TomHiggins ' Question

KeithLofstrom 's Answer

Interesting question. A garden variety Alkaline D cell will produce 10 Amp-hours. These have a voltage of 1.5V, and drop to 1.2V near end-of-life; an energy store of 13.5 Watt-hours. So 6 of those would produce 9V new, 7.2V end-of-life, just enough voltage to keep a WAP going. At an average of 5.5 Watts for the WRT54G, that is about 14 hours of running time for one WAP, 7 hours for 2.

The cheapest way to power a WAP for a LONG time is to use a car battery. While it is probably NOT a good idea to connect the WAP to a car battery while the car is running (alternators and ignition systems make nasty voltage spikes), wuth the engine off it should be OK. A typical car battery stores >12 volts at 50 Amp-Hours, or 600 watt hours. That's 4 days of running time for a WAP.

As long as you can get the hood opened and the battery out of there before the owner comes back, that is less than 2 crimes per week

KeithLofstrom 's Battery AP Observerations

The Linksys BEFW11S4 V4 802.11B WAP/Router, and the WRT54G V2 802.11G WAP/Router, are designed to be powered off a 12 Volt, 1 Amp, center-conductor-positive 5mm diameter power plug.

The power tolerance for these units is quite wide, though. I did some work with a regulated bench power supply, and looked at the circuit inside, and learned that you can power these units off a *regulated* supply anywhere between 4V and 27V, with a safe area probably between 7V and 22V. Among other things, this means I can power one of these WAPs off the AC adapter for my Thinkpad laptop (16V nominal). It also means I can power one off my 130Watt-hour NCharge portable battery for a LONG time ( 20 hours for the WRT54G, 50 hours for the BEFW11S4 ).

Inside the WAPs is a circuit called a "bucking" or step-down regulator, based on the Anachip AC1501-33, capable of 40V operation. The input capacitor is a 25V device (if this was replaced with a 50V device, and there is plenty of room for it, the WAP could take 40V). The regulator is designed to provide a constant 3.3V for internal circuitry, using power from any external voltage in the above-mentioned range. This allows the unit to work with the cheezy 12V/1A wall warts provided by Linksys (which can ripple between 9V and 16V under load). In general, wall warts are pretty sloppy, and most of the numbers below assume something better controlled.

The BEFW11S4 draws about 2.4 watts idling (with ESSID broadcast) and 2.6 watts transmitting (an FTP download), while the WRT54G draws 5.4 watts idling and 5.6 watts transmitting. The current drain appears "bursty", matching the increased power into the transmitter during packet transmission, as you would expect. The buck regulator inside draws enough current at any given voltage level to supply that power, so an idling BEFW11S4 draws around 400mA at 6V and 100mA at 24V. This is called a "negative differential resistance" in electronics-speak, and can cause instabilities with the wrong sort of power source.

One typical "wrong power source" that people here might see would be a Power-Over-Ethernet kludge with too much resistance in the wire. For example, if you power the WRT54G unit from a 24V source with 25 ohms of series resistance, you can theoretically provide the needed 5.6W as 12V. However, during startup the input voltage has to move from 0V up to that 12V, and while passing through 3.3V, for example, the unit might be attempting to draw 1.7A, which is more than the 25 ohm resistance can provide. This is probably why some folks have had better luck with a regulated power source - those have lower internal resistance added onto the resistance of the interconnect wire.

My guess is that the power source will need to temporarily provide 2 amps at 3.3V to the WRT54G, and 1 amp at 3.3V to the BEFW11S4, during startup. So with a 24V source, that is a maximum resistance (including connectors and eventual corrosion) of 10 ohms for the WRT54G, and 20 ohms for the BEFW11S4 . With a 12V source, the maximum resistances are 4 ohms and 8 ohms. An RJ45 with "green scum" on it, copper corrosion pretty common in Oregon, will have higher resistance. The Linksys 12V/1A wall warts have an internal resistance of 2 to 5 ohms that adds to that external resistance. Yikes!

Well, enough calculation nonsense;

What does this all mean?

First, it means that for POE situations you may be able to use a >15V laptop AC adapter as a power source. Even the older ones have significantly lower internal resistance than Linksys wall warts. All the IBM Thinkpads since the 560 have used the same power plug, compatable with the Linksys units, and all their adapters provide more than enough power to drive a Linksys. Prepackaged and ready to go, and there are plenty on Ebay.

Second, it means that you are compatable with the supplemental external notebook batteries. My 130 Watt-hour, 3 pound NCharge battery from Valence Technologies (, $300 from Fry's with one adapter) can be connected to different adapters for different laptop plug/voltage combinations. So you could get one $25 adapter that works with your laptop, and if that is not suitable for the Linksys, you could get a second adapter that is Thinkpad and Linksys compatable, and interchange adapters as needed.

Third, this makes projects like Sam Churchill's Wifi bike project pretty darned easy. Sam could use 2 WRT54G's running the custom EWRT code, one device configured in client mode and connected to a high-gain antenna for the uplink, the other configured as a Personal Telco portal. A Y splitter could power both units off the same NCharge battery, and multiple batteries can be connected in tandem if Sam needed more than 10 hours.

Sam, if you want to follow this up, you might try contacting Valence for a donation of a battery or two; they will be pleased to know about this new application of their product. They might appreciate it even more if you include some JPEGs of photogenic 20-something females posing with your wifi bike.

Enough for now. Forgive the long-winded writeup, but the possibilities that all you imaginative folks may come up with based on this info has me really jazzed ...


Great info. I have a wrt54g, and access to solar panels(15W and a 5W) and a few deepcycle batteries. I'm also going to pickup a microcontroller soon. I plan on doing some testing the summer running the AP 12-16 hours a day using the micro to shut down the AP and monitor power usage. The 15W panel should be enough to run the 5.5W WRT54G for 12-16 hours a day(about the same about of daylight here in the summer). I plan to have it go online around 1pm in the day and run till 12 or so at night. That will give it 6 hours or so in the morning to charge the batteries before the AP goes online.

I haven't worked with solar panels much before. Can the panels charge the batteries while a load is connected? What dose the circuit look like?(how can you model the the panel with RLC and source components) -- Chris McDonald (Electronics Engineering Technologist Student)

ptrck's answer

I would just hook the solar panel(s) up in parallel to the battery and power your AP off that. That way, any unused power goes into charging the battery, which powers the AP overnight. There are probably better ways to do this, and you may end up killing your battery after a few years (because most lead-acids sulphate when they sit at a float voltage for a while), but it should work, assuming the panel can charge the battery enough each day to last the night. Then you dont need to worry about a PIC doing power managment.

I'm interested in setting up a mesh network of weather stations over a farm, and powering each node (WRT54Gs with sensors attached) off a battery and a solar panel, so let us know how you go. -- ptrck

Solar / Battery / Alternator setup

I'm interested in putting a WRT54G into my motorcycle, powered primarily from the bike's electrical system. To keep from draining the bike's battery, I would mount a (6W) solar panel. Do you think I should put in a few diodes to keep from reversing the polarity on the solar panel? Would the WRT54G handle starting surges or should it be protected? I realize that any advice comes with a standard disclaimer. Thanks! -- cton


We are using many battery powered Linksys routers (WRT54G). It would be nice to know when the mains power is down. In other words, when the battery is being discharged rather then being charged. This will give us some days to reach that AP before the battery is depleted and service is stopped. Some of these APs are located on very hard to reach mountains (incl. one at 3300meters above see level). What would be a good way to interface with the WRT54G, so that we can read from remote the status of battery. Maybe flip some GPIO line from 1 to 0, when voltage drops below 11.5v or something like that ? Also - how can we check that GPIO state with software ?



Smart Battery Monitor

Hi everyone. We implemented our ad hoc solution for this in havana, for our ants experiments:

The big picture is as follow: you have different power sources, each with different availability in terms of both time and power. Say the grid (unstable one), solar, wind, you name it. Then you need to charge a battery in a controlled fashion (voltage, temperature, power, current, etc) depending on whether you use Li-Ion, lead-acid, or other type of battery.

For example for lead-acid, we control charge voltage and temperature. Idealy a constant voltage switch regulator will provide best charge effort.

Yet one'd like to make the network power aware, so we came up with Dallas Maxim DS2438 (a 1-wire smart battery monitor: and a battery switch MAX1614 ( to protect the battery from deep discharges and provide greacefull power down.

We control power via the DS2438 from the SNAP ( but it looks very appealing to taylor this for the WRTs. Since there are already some 1-Wire API funtionalities available over openWRT (check and

You can have a flavor of our past experiments at We'd be happy to help shape this out. I'd put all the info on my page for you guys to hack your own.

Still some glue coding would be necesary to get a power aware wrt for a mesh network node. Cool, isn't it?

Any volunteer/feedback?




BatteryPoweredAP (last edited 2007-11-23 18:02:40 by localhost)