u-boot power-up behaviour (GTA02)

[Werner Almesberger]

[ This is the contiuation of a discussion about how we can make u-boot
  come up such that a) it doesn't fry any upstream ports, b) will
  charge the battery - which may be empty - if it can, and c) do all
  this in way that has a low risk of breaking anything else, including
  our schedule :) ]

Andy Green wrote:
> It will trickle charge at (100mA @ 5V - power needed for device) in
> fact.  Do we know what power is pulled booting the kernel with backlight
> off?  

Now we do :-)
http://people.openmoko.org/werner/lay-low/usb-batt.html

This is with USB disconnected from the host, so we have no signaling.
(The PSU still delivers USB power.) So this would be equivalent to a
host that's crashed or a 100mA-only upstream port.

u-boot has the don't smoke patch and we can see that it works
beautifully, except that one spike around samples 300/1000. Not sure
yet what caused that - might be PMU register initialization order.

What I did was that I brought up the boot menu, switched to the serial
console, hurried over to the PC, and did a

neo backlight off

followed by 

nand read.e 0x32000000 kernel

Reading from NAND doesn't increase power consumption. Nominal battery 
voltage was 3.6V, as usual.

So we draw about 200mA in total, even with the backlight off. I suppose 
we could get this down by also turning off LCM and Glamo.

I'll now add the wait-for-power loop, which will give us an indication 
of the minimum power the device will be content with.

- Werner