Solar powering a linksys router?
March 12, 2006 1:28 PM   Subscribe

Solar powering a linksys router?

So... the city of Tempe, Arizona is getting city-wide wireless access. Which is exciting -- I should be able to get really great bandwidth -- except that I live a few hundred yards from the nearest wireless access point.

I've got a cantenna that I've been dying to use and a linksys router with DD-WRT installed, so I thought I'd just throw it up on the roof, where I can get line of site... but the wife would rather that I not drill a hole in the roof to set it up.

One thing we've got out here is plenty of sun... is there something simple I can do to solar power this thing?
posted by ph00dz to Computers & Internet (21 answers total)
For as much as I LOVE solar.. its not the way to go.. You'll have to drill something, somewhere at some point anyway.. I would suggestion getting this Linksys POE (Power Over Ethernet) kit.
posted by cowmix at 1:52 PM on March 12, 2006

Don't you want just the cantenna to be on the roof, and the actual router in your house, so that it doesn't have to try to re-broadcast through the roof?

So you don't need a hole for power, you need a hole for antenna wire to get on the roof. Right? Or am I misunderstanding the setup?
posted by misterbrandt at 1:53 PM on March 12, 2006

PoE is probably going to be easier... but make sure you get a PoE-enabled device! Sending voltage down an Ethernet cable, when a device isn't expecting the power, is Bad.

The thing about solar, of course, is that it's only on part of the day. So you need the solar cell and batteries. You need to charge the batteries during the day, without overcharging them, and then discharge them as little as possible at night, because most batteries last longer if you don't discharge them. And you'll need devices to prevent short circuits and/or too much current draw. None of this is rocket science, it's all quite doable, but it's not terribly cheap. And you need to protect all of that gear from the incredibly intense heat in Tempe.

I'd suggest just setting up with PoE first, and then if you're really really determined to solar-power it, rig that up separately, and switch over when you're ready.
posted by Malor at 3:02 PM on March 12, 2006

misterbrandt: it's important, with wireless devices, to have the device as close to the antenna as physically possible. Wireless transmissions are analog and lossy already, and every extra foot of wire between the AP and the antenna loses more signal. So you want your AP about one inch away from the antenna if at all possible. Gain is life in wireless... and that means the shortest possible signal paths. Once you've converted back to digital, it's much easier to move the bits around.

Being digital, Ethernet can go 200m without problems. Sending power that far will reduce the voltage a great deal, but the loss can be calculated, and most PoE injectors will let you decide how much voltage to send. If you're losing, say, 18v over distance X, and your device needs at least 9v, you'd send about 30v with your PoE device.

All of this is WAY easier to figure out than solar/battery power. :)
posted by Malor at 3:11 PM on March 12, 2006

Response by poster: I was thinking that I could have the linksys box act as a bridge. So, the box running as an access point client under DD-WRT would connect to the city network. Then, my local machines would connect to my access point.

Eventually, I'd have to do this anyway, outside of any issues with proximity to the access point. You only get one network login for x amount of money with this service. So, since we have more than one wireless device, a local AP would have to be involved, regardless of my vendor's preferred method of delivering service.

My signal into the house should be just fine from the roof, I'd think. If worse came to worst, I could just add another linksys running dd-wrt to connect that that... it seems easier than drilling into the crawl space... but maybe it's just easier to do this without the whole solar power factor.
posted by ph00dz at 3:21 PM on March 12, 2006

Are you sure your linksys box will bridge in this fashion? I'd be a little surprised if it did -- most wireless devices are meant to bridge between EITEHR a wireless internet service and a wired local network, or a wired internet service (DSL/Cable) and a wireless local network. I'm not sure I've seen one that does wireless-to-wireless but I admit never having looked, since I've never had wireless internet access. If I'm wrong, you're fine, otherwise, you'll need two devices, one to connect to the cities network, and the other to connect your local devices to.

You may not need to drill in order to get a wire in/out. My attic fan has enough clearance on the side to run wiring through, out the vent. You may have something similar. You would want to attach the wires securely to the roof, probably with staples or something.
posted by RustyBrooks at 3:52 PM on March 12, 2006

Also, if you're handy, you can tap power off the attic fan (if it's powered, mine is). You can basically cut the power line to the fan, install an outlet, and then connect from there to the fan. Provided your fan does not draw a lot of amperage (probably it does not) this should be fine.
posted by RustyBrooks at 3:56 PM on March 12, 2006

I've had nothing but trouble with wireless bridging on the Linksys hardware.

You'd probably want to do NAT instead. You'd set up the Linksys in client mode. (which would require DD-WRT or OpenWRT, I don't think the stock Linksys firmware will do this.) And you'd have it do NAT and firewalling for you, just like any other router/firewall.. the only difference being that your upstream is the wireless. (usually it's the other way around.) NAT allows you to multiplex as many devices as you like down to one external IP address, so you can support all the computers in your house on just one account with the city.

As far as where to put the AP... you really have two options. Put the AP right next to the antenna, and run a PoE cable up to power it, or keep the AP in the house, and just run a long antenna wire.

The upside to roof placement is that you'll get a better signal and possibly faster speeds from the Net connection. The downside is that the Linksys unit may not handle the climate in Tempe very well. 120F in direct sun is usually not good for electronics.

If you just run an antenna wire up, you'll get a substantially poorer signal. This could mean it wouldn't work at all, or if it did, it would be slow. (if the city's signal is strong enough, it could be just fine, too... something to keep in mind.) But your consumer-grade hardware could stay in your consumer-grade-climatized house, which would probably be a good idea.

If you have local wireless devices too, you'd set up a second AP on another channel. Remember that you need to be 5 channels away from other signals... so 1,6, and 11 are the only channels that should be used under normal circumstances. And you'd configure your local laptops to connect to your local AP.

Physically, you'd go roof to external AP to switch to internal AP to your wireless clients. (Any wired computers you have would plug into the switch.)

Electrically, you'd go wireless to wired to wireless.

Logically, you'd go 1 external IP address to the NAT box to as many local, private IPs as you need.

Does all that make sense?
posted by Malor at 3:58 PM on March 12, 2006

Response by poster: I just wandered out front, and I think you're right -- going out the attic vent is the perfect way to approach it. I've got a little overhang right there that'll protect the linksys box from the elements and direct sun.

It looks like you're right about the wireless bridging as well, which isn't really a big deal, given how cheap these routers have become.

It'll be interesting to see how well this all works, particularly given that slower speed connections seem to be able to slow the whole network down.
posted by ph00dz at 4:17 PM on March 12, 2006

I can't get all my thoughts straight at once, as you can see...

Some other ways to get either power, or PoE into the house without drilling the roof:

* if you have cable, particularly the satellite kind, there is an entry point somewhere for that. You might be able to go through there.

* If you have a sprinkler system, there is typically someplace near the sprinkler control unit that goes from inside (probably your garage) to outside. I think mine is a little PVC tube. It's near the electric meter outside, you can see some wires come out of the house and go into the ground, these are, I assume, to control, uh, something. I guess there are some electronic valves situated at various places in my yard. Anyway.
posted by RustyBrooks at 4:18 PM on March 12, 2006

Also: if you go into your attic, you'll see that there are screened vents under the eaves, to allow air to flow in (otherwise, the fan in the roof doesn't really work). These could have wires threaded through them also. You'd have to probably remove one of the screens, run the wire, reattach the screen.
posted by RustyBrooks at 4:21 PM on March 12, 2006

Response by poster: Yeah, that's pretty much exactly my plan, Malor... but you may have a point about the potential effects of the summer heat furnace. Throwing it in a thing of tupperware may not work very well... and I certainly don't want to worry about putting a fan inside of it, either.

Well, cool. I guess we'll see how this goes as they roll it out in our city. The more I think about the limitations of wifi signals, th more I wonder if this is going to work in a city like Tempe. I'm 217.05 yards from the AP, which is up in a lamp post in a right down the street from me. Pretty much the optimal conditions, one might think... Still though, they reckon that I'm out of range by their signal finder.
posted by ph00dz at 4:31 PM on March 12, 2006

200 yards with a clear line of sight shouldn't be any big deal, particularly if you have a good directional antenna on your end. It's quite possible to send these signals for miles. It's the line-of-sight part that's really important.

The open firmwares will also let you boost your signal strength... I assume that works in client mode as well as AP mode.

I very strongly suspect this will work just fine.

BTW, how did you measure down to a hundredth of a yard accuracy?
posted by Malor at 4:43 PM on March 12, 2006

Response by poster: They have the AP's listed in some kind of proxity locator program and on google maps. Dunno what it uses for a backend to find distance..., maybe?
posted by ph00dz at 5:32 PM on March 12, 2006

One last thing... be sure to buy WRT54GLs. They are the Linux flavor and will work with the open firmware. The G and GS have both converted to an inferior OS, don't have enough RAM and flash to support Linux anymore, and by all accounts, suck. The GLs are still good.

If you have any interest in NAS (network-attached storage), Asus makes a wireless AP/router that also works with the open firmware, and has two USB ports. So you can attach pretty much anything you can imagine. Putting drives on it and sharing them is pretty common, apparently.

You probably wouldn't want that one for the outside AP, though. :)
posted by Malor at 5:49 PM on March 12, 2006

Malor: PoE is probably going to be easier... but make sure you get a PoE-enabled device! Sending voltage down an Ethernet cable, when a device isn't expecting the power, is Bad.

Do you have a reference for that? I'm not certain, and it is worth checking more thoroughly. Regardless, I expect there are lots of DIY hacks to let you add PoE to any router/hub. All you need is an appropriately sized cap to block the DC voltage, another big electrolytic to buffer the power supply, and maybe a few diodes to protect the input. Probably very easy. just ask for details, if you need them
posted by Chuckles at 6:06 PM on March 12, 2006

Oh stupid me! I'm not sure WRT54GLs are directly PoE enabled. You can probably do some kind of adapter kit, but I don't know if they do it directly. Oops!

Chuckles, I'm sure I heard stories about dead equipment from PoE, but if what skallas is saying about handshaking is true, they're probably just urban legends. I don't see anything in a quick Google search.

A PoE hack wouldn't be just a cap.... these devices are designed to take a broad range of voltages, because voltage drop is quite rapid on the thin Cat5/6 wires. So a direct mod to a piece of equipment would require some kind of regulator circuit, not just a capacitor.

I suppose you could measure your specific wire and precisely tune the injected voltage to suit, but that seems very fragile.
posted by Malor at 7:28 PM on March 12, 2006

The idea that there is handshaking required to enable the power - wow.

As for the voltage drop.. Isn't ethernet specked out to 100m or something (wikipedia agrees)? I bet it wouldn't be a big deal upto about 10m. You would have to test it all out carefully and everything, but you have to do that anyway. So ya, probably not _that_ easy, but...

Regardless, I think skallas is right, putting it on the roof is a little nuts :P

Figure out where the existing penetrations are (for hydro, phone, cable) pull out the caulk, feed a piece of coax through there and up to your antenna, and re-caulk. Sure, there will be losses from your 10-30 feet of antenna wire, but thems the breaks.. (I would say put it in the attic, but I bet it gets crazy hot in there!)

Make sure the coax has the right impedance rating for your antenna/router!
posted by Chuckles at 8:38 PM on March 12, 2006

No, it doesn't work that way. Ethernet is specced to either 100 or 200m, I forget which. (I thought it was 200m up yonder.) But that's just a signal, it's not trying to pull actual power through the cable to do real work. It just needs to sense the bit data. It's thinking in terms of decibels, not volts... the signal levels are very tiny.

When you're trying to do _work_ with the power, the voltage drop gets very significant. I just searched a list I follow that deals with a lot of this. Over a 50m length, one guy, from memory, thinks that the voltage loss will be about 2v, just from the cable. In addition, you have voltage drops from the device does work and draws power from the line, the voltage drops. Lower voltages are affected more, because to do the same amount of work with 12v, you have to draw twice as much current as you do at 24. So the absolute voltage drop is twice as high, and the PERCENTAGE voltage drop is something like a square function.

Consumer devices generally run on either 9v or 12v. If you tried what you're thinking, which is to run the standard power supply over a 10m cable, you'll lose about 0.4 volts just from the wire. Then the device will draw current, dropping the voltage even more. It might work if A) it can tolerate 8.6v instead of 9, and B) if the power supply brick is good enough to supply quite a bit more current than it normally provides. Otherwise, you'd have problems, up to and including burning up your device. (it could pull too much current at the lower voltage and burn itself out... current is what generates heat.)

PoE standard is 48v, so over 100m, the 4v drop is not that big a deal. And 44v allows quite a bit of power to be delivered without too much current, so it won't fluctuate that much from load. Obviously, to use a 44-48v signal on consumer gear, you'll need a voltage regulator.

That's what it would take to do it properly... trying to do it with the existing power supply and just a capacitor would be begging for dead equipment.
posted by Malor at 10:44 PM on March 12, 2006

Oh, that same list talks about dead equipment from PoE. Apparently, 802.3af added in the autonegotiation that skallas is talking about. Earlier devices apparently didn't do it, and it was easy to smoke hardware. There's at least one case of magic smoke on the Soekris-tech mailing list.
posted by Malor at 10:47 PM on March 12, 2006

Of course the length spec is about signal level. But if you were going to develop a power standard, you would at least try to make it compatible with the data specification - that is why I said 100'.

If the voltage is DC and the current draw is zero, the voltage will be the same at both ends of the cable no matter how long it is. The resistance of 22 gauge wire is 15ohms/1000'. Say the router takes 10W at 12V (0.83Amps), and it is 30' away from the power source. You have 0.83 Amps x 15 ohms/kfeet x (30'/1000') x 2 (out and back) = 0.75V. That is pretty close to workable..

Ya, 48V would definitely fry the input transformer.. I guess 5v probably would too actually.. In theory the DC voltage on the transformer must be zero, of course, but in practice there will be some threshold. (Err.. just been thinking about this... It will be governed in part by the DC winding resistance and current carrying capacity of the wire. However, the AC voltage still has to be able to swing the magnetic field, so it must be substantially greater in magnitude than the DC voltage or the core will stay saturated all the time, and you will never see any output.)
posted by Chuckles at 1:50 AM on March 13, 2006

« Older Giant silent clock?   |   font filter - proof movie Newer »
This thread is closed to new comments.