The way that access point you show is set up is the 'obvious' way from a radio engineering viewpoint. So called omni-directional antennae are not truly omni-directional. This kind of antenna is designed to be (roughly) vertical, and when so positioned they actually focus the transmission sigtnificantly in the horizontal plane, at the expense of transmitting less vertically. Generally this is what you want - your access point is typically trying to reach stations elsewhere on the same floor of the building (or in this case on the same platform) and not trying to use it to connect users on other floors of the building, which would be better served by an access point on that floor than by signals that would anyway have to pass through lots of concrete and metal.

This kind of antenna will have a quoted gain of maybe 3db-4db. What does this mean? It means the signal is twice as strong or even a bit more _in the direction(s) it focusses the transmission_ compared with what would you expect from a perfectly omni-directional antenna (a so-called isotropic radiator). How is this possible? Obviously the antenna is just a piece of metal and plastic, it can't actually make radio energy out of nothing! So how does it achieve this seemingly magical field of doubling the amount of radio power? Ofcourse, it doesn't. It's transmitting twice as much power as you would expect in the horizontal plane because it's transmitting a lot _less_ than you would expect above and below the horizontal.

So given you know that, you now understand why the first instinct of a radio engineer is to make the antennae vertical, And vertical is certainly the right thing to do if you only have one antenna.

But what do you do when you have 6 antennae? And why do we have six antennae anyway? Well, the first part of the answer is that this will almost certainly be a simultaneous dual band access point. It was two independent radios, one for 2.4GHz and one for 5GHz. These are completely independent. But each radio still has three antenna. Why? Well, one reason is diversity. The vagaries of radio propagation mean that for any particular staton the access point is communicating with, one antenna might pick up a better signal than the others. But there's more. 802.11n and 802.11ac use MIMO, which actually relies on the fact that the signals from the different antennae will take different paths to the station. So for both these reasons, it makes sense to vary the angles of the antennae somewhat, to exagerate the differences in signal picked up by each antenna. You certainly don't want any of the antennae horizontal, because then the blackspots, where the antenna transmits a very much weaker signal, would be on the floor you are trying to serve rather than above and below. (Don't worry too much about someone directly below the AP, they will be so close they will get a good signal no matter what.) But is *probably* does make sense to angle them away from the vertical (in different directions) so as to maximize the effects of diversity and to help MIMO do its job. So the picture that TP-link posts here makes sense.

http://www.tp-link.com/no/FAQ-455.html

So, perpendicular doesn't make much sense, but angled away from the vertical does. Probably. But as you say, the best way to find out what works best is really to experiment.

roy