The "wall wart" will be out in the rain, humidity, freezing cold -- This is not some comfy office environment. UL will not approve virtually any connection to 120VAC without some form of protection.

Sorry for some of my latests posts. I had a troubling weekend and my rough edges showed. I apologize.

Look for IP67 or IP68 rated power supplies. IP67 means dust proof, and that the device can handle immersion in 1m of water for half an hour at least (indefinitely for IP68). Power supply temperature ranges vary, but -25°C to +40°C is typical; just check. (here are some 12V examples geared towards outdoor LED use.)

You'll also need IP67 on IP68 rated connectors both on the 120VAC/240VAC side, and the device 5VDC/12VDC side (unless you wire the power supply directly inside the enclosure). The current levels are so low, less than an amp on the mains side, and a few amps on the device side, that these should pose no problem either.

Working temperature range for example for the SAM9G45 boards is -40°C to +85°C. Remember that the board itself produces heat; -40°C to +40°C ambient should be okay with a sensible enclosure. Get a suitable, comfortably larger than the board, IP67 or IP68 rated aluminium enclosure (see examples here) to put the board into. As the device will put out 5-40 watts of heat, depending on the board, overheating is the only real danger. Aluminium does not corrode easily, and conducts heat well, and is therefore usually the best option. A plastic enclosure should also work well, but it should be comparably larger -- more surface area, as it does not conduct heat as well as aluminium -- to transfer the heat.

Additionally, if you put the enclosure in a greenhouse, you could use a Peltier element to transfer excess heat from inside the chassis to the outside, if you use an aluminium enclosure. A Peltier element is not very energy efficient (it'll basically double the power used by the device, but it'll keep it cool), but it is certainly better than some complex radiator arrangement. You can even get Peltier elements in IP68, if you wish to transfer the heat from the enclosure outside the enclosure, say to a large metal surface.

For data storage, you'll have to rely on either a remote server, or flash media. If you are worried about wearing out the Flash media, just use a bigger one. Wear leveling means the device will write to the least written to unused sector, more or less. (Pick the writeout interval large enough so that the device does not need to do many read-modify-write cycles. Don't use O_DIRECT; let the kernel and the device manage the details.)

It makes a lot of sense to use separate devices for OS, temporary files and logs, and data. Mount the OS read-only by default, and just remount it read-write when updating.

So, weatherproofing should not be a problem. It does cost, but not as much as you might expect. Just make sure you use IP67 or IP68 enclosures and connectors. Many suitable power supplies and ARM boards run in a wide temperature range, up to -40°C to +85°C, so check for that, too.

2 members found this post helpful.

IMHO, if your board dissipates greater than 2 watts, it is the wrong board. Sub-1-watt boards are fairly common. I use an ARM board in a fully enclosed (but not hermetically sealed) aluminum enclosure, and it does not feel perceptively warm on the exterior. I just got a batch with built-in temperature sensors, and will shortly have some actual temperature data to provide.
Peltier coolers are only good for transferring heat from one place to another. They are horribly inefficient, and would add way more heat overall (net power dissipation) than the good they would do. Adding one would almost certainly impose a requirement for more elaborate cooling fins.

And, do mount the flash media with noatime attribute, especially if there is a lot of reads from files already created. It isn't obvious that reading from a file will result in writes to the filesystem metadata, but such is the case. That's how you can tell when a file was last accessed.

Is this matter of power supply related to the Linux-based controller, the satellite nodes, or both? Have you determined how the satellite nodes will be powered? I'm certainly interested in this kind of discussion, and it is a bit divergent from anything Linux-specific, although tuning the OS to accommodate the possibility (or not) of spontaneous power interruptions is certainly on-topic.

--- rod.

The devices at the end of the RS485 line are connected to 24 VAC transformers which in turn are connected to 120VAC. Having the device run at 24 VAC means that they will not need UL approval. Internally to the device is a Bridge rectifier that converts to 33 VDC (peak to peak) and an emitter follower linear regulator down to 12 VDC. A 7805 converts the 12 V to 5 volts for the PIC and other logic. There is a new switching supply (MORNSUN K7805-500 (ma) $3.95 at quantities of 100) that will take the 33 VDC down to 5V without the heat problems of the linear regulators that I will use in the next version of the mist sensor devices. The 24 VAC will run over the cat5 and dielectic grease will be used on the RJ-45 connectors to prevent oxidation problems.

Also IP67 is NEMA 4 in the US.

So, you are distributing power alongside the RS-485 in one cat5 cable? Have you tested the effect of the AC power inductively coupling onto the comm's conductors? I suppose the correct use of twisted pairs should eliminate that, but I'd be remiss to not at least ask the question.

--- rod.

i didnt read all of the posts, but DC doesnt have "peak-to-peak". rectified AC will have peak and a rms voltage. if oxidation is an issue, use gold plated contacts and inject silicone or epoxy in where the cat5 wires enter the cat5 connector. unfortunately, unless the push-down blades of the cat5 connector are also sealed after clamping into the cat5 wire then moisture & air will get to the copper of the cat5 wire, might take some time, but you get the point. if oxidation is an issue then hermetically sealed connectors are a must, etc.

1 members found this post helpful.

Sorry about the peak to peak. The rectifing process is taking the 24VAC and dividing it by .707 to get a peak to peak ouput of 33.9 volts. The output of the bridge rectifier is 33.9 volts especially when filtered by the output capacitor across +- terminals of the bridge. I am going to take the 33.9 and reduce it to 5 volts by the K7805-500 which is a new three terminal switching supply by MORNSUN. The AC is by twisted pair - I have not tried looking at the data lines with this arrangement but what I would have to worry about are the transients produced by the inductive load of the solenoid valve which runs at 24 VAC. I am switching the valve off at zero volts by a triac so the transients should be small. The blades and wire connectors of the RJ-45 end will have the dielectric grease on them to prevent oxidation. The point about using silicone is well taken.

33.9vdc is too high for k7805-500 ?? if so then perhaps clamp the 24vac down to say 20vac using zener method.

just as fyi
24vac/.707 = peak
peak-to-peak = 2*(24vac/.707)

or, since division and subtraction is really multiplication and addition

24vac*1.4144 = peak
peak-to-peak = 2*(24vac*1.414) = ~67.872v


and..... i read through most of the posts. from last year there was question about running this in *nix or windows. the *nix route gives most control and ease of programming, but with the user stations being windoze there is a snafu. so why not bundle the application up into a virtual machine, like *nix with VMplayer. the virtual layer should be able to map all of the devices correctly so your *nix can address them as if they were local, etc. just a thought.

1 members found this post helpful.

To Kidd: The 33vdc was measured. A typical 7805 linear regulator would blow at 33vdc. That is why they came up with the K7805-500. It converts 33vdc to 5 v using a switching regulator. It is a three terminal device (in,out,ground) and the inductor and all parts are in the regulator block. Probably 80% efficient. If I have to lower the voltage to 32 vdc I will use a 5w resistor. I have power to spare. See: http://www.mornsun-power.com/UploadF...8XX-500_EN.pdf

32vdc is max, as per the datasheet, you would be operating the device at a limit, which is not something i would do. i would clamp the ac down to ~20. or, seek a different switching regulator. the zener method to clamp the ac also provides transient protection. if the ac is from a plain 'ol transformer then it has no protection from line swings or transients. you'll see dc swing when the line ac swings or has a transient! zener clamp will provide some protection if the ac decides to wander high, etc.

as example:
http://sound.westhost.com/appnotes/an008-f2.gif

Thanks Linux_Kidd for your input. I will consider what you have presented. Anyone know of a cheap mesh network (<$25). I am considering zigbee and synapse. In place of the rs485.

I signed up for the Embedded Linux Confrence in Febuary 2012 at the Sofitel in San Francisco. Anyone going or has been before? Since this post I changed my flight to one day early to attend the YOCTO project all day build class. Using yocto has given me a better feeling toward an ARM based system.

"mesh network"? looking for a bus to communicate on? what about CAN?
http://en.wikipedia.org/wiki/CAN_bus