Dynamic magnetic information storage or retrieval – General recording or reproducing – Specifics of equalizing
Reexamination Certificate
1998-03-04
2001-06-19
Hudspeth, David (Department: 2651)
Dynamic magnetic information storage or retrieval
General recording or reproducing
Specifics of equalizing
C360S046000, C375S290000, C375S341000, C375S350000
Reexamination Certificate
active
06249398
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention is in the field of digital data storage technology and, more specifically, pertains to improvements in partial response, maximum likelihood detection systems of the type commonly used in read channels of magnetic recording systems, particularly disk drives and tape systems.
2. Description of the Related Art
Performance of magnetic storage devices has improved dramatically over the past decade. Strides have been made in reduced error rates, lower cost, faster operating speeds, and increased storage capacity on magnetic tape and disk systems. There are various ways to increase storage capacity on disk drives, for example. One way is to increase the number of concentric tracks on the disk. Also, data can also be encoded more efficiently before writing to disk, thereby increasing the ratio of stored data to error correcting data—or more generally, reducing encoding overhead. Another important way to increase capacity has been to increase the bit density on the magnetic platters of the device, that is, pack the data more tightly on a given track. When this method is implemented, writing data is still relatively simple using standard inductive head technology. Reading the data back, however, becomes more challenging as spacing between flux transitions on the medium is reduced.
To ensure accurate data reads, several methods are used. First, it is known to use two separate heads; one for reading and one for writing. Standard inductive heads are used for writing, and ultra-sensitive magneto-resistive (MR) heads are used for reading. The read heads generate analog signals in response to flux transitions on the medium, and how this “read signal” is interpreted or “detected” by the drive electronics is the focus of the present invention.
In the past, a method of detection known as “pulse-peak” detection was used. Pulse-peak detection works in the following (highly simplified) manner. The signal from the read head is rectified, then threshold qualified to determine a window around the signal pulse peaks. Next, it is differentiated, which causes the flux transitions from the original input signal to appear as zero-crossings. Apply this final signal to a zero-crossing detector, and the circuit generates a pulse each time there is an input signal peak. This method worked well enough, as long as there were spaces between transitions on the medium, but as drive capacities have increased its limitations have surfaced. Inter symbol interference (ISI) caused by adjacent transitions on the medium can shift the timing of detected pulse peaks, and generate read errors.
To avoid ISI, transitions must be physically separated on the disk. This spacing wastes precious disk space, however. For example, with standard pulse-peak run-length-limited coding of (1,7), the ratio of user data to stored data on the drive is only 2:3. On a drive with a 6 GB capacity (around the current hard drive average), this wastes a full 2 GB of drive space—an unacceptable amount. Also, differentiating the incoming signal amplifies high frequency noise, which degrades the signal-to-noise ratio of the read channel and again increases the number of errors. As bit densities increase, the bit-error rate of pulse-peak detection increases, as does the amount of wasted space.
It is also known that partial response signaling better tolerates ISI and allows more efficient use of the bandwidth of a given channel. Specifically, partial response channels allow transitions in adjacent bit cells, and thus accommodate higher code rates than peak detection. Since the nature of ISI is well known, it can be taken into account in the decoding/detection process. Partial response transmission of data lends itself to synchronous sampling and provides a good compromise between error probability and the available spectrum.
Briefly, partial response, maximum likelihood (“PRML”) read channels work by sampling a pulse—e.g. the read signal pulse resulting from a magnetic flux reversal—several times instead of just detecting a peak. The pulse is wider than just one bit-time, i.e. more than one non-zero sample results from each flux reversal. Since an ideal partial response system is equivalent to a discrete-time filter, the analysis techniques used on filters can be applied. The D-transform is commonly used in magnetic disk storage. Here, the transform variable D represents unit time delay, the same delay as the inverse of the z-transform, i.e. D=z{circumflex over ( )}−1. The D-transform of the unit pulse response of a digital filter is the transfer function of that filter, and can be used to reveal its characteristics in the frequency domain. For a finite impulse response filter, the D-transform is simply the polynomial in D whose coefficients are the same as the unit pulse response sequence. So, for example, the EPR4 unit pulse (or dibit) response sequence of (1,1,−1,−1) is described by the polynomial 1+D−D
2
−D
3
. This is referred to as the EPR4 partial-response polynomial. Partial response systems described by the polynomials 1+D, 1−D and 1−D
2
are known as duobinary, dicode and Class IV (or “PR4”), respectively.
The detector attempts to match read signal samples to possible PR channel output sequences.
FIGS. 3 and 4
show typical PR channel pulse responses (i.e. response to magnetic transitions in adjacent bit cells.) For EPR4, an 8-state Viterbi detector generally realizes this detector, and essentially works by using an iterative method to determining the most likely route along the branches of a trellis, “most likely” generally meaning the path creating the minimum square error. PRML detection allows RLL (0,4/4) encoding, which improves the ratio of user data to stored data to about 8:9 (an example of a commercial implementation of this encoding is Quantum's Empire series) or better. This increased ratio allows more efficient use of platter area as well as increased data transfer rates. A new rate 24/25 encoding is disclosed in commonly-assigned U.S. patent application Ser. No. 08/774,412 filed Dec. 31, 1996, now U.S. Pat. No. 5,757,294.
In its current implementation, PRML presents limitations. It is generally known that for any given magnetic recording product, variations exist in the actual head/media response on a per head/disk basis as a function of many parameters, including manufacturing tolerances on the head and disk components, fly height of the magnetic head, component aging, environmental conditions, radius of the particular track, etc. This variation manifests itself mainly in pulse width and signal-to-noise-ratio variations. These variations mean that not only is an adaptive or programmable filter necessary, but also the optimum partial response target varies over the range of heads and disks, aging, etc. The optimal partial response target would be one that adapts as a function of channel variation to jointly maximize the minimum-distance (known as “Dina”) between all allowable sequences of idealized channel outputs (ICO's), while also minimizing the noise and equalization error for Dina-type Viterbi detector error events (known as “sigma”), i.e., maximize “Dina/sigma”. Unfortunately, there are no known algorithms today for automatically doing this optimization over all possible types of Viterbi detector error events.
System designers are currently forced to choose between canonical (1−D)(1+D){circumflex over ( )}N fixed partial response type targets and minimum mean square error (MMSE) type targets. Canonical targets, even the modern E
2
PR4, are an approximate match to the magnetic channel and are relatively simple to implement, but suffer from the existence of “quasi-catastrophic” error sequences due to spectral nulls at the Nyquist frequency, as well as significant equalization losses over relevant channel densities. MMSE targets are a good match to the channel, but only minimize “sigma” (noise and equalization error). They do not necessarily maximize the minimum-distance (
Fisher Kevin
Fitzpatrick Kelly K.
Modlin Cory
Patapoutian Ara
Sayiner Necip
Davidson Dan I.
Fish & Richardson
Hudspeth David
Maxtor Corporation
LandOfFree
Class of fixed partial response targets in a PRML sampled... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Class of fixed partial response targets in a PRML sampled..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Class of fixed partial response targets in a PRML sampled... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2492045