Proper way to combine same-frequency PCB antennas

Question

I want to combine 2 identical PCB antennas to increase gain. The antennas are designed to work on 2.4GHz and have 50ohm impedance, linear polarization with Omni-Directional radiation pattern, they are supposed to be used in this way: a combined couple horizontally oriented in the bottom of a weather balloon, and another couple in ground station (this one could be used vertically oriented eventually). The goal is to improve communication between the weather balloon and the ground station. I'm wondering what would be the proper approach to do this without mess with antennas frequency range reception and worse performances in desired frequency range rather than improve them.

E.g. Consider the following scheme.

Is combining them in following way correct? If no, what should I expect from this configuration and what would be the proper way?

NOTE

I'm not talking about "antenna diversity" (that is based on the use of multiple different antennas that work separately and it requires additional hardware and integration), but to create a single bigger and more powerful PCB antenna combining 2 smaller antennas together.

Answer 1

There is no one correct, there's just correct for your situation.

We live in a universe where energy (and, hence, power) is conserved. So once you've made your antenna reasonably efficient, the only way to increase gain is to make it directional.

Making an antenna directional means that in transmit, energy is radiated in a restricted number of directions (usually one by design, with some parasitic sidelobes). In receive, the antenna has the same pattern, it's just that it doesn't respond to signals in the "off" direction, only in the directions where it is sensitive.

The more you make your antenna directional, the more important it is to have it pointed in the correct direction. If you had a purely omnidirectional antenna, then it would radiate and receive equally in all directions. If you make an antenna with 20dB of gain, then it must have an effective main lobe area of roughly \$0.04 \pi\$ steradians (out of \$4\pi\$); e.g. one that radiates to a circular patch would have a beamwidth of about 36 degrees if I'm getting my math right.

If you put a highly directional antenna somewhere that the orientation is changing in an uncontrolled manner -- such as a weather balloon -- then you only get a strong signal back when the antenna happens to be pointed your way -- the rest of the time it's worse than an omnidirectional antenna.

What you need to do, in your situation, is to first design the desired antenna pattern for each station. You need to keep in mind that you have a balloon that's going to take on fairly random directions. Then you need to look hard at the requirements for the ground station -- i.e., whether you can actively point the ground antenna, what you would do if a high-gain antenna loses the direction to the balloon, etc.

Once you have your candidate antenna patterns designed (and for any old balloon and any old ground station it may be that omnidirectional is, indeed, best) then you can worry about how to achieve that pattern.

Answer 2

If the individual elements are linearly polarized and "omnidirectional" then what they really have is a doughnut-shaped pattern. With the antennas vertical, then they are omnidirectional in the horizontal plane.

If you mount the antennas in a parallel configuration, as you drew, it will result in a directional pattern with peaks and nulls.

If you mount them in a straight line, one above the other -- called a collinear array -- then you will still have an omnidirectional pattern, but the "doughnut" will be flattened somewhat as it's been stepped on. If you look at TV and FM antenna transmitting towers, you'll see multiple elements stacked vertically in this manner. It provides more power along the horizon by wasting less power heading up into the sky and down toward the ground.

But you must also create a divider/matching network to feed both antenna elements in phase. You can't just connect them in parallel, because the result would be a 25 ohm feedpoint impedance. The network can be designed in microstrip but that's beyond the scope of the question.