Anything with RF is complex, and wireless in the home is especially so. Our many false starts forced us to do some research to establish a solid context for our tests.
Many variables affect the speed and reliability of wireless data transmission, whether used for cellphones or Wi-Fi in the home. We have tried hard to understand and control these variables in the tests we report here.
Wi-Fi Variables -- What Impacts Speed?
Here are some of the variables which can affect the speed and reliability of Wi-Fi in the home:
The particular "flavor" of Wi-Fi (802.11b, 802.11a, 802.11g) has a major impact. Each establishes the frequency band, protocols, modulation schemes, range of speeds, and specific choices for channels. The 2.4 GHz frequency band used by 11b and 11g is shared with other consumer devices such as microwave ovens and portable phones. The 5 GHz band used by 11a is less congested but the higher frequency probably results in a shorter useful range.
The rated output power of the access point, and to some extent of the network adapter directly affects the useful range of the signal. This is a characteristic of the radio chips and the antenna in each device, and is expressed in milliwatts (mW) or dBm (the two are directly related by a log formula: 0 dBm is 1 mW). The actual output power may differ from the rated power, either with a power control mechanism or some other cause. We believe variations in output power was a major factor causing variability in our speed measurements.
Receiver sensitivity in the network adapter and to some extent in the access point is the other major factor.
The operating channel probably has an impact, since other wireless devices may interfere more with one channel than another, and the access point and network adapter may operate better in certain channels.
The radio chip set represents an implementation of the selected Wi-Fi standard(s) and establishes many elements of performance.
The product implementation includes elements such as the power supply and antenna.
Software and firmware versions for the radio and the implementation can impact the operation of both the access point and the network adapter.
The house design determines the location and construction of walls, floors, windows, ducts and other elements which reduce the signal level and/or create "multipath" distortion.
The house may contain other wireless devices including microwave ovens and portable phones which operate in the same 2.4 GHz frequency as 802.11b and 11g.
The ambient noise level varies over time with other signals operating in the same frequency bands. These may well be coming from outside the home.
The location of the access point determines what walls and floors lie between the access point and the desired receiving locations. The easist location -- typically where the broadband modem connects to the primary PC -- may not be optimum for an access point.
The antenna orientation of both the access point and the network adapter can affect the received signal strength. Turning a notebook PC 90 degrees can sometimes change whether or not a reliable signal is received.
Notebook computer speed and memory has some effect on Wi-Fi performance, especially with faster protocols such as 11g and 11a.
People sitting or walking in or near the wireless path between the access point and the network adapter.
Managing The Variables
We attempted to manage these variables while conducting the Wi-Fi evaluation:
We measured the signal strength of each access point over long periods of ten or more hours and with short surveys before and after each measurement series. This provided some indication of how variations in the signal strength affected the test results.
We optimized the access point location in approximately the center of our house on the main floor. Had we located the access point at one of our main PCs -- on the upper and ground floors -- the signal would have had to travel through two floors and several walls to reach the other location.
During our tests, we tried to maintain the same position for each access point and network adapter. We placed each access point in nearly the same place within a foot. We placed the test notebook in the same place with the same antenna orientation within a few inches.
We let the access points "warm up". We found that the 11g access points changed the output signal level after half an hour or so, so we let them stabilize before starting the test measurements.
We tried not to interfere with the signals by staying as far as possible from the wireless path during each test measurement series.
We turned off all wireless devices and did not use the microwave oven or portable phone during the test measurements.
We studied the construction of our house to explain some of the varibility in performance.
Things We Did Not Do
From talking with other people involved in similar measurements, we understand that we could have done more to control the variables. For example:
We did not rotate the test PC. The standard procedure for these measurements is to rotate the test notebook PC slowly on a turntable, so the antenna goes through all possible orientations during the test. This might have provided more consistent measurements, but we did not think it was realistic in a home environment.
We did not monitor the signal strength during each test. We would have liked to do this, especially as we suspected that the signal was varying, but we used the same notebook PC for the surveys and for the test series.