BBH Central IconBBH Report Home PageSandy and Dave
  CENTRAL home  |   REPORT home About/Contact Us  |   Subscribe  |   Index by Topic  
The June 17, 2003 Issue Provided by System Dynamics Inc.
Table of Contents Print this article Email this article to a friend

"Wireless Is Magic" -- Our Evaluation of 802.11g Wi-Fi

We have been conducting an extensive evaluation of current Wi-Fi wireless networking equipment based on the IEEE 802.11 standards. We wanted to see how well Wi-Fi worked in our home, how it compared with HomePlug powerline networking, and to what extent the faster versions of Wi-Fi would be suitable for "whole home" networking including video.

While we have been testing all three "flavors" of Wi-Fi, we decided to focus our first round of tests on 802.11g, the newest version. The 802.11g standard was approved last week, and we expect Wi-Fi certification will come next, followed by a flood of products. We were especially interested in how well 11g devices interoperated with 11b devices, since a key goal of the 11g standard was to provide much faster data rates in products that would be back-compatible and interoperable with 802.11b products.

This article summarizes our completed tests of 802.11g access points with both 802.11g and 802.11b notebook adaptors. Next month, we will report on 802.11b and 802.11a access points and notebook adaptors, and on mixed-mode networks with both 11g and 11b devices. This article attempts to provide the answers while sparing readers the extensive details. For those we refer readers to our Broadband Home Labs Website.

As we started to measure the range and speed of devices based on each "flavor" of Wi-Fi, we kept getting inconsistent results. When we talked with a good friend who holds an FCC First Class license, he said "Wireless is magic!" and helped us understand the variables which can affect performance. Arthur C. Clarke said: “Any sufficiently advanced technology is indistinguishable from magic.” We now know this is particularly true of wireless networking!

Although some of the magic still remains, wireless networking has made tremendous progress and we've been able to explain some of the science behind the magic. We have been using first-generation wireless LAN products for several years and had been disappointed to find that they didn't work at many of the places we most wanted to use them--especially those places where we hadn't installed Ethernet outlets. In our current tests, we saw excellent 802.11b performance at all test locations, with reduced performance only at a few locations with substantial path loss. We were also pleased to see that 11b notebook adapters worked very well with 11g access points.

We found that current 802.11g products fall short of advertising claiming "five times faster than 11b". While it was gratifying to see that 11g worked at every test location, the average data rate was little better than two times faster across all locations in our house. From our measurements, we concluded that 11g products in their current form are probably not well suited for wireless video networking.

However, it is very early in the product iteration cycle, and we expect to see rapid improvements in speed and consistency. Product announcements from Broadcom and others this week suggest that the refinement process is already well under way.

We look forward to repeating our tests on successive generations of 802.11g products. The tests reported here will serve as a "baseline" to quantify the improvement from one generation to the next.

Our testing procedure, results and conclusions are reported in depth in the Wi-Fi Evaluation pages of our website.

Summary of our test results

We measured the network speed while transferring large files between PCs at nineteen locations around the house. Fourteen locations were used in our earlier HomePlug evaluation, and we added two more inside the house and three outside. One location was close to the access point, while the others were separated by one or several walls, a floor, and/or ducting. We think this is fairly typical of North American houses.

Buffalo WBR-G54 broadband router/access point --> Click for larger pictureLinksys 54g card --> Click for larger pictureWe tested 802.11g access points and notebook adapters from Linksys and Buffalo Technology plus a Linksys 802.11b notebook adapter, in six "test combinations" of an access point and a notebook adapter. We used AirMagnet, a test tool, for signal level measurements, and our own diagnostic to measure file transfer speed.

Although the 802.11g standard has been approved in final form (on June 12, 2003), all tested 11g equipment was based on the Broadcom "54g" chip set with an early implementation of draft standards. 11g products based on the approved standard with Wi-Fi certification will likely provide different--probably better--results.

We did these tests in only one house, with one sample of each product and with one particular test tool. We have no way of knowing how well other samples and other products in other homes will perform.

Here's a summary of what we found:

  • Both 802.11g and 802.11b worked with all test combinations at all nineteen locations we tested, most of them obstructed by walls, floors and ducts, and one quite far from the access point and through a wall. This is substantially improved from our first-generation 802.11b equipment, which did not work at all at half of the locations.
  • We tested 802.11b with two test combinations. The measured file transfer speed averaged 4.34 Mbps across all nineteen locations, varying from an average maximum of about 5 Mbps at many locations down to about 3 Mbps at a few, with a low of about 0.6 Mbps at location 4 (Dave's conference table) with one combination.
  • We tested 802.11g with four test combinations. The measured file transfer speed averaged 9.07 Mbps across all nineteen locations, varying from an average maximum of about 17 Mbps (with the notebook adapter in the same room as the access point) down to about 5 Mbps at three locations, with a low of about 2 Mbps at location 16 (Dave's desk) with one combination.
  • Measured 802.11g speeds varied considerably more than those with 802.11b. The average 802.11b speed (4.34 Mbps) was about 85% of the maximum of 5.14, while the average 802.11g speed (9.07 Mbps) was only about 50% of the maximum (18.09).
  • The 802.11g measurements were quite inconsistent. Since all tested 11g equipment is based on the same Broadcom "54g" chip set, we expected roughly similar results from each combination. Instead, we saw significant speed variations at many locations, with some varying by more than 2:1.

Please see the Wi-Fi Test Results page on our website for the detailed measurements at each location.

New Product Announcements

Hot on the heels of 802.11g approval last week, Broadcom and SiGe Semiconductor issued related product announcements that encourage us to believe the product iteration process will move quickly in this highly-competitive market.

By enabling 802.11g product shipments in advance of standards approval, Broadcom's "54g" products won early market leadership. Broadcom reports the 54g is in 93% of all 802.11g products sold through U.S. retail distribution, and in 100% of the notebook PCs shipped to date.

Broadcom's first announcement following standards approval said that it was providing its partners with software to upgrade existing 54g products to the standard, and that it expected many of those partners to offer the upgrade as a software download for existing 54g products.

Broadcom's second announcement was that it is shipping standards-based technology to enhance the performance of all 802.11 based products. The new technology, which it calls Xpress, is based on WME, a key element of the forthcoming 802.11e standard. With Xpress, wireless devices can send longer packets, reducing the overhead associated with multiple packets and increasing the throughput. Broadcom says this can provide a 27% improvement in g-only networks and up to 74% in mixed b and g networks.

Finally, SiGe Semiconductor announced a new RF front-end solution for 802.11b/g. They claim their new chip will improve transmission range and speed for WLAN-enabled devices.

( ) ( ) ( ) ( ) ( )