JP2012174310A - Signal detection device, and signal detection method - Google Patents

Abstract

An object of the present invention is to enable signal detection of each track correctly even when the response of inter-track interference is time-varying.
There is provided a signal detection device for detecting a recording signal recorded on each track using a TDPR filter from a plurality of reproduction signals read from the plurality of tracks by a head having a width wider than the track width. The signal detection apparatus sets a tap coefficient generation unit that generates a tap coefficient of a TDPR filter using a preamble and a postamble of a reproduction signal, sets the tap coefficient in each state of a trellis diagram, and sets two-dimensional max-log- A survivor path detected by the MAP detection, and a survivor path detector that calculates an error between a replica signal corresponding to each survivor path and an output signal from the TDPR filter, and a survivor path detected by the survivor path detector And a tap coefficient updating unit that updates the tap coefficient based on the error corresponding to.
[Selection] Figure 11

Description

  The present invention relates to a signal detection device and a signal detection method. In particular, the present invention relates to a multi-track signal detection method for detecting a signal written to each track by shingled recording based on signals simultaneously read from a plurality of adjacent tracks.

  In recent years, research for improving the recording density of magnetic recording disks has been actively conducted. As a method of improving the recording density of the magnetic recording disk, for example, there is a method of increasing the number of tracks per disk by narrowing the track width of each track. However, when the track width is narrowed, it is necessary to reduce the size of the read head in order to read the information of each track with high accuracy. If possible, it is preferable that the width of the read head be approximately the same as the track width. However, it is said that the miniaturization of the read head has already reached its limit. For this reason, attention has been focused on a technique for accurately reading out the signal of each track using a read head having a width wider than the track width.

  If the width of the read head is wider than the track width, when the read head scans a certain track, the signals of the adjacent tracks are read simultaneously. Signals that are read simultaneously from adjacent tracks interfere with the signal of the track that is originally desired to be read. Therefore, a technique (hereinafter referred to as a multitrack signal detection method) that suppresses this interference and separates the signals of each track is required. The multitrack signal detection method is described in Non-Patent Document 1, for example. The method of this document suppresses inter-track interference using a TDPR filter, a one-dimensional max-log-MAP detector, an error correction decoder, a bit error detector, a soft interference replica generator, and a multi-track soft interference canceller. It is.

M.M. Fujii and N.J. Shinohara, “Multi-track iterative ITI canceller for shingled write recording”, IEICE, Magnetic Recording / Information Storage Study Group, IEICE Technical Report, MR2010-44, pp. 11-28. 15-22 December 2010

  The multi-track signal detection method described in this document is based on the premise that bit synchronization is established between tracks when signals are written to each track. Under this assumption, writing frequency synchronization is established between tracks. In other words, the response of inter-track interference is also time-invariant. Therefore, even if signal detection is performed using tap coefficients of various filters obtained in advance, the signal of each track can be detected with high accuracy. However, when bit synchronization is not established between tracks, the signal detection of each track cannot be performed correctly when the multitrack signal detection method described in the document is applied.

  Therefore, the present invention has been made in view of the above problems, and the object of the present invention is to correctly detect the signal of each track even when the response of inter-track interference is time-varying. A new and improved signal detection apparatus and signal detection method are provided.

  In order to solve the above-described problem, according to one aspect of the present invention, a two-dimensional partial response filter is used from a plurality of reproduction signals read from a plurality of tracks using a head having a width larger than the track width. Thus, a signal detection device for detecting a recording signal recorded on each track is provided. The signal detection device includes: a tap coefficient generation unit that generates a tap coefficient of the two-dimensional partial response filter using a preamble and a postamble of the reproduction signal; and a trellis diagram illustrating the tap coefficient generated by the tap coefficient generation unit. In each state, a survivor path of each state is detected by two-dimensional max-log-MAP detection, and an error between a replica signal corresponding to each survivor path and an output signal from the two-dimensional partial response filter is calculated. A surviving path detecting unit, and a tap coefficient updating unit that updates the tap coefficient of the two-dimensional partial response filter based on the error corresponding to the surviving path detected by the surviving path detecting unit.

  When writing is performed synchronously between tracks, the inter-track interference does not change with time, so there is no need to update the tap coefficient of the two-dimensional partial response filter. However, when writing asynchronously between tracks, the inter-track interference becomes time-varying, so the two-dimensional partial response filter that optimizes multiple track signals adaptively taps against fluctuations in inter-track interference. It is required to optimize the coefficients. According to the above configuration, since the tap coefficient is updated according to the surviving path selected by the two-dimensional max-log-MAP detection, it is suitable for the channel response of inter-track interference and the channel response of inter-symbol interference. It becomes possible to set the tap coefficient. As a result, it is possible to correctly detect the signal of each track even when the inter-track interference is time-varying.

  In addition, the signal detection device includes an error correction unit that performs error correction of the recording signal using a log likelihood ratio of each bit calculated when the two-dimensional max-log-MAP detection is performed, and the preamble And a channel estimation unit for estimating a first channel response vector representing intersymbol interference and a second channel response vector representing intertrack interference using the postamble, and the first and second channel response vectors. And an interference replica generation unit that generates an interference replica indicating an interference component included in the reproduction signal, and an interference canceller that subtracts the interference replica from the reproduction signal and cancels the interference component. Good. In this case, the channel estimation unit updates the first and second channel response vectors according to the surviving path. With this configuration, it is possible to adaptively estimate the interference channel response required when the inter-track interference becomes time-varying.

  Further, the surviving path detection unit is configured to estimate a square value of an error between the replica signal and an output signal from the two-dimensional partial response filter by a moving average corresponding to the surviving path in each state. May be. With this configuration, it is possible to adaptively estimate the noise variance value of the two-dimensional partial response filter output required when the inter-track interference becomes time-varying.

  Further, when the recording signal corresponding to at least one track is correctly decoded, the surviving path detection unit adds the trellis to the delay bits of the remaining reproduction signals corresponding to tracks other than the track where the recording signal is correctly decoded. All states of the diagram may be allocated and a surviving path of each state may be detected by one-dimensional max-log-MAP detection. With this configuration, the number of write signal replica candidates used for channel estimation increases, and channel estimation accuracy can be improved.

  Further, the signal detection apparatus estimates a first phase shift from a channel response of intersymbol interference to a channel response of intertrack interference using the preamble, and uses the postamble to determine a channel response of intersymbol interference. A phase shift estimator for estimating a second phase shift from the first phase shift to the channel response of the inter-track interference, and a write clock frequency between the tracks from the phase shift between the first phase shift and the second phase shift. A frequency difference estimation unit that estimates a difference between the clock frequency and a time constant setting unit that sets a time constant used to update at least the tap coefficient based on the clock frequency. When the inter-track interference becomes time-varying, there is a drift of the write clock frequency. With this configuration, it is possible to appropriately set the time constant parameter according to the drift. Further, since the time constant optimized for the clock frequency difference within a range that can be assumed in advance can be used for signal detection, improvement in signal detection accuracy can be expected.

  In order to solve the above problem, according to another aspect of the present invention, a two-dimensional partial response is obtained from a plurality of reproduction signals read from a plurality of tracks using a head having a width larger than the track width. A signal detection method for detecting a recording signal recorded on each track using a filter is provided. The signal detection method includes a tap coefficient generation step of generating a tap coefficient of the two-dimensional partial response filter using a preamble and a postamble of the reproduction signal, and a trellis diagram showing the tap coefficient generated in the tap coefficient generation step. In each state, a survivor path of each state is detected by two-dimensional max-log-MAP detection, and an error between a replica signal corresponding to each survivor path and an output signal from the two-dimensional partial response filter is calculated. A surviving path detecting step, and a tap coefficient updating step of updating a tap coefficient of the two-dimensional partial response filter based on the error corresponding to the surviving path detected in the surviving path detecting step.

  When writing is performed synchronously between tracks, the inter-track interference does not change with time, so there is no need to update the tap coefficient of the two-dimensional partial response filter. However, when writing asynchronously between tracks, the inter-track interference becomes time-varying, so the two-dimensional partial response filter that optimizes multiple track signals adaptively taps against fluctuations in inter-track interference. It is required to optimize the coefficients. According to the above configuration, since the tap coefficient is updated according to the surviving path selected by the two-dimensional max-log-MAP detection, it is suitable for the channel response of inter-track interference and the channel response of inter-symbol interference. It becomes possible to set the tap coefficient. As a result, it is possible to correctly detect the signal of each track even when the inter-track interference is time-varying.

  As described above, according to the present invention, it is possible to correctly detect the signal of each track even when the response of inter-track interference is time-varying.

It is explanatory drawing which showed the structure of the signal processing apparatus. It is the functional block diagram which showed the structure of the recording signal production | generation part. It is the functional block diagram which showed the structure of the recording signal production | generation part. It is the functional block diagram which showed the structure of the reproduction | regeneration signal processing part. It is the functional block diagram which showed the structure of the multitrack signal detection part. It is explanatory drawing for demonstrating the state by which information was written on the medium by asynchronous writing between tracks. It is explanatory drawing for demonstrating the operation | movement which reads information from the track | truck of the state in which the information was written in the medium by asynchronous writing between tracks. It is explanatory drawing for demonstrating the interference (delay) between time-varying tracks. It is explanatory drawing for demonstrating the interference (advance) between time-varying tracks. It is explanatory drawing for demonstrating the format of information. It is the functional block diagram which showed the structure of the multitrack signal detection part. It is a functional block diagram showing a configuration of means for setting a time constant used in the multitrack signal detection unit. It is explanatory drawing which showed the structure of the TDPR filter. It is explanatory drawing which showed the survival path on the trellis diagram which has 1 delay bit of 2 track signals as a state. It is explanatory drawing which showed the equivalent model with respect to 2 sets of TDPR filters. It is a trellis diagram in an equivalent model for two sets of TDPR filters. It is explanatory drawing which showed the equivalent model with respect to 1 set of TDPR filters. FIG. 3 is a trellis diagram in an equivalent model for a set of TDPR filters. It is the graph which compared the BER characteristic.

  Exemplary embodiments of the present invention will be described below in detail with reference to the accompanying drawings. In addition, in this specification and drawing, about the component which has the substantially same function structure, duplication description is abbreviate | omitted by attaching | subjecting the same code | symbol.

  In the present specification and drawings, a plurality of components having substantially the same functional configuration may be distinguished by adding different alphabets after the same reference numeral. For example, a plurality of configurations having substantially the same functional configuration are distinguished as the buffer 22A and the buffer 22B as necessary. However, when it is not necessary to particularly distinguish each of a plurality of constituent elements having substantially the same functional configuration, only the same reference numerals are given. For example, when it is not necessary to distinguish between the buffer 22A and the buffer 22B, they are simply referred to as the buffer 22.

<1. Basic Configuration of Signal Processing Device>
First, the basic configuration of the signal processing apparatus 1 having functions as a signal reproducing apparatus and a signal recording apparatus will be described with reference to FIG. FIG. 1 is an explanatory diagram showing the configuration of the signal processing apparatus 1. As shown in FIG. 1, the signal processing apparatus 1 includes a recording medium 4, a spindle motor 6, a head 8, a recording signal generation unit 10, and a reproduction signal processing unit 20.

  The recording medium 4 is mounted on a spindle motor 6 and is rotated by the spindle motor 6. The recording medium 4 may be a magnetic disk on which a plurality of tracks are formed as shown in FIG. Although only one magnetic disk is shown as the recording medium 4 in FIG. 1, the signal processing device 1 may have a plurality of magnetic disks.

  The head 8 functions as a recording head that records the recording signal supplied from the recording signal generation unit 10 on the recording medium 4 and a reproducing head that reads out the reproducing signal from the recording medium 4. Moreover, since the head 8 is larger than the track width of the recording medium 4 as will be described later, the recording medium 4 is scanned over a plurality of tracks.

  The recording signal generation unit 10 generates a recording signal for recording on the recording medium 4. Further, the reproduction signal processing unit 20 processes the reproduction signal read from the recording medium 4 by the head 8. The detailed configuration of the recording signal generation unit 10 will be described with reference to FIG. 2, and the detailed configuration of the reproduction signal processing unit 20 will be described with reference to FIGS. 4 and 5.

  The signal processing apparatus 1 described above includes, for example, a PC (Personal Computer), a home video processing apparatus (DVD recorder, VCR, etc.), a mobile phone, a PHS (Personal Handyphone System), a portable music playback apparatus, and a portable video. You may provide in a processing apparatus, PDA (Personal Digital Assistants), a home game device, a portable game device, a household appliance, etc. FIG.

<2. Configuration of Recording Signal Generation Unit 10>
FIG. 2 is a functional block diagram showing the configuration of the recording signal generation unit 10. As illustrated in FIG. 2, the recording signal generation unit 10 includes a CRC encoder 12, an LDPC encoder 14, an interleaver 16, and a recording signal converter 18. The CRC is an abbreviation for Cyclic Redundancy Check. The LDPC is an abbreviation for Low Density Parity Check.

  The CRC encoder 12 adds an error detection code such as a CRC code to the supplied recording information data. The LDPC encoder 14 is an error correction encoding unit for ensuring a predetermined error rate even with a low S / N ratio. The interleaver 16 interleaves the data processed by the LDPC encoder 14 with a different pattern for each track. The recording signal converter 18 converts the data after interleaving by the interleaver 16 into a magnetic polarity signal to generate a recording signal. The recording signal generated by the recording signal converter 18 is recorded on the recording medium 4 by the head 8.

<3. Scanning control of head 8>
Next, scanning control of the head 8 will be described. As described above, the head 8 is larger than the track width of the recording medium 4. Therefore, the head 8 scans the recording medium 4 a plurality of times in a state where the head 8 extends over a plurality of tracks. However, the signal processing apparatus 1 controls the position of the head 8 so that the head 8 does not protrude from a track that is not a detection target of the reproduction signal during the multiple scans. Hereinafter, a more specific description will be given with reference to FIG. However, the example of FIG. 3 is a case where bit synchronization is established between tracks at the time of signal writing. Note that the operation of the head 8 is substantially the same even when bit synchronization is not established between the tracks at the time of signal writing.

  FIG. 3 is an explanatory diagram showing how the head 8 scans the recording medium 4. More specifically, FIG. 3 shows a scanning example in the case where reproduction signals from two tracks (track # 1 and track # 2) are to be detected. In this case, the signal processing apparatus 1 first controls the head 8 to scan the recording medium 4 at a position P1 spanning the track # 1 and the track # 2, and after one rotation of the recording medium 4, the signal processing apparatus 1 controls the track # from the position P1. Control is performed so that the head 8 scans the recording medium 4 at a position P2 on the second side.

  Here, the signal processing apparatus 1 applies to tracks other than the track # 1 and the track # 2 that are the detection targets of the reproduction signal, that is, the track below the track # 1 and the track above the track # 2 in FIG. The position of the head 8 is controlled so that the head 8 does not protrude. According to such a configuration, it is possible to suppress the case where the head 8 reads the interference signal from a track other than the detection target, so that it is possible to improve the signal detection accuracy by the reproduction signal processing unit 20 described later.

  In the present specification, the description will be given with an emphasis on an example in which the reproduction signals of two tracks are simultaneously detected, but the present embodiment is not limited to such an example. For example, it is possible to simultaneously detect the reproduction signals of three or more tracks. For example, when the reproduction signals from three tracks # 1 to # 3 are to be detected, the signal processing apparatus 1 moves the head 8 for each rotation of the recording medium. However, in the signal processing apparatus 1, the head 8 protrudes from a track other than the tracks # 1 to # 3 (for example, the track below the track # 1 and the track above the track # 3) that is the detection target of the reproduction signal. The position of the head 8 is controlled so as not to occur. With this configuration, it is possible to improve the accuracy of signal detection by the reproduction signal processing unit 20 described later, as in the case where there are two tracks to be detected.

<4. Configuration of Reproduction Signal Processing Unit 20>
Next, the configuration of the reproduction signal processing unit 20 will be described in detail with reference to FIGS. 4 and 5. FIG. 4 is a functional block diagram showing the configuration of the reproduction signal processing unit 20. As shown in FIG. 4, the reproduction signal processing unit 20 includes buffers 22 </ b> A and 22 </ b> B and a multitrack signal detection unit 30.

  The buffer 22A holds a signal read from the recording medium 4 at the position P1 by the head 8. Similarly, the buffer 22B holds a signal read from the recording medium 4 at the position P2 by the head 8. Then, the buffer 22A and the buffer 22B simultaneously output the held data to the multitrack signal detection unit 30. With this configuration, simultaneous reading from the position P1 and the position P2 is realized in a pseudo manner.

  Here, at the position P1, the head 8 performs scanning while straddling the track # 1 and the track # 2. Therefore, the signal read from the recording medium 4 by the head 8 at the position P1 is the reproduction signal of the track # 1 and the track #. 2 is a multitrack reproduction signal on which two reproduction signals are superimposed (agreeing with the superimposed reproduction signal; hereinafter referred to as multitrack reproduction signal # 1). Similarly, at the position P2, the head 8 performs scanning while straddling the track # 1 and the track # 2, so that the signal read from the recording medium 4 by the head 8 at the position P2 is also the reproduction signal and track of the track # 1. This is a multi-track playback signal (hereinafter referred to as multi-track playback signal # 2) on which the playback signal of # 2 is superimposed. However, the multi-track reproduction signal # 1 and the multi-track reproduction signal # 2 have different superposition ratios of the reproduction signal of the track # 1 and the reproduction signal of the track # 2 based on the difference between the position P1 and the position P2.

  The multi-track signal detection unit 30 detects the reproduction signal of the track # 1 and the reproduction signal of the track # 2 with high accuracy from the multi-track reproduction signals input from the buffer 22A and the buffer 22B based on the repeated signal detection process. Hereinafter, the configuration of the multitrack signal detection unit 30 will be described with reference to FIG.

  FIG. 5 is a functional block diagram showing the configuration of the multitrack signal detection unit 30. As shown in FIG. 5, the multitrack signal detector 30 includes two-dimensional partial response filters 32A and 32B, Max-log-MAP detectors 34A and 34B, deinterleavers 36A and 36B, and an LDPC decoder 38A. And 38B, hard decision units 40A and 40B, CRC decoders 42A and 42B, switching unit 44, interleavers 46A and 46B, interleavers 48A and 48B, (ISI & ITI) soft replica generators 50A and 50B, Interference cancellers 52A, 52B, 54A and 54B.

  The two-dimensional partial response filter 32 processes the multitrack reproduction signal at a time and performs a process for extracting the reproduction signal of each track from the multitrack reproduction signal. Specifically, the two-dimensional partial response filter 32 is based on the inter-track interference suppression tap coefficient acquired in advance to equalize the reproduction signal of the track m to the target response from the multi-track reproduction signal having inter-track interference, A reproduction signal of track m in which inter-track interference is suppressed is extracted from the multi-track reproduction signal. For example, the two-dimensional partial response filter 32A extracts the reproduction signal of the track # 1 in which the inter-track interference is suppressed, and the two-dimensional partial response filter 32B extracts the reproduction signal of the track # 2 in which the inter-track interference is suppressed.

  The Max-log-MAP detector 34, the deinterleaver 36, and the LDPC decoder 38 function as a bit decoding unit that decodes the reproduction signal extracted by the two-dimensional partial response filter 32 into a bit string.

  For example, the Max-log-MAP detector 34A uses the reproduction signal of the track # 1 extracted by the two-dimensional partial response filter 32 and the signal after interleaving by the interleaver 46A, and performs delay superposition according to these target responses. The likelihood for each bit is calculated from the processed signal by Max-log-MAP detection. Then, the deinterleaver 36A deinterleaves the likelihood obtained by the Max-log-MAP detector 34A. Further, the LDPC decoder 38A decodes the bit string of the track # 1 using the error correction code.

  Similarly, the Max-log-MAP detector 34B uses the reproduction signal of the track # 2 extracted by the two-dimensional partial response filter 32 and the signal after interleaving by the interleaver 46B, and delays according to these target responses. The likelihood for each bit is calculated from the superimposed signal by Max-log-MAP detection. The deinterleaver 36B deinterleaves the likelihood obtained by the Max-log-MAP detector 34B. Further, the LDPC decoder 38B decodes the bit string of the track # 1 using the error correction code.

  The CRC decoder 42 detects the remaining bit error by decoding the error detection code from the bit string after the hard decision by the hard decision unit 40 with respect to the bit string decoded by the LDPC decoder 38.

  The switching unit 44 connects the LDPC decoder 38 to the interleaver 46 and the interleaver 48 when the CRC decoder 42 detects the remaining bit error. With this connection, the decoding result by the LDPC decoder 38 is supplied to the interleaver 46 and the interleaver 48. On the other hand, when no bit error is detected by the CRC decoder 42, the decoded bit string is output to the upper layer as reproduced recording data.

  The interleaver 48 performs interleaving on the decoding result of each track. That is, the interleaver 48 performs the same interleaving as that during recording on the likelihood of each bit obtained by decoding. Specifically, interleaver 48A interleaves the decoding result of track # 1, and interleaver 48B interleaves the decoding result of track # 2.

  The soft replica generator 50 converts the interleaved bit likelihood into a soft replica using a hyperbolic tangent function, and generates an inter-track interference soft replica and an inter-symbol interference soft replica based on the soft replica. For example, the soft replica generator 50A generates an intersymbol interference soft replica of the track # 1 by multiplying the soft replica related to the track # 1 by the intersymbol interference tap coefficient vector estimated in advance, and is estimated in advance. The inter-track interference tap coefficient vector is multiplied to generate an inter-track interference soft replica from track # 2 for track # 1.

  Similarly, the soft replica generator 50B multiplies the soft replica related to the track # 2 by the intersymbol interference tap coefficient vector estimated in advance to generate the intersymbol interference soft replica of the track # 2, and is estimated in advance. The inter-track interference tap coefficient vector is multiplied to generate an inter-track interference soft replica from track # 1 for track # 2.

  Interference cancellers (subtraction units) 52 and 54 subtract the interference soft replica generated by the soft replica generator 50 from the multitrack reproduction signal. For example, when detecting the signal of the track # 1, the interference canceller 52A soft cancels (subtracts) the inter-track interference soft replica for the track # 1 from the multitrack playback signal # 1, and the interference canceller 54A starts from the multitrack playback signal # 2. Soft cancel the intersymbol interference soft replica of track # 2.

  Thereby, since the signal of track # 1 remains in any multitrack reproduction signal, the two-dimensional partial response filter 32A equalizes the multitrack reproduction signal after the soft cancellation by the interference cancellers 52A and 54A.

  Here, the two-dimensional partial response filter 32A uses a tap coefficient after inter-track interference removal that is different from the first iteration when performing equalization processing on the repeated signal. This inter-track interference removal tap coefficient equalizes the playback signal of track m from the multi-track signal obtained by removing track signals other than track m from the multi-track playback signal, and can be obtained by prior training. It is.

  Similarly, when detecting the signal of track # 2, the interference canceller 54B soft cancels the inter-track interference soft replica for the track # 2 from the multitrack playback signal # 2, and the interference canceller 52B detects the track # 2 from the multitrack playback signal # 1. Soft cancel 1 intersymbol interference soft replica.

  As a result, since the signal of track # 2 remains in any multitrack playback signal, the two-dimensional partial response filter 32B removes the above-described intertrack interference from the multitrack playback signal after soft cancellation by the interference cancellers 52B and 54B. It is possible to equalize using a post-tap coefficient.

  The multi-track signal detection unit 30 performs the above-described interference cancellation processing using the interference soft replica, together with the bit string decoding processing by the Max-log-MAP detector 34, the deinterleaver 36, and the LDPC decoder 38, the CRC. Repeat until no bit errors are detected by the decoder 42.

  Even at the initial stage of repetition, for example, when detecting the signal of track # 1, inter-track interference with respect to track # 1 and inter-code interference with track # 2 can be greatly suppressed, so that signal quality is improved. When the signal detection accuracy is improved by this effect, the bit likelihood increases, and the soft replica becomes asymptotic to the hard replica. As a result, in the next iterative process, the inter-track interference with respect to the track # 1 and the inter-code interference with the track # 2 are further improved. At the same time, the useful signal component leaking to the adjacent track can be used for signal detection.

<5. Time-varying inter-track interference>
When bit synchronization is established between tracks at the time of writing a signal to each track, the signal of each track can be detected correctly by using the multitrack signal detection unit 30 shown in FIG. However, as shown in FIG. 6, when bit synchronization is not established between tracks (hereinafter referred to as inter-track asynchronous writing), the multi-track signal detection unit 30 shown in FIG. I can't. Here, a problem caused by the time variation of the inter-track interference will be described.

  As shown in FIG. 6, a bit phase difference occurs between tracks on a medium written by asynchronous writing between tracks due to a difference in writing clock frequency. As shown in FIG. 7, when the head 8 scans the medium on which the bit phase difference has occurred between the tracks, the bit phase difference gradually changes according to the scanning time of the head 8. That is, when this track is scanned by the wide head 8, the inter-track interference with the light frame as a reference becomes the time-varying inter-track interference.

  When the tracks # 1 and # 2 shown in FIG. 7 are scanned, the inter-track interference from the track # 2 to the track # 1 is gradually delayed as shown in FIG. 8 (time change of impulse response). . On the other hand, the inter-track interference from the track # 1 to the track # 2 gradually proceeds as shown in FIG. 9 (time change of impulse response). For the multi-track signal detection unit 30 that collectively processes multi-track signals reproduced from a plurality of adjacent tracks, time-varying inter-track interference occurs. For this reason, the multi-track signal detection unit 30 assuming a time-invariant channel response cannot correctly detect the signal of each track.

<6. Embodiment>
Therefore, the inventor of the present invention devised a configuration of the multitrack signal detection unit 100 shown in FIG. Hereinafter, the configuration of the multitrack signal detection unit 100 according to an embodiment of the present invention will be described. In the present embodiment, recording data having a data format as shown in FIG. 10 is used. As shown in FIG. 10, the recording data includes a preamble, user data, a parity bit of a CRC code, a parity bit of an LDPC code, and a postamble. Further, the recording signal is generated by the recording signal generator 10 shown in FIG.

[Configuration of Multitrack Signal Detection Unit 100]
As shown in FIG. 11, the multitrack signal detection unit 100 includes a channel estimation unit 101, a TDPR filter tap coefficient generator 102, a TDPR filter output noise variance initial estimation unit 103, a TDPR filter output noise variance update unit 104, A tap coefficient controller 105, a channel tracking unit 106, a multi-track soft interference canceller 107, and a TDPR filter 108 (two-dimensional partial response filter).

  Further, the multitrack signal detection unit 100 switches between a one-dimensional / two-dimensional max-log-MAP detector 110, a deinterleaver 111, an LDPC decoder 112, a hard decision unit 113, and a CRC decoder 114. Unit 115, write signal soft replica converter 116, and interleavers 117 and 118.

  The channel estimation unit 101 estimates a channel matrix using a preamble and a postamble included in the reproduction signal. The TDPR filter tap coefficient generator 102 generates a tap coefficient used by the TDPR filter 108. The TDPR filter output noise variance initial estimation unit 103 estimates an initial value of the TDPR filter output noise variance value used in the one-dimensional / 2-dimensional max-log-MAP detector 110. The TDPR filter output noise variance updating unit 104 sets a max-log at the next time based on a survival path (described later) output from the one-dimensional / two-dimensional max-log-MAP detector 110 and an error corresponding to the survival path. Calculate the TDPR filter output noise variance used for MAP detection.

  The tap coefficient controller 105 sets the tap coefficient generated by the TDPR filter tap coefficient generator 102 in the TDPR filter 108. The channel tracking unit 106 estimates channel responses of ITI (inter-track interference) and ISI (inter-symbol interference) used by the multi-track soft interference canceller 107 for each trellis state. The multi-track soft interference canceller 107 performs soft interference cancellation using a channel response estimated for each channel state when a CRC error is detected by the CRC decoder 114, and generates a residual signal vector (described later). To do. The TDPR filter 108 collectively processes the multitrack reproduction signal and extracts the reproduction signal of each track from the multitrack reproduction signal.

  The one-dimensional / two-dimensional max-log-MAP detector 110 uses the TDPR filter output signal obtained by equalizing two adjacent track reproduction signals by the TDPR filter 108 having the configuration shown in FIG. A branch metric is obtained, a path metric is calculated from the branch metric, and a surviving path for each trellis state is selected. If the CRC decoder 114 determines that there is no error in the reproduction data of one track, the one-dimensional / two-dimensional max-log-MAP detector 110 in FIG. 15 or FIG. One-dimensional max-log-MAP detection is performed on the TDPR filter output signal output by the TDPR filter 108 having the configuration shown.

  The deinterleaver 111 deinterleaves the path metric of the surviving path obtained by the one-dimensional / two-dimensional max-log-MAP detector 110. The LDPC decoder 112 decodes the bit string of each track using the error correction code. The CRC decoder 114 detects the remaining bit error from the bit string after the hard decision by the hard decision unit 113 by decoding the error detection code.

  The switching unit 115 connects the LDPC decoder 112, the interleaver 117, and the write signal soft replica converter 116 when the CRC decoder 114 detects the remaining bit error. With this connection, the decoding result by the LDPC decoder 112 is supplied to the interleaver 117 and the write signal soft replica converter 116. On the other hand, when no bit error is detected by the CRC decoder 114, the decoded bit string is output to the upper layer as reproduction data.

  The write signal soft replica converter 116 generates a soft replica signal of a write signal used for soft interference cancellation by the multitrack soft interference canceller 107. The interleaver 118 interleaves the soft replica signal generated by the write signal soft replica converter 116 and supplies it to the multi-track soft interference canceller 107.

  The configuration of the multitrack signal detection unit 100 has been described above. Next, the configuration of time constant setting means for setting the time constant used in the multitrack signal detection unit 100 will be described.

[Configuration of time constant setting means]
The multitrack signal detection unit 100 receives a time constant for detecting the multitrack signal from the time constant setting unit. This time constant setting means has a configuration as shown in FIG. As shown in FIG. 12, the time constant setting means includes channel estimators 131 and 133, shift amount estimators 132 and 134, a frequency difference estimator 135, an absolute value calculator 136, and an average value calculator 137. And a time constant table 138.

  Channel estimator 131 estimates the ITI channel response vector using the preamble of the reproduced signal. The shift amount estimator 132 estimates the shift amount of the ITI channel response vector estimated by the channel estimator 131 with respect to the ISI channel response vector. Similarly, the channel estimator 133 estimates the ITI channel response vector using the postamble of the reproduction signal. The shift amount estimator 134 estimates the shift amount of the ITI channel response vector estimated by the channel estimator 133 with respect to the ISI channel response vector.

  The frequency difference estimator 135 estimates the clock frequency difference from the difference between the shift amounts estimated by the shift amount estimators 132 and 134. The absolute value calculation unit 136 calculates the absolute value of the difference between the shift amounts estimated by the frequency difference estimator 135. The average value calculation unit 137 averages the difference absolute value of the shift amount calculated for each track by the absolute value calculation unit 136. The average value calculated by the average value calculation unit 137 is stored in the time constant table 138. An optimum time constant is selected from the time constant table 138 and input to the multitrack signal detection unit 100.

  The configuration of the time constant setting unit has been described above.

[Operation of multitrack signal detector]
Next, the operation of the multitrack signal detection unit 100 will be described.

(model)
A reproduction signal vector y (k) = [y ₁ (k), y ₂ (k)] ^T read from two adjacent tracks is modeled by the following equation (1). Here, H (k) is a channel matrix estimated by the channel estimation unit 101. S (k) is a write signal vector. Ζ (k) is medium noise from its own track, and n (k) is white Gaussian noise. The write signal vector s ^T (k) is s ^T (k) = [s ₁ ^T (k), s ₂ ^T (k)]. Further, s _m ^T (k) is s _m ^T (k) = [s _m (k + L),..., S _m (k−L)]. However, 2L + 1 is the channel response length.

The channel matrix H (k) included in the reproduction signal vector model shown in the following equation (1) is expressed by the following equation (2). However, the element vector h _{m, m ′} (k) of the channel matrix H (k) is a vector (hereinafter referred to as ISI vector) representing the ISI channel response when m = m ′, and when m ≠ m ′. Is a vector representing the ITI channel response (hereinafter referred to as ITI vector).

The elements of the element vector h _{m, m ′} (k) are given by the following equation (3). However, when m = m ′, am _{, m ′} = 1, and when m ≠ m ′, am _{, m ′} = α (off-track rate). H (t) is a channel response. Further, the channel index l is l = 0,. The time-varying parameter θ (k) is determined by the initial bit phase difference between tracks and the frequency error of the write clock.

…（１）

…（２）

…（３）

... (1)

... (2)

... (3)

(Path metric calculation)
Reproduction signals read from two adjacent tracks are first equalized by the TDPR filter 108 having the configuration shown in FIG. The tap data vector r (k) of the TDPR filter 108 is r (k) = [y ₁ (k + N), ..., y ₁ (k), ..., y ₁ (k-N), y ₂ (k + N), ... , Y ₂ (k),... Y ₂ (k−N)] ^T. Here, the tap length per track is 2N + 1. Further, the total tap length of the TDPR filter 108 is M _T (2N + 1). However, _{M T} is a multi-track number (in this example 2) represents a.

Now, the transition of trellis states of the state number _{N S} from time k-1 at time k and (S ', S). Also, among the TDPR filter tap coefficients generated by the TDPR filter tap coefficient generator 102 using the preamble and postamble of the reproduction signal, a vector of TDPR filter tap coefficients for ITI suppression is represented by w _{m, S} (k−1). ). The TDPR filter 108 uses the TDPR filter tap coefficient vector w _{m, S} (k−1) and the tap data vector r (k) to output signal z _{m, S ′} (k) is generated.

…（４）

... (4)

However, the subscript _{m of} the TDPR filter tap coefficient vector w _{m, S ′} (k−1) indicates that it corresponds to the track m. That is, the TDPR filter tap coefficient vector w _{m, S} (k−1) represents a set of tap coefficients for extracting the signal of the track m. The subscript S ′ of the TDPR filter tap coefficient vector w _{m, S ′} (k−1) indicates that it corresponds to the trellis state S ′. That is, the TDPR filter tap coefficient vector w _{m, S ′} (k−1) represents a set of tap coefficients in the trellis state S ′. Therefore, M _T N _S TDPR filter tap coefficient vectors w _{m, S} are prepared.

The output signal z _{m, S ′} (k) of the TDPR filter 108 is input to the one-dimensional / 2-dimensional max-log-MAP detector 110. The one-dimensional / 2-dimensional max-log-MAP detector 110 receives the input output signal z _{m, S ′} (k) of the TDPR filter 108 and the replica signal candidates z ′ _{m, S ′, A} square sum of errors from _S (k) is calculated (branch metric synthesis is performed), and a branch metric Γ _{S ′, S} (k) of trellis transition is calculated. The branch metric Γ _{S ′, S} (k) is calculated by the following equation (5).

…（５）

... (5)

However, σ _{m, S ′} ² (k−1) is a TDPR filter output noise variance value estimated by the TDPR filter output noise variance initial estimation unit 103 or a TDPR filter updated by the TDPR filter output noise variance update unit 104 Output noise variance value. The one-dimensional / 2-dimensional max-log-MAP detector 110 calculates a path metric corresponding to the trellis transition from the branch metric Γ _{S ′, S} (k) and the state metric at time k−1, and for each trellis state. Ask for a survival path. The survivor path information and the error of each survivor path are input to the channel tracking unit 106, the tap coefficient controller 105, and the TDPR filter output noise variance update unit 104. Then, the TDPR filter tap coefficient is updated based on each error according to the surviving path.

  The update of the TDPR filter tap coefficient can be realized using an adaptive algorithm such as an RLS (Sequential Least Squares) algorithm. For example, the followability to time-varying ITI can be adjusted using a forgetting factor set by the RLS algorithm. Further, the same processing is executed backward from the postamble. The path metric calculated by the one-dimensional / two-dimensional max-log-MAP detector 110 in this manner is input to the deinterleaver 111, and processing in the subsequent stage of the deinterleaver 111 is executed.

(Interference cancellation by repeated processing)
When it is determined that a bit error remains due to error detection by the CRC decoder 114, interference cancellation by iterative processing is executed. Soft replica signal of the write signal generated by write signals soft replica converter 116 based on the bit string decoded by the LDPC decoder ^{_{112 s 'm T (k)}} = [s' m (k + L E), ..., s _{'m (k + L E)} ] to. However, _2L E +1 is the channel estimation tap length. The channel tracking unit 106 estimates the ITI and ISI channel responses used in the multitrack soft interference canceller 107.

When the partial matrix H ′ _{m, S ′} (k−1) of the channel matrix estimated in the trellis state S ′ at time k−1 is expressed by the following equation (6), interference cancellation by the multitrack soft interference canceller 107 is performed. Is expressed by the following equation (7). Here, m ″ is a value that is 2 when m is 1 and 1 when m is 2. The multitrack soft interference canceller 107 performs interference cancellation expressed by the following equation (7) at a point in time. Run from k-N to time k + N to generate a multitrack residual signal vector r ″ _{m, S ′} (k). The multitrack residual signal vector r ″ _{m, S ′} (k) is input to the TDPR filter 108.

…（６）

…（７）

(6)

... (7)

The TDPR filter 108 equalizes the multitrack residual signal vector r ″ _{m, S ′} (k) generated by the multitrack soft interference canceller 107 as shown in the following equation (8). Output signal of the TDPR filter 108 Is input to the one-dimensional / 2-dimensional max-log-MAP detector 110. The one-dimensional / 2-dimensional max-log-MAP detector 110 uses the output signal of the TDPR filter 108, and the following equation (9 ), Branch metric Γ _{S ′, S} (k) is calculated by branch metric synthesis _{, and} the second term on the right side of the following equation (9) is for each bit obtained by the previous iteration. Likelihood.

…（８）

…（９）

(8)

... (9)

The one-dimensional / 2-dimensional max-log-MAP detector 110 performs path metric calculation and survivor path selection in the two-dimensional max-log-MAP detection. At this time, the one-dimensional / 2-dimensional max-log-MAP detector 110 generates an error signal em _{, S} (k) corresponding to the surviving path in each trellis state S. The survivor path and error selected by the one-dimensional / two-dimensional max-log-MAP detector 110 in this way are input to the channel tracking unit 106, the tap coefficient controller 105, and the TDPR filter output noise variance update unit 104. .

  The channel tracking unit 106 performs ITI and ISI channel estimation (hereinafter referred to as multitrack channel estimation) using a surviving path sequence including a multitrack bit sequence for each trellis state. The tap coefficient controller 105 controls the tap coefficient using an RLS algorithm or the like. Note that the LMS (least mean square) algorithm can be used for updating the multitrack channel estimation, and the followability to ITI can be adjusted using the step size set by this algorithm. The same processing is also performed backward using the postamble.

  Further, according to the surviving path selected for each trellis state, the TDPR filter output noise variance updating unit 104 calculates the TDPR filter output noise variance value from the output signal of the TDPR filter 108 and the error signal corresponding to the surviving path. Calculate the moving average. The moving average value calculated by the TDPR filter output noise variance updating unit 104 in this way is used for max-log-MAP detection by the one-dimensional / two-dimensional max-log-MAP detector 110 at the next time. The

(Specific example of parameter update)
Here, the update of the TDPR filter tap coefficient, the update of the ITI and ISI channel responses, and the update of the TDPR filter output noise variance value will be described with reference to FIG.

FIG. 14 shows a trellis diagram having 1 delay bit (b _{1, k−1} , b _{2, k−1} ) of a reproduction signal read from two tracks as a state. The solid line indicates the surviving path of each trellis state at time k. Let the signal vector of the surviving path be s ′ ^T (k) = [s ′ ₁ ^T (k), s ′ ₂ ^T (k)]. Further, s ′ _m ^T (k) = [s ′ _m (k),..., S ′ _m (k−2L _E )]. As parameters in each trellis state, a TDPR filter tap coefficient vector, an ITI channel response vector, an ISI channel response vector, and a TDPR filter output noise variance value are held.

Update these parameters according to the survival path. In the example of FIG. 14, state 0 at time k is updated from the parameters of state 1 at time k-1. Similarly, state 1 at time k is from state 3 parameters at time k-1, state 2 at time k is from state 1 parameters at time k-1, state 3 at time k is state 1 at time k-1. It is updated from the parameter. Let S ′ (S) be the state of the surviving path in state S, and h ′ _{m, S} ^T (k) = [h ′ _{m, 1, S} ^T (k), h _{Assuming that} ' _{m, 2, S} ^T (k)], an error em _{, S} (k) between the reproduction signal and its replica is expressed by the following equation (10).

The vector h ′ _{m, S} (k) is updated based on the following equation (11). Furthermore, TDPR filter tap coefficient vector _{w m,} S (k) is based on the following equation (12), the gain vector _{k m} determined by RLS _algorithm, is updated using the S (k). Further, the TDPR filter output noise variance value σ _{m, S} ² (k) is updated based on the following equation (13). However, N is a moving average length.

…（１０）

…（１１）

…（１２）

…（１３）

(10)

... (11)

(12)

... (13)

(Repeat processing when there is no error in one track)
If the CRC decoder 114 determines that there is no bit error for the reproduction data of one track, the one-dimensional / 2-dimensional max-log-MAP detector 110 performs the one-dimensional max- Log-MAP detection is used. That is, when there is no bit error in the reproduction data of one track, signal detection is executed using a trellis transition composed only of delay bits of one track signal. In this case, the output of the TDPR filter 108 is expressed by an equalization model shown in FIG. Note that the noise component is omitted here for the sake of simplicity.

  The reproduction signal of each track is extracted by a desired response (1 + D) by the TDPR filter 108. When signal detection is performed from these extracted signals, a model in which one delay tap having a tap coefficient of 0 is added can be used to enhance the channel estimation function. In this case, the trellis diagram for the equalization model shown in FIG. 15 is composed of the 16 trellis states shown in FIG. A branch metric is calculated by combining the square value of the difference between each TDPR filter output and the TDPR filter output replica given to each branch of the trellis diagram. A path metric is calculated by two-dimensional max-log-MAP detection.

  At this time, in order to perform channel estimation in which the number of taps of ISI and ITI is 2 or more, it is necessary to use surviving paths in each trellis state. As a result, channel estimation accuracy may not be sufficient. However, once one of the tracks is correctly decoded, it is no longer necessary to assign that signal to the trellis in subsequent iterations. Therefore, when the same 16-state trellis is used, a model in which three delay taps with an equivalent tap coefficient of 0 as shown in FIG. 17 are added can be used. The trellis diagram corresponding to this model is as shown in FIG. In the trellis diagram shown in FIG. 18, since 4 bits are held in each state, channel estimation accuracy is improved.

[effect]
FIG. 19 shows a comparison result of the BER characteristics when the off-track ratio of the reproducing head to the adjacent track is 50% and the inter-track write clock frequency difference is 0.02%. When there is no time-varying ITI tracking function (w / o channel tracking & tap-weight control), the BER is greatly deteriorated. On the other hand, when the technique of the present embodiment is applied (w / channel tracking & tap-weight control), the time-varying ITI is followed and the BER characteristic is improved as the number of repetitions increases. Further, as the number of repetitions increases, there is no inter-track interference and there is no write clock frequency difference (α = 0, Δf = 0%, λ = 0.999, μ = 0.001), which is close to the BER. A BER characteristic is obtained. As described above, the technique of this embodiment contributes to the improvement of the reliability of magnetic recording signal reproduction.

  As mentioned above, although preferred embodiment of this invention was described referring an accompanying drawing, it cannot be overemphasized that this invention is not limited to the example which concerns. It will be apparent to those skilled in the art that various changes and modifications can be made within the scope of the claims, and these are naturally within the technical scope of the present invention. Understood.

DESCRIPTION OF SYMBOLS 1 Signal processing apparatus 4 Recording medium 6 Spindle motor 8 Head 10 Recording signal production | generation part 20 Playback signal processing part 22 Buffer 30 Multitrack signal detection part 32 Two-dimensional partial response filter 34 Max-log-MAP detector 36 Deinterleaver
38 LDPC decoder 40 Hard decision unit 42 CRC decoder 44 Switching unit 46, 48 Interleaver 50 Soft replica generator 52, 54 Interference canceller 100 Multitrack signal detection unit 101 Channel estimation unit 102 TDPR filter tap coefficient generator 103 TDPR filter Output noise variance initial estimation unit 104 TDPR filter output noise variance update unit 105 Tap coefficient controller 106 Channel tracking unit 107 Multitrack soft interference canceller 108 TDPR filter 110 1D / 2D max-log-MAP detector 111 Deinterleaver 112 LDPC decoder 113 Hard decision unit 114 CRC decoder 115 Switching unit 116 Write signal soft replica converter 117, 118 Interleaver 131, 133 channels Estimator 132, 134 Shift amount estimator 135 Frequency difference estimator 136 Absolute value calculator 137 Average value calculator 138 Time constant table

Claims (6)

A signal detection device that detects a recording signal recorded on each track using a two-dimensional partial response filter from a plurality of reproduction signals read from a plurality of tracks using a head having a width larger than the track width. And
A tap coefficient generation unit that generates a tap coefficient of the two-dimensional partial response filter using a preamble and a postamble of the reproduction signal;
The tap coefficient generated by the tap coefficient generation unit is set in each state of the trellis diagram, and a survivor path in each state is detected by two-dimensional max-log-MAP detection, and a replica signal corresponding to each survivor path A survivor path detector that calculates an error from the output signal from the two-dimensional partial response filter;
A tap coefficient updating unit that updates a tap coefficient of the two-dimensional partial response filter based on the error corresponding to the surviving path detected by the surviving path detecting unit;
A signal detection device comprising: An error correction unit that performs error correction of the recording signal using a log likelihood ratio of each bit calculated when the two-dimensional max-log-MAP detection is performed;
A channel estimator for estimating a first channel response vector representing intersymbol interference and a second channel response vector representing intertrack interference using the preamble and postamble;
An interference replica generation unit that generates an interference replica indicating an interference component included in the reproduction signal using the first and second channel response vectors;
An interference canceller that subtracts the interference replica from the reproduction signal to cancel an interference component;
Further comprising
The signal detection apparatus according to claim 1, wherein the channel estimation unit updates the first and second channel response vectors according to the surviving path. The surviving path detection unit estimates a square value of an error between the replica signal and an output signal from the two-dimensional partial response filter by a moving average corresponding to the surviving path of each state. Item 3. The signal detection device according to Item 1 or 2. When the recording signal corresponding to at least one track is correctly decoded, the surviving path detection unit adds the trellis diagram to the delay bits of the remaining reproduction signals corresponding to tracks other than the track where the recording signal is correctly decoded. The signal detection apparatus according to any one of claims 1 to 3, wherein all the states are assigned and surviving paths in each state are detected by one-dimensional max-log-MAP detection. A first phase shift from an intersymbol interference channel response to an intertrack interference channel response is estimated using the preamble, and an intersymbol interference channel response to an intertrack interference channel response is estimated using the postamble. A phase shift estimator for estimating a second phase shift;
A frequency difference estimator for estimating a write clock frequency difference between tracks from a phase shift between the first phase shift and the second phase shift;
A time constant setting unit for setting a time constant used to update at least the tap coefficient based on the clock frequency;
The signal detection device according to claim 1, further comprising: This signal detection method detects a recording signal recorded on each track using a two-dimensional partial response filter from a plurality of reproduction signals read from a plurality of tracks using a head having a width larger than the track width. And
A tap coefficient generation step of generating a tap coefficient of the two-dimensional partial response filter using a preamble and a postamble of the reproduction signal;
The tap coefficient generated in the tap coefficient generation step is set in each state of the trellis diagram, and a survivor path in each state is detected by two-dimensional max-log-MAP detection, and a replica signal corresponding to each survivor path A surviving path detection step of calculating an error from the output signal from the two-dimensional partial response filter;
A tap coefficient updating step of updating a tap coefficient of the two-dimensional partial response filter based on the error corresponding to the surviving path detected in the surviving path detecting step;
The signal detection method characterized by including.

Images