You are here

Node Performance

13 posts / 0 new
Last post
Node Performance

Good Morning:

I had an interesting situation as our Inland Empire mesh builds out. Our current hub if you will is a Rocket M2 with a 120 degree sector located atop a water tank at the highest point in town. The majority of nodes looking back at it are NanoStation M2 that are within 2-5 miles of the tank location. One user, approximately 2.5 miles out has a RocketDish M2 pointed back. Until just recently, all were seeing good performance despite a couple spots having less than optimal line of site to the tank.

Last weekend a new node came live, again, a NanoStation M2, located about 7 miles out. We had solid LQ at 100% but junk NLQ dancing anywhere from 0 to 50% completely unusable. Another Nanostation in place confirmed the numbers so we went back to cabling, etc. I adjusted the sector towards the site as I knew it was on the fringe with marginal improvement. Finally the member at this site dropped some cash and put up a RocketDish M2 feeling the distance may have been on the fringe. This made some improvement but only marginal considering the significant increase in gain that the dish provides. This link actually has the best line of site of any existing node so should be the best overall connection, at least in theory. The system is running channel -2 at 10mhz. Any ideas that we might dig into this weekend? An interesting note too is that other links now appear to be suffering a bit as well. I do have a second 120 sector prepared for deployment at the tank providing 240 degree coverage. This should happen within the next two weeks.



N8NQH's picture
may not matter, but...

we were recently told by a WiFi networking guru that Ubiquiti equipment is sensitive to the distance setting; more so than other mfgrs.  The Rockets distance setting should be set  to the node furthest away; no more and no less that this most distant node. And this most distant node set accordingly.

AE6XE's picture
This is physics with the

This is physics with the speed of RF and the 802.11 handshake timing sort of thing, governed by software/firmware.   Maybe some other vendor's firmware is dynamically adjusting the distance settings and AirOS is not?    In AREDN it's a specific static setting.

Distance paramater is a good

Distance paramater is a good one to make sure its set correctly.

Next item is to check if your overdriving central node with all the signals.  A high LQ means the link is good on its own,  the low NLQ means the "Central" node is having trouble hearing you, this could be because all the NanoStations at 2-3 miles are drowning out the farther node,or more likely the Rocket Dish at 2.5 miles away is beaming full RF power into the node the node may be backing down its receiver pre-amp to maintain quality of the signal.  Make sure this node isn't thrashing the frontend and have it back down its power if needed.

As for "Ubiquiti" being more sensitive to the Distance parameter, the code we use is used across devices, don't expect this to be different for one device over another, the distance parameter does need to be set correctly (its better long vs short, too long and it slows down a bit, too short and you thrash the RF, go for long over short always)

How Strong is Too Strong?

The implication here is that a signal could be too strong.
How strong should we liit the signals? Using Charts, we could adjust as needed.



Distance Parameter

Looking from the remote node pointing back to the central I was seeing a good LQ and diminished NLQ. We have turned the power back on everything expect the central node a this time. Additionally, the distance settings have been adjusted accordingly to best represent the distance to the central and the central has been adjusted to be truer to the most distant. All are set erroring on the side of greater distance as noted above. Honestly not much of a change was seen. 

I did change channels to-1 on all but the most near-field rocket dish and saw some gains but do still notice fluctuations from day to night.

kj6dzb's picture
you need to separate the

you need to separate the channels. You cant have 2 radios on the same ch at the same site, like that. Try placing the the dish on -1 and set the distance properly. technical you should have 20mhz spacing between the channels, that difficult to get on -2 , -1 , 0. you could go to ch 0. There was just a post about how it was done on 5ghz were we have more spectrum. 
you will need to DTD link the nodes, you will not need a Vlan switch, just plug them into each other.

73 Mathison
AE6XE's picture
The "O" in "OFDM" modulation

[Edited to correct.]

The "O" in "OFDM" modulation was designed so that we can have adjacent channels with really good mitigated interference.   Now, this will certainly break down if you have 2 unshielded transmitters 5" apart at the site in near field.     However, the San Diego folks just tested on Sunday to show that there was no detectable impact on ch -2 @ 5 Mhz when running a 2nd node on ch -1 @ 5 Mhz on the same tower--it can be done--Rockets with 120 degree unshielded Sectors.

Put this into context that a channel has 64 individual modulated carriers.  There's already built in null guard carriers between channels.  The design of the symbol timing and spacing of the carriers in the channel has a nulled-out interference pattern between carriers (this is what is expected of  'orthogonal' signals).  Cutting the bandwidth in half is still using 64 carriers, but the symbol length is doubled and spacing cut in half to archive Orthogonality.  With 2 adjacent signals from different transmitters are received, the power and phases will not be identical between signals to be orthogonal, but the additional digital pass filters and null carriers are a built in gap between channels.   This is why OFDM is taking over as the dominate modulate for digital communications.  It is the best know efficient use of bandwidth.

The more RF shielding, directionality of the antenna, the closer you can get the nodes on the tower and not have impact with adjacent channels.  It's more a question of how close is too close vs. one should never do.


K6AH's picture
Yep... it surprised me

In know it's hard to believe.  I'm the one who has been championing guard channels to separate collocated devices.  Conrad, KG6JEI, and I conducted the test Joe describes, I wouldn't have believed it, if I hadn't seen it for myself.  It's also contrary to advice I've been given by respected WISP technicians and on the UBNT forum.

This is good news though.  It means we have more channels available to us at a single site if we remember to keep adjacent channel devices configured to the same bandwidth.

Now I'll need to go back and edit my earlier posts to the contrary.

Andre, K6AH

AE6XE's picture
Andre, thinking about this

Andre, thinking about this some more, I have it wrong on the orthogonality between channels :) .  The built in carrier/bin guard gaps and pass digital filtering for a channel must be enabling adjacent channels by themselves.   I'm forgetting about the timing of the carriers between channels of 2 different transmitter signals--they are not time synced in phase and thus not orthogonal.  In short, It shouldn't make much difference if it is same bandwidth or different bandwidth adjacent signal, but this stuff is complex enough, should test to confirm.   I'm dusting off some cob-webs on impulse responses, convolution, and digital signals... 

Andre and Joe

This is great info for my own personal learning and greater depth of understanding. Thank you for putting this dialog up as it will be most helpful in moving forward with the mesh build out. Based on Joe's breakfast comments regarding the 5ghz equipment and greater number of open channels I can see real benefit to the use of 5ghz for the last mile stations where we want to get 360 degree coverage from a single tower. I had originally been hoping to accomplish a 360 degree coverage wtith 2 ghz however given the wifi scanning I this past weekend I truly am limited to the -1 and -2 frequencies in the area of my main tower. Looking at it now, I will likely cover the front 120 with the existing sector and then use 5 ghz equipment for the remaining two angles. That is, unless your experiences in other tight areas supports another alternative that I am missing. At the end of the day, I want the best 360 degree coverage of the valleys and canyon below for emergency communications and services as this is the best real estate available.

What type of throughput was achieved with the -1 and -2 operating at 5mhz? would it be acceptable for the last 1-5 mile links?

separate the channels


I did split channels at the repeater site where I have a link node pointing down at my repeater radio building. Ultimately this will become a DTD but have to run some conduit and do some trenching first. I have -2 on the sector at 10mhz with a dtd to a nanostattion running low power pointed at my repeater 100' below. I did notice some improvement after doing this though not what I would have hoped for.

Great feedback so far

thanks for the comments thus far. I will definitely start looking at the distance settings and see about getting the gain turned down on the close in RocketDish. At this time I do not have dual nodes at a site on the same channel. I did run into this problem in my Linksys days so will be using the dtd and rf shields when the second sector goes up as well as separating channels assignments.

i will provide an update after tweaking other settings as suggested. Keep the thoughts coming and many thanks!


Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer