who where wrote:
On Sun, 10 Jan 2010 11:34:06 +0800, who where wrote:

I'm no stranger to Li-Ion behaviour, having studied them a fair bit
and having designed commercial chargers. I think I know what the
issue is here, but would appreciate informed comment from others.

Laptop pack, "Li-Ion 10v8 @ 4500mAh" Haven't opened it up yet, but
NiXX chemistry wouldn't achieve the capacity in that size anyway.

Charge in laptop (AC on, machine off), indicator LED goes green after
~ 3 hours. After 1/2 hour "rest", boot to Windoze on battery power,
on-screen gauge shows 100% declining to 54% after 2 hours operation.
Resume AC supply, turn machine off, battery recharges to green LED
status (not timed). Rest 1/2 hour, measure pack terminal volts =
12v51. Leave pack out of machine.

Ten hours later, measure 12v48. Alles ist gut. Insert pack into
machine (no AC) and boot. Reach desktop, then machine shuts off.
Restore AC, reboot and check on-screen gauge - "2% and charging".

Clealry the pack has usable capacity as it sustained 2 hours'
operation without drama. Clearly

Clearly??? The initial conclusion from the experimental evidence
suggests the "it did".

the pack did *not* self-discharge in
the ten hours. Right now it is "recharging", meaning going through
the motions to satisfy the electronics - while the pack is already
effectively at full SOC.

What I failed to do is measure the pack voltage *after* the premature
shutdown. Will do that on a later test.

Pack voltage is mostly irrelevant. Bad voltage implies a bad pack.
Good voltage implies nothing.

Follow-up: After another premature shutdown (which was prefaced by a
warning message "you should shut down immediately to save your work"
or similar), the pack was removed and measure 12v52.

Possibilities that I see a

(a) the pack protection module decided prematurely (on some basis)
that the pack was exhausted.

(b) the laptop, based on data (or lack of) from the pack's
electronics, made that call. The pack is "brand X", and the on-screen
SOC panel tags it "standard APM battery" and "Manufacturer: unknown"

Anyone seen this type of behaviour before? Comments? Ideas?

Battery controller IC's are not designed to optimize the life of the
battery. They're designed to limit the liability of the vendor.
They're designed to protect you from suing the vendor.
They're designed to be lowest cost.
They sell more high-margin batteries, so there's little incentive to fix it.

Lithium batteries increase in series resistance as they age.
I've taken failed packs that won't even run the computer long enough to
boot and discharged a significant portion of the capacity from them.
Problem is that the resistance is high and lapotp peak currents are high.
The voltage dips and the machine shuts off. The electrons are in there,
but the system won't let you get them out. You could set the trip
voltage lower, but then you'd overdischarge a good pack. There's
considerable opportunity to have better gas-gauge performance,
but it comes with risk of being sued. Keep it safe, no matter what
it costs the customer.

The gas gauges have varying sophistication, but measure charge in/out
and voltage on a cell basis. Some have memory that remembers the current
capacity of the
cell to more realistically represent the available run time left.
Bad news is that these can get out of sync. Some can get back in
sync, others can't. In many cases, you can replace the cells
and the chip still remembers that the pack is bad. This can be
reset, but nobody will divulge how...to protect you from hurting yourself.

One limp cell can cause the pack to be bad even though the pack voltage
seems to be ok. And a bad cell surely can self-discharge overnight.