There are a lot of weird articles out there. Out of respect, I’m not going to reference anything now, but if you’ve been around a bit, you know what I’m talking about. Specifically, in the field of Antennas, some person designed an antenna in a given time or place and wrote an article about it.
Disclaimer: I am mostly discussing printed antennas here.
Let’s get down to actual examples. In some WiFi\BLE\Cellular Antenna designs (Be it multi-band or single), a similar design to the one below will show up:
Wow, much bands, small size. I will definitely call this a cactus Antenna. For the purpose of this demonstration, I aimed for two bands: 
The Sin
Quite a big difference there, especially in the higher frequencies. Well, why is that? There are several issues to discuss here. One, is what you might call an “unbalanced feed”. As I stated before, port definition is one of the most important aspects of electromagnetic simulation. The port defined in the simulation above is a discrete (or lumped) port. Namely, a time-harmonic current source with a fixed impedance.
Apart for the fact that there is no such thing as lumped in a high-frequency environment, this port does not emulate a feasible feed. So on top of the discontinuity caused by the transition, we are also getting an unbalanced feed. Realistically speaking, such a single sided grounding doesn’t allow soldering on a connector reliably. I tend to like my designs repeatable and measurable.
So what is the reason for the first result set, then? Well, basically this is a differential feed for this Antenna. Anything that goes between a differential and a single ended (e.g. coaxial cable) device has to go through a BALUN (Balanced-Unbalanced). Not saying this is impossible, but let’s consider repeatability again. The coaxial cable has to be exposed each time, in the same length, the same lead has to be kept out. Then, it needs to be soldered on at the exact same position. Possible, but isn’t a connector just that much simpler?
The Solution
So what am I suggesting? Well, do yourselves a favor and start your designs from something predictable. A well characterized transmission line, for example. At some day and age someone may want to use this design. For that, physical connections are a must. Use well grounded lines with no chance of surprise. Don’t let the ground be the radiator. If necessary, design a BALUN afterwards, or solder on one, whatever floats your boat.
Well, as my Sensei says, stick to the basics. Print well defined transmission lines, Shunt vias to keep return signal paths valid, maintain sane design rules.
The Caveat
Fine and dandy, but there is more than one printed Antenna type. What about flexible Antennas? The grounding there is anywhere between very hard and nonexistent. Material characterization is also not always reliable and stackup thickness also tends to be unpredictable. There, the keyword is assembly. You have to figure out a good assembly solution. Also, this is usually not relevant for integrated Antenna design, so I will speak of it no more (here).
Here was my attempt of a practical Antenna design post. Hope you enjoyed!
The Gratitude
A special thank you for the InventSim team, that allowed me to try out their software. It has proven to be a reliable and very useful tool!