Intresting, but I don't have two abttery packs...

How about hot swapping the sim card? Maybe I'll try this out later... ]:->

There's a lot of usecases where it may suffice to keep WLAN connectivity, which usually should work, esp if you haven't switched 10mW->100mW in connections advanced setup.
For GPRS/3G data it depends where you are (how far from base tower) and what the modem is doing. It may work with modem on, as long as there's no cell handover and you are close to the base station tower so the TX power the modem needs to use is low.
Checking the air network situation by querying modem about some service parameters - which Konttori 'promised' we get access to, eventually - might help in deciding if we may dare to keep modem on. Reducing risk and impact to system (e.g. by sync, stop certain processes, umount), might be another way to cope with occasional case *when* modem actually hardkills the power. If it fails one out of 100, and no filesystem damage - where would be the limit you'd be willing to accept?


The idea of a script managing all the things like making sure there's a charger and device actually switched to charge mode, to handle all the power hog kicking, and also do things like mentioned above for modem - this will for sure increase safety of the whole procedure. But only if the script is correctly working and tested. For now it's much safer and gives better feedback for developing such script, when users try the manual method and report here about success and particularly about problems they had. That's the reason why I hadn't included method-2 straight from beginning.
For the 'in the field change' I promised I'll get something for you. It obviously will need some hackerish DIY hardware (maybe I'll produce a small bunch of those dongles), and it will need a *script*, so see above... ;-D


See above for rationale about modem keep active during battery hotswap. Also I doubt modem will play nice with SIM changed while powered up. For swapping SIM on a deactivated modem - while rest of system stays on - I see no particular reason why it shouldn't work. Except maybe from cellmo drivers etc *main system* software not capable of dealing with that. This will give nice errors then :-P


cheers
jOERG

In The Field Hotswapping:

You need a dongle *) that has two slots for contacting 2 BL-5J batteries, a 2way switch, and a 1000uF 6.3V capacitor. And a USB-A-receptacle with D+/- short plus CA-101, or a charger-cable stub.
[edit] Depending on component variance this might not work with voltage of empty battery, so some sort of booster might be needed to step up voltage of even an empty battery to a reliable V-IN(min) level of >4.0V for BQ24150

Procedure:
Plug in fresh battery to dongle,
Set 2way switch to connect cable/receptacle of dongle and capacitor to the slot with fresh battery.
Plug in dongle to N900.
Do the hotswap procedure as described before, 1st part, to remove weak battery.
Insert weak battery to dongle.
Throw switch to connect capacitor and receptacle/cable with weak battery.
Remove fresh battery from dongle and insert to phone.
Finish hotswap procedure (2nd part)
Unplug dongle


*) [edit] just for a few bits of info to base my decision on: how many of you were interested in such a dongle? I might offer DIY kits with prepared components (I'm NOT going to sell complete devices, with warranty and whatnot)


jOERG

LOL!!!! that's the thing about linux.........people do things just because it can be done. we'll figure out the practical uses of it later....haahah.

According to my tests the CA-146C charge adapter dongle that comes (or came?) with N900, serves nicely for such boosting purposes. Continues charging N900, when I switch electronically stabilized power supply plugged in to it, to 3V. So should be OK with even comlpetely nominally discharged cells, which still should have >3V.

So - in theory - replace the USB-receptacle or cable stub in the DIY instructions by a cable plus barrel type plug fitting one of the CA-146C receptacles - minus=outer,+=inner. And you should be done. Alas in real tests it didn't work for me yet, maybe needs tighter control of maximum current taken by device, as it seems the CA-146C can provide only ~150mA on 3V input power, and also has no big capacitor to buffer spikes. So might work, but needs further R&D.

cheers
jOERG

(if some users might bother to click thanks on this post, I might bother to build a prototype and publish a video)

Red Green already did this Quite the contraption.

Howabout a hardware hack, where small resistor is connected parallel between 3rd and 2nd (?) contact of the battery socket, so bme would be fooled to think the battery is still in place although would be quickly changed to a fresh one?

At least, with Mugen 2400 mAh battery, which has thermistor instead of resistor there, seems to fool BME correctly to think the battery is there all the time although the resistance is changing with the inner temperature. I mean, BME seem not to be too picky about the resistance and some resistor parallel to battery and without a battery could fool it.

What I am (would like) to experiment, is how to make bigger 2640 mAh battery from two original BL-5J batteries. I have extra bigger back cover, which came with Mugen Power battery. Under this bigger back cover I can have two BL-5J batteries currently, but nowadays I have to reboot to use (switch) both of them during the day. (My original Mugen 2400 mAh battery was stolen with my previous N900 in Lisbon last month.)

I wonder, if just connecting the two BL-5J batteries parallel on their [+] and [-] pins would work? The internal resistance of the battery system then would be half of that of the single battery. I wonder would that be a problem or break something?

Couldn't you also use one of the "emergency" (2xAA or 9V type) instead of swapping batteries with a custom dongle?

I don't think the battery system cares that much about internal battery resistance.

Actually it might sound the battery low warning earlier/later than it should..

shadowjk, this was about BSI aiui. Depending on algo used in BME the parallel Rs of 2 pcs BL-5J might actually result in a better BSI reading and more correct capacity-designed value reported. For sure it won't hurt or break anyhing. Paralleling 2 LiIon is yet to evaluate. Not sure if it's ok, for NiMH e.g. it's absolutely forbidden. For LiIon my educated guess would be it should work.

[edit]
oops, of course the charge termination current threshold SHOULD be adjusted to be (twice as high - "nominal system load"), otherwise bq24150 might never detect charge end condition