The paper presented a study of packet delivery performance for 802.15.4 sensor networks. The tests were aimed around the common conceptual model generally used by protocol designers when developing communications protocols. The authors conducted experiments using two different kinds of motes and three different testbeds that vary in the number of motes, the size of the network, and environment. The first was in a 6000 sq ft office, the next in a 2500 sq ft space on a University campus, and the final on a dry outdoor lakebed. Based on the results of the experiments the authors developed theories on network performance due to internal and external conditions, with the environment playing a large role in packet performance. The first test consisted of studying the effects of modifying the inter-packet interval(IPI) between packet transmission. The results showed that over time the number of intermediate links increased and the numbers of poor and good links were reduced. An intermediate link is defined as a link with a packet reception ratio(PRR) between 10-90%. The next test was an observation of PRRs across all receivers observing each channel from a single transmitter. The authors were able to conclude that reception rates vary across channels. For example, a single receiver may have excellent packet reception for channel 26, but have poor to no reception for channel 17. A statistical analysis of packet success/failure was performed to identify any predictability for packet performance. Plotting consecutive successes and failures, the authors used the conditional packet delivery function(CPDF) and found that losses over time are independent, but over short periods contained temporal correlation. Further tests showed this was due to the received signal strength(RSSI) maintaining similar values over short periods, but varying greatly over longer periods. The authors also researched external noise(802.11,Bluetooth) and link asymmetries as reasons for poor packet reception.
During the discussion several topics were discussed including strengths and weaknesses of the article. One observation was around external noise. The authors provided the frequency spectrums of 802.15.4, 802.11, and Bluetooth, which explained the different reception rates across channels. Another question in the same topic was how and why does “channel” or frequency hopping affect performance. Going back to the frequency spectrum, we could see that Bluetooth completely overlapped the 802.15.4 frequency range, but didn’t not cause as much interference as in 802.11 due to the fact that is moves across the frequency spectrum very quickly(ever 625usecs). A question on link asymmetries was asked regarding noise floor differences and RSSI asymmetries and the effects on asymmetric links. After quickly perusing the paper, we concluded that the authors skimmed over the topic and didn’t provide any insight in the reasons behind the occurrence.