As my cabin project is slowly evolving from one small shelter to a few tiny structures with the potential to include automated small-scale farming in the future, I was looking at the possibility of connecting them together in some fashion that didn't involve wires. I dislike tech for tech's sake so everything needed a purpose. More importantly, I didn't like the idea of babysitting the network. Writing and enjoying a wood stove flame seem like a better use of my time.
Off-the-shelf access points with open source firmware was the easy choice. Probably those capable of OpenWRT or DD-WRT. The device and antenna would likely need to live outdoors for better connectivity, which means I'd rather have a single wire coming through the walls of the structures. That narrows it down to Power over Ethernet capable devices.
These "connectivity nodes" will then have the following constraints:
They must be remotely updatable
They must be powered off-grid
They must survive exposure to sunlight
Thay must withstand heavy wind, rain, and snow
They must be immune to occasional smoke (I.E. from a wood stove)
The last requirement might be particuliar to my situation, but I don't plan on burning a lot of wood. If anything, my own stove experiments have shown that it's possible to be efficient and have a relatively clean-burning fire with properly seasoned fuel and a good stove design.
I would likely have to create my own antenna to get decent reception since most stock devices come with omnidirectional antennas. I probably won't need modify the rest of the device functionality.
The case would probably need to be made of UV resistant resin, transparent to RF, and still durable in the outdoor environment. I'll have to conduct some experiments to see what works.
Meanwhile, I came cross this particular antenna, featured by Andrew Mcneil, that seems relatively easy to make with plain copper sheets. Provided, I use the correct measurements.
I thoroughly enjoy that channel, not just for the information, but his manner of presentation too. Andrew is a bit like the Bob Ross of antennas.
I don't think I'll need the extra elements since I plan on keeping all structures within line-of-sight. I prefer the 2.4GHz frequency range since it's still relatively good in bad weather, even heavy snow.
What I like the most about this design is its solid construction and lack of a PCB. This means I don't have to fiddle with etching to make these and instead, simply cut copper sheets to size and shape. Etching PCBs is always a messy affair, especially with the harsh chemicals involved, and I'd rather avoid it if I can.
The design itself is relatively flat, compared to most other Yagi-based antennas, which means it will be less prone to collecting ice and snow and less likely to act like a sail in high winds. I can mount it directly to the side of each structure that needs connectivity, like a porch light, while aiming at the next closest node to create the wireless mesh.
I'll probably need to copy the overall shape of the enclosure seen in Andrews' original presentation of the antenna.
The node hardware platform will likely revolve around a small travel router or access point board, with an external antenna and PoE capability.
I have several in mind like the GL-AR150 and similar devices. At present, OpenWRT 18.06 seems to be supported on that device. I want to make sure whichever device I settle on as a base for this will be at least be secure for the foreseeable future, if not feature rich. If I use two antennas, I can dedicate one to transmit and the other to receive, creating a true duplex link.
There's an old truism in marine research: "Don't put anything in the ocean you don't plan to lose". Or something to that effect. The harsh environment of the ocean and its vast depths being the biggest enemies. There's a much less intense parallel on land; Anything exposed to the Sun and the elements will eventually degrade and malfunction.
The enclosures that come with most access points are no match for the elements, which means I would either have to make a DIY case for it with an outlet for the ethernet cable or buy a NEMA-rated enclosure and modify it to my needs. I'd rather not have a large, ugly, portruding "thing" outside my structures, even in the name of better connectivity, so some customization will be necessary anyway. The devices themselves will need to be tested independently to ensure they can survive the temperature swings. I can at least control the humidity by using adequate seals on the cases.
The cases are probably going to be the most complicated parts. UV resistant resins are available in the market, usually labled "marine grade", but I have no experience casting shapes from these.
Another useful skill to learn.
All in all, it's been fun exploring what options are available to me should I pursue this path to connectivity.
Yesterday was one of the hottest days in New York in several years. Most people in my line of work have the respite of office air conditioning, but a lot of others were caught in terrible circumstances. Considering what they had to go through, my complaints seem like trifles.
But they're my trifles.
Once upon a time, this site and its siblings were on an old Thinkpad with a dead battery and cracked screen. It was at home anywhere as long as WiFi and power was available. Because of a possible hardware failure that affected the mainboard, I had to move it to a "proper" server; Another obsolete hunk that was taking up space. This was connected to a battery backup with 2 hours of reserve.
Unfortunately, yesterday, on the hottest day of the week, the power went out for 4 hours and only came back sporadically until this Monday morning. To prevent data loss, I shut down the whole thing for most of the day.
I appreciate the nimble notebook even more now.
This made me think of what would be a viable alternative in a mobile web platform. Without a fixed IP address, hosting a clearnet site is cumbersome, but Hidden Sites have no such ailment. I've been considering moving all this to a simple box with a battery and solar panel, leaving the connectivity purely on WiFi. The benefit is that it doesn't matter where the connectivity comes from. Even a shoddy public hotspot will suffice in a pinch. Hidden Services in Onion Land only need to stay connected for an hour at a stretch at most for other nodes to be aware of it.
A while ago, when I was still using social media, I had a chat (one of many there) about the possibility of using cheap and easily produced connectivity nodes for a ubiquitous WiFi mesh. Unfortunately, Mastodon didn't save some of the early images from back then, but I'm sure I have them somewhere in my backups.
There are lots of possibilities to consider, including an eventual upstream connection to the open web via internet exchange at some point. But the technology to pursue this is more available and affordable now than even a few years ago.
I'm also considering combining this idea with bringing back telephone booths. With the obvious privacy concerns of public kiosks, if they can still be allocating unused bandwidth toward the internal mesh, then Tor encryption can still serve to prevent each node from discovering the final destination of its neighbors' traffic.
The best part is that because the electronics on each device require so little power, they can be individually powered with distributed sources. Wireless routers themselves can provide hop connectivity while being completely off-grid. No more reliance on utilities that must maintain decades of cobwebs that melt in the heat.
Sites like these that sip bits don't need high-speed highways and can remain primarily mesh-connected. But streaming videos and games are another matter. I'm probably going to maintain a high-bandwidth connection in addition to a mesh link, even if it one day becomes a thing. Nice to dream what cabin life on a mesh would be like.
I've mostly gone back to using my phone for calls and text messages. I do sometimes still browse the internet on it, but I fire up the laptop more often for that these days. The larger screen is kinder to my eyes as well. Public transport is usually a respite for me so I try to turn off the rest of the world while on the road, internet included. Silence is a rare commodity in modern civilization, with an exponentially increasing value.
I remember in high school, I got my first mobile flip phone for emergency use. Most other kids with some form of remote communication had beepers. Again, for emergencies, and the school payphones were in high demand.
Communication is a lot cheaper now, but that has come with the cheapening of conversations. There's a noticeable lack of warmth in the smiles I see every day. More people greet each other by bumping into them while looking down on their phone. I don't think we've faced a period in time where people are so connected and yet be so alone. The immediacy of feedback rarely carries the weight of longing and so most conversations, even across continents, feel trivial.
While it's naive to think that having more phone booths will prevent people from accessing their personal supercomputers as long as they have them in their pockets, I think there's still a need for a dedicated, permanent access point for communication.
Phone batteries die all the time. Connection can still be spotty. Privacy is harder on a device that stays with you and sometimes, you really do need to use a screen at least the size of a tablet. Not carrying any sort of complex electronics that can track you might also come into fashion at some point in the future.
When I think of "booths", I usually envision the classic, red phone booths in the UK and the colonies.
But let's be more practical here. These are going to be expensive to recreate and, in an age of lax moral discretion, far more likely to be vandalized. An alternative is to create a monolithic pillar with few protrusions and a simple screen for interaction. While it can serve primarily as a contact point, it can double as a browsing screen with a bigger view. In that regard, it would make more sense to call these "nodes" instead.
The idea for this is of course derived from Blade Runner. Users Staar and Toa Quarax posted these on a PropSummit forum thread a while ago.
Isn't it beautiful?
These are apparently operated by a fictional version of Ma Bell, back when AT&T held on to its roots as the Bell Telephone Company, noted by the conspicuous bell logo. And they're called "Vid Phon".
This being a "Vid Phon", there is a camera next to the speaker. I think that's a speaker, but it could also be a futuristic "depth thing" a-la Microsoft Kinect.
Monopolies belong in cyberpunk fiction, but the real world breakup of Ma Bell was obviously unforeseen. AT&T is now a monopoly again, but that's another topic.
On the same thread, user nickdaring posted this possible concept version.
Unlike the concept, they've minimized the final design to be more streamlined. While that makes it less inviting for privacy, it's also less prone to damage outdoors and "in the wild". The front interface is much wider and the keypad is more prominent in the final design as well.
The screen appears worse for wear. Being the most obvious focual point, it remains as the last vandal magnet on the entire device.
It seems to be a little over 79cm tall, which actually is within the range of a typical payphone.
Note, the back of this is actually made of wood. Being a prop, it doesn't actually need to be vandal-resistant.
I think we can dispense with the phone handset altogether and instead try to use speaker tricks to converge sound to a single point directly in front. This will give some privacy while eliminating yet another potential vandal magnet.
What I especially like about them is the appearence of near-bullet-proof construction. Almost like a bollard turned into a node; Which itself may not be a bad idea. Instead of a row of bollards which do nothing but prevent cars from jumping the curb, why not extend upper portion into an access point.
That's not to say, the rest of the device won't face some damage. The care gone into the detail here shows that the cover is meant to resemble a detatchable part whereas the inner core is another "metallic" object. This would make refurbishment easier on an actual device by simply replacing the cover.
I mentioned earlier about not carrying complex electronics which can track you. Disposable phone cards are one place where this might be viable, provided the cards can still be purchased with cash. The Vid Phone cards are gorgeous as well.
I've also been thinking about a potential technology stack to make this work; The set of features for a communications platform that can be left unattended for long periods of time. Instead of a typical desktop computer, I was thinking of single-board computers which can easily be solar powered and run on batteries with comparatively little maintenance. Most can be configured to run fanless, which makes them quiet, and they're designed for industrial enviornments which aren't friendly to delicate electronics in general. That also makes them more resilient to bad weather. Many have ample GPIO (General Purpose Input Output) pins which enable environmental monitoring, with appropriate sensors, at the node location as well.
The Pine64 series of single-board computers seems the most viable for availability, cost, and foreseeable long-term support. There are many others out there, but this seems the most cost-effective while being supported by the Operating System I've also been thinking of for this: OpenBSD. The current arm64 port does support the Rockchip RK3399 platform so the main board will likely be a ROCKPro64. Even though it's the most power-hungry of the lot, it can run fanless with a tall heatsink and should have enough processor power and memory capacity to function as a decent kiosk. At the time of this post, it costs $80.
The other option is the newly released Raspberry Pi 3 with its dual monitor support and more attractive price. The down side is that there are some USB-C issues caused by an implementation that didn't follow the spec. Besides this, the board is barely a few weeks old and I don't know how well it will perform in the long run. It hasn't been proven in the field yet.
The ROCKPro64 could also potentially be coupled with an ESP32 for nearby Wifi and Bluetooth as a 1-2 user hotspot. This too is a very cheap component.
I'd like to implement these in pairs of computers. One functioning as the node out in the field, which connects to a "handler" at a central location.
The "handler" will likely be the Sopine compute modules, which can sit on a clusterboard, with seven modules per cluster. Making one Sopine compute board responsible for one node makes running the system more efficient in terms of resources instead of dealing with one big server handling many nodes. Both the node and handler location can be solar powered since there aren't any severe cooling requirements with a cluster of Sopine boards besides a fan or two per cluster.
The nodes out in the field are are essentially "fat clients", the opposite of "thin clients" usually seen on kiosk duty. Fat clients have more resources to deal with the applications they're currently running, and they do run them locally most of the time with only data saving and communications sent back.
The "handler" only dispatches the boot image, provides application copies, system updates etc... This allows only a minimal system to be installed on the nodes, making updating the whole thing easier. If there's any problem with a node, try sending a reboot command while swapping out the Sopine compute board. It also gives a degree of resiliency and decentralization since there isn't a single server responsible for a whole swath of nodes. And since each clusterboard can be on its own ethernet connection, only a single cluster with seven nodes will go down if the port fails. This also enables the handlers and nodes to be independently updated in staggered rollouts without taking down the entire node system.
Speaking of the network, media converters can be used to maintain Ethernet connectivity over much longer distances. There are media converters with 2km to 20km range at reasonable prices now and most don't consume much power, keeping that solar power option open. Using fiber out to the field also makes it less attractive to copper thieves while maintaining high bandwidth. The whole package for the node - case (bulk of the cost), mainboard, media converter, batteries, solar panel - will probably be around $1000. Adding it on top of a bollard will make installation simpler so that's an additional reduction in cost.
There are still a great number of details to work out in this project, the least of which is the software configuration. OpenBSD makes for a very simple desktop, especially when coupled with Xfce in kiosk mode.
A lot to digest, but I've been thinking strongly about all of this for some time. All of this will have to come after my cabin project, of course.