincides with each other. The demodulator of the magnetic recording / reproducing device according to any one of the above.