Nanostations, OLSR, and OpenWRT

Well, it has been an interesting period of time since my last post.

Please understand that I have been simultaneously working on the technical aspects of the Ayrstone AyrMesh(TM), updating the web page, writing the business plan, and preparing to show the product at the World Ag Expo in Tulare, CA.

So, first, the technology. While OLSR does definitely work on Ubiquiti's AirOS, the odd fact is that they have somehow messed with MadWiFi so that you cannot create a VAP running in adhoc or ahdemo mode. This, of course, effectively prevents the creation of a mesh network using AirOS.

Ubiquiti must have had some very good reason for doing this, but I have not yet had the temerity to ask them.

I decided at this point to start working with the open-source OpenWRT OS, a Linux distribution originally built for the Linksys WRT54G. It has been ported to the Atheros chipset used by Ubiquiti and runs very well on the NanoStations while providing a lot more control. I started by studing two existing meshing packages: RoBIn and NightWing. From them I was able to figure out how to build a version of OpenWRT that (1) fits in the 4 MB space on the NanoStation and (2) does what we want it to do. So far it's working very, very nicely.

A very nice added bonus is that it works perfectly on the new Ubiquiti Bullet2HP, which is better-suited to our purposes than the NanoStation. I have one of the early units, and it is working very well.

In the meantime, we showed the Ayrstone AyrMesh at the World Ag Expo and were very well-recieved there. People asked what we had and, when we explained, were simultaneously very surprised and very interested. Those who were able to fully understand the capabilities and potential of the concept were eager to try it out, although, in fairness, not everyone really understood it. This is the bane of the high-tech marketer's existance: figuring out how to explain technology so that people understand the full potential of the device - the facsimile machine being a "classic" example.

Meanwhile, we are polishing up the business plan and starting to look for funding. Anyone with a few hundred thousand that you'd like to use to buy into the next high-growth agricultural technology company, please form an orderly line...

OLSR on NanoStation 2

As noted earlier, I built the Nanostation firmware for the Nanostation 2 using the code and directions found here: https://wiki.graz.funkfeuer.at/UbntStations

You can get the code for the NanoStation 5 there, of course, and I have posted my build of the code for the NanoStation 2 here.

Please let me know if you want to try it out - bmoffitt "at" bmoffitt "dot" com

More from Radio Boy...

A friend of mine, a few Christmases back, bought me a couple of old serial books, part of the "Radio Boys" series, about a group of boys experimenting with radios (new-fangled, high-tech stuff). They make me smile every time I look over at them - from the 1920s, the language is somewhat odd, but the stories are cute.

I'm trying to find the "perfect" radio system for this product. The perfect radio system will be no-config - just power it up and it will figure out its role. Each node should automatically configure itself as a gateway, repeater, or client (or, even better, repeater-client). I know this is possible, because the Open-Mesh boxes do it. They're just not suitable for outdoor use.

So far, the Ubiquiti radios (I have been testing the NanoStation 2) have won the day - mostly open-source firmware, with a very nice SDK for developing new extensions. The only problem is that the only "meshing" protocol that comes with them is WDS with STP. It actually works pretty well, and will get the job done. However, it requires more setup than I like (not automatic configuration) and doesn't use the "best" encryption algorithm (WEP vs. WPA).

The Open-Mesh boxes run an interesting mix of software, and they can use either B.A.T.M.A.N. or OLSR as the meshing protocol. I recently changed my Open-Mesh boxes from B.A.T.M.A.N. to OLSR and noticed no difference in performance. Running either of these protocols would seem to be a possible solution, but I'd have to change my NanoStations over to OpenWRT, which I am somewhat reluctant to do.

However, I recently found a fellow in Austria who has ported OLSR to Ubiquiti's AirOS as part of a "free wi-fi network" project - see https://wiki.graz.funkfeuer.at/UbntStations.

So I downloaded the SDK, code, patches, and other files from ubnt.com and from funkfeuer.at, and made myself a firmware file for the NanoStation 2 (on about the fifth try - hey, I haven't done real software development for over 15 years!) I loaded it on to one of the Nanos, fired it up, turned it on, and it was running happily.

There are still some small glitches to be worked out before this solution is ready for prime time, so, for now, we'll keep going with WDS. However, the more I read about OLSR the more excited I am about it. The benefits for the Ayrstone customers are great.

Nanostations in the "field"

As noted in my previous post, I have been doing quite a lot of work with wi-fi radios. I took the Ubiquiti Nanostation 2 radios out for some in-field testing, and discovered some interesting things.

First, I set one up on my roof to try my "integrated solution, Mark I:" Nanostation 2 hooked up to a 7 Amp-hour (A-h) 12 volt battery, with a 5 Watt solar panel and a SunGard charge controller. It worked very well for about two days, and then the battery went dead. I then replaced the 5 watt solar panel with a 10 watt unit, and it went until the first storm before dying. Bottom line: 7 A-h is not enough battery.

In my testing, I have found that the Nanostation 2 pulls about .2 - .25 amps, or about 3 watts. Hence, to run all day requires about 72 watt-hours, or, at 12 volts, almost 6 A-h. If you don't get sunlight for a day, the battery is toast.

My next experiment was to replace the Nanostation with a 63 ohm resistor and try taking out the charge controller and see what effect that would have. The battery has been progressively losing voltage and not coming back up to full charge, and it's at 8 volts today. This indicates that the 10 W solar panel might not be quite powerful enough to keep up with the Nanostation. Interestingly, the resistor is rated for 5 watts, but it's mighty warm.

I found you could get a 50 A-h marine battery at Costco for about $50, and that should provide enough storage to get through a storm, but it won't fit into my scrounged-from-Halted NEMA 4 weatherproof box.

My next adventure involved some radio range testing with the Nanostations. I cajoled a friend of mine to come down to the fields in the Coyote Valley on Santa Teresa Blvd. We didn't have a lot of time, so I put simple "9 dBi" rubber-ducky antennas onto the radios, put one up on the truck, and sent Ethan out for a half-mile walk with a GPS and another radio. I watched the signal strength on my laptop from the truck, and was encouraged by the results.

I also tried replacing the 9 dBi rubber ducky on my end with a 15 dBi, 6+ foot antenna I had purchased. The signal strength diminished noticeably, which was disturbing, but we were out of time so we packed up and went home.

Then, last week, I went afield again to do some testing. I put one Nanostation on a 10' pole on a fencepost, and then drove to another point .85 miles away. What was intriguing about this experiment was the behavior of the "remote" radio. The stationary radio was set on Channel 3, using Wireless Distribution System (WDS), in AP mode, with the MAC address of the remote radio "hard-wired" (entered manually into the WDS table).

The remote radio was started in WDS-AP mode, but, despite having better than -80 dBm signal, would not "link" with the stationary radio. Reset into WDS-Station mode, it linked up perfectly and was able to transfer information back and forth at more than a Megabyte per second.

Why could it not connect in AP mode? Is it more sensitive to distance? (I found the distance parameter on the "stationary" radio was set to half a mile, but I can't imagine that could make too much difference) These are the mysteries in my life.

Incidentally, I tried the 15 dB antenna again and, again, it was outperformed by the rubber duckie. I thought it could have been the frequency (I had tested it on channel 11 the first time), but it was almost as bad on channel 3. I thought it could have been off-vertical (it has a pretty narrow beamwidth), so I moved it around vertical for a while to see if it got better. It didn't.

So I'm decidedly disenchanted by the 15 dBi antenna, and I'm still not sure about hooking up the NanoStations as a pseudo-mesh using WDS, half a mile apart.

Watch this space, as they say.

Three good months of learning about wi-fi, writing and re-writing business plans, and networking, and I'm approximately where I was when I wrote the last entry.

But I'm learning a lot.

I started out by evaluating Meraki - they have very nice equipment and a system that is really both beautiful and effective. I was able to get a couple of Meraki nodes on Ebay and put them up on my home network. I can't say enough good about them - the system really works well for providing well-isolated internet access across a geographical area. However, my problem is that I want to distribute a LAN, not just internet access, and Meraki's system does not enable access to your LAN. This is a good thing in a "metro" network (you don't want random folks wandering through your LAN) but doesn't fit into my needs.

I then evaluated Open-Mesh, buying two nodes from Open-Mesh.com. I found some basic reliability problems with them, particularly the well-documented problem of the nodes losing connectivity after a brief poser outage. I actually found mine would not connect if the power was off for less than 20 seconds. When I move or reboot a node, I have to remember to unplug it, go do something else, then plug it in again.

I initially decided that they couldn't access my local network because, while they posessed 192.168.1.x addresses (my local NAT) and I could SSH to them at that address, they were providing DHCP addresses in a different net. The Open-Mesh nodes are also very low-power and not suitable for outdoor use. Given those disadvantages, I decided to put them aside.

I took a look at Ubiquiti's radios, particularly the Nanostation 2. It's a terrific radio - sealed, weather-resistant, with an open-source-derived OS called AirOS for which they provide an entire SDK to suppport the creation of third-party software. It also puts out 26 dBm to either an external antenna (now attached with the "standard" SMA-RP connector) or an internal "adaptive polarity" 10 dBi directional patch antenna. One of the standard features is a Wireless Distribution System (or Wireless Domain System, WDS) implementation with an "Auto" feature that can automatically associate with other nearby WDS radios. They also implement Spanning Tree Protocol (STP) to prevent looping in this sort of an ad-hoc network. This provides the same benefits of B.A.T.M.A.N., the protocol used in Open-Mesh, and just distributes the LAN instead of providing "public" hotspots, client isolation, and other features I don't want or need.

It turns out that there are two problems with this approach. First, the "Auto" feature generates a lot of traffic as the number of nodes increased. With three nodes on "Auto" WDS mode, the traffic overwhelmed my little (factory-spec) Linksys WRT54GS. I was afraid of what might happen when I pushed that up to 10 or 20 nodes.

The second problem is that WPA cannot be used in conjunction with WDS. This is a technical limitation of at least their implementation (see here and numerous other places in the Ubiquiti forums) if not WDS and WPA in general.

This brings me to a conundrum: WEP is not very secure (better than nothing, but still...) and the only way to get WDS to work reliably is to set up the partnering manually.

Now, going back to Open-Mesh, it turns out that you can access the LAN, but it does not seem to be reliable in my network. Sometimes it works, sometimes for some nodes, sometimes it doesn't work, for any nodes. However, if I could get the B.A.T.M.A.N. - RoBIn combo working the way I want, I could load OpenWRT onto the NanoStations and try building a NanoStation OpenMesh node. In fact, I might even try just running B.A.T.M.A.N. and see if I could build what I want without the other parts.

That's where the ag mesh network is as we swing into the Thanksgiving holiday.

More to come...

Another fine mesh

Greetings from my laboratory. My house has become a hotbed of microwave radiation. I now have my old, reliable D-Link router/switch/access point, two Meraki mesh routers (one "mini" and one "outdoor"), and two Open-mesh routers working in and around my very small house. There are also a Linksys WRT54GS and a $40 OpenWRT box (purchased from the internet) kicking around here, as well. Oh, and just to be complete, I got myself a FON router and directional antenna, because it seemed like a handy thing and was very inexpensive.

While the popcorn is not quite popping itself, there is a lot or RF in the air here. Technically, all the Meraki, open-mesh, and FON boxes are roughly the same, running slightly different software. And, based on some informal testing, they all seem to work pretty well. There are some interesting differences: the FON box has two ethernet ports (which may make it impossible to turn it into an open-mesh box should that be my eventual goal). The Meraki mini is identical to the open-mesh routers, but the Meraki outdoor has a more powerful transceiver (200 mW instead of the 60 mW in the "minis").

Results to come...

...and then 11 months pass...

...and we're back. Briefly, Cisco didn't work out, and I'm glad to be back out on my own again. But my time there was valuable, and I am very grateful to them both for employing me for a few months and for letting me go gracefully.

I don't believe in a deterministic world, but I do believe that fate will sometimes whisper in your ear, and you should pay attention to it.

When I went to Cisco, we had been talking about the concept of farm wireless networks. There is value in having a wireless network on a farm, for data collection (soil sensors, irrigation systems, weather stations, yield monitors, telematics) as well as for internet access (being able to use your laptop from the tractor, for instance). In many ways, Cisco was a very good place for me to be surrounded by networking technology while I was thinking about this concept.

So I'm here, writing a business plan for a new chapter for Ayrstone, based on wireless networking and web technology. There are a lot of nice things out there that the agricultural world isn't using yet, and I want to bring some of the best of "silicon valley" technology to the "great midwest."

In particular, I'm excited about the possibility of using relatively low-cost "repeater" technology from Meraki, particularly their outdoor repeater and perhaps their upcoming solar model. I think that, by mating these to 12 dB gain antennas, we can actually light up a pretty fair area (one per section???). I'm thinking that, by using a network of 5-10 of these, covering a 2000+ acre farm should be possible for a few thousand dollars. My thinking is that, even if the signal goes three hops, you're still in the multi-megabit range with 802.11g, which is just fine.

The devil's in the details, of course, so we need to get out and prove the concept.

My Meraki equipment (and a couple of routers from open-mesh, and erstwhile competitor to Meraki) are on order. Should be fun...

Stay tuned!

A movement to the side, perhaps to move ahead

As the summer progressed, it became clear that the current operations for Ayrstone weren't going to start making a living for me, and the opportunities we were seeing weren't looking like they'd "blow up" fast enough and large enough to deliver enough money to keep our household afloat.

Furthermore, while I'm still excited about the idea of a farm equipment review site, RateItAll, while intriguing, just doesn't seem to be robust enough to do what I had hoped (I did an experiment to rate precision ag sites). That leaves me with the option of building a ratings engine or using a pre-built one (like RedQueen), which is very time-consuming. Since I didn't have the time to consume, I decided to look for a paying gig.

Happily, about that time I was contacted by a recruiter from Cisco Systems, who was looking for some help in their Networking Academy. The Networking Academy is a fascinating group at Cisco: they create curricula in networking technology to help schools train students to get good jobs in networking, and they give it away, for free, and help the schools adopt it. It's self-serving, of course, but it actually does good for the schools and students - people learn critical skills to get a good job and move up the economic ladder.

This seems like a step sideways, but I don't see it that way. I think one of the most exciting opportunities coming up is networking the farm and getting more farm operations onto the web. This gives me a perfect seat to understand the technology and products available better.

So I'm excited, but I still want to build the review/rating site as I have a chance. Watch for more on that...

Here's a new twist...

As you can see to the right, I have added a new widget from RateItAll.com, which allows you to rate this blog without even going to RateItAll. This widget is currently in beta, so I encourage all the readers out there to go ahead and cast your arrows at me using this tool, and then let me know what you think of the widget itself via email at bmoffitt (at) ayrstone.com.

Beginning the next fun thing...

As noted in the last entry, I am looking at starting a rating service for agricultural products. As a beginning, I have started a list on rateitall.com for Precision Ag web sites. See it here: http://www.rateitall.com/t-23470-precision-ag-web-sites.aspx

I think this is kinda cool - even better is to move this all to a promotable site and populate that site with the flash widgets for each piece of equipment so users need never go to rateitall.