Recently I had a problem with a new 7km link to a tower using two QRT5ac. The overall perfomance was quite bad. After several days of tweaking (changing frequencies and parameters, I also tried to switch to N-only), I decided to mount a QRT5 (with N-only) at the tower - and all problems are gone. The link performance improved to 100MBit up/down (UDP-BW Test) at the same frequency and bandwidth (40MHz), before especially the uplink quality to the tower was quite low (at about 10MBit, very fluctuating). The 5GHz band at the tower is quite full (many other links mounted here). It seems that the AC-Devices are much more sensitive to disturbances from other 5GHz equipment than the N-Devices.
Another point: It seems that the links are also need some "training". I also installed a QRT5 at the client side. At the day of installaion the rates are quite ok (CCR at 70% and about 70MBit/s UDP-BW test), one day later, I got CCR at 95% and 100MBit/s at the UDP-BW test, and now 5 days later it is still the same - so it seems there must be some link training in the N-Devices...
My settings are a little bit special: to have latencies with low jitter I always set nv2 to 1ms and I diabled all HT MCS above MCS5 (above MCS13) so that my max connection rate is 240MBit/s. I now get about 60MBit/s TCP speed (transfer rate of a browser in win 7) over this link.
My question: What is wrong with the AC chipsets? Does anyone here have similar experiences?