Rohn25g tower, 3x 2.4 AP with RF Armor. Self Interference?

[TheHox]

tower.png (463.01 KiB) Viewed 30175 times

I have a 90ft rohn 25g tower. I started of small with just a 2.4 omni. I started to get many customers so I finally took the omni down, and instead installed 3, 120 deg sectors.

I have RF armor on all of them. I have them each at 10mhz channel width, separate channels.

I have had this up and running for almost 2 months. I've noticed the connections to the customers have not been the greatest, but seemed to work still. Someone else said I might have the sectors too close to each other and to move them vertically. I really would hate to do that because I need them all as high as possible.

So this morning I get a phone call from a customer saying her connection is just "wobbly" she has rebooted devices and etc. No improvement.
I look into her CPE and power levels are strong, but CCQ is high 70s, sometimes up to mid 80s.

The TX and RX rates fluctuate as well. Here are 2 screen shots.

cpe2.png (89.54 KiB) Viewed 30175 times

cpe.png (46.12 KiB) Viewed 30175 times

Is this self interference? Local? Or just other noise?
I do have 3 other towers that also have 2.4 sectors on them. 2 of the towers only have 2 sectors, 1 has 3. I feel I have them aimed and setup where they do not interfere but may be wrong.

[sirhc]

2.4 GHz is a bear, and the spectrum is crowded, but many people (including me) use our cluster mount like this, so no your install is possible.

I do run some 2.4 GHz sectors but I only put up to (3) 2.4 GHz sectors on a tower and I ALWAYS run channels 1,6,11 even though I run 10 MHz wide channels.

I scan to do channel planning in both 10 and 20 MHz wide channels making sure I only see hostile and or my other towers signals in both the 10 and 20 MHz wide channels of -80 dB or less if at all possible in that sectors channel width.

As I said even though you are using 10 MHz wide it is important to stick which channels 1,6,11 on sectors otherwise you double the chance of hostile noise being within your channel plus the channel bleed at 2.4 is horrible.

This is why most people try to use 5 GHz as their PTMP delivery and only use limited 2.4 GHz and then I only try to put less than 10 clients per AP.

If you can not achieve -65 dB or better signal at 2.4 GHz I do not even try.

Plus at 2.4 GHz local interference at your customer location from internal AP's and those of their neighbors becomes an issue which is why I only use NanoBridges or Breams with shield kits on customer radios to cut down on local noise.

I hope this provided some help but 2.4 is a bear! This is why I am so against PTP links like airFIBER 5 and MIMOSA's new 5 GHz ultra wide channels or even if people use the new AC radios with 80 MHz wide channels. Where do WISPs go if the 5 GHz spectrum gets trashed like 2.4?

[Ant]

Enable channel shifting on all your units and do not upgrade above the 5.5.x firmware branch. (Channel shifting works totally different on the 5.6.x branch). That will help some. If you can, when the rates drop low like that, power off all three sectors and then one by one power up and run airview for a couple of minutes. Make sure to match up which airview scan matches with which unit and the direction it is facing. Then shift your channels according to the least amount of interference. It's crucial to power off the other sectors or the results will be inaccurate.

I run a three sector 2.4ghz tower too. Each sector runs shield kits, channel shifted, and 20 mhz channel width. Works perfect. The only difference between your tower, I staggered the units 6ft apart from each other vertically. In the grand scheme, the height sacrificed is minimal to increasing overall throughput to all customers. Lowest sector North facing, 2414mhz, next sector SE facing 2464mhz, and top sector SW facing 2439mhz or something like that.

[TheHox]

Enable channel shifting on all your units and do not upgrade above the 5.5.x firmware branch. (Channel shifting works totally different on the 5.6.x branch). That will help some. If you can, when the rates drop low like that, power off all three sectors and then one by one power up and run airview for a couple of minutes. Make sure to match up which airview scan matches with which unit and the direction it is facing. Then shift your channels according to the least amount of interference. It's crucial to power off the other sectors or the results will be inaccurate.

I run a three sector 2.4ghz tower too. Each sector runs shield kits, channel shifted, and 20 mhz channel width. Works perfect. The only difference between your tower, I staggered the units 6ft apart from each other vertically. In the grand scheme, the height sacrificed is minimal to increasing overall throughput to all customers. Lowest sector North facing, 2414mhz, next sector SE facing 2464mhz, and top sector SW facing 2439mhz or something like that.

Finally got some free time to work on this some more.

Enable channel shift on all devices?
All 3 APs and CPEs as well?

I threw up a 5g sector to get some of the clear LOS users off 2.4, but many still need 2.4.
Thanks

[Ant]

Yeah, enable channel shifting on the cpes first then change the APs otherwise, you'll be doing a truck rolls to the cpes. The big issue running 2.4ghz is taking the time to coordinate the outside unit with the indoor wifi unit so they don't stomp on each other. Ideally, if the outdoor unit is 2414mhz then the indoor needs to be 2462 mhz, or outdoor 2464mhz then indoor 2412mhz. The only exception is if the outdoor unit is running 2439mhz, then the indoor needs to run 2412mhz.

[mhoppes]

@ANT - What do you mean channel shifting works differently in 5.6.x?

[Ant]

In the 5.5.x versions, enabling channel shifting was done in the config file. Doing it this way, the units did not perceive normal channels as noise because every interval frequency was also shifted. In the 5.6.x firmware, channel shifting is not actually being done but the center frequencies and intervals are added to the country code at the eeprom-regdomain list level. The software is taking into account now both the shifted frequencies and normal frequencies. The effect this has is now the units think a low noise unit at say -85 is significant interference. While, the older firmware would keep right on at the higher data rates and achieve much better throughput performance. I did a before and after test on a link and posted the screenshots to UBNT on their forum. It was largely ignored. The test showed 31% improvement on the 5.5.x channel shifting method vs the newer 5.6.x method. A 13.93 Mbps (5.5.6) download speed became 9.51 Mbps on (5.6.x).

[mhoppes]

This is news to me -- but admittedly I've been swamped for the last 6-8 months and am falling a little behind on keeping up on everything going on at UBNT.

So channel shifting is effectively gone and now it's just pick the channel you want? (To me that's a good thing... don't know why they didn't do it years ago).

BUT... you're also saying that with the same noise present the radios performed better when they didn't detect it as noise?!? That's a pretty huge issue... whether it's detected as noise or not... if the data was getting through like it should, there shouldn't be a reason to step down.

[rebelwireless]

I'm not sure what is going on at UBNT, they are having a lot of regressions as of late. I did read on the thread about 5.6.x being less aware of the actual noise, which really kills when doing a 40Mhz channel because it only measures the noise level at the center.

There are a few posts about the PBEAC and RAClite not being able to get good connections at 60-80Mhz channels even though the noise level says all is well. AirView shows noise all over the channel, except at the center frequency.

Then there is Site Survey being 1/2 useless on NBE, PBE radios :/

they're going senile over there!