Department: Center for Magnetic Recording Research (CMRR)
Paul Siegel | Jack Keil Wolf
Name: Zeinab Taghavi
Email: ztaghavi @ ucsd.edu
Grad Year: 2008
The intensity model is one of the models widely used for optical channels. In this model, the output of the channel is equal to the absolute value of the sum of the input bits and the white complex Gaussian noise. Therefore, for this non-coherent channel, if the value of the bit is known, the probability distribution of the output is Rician. Furthermore, in many applications, such as data storage, the channel suffers from intersymbol interference (ISI). In these cases, trellis-based detectors provide near optimal performance, and these methods are well-studied in the context of partial-response channels with additive white Gaussian noise (AWGN).
In this work, we study the performance of trellis-based detectors for Rician partial-response channels. Trellis-based detection on a Rician channel uses branch metrics based upon the logarithm of a Rician distribution function. However, this metric requires evaluation of the logarithm of a Bessel function, which makes the implementation of the algorithm very complex. In addition, the analysis of the detector incorporating this metric is very complicated. Therefore, we propose to use a practical approximation for the branch metric. Our simulation results indicate that the bit error rate of the detector using this approximation is very close to that of the detector with the exact metric.
To analyze the performance of the modified detector, we compared the likelihoods of different error events in the detection process. By deriving the probability of occurrence of the dominant error events, we were able to bound the bit error rate of the detector. This bound is asymptotically tight as the signal to noise ratio (SNR) increases.
<< Back to Posters or Search Results