Speaking of antennas for 3G 4G and Wi-Fi, we cannot ignore the patch antennas. The emergence of such antennas arose with the development of microstrip microwave technology, in particular the theory of long microstrip transmission lines. The idea of such an antenna was first proposed back in 1954, but at that time there were no substrates for printed circuit boards with low losses, and it was forgotten. The second birth of the patch antenna took place in 1972 after the report of the Robert E. Munson at a scientific symposium. The antenna was designed by him for moving objects, and even now it is especially popular with FPV fans.
Microstrip antennas on a printed circuit board are used in almost every modem or other device as a built-in antenna. Such antennas are very technologically advanced and have a low cost for mass production. Apart from fractal structures, printed antennas can be dipole, slot, flat two-dimensional structures. The latter are a printed emitter in the form of a square (rectangular) or round (elliptical) plate located above a dielectric layer with a large metal screen, in other words, made on the basis of a double-sided PCB. These are true patch antennas.
Consider what a rectangular patch is.
This is a resonant system with side lengths L and W close to half the wavelength, taking into account shortening in the dielectric and the effect of the screen. This design can be considered as a cavity resonator open on all four sides. Without going into theory, we will keep in mind that both sides with length L do not radiate, but the slots of the sides W do radiate. Thus, such a patch is a stack of two elementary half-wave emitters. On the other hand, it can be considered as a radiating rectangular loop, in which instead of an electric current (as in a wire loop), an imaginary “magnetic current” flows in a slot along the perimeter of the patch.
The disadvantages of printed antennas include their low efficiency and relative narrowband 2Δf/f0 < 5%. To expand the band, special tricks are used, for example, a U-shaped cutout in the center of the patch. If air is used instead of the substrate and the distance from the screen is increased, then the losses in the antenna are significantly reduced, and the band is expanded up to 15%. The bandwidth is directly proportional to the h/W ratio. It is antennas with an “air” substrate that should be used in the 3G-4G bands. Well, the next logical step. If we add a collecting structure in the form of the same square passive elements to such an air patch, like directors in Yagi-Uda, we can significantly increase the gain of such an antenna.
The gain of a patch antenna varies from 5 to 9 dBi and is highly dependent on the parameters of the dielectric. The higher its dielectric constant, the smaller the patch sizes and the worse the directional properties of the antenna. In addition, the presence of a dielectric reduces the efficiency of the antenna. When powered from the edge of the patch, the input impedance of the antenna is close to 100 ohms for the “air patch” and 2-3 times higher than this value for the printed one. 
To match the 50/75 ohm feeder, the feed point is moved closer to the center of the patch. At the same time, tuning to resonance is maintained and there is no need for special matching devices, which is another advantage of the antenna. Although, no one bothers to use circuits with quarter-wave impedance transformers on microstrip transmission lines. In addition, it is relatively easy to create a high-gain multi-stage antenna array from such patches. It is these designs that are sold under the name of panel antennas.
If we connect a second feeder along the Y axis to the patch at the same distance x0 from the edge, we will get the simplest MIMO antenna that can be used in the near reception area or as a dish antenna feed in the case of a far area from the base station. Just by cutting off two opposite corners of the antenna, we get circular polarization, that might be useful for GPS or FPV, but for 3G-4G, Wi-Fi antennas this is not relevant.
In conclusion, a few words about the calculation of such antennas. In principle, they can be calculated using 4NEC2 and similar programs based on the NEC kernel. For this, the patch and screen are presented as a mesh with a very large number of intersecting wires. But not the fact that you will get the correct result in the output. So it is better to use professional electromagnetic simulators that use the finite element method for calculation, for example: CST Studio, HFSS, Comsol Multisystems, EM.Cube etc. A simplified prior calculation can be done on our website using the online patch antenna calculator.
Related links:
- Panel antenna "Ellipse-patch 2x2" for the 2100 MHz band with a gain of 15 dBi;
- Panel antenna “Double Ellipse-rectangle” for all 3G/4G bands;
The site has comments on social networks, which have much more opportunities for communication.