[Power] [support-gang] Questions About Solar Panels

[Richard A. Smith]

I'm Duplicating this response to power@ so that this is archived for 
future reference.  I don't do this by default in my response because 
lots of people who might reply aren't subscribed to power at .

On 11/02/2012 01:27 AM, Caryl Bigenho wrote:

> Hi Folks,
>
> I have some solar panel questions. Today I started doing some
> testing of the solar panel I bought from Adam at the SF Summit.

Hi Caryl,

Thanks for the testing. I've merged my responses to your emails.

As mentioned there's a slight chance that the panel will not work 100%
on an XO-1.  You can't damage it but it might either not charge at all
or might stop charging very late in the charge cycle.  This is due to
the possibility of the panel producing a voltage higher than the over
voltage threshold of the XO-1.  Its highly likely that it will Just Work
but its not guaranteed to work for XO-1.

> Thanks folks for the info.... now I am wondering... if I were to put
> the solar panel in the shade on a bright, sunny day would it work on
> the XO-1?

Depends.  Under full sun conditions with the panel disconnected the
output voltage of the panel will be at its highest.  If this is high
enough to trip the XO's protection circuit then when you connect it up
to the XO nothing will happen.  In this case you can work around by
covering the panel connecting it to the XO and then exposing it to the
sun.  This way the voltage ramps up much more slowly and the XO will
pull more load as the voltage ramps.  This load should be enough to keep
the voltage below the trip point.

When charging the battery there is a period near the end of the charging
cycle power draw of the battery begins to decrease until it stops.  If
during this cycle you have good strong sun the the laptop may not draw
enough power to keep the voltage below the over voltage trip point. So
charging might just suddenly stop.  The work around in that case would
be to shade part of the panel so the output voltage drops a bit.

> Also, I still would like to know what the high-pitched whining sound
>  (guessing it was around 1800cps) was and whether the XO could be
> damaged? I won't try it until I have assurance it won't damage the
> machine.

I can assure you that the whine will not damage the panel.  The whine
indicates that the switching power supply on the XO is trying very hard
to maintain a stable voltage but there just isn't enough power input to
allow that to happen.  The whine is the result of the XO's power supply
shutting off and on very quickly.  It turns on, tries to supply power
but the input voltage plummets and it turns off, which allows the input
voltage to rise back up, it turns on, rinse, repeat.

In 1.5 and beyond we added additional smarts to the laptop to try and
prevent this case from occurring.

> I have access to an XO, XO-1.5 and XO-1.75 so I wanted to test the
> performance will all 3 models.

If you want to measure performance then you need my measurement tools.
I have a software tool specifically for this purpose.  Its called
olpc-panel-pwr.  You can find it here:

http://dev.laptop.org/~rsmith/pwr_scripts/olpc-panelpwr-log

The script tries to put the laptop into the lowest power mode possible 
and then wakes up periodically to take a reading from the battery on the 
state of things and goes back to sleep.
When you run it the screen will go dark.  Pressing a key will wake the 
system back up briefly and to stop it you use control-C.

I say it will try to put the laptop into the lowest power mode because 
I've not actually tested it with our newer builds and lots of our 
plumbing has changed.  Its possible its broken.

What OS release are you using on your XO laptops? Tell me and I'll make 
sure it works on those releases.

> I started by leaving them unplugged and running all night to run
> the batteries down. That didn't work too well. They shut themselves
> down with power remaining.

Yes. powerd by default will turn off after 4 hours of inactivity.  When 
I deplete batteries I either disable powerd, turn off powerd auto 
shutdown, or stop at open firmware and run watch-battery until the 
laptop powers off.

> This morning, after the fog burned off I put the XO-1.75, which I
> thought had a drained battery, and the panel outside in very bright
> sunlight. It was fully charged within an hour. I suspect that it
> still had some charge left when I started, so I will have to test it
> again.

See above about my measurement tool.  We can use it to figure out how 
much juice was delivered during your charging.

> Then I took the battery out and tried to run it with just the solar
> panel and no battery. I pressed the on button… it started to boot,
> but quit after the second chime. Wouldn't go beyond that.

I discovered that the 1.75 would run on our solar panel without a 
battery by accident.  During testing I swap in many different batteries 
with different charge levels.  On XO-1 and 1.5 the laptop always lost 
power when I swapped. While testing 1.75 I discovered that in many cases 
it would stay running when I swapped out batteries.  Thats really 
impressive and I think still unique to the XO so we decided to make a 
video showing that.

What never occurred to me was that people trying to duplicate that video 
would try to boot the system on only a solar panel.  It should work but 
when you boot you are going to hit near maximal power draw of the 
system. Cpu is cranking away, backlight is on full, storage device 
activity, audio plays during startup, etc.  Others have also reported 
boot fails.

When dealing with power there is always one more level deeper.  All of 
our power numbers are relatively static measurements.  They are averages 
over time.  Dynamic power draw is a whole different game.  I haven't 
measured the dynamic draw during boot but it will certainly have peaks 
that are greater than 5W.  I would not be surprised if I found very 
short pulses that were even close to 10W.  Give that under normal 
operating conditions (after thermal de-rating, full sun) the panel you 
have will around 10W of output I'm not surprised to see it fail to meet 
the booting power requirements.

Since it starts to boot and then dies that suggests that you just don't 
have enough sun to support a boot.

There is one small detail that I've discovered since then.  If the 
output of the panel is 20+ Volts.  Then the 1.75 may fail to turn on at 
all if there isn't a battery.  Under certain conditions the high voltage 
makes our power supply circuit do some bad things on startup.  Fixing it 
was not easy and since it doesn't happen if a battery is installed we 
didn't try too hard.  Solar without a battery is a fun powerful and fun 
video demo but you wouldn't ever want to run that way because at any 
moment your input power can drop below the minimum required to power the 
system and it will power off or crash.

That is not happening to you since it won't power on under those conditions.

>
> 1) Why the XO-1.75 would not even complete the start up using the
> solar panel with the battery out?
>
> 2) What the high pitched whine on the XO could be when I tried to
> charge it and whether it indicates a malfunction that could damage
> the machine?

Hopefully I answered those.

> Back to the XO-1.75... The sun was really bright yesterday when I
> was experimenting, and the panel was tilted at an angle to catch the
> full sun. XO-1.75 still wouldn't do anything. I really wanted to
> replicate Richard's "show" with the idea of doing it at SCaLE 12X in
> February. But, then, it is usually pretty foggy or rainy near LAX in
> Feb when they have SCaLE.

Boot with battery then remove after sugar comes up.  If it dies then you 
just don't have enough input power to run.

An approximate test to see if you have enough power is to boot on a 
fully charged battery and then stop at the openfirmware ok prompt.  Run 
'watch-battery' . The 5th item in the display is the battery current. If 
its negative then you don't have enough power to run and certainly can't 
boot.  If its zero and stays zero then you _might_ have a chance.  Solar 
power input changes second to second so you never really know.

-- 
Richard A. Smith  <richard at laptop.org>
One Laptop per Child