Just to be pedantic, I'm using the N800.

Short version: I have switched my wifi router to serve 802.11g only, rather than the b & g I had set before, and it's saving me battery life.

Long version: I noticed that sometimes my battery charge would last days just sitting around on standby, and sometimes it would go flat overnight. I've been passively observing to see if I could find a common set of conditions causing this. Originally I thought some dufus in an apartment near me had a Bluetooth device that he was trying to connect - deliberately, or not.

What tipped it for me (assuming I'm right), was my little Acer laptop (Aspire 3624), running XP Home, giving me the BSOD several times a week all of a sudden. Rather than reinstalling XP right away, I thought I'd give Ubuntu 7.10 a try (I've been using Linux since the only distro was Slackware pre v1, and it was a 50+ floppy download). 7.04 didn't detect all the hardware on my Acer from the live CD, but 7.10 is fantastic, and the stupid little Intel video chipset runs the compiz-fusion window decorator just fine, so I'm in hog heaven. Anyway... I digress.

On Windows the Acer had no trouble connecting to the Router, but under Ubuntu it kept switching to an unlocked router near me. I eventually tracked down the problem, and generally speaking it's the lack of a way to specify preferred networks under Linux. This isn't entirely true, however. If you drill down in the gconf structure for the wifi setup you'll find a directory containing information on how to connect to the various networks your laptop knows about, and all you have to do is delete those you're not interested in, and you'll stay connected to your own router. Essentially, putting a setup in that config folder tells the wifi modules that it's ok to connect to anything it finds there. Removing it means removing that permission.

Why do I go on about this? In the meantime, I discovered that the laptop switching back and forth on the router eats a lot of CPU cycles. Off A/C power that equates to battery charge, so I assumed the same for the N800. As I had the router on factory defaults - auto, auto, auto - the various wifi devices in my home would have the router switching back, and forth between b and g, and different encryption types. Every time one device needed a change, the router fell back to the next thing, and every other device had to reconnect. While all this reconnecting happened fairly seamlessly, it still burned cycles on the various CPUs. On the A/C-only devices it wasn't even noticeable, but on the N800 it meant battery drain. Anyway... It's feeling much better now

I also have noticed huge differences in battery life depending on wifi connection. At work, I have a rather fresh Netgear 802.11g router, which also happens to sit right on my desk. I can have my N800 on all day with quite a bit of usage, and the battery won't go down more than a tiny bit. At home, I have an ancient 802.11b access point (I think it's also a Netgear) with the signal passing through a lot of walls, and I can't seem to get more than 2 or 3 hours of battery life. I've resorted to going to offline mode whenever I don't use my N800, but as soon as I'm online, battery estimate goes down to 2 hours in regular surfing mode.

Until now I blamed it on the distance from the tablet to the access point (and the amount of bricks and concrete between the two). But it seems quite plausible that it might have to do with the protocol, too. Maybe 802.11b has less efficient power saving features than later protocol variants?

Not quite what I thought. I'm blaming it on the constant need for reconnecting with the b version. In other words, you have a bad signal, your N800 tries to roam a bit, can't find anything else to connect to, then comes back, but the roaming burns CPU cycles, and that burns battery power. On the other hand, I don't know for sure that the b version doesn't actually burn more power.