More information: [pdf]
More information: [htm]
In this new concept, the bits are organized in a broad spiral. Such a spiral consists of a number of bit-rows stacked upon each other with a fixed phase relation in the radial direction, so that the bits are arranged on a 2D lattice. A 2D closed-packed hexagonal ordering of the bits is chosen because it has a 15% higher packing fraction than the square lattice. Successive revolutions of the broad spiral are separated by a guard band consisting of one empty bit row. A multi-spot light-path for parallel readout is realized, where each spot has BD characteristics. Bits with a value "1" are represented physically as circular pit-holes on the disc, with a pit-hole size that is smaller than the hexagonal bit-cell in order to reduce the non-linearity of the channel.
The signal processing with equalization, timing recovery and bit-detection is carried out in a 2D fashion, that is, jointly over all the bit-rows within the broad spiral. A stripe-wise Viterbi detector is proposed to perform 2D bit-detection with a limited state complexity of the trellis. Only two iterations are required to reach a BER of 1E-04 at practical SNR levels. Other aspects for a limitation of the hardware complexity will also be discussed. A 2D modulation code is applied to eliminate "worst-case bit-patterns" that yield a high probability of erroneous detection: the modulation code has a very high code rate, and enumerative coding is applied together with a Bliss-like scheme for the combination with the error-correction coding.
Viewgraphs: [ppt]
Recording codes that are based on RLL sequences have found almost universal application in disc recording practice. We have the EFM code (rate = 8/17, d=2, k=10), used in the Compact Disc (CD), the EFMPlus code (rate = 8/16, d=2, k=10) used in the DVD, and 17PP (rate =2/3, d=1, k=7) used in the BluRay Disc.
In our presentation we will discuss state of the art coding technology, and we will show how the information theoretical limits on code performance can be approached.
Viewgraphs: NA
Viewgraphs: [ppt]
Viewgraphs: [pdf]
This is joint work with Panu Chaichanavong.
Viewgraphs: [pdf]
The talk will focus on two major issues related to important open problems in molecular biology and information theory. The first is concerned with investigating the characteristics of the DNA error-correcting mehanism and its possible connection to codes on graphs. Codes on graphs and iterative decoding represent recent conceptual advances in communication theory, and their close relationship to neural networks, self-organizing systems and spin glasses makes them universal models for a wide variety of phenomena. It seems plausible that the similarity between interactions of genes within regulatory systems, modeled by Boolean networks with sparse connectivity, and iterative decoding on graphs is not accidental; hence, this new development in coding theory could be utilized to provide partial information about the intricate and multifaceted error-correction mechanisms of the genome and its relationship to the gene regulatory network. The theory of structured Boolean networks, attractor and graph sensitivity analysis will be used for infering the gene network structure, its error-correcting potential as well as its operational influence on disease development and classification. The second problem to be addressed in the talk is concerned with developing specialized methods for DNA storage. Since approximatelly twenty new DNA strands are sequenced every day, efficient compression of DNA codes has become a problem of great interest to genetic databases and researchers trying to reconstruct the tree of life. Many known compression techniques, including the Ziv-Lempel, Burrows-Wheeler and grammar based algorithms, were proposed for this task. All of the described algorithms showed poor performance when applied to genetic sequences. This finding can be explained through the fact that the DNA code has an extremely varying statistics for coding and non-coding regions. For example, an overwhelming percentage of the DNA code has a fractal nature, and such long term correlations are not captured by standard compression algorithms. In the talk, problems related to the statistical analysis of DNA sequences, applications of Tsallis entropy, DNA distance measures and their use in cross-reference compression will be discussed and new research problems emerging in this area will be proposed.
This is joint work with Bane Vasic.
Viewgraphs: [ppt]
Viewgraphs: [ppt]
This is joint work with G. Wright
Viewgraphs: [pdf]
As in the case of one-dimensional channels, the capacity of this two-dimensional channel is defined as the mutual information rate between the input and output, maximized over all input distributions. When the input is i.i.d., equiprobable, the mutual information rate is called the symmetric information rate (SIR). The capacity and the SIR provide useful measures of the storage density that can, in principle, be achieved with page-oriented technologies, and also serve as performance benchmarks for channel coding and detection methods.
Recently, several authors independently proposed a new Monte-Carlo approach to calculating a convergent sequence of lower bounds on these information rates for one-dimensional ISI channels. The method requires the simulation of an a-posteriori probability (APP) detector, matched to the information source and channel, on a long sample realization of the channel output. The forward recursion of the sum-product (BCJR) algorithm, applied to the combined source-channel trellis, can be used to reduce the overall computational complexity of the APP calculations. The method was further extended to evaluate the information rates of multi-track recording channels.
In this talk, we investigate the extension of this approach to two-dimensional ISI channels. The main problem is that there is no direct counterpart of the BCJR algorithm that can simplify the APP calculations for a large sample output array in the two-dimensional setting. To overcome this difficulty, we derive upper and lower bounds on the entropy rate of a large output array based upon conditional entropies of smaller output arrays. By adapting the one-dimensional Monte Carlo technique to the calculation of these conditional entropies, we are able to compute fairly tight upper and lower bounds on the SIR for two-dimensional ISI channels with small impulse response support. We also analyze the convergence of these bounds.
The computational complexity of these bounds grows exponentially with the size of the support region of the impulse response. We therefore introduce a modification of the upper bound that allows us to consider channels with somewhat larger ISI. This alternative upper bound permits a tradeoff between the tightness and the computational complexity of the bound.
Finally, we discuss extension of these bounds to d-dimensional ISI channels, with d>2.
This talk describes work done jointly with Jiangxin Chen.
Viewgraphs: [ppt]
This is joint work with Andy DeWan, Chad Hayes, Josephine Hoh, Jurg Ott, Tony Parrado and Richard Sackler.
Viewgraphs: NA