I'm looking for feedback on real world results with deploying ePMP in noisy environments. We are wanting to replace all of our Ubiquiti AP's with ePMP, even expand on a few 450 deployements where cost is a concern, but are unsure of how the platform will perform with a -80 noise floor in 5 GHz. I'd love to hear feed back on this from you guys...especially when running VoIP :)
what are your signal levels at the AP?
Our current subs average -55 and -60.
It depends. The problem is that a noise floor of -80dBm doesn't actually mean much in the real world. If, theoretically your noise floor was a constant -80dBm (and it almost never is) then you should expect to see MSC15 at about -50dBm. But as I said, when a device estimates the noise-floor it's just that, and estimate, averaged over some time-period. In the real world that -80 could be -90 for one moment and -60 at another. Your RF environment of a -80 noise-floor is going to be different than my RF environment of a -80 noise-floor.
With that said, the performance of the ePMP in busy/noisy situations has been what I would expect to see (I have high expectations :), which is much better than what I've seen in some of the competing gear.
ePMP has much more resistance to interference then UBNT. ePMP has link adaptation feature that set proper modulation in accordance to measured link packet error rate (PER), that depends on CINR= Signal/(Interference+ termal Noise. It means that ePMP always in any noisy environment set such modulation ( data rate) , that provides low required PER without packet losses. For this reason, for example, we very rarely ( almost never) may see ping packet losses in ePMP link.
UBNT airmax also can somehow adapt it's modulation to interference level, but makes it unproperly, because it does not measure PER and adaptation criteria is unknown. May be it is CCQ or something else. And we can often to see that ubnt working data rate ( modulation) does not match to required for this modulation current CINR, that causes ubnt link packet losses and throughput, latency and jitter degradation.
VoIP is very sensitive to latency and jitter. UBNT is not able to provide low ( required ) link PER , and latency, jitter in noisy environment very often may be unacceptable for VoIP service.
ePMP due to it's link adaptation ALWAYS provides in noisy environment and also in NearLOS ( impact of multipath signal interference ) link parameters, that are required for data/voice/video/game service .
Just to add my 2 cents here:
According to cambium, the modulation adapt algorithm is "Hitless".
What this means is that when the radios change MCS due to noise/signal fade there is no data loss.
In UBNT's Airmax, sometimes we had to lock radios at a given MCS because the constant flapping caused a unacceptable packet loss rate.
Cambium's algorithm properly deals with these changes, making co-existence in a noisy environment less of a problem.
I had epmp + sectors outperform "other" gear + dish antennas several times. (The sectors being at the towers, co-located with competitor's equipment).
My point is, Epmp's performance in noisy environments is so greater than the stuff we used to work with, it's just game-changing.
since we start deploying Epmp.... our problems are in other places...not in an tower with epmp deploid.
our noise floor is about -90 dbmn.... but the area is plenty of competitors.... ubnt...mkt...even pmp450...
my ubnt gear...was restarting every day 2 times... we place an epmp.... instead of 40mhz channel bw ... we use 20 ... results are amaizing ...since we connect client there...we forget about him , we are moving all of them to epmp ... ubnt for us ...is dead... many performance problems... lan port start to drop packages in powerbridges m10 and rocket ac...quite problematic in "combat areas" (interfertence places) ...and range too short...3 years a go we placed a client at 6km away from to tower ...LOS...and was reaching AMC 98% AMQ 95% ....today is amc 6% amq 1% ...tomorrow we place it to epmp with force 110 .
hope it helps