josh@polaris.UUCP (Josh Knight) (02/25/86)
Reposted because of apparent network prolbems... %A Arvind M. Patel %T Adaptive Cross-Parity (AXP) Code for a High-Density Magnetic Tape System %J IBM J R&D %V 29 %N 6 %P 546-562 %D NOV 1985 %X This paper describes an error-correction system, called adaptive cross-parity (AXP) code, for the IBM 3480, a new high-density 18-track tape storage subsystem. Redundancy is applied to two interleaved sets of nine tracks in the same proportion as that in the previous IBM 3420 tape machines. The coding structure, however, is simpler, for it avoids the complex computations of Galois fields. The coding structure is based on a concept of interactive vertical and cross-parity checks, where the cross-parity checks span both sets of tracks and are used in either set in an adaptive manner. As a result, the overall error-correcting capability is substantially improved without increasing the redundancy. Decoding, in which simple parity equations are preprocessed, is designed to progress iteratively. By means of adaptive use of redundancy, the new method corrects up to three known erroneous tracks in any one set of nine tracks and up to four known erroneous tracks in the two sets together. The code also identifies the first unknown erroneous track in each of the two sets, and subsequently identifies the second unknown erroneous track in one of the two sets while providing correction for all these tracks. The result is generalized for a system with any number of tracks divided into a multiple number of unequal sets. %A Richard C. Scheider %T Write Equalization in High-Linear-Density Magnetic Recording %J IBM J R&D %V 29 %N 6 %P 563-568 %D NOV 1985 %X For many years, equalization has been used on the read side of a magnetic-recording channel to obtain a desired signal shape at the detector. Compensation on the write side has, for the most part, been limited to moving transition locations to offset read-signal peak shifts. This paper presents a new method of equalization on the write side through the addition of pulses at strategic locations on the write waveform. The resulting write current continues to be a two-level signal, so ac bias is not required. A linear transfer function can be derived for these write equalizers. This enables the recording-channel designer to partition the equalization more optimally between the write and read sides. The principal benefit of write equalization is that the read-flux-amplitude differences between high and low densities are significantly reduced. This permits maximum use of the linear operating region of the magnetoresistive read head. By providing high-frequency boosts on the write side, write equalization can reduce high-frequency noise at the read detector. Test results of channel linearity, as well as read signal waveshapes, are presented. %A P. Carnevali %A L. Coletti %A S. Patarnello %T Image Processing by Simulated Annealing %J IBM J R&D %V 29 %N 6 %P 569-579 %D NOV 1985 %X It is shown that simulated annealing, a statistical mechanics method recently proposed as a tool in solving complex optimization problems, can be used in problems arising in image processing. The problems examined are the estimation of the parameters necessary to describe a geometrical pattern corrupted by noise, the smoothing of bi-level images, and the process of halftoning a continuous-level image. The analogy between the system to be optimized and an equivalent physical system, whose ground state is sought, is put forward by showing that some of these problems are formally equivalent to ground state problems for two-dimensional Ising spin systems. In the case of low signal-to-noise ratios (particularly in image smoothing), the methods proposed here give better results than those obtained with standard techniques. %A R.J.S. Bates %A L.A. Sauer %T Jitter Accommodation in Token-Passing Ring LANs %J IBM J R&D %V 29 %N 6 %P 580-587 %D NOV 1985 %X In a token-passing ring Local Area Network (LAN) each message accumulates phase jitter as it travels around the ring. Unlike typical digital transmission systems, which tend to have random data traffic, ring systems carrying computer-generated traffic may have long strings of repetitive-pattern data. This traffic produces a jitter amplitude which is a function of the message statistics and the transmission characteristics of the physical layer. For the system to be stable, this jitter must be controlled. This paper describes the repetitive-pattern jitter generation and accumulation process and gives a methodology for designing the physical layer components to accommodate it. %A Haruhio Matino %T Analysis of the Holding Current in CMOS Latch-Up %J IBM J R&D %V 29 %N 6 %P 588-592 %D NOV 1985 %X The holding current in CMOS latch-up with or without well and/or substrate bias has been examined. Measurements indicate that the holding current increases significantly with reverse bias and low shunting base resistance. It is shown that a previous equation for the holding current is inaccurate, and a new equation for holding current with bias is presented. -- Josh Knight, IBM T.J. Watson Research josh at YKTVMH on BITNET, josh.yktvmh@ibm-sj.arpa ...!philabs!polaris!josh