Community discussions

MikroTik App
 
mikepratt
newbie
Topic Author
Posts: 30
Joined: Wed Jan 11, 2012 7:31 pm

best network design for multilink, UNDERGROUND backhaul...

Mon Apr 22, 2013 6:25 pm

For the real wizards in wireless out there in the wilderness, here's an interesting challenge.

I'm considering an underground (non-gaseous, non-explosive) wireless PtP network, intended to serve as a wireless back-haul for other services. Actual wireless communications is accomplished using a different, dedicated voice radio system (low frequency), whose repeaters are networked via our wireless PtP network, as illustrated by the 'cells' in the included drawing.

Imagine if you will, extremely long and quite regular corridors, in which we'll establish PtP links. At certain corridor junctures, we'll turn at a right angle to extend the network. In some corridors, we'll need to have PtP links back-to-back, so we're able to add right-angle 'spurs' to the network.

Image

I've searched the forum for input on what is the best solutions for mulitlink PtP links, and see only the thread from October 2012 regarding nv2 issues. It does seem that a TDMA structure might be best for an exclusively back-haul link (no multi-access required), but capacity is not the foremost requirement. Network robustness and minimal, consistent delay are the top priorities.

Radios used at PtP nodes will be directional MIMO, with the back-sides shielded to minimize interference from the link radio in opposite/other direction. Directional antennas are chosen to direct signal and enable minimum transmit power - hopefully improving interference mitigation.

Frequency use/re-use will be trialed to minimize interference/co-channel interference, with particular concern on extended links (back-to-back, PtP links) in the same corridor, obviously.

Now wizards, step up, please. :D
First, I have found almost no postings regarding multilink PtP bridge links in general in the forum (the October, 2012 nv2 post being the only one I've found), so any pointers to relevant posts I've missed would be a great start.

Second, what is the best/recommended protocol/configuration construction of a many link, multilink PtP bridge? (I've seen this referenced as a 'string-of-pearls' but can't find much useful info) I gather that nv2 is not a good choice, except for perhaps the end link(s). The longest multilink I've constructed to date is 3 links in length - and it is nv2. I've not tested it's max throughput, but it is very stable.

Third, what's the longest - most end-to-end links - PtP bridge link you've installed, and what were the issues/best solutions? Some of these bridge links could be 10-12 PtP links in total end-to-end length.

Fourth, has anyone done an underground wireless installation? Experiences to share?

Fifth, could a 'bridge' be constructed so that all radios/routers (and voice repeaters on ether ports) would be on the same sub-net, to minimize packet processing and delay, or is routing required, given the 'complexity' of this network? Preserving MAC addresses is desirable.

Sixth, an alternative architecture could be to use WDS, with each 'node' an Access Point, but then I lose directional antennas and have to operate on a single frequency, correct? This seems problematic as well, as self-interference in these long corridors looks likely to be a big problem, and Omni antennas don't provide any 'directional' selectivity. But, maybe WDS MESH might be a plausible alternative to multilinks. Thoughts or experience on this?

Here's a thought, maybe crazy - anyone ever try to build/use a MIMO Access Point with directional, single polarity antennas connected, and pointing in different directions? You could get H vs V polarization differentiation on back-to-back, linear multilink connections, which should help with self-interference.

Thanks (and much Karma) in advance for any and all input.
 
samsung172
Forum Guru
Forum Guru
Posts: 1191
Joined: Sat Apr 04, 2009 3:45 am
Location: Østfold - Norway
Contact:

Re: best network design for multilink, UNDERGROUND backhaul.

Tue Apr 23, 2013 1:50 am

Consider the fresnell sone in a underground setup (link don't use free line of sight, but goes up and down)

Its possible to use regular mirrors in corners. - Ore you need power to equipment. And, indeed, if you have a power cable. Why don't also have a fiber going with the powercable?

NV2 and TDMA is not the best solution if you need low ping time (latency). Use old fashion nstreme.

Interference is no problem. Just use channels in both ends of band. And try to have the antennas 50cm apart. (in horizontal and vertical position)

How many radios you use in a long link, don't really matters. What matters most is how many radioes do you use in a bcast domain. And how many "clients" do you have "on the way". Its a bit like a water pipe. If you have a straight pipe, and don't take any water, until its end. How many pipe connection used on "the road" don't matter to much. if you have drains all the ways, the water coming out of the end, is not much left.

And yes. We have lab tested your "MIMO" antenna by changing polarity at 2 physical antennas. Its not that much different from using one dual pol antenna.


I think you biggest issue is fresnell sone, and it will destroy your setup completely. Its not that far away (you can calculate it) and fresnell (up and down) are 6 meter from center line. then your "tunnel" need to be 12 meter high. And more more more if really long tunnel. I have seen so many wired installations that don't take the fresnell sone in its calculations. most used mistake, is to mount a client 20 cm over the roof of a building (or 1,5 in a pipe etc on a big bulding) having 15-20 (or in the big building 50-60) meters to the edge.
 
mikepratt
newbie
Topic Author
Posts: 30
Joined: Wed Jan 11, 2012 7:31 pm

Re: best network design for multilink, UNDERGROUND backhaul.

Tue Apr 23, 2013 6:06 am

Thanks for the input, samsung172.

Tunnel 'corridors' are 35 ft wide by 25 ft tall. Max distance I'd attempt would be 2000 ft - most shorter. Directional antennas at really low power.

Another thought would be WDS at a single frequency, but then I'm using Omni or 'splitting' the radio chains, as I mentioned. Of course, throughput gets cut in half (or worse on multiple links?) due to re-transmission. Doesn't sound like a good idea in a long, PtP multi-chain link, but maybe I'm wrong. Max capacity is not a big concern - latency and stability are tops.

900MHz FZ is 16ft radius at that length - 5GHz FZ is 10ft. Antennas would be around 5ft from ceiling. Might get 'iffi' at max 2000ft. We'll try both and see how they perform.

Power is already there (hanging from ceiling)- 480 volts 3 phase, so transformers required.

Aerial fiber (everything must be up on the ceiling) is $2/ft, not counting hangers and terminal management hardware. Fiber termination gear is a 'wash' with radio costs, so all incremental costs of fiber over wireless would be the fiber and misc. hardware - a lot.

Would you use Nstreme with a single radio chain and not Nstreme dual? (I've never used Nstreme) How about 'fllipping' polarizaiton every link - even with frequency channel separation?

With respect to the thoughts about using a MIMO radio with the two chains on separate antennas - the thought was to use on chain pointed east and the other pointed west with opposite polarity - but, I realize they'd still be on the same frequency, so common channel interference coming back from the other end would still be an issue at the next linear link node (same frequency, but opposite polarity) to some degree - probably a lot.

Thanks again for the input.
 
samsung172
Forum Guru
Forum Guru
Posts: 1191
Joined: Sat Apr 04, 2009 3:45 am
Location: Østfold - Norway
Contact:

Re: best network design for multilink, UNDERGROUND backhaul.

Tue Apr 23, 2013 9:59 am

Its possible to flip polarisation , even have "1/4" polarisation (X).

To have good latency, I would use nstreme. To have high speed, use nv2.

I don't know about your place, but I can get fiber at 1$ per 3 meter. I can hire splicing unit for one day, at 150$.

If you have 100% free line of sight, and no nothing in the path and fresnell sone, a dual chain mimo would not help that much. Mimo wants "reflection" etc to get best speed.

I think I would try this setup. Use a radio link and test. I have no experience using 900Mhz (I have 2 pieces of 4 meter long Omni 900mhz antennas at store, but don't know for what to use this.)
 
mikepratt
newbie
Topic Author
Posts: 30
Joined: Wed Jan 11, 2012 7:31 pm

Re: best network design for multilink, UNDERGROUND backhaul.

Tue Apr 23, 2013 4:25 pm

Yes, samsung172, you can get fiber pretty cheap - but in this application it needs to be hung from the ceiling - like going from telephone pole to pole - so it needs to be aerial fiber to support its hanging weight and maintain bend radius - a lot more expensive. But still, $10K a mile is pretty cheap for what you get (6 fiber), but a lot more than free, as with RF spectrum.

I've seen direct burial fiber cable literally 'stapled' from tree to tree with fence post staples in India, so I know you can sometimes get away with less than textbook installation, but I would not attempt to use anything but aerial cable, as it might work initially, but fail as its weight 'bends' the cable over time.

The 'corridors' are 'broken' every 70 feet, by crossing 'corridors' so I expect a lot of 'scatter' if not simple reflection. Don't know how MIMO will respond, but I will probably try it, to compare nv2 2x2 MIMO with Nstreme 1X1. Well see.

When the antennas are hung approx. 5 feet from ceiling, there will certainly be some FZ intrusion at 900MHz on longer runs, and maybe some at 5MHz and on shorter runs, if ceiling is irregular or not perfectly planar. I don't know exactly how planar the ceiling/floors are, as I have not 'laser surveyed' them, but they look pretty true. Anyway, I expect some FZ intrusion, but 'hope' to maintain 60% open.

I'll post results. Thanks for the input!

Who is online

Users browsing this forum: redjahred and 28 guests