The Triple Feed Patch antenna is a directional circularly polarized antenna that I created in collaboration with Robin Theunis, one of my co-workers. The main design goals were:

Good radiation efficiency. Most FR4-based patches have terrible efficiency because FR4 is very lossy at 5.8 GHz.
Good axial ratio, not just directly perpendicular to the patch but also off-center.
Cheap and easy to manufacture with high accuracy, even for a hobbyist.

(strictly according to Maarten Baer design requirements, learn more, please click: http://www.maartenbaert.be/quadcopters/antennas/triple-feed-patch-antenna/)

PRODUCT BASIC INFORMATION

Center frequency: 5.8 GHz
Bandwidth: 660 MHz (5.47 – 6.13 GHz)
Matching: S11 < -20 dB, VSWR < 1.22 (at center frequency)
Axial ratio: < 1.3
Antenna gain: 9.4 dBi
Half power beam width: 55° (horizontal and vertical)
Radiation efficiency: 83.5%

These values are based on simulation, actual measurements are below (they are very similar). The center frequency may shift a bit between batches as a result of variations in the dielectric constant of FR4. Also, not all brands of FR4 have the same dielectric constant. Frequency shifts up to 100 MHz are normal.

Simulation model. The two plates are wave ports that act as stand-ins for the SMA connectors.

3D radiation pattern. Look at the 2D version if you actually want to see the numbers.

Vertical cross-section taken at various angles.

Measurements

VNA measurements (S-parameters). S11 and S22 indicate the reflected power at port 1 and 2 respectively. S12 and S21 indicate how much power is leaking from one port to the other.

This gives you an idea of the kind of variations you can expect for hand-assembled antennas. I suspect that most of the variation is caused by the surface-mount SMA connectors since those were soldered by hand, and it’s difficult to accurately control the amount of solder when you can’t even see the center pin. This would explain why there is a lot of variation in S11 and S22, but almost no variation in S12 and S21. Based on S12 and S21 (which should be more reliable), the center frequency appears to be close to 5.7 GHz, which I suspect is a result of inaccurate spacing of the PCBs. The 4 mm spacers which I used to construct these units somehow resulted in a final spacing closer to 4.2 mm in reality.