I forgot to mention my extended post for the same matter;
viewtopic.php?f=7&t=124499
Here it is explained what the setup and performance test was.
All the CPEs from one AP simply MUST have very good isolation from other AP(s) and vice versa.
Well, imho the whole Idea of running two AP's (or more) is that there is no more interference between the AP's on the same tower in the same frequency.
The AP sends and all stations listen, the stations can then send when the AP's are listening.
Off course now it could be that a client of AP1 while transmitting his signal will hit AP2. And vice versa, client of AP1 while in receiving mode might also receive signal from AP2.
We only have to hope that client only communicates with its designated AP's signals and not the other.
With Mikrotik SXT it is almost impossible to have 'very good isolation' between the signals of both AP's.
An SXT that sits 100 meter away from AP1 and is designated to pick its signal, will also pick a signal from AP2 on the same tower even if that AP might be directed in the opposite direction.
The question now is more, what S/N distance (or pure signal difference) will client get/send to AP1 compared to AP2? Is 6dB enough? (twice the signal), 10dB?, 20? 30 or more? What level are we seeking for to be considered as a good 'isolation'?
I could imaging replacing the two SXT APs' with two netmetals and RF elements horns for a 'very high isolation' at the tower. But that would then only work in an 'open field' scenario. In an urban environment there will be signal reflections and bouncing so SXT-CPE A sending signal to AP1 will still 'hit' the AP2 it is not associated to. Due the sync this AP will pick this signal! And that can be pretty strong!
Some clients can be located close to the edge of one AP's signal sector and thus also close to the border of the next AP. If they are both in sync, the signal difference between the two AP's from or to that client is marginal, no matter for instance the ftb ratio of the AP's or the limited side lobs.