Electrical computers: arithmetic processing and calculating – Electrical digital calculating computer – Particular function performed
Reexamination Certificate
1999-10-01
2003-09-09
Ngo, Chuong Dinh (Department: 2124)
Electrical computers: arithmetic processing and calculating
Electrical digital calculating computer
Particular function performed
Reexamination Certificate
active
06618740
ABSTRACT:
BACKGROUND OF THE INVENTION
Field of the Invention
This invention relates to improvements in methods or operating a mass data storage device having an FIR filter in a read channel thereof that uses PRML equalization and dynamic coefficients, and more particularly to improvements in methods for modifying the dynamic coefficients thereof, and to improvements in mass data storage devices, filters used therein in PRML equalization environments, and more particularly to circuits for modifying the dynamic coefficients thereof.
RELEVANT BACKGROUND
The RML equalization of a magnetic recording read channel is usually accomplished by a continuous-time filter, CTF, that is implemented using analog techniques, followed by a finite-impulse response filter, FIR. The FIR filter can be implemented in one of two ways: either in a sampled analog fashion or in a pure digital fashion. In the former case, a sample-and-hold circuit separates the CTF and FIR, whereas in the latter case, an analog-to-digital converter, ADC, separates the two blocks.
The chief purpose of the CTF filter is to provide an anti-aliasing filter to limit the spectral contents of the signal and noise beyond the Nyquist frequency. However, CTF filters are inherently difficult to tune and adapt. It the FIR filter, nowadays, that plays he major role achieving a fine signal equalization to the desired PRML target. With the advent of deep submicron CMOS processes, the digital FIR filters with sophisticated Least Mean Square, LMS, coefficient adaptation algorithms are becoming less and lest costly, whereas the analog CTF circuits do not scale down very well. The LMS adaptation of FTR coefficients is customarily done using a sign-sign LMS algorithm. Here, only a sign bit of the unequalized input data sample and a sign bit of the corresponding equalized output data sample error are used, instead of their full value representations.
Magneto-resistive heads have been recently more popular than their thin-film inductive head counterparts. This is due mainly to their lower cost. However, their response transfer function is nonsymmetric for positive and negative pulses, as well as the leading and trailing edges, and it exhibits a larger amount of nonlinearities.
There are two kinds of FIR configurations that are used to ensure its generally linear phase characteristic: odd- and even-symmetricity of the unit pulse response sequence. The linear phase property is required to minimize the frequency dispersion effects by nonlinear phase response in any data transmission application. However, a slight deviation from the perfect coefficient symmetricity is beneficial to compensate for the non-linear characteristics of MR heads and CTF filters.
Due to the problems of loop contention between the timing recovery and FIR coefficients adaptation loops, a symmetric-manner of coefficients adaptation is usually enforced. Not doing so is likely to cause a “runaway” situation for both loops. Even though the initial FIR coefficient settings do not have to be symmetric, the LMS adaptation will be incomplete and not able to automatically compensate for the signal nonlinearities.
SUMMARY OF THE INVENTION
The invention bridges the gap between the unrestricted coefficient adaptation that is capable of compensating high amount of signal nonlinearities but is likely to cause system instabilities, and the symmetric coefficient adaptation constraint that ensures FIR group delay stability.
Thus, in accordance with a broad aspect of the invention, a method for controlling adaptation asymmetry of coefficients of an FIR filter is presented in which the FIR coefficients are grouped into symmetric pairs. Correlation results between unequalized FIR filter input data samples and FIR filter equalized output error samples are accumulated. Coefficient update information is generated from the accumulated correlation results, and the coefficients within each symmetric pair are updated on the basis of the generated coefficient update information only if a nonzero limit of permitted coefficient asymmetry would not be exceeded by the update. The correlation results may be accumulated as successive sign values of an error between input samples to and output error samples from the FIR filter, and the coefficient update information may be generated by requesting an increment and decrement of a coefficient value of the FIR filter on the basis of carry-out and borrow-in operations of the least significant bit register, respectively.
According to another broad aspect of the invention, a method is presented for controlling adaptation asymmetry or coefficients of an FIR filter following a continuous time filter for limiting signal aliasing beyond the Nyquist frequency. The method includes grouping the FIR coefficients into symmetric pairs to follow a unit pulse response property of a linear phase response filter, and accumulating correlation results between unequalized input data samples and equalized output decision error samples in a correlator filter. The method also includes generating FIR filter coefficient update requests from information from the correlator filter, and updating the coefficients within each symmetric pair or the basis of the generated update requests only if a nonzero limit of permitted coefficient asymmerty in the FIR would not be exceeded.
According to yet another broad aspect of the invention, a method is presented for generating coefficients for an FIR filter in a sign-sign LMS algorithm. The method includes accumulating in a sub-least significant bit register successive sign values of an error between input data samples to the FIR filter and output error samples from the FIR filter. The method also includes, on the basis of carry-out and borrow-in operations of the Least significant bit register, respectively requesting an increment and decrement of a coefficient value of the FIR filter.
According to still another broad aspect of the invention, a circuit is presented for controlling adaptation asymmetry of coefficients of an FIR filter. The circuit includes an accumulator for accumulating correlation results between unequalized FIR filter input data samples and FIR filter output equalized error samples and a circuit for generating coefficient increment and decrement requests from the accumulated correlation results. A circuit updates the coefficients within a symmetric coefficient pair on the basis of the increment and decrement requests only if a predetermined nonzero coefficient magnitude difference between the coefficient pair would not be exceeded by the update. The circuit for updating the coefficients may include a circuit for determining a coefficient magnitude difference between a coefficient pair, a decoder to determine if the coefficient magnitude difference has not exceeded a predetermined value, and an arbiter circuit for issuing a coefficient update instruction if the decoder has determined that the coefficient magnitude difference has not exceeded the predetermined value.
According to still yet another broad aspect of the invention, a circuit is presented for generating coefficients for an FIR filter in a sign-sign LMS algorithm that includes an accumulator for accumulating in a sub-least significant bit register successive sign values of an error between input data samples to the FIR filter and output error samples from the FIR filter an increment/decrement circuit requests an increment/decrement of a coefficient value of the FIR filter on the basis of carry-out and borrow-in operations of the least significant bit resister.
In accordance with yet another broad aspect of the invention, a mass data storage device, having a read channel with an FIR filter, is presented. The circuit has an accumulator for accumulating correlation results between unequalized FIR filter input data samples and FIR filter output equalized error samples. A circuit generates coefficient increment and decrement requests from the accumulated correlation results. A circuit updates the coefficients within a symmetric coefficient pair on the basis of the increment and decrem
Brady W. James
Ngo Chuong Dinh
Swayze, Jr. W. Daniel
Telecky , Jr. Frederick J.
Texas Instruments Incorporated
LandOfFree
Method and architecture for controlling asymmetry of an LMS... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and architecture for controlling asymmetry of an LMS..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and architecture for controlling asymmetry of an LMS... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3049106