CL1B power distribution

[John Watlington]

On Apr 25, 2009, at 7:29 PM, pgf at laptop.org wrote:

> wad wrote:
>>
>> This is the current power distribution diagram for A-phase CL1B,
>> identifying what we can power, when, and how.
>
> wad --
>
> a few questions -- for some i can guess at the answer, but better
> to ask and be sure:
>
>     - if there are no USB devices inserted, is there an advantage
> 	to powering down USB?  i.e., does it affect anything more
> 	than the devices themselves?
I think most USB ports these days don't actually turn their drivers
on until they detect something inserted.   We can't really power
down the controller due to Via's requirements, but it should be in
the low single mW range unless it is being used.

>     - same question for SD?
In the case of SD, a CMOS implementation shouldn't draw power
unless it is being used.

>     - the keyboard/touchpad are now optionally powered in suspend
> 	and run.  do you have a specific use case in mind?  the
> 	only case i can think of is turning them off if we're
> 	suspended and don't want their wakeups anyway.
For example, there is no need to power the keyboard and touchpad in S3
with the lid closed or in ebook mode.

>     - will we have (approximate) numbers at some point for how
> 	much power any given subsystem takes?  (e.g., for the
> 	above case, how much would powering down the kbd/tpad
> 	save?  this will inform decisions like "how much effort
> 	is it worth?")
I have estimates for the new sections, and we have numbers from
Tinderbox for the portions that aren't changing (keyboard, touchpad,
DCON, display).

>     - i can't believe i'm asking this, but is it feasible to only
> 	power half the ram?  would that help the power budget?
We have several options to test in A-phase with regards to obtaining
lower power from the RAM (lower speed, lower voltage) but are
waiting to see where we stand.

> 	i have no idea how that feature would be put to use.
> 	or, perhaps more manageable:  half the flash?  if half
> 	were unmounted when not in use, for instance.
Difficult to do, as a good flash controller stripes across both devices
to equalize wear and improve performance.   The 4GB machine will
just have a single NAND Flash chip.

>     - comparing with http://wiki.laptop.org/images/1/1c/ 
> Tinderbox_C2.png
> 	(which i'm assuming is "correct" for XO-1), i see the audio
> 	amp could be powered down before.  is that integral to the
> 	"HD Audio Codec" box now?
The amp can still be powered down (that is the "red" part of the
HD audio codec).  The remainder of the codec remains powered
in suspend to implement wake-on-jack insert (although I have a
populate option to turn off the codec power in suspend if it isn't
low enough power --- the specs say it will be.)

>     - also comparing with that page, the RTC battery charger is
> 	always on now, and wasn't before.  from our conversation
> 	on IRC, it sounds like the circuit hasn't changed -- is that
> 	right?  (and that it's less a "charge" circuit than an
> 	"anti-discharge" circuit.)
Bonus points for pointing out an error in the CL1B diagram ---
the RTC battery charger isn't powered outside of suspend and run.

wad