Depends.
Anything from totally mess everything up, to nothing... depending.
Yes, but only one antenna at a time, and only transmit or receive at one time... and on which antenna, again depends.
If you starting to get the idea this *all* depends... It *does!* (And eventually, after all of these comments, I'll try to explain some of it.)
Depends.
Depends.
Probably not though. Think about that for a minute. If you crank it all the way up, that's 255 mW of power, or 1/4 of a Watt. If that chip is only 25% efficient, that means it is dissipating 3/4 of a Watt.
So reduce the power all the way down to the default 28 mW, and if that chip is still only 25% efficient, the chip will be using
112 mW of power, 28 wW will be RF output, and 84 mW will be heat. So you've changed the heat dissipation from .750 W to .084 W, or a reduction of 2/3rds of a Watt. 2/3rds of a Watt isn't what's heating up the box... and if you reduce the standard 28 mW, you obviously won't get anything close to any 2/3rds of a Watt reduction.
I have a wrt54g that has a single antenna, and of course both tx and rx are locked to that one (if you do that, use the antenna socket nearest the power socket, as it has a much shorter connection inside the box).
As I'll explain down a ways, you might be better off moving the
1st Floor units to the 2nd floor, and the 3rd Floor units to the 5th floor. Hopefully (you'll have to test and see), each unit can cover one floor above and one floor below. The question is, will it go through the roof, and do you need that unit at all.Since you are going up and down, horizontal polarization might be better, depending... if there is any location where you can point the two ends of the antenna toward areas that will never be used by a client, it will be better. That is particularly true because you are using one unit in the center of each floor.
An example that does not fit your situation, but which might illustrate the point, would be one large rectangular room with two AP's. If they are located about 1/3rd of the way into the room from each end, perhaps vertical polarization would be best. I they are at the ends of the room, horizontal polarization would be better (especially if a reflector could be added to bounce what would go out the wall back into the room).
Think about the "donut" shaped pattern around an antenna, and see if that makes sense. If it doesn't (for example if the idea of a "donut" shaped pattern doesn't ring a bell), just say so and I'll let Jeff Liebermann try explaining it! :-)
Wow! You've got a *lot* of stuff that this depends on!
First off, lets understand what those antennas are doing and how they are switched, so that you can get a better idea of the effect of any given configuration.
Either the transmitter or the receiver, connects to one or the other of the antennas. When a receiver is connected, it will switch antennas at regular intervals (whether it is receiving a signal or not), looking for the best signal. Hence if you have two very different antennas, it will eventually pick the best one for any given signal.
But the transmit *always* starts with the antenna last used to receive. That assumes it will also be the best signal, but it just ain't necessarily so! Lets say you have two highly directional antennas, one pointed north and one pointed south. If a client to the south transmits, it may take a bit, but the receiver will eventually pick the south antenna and will see the signal. At that point, no signal from the north will be received until at least the interval it takes to test the other antenna, so a second client to the north is going to wait much longer to connect than would another client to the south.
Worse though, is if the receiver is seeing a steady signal from south (somebody doing a big file transfer for example), and a packet of data needs to be sent to a north client. The transmitter will *always* use the last antenna used for receive, which in this case is *always* going to be the south antenna. The north client will *never* see it, until it happens to transmit something itself and cause the north antenna to be switched on for receive.
Which is to say, except for point-to-point links, you may want only antennas that are very nearly the same. If you have only one client, and that client roams widely, then the increase coverage would be fine, and there would be little effect from switching disparate antennas. But if this is an office building or an hotel, and there might be many users connected over this wide coverage, disparate antennas being switched might cause them significant annoyance.
That applies to polarization in ideal situations, but less so in real situations. In your case, the amount of multipath from various reflecting surfaces is almost certain to cause polarization of received signals to be all over the map.
Specifically, since you are getting good coverage, you obviously have a workable configuration! The only thing I'd suggest is checking to see if that rooftop unit really needs to be on the roof. On the roof it probably only provides coverage for one floor below. But if you put it one the top floor, will it then cover the roof, the top floor, and the next one down? If so, you get three floors with one unit, and you might lessen the amount of off premise exposure too. Of course, if the roof is metal, that won't work...