we’ll start with our first 11n segments soon. I report some observations and may be someone
can jump in and give some hints. As I want to do ptmp, exes configuration does not help for
this case.
Setup will be: StellaDoradus Dual Pol Sectors, RB411AH, R52N, ROS 4.9
Main reason using these antennas: they have a housing with metal backplane
included so I hope to reduce interference and avoid long cable runs.
As we are in ETSI we’ve to avoid db cable loss in receive path.
I tested in lab and got some bad results using nstreme. After running some tests I ran
into situations where client do no longer connect to the ap. Disabling nstreme allows
connecting again. I got the “polled for to long” logs. hw-retries are 8.
So I start without nstreme using RTS/CTS instead.
Using both chains for sending I get very unstable results. Best results I got is using
both chains for receiving and only vertical polarisation for sending.
I got nearly 100MBit in one direction UDP and 60 FDX in lab with some thin walls
between.
Hi, we have one site running 802.11n. We use it for a village (3-4km distance), base station is on a hill.
We use antenna from UBNT (AirMAX MIMO, 16 dBi, 120°), RB600A, R52N card, both chains for RX and TX. NStreme off, HW retries 15, EIRP 27dBm. On client side we use UBNT NanoBridge M5 (AirMax MIMO) and NanoStation M5 (AirMax MIMO).
With NanoStation M5 it is working but results are very awfull. TCP gets only 8Mbit/s, with very increased ping up to 60ms afrer 5Mbit/s. Without traffic it’s avg 4ms.
With NanoBridge M5 we can get 20Mbit/20Mbit (FD), TCP, with avg ping 20ms. 30Mbit/30Mbit (FD),TCP, can be achieved but ping raises dramatically (400ms). It’s still not that much, but on that distance PTMP it’s more than with 802.11a. The only thing that is bad is the latency.
I wish there was a cheap antenna for MIMO which I could use with RB411 and R52N/R52Hn for clients. Maybe the results would be better then.
I’ve UBNT for testing (rocket+120degree + NSM5). I dont want to use them now as they are not
implementing DFS until now (I’m not sure they will in future) and there are still some tx-power
anomalies with lower/allowed power settings.
They are “all-plastic” so I expect interference problems in medium/high density areas.
Our clients are Mars 23 dbi with metal housing on the back. Only vertical polarisation at the moment.
Older models RB411/R52 newer with RB411/R52n. We’ve Mars 23dbi/dualpol too but they are to big
40x40cm for a standard CPE and too expensive.
Have you tried to decrease hw-retries and sending on only one chain?
I’m not sure the (MI)MO benefit is big enough to equalize the -3db each chain is reduced in ETSI.
And there seem to be some problems with mimo as I get strange effects using it. Dont know wether
this are signal problems or bugs in ROS.
Got Problems using rts/cts. One cpe kills the whole Segment.
The Client is Limited to 3M/384k but when he is torrenting the other
Subscribers get Bad latency. Disabling rts/cts on this Client
all is Fine. He gets Full bandwith and the other subscribers are happy
again.
So something is wrong with rts/cts implementation.