Help me understand how MU-MIMO will work with RF Elements 4x4 horns. From what I can tell they are just two horns of the same beamwidth covering the exact same areas. I thought for MU-MIMO to work best you would want the antennas to cover different areas so it can make groupings. That way it can use one set of 2x2 to talk to one SM while the other 2x2 set talks to another SM, and the different antenna coverages/beamforming help make that possible. So wouldn’t a 4x4 horn setup that is two 45deg horns offset to cover the full 90deg be better than two 90deg horns covering the same area?
We did some testing using an e3k 4x4 AP and the Cambium OEM 4x4 antenna VS the RF Elements 90deg 4x4 horn array and found that the RFE horn array had slightly better performance (like 1 point on avg better) then the OEM antenna BUT for reason we don’t fully understand, it was very hard to get MU-MIMO to trigger successfully.
The OEM antenna seemed to reliably produce MU-MIMO groupings much more easily and frequently.
That would make sense to me as the RF Elements 4x4 are two antennas that cover the same 90deg. This would virtually eliminate the ability to get the spatial diversity you need to create good beamforming and MU-MIMO groups. Not sure why they did not package them as two 45deg horns offset to create a 90 deg coverage area. That to me would seem to work much better as you now have two diverse coverage areas which would allow for relatively easy MU-MIMO grouping.
Maybe I am wrong as I am not an RF engineer. That is why I asked how MU-MIMO worked with the 4x4 horns.
Hi, Any update on this one?. I would’ve thought exactly the same thing than you.
I have not seen any other posts from other operators that have tested MU-MIMO sector vs dual horn performance. Based on my own testing, I don’t feel as though MU-MIMO is optimal on these dual horn configurations. So far I’ve had the most reliable results from panel antennas.
We put up the RF Elements 4x4 at a test site and confirmed great difficulty in getting any MU-MIMO to happen during tests. With the regular OEM sector, no problem.
Unfortunately haven’t had time to chase this issue further with RF Elements / Cambium.
I mentioned my findings to Tasos @ RFE and he said he’d look into it and I never heard back from him.
Just curious, what method do you use for testing MU-MIMO? I have two 4600 AP’s with 5.9 and 5.9.1-RC4 and currently I don’t see any MU-MIMO throughput improvements using the dual sm speed test or just manual speed tests. Definitely seeing the beam forming working though for the higher modulations. These are both using 4x4 panel antennas, one Cambium and one Alpha Wireless.
Reason I bring it up here is that I have two 4500c on order, one with dual 60 degree RF Element 4x4 antenna arrays and the other with 90 degree. Currently looking at doing the split sector deployment on a tower with them for 120 degree coverage, but would be nice to see if there is a way to test MU-MIMO. I am thinking of going with the 90 degree horns vs the 60 degree horns and then overlapping the coverage in the middle where most of our members are. Gets me the 120 degree coverage I am after, but if MU-MIMO works, no performance hit vs doing split sectors. (Well, besides antenna gain)
MU-MIMO isn’t operational on 4500 or 4600 yet.
1 Like
We used an e3k 4x4 AP and ran dual SM speed tests from the AP.
1 Like
@Jacob_Turner @Eric_Ozrelic Thanks! I keep getting Beam forming and MU-MIMO mixed up evidently.
My understanding of MU-MIMO is that to be able to have the AP talk to more than one SM at a time you need SMs that cannot hear the TX meant for the other SM. So when you have two 90deg horns covering the same azimuth you will not get any MU-MIMO as when antenna A is transmitting all SMs in the 90deg azimuth can hear the AP equally across both antennas.
In my opinion for MU-MIMO to work on the 8x8 4500 AP you could do a setup with four 45deg horns. Horn A would cover 0-45deg, Horn B 45-90deg, Horn C 90-135, and Horn D, 135-180deg.
What that allows is for the AP to communnicate to an SM on Horn A and the other SMs on Horn C or D cannot hear that transmission or hear it at such a low level it does not matter. In this setup the AP could communicate at the same time to SMs on A and C, or B and D, or A and D. This keeps at least a 45deg null between antennas by the AP ‘turning off’ 2 horns.
This setup is still not ideal as you can only talk at most to two SMs at a time to elminate self interference. That is why it seems most, if not all, MU-MIMO systems use panel antennas. On a proper MU-MIMO panel you can have the AP change the antenna patterns as needed (within design limits) to get much smaller azimuths to allow communication to more SMs at the same time with ‘nulls’ in between.
My guess is the 90deg 4500 panel AP can create up to eight 11.25deg patterns or maybe 4 20deg which would then allow the AP to talk to up to four SMs at the same time while keeping a null between each pattern.
I suspect that if you want to use horns you will have to give up on full performance MU-MIMO (4 SMs at a time) and settle at best for half performance MU-MIMO (2 SMs at a time), but only if you setup enough narrow beam horns.
This has been discussed in the past:
You are correct about needing isolation between data streams for MU-MIMO to work, and you are correct about the 4x4 5 GHz Asymmetrical Horn Antennas not having enough isolation.
If you use four 45° horns to cover a 180° sector with an ePMP 4500c, you will indeed have enough isolation for MU-MIMO to work, but the spectrum analyzer will only work on one sector, and you will pick up noise from all four. At that point, you would be much better off using four 2x2 radios operating on separate channels. This setup would provide even more isolation with channel separation, better modulation rates, and greater capacity. Four separate radios would also provide a spectrum analyzer for each sector, allowing you to optimize each one individually. If you use the ePMP 4500c with four split sectors, the spectrum analyzer will still only work on one sector while picking up undetected noise and interference from all four sectors.
MIMO works by isolating antennas—the more the isolation, the better the performance. My company, IsoHorns, leads the industry in port isolation (>30 dB), which is why we achieve the highest modulation rates with our 2x2 antennas.
We also offer a 4x4 Double Horn, but we don’t claim it can do MU-MIMO because we haven’t tested it yet. If you’d like to test our 4x4 antenna, you can use coupon code “casestudy” to receive 50% off and free shipping worldwide. We just ask that you compare it with your existing antennas, share the results, and let our products speak for themselves.
@terintamel I invite you to read the ePMP 4500C & High Performance Horn Antennas brochure. Notice that Cambium and RF Elements ascribe MU-MIMO capabilities to the radios but not to the antennas.
I apologize for not following up on this topic earlier, it was somehow “lost in translation” on our end. Let me address some of the observations in this discussion.
In 4x4 dual horn setup, both horns have the same area of coverage, and are mounted side by side. If you look under the hood of e.g. 4x4 patch array antenna, there are usually two “regular” 2x2 patch antennas placed side by side. According to Cambium recommendation, the two horn antennas must be separated in azimuth plane (= mounted side by side) in order o support 4x4 MuMIMO in ePMP3k.
MuMIMO is to my understanding one of the modes (or features) that ePMP radio operates. It is a specific function of ePMP radio, not the specific function of an antenna. So we have to rely on the level of knowledge about ePMP we have as outsiders. I think, that one of the trade-offs the radio has to decide on, is either a) use single user transmit beamforming or b) split it into multi user MIMO. The trade-off is that with MuMIMO you are serving more users but also splitting the power and I would guess you lose few dB / modulation step in comparison to single user beamforming.
This is just an assumption, but because the horn antenna provides strong “old school” SNR benefit, it is probable that ePMP radio opts to go into single user beamforming more often than with patch array. This might be a reason why you see a better throughput result but much less MuMIMO with 4x4 horns compared to patch array. It does not necessarily mean that MuMIMO is impossible with 4x4 horn antenna, it means that it is the radio that decides what is the best way to achieve the goal.
Let me add for those of you not experienced with using horn antennas: Narrow beam 2x2 horns deliver excellent performance as 2x2 sectors in Split sector setup covering different areas. 2x2 narrow beam horns are not recommended for creating 4x4 or 8x8 single sector setup. These require wider beam widths - we offer 4x4 Horn antennas with 60° and 90° azimuth beam widths.
1 Like
Hello everyone,
I also apologize for not jumping in to this discussion earlier.
I want to give Eric’s explanation an upvote - the MU-MIMO groups are formed in the radio. It’s a function of beamforming, combined with the ability of the radio to schedule resource elements to various OFDM carriers at specific moments of time. To form a beam, you need at least 2x2 MIMO streams. Normally, the phase relationship between the resource elements in each stream 0 degrees. ASIC’s today are so fast that the scheduler in the radio can change the phase slighlty between each stream. This has the effect of increasing the signal strength in one direction, and creating nulls in other directions.
Did anyone take high school physics? Remember your study of nodal waves? Two point sources creating waves in a tank? When the wave source is in phase, the nodal waves are evenly distributed across the tank, radiating out from the source. Could this be just like a multiple section antenna transmitting?
In the drawing, the blue lines show waves reinforcing each other (gain) and the red line shows waves interfereing with each other (attenuation).
By varying the phase of the 2 point sources, we can ‘move’ the reinforcing areas across the pattern. If we control this phase relationship properly, we can aim the reinforcing “beams” where we want. This is phased array beamforming simplified!
Now, imagine that we have a large tank. We have 4 x 2 point sources (8x8 antenna equivalent). The point sources are different by frequency (not every beam uses all the OFDM carriers) and vary in time (a data symbol in modern radios varies between 70 and 110 microseconds), and then we have our TDD split…the frame is 5 ms long, so maybe we have 3 ms in the DL, and 1 ms in the uplink, and 1 ms of guard time between DL and UL. In 3 ms of DL, at 100 microseconds per symbol, I can have 30 symbols transmitted in my imaginary radio system here. If I have 720 OFDM carriers with data, I can have about a thousand different resource elements in the DL portion of this frame. 720 / 2 streams x 30 symbols / frame) My switching is not fast enough to have 360 beams, but maybe I can have 5 or 8 beams. If the base station detects SM’s with similar RF phase and amplitude characteristics, it can form them into a MU-MIMO group. If traffic is light, the MU-MIMO group may not be warranted, and the base station will continue to talk individually to each SM.
Eric is also correct in that we have not talked about the antenna beamwidth yet at all.
If we connect our base station to narrow beamwidth horn antennas, the phase relationship between SM’s is very small, and the base station WILL choose to not use MU-MIMO. If the antenna beamwidth is wider, say 90 degrees, the phase relationship between SMs can be significant, and the base station WILL choose to use MU-MIMO to group the SM’s. Automatic transmit power control in the SM’s is very important to make each SM’s signal arrive at the base station as close to the same amplitude as possible so that phase is the only variable in the signal as the base station characterizes the signal from a specific SM to determin what to do with it from the base station’s scheduler point of view.
I hope this helps understand a bit more about beamforming, MU-MIMO, and antenna characteristics.
1 Like