[gta02-core] Backup battery underdimensioned

[Werner Almesberger]

Rask Ingemann Lambertsen wrote:
>    I looked at the specifications for the backup battery[1]. The charge
> current is specified as 25 uA in the charge/discharge graph, but the
> pcf50633 can only be configured for 50, 100, 200 or 300 uA (with 200 uA
> being used in the GTA02).

The data sheet says:

| 4-6 Maximum Charging Current (mA)
|       The battery voltage is 3V or more   0.1
|                            Less than 3V   0.5
|                          Less than 2.8V   5.0

So I think we're okay. Not sure about the discharge current. The
battery's data sheet only specifies a "Standard Discharge Current"
(5 uA) but doesn't say what happens if we exceed it.

Also, the PMU's data sheet specifies only a maximum total supply
current, leaving it unclear on which path this goes, and if it
also applies if there's no main battery. Also, this doesn't say
much about the average current.

So I'm not sure we have a problem with the charge/discharge current.

However, the backup battery has been trouble. First of all, it
doesn't solder well. Second, it does seem prone to failure, but
nobody knows exactly why. I'm also not sure if this happens a lot
in the field or not. Third, if it works, it helps to increase the
number of PMU configurations we have to worry about after reset.

There's also the question of what it really does. We don't actually
want the PMU to remember all that much state, and we have to be
prepared to have a functional system in case the backup battery dies
anyway. We do want the time, but we also have at least three other
RTCs in the system (GPS, CPU, GE865), plus we have a number of
potential time sources (GPS, GSM, NTP if we have IP connectivity),
so resetting the clock after a battery change may be an option.

When planning GTA03, we considered just removing the backup battery
and replacing it with a supercap. This idea was eventually rejected
as being too risky.

The maximum backup battery charge voltage is 3.0V (typ), the backup
battery presence detection voltage is 1.3V (typ). If we assume that
the RTC does its worst and continuously draws 50 uA, a 100 uF cap
would only last about 3.5 seconds.

I'd love to get rid of the backup battery. I'm just not sure we can.

- Werner