View Full Version : XR9 network design
I work with a nonprofit organization and help organize long distance trail runs in various places here in the US. The races take place in the eastern woodlands where there are many hills and valleys. Each race has a series of checkpoints with laptops to track participants. I'm working on networking these sites together to convey logistics and tracking data, among other things, in a reliable way. In many areas, cell service is poor or non-existent, which rules out any cell-based solution. The UHF radio modems I've looked at aren't fast enough -- I'm looking for minimum 1Mbps across the network.
Because the race only lasts one or two days, the network is transient. This means no fixed structures, and the gear must be transportable with a pickup or small truck. We'll use Radio Mobile to evaluate where to put relay points depending on the topography of the particular course. We'll drive to each location, set up the site (powered by batteries), configure it, then remove it afterwards.
Below is an elevation profile showing a 10-mile section of one of the courses. The blue stations (checkpoints) are all in valleys (typically where the roads and towns are), and the red relays are on the hilltops. I'm using 900Mhz for the relay links, because of the hills and forests. Each station has both 900Mhz for the uplink and 2.4Ghz for local (~150 ft) laptop access. Radio Mobile shows acceptable connectivity between everything.
Assumptions:
1. The relays and stations use XR9s.
2. The 900Mhz links use 6dbi omni antennas.
http://i44.tinypic.com/2urmtcl.gif
So, now to the questions:
1. What platform(s) should I look at? Remember, for the stations I need both 900Mhz and 2.4Ghz.
2. What protocol would support this config? WDS? or something like Open Mesh?
3. Where possible, one of the relays could connect to the cell network to gain Internet access. What interface would be most flexible for that? USB?
4. For the 900Mhz links, it seems sector antennas would be better at the relay points. But since we rarely would need 360 coverage, it may actually be cheaper to get three XR9s and three yagis. Or am I missing something?
Thanks for any insight.
So, now to the questions:
1. What platform(s) should I look at? Remember, for the stations I need both 900Mhz and 2.4Ghz.
Answer: For the backhauls, a single XR9 (connected to a cellular modem) with 25 dBm TPO and 11 dBi omni (36 dBm EIRP) will get you a 28 dB fade margin at six miles with an XR9 and 13 dBi Yagi at the far end. At the far end points. connect the XR9 to a NS with an omni antenna for 150 ft. coverage.
2. What protocol would support this config? WDS? or something like Open Mesh?
Answer: Mesh has its own problems, keep it simple as you really don't need it. WDS isn't going to work unless the whole network is on 2.4 Gig, although you could WDS the 900 Meg to 2.4 Gig stuff for a little better performance, but not that much more to make a difference if you're going to use a less than 3 Mbps cellular connection.
EDITED: You could use WDS at the far end repeaters, I just looked back at your drawing.
3. Where possible, one of the relays could connect to the cell network to gain Internet access. What interface would be most flexible for that? USB?
Answer: Even thought several cellualr modems have USB connections, I'd avoid USB as it requires a laptop (with Internet Connection Sharing enabled). 3Gstore http://3gstore.com has reviews of the of the Cradlepoint products http://www.cradlepoint.com
4. For the 900Mhz links, it seems sector antennas would be better at the relay points. But since we rarely would need 360 coverage, it may actually be cheaper to get three XR9s and three yagis.
Answer: A sector antenna isn't going to have that much more gain over an omni. See answer #1. On the other hand, using three XR9 radios and yagis means you'll have a PtP set up and you can run them at 26 dBm (the "add 3, subtract 1" rule) with a 13 dBi yagi and get a 31 dB fade margin. But for a mere 3 dB more gain, it may not be worth the trouble aligning the base station antennas.
Thanks for sharing your knowledge! To clarify, the network won't require a backhaul connection. An Internet connection is not essential, but connectivity between all the stations is.
Regarding the platform, I was thinking about the base hardware. I've seen the RB411 mentioned, but UBT has a RouterStation that looks decent. Any thoughts in this area? I like the 3 slots available in the RouterStation.
I'd like to minimize the different hardware needed, so my first thought for the local wifi would be to just get something like an SR2 card for each station and plug that into an available slot. But I'd need an antenna too, so the Nanostation Loco makes more sense cost-wise. I'd think the Loco would suffice for the 150 ft. wifi links as it's outdoor and mostly LOS.
I'm not clear how wireless repeater links work. In particular, below is the Radio Mobile diagram of all the stations and relays for one particular course. As you can see, Relay 2 is visible to three stations and three relays. If a packet leaves Station 3 destined for Station 6, there are several paths it can take once it gets to Relay 2. What determines its route?
http://i39.tinypic.com/2jfkhkz.gif
Thanks for the lead on the Cradlepoint devices. I'll check them out.
In your Radio Mobile, make sure the Role of each member is Slave. It looks like you have a few set as Master. Or you could set just one as a Master if its the primary originating AP.
Also in your Topology tab, use Data Net - Star Topology (Master/Slave)
That gets to the heart of my question. When I actually set up the relay (for real, not in Radio Mobile), do I configure it to connect only to specific relays? In this case, Relay 2 should only connect with Relay 1 and Relay 3 (and also Station 2 and 3?). I gather the routing informatoin is propagated automatically?
If so that's fine, but is there a way to have the relays dynamically connect and route depending on path latency? It'd be quite useful to have redundancy in the network in case a relay fails.
That would be what a mesh network does. While you could use third-party firmware on all the stations, you won't be able to use directional antennas. At the distances you are looking at, omni-to-omni most likely isn't going to work.
So what you want is to have everything on the same network where all laptops can communicate with each other, right?
Yes, that's right. All laptops should be able to talk to each other across the net.
I set up Radio Mobile with 6 dbi omnis everywhere with 28 dbm xmit and -95 dbm receive, 25% additional forest loss. It showed acceptable connectivity. The image above is with 11 dbi omnis.
When you look at the Radio Link, what is the signal level in the far right side at the top?
This is with 11 dbi omnis:
http://i44.tinypic.com/r2sr2r.gif
That means you have a 10.9 dB fade margin. For a one or two mile link, that's somewhat acceptable. For a six mile link, it could be iffy.
BTW...look at your Fresnel zone clearance, its only .3, you need AT LEAST .5 to make it work.
Thanks. I can improve both by increasing the antenna heights to 30 feet (fz=0.6F1, rx rel=17.4db). Now all I need is to find a portable way of getting an antenna that high.
Mesh has its own problems
Can you elaborate on that statement? What mesh technologies have you used, in what scenarios, and what did you find?
Mesh doesn't scale very well. For each hop you half your speed. Its best used when you have a compelling need to use it, compared to simply deciding on using it because its cool.
Of all the city wide muni mesh projects I've read about, they don't provide the expected coverage, end up costing far more than anticipated...and most have failed to get finally deployed.
