|
Interleaving is a well-known technique utilizing temporal diversity to mitigate burst errors in a communications link. Interleaving provides robustness at the physical layer of the network stack, at the expense of an increase in latency. An important parameter used in characterizing interleaver performance is the temporal diversity L, defined as the number of statistically independent fade events experienced by an individual codeword. This temporal diversity L characterizes the interleaver’s ability to whiten a channel so that forward error correction codes designed to mitigate additive white Gaussian noise function effectively. For large values of L, the interleaver is effectively “infinite”, and deeper interleaving produces no additional benefit. As L is decreased, the power penalty grows, and increasingly more average power must be transmitted to ensure error-free decoding at the receiver. Here we investigate interleaver performance in the regime where L is decreasing and the interleaver is not effectively infinite. Our goals are to (1) determine the regime where the interleaver is no longer effectively infinite, and (2) characterize the power penalty as a function of L to understand how the performance degrades as L is reduced. A theoretical model is presented that allows us to investigate interleaver performance in the presence of weak, moderate, and strong turbulence (scintillation indices of 0.2, 0.5, and 1.0, respectively). The model results show a gradual increase in penalty for reduced L for the weak-fading conditions. The moderate and strong fading conditions show similar dependence, with significant penalties over 3 dB developing as L drops below 100, and these penalties grow more rapidly when L drops below 10. We show close agreement to these models with experiments using a fiber-based, optical modem. This optical modem utilizes rate-1/2 coding together with a variable interleaver delay spanning three octaves. The optical data signal is transmitted through a fade emulator that applies the same fade profiles used in the theoretical model. This faded signal is sent to an optical-to-electrical receiver, and then into a digital electronics unit that processes the electrical signal to assess the coded communications performance.
|
