I started to think on this a little more and tried to see where they were going with this train of thought. There are instances where you need a patch, yagi or otherwise sectorized antenna - usually when getting the longest range is the goal of your wifi deployment. This being said, I'd hardly agree with the statement that omnidirectional antennas suck.
If you're deploying a wireless infrastructure to support location tracking - omnidirectional antennas are your best friend. Since the horizontal lobe of an omnidirectional antenna spreads out in a 360 degree fashion - this RF spread pattern aides the access point in reporting client location data back to a Cisco Mobility Services Engine (MSE) or an AeroScout Engine (for non Cisco shops).
The Cisco Wi-FI Location Based Services 4.1 Design Guide has a good graphic that explains mathematically how the location of a client device is calculated based on tri-lateration (using the combined client RSSI data from at least three or more APs).
- ...the concept of ToA tri-lateration. The amount of time required for a message transmitted from station X to arrive at receiving sensors A, B, and C is precisely measured as tA, tB, and tC. Given a known propagation velocity (stated as c), the mobile device distance from each of these three receiving sensors can then be calculated as DA, DB, and DC, respectively. Each calculated distance value is used to construct a circular plot around the respective receiving sensor. From the individual perspective of each receiver, station X is believed to reside somewhere along this plot. The intersection of the three circular plots resolves the location of station X as illustrated in Figure 2-3. In some cases, there may be more than one possible solution for the location of mobile device station X, even when using three remote sensors to perform tri-lateration. In these cases, four or more receiving sensors are employed to perform ToA multi-lateration.
OK - now that we know how a group of APs send the client RSSI information to an MSE or AeroScout Engine to calculate the approximate location of a client device - how would an access point that uses sectorized antennas calculate a client location using this same type of RF location calculation? I would think that the access point using sectorized antennas would have to have a central "brain" in the array to combine all the client RSSI information that could be "heard" by the separate arrays. Every sectorized antenna has some sidelobe to its RF beam, and surely antenna arrays arranged close next to one another would have some overlap, and therefore neighboring array antennas could potentially "hear" the RSSI of a given client device.
colorized array RF footprint to show channelization of the arrays' antenna orientation
My point is - from the information at hand regarding how client locations are triangulated, I believe that using patch, yagi or sectorized antennas would adversely affect the accuracy of location tracking. The RF footprint has been changed to such a degree by the use of a sectorized antenna that the mathematical calculations used to determine client locations are no longer valid.
The same could be said of using sectorized antennas in a deployment utilizing Radio Resource Management (RRM). RRM calculations are based upon each AP "hearing" a given number of neighboring APs at a given RSSI signal strength (usually -75dBm). If antennas that have non omnidirectional lobes (patch/directional) the access points nearer the smaller back lobe of a patch/directional antenna would most likely sense this neighbor with a patch antenna at a lower RSSI and attempt to adjust RRM to power up in order to compensate for their "less heard" neighbor.
Point being - if you're using RRM and patch antennas - you should be very aware of how your access points are deployed and perhaps statically assign power levels to APs that are RF neighbors of the AP using sectorized antennas.
-- I invite additional information about how an array type access deployment handles calculating clients' RSSI and turning that into an X,Y location.. I'd like to understand how that works.