In my previous blog entry, I set out to answer the question of how much WiFi is deployed. Here, I like to focus on the question of interference. Unfortunately, I think there are fewer studies that characterize WiFi usage and interference than there ought to be. I will show here a few of the results and conclusions and present my own model and take on interference.
Geosimulations Research Laboratory conducted a project to record and map WiFi transmissions in Salt Lake City, UT which showed very high activity and penetration in the commercial core where as many as 43 access points are within range in parts of the downtown core and 7 on average throughout the city.
The UK regulator Ofcom has completed a study in 2009 on ISM band utilization in several locations including London and other cities and towns with lower population density. They data frame rate as proxy for utilization and concluded that interference tends to increase with population density but is not necessarily associated with high WiFi usage, rather it is mainly the result of other wireless devices such as baby monitors, security cameras, audio video senders, cordless phones, garage door openers, wireless game controllers, home automation devices, and other communication devices (Bluetooth, Zigbee). In many parts of London, there is no significant congestion.
An earlier Ofcom report from 2007 measured the relative time and frequency utilization which was found to be 9% for the 2.4 GHz ISM band, a low figure attributed to propagation characteristics that tend to localize transmissions. So, the attenuation is significant enough in this band that interference is localized to nearby interferers.
To illustrate, I developed my own, rather simple, interference model. Taking an area of 1 sq. km, I can define a certain uniform density of WiFi access points, activity factor, and a path loss dependency to account for propagation losses. The graph below shows results for 500 BSSID/Channel/km2 and propagation decay factor of 4 which I think is appropriate given the proximity of transmitters to ground. Wall losses were not factored in, so the amount of interference can be lower. The 90% percentile of interference occurs from nodes within 20 m of the victim access node, whereas far away access nodes contribute little interference power. However, if the propagation decay factor was 2.5, more in line with near line of sight communication, the 90% percentile of interference contribution is from within 200 m of the victim AP.
However, received interference may not in itself a critical factor as long as the signal quality remains high enough for communication. In typical WiFi networks, the subscriber is close to the access point which results in high enough signal to interference ratio for good communications.
What this all means is that in WiFi operation will migrate to the 5 GHz band for several reasons. While intra-WiFi interference may not be the main driver, interference from other devices is a motivation since the 5 GHz band is “cleaner.” Also, less WiFi devices in the 5 GHz band means less congestion and higher data rates. It will not take long before activity in the 5 GHz band increases (on this note, Ofcom reported 0% utilization of the 5 GHz band in 2007 when it was made available. Part of the reason is also related to higher propagation loss factor for 5 GHz which would localize signals more than in 2.4 GHz).
I set out to explore the topic of how much WiFi is out there with an eye on better understanding potential performance for carrier WiFi as well as other personal communication networks. It is clear that in many occasions, the type of application scenario is a dominant factor in determining interference. So, while a channel could be heavily polluted, it may still be useable if the application scenario minimizes the effects of interference.