It is true that the battery module does not control charging. This is a good thing, because that means any observations we make through it do not change bme's behaviour.

bme's charge control actually seems to run independently of bme's battery meter in a number of ways.
First, when charging it will display "battery full" before charging actually stops or reaches full. Then it will display green forever even if it charges and stops charging.
Secondly, if the battery is near full, like 90-100% in real terms (not the percentages reorted by bme, hal and so on), and you connect charger, it will often go straight to "Battery full" green led. This doesn't mean it's not charging, it does charge, and typically the charge rate is 2 - 3 times higher than when green light is triggered during a charge started from lower initial state of charge.

How "full" the battery is charged depends on a few factors:

Before PR1.3:
CC/CV @ 950mA to 4.2V, Charge termination at 50mA during CV stage.
The charge chip current sensor can not distinguish between system load and current going into battery. (Check leaked schematics and you'll see it).
Additionally there are some PCB losses between charging chip and system+battery amounting to about 0.125 Ohms.

Let's investigate two scenarios here: Idle phone, and not quite idle phone.
An idle N900 has a power consumption of about 6mA.
This decreases the termination current (at the battery) to 44mA, while the charger chip is outputting 50mA @ 4200mV.
By ohms law, The voltage drop across .125 Ohm is 6.25mV. Let's call it 6. The voltage at the battery will then be 4194mV.

So, charge ended at: 4194mV, 44mA.
Manufacturer recommended charge is CC/CV to 4200mA @ 0.7C with C/20 cutoff. Or, 924mA, 66mA.
The end voltage was lower, but so was the termination current, meaning more power was put into the battery.
Very close to the specs recommended by battery manufacturers anyway.

For the highly loaded N900, let's take 350mA system load. This is typical of heavy browsing on edge or wifi, or heavy IM activity on 3g, screen on.

Charge will never terminate, as 350mA > 50mA.
Voltage drop due to losses in path between charger and system: 0,35A * 0,125 Ohm = 43,75mV. Let's call it 44mV.
The battery voltage eventually stabilizes at 4200mV - 44mV, 4156mV, with the charger chip float charging it indefinitely.

The end voltage is lower, but then it has floated at that voltage for much longer than in the previous example.

After PR1,3 (but not with some community kernels) there's a minimum system load of about 60-80mA, which means that the charge never ends.


From a battery longevity point of view, time spent at 3.7-3.9V degrade the battery the least. Time spent over 4.1V accelerate aging. Time spent at 4.15 trickle charging (charging never stops) accelerates aging even more.
Likewise, time spent at low voltages also accelerate aging.

What this mean is that the battery is happiest when it's at 40-80%.

Anyway, after 100 cycles (where a partial cycle from 60-40-60 would be counted as one fifth of a cycle) and spending 8-12 hours a day on the charger, my battery tests out as 1261mAh (was 1300 as new). Seems to hold up much better than N8x0 BP-4L and BP-5L batteries