Sudo Room

Mesh/Power

Revision as of 19:24, 8 August 2016 by Juul (talk | contribs) (→‎Battery)

Low power equipment

This section is about very low power equipment. The rest of this page is more focused on running stuff like TP-Link dual band routers and Ubiquiti AirMax gear on solar which is a different scale of cost/difficulty.

Low power nodes

The hardware experiments page lists some nodes that can be used with low power.

We can probably assume power usage of about 1 watt for most of the lower power stuff but probably not a lot less.

Solar

A small 4 watt panel that outputs 5v directly is available for $14 on ebay. It comes with a female USB A plug but the plug is of visible crappy quality and it failed after a few uses, however it's trivial to just solder on e.g. a male micro-usb end instead. We haven't yet tested what the actual power output is for these panels over a day of california sun but will do so soon.

Battery

If we have the solar hooked up straight to the node then any passing cloud might take it offline. Better to have a small battery acting as a UPS. Since we can use a 5v solar panel why not use a cheap Lithium Ion portable USB battery packs. We've been lead to believe that it will wear out Li-Ion batteries faster if they're kept at 100% charge all the time. We're just going to try anyway and see how long the last.

Some USB battery packs don't allow charging and discharging at the same time. We ordered a selection of the cheaper units and tested them. Here are the results:

Assuming that these batteries actually have a capacity of 2600 mAh at 5v that's 13 Wh which is enough to power a 1 watt router like the NEXX WT2030F for 13 hours. That means that 2600 mAh is not enough to carry such a device through the night during winter in Oakland since the longest night is just under 14.5 hours long. It's probably a good idea to double the capacity since the capacity will drop over the lifetime of the battery.

Then the question remains if a 4 watt panel is enough to charge 14.5 hours worth of power in the ~9.5 hours of "sun" on the shortest day. We need 14.5 Wh generated in 9.5 hours. So we need a bit more than 1.5 watts of charge. Let's say 2 watts given conversion losses (how big are losses?). So if we have a solar panel that's theoretically generating 4 watts in full noon sun that's definitely not going to cut it on a cloudy winter day. We need some numbers on the efficiency degradation of standard solar panels on a cloudy day.

Power usage

Ubiquiti routers all use 8 watts (actually maybe only 6.5 watts for current gen gear, and we should test that). That comes to 5856 watt-hours per month. As of May 2013, the average price for electricity in the bay area was 22.8 cents per kilowatt-hours. $0.228 * 5.856 kilowatt-hours ~= $1.335. So less than $1.5 per month.

Voltage

The Ubiquiti gear takes between 10.5 volts and 25 volts. I believe most of their official numbers state something like 12 to 24 volts, but one of their engineers posted a spreadsheet on their forum which listed the actual limits.

TP-Link N750 routers use a MP1482 voltage regulator which takes between 4.75v and 18v input.

Western Digital N600 routers use a RT8293B voltage regulator which takes 4.5v to 23v input.

Solar

The shortest day of the year in Oakland is about December 2013 and has about 572 minutes of sun. December also has 14 days of average overcast days (coming in second after January that has 15).

It's difficult to find info on the performance of solar cells on overcast days, which is likely because there is no such thing as a standard overcast day. Let's assume that the performance of a solar panel in Oakland is falls to 20% during overcast daylight hours. On the shortest day of the year, that means we have an average effectiveness over a 24-hour period of: 

 (572 minutes / (24 * 60) minutes) * 0.2 = 0.0794

So approximately 8% efficiency in the worst-case scenario. If the overcast efficiency is 10% then that's a 4% efficiency.

Given that the Ubiquiti gear needs 8 watts, we'd then need: 

 8 watts / 0.08 = 100 watts

Or 200 watts if the efficiency is 4%.

A 50 watt panel can be bought for a bit under $100. A 100 watt panel can be bought for about $150.

On top of that some electronics are needed for battery charge control. I'm not sure what the price for that is, but probably around $20. And then there's the battery.


Battery

Shortest day of the year in Oakland has (24 * 60) - 572 = 868 minutes without sun. At 8 watts that's: 

 8 watts * (868 * 60) seconds = 416640 joules = 115.73 watt-hours

You can get a small 10 ampere-hour 12 volt led-acid scooter battery for about $23. That's 

 10 ampere-hours * 12 volts = 120 watt-hours.

That should be just enough, but since the battery will degrade over time, it might be prudent to go with something with a bit more capacity. I also don't know if these types of batteries will quickly degrade if they're routinely discharged to near depletion (I think they will).

Total cost

A very approximate answer to "how much will it cost to make a node off-grid" is:

  • About $200 + the price of the router.
So maybe $300 to $350 including the router.

Car battery cable

Rationale: We have had this idea that the mesh could be more disaster resilient than other connectivity but in reality installing solar and battery systems is expensive and simply isn't going to happen for most nodes. The ony power source that most people will have immediately available following a disaster is their car and car battery. It would be cheap and simple to make kits containing a long cable with cigarette plug and battery clip connectors on one end and a PoE on the other. This would allow people to power on their nodes after a power outage. It would also make it simple to hook up directly to a 12v solar panel.

Marc ordered the following to make 10x test units:

  • The schottky diode protects from inverted polarity (in case someone attaches the battery clips the wrong way around).
  • The capacitor protects from voltage spikes (in case someone attaches battery clips while node is plugged in)
  • The zener diode protects againts over-voltage (in case someone attaches it to a voltage source higher than 22vdc)

Why not cut off at 25 volts with the zener diode so people can use 24 volt batteries as well? Because 24 volt batteries are often actually outputting voltages above 26 volts (especially when being charged) and Ubiquiti gear dies at above 26 volts. Better to have a higher margin of error and just be explicit that it doesn't support 24 volts.

Total cost for enough to make 10 units (with some extra components): ~$75 (that's without fuses, bags, cost of CAT5 and instructions)

If we sell the kits for $10 we're more than breaking even.

We should put the electronics in some common tin/plastic container (dollar store or candy container (Alex's idea)). Maybe we just hot glue that box under the passive PoE.

Apart from the PoE there should be two additional output wires with 5.5 x 2.5 mm male barrel plugs and we should include an additional passive PoE for people who have two extender nodes. This way there will be enough plugs to power both home and extender nodes.

We should include a few extra fuses in the kit and a guide on how to replace them.

Cable

We probably want less than 1 volt voltage drop over the cable since there will be additional voltage drop over the ethernet cable. We probably want at least 50 feet of cable so people can realistically reach their car from their home (or maybe we just offer different lengths).

This calculator says that 50 feet with 16 awg cable at 12 volts and 1.5 amps will have a 0.6v drop. That's probably acceptable, that means we can support up to 82 feet of cable while staying under a 1 volt drop: http://www.calculator.net/voltage-drop-calculator.html?material=copper&wiresize=13.17&voltage=12&phase=dc&noofconductor=1&distance=50&distanceunit=feet&amperes=1.5&x=59&y=11

You can get 1000 feet of 16 awg cable for about $50. That means a 50 foot cable costs $2.5

Compare this to using all pairs in 26 awg CAT5 cable which gives a drop of 1.53 volts over 50 feet: http://www.calculator.net/voltage-drop-calculator.html?material=copper&wiresize=133.9&voltage=12&phase=dc&noofconductor=4&distance=50&distanceunit=feet&amperes=1.5&x=90&y=10

This calculator says that the drop over the nice 24 awg outdoor cable (using two pairs for PoE) with a 12v input over 50 feet and assuming about 7 watts of draw for a nanobridge/nanostation results in a 0.77 volt drop: http://www.calculator.net/voltage-drop-calculator.html?material=copper&wiresize=84.22&voltage=12&phase=dc&noofconductor=2&distance=50&distanceunit=feet&amperes=1.5&x=0&y=0

So total drop from car to extender node assuming 50 feet from car to home node and 50 feet from home node to extender node will be about 1.4 volts. If the extender nodes stop working at 10.5 volts that means that 11.9 volts will be the minimum allowed output from the car battery. Many car batteries will probably drop below that point at some point. Some say that Ubiquiti gear will work down to 9-10 volts though.

Low voltage cutoff

  • We know that completely discharging a car battery is bad for the battery (also then your car won't start)
  • What happens when home and extender nodes drop to below their minimum voltage. Do they still draw power?
  • What is the minimum voltage for home nodes?

Either way we shouldn't discharge car batteries below 11 volts. How do we prevent that? What is the simples/cheapest low voltage cut-off circuit?

Twisted_Haywire has some ideas for super simple circuits. Another option is to use a voltage comparator + divider + mosfet. You can get cheap comparators for $1 or less on ebay/aliexpress.

PCB

If we end up wanting low voltage cut-off then if makes sense to put it on a PCB. Dirty PCBs charge $14 for ten 5x5 cm PCBs, so $1.4 extra per unit, and it drops to $75 for 100 units.

Higher voltage extended version

This version allows voltages up to 36 VDC as input. It has the normal circuit and in addition it has a double throw switch that disengages the 22V zener diode and instead engages a 36v zener and a LM2596S buck converter board (less than $1 from china, but we should add a heat sink). The reason why the switch is important is that the buck converter needs 1.25 volts more input than the output, so if the output is set to 12v then the input would have to be 13.25 minimum. If powering from at 12v battery then any voltage drop is bad, since ubiquiti absolute minimum is 10.5 volts and long ethernet cable runs introduce an additional voltage drop. We might want to use the same kinds of switches that switch between 110v and 220v input on the back of computer PSUs to avoid accidentally switching (which could fry the fuse).

  • Ordered 36V 5W zeners
  • Already have some LM2596S buck converters (we should add tiny heat-sinks to them)
  • Ordered a few double-throw switches (but the "easy to accidentally switch" kind)

Hooking up to solar panels

A home node + extender node probably draws about 12 watts total (need to test this) so a 20 watt panel might be enough but a 50 watt panel would be better.

  • 20 watt panels are about $50
  • 50 watt panels are about $75
  • 20 watt panels with a charge controller are about $70
  • A 12v 18Ah deep cycle battery is about $35

Total cost is probably between $100 for something minimal to use in an emergency and probably closer to $300 (or maybe more) for something that is reliable enough to be in operation year round. Not sure what local laws are regarding having something like this permanently mounted on/at your house.

Of course actual minimal setup is just the 20 watt solar panel and no battery and that might actually work for several hours a day.

Retrieved from "https://sudoroom.org/mediawiki/index.php?title=Mesh/Power&oldid=10285"