Are 802.11 link quality broadcast measurements always reliable

This paper describes an unexpected nding concerning link quality measurements carried out through broadcast frames. We found out that, with experimental equipments/drivers widely used by the scientic community (Atheros WLAN cards with MADWiFi driver), broadcast-based link qual-ity measurements seem to be aected by a NIC-proprietary power saving mode. The striking conclusion is that some packet losses, typically attributed to bad channel charac-teristics, rather may depend on the broadcast measurement methodology. Unicast measurements are shown to be ex-empt from such a problem. A very simple test is further-more proposed to verify whether measurement results are aected by the described phenomenon. 1