Issue TOC Print full article Print this page Download as PDF E-Mail this URL Flame the author     In this article Introduction Far-Reaching Effects Products Reviewed How We Tested How Far Can You Go? Report Card

We tested the fixed wireless point-to-point systems in our Syracuse University labs using calibrated radio-frequency attenuators to simulate the free space path loss that would occur in a typical outdoor installation. Although our test bed let us simulate longer links, our objective was to simulate a 1-mile connection. At 5.3 GHz, the path loss for a 1-mile link is approximately 111 dB. Because some of the products include integrated antennas and others work with a variety of external antennas, we controlled for the effects of antenna gain by assuming that each product used an antenna on each end with a gain of 23 dBi, a number typical of several commercially available directional path antennas. We also corrected for path loss in our cabling system. For the products that included integrated antennas, we requested vendors provide an RF connector that bypassed the internal antenna.

click to enlarge

Because an antenna on each end would provide 46 dBi of total signal gain and our expected 1-mile path loss is 111 dB, we expected each product to operate properly with 65 dB of attenuation (111 minus 46). All the products met that standard. We also measured the level of attenuation at which each product stopped operating and compared that result to what we would expect based on manufacturers' claims of radio output power and receive sensitivity. We found that all the products operated within 5 percent of these specifications. Thus, we're confident in stating that, when properly installed with appropriate antennas, all these products will operate at the distances stated by the manufacturers.

For performance testing, we used NetIQ Chariot's filercvl TCP test using 100 iterations of 1-MB records to evaluate the maximum throughput under optimal conditions. The test bed consisted of four 1.2-GHz Toshiba Satellite notebook computers with 256 MB of RAM running Microsoft Windows XP. All computers and wireless bridges were connected to a Cisco 3500 switch with VLANs defined for each bridge and its computer pair (see diagram). We measured peak and average performance in unidirectional half-duplex mode and average aggregate performance in bidirectional full-duplex mode. We also ran the same tests with the notebook computers connected directly to the switches to establish a baseline Ethernet measurement and ensure that the notebook computers themselves were not a bottleneck.