Examining Relationships Between 802.11n Physical Layer Transmission Feature Combinations

To increase throughput the 802.11n standard introduced several physical layer transmission features including a short guard interval wider channels, and MIMO. Since obtaining peak throughput depends on choosing the combination of physical layer features (configuration) best suited for the channel conditions, the large number of configurations greatly complicates the decision. A deeper understanding of relationships between configurations under a variety of channel conditions should simplify the choices and improve the performance of algorithms selecting configurations. Examples of such algorithms include: rate and channel width adaptation, frame aggregation, and MIMO setting optimization. We propose a methodology for assessing the possibility of accurate estimation of the frame error rate (FER) of one configuration from the FER of another. Using devices that support up to 3 spatial streams (96 configurations), we conduct experiments under a variety of channel conditions to quantify relationships between configurations. We find that interesting relationships exist between many different configurations. Our results show that in 6 of the 7 scenarios studied at most five configurations are required to accurately estimate the error rate of all remaining 91 configurations and in the other scenario at most 15 configurations are required. Although we show that these relationships may change over time, perhaps most surprising is that relationships have been found over periods of up to one hour. These findings suggest optimization algorithms should not need to measure the FER of many configurations, but instead can sample a small subset of configurations to accurately estimate the FER of other configurtions. To demonstrate this possibility, we make simple modifications to the Minstrel HT rate adaptation algorithm to exploit relationships and observe improvements in throughput of up to 28%.