It seems every few months somebody brings this topic up and we beat it again. Perhaps we should have a sub-forum filled with stickies of 'dead horse' threads. Let me take a stab at beating this one to death.
In reality this is simply one variation on a technique that has been used for many decades in many domains. VLBI, direction finding, interference cancellation and directional antennas are common examples. In fact it wasn't uncommon to assemble arrays of TV antennas to improve fringe reception not that long ago. That was easily accessible to hobbyists because the wavelengths were long enough that RF combining was practical. Things are a little different at 12 GHz, compared to 200 MHz.
I have argued a number of times that this technique is for the most part economically irrelevant to FTA. That does not mean it would be impossible to do, just far more expensive than putting up a bigger dish, which is what I think brings most people to the topic on this forum.
However I suppose there could be some situations where someone might legitimately consider it. If you have HOA or government rules stopping you from putting up an adequate sized dish, you would have few other choices. This also turns out to be an incredibly selective way of canceling interference from adjacent orbital positions. While bigger dishes typically address this more economically, it still could be a reason to consider this. As the mini-BUDers have found, there is often a lot of signal in a C-band footprint, but adjacent satellite interference cannot be addressed with a single small dish.
So to clear up misconceptions, I thought I would review a bit of the theory. To start with, lets look at a 12 GHz signal. We could do C-band, but the result ends up about the same. A 12 GHz signal has a wavelength of 2.5cm. If we want to have maximum combining efficiency, let's try to add the signals within 5 degrees of phase of one another. That would be within about 0.35mm of path difference at RF. Not very easy.
However if we use a stable frequency source we could use true PLL LNBs from more than one dish to take this down to around 1 GHz. Stable frequency sources are pretty easy to come by. A few years ago I picked one up for $10 surplus and only had to make some minor fixes to the oven drive circuitry. After that I could lock it to WWVB with a simple but extremely long time constant PLL and it would hold better than 1 part in 10^12 (1 second in 30000 years). True PLL LNBs are not cheap, however.
What does this buy us? Well, phase is what is important, not frequency. That's one reason why WWVB, which broadcasts at only 60 kHz, can be used to make such incredibly accurate adjustments to a frequency source. At 1 GHz, a 5 degree phase error is about 4.2mm. That's starting to look possible, but probably not practical for the hobbyist to trim cables. However if you had an analog phase shifter that worked at this frequency, that might be good enough.
It's also important to note that the paths from each PLL LNB don't need to be exactly the same length, but only be in phase, unless the path differences start to become a major fraction of a symbol. If we take a SR of 30000, the wave length is 10m. Applying our stringent 5 degree criteria, we need the maximum path length difference to be less than 14cm. Not a big deal.
Since we're heading the right direction, let's mix and filter the LNB outputs in the comfort of our home with a generator, say at 950 MHz and also locked to our reference. Now the signals will only have to be within a phase difference of 8.4cm because the IF is 50 MHz. I could trim cables to do that, but would need a way to vary this for different signals. We also have to compensate gain, but that's not very hard. At this point the signals could be easily combined in a number of ways. Mix the result with our 950 MHz generator and we're back to 1 GHz and the input of a standard FTA receiver.
None of this would be easy, but a well motivated hobbyist with a good understanding of the theory could likely do this for less than $2000, and perhaps half of that, with careful selection of surplus equipment and finds on eBay. You might have to buy the LNBs new. Still people here buy receivers at around $400 fairly frequently, so we're not totally out of the ballpark.
I've glossed over the details, but commonly available equipment could do this. It just costs. You don't have to do this at the LNB. You don't need to get down to 5 degrees for useful combining. And if you did this digitally, you could obviate many of the complexities and compensate in software. I'm not encouraging anyone to do this, and still believe there are easier solutions in most cases. But it is possible. For most cases, use a bigger dish.
