Bit error distribution and mutation patterns of corrupted packets in low-power wireless networks

It is well known that wireless channels produce higher bit error rates than wired connections. However, little knowledge exists about how bit errors are distributed within messages. In this paper, we present results from our experiments in an 802.15.4 sensor node testbed investigating the distribution of errors within erroneous frames. We identify three effects that can only partially be explained by coding and channel conditions: (1) errors are not independently distributed, but to a certain extent bursty, (2) coding leads to some bits being more stable than others, and (3) some content is inherently more stable than other during transmission. We discuss hypotheses on the origins of these effects and give some preliminary ideas on how to leverage them.