Cisco Validates Smarter Wi-Fi, Sort Of.

Informants have hinted that Cisco is planning to unveil a new line of 802.11n dual band access points next week.

What's more Cisco says these APs will use the "built-in" 802.11n beamforming functionality for the first time. 

Beamforming is an option in the 802.11n standard and has been integrated into the 802.11n chipsets provided by Atheros and Broadcom.

No surprise to most, Cisco seems to be following the path of least resistance by relying on chip suppliers for each and every morsel of RF technology advancement.

That said, Cisco's validation of beamforming is a big acknowledgment to the industry that more needs to be done to make Wi-Fi more reliable at the physical layer. Ruckus was conceived around this concept - making WI-Fi more reliable.

The problem is, the "beamforming" called out by the 802.11n standard does very little to solve this problem.  In other words, not all beamforming was created equal.

Beamforming can essentially be performed in two ways:

1. Mathematical beamforming
   This is achieved through digital signal processing in lower levels of the chipset (baseband and multiple radios). This gets all the academic attention and what most people mean when they mention beamforming.
    
2. Physical beamforming
   This is achieved through the use of adaptive directional antennas and best path selection algorithms that dictate that actual form and direction of radio signals through the RF domain using thousands of antennas and actual client feedback (click on figure below) to optimize things.

Mathematical beamforming at the chip level tells the system which antennas to use for a given client and has theoretical maximum limits (eg. 3dB of gain for two radio chains). But nearly every 802.11n access point on the planet uses omni-directional (rubber duck) antennas that constantly blast out and receive signals in all direction providing next to no way to combat environmental problems.

This type of beamforming can't optimize the actual form and direction of Wi-Fi signals and has no real-time adaptive capabilities. Therefore it can't determine the actual performance of a given path, change it if there's a problem or reject or avoid interference as it is experienced.

Testing has shown that the slightest change in the way 802.11n antennas are pointed or the AP is positioned results in wildly different performance levels.  We've seen fluctionation from 5 Mbps to 80 Mbps in performance by simply moving the AP 90 degrees or the client to a different locations.

Physical beamforming goes waaaaaay further. Physical beamforming adds a whole "subsystem"  on top of the standard chipset that allows complete control over Wi-Fi signals. The basic idea is to improve performance by creating several independent signal paths between the transmitter and the receiver.

What's important here is that physical beamforming is adaptive - constantly adjusting Wi-Fi performance based on real things happening in real time.  Here's how it works (for the most part):

A miniaturized antenna array provides thousands of antenna combinations that smart software algorithms use to form very concise and optimized signals. These same software algorithms use actual feedback from each client to select the best performing signal path at any given time.  We use the analogy of holding a flashlight in your hand in a dark room vs. turning a flood light on overhead.

This translates into three very important benefits for users:

1. better (read more consistent) performance over longer distances
2. more reliable connectivity (interference rejection helps avoid Wi-Fi "flakiness")
3. Non disruptive (no client support required to achieve these benefits)

While mathematical beamforming requires chip-level cooperation from both sides and hence requires standardization, physical beamforming gets that cooperation for FREE from the 802.11a/b/g/n protocol. And with physical beamforming there are effectively no theoretical maximum gains limitations. With our "smart antenna" system we've seen system gains of 9dB and interference rejection of 17dBi.

So now you know.