[Community-news] OLPC News (2008-02-09)

[Richard A. Smith]

Hal Murray wrote:

> How long does it take to turn the radio on and off?  (I'm assuming the EC can 
> do it without help from the main CPU.)

It can, but only when the CPU is off.  Turning off the WLAN turns of 
3.3V and will crash the machine if things are not in suspend.  The EC 
can also put the WLAN into reset where it would be pretty close to off. 
  After you turn off (or reset) the WLAN the host has to down load the 
firmware to it again.

> Is there any communication path from the radio to the EC?  (to tell it when 
> to turn off and how long to wait)

There are 2 IO lines to indicate battery level.  (I suppose you could do 
clock and data and send it a serial stream) and an IO to tell the WLAN 
that the host has woken up.

> If the radio gets turned off, is there any place for it to save state?  Or 
> will it have to start from scratch each time it gets turned on?

I don't think we have any state storage for the WLAN.

> Maybe the CPU in the radio shouldn't turn off, but just switch to a (very) 
> low power mode.

This seems like the best option.

Before this turns into a big long discussion on things to do in the WLAN 
idle case I'll point out that the backlight and dcon powerdowns and 
perhaps tweaking some knobs that reduce the WLAN power are pretty easy 
low hanging fruit that can extend the upper bound marker.  But its not 
the primary use case.

Getting into the more complex plans to reduce the WLAN power is probably 
a bit premature.  by _FAR_ the biggest issue is that our fast 
suspend/resume is not near fast enough.  Right now the ohmd timeout is 
set to 30 seconds because thats a fairly reasonable compromise.  But 
thats 30 seconds of about 3.5W of power that in a lot of cases was 
unnecessary. The Ohmd timeout needs to head toward zero in all but 
special cases. Wake up, do the work that needs to be done, then go back 
to into suspend.

Getting our suspend speed to that magic < 200mS mark we have talked 
about is where I see the focus of the efforts need to be going.  Thats 
the metric that will mean the most for what the average user of the 
laptop will see.

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