From: Peter Kranz [mailto:pkr...@unwiredltd.com] Sent: Friday, May 09, 2014 12:45 PM To: af@afmug.com <mailto:af@afmug.com> Subject: PMP450 Beamforming testing - Introduction and lab test results
We recently completed lab testing of the ET Industries 8 beam system in combination with Cambium PMP450 radios and are now moving forward to an 8 radio field test system. I wanted to share results with the list. Problem statement For an upcoming PMP450 deploying covering the San Francisco region, Unwired was facing an environment with very high noise floors, where achieving the SNR ratios required to utilize higher order modulations, and avoiding interfering carriers was seen as an almost insurmountable problem. Traditional 90 degree and 60 degree sector designs were able to see far too many competing systems, and the prospect of using many physical antennas with smaller patterns presented a mechanical challenge. Tested solution Recently the cellular industry has started to deploy multi-beam panel systems to achieve channel re-use in crowded cellular environments (stadiums for instance). These multi-beam systems start with a panel made of many 120 degree antenna elements and use electrical beamforming techniques to create electrically steered sectors with narrow beam patterns (15 degrees x 10 degrees for instance) with higher gain than the individual 120 degree sectors. The ET Industries panel we tested is roughly 2’ x 2’ in size and has 16 inputs, 8 vertically polarized 120 degree inputs and 8 horizontally polarized 120 degree inputs. These 16 inputs are connected to a beam forming network (or BFN) which applies electrical delays to the inbound or outbound signal to create multiple effectively higher gain and narrower coverage sectors. The panel can be configured as (1 120 degree sector, 2 60 degree sectors, 4 30 degree sectors, 8 15 degree sectors) by varying the delay characteristics inside the BFN. Test results Disclaimer: Unwired used off the shelf Cambium equipment for this test, so our results are not ‘calibrated’ to any known good source, we simply used the dBm readouts provided in the software to generate results, so I imagine they may be off by a dB or 2 from calibrated results. The gain of the 15 degree beams is highest for the beams pointing toward the front of the panel, and lowers for beams pointed to the side. If you assume pointing directly forward from the panel is 0 degrees, the first 15 degree beam centers will be at -7.5 and +7.5 degrees in the below results. (These gains include the insertion loss from cables and the beamforming network, so are worst case numbers) +/- 7.5 Beams : +23 dB +/- 21.5 Beams: +21 dB +/- 37.5 Beams: +19 dB +/- 55.5 Beams: +16 dB So essentially 6 of the 8 beams have gains that meet or exceed the gains of the best sectors WISPS have available to them. Caveats To meet power output limit requirements, you may have to dial down your TX power for the higher gain beams. You can use the insertion loss of the BFN in this calculation. So for the +/- 7.5 Beams = +16db -2db +23db = 36db .. I am assuming you have to use the PMP rule here, and not the less restrictive PTP rules (which you might be able to argue for given the directivity of the system, and was successfully argued for by at least one company I know of). The beam former antenna system isolation between beams varies depending on which beams you are comparing, with it being absolutely necessary for the PMP450 platform to be operated in sync mode when using the platform to avoid TX energy from another beam being seen by another beam. The PMP450 radio when run in the DFS bands has a listen before transmit phase. Depending on the output levels pushed into the panel, radios in the listen before transmit mode may see power levels on their input that exceed the design spec for the backend IC of the radio. Cambium engineering has not tested this, so time will tell if this is really an issue. If you don’t run in DFS bands, and don’t run AP in spectrum analysis mode while others are transmitting, this is not an issue at all. Cost - Buying 8 PMP450’s is going to drain your wallet and the ET panel and BFN is not inexpensive itself.
