I used Uberman with 24 minute sleep periods and a 2 minute fall asleep window. I'd sleep 6 periods in the day every 4 hours on the hour: 2 am, 6 am, 10 am, 2 pm, 6 pm, 10 pm. Staying awake was not an issue past the initial acclimation period. I'd feel just as awake at 4AM as I do during the middle of the day; though I've always been a night owl. My biggest struggles always came from trying to 'cheat' (skip a period, or push it back) and getting off schedule by being too tired and thus oversleeping the next period.
I think what allowed me to succeed was that I partitioned my day into 6 separate days consisting of 4 hours (240 minutes) instead of 1 day consisting of 24 hours.
1-26: Sleep
27-60: Do one of 3 groomings (brush teeth, shower, or shave/nails/facial) and eat 1/2 a meal
61-180: Perform my productive work for the 'day'. Eg working out, work around the house, go out shopping, programming, etc.
181-240: Relax. Usually in the form of watching something, socializing, or reading. (I quickly learned Videogames would leave me too restless to sleep immediately after and would push back my sleep schedule, so they would have to go into the productive period if I wanted to partake.)
My girlfriend assisted me in waking up for those hours she was awake for, though she didn't attempt the schedule herself. I imagine it would have been much easier if she had. Waking up at 2 and 6 AM was hard as hell; for those I wouldn't be able to sleep in my own bed (since any alarm would wake her) so I used college alarm clock and slept near my computer. (NB: Haven't checked that download, it's the second google hit and looks like the one I used though.)
I was doing it back before blogging was popular, although Mr Pavlina's blog is a very, very similar description (though he adapted much more quickly than I did). Unlike him I do still fancy going back to the schedule once I'm self employed and can convince my current girlfriend that it's not unnatural. However, I've tried multiple methods (like Everyman) to make it work, but my reasons for quitting and never stably returning echo his exactly, "The rest of the world simply isn’t polyphasic."
(not yet studied in mammals)
The ratio of the strength of a synapse between neurons and the total potential (from all incoming synapses) needed to activate a neuron may be all that figures; the absolute values may not be important (this is the basis for computer neural networks, though temporal effects, firing rates, and who knows what else also matter in real brains).
So you can renormalize (multiply by some constant 1) and see almost no difference except for perhaps greater susceptibility to noise. But at least the amount of physical material needed is smaller, and the energy needed is smaller. It's more efficient.
In flies and other simple animals studied so far, this definitely happens during sleep. Maybe it happens in humans also. (remains to be studied).
In any case, be careful committing to some REM-sleep only 3hr/day-with-naps sleep schedule, just because you may feel fine at first, when the exact utility of non-REM sleep isn't completely known.
Via.