« Let's Get Small | Main | Wi-Fi for Cellular Backhaul? Really. »

November 26, 2011

Channel Changing Done Right

Remote-control

Since the inception of Wi-Fi there has been an ongoing debate about how APs should pick the best RF channel to yield the highest capacity for clients. Of the 11 channels available in North America, conventional Wi-Fi wisdom states that only channels 1, 6 and 11 should be used but those could actually be the worst channel choices.

Channel Basics

Think of each channel as a lane on a freeway and each lane has a defined width. Every Wi-Fi channel has a width of 5 MHz. The problem is, a typical Wi-Fi transmission takes up 4 lanes (20 MHz) of traffic. 

Wi-fi-spectrum

This chart shows channels 1, 6 and 11 and their respective frequencies. If you were to have two access points, one on channel 1 and the other on channel 2, they will overlap. Depending on the attenuation (signal loss) between each AP, this can cause some interesting issues. If the APs are in signal range of each other they will actually cause interference. 

The way to combat this channel overlapping problem is to skip channels when deploying Wi-Fi. This is why Wi-Fi deployments typically choose channels 1, 6 and 11. Choosing these channels results in 3 non-overlapping (non-interfering) channels. In a perfectly clean environment where no sources of outside interference can be guaranteed, the chosen 3-channel model is the right way to go. 

But what is the right channel when deploying a WLAN in Manhattan, San Francisco or a noisy hospital or school with sources of interference? Other Wi-Fi systems or sources of interference can dramatically reduce performance of any system. Beyond fancy signal path section technology using adaptive antenna arrays, intelligent channel selection is another important way of dealing with interference and increasing the capacity of Wi-Fi networks.

Every vendor has a different take on how to choose the proper channel. Some with built in spectrum analyzers will scan the spectrum and choose the best channel based on what it sees. On the surface this may seem like a good way to determine the right channel but in practice, it doesn’t work very well. Here’s why.

Let’s say you are on stage about ready to speak. Papers are rustling and there are other people having whispered conversations. If you speak, will people hear you? How well will they hear you? You can’t know if they will hear you just by listening. The only way to find out is to talk and observe the results. 

An AP that just listening will give you information about the noise but doesn’t really know how well it will be able to communicate with the clients. Why? Because the AP doesn’t have the same perspective of the client.

Today's Channel Changing Convention

Almost all of today’s channel optimization approaches use background checking methods to identify a better channel in case of throughput degradation in the current channel.  When conducting background checking, the AP jumps to other channels to passively listen for packet retries, transmission errors and RF interference when they it is idle — not sending or receiving data. 

These approaches provide incomplete information for channel selection, as they are not able to measure realizable channel capacity and are a equivalent to a random guess at which alternative channel might provide better results.  In addition, the process of background checking creates “dead time” when the AP is busy checking other channels and may miss transmission from clients.  Although the dead time is short, it can seriously affect client activities and channel performance when bandwidth demands and utilization are high.

The only thing that really matters when choosing channels is channel capacity. Why not choose the best RF channel with the highest capacity?  A new technology called ChannelFly does just that.

Introducing ChannelFly

ChannelFly is a statistical, adaptive channel selection technique optimized around maximum system throughput.  It relies on real time observed capacity on all channels in both 2.4GHz and 5 GHz frequencies to automatically move clients to a better channel with less interference and higher capacity.

Borrowing patented techniques from our popular BeamFlex technology that automatically selects the best Wi-Fi signal path based on actual acknowledgements from each client, ChannelFly learns the proper channel because it knows, at all times, what the capacity is of the channel.  Our new and unique approach involves assessment across all available channels, based on actual realizable capacity, before making any decisions about changes. 

ChannelFly achieves superior results by utilizing actual channel throughput measurements rather than just listening and guessing. The AP will always be in service for its clients.  ChannelFly learns each channel’s conditions by periodically changing channels across 2.4GHz or 5GHz bands. This can be done with or without active clients.  

Once representative statistical capacity data for each channel is learned, the AP will select the best channel less frequently but continue to monitor channel capacity. If a significant drop in capacity occurs on the current channel, ChannelFly can immediately react by switching to a better channel. Operating with or without client activity enables administrators to allow APs to self-tune before any client devices are connected, therefore greatly improve overall channel performance with minimal network disruption.

With ChannelFly in very congested environments, customers have observed a 25% or more increase in channel capacity. ChannelFly, like adaptive antennas, is another advancement by Ruckus to improve the capacity and reliability of Wi-Fi. Because in the end, people only care about two things: 1) connecting reliably and 2) going fast. And we couldn't agree more.

Comments

tom costa

nice information, thanks for sharing

Clint

Can you give more details on how picking an overlapping channel will ever be the best choice? Choosing anything other than 1,6 and 11 will give you only 2 channels available that don't overlap. And picking one of them in an area of high interference means you're interfering with 2 neighbors rather than one.

It seems like a total non-sequiter to go from saying that the standard non-overlapping channels aren't good to how Rukus does automatic channel selection.

Can you give some real data on why Rukus wouldn't be better off doing it's automatic channel selection limited to channels 1, 6 and 11 to choose from?

Lastly, systems like yours make it a lot harder for the rest of us trying to find a channel to use because your APs are adding interference on 2 channels that we could have used rather than just one.

TheRuckusRoom

Choosing channels 1, 6 and 11 has long been the standard and in many cases are the proper channels to choose. If you are in a perfectly clean area ChannelFly will choose these channels. However, when deploying Wi-Fi in certain environments such as heavily congested public areas (e.g. New York City) choosing channels 1, 6 and 11 are usually the exact WRONG channels to use.

The right channels to choose in any situation is the channel that gives you the best capacity. The interesting thing is, if an AP chooses the channels with the best capacity (like ChannelFly does) it has the LEAST impact on surrounding systems. In a given area, you can’t look at Wi-Fi as multiple APs and clients on multiple different networks. In that given area, all Wi-Fi works together. Different APs, SSIDs etc can’t be viewed as separate systems; they all have an effect on each other.

So the next question is, how can a transmitter (AP or STA) on an overlapping channel be better than a transmitter on 1, 6, or 11? It’s like many things in Wi-Fi, it’s all about signal to noise ratio (SNR). A device can successfully transmit on an overlapping channel if the SNR is adequate. SNR is achieved through channel separation (even between channel 1 to 3 you get some separation) and attenuation via distance or obstruction.

ChannelFly can calculate the capacity on a channel without actually running a throughput test. This is what ChannelFly is doing all the time and using that information to choose the right channel. As an experiment, run a throughput test on each channel (1-11) with one AP in a a given environment. If you do it in an ultra saturated area you’ll see that channels 1, 6, and 11 will have the worst throughput and therefore are the worst channels that could be chosen.

Post a comment

Comments are moderated, and will not appear on this weblog until the author has approved them.

If you have a TypeKey or TypePad account, please Sign In.