We have a setup with two KP duel pole 4port omni’s horizontally separated about 45". E3000 on two of the ports and LTE on the other two ports of each omni. The idea is we get some spatial diversity for better NLOS for the LTE system while maintaining the MU-MIMO capabilities of the E3000.
For the E3000, when we use the Omni / Split Sector setting we get 0 Grouping. When we set to Sector we get roughly a 50 / 50 split on grouping. Has anyone thoroughly tested this type of setup?
If set to Omni are we in effect getting a 4x TX / RX so we would have better gain vs a two separate 2x TX / RX for more potential throughput? I am inclined to leave it on Sector as the base assumption is that it is a 4x TX / RX except when loaded and then splits the TX for more throughput. Using the wireless link test we do not see much if any MU-throughput on ether setting.
Any thoughts or others testing of a similar setup are appreciated.
Anthony,
As an RF engineer with lots of MIMO experience, I’m thinking that there are a couple of things that I would do differently on your setup.
I connect all 4 ports of the ePMP 3000 to all 4 ports of a single antenna, and connect all 4 ports of the LTE base station to the other antenna.
Instead of placing the antennas on the same horizontal plane (if I read your preamble correctly), since omni’s have the most isolation out of the top or the bottom of the antenna, I’d stack the antennas in the same vertical plane, separated by 3 or 4 ft (as much as you can get).
There is no timing common between the ePMP and the LTE, and using ports on the same antenna is likely throwing the sounding frames off on the ePMP - the frequency isolation between the LTE and ePMP is why the system is working as well as it is. Beamforming works best when the antennas are as close as possible, while MIMO B (or traditional 2x2 MIMO works best with the antennas spatially separated. In your case, I think it’s most important to have BF and MU-MIMO working correctly on the ePMP to enhance it’s performance as much as possible.
Is this the KPP Omni built for the 3000 ? I don’t know what the antenna pattern looks like for the other 4 port omni’s but the 3000 Omni is basicly doing ABAB (only with the same frequency normally so I guess actually AAAA).
The point is that when you connect it as intended Ports 0 and 1 are basicly looking (for example) North and South and ports 2 and 3 are looking East and West the patterns do not overlap. Given that others are using 2 x horns aimed the same direction (completely overlapping) with 3000 and getting grouping/MU I’m guessing that when you set a 3000 to Omni it expects 2 sectors/patterns not overlapping and when you set it to sector it expects the patterns to be completely overlapped.
Again, everything assuming 3000 Omni I don’t know anything about the patterns of KPP’s other 4 port Omnis, since the 3000 Omni is bascially doing ABAB patterwise if you connect 1 3000 to two ports on 2 Omnis whether or not the patterns overlap or do ABAB will depend as much on how the Omni’s are rotatated / orinented as it does what two ports it’s connect to. So the reason for your grouping might suggest you have the omni’s connected/oriented in a such a way that the pattern from the two omni’s is overlapping instead of N/S and E/W ? It’s a question because I’m just guessing here, I haven’t done any testing on how the Omni / Sector setting in the 3000 actually works / affects things.
If you aren’t using the 3000 omni’s then none of this may be relevent though I think the other Omni’s are likely set up simular.
Brubble1:
We are not using the 4port “cambium omni” I believe your on to something though on the expectation of the “Omni / Split” config and the expectation of no overlap. By the fact if we use that setting we do not see any grouping. But with “sector” setting and the expectation of the overlap we are seeing grouping. If you assume a perfect Iso Omni and the overlap of the propagations the radio would see this the same as it would a sector antenna.
Additional testing has shown that we are seeing about at 15% increase in capacity with the utilization of the sector setting and MU-Mimo. That obvious is not great but it is something.
DaveClelland:
I considered doing two sperate 4 port single frequency antennas instead of two duel frequency antenna. I went the way I did because I believe the spatial diversity would likely provide better NLOS performance on the 3.5ghz LTE system, with the hope that the improved CiNR would be an overall system gain and thus capacity gain. We have been pleasantly surprised at the range and effectiveness of the LTE system in the duel omni config vs our traditional sector deployments.
Stacking the antenna would actually be easier to implement then the current implementation of Horizontally deployed antenna. Except that we would have to mount it in such a way as to avoid shadowing from the mount / tower leg. I also wonder if we would see similar NLOS performance. Trees tend to be vertical structures.
I have several more sites we are going to deploy with an Omni setup. I am going to try vertical though I am concerned about getting the Omni away for the mount to avoid / minimize shadowing.
We are unable to take advantage of beam forming on the upload because it is an omni so the BF antenna is not deployed in this setup. I believe the MU-Mimo implementation in wifi / e3000 is by frequency spatial shifting and not beam forming vs the PMP platform is truly a beam forming implementation that also likely incorporates frequency spatial shifting for further isolation. I believe that our LTE platform also utilizes frequency spatial shifting for it’s MU-Mimo but we utilize Carrier aggregation or all 4 ports on TX and RX in our implementation instead of MU-Mimo.
Thanks again for both of your incites. I am hoping someone else out there or a Cambium rep might have some additional idea’s on how best to implement this.