EDIT: The pcitures were huge and were getting cutoff, so I resized them quickly only to make them look really bad, and hard to read, so instead now they are thumbnails — click on them to see full size images.
———————————————————————————————–
I came across this article on an antenna intended for MIMO (multiple input, multiple output) applications, that I feel is impressive from the standpoint of small size, but not for the dual feed system it purports as being revolutionary. The antenna is from the company Skymap, and the product is their iMAT antenna .

Now, for MIMO systems you want to be able to used different bands or spatial regions within which to send or receive signals, since this allows capacity of the system to be increased – for example, allocating different spatial regions for different channels allows the reuse of the same frequency band. A good example of this is the cell phone network — by dividing up geographical regions into “cells” with their own set of antennas, the same freqeuncy bands can be used at the same time by callers in nonadjacent cells, in contrast to having one large powerful transmitter cover all of the area.
Anyway, a lot of articles I find talk up the discovery that you could use two feeds on a single antenna to tap into either different polarizations or frequency tuning by placing feeds at voltage nulls of the excited mode of the other probe. This is a bit misleading, however, since this idea is not new, and has shown up in many types of antennas. Since I have spent a lot of time working with microstrip antennas, I threw together a quick example of an antenna that utilizes two feeds to tune to 2 separate frequencies, and each frequency also has a different polarization – this was just done to show two separate polarizations and two different tunings. With microstrip (patch) antennas, you can also tune multiple frequencies on the same polarization, or multiple polarizations on the same frequency, etc. Below is an image of the antenna, drawn up in Ansoft HFSS.
This is a simple rectangular patch on 30mil duroid substrate(relative permittivity of 2.2) that is 5cmX4cm, and is fed by coaxial probes from underneath along each axis, as shown below:

The center conductor of each probe is brought up through the ground plane and is terminated at the patch. They are positioned to match the antenna impedance to 50Ohms, and the coaxial cable feeding the antenna has a 50Ohm impedance. Each probe is along the center line of the rectangular patch, which places it at the voltage null for the mode the other probe is exciting. As a result, each probe can’t really “see” the other probe, since the other probe is located at a virtual short, and so each probe acts as though it is exciting it’s own antenna. This reduces coupling between probes and ensures each probe excites the dominant mode of the patch along it’s dimension and will have a linear polarization with very low crosspolarization (which can arise from asymmetric feeding under the patch). Below is a diagram showing how the field polarization and frequencies are orientated along the orthogonal x and y axis:

It should be noted that the x and y polarizations are themselves orthogonal, and would allow individual polarizations to send signals on the same frequency, and the separate frequencies could allow multiple channels on the same polarization. Anyway, just to show some technical specs, below is the Return loss of the antenna:

The red curve shows the response of the X axis coax probe, and the purple shows the Y axis coax probe. A return loss of -10dB is a match of 2:1 VSWR, which is a typical bandwidth spec, giving each band 10MHz and 18 MHz of bandwidth. What this shows is that each probe only tunes at one frequency, severly reducing coupling between the probes, which was below -30dB in this simulation (although isn’t shown). Shown below are the impedance loci on the smith chart: 
Next art the gain patterns for each band, first the 1.962GHz band and then the 2.422GHz band.

Here Brown is the E-plane pattern, and Red is the H-plane pattern , with crosspol shown as the small circles down below -25dB, which is typical for a patch antenna.

Here Blue is the E-plane pattern, and Purple is the H-plane.
The bottom line is that this example is very simple, and has been exploited for lots of applications where dual frequency, dual polarization, etc, are desired. So feeding a single antenna with two feeds is not new as these articles make it sound. Don’t get me wrong, the antenna in the article is impressive, and is very small and has great performance, but I think they are wrong in saying this is a new idea.