Assumption: your antennas are placed such that they don't directly couple, i.e., in a way that they are practically in a null of the other antennas' patterns. Otherwise, you're not designing a stacked antenna system, but one giant 3D antenna, and the only way to do that is honestly simulation and loads of experience. Rule-of-thumbs quickly break down for complex 3D antennas, and a lot of the "results" you find online are very questionable. This assumption is, however, well justified, as Yagis can well be placed parallely with > 1/2 wavelength distance.
For a yagi, you have no choice, essentially: you feed the driver element of each individual antenna. You'll hence need a power divider that takes your one feedline (70cm: probably coax) and divides it for the multiple antennas, such that the impedance still matches. Same for quad, the feedpoint is inherent to the antenna design.
I've seen such antennas being stacked for reasons of increasing gain / narrowing beamwidth (that's essentially the same thing). The question that arises is
In which direction to you want to decrease the beamwidth?
You'd need to stack in that same direction. Want your beam to be narrower vertically? Stack vertically. Want it to be narrower horizontally? Stack horizontally.
From that, what remains to be answered is
- How do you divide the power between your $N$ antennas, and
- How do you want them to be phased?
Let's address 2. first: Imagine you put two Yagis on top of each other, such that they don't directly couple (not too hard, as a yagi has directivity "up front", not to the top). Now, if you supply them both with exactly the same signal (i.e., "in phase"), the waves emanating from them add up constructively exactly in points that are exactly equally distant from both antennas' phase origin, because they have the same phase there. In other words: in the ray in the middle between these two antennas main axes.
Now, add, say, 1/8 of a wavelength in additional cable leading to the upper one. Now, the points where the two wavefronts emanating from both antennas add up constructively is still where the waves are in phase. Only that the wave from the bottom antenna has 1/8 wavelength of an advantage – therefore, the points where the waves add up constructively end up on curve where each point is 1/8 wavelength further from the lower antenna's phase center than from the top one's.
Congratulations! You just implemented beamforming through means of phased array.
So, your antennas' relative phase (this extends to arbitrarily many antennas) defines the antenna pattern you get. You define that phase through cable lengths, or other means of delaying the signal.
So, in conclusion:
trying to figure out how/where to feed these antennas
- how: using a matched cabling, and a power divider, with the desired length of cabling. If you "just" want to increase gain in the direction they already have, stack them equidistantly, and feed all with exactly the same length
- where: like you'd feed any individual antenna.
The question of "oh, I get a lot of freedom when I feed antennas, how to use that optimally" is answered by designing the antenna array to have the so-called array-factor you want. The array factor is like an antenna pattern, only that you multiply it with the antenna pattern of your individual antenna, which in this context is called element factor.
I really can't (not good enough, and also, too much for an answer) introduce you to antenna and multi-antenna design, but it happens so that if you put your antennas in a regular distance on a straight line in space, that antenna factor can be calculated through the DFT of a vector of the individual antenna's excitation factors, i.e., a complex value having the magnitude representing the amount of amplitude of the signal they're getting and a phase representing the signal's phase.
