JP3646684B2 - Data recording / reproducing apparatus using partial response demodulation method - Google Patents

Description

表中に示す孤立パルスの距離は孤立パルスの有する電力そのものである。最小距離は与えられた係数を持つパーシャルレスポンス信号のトレリスダイアグラム上の距離の内、最小になるものである。したがって、最小距離／孤立パルスの距離はこれを与えるパーシャルレスポンスのエネルギの利用効率を与える指標となる。本発明による係数を有するパーシャルレスポン方式はいずれも通常のＥＥＰＲＭＬのものよりこの点で優る。結果として、対称係数を有するＥＥＰＲＭＬに対し、Ｓ／Ｎを効果的に改善できていることが分かる。表２に３２状態のパーシャルレスポンスの代表的特性を示す。
【００３８】
【表２】

この場合にも、同様に特性の改善が顕著である。なお、表１および表２の特性は規格化線密度が２．５の場合に相当する。本発明は、さらに、ＳＮ比の改善だけでなく、符号誤りの長さも従来のＥＥＰＲＭＬやＥＥＥＰＲＭＬ方式の持つ長く連続する誤りから１ビットかあるいは３ビット長さの誤りが主体のものに改善できる。したがって、本発明は、少なくとも１ビットおよび３ビット連続誤りに対して符号誤り訂正能力を有する誤り訂正符号と組み合わせることで効率的な誤り訂正が可能になるという特長も有する。
【００３９】
【発明の実施の形態】
本発明にしたがった実際の回路構成例を図１２に示す。まず、磁気ヘッド出力はプリアンプを経て、ＡＧＣ（自動利得制御回路）とＬＰＦ（低域通過フィルタ）１５に供給される。ＡＧＣにて、信号振幅が一定値になるように制御された後、ＬＰＦにより所望周波数帯域以外の雑音成分が除去される。このＬＰＦ出力信号はＡＤＣ１６により離散量子化され、等化器１０に入力される。等化器１０では前述したように磁気ヘッドからの再生信号を、（１−Ｄ^２）なるパーシャルレスポンス特性となるように等化する。
【００４０】
この等化器の出力からＡＤＣ１６を動作させるために必要なクロック信号をＰＬＬ回路２０により生成する。同時に、ＡＧＣ１５の制御信号もＡＧＣ制御回路２１から得る。つぎに、等化器出力を離散フィルタ１８に加え、その出力に（１−Ｄ^２）（ｃ_０＋ｃ_１Ｄ＋・・・＋ｃ_ｎＤ^ｎ）なる応答特性を持つ波形を得て、これを最尤復号器１９に加え、データ識別を行う。この識別データを１６／１７変換もしくは８／９ＥＮＤＥＣにより復調し、その出力に元のユーザデータを得る。なお、等化器出力をＰＲ４の最尤復号器２３に供給することにより、通常のＰＲＭＬ復調データが得られる。つぎに、離散フィルタの構成を示す。
【００４１】
図１３は（ｃ_０＝３、ｃ_１＝２、ｃ_２＝１）なる係数を持つ離散フィルタの構成例である。等化器１０の出力を離散フィルタの入力端３０に加える。この信号を３倍の係数乗算器３１を通した出力、遅延回路３３により１ビット遅延した信号を２倍の係数乗算器３２に通した出力、２ビット遅延回路を通した出力を加算器３４により加算することで所望のフィルタ特性を出力端３５に得る。他の係数に関しても同様に構成できることは明らかである。
【００４２】
つぎに、本発明によるトレリスダイアグラムの構成法を示す。最尤復号器への入力ビットの値ａ_ｋと各状態Ｓ_ｋおよび出力ｙ_ｋは次式の関係がある。
【００４３】
Ｓ_ｋ=ａ_ｋ−５，ａ_ｋ−４，ａ_ｋ−３，ａ_ｋ−２，ａ_ｋ−１
ｙ_ｋ=ｃ_０ａ_ｋ＋ｃ_１ａ_ｋ−１＋（ｃ_２−ｃ_０）ａ_ｋ−２＋（ｃ_３−ｃ_１）ａ_ｋ−３−ｃ_２ａ_ｋ−４−ｃ_３ａ_ｋ−５ （数４）
最尤復号器ｃ_３＝０の場合には１６状態、ｃ_３の値が非零の場合には、３２状態を有する。（ｃ_０＝３、ｃ_１＝２、ｃ_２＝１）なる値を持つ１６状態最尤復号器のトレリス線図の構成例を図１４に示す。ここでこのような係数を有するパーシャルレスポンスをＭＥＥＰＲＭＬと呼称する。図１４の１６状態最尤復号器の一実施形態を図１５に示す。本処理回路は、ブランチメトリック生成部４０、ACS回路４１、パスメモリ４２から構成されており、図１４に示したＭＥＥＰＲＭＬトレリス線図に基づき回路が構成されている。ブランチメトリック生成部４０は、ＭＥＥＰＲＭＬトレリス線図の各状態から発生する状態遷移のブランチメトリックを与えるものである。
【００４４】
ACS４１は、１６状態のパスメトリックとブランチメトリック値との加算、比較、選択を行い、もっとも確からしいパスに対するパスメトリック値を生成する。パスメモリ４２は、各状態の比較結果をもとに、復号データの生成を行う。なおパスメトリックの初期化を初期設定回路４３により、本回路起動時に行なう。
【００４５】
次に、本発明のデータ復調回路を用いた磁気記録再生装置の一実施例を図１６に示す。パソコン等の外部装置は、磁気記録再生装置内のコントローラ１０２を介して、データの授受が行われる。まず、外部装置からのデータを記録する場合について説明する。コントローラ１０２は、データの記録命令を受けるとサーボ制御回路１０３に対し、記録すべき位置(トラック)に記録再生ヘッド１０６を移動する命令を発行する。記録再生ヘッドの移動が完了後、記録データは、記録データ処理回路１０４、Ｒ／Ｗアンプ５、記録再生ヘッド４を介して記録媒体3に記録される。
【００４６】
記録データ処理回路１０４は、エンコーダ２３、シンセサイザ１１２、プリコーダ９、記録補正回路１１４で構成され、エンコーダ２３は、記録データをコーディング規則に従ったコーディング処理、例えば、８／９ＧＣＲ（０，４／４）コード変換を行う。エンコードされたデータ列は、シンセサイザ１１２の記録ビット周期にしたがって送り出される。プリコード９は、データ列に一定の拘束条件を与えるため、再度コード変換される。記録補正回路１１４は、磁気記録固有の記録処理の非線形性を除去するものである。以上の動作により記録処理が行われる。
【００４７】
次に、データの再生動作について説明する。コントローラ１０２は、データ再生命令を受けると、サーボ制御回路１０３に対し、記録再生ヘッド４を該当するデータが記録された位置(トラック)へ移動する命令を発行する。記録再生ヘッドの移動が完了後、記録媒体３に記録された信号は、記録再生ヘッド４、Ｒ／Ｗアンプ５を介して、データ復調回路１に入力される。
【００４８】
データ復調回路１で復調された復調データは、コントローラ１０２に出力され、復調データの正当性を確認したのち外部装置にデータを転送する。データ復調回路１は、ヘッド再生波形の振幅を一定にするＡＧＣ回路１５、信号帯域外の雑音を除去する帯域除去フィルタ（ＬＰＦ）１５、再生信号をサンプリングするＡＤＣ１６、再生波形の符号間干渉を除去する等化器１０、ＡＤＣ１６のサンプリングタイミングを決定するＰＬＬ２０、本発明の主眼となるデータ復調回路１、復調データのデコード処理（８／９ＧＣＲデコーダ）を行うデコーダ２３から構成される。
【００４９】
マイコン１０１は、コントローラ１０２、データ復調回路１などの装置全体の処理をソフトウェアで行うものである。ここでは、マイコン１０１は、コード違反検出回路１２８の検出結果の検知、ＰＲＭＬ処理回路２３とＭＥＥＰＲＭＬ処理回路１９を切り替えるマルチプレクサ１２９に情報を与えるレジスタ１３０の設定などの処理を行う。さらに、ＰＲＭＬ処理回路の代わりに、表１に示す係数を有する他のＭＥＥＰＲＭＬ処理回路を用いて、記録密度に応じてこれらを適応的に切り替える構成にすることも可能である。
【００５０】
さらに、前述したように本発明では最尤復号器の出力データ中に生じる誤りの長さが１ビットあるいは３ビットが支配的になるため、これに適する誤り訂正を行なった後、デコーダ２３にて８／９ＧＣＲ等のデコード処理を行なう方が、符号誤りの拡大を防止する点で望ましい。このために、本発明によるＬＳＩを構成する際に、ＬＳＩ出力端子にデコード前の最尤復号器の出力を直接出すことを可能にする配線を行なうことも有効である。
【００５１】
【発明の効果】
本発明は、磁気記録装置の再生孤立磁化反転波形を非対称波形に変更することで、ＥＥＰＲＭＬやＥＥＥＰＲＭＬ等で問題になっていたある特定パターンにおいて発生する誤り伝播を抑圧する。ＭＥＥＰＲＭＬ方式はＥＥＰＲＭＬ方式に比較して、磁気記録装置の再生孤立磁化反転の半値幅と記録信号の半値幅の比が装置の実用範囲である２．５程度の場合には、約１．５ｄＢ以上のＳ／Ｎの改善が見込まれる。
【００５２】
本発明は、さらに、ＳＮ比の改善だけでなく、符号誤りの長さも従来のＥＥＰＲＭＬやＥＥＥＰＲＭＬ方式の持つ長く連続する誤りから単一ビットかあるいは３ビット長さの誤りが主体のものに改善できる。
【００５３】
【図面の簡単な説明】
【図１】本発明のデータ復調回路の一実施例を示した構成図である。
【図２】 PRML復調と磁気記録再生系の関連を示す図である。
【図３】各種パーシャルレスポンスの孤立波形応答を示す図である。
【図４】 EEPR4の孤立波形および孤立パルス応答を示す図である。
【図５】 EEPRMLのトレリス線図を示す図である。
【図６】 EEPRMLの信号間距離６を与えるパターンをトレリス線図上に示した図である。
【図７】 EEPRMLの信号間距離６を与える波形の例を示す図である。
【図８】 EEPRMLの信号間距離８を与える波形の例を示す図である。
【図９】 EEPRMLの信号間距離が６に縮退する原因を示す図である。
【図１０】パーシャルレスポンスの孤立波形応答のエネルギを集中させる基本原理を示す図である。
【図１１】最小位相推移波形の例を示す図である。
【図１２】本発明を実施する回路構成の一実施例を示す図である。
【図１３】本発明の離散フィルタ回路構成の一実施例を示す図である。
【図１４】本発明の係数を持つトレリス線図の例を示す図である。
【図１５】本発明による１６状態最尤復号器の一実施形態を示す図である。
【図１６】本発明を用いた磁気ディスク装置のデータ復調方法を示す図である。
【符号の説明】
1：データ復調器、2：誤り訂正符号復号器、5：記録再生アンプ、6：データ変調器、7：誤り訂正符号器、9：プリコーダ、10：等化器、11：最尤復号器、12：パスメトリック、13：ブランチメトリック演算器、14：全域通過フィルタ、15：ＡＧＣ,16：ＡＤＣ、18：離散フィルタ、19：高次パーシャルレスポンス最尤復号器、20：ＰＬＬ、23：16/17ENDEC、31：３乗算器、32：２倍乗算器、33：遅延回路、34：加算器、40：ブランチメトリック生成回路、41：ACS回路、42：パスメモリ、101：マイコン、102：コントローラ、103：サーボ制御回路、129：マルチプレクサ、130：レジスタ。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a signal processing method in a data recording / reproducing apparatus such as a magnetic disk or an optical disk device, and more particularly to a high-efficiency demodulation method of a high-order partial response method such as an EEPRML (Extended EPRML) or an EEEPRML (Extended EEPRML) signal processing method. .
[0002]
[Prior art]
In a magnetic disk device, a partial response class 4 (PR4) and a maximum likelihood decoding method (Partial Response Maximum Likelihood, hereinafter abbreviated as PRML) have been put into practical use as a high-efficiency signal processing method. The high-efficiency signal processing system means a system that can realize a desired data error rate with low S / N. Recently, as a signal processing method capable of reproducing a signal with an S / N lower than that of the PRML method, an EPRML method combining an EPR4 (Extended PR4) and a maximum likelihood decoding method, an EEPR4 (Extended EPR4) and a maximum likelihood decoding method. Higher order partial response methods such as an EEPRML (Extended EPRML) method combining the above are put into practical use.
[0003]
FIG. 1 shows a configuration example of a general magnetic disk device using the PRML signal processing method. The original data is supplied to the error correction encoder 7 via the interface circuit 8, and redundant data necessary for error correction is added. Next, modulation necessary for the PRML system is performed by the data modulator 6, and this is recorded on the magnetic disk 3 by the magnetic head 4 via the recording amplifier 5.
[0004]
A signal reproduced from the magnetic disk is subjected to PRML processing by a data demodulator 1 through a reproduction amplifier 5. The demodulated data is error-corrected by the error correction decoder 2 and then converted to the original data through the interface circuit 8. Low S / N signals are reproduced by such recording and reproduction processing. The operation and configuration of the data modems 1 and 6 will be described in detail with reference to FIG. 2 showing the relationship between the magnetic recording / reproducing system and the partial response system.
[0005]
First, the processing on the recording side will be described. Data from the error correction encoder 7 is recorded on the medium via a recording amplifier 5 via a precoder 9 comprising a delay element and modulo 2 (Mod. 2). This precoder 9 is for performing a measure for preventing error propagation of data occurring at the time of demodulation.
[0006]
Next, processing on the reproduction side will be described. Magnetization on the recording medium is reproduced as a waveform having differential characteristics by a reproducing magnetic head. PR4 regards this differential characteristic as a difference system of (1-D). Here, D means a 1-bit delay operator. The reproduced waveform is supplied to the equalizer 10 and equalized so that the waveform response becomes (1 + D). As a result, the total transfer characteristic at the output of the equalizer is (1-D²)become. Thereafter, the maximum likelihood decoder 11 identifies the data.
[0007]
FIG. 3 shows a response of a reproduction isolated waveform (step response is hereinafter abbreviated as an isolated waveform) when a step waveform is magnetically recorded. PR4 regards an isolated waveform as a waveform expanded to two time slots as shown in FIG. This waveform has a characteristic of (1 + D). EPR4 is regarded as a waveform expanded to 3 time slots as shown in (b). This waveform is (1 + D)²It has the characteristic that Further, EEPR4 is regarded as a waveform expanded to 4 time slots as shown in (c). This waveform is (1 + D)³It has the characteristic that
[0008]
Hereinafter, taking the EEPR4 method as an example, a high-order partial response method will be outlined.
[0009]
The total transfer characteristic of the EEPROM 4 is the product of the transfer characteristic of the isolated waveform and the transfer characteristic of the magnetic recording system.³It becomes. The impulse response of the EEPR4 system determined by this is shown in FIG. As can be seen from FIG. 4A, the isolated waveform of EEPR4 has amplitude characteristics of 1, 3, 3, and 1 for each bit period. Therefore, the isolated pulse response is obtained by superimposing the vertically inverted isolated waveforms by shifting them by 1 bit period as shown in FIG. 4B. That is, the isolated pulse response is 1, 2, 0, -2, -1. FIG. 5 shows an EEPRML trellis diagram in which a maximum likelihood decoder is combined with EEPR4.
[0010]
As is well known, the operation of the EEPRML system is explained by a trellis diagram. In the figure, a_kRepresents the input signal to EEPRML at time k. Here, 12 indicates a state, and 13 indicates a state transition. Label (a_k/ Y_kThe upper and lower tiers indicate the input signal value and the output signal value, respectively. The state of each signal processing method is determined by past input signal sequences. In EEPRML, the playback signal level at the current time is affected by the signal over the past four time slots. The state at time k is S_kS_k= ((A_k-4, A_k-3, A_k-2, A_k-1) | A_{k (1,0)}) And the number of states is 16. State transitions originating from a plurality of states at time k-1 are collected in a specific state at time k.
[0011]
For these state transitions, the square value of the difference between the output signal and the input signal shown at the bottom of each label is called a branch metric. In addition, the cumulative value of branch metrics up to the current time for each state is called a path metric. Of the state transitions gathered in a specific state at time k, only the state transitions with the smallest sum of the path metric up to time k-1 and the branch metric corresponding to each state transition are the maximum likelihood conditions (most likely ) Is selected as a state transition (path) that satisfies This process is divided into the following steps.
[0012]
That is, the path metric and the branch metric are added (Add). Next, these added values are compared for each state (Compare), and the state transition that is the minimum value is selected (Select). A series of these operations is abbreviated as ACS. Maximum likelihood decoding is a well-known technique in which the ACS operation is repeated for each time and for each state, and data is finally determined when one path converges on the trellis diagram.
[0013]
The performance of EEPRML is determined by the minimum free distance (Dfree). Here, Dfree is the one having the smallest path metric difference among various combinations from a specific node to another specific node on the trellis diagram shown in FIG. It is known that Dfree of EEPROM is 6. Furthermore, the distance between signals following Dfree is 8 and 10. The distance between these signals of EEPRML is determined by the data pattern input to the maximum likelihood decoder. In particular, the signal-to-signal distance is defined by the number of continuous changes from 0 to 1 or 1 to 0 in the pattern. As will be described later, for example, when the inversion position in the pattern is represented by p, the distance between these patterns becomes Dfree when two types of patterns having inversion positions continuous with ppp are shifted by 1 bit. give.
[0014]
In order to further improve the performance of these EPRML systems and EEPRML systems, a Maximum Transition Run Code (abbreviated as MTR code) has recently been proposed. For example, “Maximum Transition Run Codes for Data Storage Systems”, IEEE Transactions on Magnetics, vol. 32, No. 5, September, 1996, pp3992-3994 is known as a known example.
[0015]
The MTR code has a function of restricting occurrence of pattern inversion three or more times. When this MTR code is used, it can be limited to only those having an EEPROM distance of 10 or more. Therefore, the signal S / N can be improved equivalently. However, the MTR code has a code rate of 4/5 or the like, and this value is lower than that of 16/17 GCR (Group Coded Recording) or 8/9 GCR which are normally used. For this reason, the code rate loss is large, and the total coding gain is not always satisfactory. Specifically, the gain generated when the inter-signal distance is improved from 6 to 10 is about 2.2 dB.
[0016]
On the other hand, although the code rate loss depends on the recording density of the magnetic disk, for example, when the normalized linear density (the half width of the reproduction waveform is normalized by the width of the recording pulse) = 3, it becomes about 1 dB or more. The total coding gain is at most about 1 dB.
[0017]
[Problems to be solved by the invention]
An object of the present invention is to provide a general technique for extending regardless of a code using the inter-signal distance of a higher-order partial response system, particularly an EEPRML system or an EEEPRML system. In other words, the 16/17 GCR and 8/9 GCR used in the PRML signal processing for the magnetic disk device can be applied as they are, and the inter-signal distance is equivalently expanded without causing a new code rate loss. It is to provide a method.
[0018]
[Means for Solving the Problems]
The present invention expands the inter-signal distance by changing the response of the isolated pulse waveform from the original EEPRML or EEEPRML response in the higher-order partial response method, particularly the EEPRML method or the EEEPRML method. In the higher-order partial response method, the response of the isolated pulse is selected to be an odd symmetrical waveform. For example, in the EEPRML system, as described above, the response of the isolated pulse waveform is 1, 2, 0, -2, -1.
[0019]
In the present invention, first, the inter-signal distance is expanded by relaxing the odd symmetry of the response of such a high-order partial response type isolated pulse waveform. This inter-signal distance determines the S / N ratio at the time of signal identification. It means that the larger the distance, the larger the signal amplitude equivalently. Second, noise power is reduced. Partial response noise is correlated with each other over a plurality of times. The performance of the maximum likelihood decoder is degraded by the influence of this noise correlation. Therefore, the noise can be substantially reduced by suppressing the correlation of the noise. That is, the S / N of the high-order partial response signal is defined by the following equation.
[0020]
S / N = signal distance / (noise power / noise correlation coefficient) (Equation 1)
The present invention will be specifically described by taking the EEPRML method as an example. The sign is a binary number of {1, 0}. Now, in order to define the magnitude of the code error, a value of 1 when 1 is erroneously set to 0, -1 when 0 is erroneously corrected to 1, and 0 when no error occurs. To correspond. The following is a classification of error patterns of the EEPRML method according to this definition.
(A) Distance between signals = 6
(1, -1,1)
(B) Distance between signals = 8
1) (1, -1,1,0,0,1, -1,1)
2) (1, -1,1, -1,1)
(C) Distance between signals = 10
(0,1,0) etc.
The actual code error pattern of (A) is when (a, b, 1, 0, 1, c, d) is mistaken for (a, b,-, 1,-, c, d) or vice versa. is there.
[0021]
(B) The actual code error pattern of 1) is (1,0, 1, a, b, 0, 1) or (0, 1, 0, a, b, 0, 1, 0) or The reverse case.
[0022]
(B) The actual code error pattern of 2) is when (1, 0, 1, 0, 1, 0, 1) is mistaken for (0, 1, 0, 1, 0, 1) or vice versa. is there. Here, a, b, etc. are arbitrary.
[0023]
(C) is a 1-bit isolated pulse error. As described above, in the pattern common to (A) and (B), the inversion of the signal continues at least three times or more.
[0024]
Therefore, in the data pattern, ab1010cd or ab0101cd and these are continuous. FIG. 6 shows an error of the inter-signal distance = 6 on the trellis diagram. The two types of data systems α and β shown in FIG. 6 have values of 010abcde and 101abcde, respectively, and only the first three bits are different. The corresponding waveform is shown in FIG. As can be seen from this figure, the distance between the signals of these two patterns is 6. Similarly, FIG. 8 shows a waveform corresponding to (B) 1).
[0025]
Now, EEPRML has transfer characteristics (1-D) (1 + D)³Thus, as shown in FIG. 4B, the impulse response is determined as 1, 2, 0, -2, -1. Therefore, when a 1-bit error occurs, the signal-to-signal distance between the erroneous pattern and the pattern without the original error is 10, which is the sum of squares of each value of this impulse response. This inter-signal distance is the signal energy itself of the impulse. However, the reason why there is a pattern of inter-signal distance = 6 or inter-signal distance = 8 as shown in FIGS. 7 and 8 is that the combination of these patterns cancels the signal energy of the original impulse. . In other words, the EEPRML system has a pattern in which error propagation is likely to occur.
[0026]
Using the example shown in FIG. 9, the cause for reducing the inter-signal distance will be further considered. This figure shows a pattern having a distance 6 between signals in the EEPRML shown in FIG. In this pattern, signal inversion continues three times with P1, P2, and P3. Therefore, as shown in FIG. 9A, isolated waveforms having responses 1, 3, 3, and 1 are alternately repeated positively and negatively.
[0027]
As a result, a response of 1, 2, 1, 1, 2, 1 is obtained, and the energy of this signal is the sum of squares of each value (1).²+ (2)²+ (1)²+ (1)²+ (2)²+ (1)²= 12. On the other hand, the signal energy of each isolated waveform is (1)²+ (3)²+ (3)²+ (1)²= 20. Therefore, in a pattern in which signal inversion is continued three times with P1, P2, and P3, a total of 60 signal energies of three isolated waveforms are degenerated to 12.
[0028]
For example, as shown in FIG. 9B, the distance between signals is improved to 15 by removing the response having the amplitude of 1 at the right end from the response of the isolated waveform. This means that since the EEPRML isolated waveform responses 1, 3, 3, and 1 are spread over a plurality of bits, the special signals shown in (A) and (B) described above are used. The energy they have is offset. This essentially reduces the inter-signal distance, resulting in error propagation.
[0029]
Based on this consideration, the fundamental guideline for increasing the distance between signals is to devise a method of concentrating energy without losing the energy (power) of the isolated waveform. In general, as shown in FIG. 10, it is clarified by the communication theory that the means for concentrating the energy of the signal should pass the isolated waveform through the all-pass filter 14 and satisfy the minimum phase transition condition.
[0030]
Here, the minimum phase transition condition is that a zero and a pole of a transfer function of a signal given by a rational function exist within the same unit circumference. By setting the phase filter to satisfy this condition, the signal energy can be concentrated on the first half of the impulse response while the signal energy is preserved. In magnetic recording, it is well known that an isolated waveform can be approximated by a Lorentz waveform. When this is given by L (t), it is expressed by the following equation.
[0031]
L (t) = 1.0 / (1+ (2t / TW)²(Equation 2)
TW gives the half width.
[0032]
As apparent from Equation 2, L (t) is a symmetrical waveform. Here, the ratio (TW / T) between the half width and the time width T of the pulse to be recorded is defined as the normalized linear density. The larger the TW / T value, the higher the density recorded waveform. In normal magnetic recording, a standardized linear density of about 2.5 is used. Let Lmin (t) be a waveform obtained by passing a Lorentz waveform having normalized line densities of 2.5 and 3.0 through a minimum phase transition filter.
[0033]
FIG. 11 shows Lmin (t). As is apparent from FIG. 11, the waveform is asymmetrical, and it can be seen that the energy is concentrated in the first half of the response of the isolated waveform. However, it is generally very difficult to extract a clock signal (timing signal) necessary for signal discrimination from an asymmetric shape. One reason for this is that pattern-dependent jitter (time fluctuation) increases due to phase distortion. Another reason is that since the signal amplitude becomes multi-valued, the clock signal (timing signal) extraction circuit becomes complicated, and this realization becomes difficult.
[0034]
Therefore, in the present invention, in order to eliminate this contradictory condition, the higher order partial response polynomial PR (D) is factorized as follows:
PR (D) = (1-D²) (C₀+ C₁D + ... + c_nDⁿ(Equation 3)
Timing extraction is performed in the state of the previous term on the right side, and thereafter, the asymmetric response given in the latter term on the right side is given by a discrete time filter to give the above asymmetry to the waveform. At this time, the asymmetry coefficient c that maximizes the S / N given by Equation 1₀, C₁, ..., c_nSelect.
[0035]
Next, a method for obtaining an actual asymmetric coefficient will be described. First, in the case of 16-state EEPRML, the coefficient is (c₀= 1, c₁= 2, c₂= 1). C₁C₀And c₂The value of is the symmetry coefficient. On the other hand, in order to obtain the asymmetry coefficient, first, the latter term on the right side of (Expression 3)₀= 1 monic polynomial, c₁And c₂Is regarded as a real two-variable function, and an optimum coefficient is obtained according to the evaluation criterion of Equation 1. Thereafter, the integer coefficient closest to the real number is obtained.
[0036]
The method of obtaining the inter-signal distance, noise power, noise correlation coefficient, etc. shown in Equation 1 is described in the document “Maximum Likelihood Sequence Estimation of Digital Sequences in the Presence of Intersymbol Interference”, IEEE Transactions on information Theory, vol.IT-18. , No.3, May, 1972, pp363-378, and are omitted here. Table 1 shows typical characteristics of 16-state partial responses.
[0037]
[Table 1]

The distance of the isolated pulse shown in the table is the power itself of the isolated pulse. The minimum distance is the smallest distance on the trellis diagram of the partial response signal having a given coefficient. Therefore, the minimum distance / isolated pulse distance is an index that gives the energy utilization efficiency of the partial response that gives it. All of the partial response systems with coefficients according to the present invention are superior in this respect to those of ordinary EEPRML. As a result, it can be seen that S / N can be effectively improved with respect to EEPRML having a symmetry coefficient. Table 2 shows typical characteristics of the 32-state partial response.
[0038]
[Table 2]

In this case as well, the improvement in characteristics is remarkable. The characteristics shown in Tables 1 and 2 correspond to the case where the normalized linear density is 2.5. The present invention can improve not only the SN ratio but also the length of the code error from the long continuous error of the conventional EEPRML or EEEPRML system to a one-bit or three-bit length error. Therefore, the present invention also has an advantage that efficient error correction becomes possible by combining with an error correction code having a code error correction capability for at least 1-bit and 3-bit continuous errors.
[0039]
DETAILED DESCRIPTION OF THE INVENTION
An actual circuit configuration example according to the present invention is shown in FIG. First, the magnetic head output is supplied to an AGC (automatic gain control circuit) and an LPF (low-pass filter) 15 through a preamplifier. After the AGC is controlled so that the signal amplitude becomes a constant value, noise components other than the desired frequency band are removed by the LPF. This LPF output signal is discretely quantized by the ADC 16 and input to the equalizer 10. As described above, the equalizer 10 converts the reproduction signal from the magnetic head to (1-D²To equalize the partial response characteristics.
[0040]
A clock signal necessary for operating the ADC 16 is generated by the PLL circuit 20 from the output of the equalizer. At the same time, a control signal for the AGC 15 is also obtained from the AGC control circuit 21. Next, the equalizer output is applied to the discrete filter 18 and the output is (1-D²) (C₀+ C₁D + ... + c_nDⁿThe waveform having the response characteristic is obtained and added to the maximum likelihood decoder 19 to perform data identification. This identification data is demodulated by 16/17 conversion or 8 / 9ENDEC, and the original user data is obtained at the output. Note that normal PRML demodulated data is obtained by supplying the equalizer output to the PR4 maximum likelihood decoder 23. Next, the configuration of the discrete filter is shown.
[0041]
FIG.₀= 3, c₁= 2, c₂= 1) A configuration example of a discrete filter having a coefficient of 1). The output of the equalizer 10 is applied to the input 30 of the discrete filter. This signal is output through a triple multiplier 31, the signal delayed by 1 bit by a delay circuit 33 is output through a double multiplier 32, and the output through a 2-bit delay circuit is output by an adder 34. A desired filter characteristic is obtained at the output end 35 by adding. Obviously, other coefficients can be similarly constructed.
[0042]
Next, a method for constructing a trellis diagram according to the present invention will be described. The value a of the input bit to the maximum likelihood decoder_kAnd each state S_kAnd output y_kHave the following relationship:
[0043]
S_k= a_k-5, A_k-4, A_k-3, A_k-2, A_k-1
y_k= c₀a_k+ C₁a_k-1+ (C₂-C₀A)_k-2+ (C₃-C₁A)_k-3-C₂a_k-4-C₃a_k-5    (Equation 4)
Maximum likelihood decoder c₃16 states if = 0, c₃If the value of is non-zero, it has 32 states. (C₀= 3, c₁= 2, c₂FIG. 14 shows a configuration example of a trellis diagram of a 16-state maximum likelihood decoder having a value of = 1). Here, the partial response having such a coefficient is referred to as MEEPRML. One embodiment of the 16-state maximum likelihood decoder of FIG. 14 is shown in FIG. This processing circuit includes a branch metric generation unit 40, an ACS circuit 41, and a path memory 42, and the circuit is configured based on the MEEPRML trellis diagram shown in FIG. The branch metric generation unit 40 provides a branch metric of state transitions generated from each state of the MEEPRML trellis diagram.
[0044]
The ACS 41 performs addition, comparison, and selection of the 16-state path metric and the branch metric value, and generates a path metric value for the most likely path. The path memory 42 generates decoded data based on the comparison result of each state. The path metric is initialized by the initial setting circuit 43 when the circuit is activated.
[0045]
Next, FIG. 16 shows an embodiment of a magnetic recording / reproducing apparatus using the data demodulation circuit of the present invention. An external device such as a personal computer exchanges data via the controller 102 in the magnetic recording / reproducing device. First, a case where data from an external device is recorded will be described. When receiving a data recording command, the controller 102 issues a command to move the recording / reproducing head 106 to a position (track) to be recorded to the servo control circuit 103. After the movement of the recording / reproducing head is completed, the recording data is recorded on the recording medium 3 via the recording data processing circuit 104, the R / W amplifier 5, and the recording / reproducing head 4.
[0046]
The recording data processing circuit 104 includes an encoder 23, a synthesizer 112, a precoder 9, and a recording correction circuit 114. The encoder 23 performs a coding process for recording data according to a coding rule, for example, 8/9 GCR (0, 4/4). ) Perform code conversion. The encoded data string is sent out according to the recording bit period of the synthesizer 112. The precode 9 is subjected to code conversion again in order to give a certain constraint condition to the data string. The recording correction circuit 114 removes non-linearity of recording processing inherent to magnetic recording. The recording process is performed by the above operation.
[0047]
Next, the data reproduction operation will be described. Upon receiving the data reproduction command, the controller 102 issues a command to move the recording / reproducing head 4 to the position (track) where the corresponding data is recorded, to the servo control circuit 103. After the movement of the recording / reproducing head is completed, a signal recorded on the recording medium 3 is input to the data demodulating circuit 1 via the recording / reproducing head 4 and the R / W amplifier 5.
[0048]
The demodulated data demodulated by the data demodulating circuit 1 is output to the controller 102, and after confirming the validity of the demodulated data, the data is transferred to an external device. The data demodulating circuit 1 includes an AGC circuit 15 that makes the amplitude of the head reproduction waveform constant, a band elimination filter (LPF) 15 that removes noise outside the signal band, an ADC 16 that samples the reproduction signal, and intersymbol interference of the reproduction waveform. The equalizer 10, the PLL 20 that determines the sampling timing of the ADC 16, the data demodulator circuit 1 that is the main object of the present invention, and the decoder 23 that performs the decoding process (8/9 GCR decoder) of the demodulated data.
[0049]
The microcomputer 101 performs processing of the entire apparatus such as the controller 102 and the data demodulation circuit 1 by software. Here, the microcomputer 101 performs processing such as detection of the detection result of the code violation detection circuit 128 and setting of the register 130 that gives information to the multiplexer 129 that switches between the PRML processing circuit 23 and the MEEPRML processing circuit 19. Furthermore, instead of the PRML processing circuit, another MEEPRML processing circuit having the coefficients shown in Table 1 can be used to switch these adaptively according to the recording density.
[0050]
Further, as described above, in the present invention, the length of errors occurring in the output data of the maximum likelihood decoder is dominated by 1 bit or 3 bits. Therefore, after performing error correction suitable for this, the decoder 23 It is desirable to perform decoding processing such as 8/9 GCR from the viewpoint of preventing expansion of code errors. For this reason, when configuring the LSI according to the present invention, it is also effective to provide wiring that allows the output of the maximum likelihood decoder before decoding to be directly output to the LSI output terminal.
[0051]
【The invention's effect】
The present invention suppresses error propagation that occurs in a specific pattern that has been a problem in EEPRML, EEEPRML, etc., by changing the reproduced isolated magnetization reversal waveform of the magnetic recording device to an asymmetric waveform. Compared with the EEPRML system, the MEEPRML system is about 1.5 dB or more when the ratio of the half width of the regenerative isolated magnetization reversal of the magnetic recording apparatus to the half width of the recording signal is about 2.5, which is the practical range of the apparatus. S / N improvement is expected.
[0052]
The present invention can improve not only the SN ratio but also the length of the code error from the long continuous error of the conventional EEPRML or EEEPRML system to a single bit or 3 bit length error. .
[0053]
[Brief description of the drawings]
FIG. 1 is a configuration diagram showing an embodiment of a data demodulation circuit of the present invention.
FIG. 2 is a diagram showing the relationship between PRML demodulation and a magnetic recording / reproducing system.
FIG. 3 is a diagram showing isolated waveform responses of various partial responses.
FIG. 4 is a diagram showing an isolated waveform and isolated pulse response of EEPR4.
FIG. 5 is a diagram showing an EEPRML trellis diagram;
FIG. 6 is a diagram showing a pattern for giving an inter-signal distance 6 of EEPRML on a trellis diagram.
FIG. 7 is a diagram illustrating an example of a waveform that gives an inter-signal distance 6 of EEPRML.
FIG. 8 is a diagram illustrating an example of a waveform that gives an inter-signal distance 8 of EEPRML.
FIG. 9 is a diagram illustrating a cause of the EEPRML inter-signal distance degenerating to 6. FIG.
FIG. 10 is a diagram showing a basic principle for concentrating the energy of an isolated waveform response of a partial response.
FIG. 11 is a diagram illustrating an example of a minimum phase transition waveform.
FIG. 12 is a diagram showing an embodiment of a circuit configuration for carrying out the present invention.
FIG. 13 is a diagram showing an embodiment of a discrete filter circuit configuration according to the present invention.
FIG. 14 is a diagram showing an example of a trellis diagram having coefficients according to the present invention.
FIG. 15 shows an embodiment of a 16-state maximum likelihood decoder according to the present invention.
FIG. 16 is a diagram showing a data demodulation method of a magnetic disk device using the present invention.
[Explanation of symbols]
1: data demodulator, 2: error correction code decoder, 5: recording / reproducing amplifier, 6: data modulator, 7: error correction encoder, 9: precoder, 10: equalizer, 11: maximum likelihood decoder, 12: path metric, 13: branch metric calculator, 14: all-pass filter, 15: AGC, 16: ADC, 18: discrete filter, 19: high-order partial response maximum likelihood decoder, 20: PLL, 23: 16 / 17ENDEC, 31: 3 multiplier, 32: double multiplier, 33: delay circuit, 34: adder, 40: branch metric generation circuit, 41: ACS circuit, 42: path memory, 101: microcomputer, 102: controller, 103: Servo control circuit, 129: Multiplexer, 130: Register.

Claims (6)

A data recording medium for recording data;
A head for reproducing a signal from the data recording medium;
An equalizer that equalizes and outputs the signal reproduced by the head to a partial response characteristic;
A discrete filter that converts the output of the equalizer into an asymmetric waveform and outputs the asymmetric waveform;
A data recording / reproducing apparatus comprising: a maximum likelihood decoder for decoding the output of the discrete filter. The equalizer converts the signal reproduced by the head to (1-D ² And the discrete filter outputs the output of the equalizer to (1-D). ² ) (C ₀ + C ₁ D + ... + c _n D ⁿ The data recording / reproducing apparatus according to claim 1, wherein the data recording / reproducing apparatus converts the waveform into Coefficient c ₀ , C ₁ ... c _n The data recording / reproducing apparatus according to claim 2, further comprising a controller for setting Coefficient c ₀ , C ₁ ... c _n The data recording / reproducing apparatus according to claim 2, wherein is an integer. An A / D converter disposed between the head and the equalizer, A / D-converts a signal reproduced by the head and outputs the signal to the equalizer, and a timing signal generation from the output of the equalizer The data recording / reproducing apparatus according to claim 1, further comprising a timing extraction circuit that outputs to the A / D converter. 2. The data recording / reproducing apparatus according to claim 1, further comprising error correction means for correcting at least one or three bits of continuous error with respect to the output of the maximum likelihood decoder.

Images