TY - JOUR
T1 - Near-optimal parallel distributed data detection for page-oriented optical memories
AU - Chen, Xiaopeng
AU - Chugg, Keith M.
AU - Neifeld, Mark A.
N1 - Funding Information:
Manuscript received March 31, 1998; revised July 16, 1998. This work has been supported in part by the National Science Foundation under Contract NCR-9616663 and Contract NCR-9628405.
PY - 1998/9
Y1 - 1998/9
N2 - Volume optical storage systems suffer from numerous sources of noise and interference, the effects of which can seriously degrade retrieved data fidelity and produce unacceptable bit-error rates (BER's). We examine the problem of reliable two-dimensional data retrieval in the context of recently developed soft-decision methods for iterative decoding. We describe a novel near-optimal algorithm in which each pixel on the page is treated as a starting point for a simple iterative procedure so that a highly parallel, locally connected, distributed computational model emerges whose operation is well suited to the page-oriented memory (POM) interface format. We study the use of our two-dimensional distributed data detection (2D 4) algorithm with both incoherent (linear) and coherent (nonlinear) finite-contrast POM channel models. We present BER results obtained using the 2D 4 algorithm and compare these with three other typical methods [i.e., simple thresholding (THA), differential encoding (DC) and the decision feedback Viterbi algorithm (DFVA)]. The BER improvements are shown to have a direct impact on POM storage capacity and density and this impact is quantified for the special case of holographic POM. In a Rayleigh resolved holographic POM system with infinite contrast, we find that 2D 4 offers capacity improvements of 84%, 56%, and 8% as compared with DC, THA, and DFVA respectively, with corresponding storage density gains of 85%, 26%, and 9%. In the case of finite contrast (C = 4), similar capacity improvements of 93%, 18%, and 4% produce similar density improvements of 98%, 21%, and 6%. Implementational issues associated with the realization of this new distributed detection algorithm are also discussed and parallel neural and focal plane strategies are considered. A 2 cm 2 λ = 0.1 μm digital VLSI real estate budget will support a 600 × 600 pixel 2D 4 focal plane processor operating at 40 MHz with less than 1.7 W/cm 2 power dissipation.
AB - Volume optical storage systems suffer from numerous sources of noise and interference, the effects of which can seriously degrade retrieved data fidelity and produce unacceptable bit-error rates (BER's). We examine the problem of reliable two-dimensional data retrieval in the context of recently developed soft-decision methods for iterative decoding. We describe a novel near-optimal algorithm in which each pixel on the page is treated as a starting point for a simple iterative procedure so that a highly parallel, locally connected, distributed computational model emerges whose operation is well suited to the page-oriented memory (POM) interface format. We study the use of our two-dimensional distributed data detection (2D 4) algorithm with both incoherent (linear) and coherent (nonlinear) finite-contrast POM channel models. We present BER results obtained using the 2D 4 algorithm and compare these with three other typical methods [i.e., simple thresholding (THA), differential encoding (DC) and the decision feedback Viterbi algorithm (DFVA)]. The BER improvements are shown to have a direct impact on POM storage capacity and density and this impact is quantified for the special case of holographic POM. In a Rayleigh resolved holographic POM system with infinite contrast, we find that 2D 4 offers capacity improvements of 84%, 56%, and 8% as compared with DC, THA, and DFVA respectively, with corresponding storage density gains of 85%, 26%, and 9%. In the case of finite contrast (C = 4), similar capacity improvements of 93%, 18%, and 4% produce similar density improvements of 98%, 21%, and 6%. Implementational issues associated with the realization of this new distributed detection algorithm are also discussed and parallel neural and focal plane strategies are considered. A 2 cm 2 λ = 0.1 μm digital VLSI real estate budget will support a 600 × 600 pixel 2D 4 focal plane processor operating at 40 MHz with less than 1.7 W/cm 2 power dissipation.
KW - Distributed detection
KW - Iterative method
KW - Maximum likelihood detection
KW - Optical memories
KW - Parallel algorithm
UR - http://www.scopus.com/inward/record.url?scp=0032154698&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0032154698&partnerID=8YFLogxK
U2 - 10.1109/2944.735773
DO - 10.1109/2944.735773
M3 - Article
AN - SCOPUS:0032154698
SN - 1077-260X
VL - 4
SP - 866
EP - 879
JO - IEEE Journal on Selected Topics in Quantum Electronics
JF - IEEE Journal on Selected Topics in Quantum Electronics
IS - 5
ER -