AAAx3 800mAHr NiMH. Charge via USB PC or Wall adapter. 0.7V diode drop from 5V. Charge overnight (at least 10 hours).
Load current between 50mA to 100mA. Stay on during the day. But the
customer might forget to charge it at night.
So, better to allow capacity for several days.
There you go again. "Several" is NOT A SPEC. "During the day" is NOT A
SPEC.
"50 to 100mA" is NOT A SPEC.
You need to decide on numbers that serve as your design goals.
Make the decision BEFORE you design it, not after.
Does it REALLY, REALLY have to run of a computer USB port?
There is no excuse for a USB coffee cup warmer, but you can buy 'em.
Many battery powered devices with USB ports won't be able to charge your
device from the battery, and still have juice left in the charging
device the next day.
You're stuck powering the laptop from AC. Skip the middle man and plug
in your device's wall wart.
IF a wall wart will do it, your problem gets very much simpler, and
your liability becomes very much less. You can put your design effort
into the device performance instead of trying to squeak by a marginal
power supply.
My definition of several is five.
My day is 12 hours long.
So, you need 6 amp-hours of battery capacity.
And to charge it in 10 hours, you need more than 600mA of current,
depending on, temperature and a zillion other efficiency considerations.
In this case, the term "zillion" represents more issues than you'll
probably consider. And that's an average of 600mA over the whole charge
cycle.
And, because your voltage margins are insane, you won't get nearly full
capacity of the battery without using something like a SEPIC or
transformer-based switcher.
Based on MY assumptions, you can't reliably get there from here.
Your assumptions may be different...write them down.
If i can find one with adjustable drop out down to zero. The battery could drop below 3V.
Not if you expect reliable operation. I'd shoot for higher terminal
voltage and lower effective battery capacity.
You can get regulators with pretty low dropout.
Remember, that parts of your circuit might be able to run off the
battery. The regulator may have to supply only part of the system.
Yes, perhaps we need to add replaceable fuse.
If you were on my engineering staff, I'd be slapping you silly about
now. Quit sticking on more band-aids and analyze the entire system.
Since you have access to only half of the system you're gonna need
a lot of relevant experience. Do some verification experiments.
"Hey Sally, can I borrow your $1000 tablet so I can make sure this
coffee cup warmer doesn't break it?"
A non-replaceable fuse may be a good idea for catastrophic shorts.
It won't help you a bit with a limit-cycle oscillation caused by the
USB current limit and your device's charging circuit.
Listening to your USB disk drive heads slam back and forth against the stops
as the platters spin down is not a fun experience.
A replaceable fuse is an admission that your design sucks.
"Sorry boss, I can't go to the job site today. This piece of crap
device blew a fuse and didn't charge last night. And I'm fresh out
of these microfuses, so we may be down till I get some from Digikey."
And, don't quote me on this, look it up, I think some USB host ports
communicate with the client to determine how much power to allocate.
You may need to do something about that. And be assured that the
next generation of USB chips will do something different.
Bottom line is that people think power supply design is trivial.
It's NOT! They think a USB port is a free 5V power supply. It's NOT!
On a good day, with an experienced power supply designer,
your particular situation has more gotchas than a bucket of snakes.
I can tell you horror stories until your ears bleed.
And you want me to plug it into my tablet computer.
Are we having fun yet?