• Wireless newbie here…

    I have a router (R1) with (4) vertical external antennas and an access point (AP2) with 4 vertical external antennas placed at ceiling height on the same wall to extend coverage to a couple of adjacent rooms.  Literally all the antennas are lined up 10' apart at the same height on the wall such that it's

    IIII - - - - - - - - - - IIII

    When I purchased the router + AP combo the vendor swore up and down the appliances wouldn't interfere with one another so long as one was operated on channel 36 and the other channel 165.

    When appliance (R1 or AP2) transmits traffic with a single client (C1 or C2) it does so at 55MB/s.  Sadly when both (R1 or AP2) pass/transmit WLAN traffic for 2 clients at the same time the achieved bandwidth drops to 5-10MB/s.  To be clear C1 is connected to R1 and C2 is connected to AP2.  The clients are just doing an iperf3 file transfer among themselves.

    As soon as I placed R1 on the floor and out of line of site of AP2 (some boxes of stuff attenuating) the achieved transfer is 55MB/s like one would hope.

    I don't have any reason to believe the vendor had a batch of units with bad band-pass filters causing spurious emissions outside of the spectral mask.  When I check the FCC docs for the appliances, the FCC shows pass for "frequency stability" and "spurious emissions" at a 3M or 9' test distance which is akin to my setup.

    Because of attenuation to necessary coverage areas I really do need to place these appliances on the same wall, near the ceiling no more than 10-14' apart.

    Is there a wireless principle I'm unaware of that explains why R1 and AP2 are diminishing each other so greatly when transmitting at the same time even though the channels are "supposed" to be completely isolated???

    Is placing a small strip of RF shielding fabric obstructing the line of site between the router and access point a feasible solution???

    I think a strip of RF shielding fabric between them may look tacky but if that is the only solution I'll gladly do it.  If that is a feasible solution does anyone here recommend a particular RF shielding fabric to buy???

  • Placing APs 10' or 3m apart is close, but should be sufficient. Of course 14' would be better.

    Channels 36 and 165 are at the opposite ends of 5GHz so there shouldn't be any interference problem either.

    Of course you could have a faulty device, it happens. To test your hypothesis you would need a spectrum analyzer. It shows the radio energy on each frequency. It doesn't have to be a precision device, those channels are so far apart. Even some access points have a spectrum analyzing mode. (If you need to buy one, Ubiquiti Bullet with its airView is one of the less expensive options, but you also need to buy an antenna with an N connector. Any cheap indoor rubber duck type will do for this use.) Just test C1/R1 and C2/AP2 separately and you should see clear spikes on their respective frequencies.

    Just to be clear: When C1 is connected to R1 or C2 is connected to AP2 the iPerf shows 55MBps. Where is the iPerf server at? You write that when C1 and C2 are both connected and running iPerf between them you get 5-10MBps. How about using the same server as in the first case? Do they still report the same speeds? I am just thinking there might be a bottleneck somewhere else, i.e. you are measuring a wrong thing.

    However, you also write that isolating the devices increases the throughput. Do both C1 and C2 report 55MBps then? Does the throughput drop immediately if you remove the isolating boxes during the test?

    If you need to isolate the devices after all, any sheet of metal does the job. Even better if you can ground it. The metal will cause reflections, though, but n and ac can cope with that. Better yet if you can avoid placing it perpendicularly and slant it at an angle instead.

  • I have been very curious what a spectrum analyzer would reveal but was hoping to escape that cost since I don't think I'd get much use out of it other then my curiosity with this specific phenomenon.  Funny enough I have a Ubiquity bullet unused on the shelf but only a 2.4GHz version so I don't expect that can help??

    Running iperf tests in numerous scenarios revealed that whether I run the iperf server on the router or AP themselves or on the clients that 55MB/s is achieved so long as both R1 and AP2 are not simultaneously transmitting.

    E.G.     C2 -> AP2 WLAN -> an iperf server (on AP2 or R1 or C1 on the LAN) all achieve 55MB/s

                C1 -> R1 WLAN -> an iperf server (on AP2 or R1 or C2 on the LAN) all achieve 55MB/s

    Both C1 and C2 can connect to a singular WLAN (R1 or AP2) and iperf tests in either direction achieve 25MB/s solid.

    It's only when I have one client connected to one AP (doesn't matter if it's R1 or AP2) and the other client connected to the alternate AP (hence both APs are transmitting wireless simultaneously) do I observe the 5-10MB/s.  E.G.  I run iperf tests between C1 -> R1 WLAN -> R1 LAN -> AP2 LAN -> AP2 WLAN -> C2 (ergo both R1 and AP2 are transmitting simultaneously) is when only 5-10MB/s is achieved.

    Literally as I physically pull R1 down from where it's placed in line of sight of AP2 then the iperf tests jump up to 55MB/s.  This happens in either direction (running iperf with the  --reverse switch).  Equally, as I physically place R1 back up on the wall near the ceiling in clear line of sight of AP2, then the iperf results drop back down to 5-10MB/s.

    I appreciate your suggestion about an angled sheet of metal.  I thought about that but I have neighbors above and below me using 5GHz wireless as well so I thought it would be more courteous if I just used something that would absorb the signal instead of deflecting it.

    I'll go ahead and run another test with an angled metal baking sheet between R1 and AP2 to see if I'm able to keep them elevated on the wall like I need them to be and still achieve 55MB/s.  If so, RF shielding fabric should work.

    Given my having to place RF shielding between a CH 36 AP and a CH 161 AP reasonable, or should I be pursuing the vendor (Linksys) to replace the units on the prospects that replacement units may not exhibit this behavior??  (which I'm only guessing is spurious emissions)

  • I ran a test with R1 & AP2 near the ceiling on the wall and inserting a baking sheet between them.  Result  55MB/s.  Pull the baking sheet away and transmission regresses to 5-10MB/s.

    Given my having to place RF shielding between a CH 36 AP and a CH 161 AP reasonable??

    Should I be pursuing the vendor (Linksys) to replace the units on the prospects that replacement units may not exhibit this behavior??  (which I'm only guessing is spurious emissions)

  • Your tests appear to be solid and conclusive - good.

    You would need a Bullet 5M on this case (or some other 5GHz capable spectrum analyzer). The Bullet costs about an hour's worth of work, in this case it could save several hours. I would use a spectrum analyzer to gather evidence at least before blaming the vendors.

    I am thinking of some exotic setting on either device. Some mesh APs use a secondary channel for backhaul. Are you sure there is nothing of the sort using the same channels?

    One more thing to try out: Turn down the transmission power. I always start with this, so it slipped my mind to check for it. If you are blasting at full power, the devices might be interfering with each other at such close range. In general you should only use the same power level as the least powerful client. Cell phones typically use only 15mW or 12dBm max. You should only use full power for long PtP links when you really need the range and can match the power level at the other end.

    No, you shouldn't need to use shielding between the APs. A baking sheet bolted on the wall would make a great conversation piece, though.

  • It may just be that the radios are interfering with each other due to their high levels of output power - essentially swamping the others circuits.

    What power levels are you running at ?

    Are you using some variation of RRM?   If so, try turning it off.    Devin has a good CWNP-TV video on this subject.

    I agree that that spacing should cause no problems. 

    Is there any chance that there is interference on the Ethernet cables due to their (original installation) proximity to each other?    Was the cable installation tested properly,   I have seen some very strange Ethernet problems over the years caused by cable routing.

    Was NEXT testing etc. performed?

    Good luck, please let us know your findings.

  • @Petri

    Well said about the affordability of bullet's.  When I went to buy one however I see the US version only supports channels 149-165.  Are you aware of what country version supports the lower channels?  and where to buy it?

    Like you I totally wondered if unseen to me these were using secondary channels like DFS after a large transmission is initiated.  Hence if both were using the same DFS channels at the same time, this kind of behavior is exactly what I'd expect.  Though I double checked the channels of the broadcasts with free apps like InSSIDer and NetSpot I wondered if they might not accurately present extension channels that are being used.

    Agreed about transmit power and clients not being able to sound back.

    I'd tested the units at lock step transmit power of 0, 4, 7, 10, 13, 16, 19 dBm. Unfortunately the behavior remained unchanged. Only breaking the line of site between the units (metal baking sheet etc) achieves function of 55MB/s. As soon as I pull the metal baking sheet away restoring their line of site, transfers regress to 5-10MB/s.

  • @Howard

    I'd tested the units at lock step transmit power of 0, 4, 7, 10, 13, 16, 19 dBm. Unfortunately the behavior remained unchanged. Only breaking the line of site between the units (metal baking sheet etc) achieves function of 55MB/s. As soon as I pull the metal baking sheet away restoring their line of site, transfers regress to 5-10MB/s.

    I'm not aware of these Linksys appliances supporting RRM even if DDWRT, OpenWRT, LEDE were to be flashed onto them.  If someone knows that to be the case, please let me know.

    I hadn't done special ethernet cable testing other then iperf3 testing from R1 LAN to AP2 LAN which yielded 112MB/s.  I understand USB 3 appliances can diminish wifi 2.4G.  It hadn't occurred to me that Ethernet could be an interference carrier.  I'm guessing that's just when POE is used??  Fortunately in this case I'm not using POE.

    Showing my inexperience...I don't know what NEXT testing is.  I googled it and didn't come up with any clear results that speak to a wireless testing process.

  • By Petri - edited: October 2, 2017

    Hmm.. Ubiquiti spec sheet just says "Only 5725 - 5850 MHz is supported in the USA". I don't know. Perhaps you can just change the country setting to Uruguay/Paraguay that are very permissive. You would only be receiving, so you are not violating any regulations.

    I have an RF Explorer so I am not experienced with Bullets. If I had seen one in action early enough I might have chosen otherwise. It is some hassle with PoE and a crossover cable to a laptop, but you can get a decent view for less than a hundred bucks. They are much better at what they were intended for, though.

    DFS would just cause the radios to either go silent or switch to another channel abandoning the original channel. Not the behaviour you have described.

    Extension channels are used for wide channels and they are consecutive to the nominal channel. You cannot bond 36 and 165 in any way. Eight is the maximum number of bonded channels.

    I guess spectrum analysis is the only way to see what is going on. If you pursue it, please post your findings.

    The other option is to get used to the baking sheet.

    Oh.. NEXT stands for near-end crosstalk and is an Ethernet cable phenomenon:

  • The Ethernet interference I have seen, and heard of, relate to noise picked up on the cables themselves.     Poor shielding, improper twist-rates among pairs, running Ethernet cables over suspended ceiling light fixtures, bundling cheap cables next to each other , or power cables etc.

    Gigabit cable installations (especially), are supposed to be fully tested by installers.     Even the last inch of cable at the connector/punch-down-block can be a problem today.    NEXT (Near End Cross Talk) is just one of the many tests performed by the several thousand dollar test instruments that professional installers use.   

    Things like my $20 cable tester (continuity only)  just don't hack it any longer.     Which is one reason I don't personally crimp connectors anymore..

    There was a report, just a couple years ago (from Tessco ?), that talked about a large amount of counterfeit Ethernet cables being shipped around the world.

    I have also seen air-conditioning duct work causing odd RF situations.     AC ducts also channel RF signals - but that doesn't sound like your problem at all.

    If I didn't know better, I'd say you were running your AP's on the same channel.

    Just for the h--- of it, do you have a DAS network running too ?

Page 1 of 3