#8442 NORM Not Tri: Backlight dim/off timers are too conservative

[Zarro Boogs per Child]

#8442: Backlight dim/off timers are too conservative
---------------------------------+------------------------------------------
 Reporter:  gnu                  |         Owner:  cjb          
     Type:  defect               |        Status:  new          
 Priority:  normal               |     Milestone:  Not Triaged  
Component:  power manager (OHM)  |       Version:  not specified
 Keywords:  power                |   Next_action:  design       
 Verified:  0                    |     Blockedby:               
 Blocking:                       |  
---------------------------------+------------------------------------------
 Powering off the screen saves at least a quarter of the power consumed by
 an average running XO (1.5W out of 6W).  We currently wait *twenty
 minutes* before we do this, though we dim the backlight after a single
 minute.  Dimming saves some of that power, but dimming all the way to
 monochrome only saves 1W.  Turning the screen off after dimming saves
 another 0.5W or more.

 I think that dimming after a minute is fine (though it should be
 adjustable with the xset command and/or the Sugar control panel).  I think
 that after we dim the screen, perhaps one or two minutes after, we should
 turn off the screen.  If a human is looking at it when it dims, they will
 likely press a shift key or stroke the touchpad to un-dim it.  If no human
 is looking, we shouldn't wait 19 more minutes to double the power savings.

 (Turning the screen off would produce even greater power savings once we
 also power-down the DCON; see #2765.)

 As I proposed in http://dev.laptop.org/ticket/8006#comment:34, I think
 this should be done by standard X DPMS support, which would allow querying
 and changing the timeouts with the standard X "xset dpms" command.  The
 "standby" setting would dim the backlight, and the "suspend" or "off"
 setting would turn the screen off.  This is a slight perversion of DPMS
 (normally, "standby" turns off the display, but comes back to life
 slightly faster than "suspend").

 Note:  currently in 8.2-759, dimming is a bit bizarre.  With no power
 management enabled, we appear to never dim the screen, we merely blank it
 after 20 minutes using X facilities.  With "advanced" power management
 enabled, we dim it with Ohm after 1 minute, then suspend the CPU if the
 CPU is idle.  If we do suspend the CPU, we then don't blank the screen
 after 20 minutes (because we're suspended, and suspend doesn't yet wake up
 based on pending timer queue entries).  You can observe both the 1-minute
 dimming and the 20-minute blanking by running something that burns CPU
 time ("while true; do true; done").

-- 
Ticket URL: <http://dev.laptop.org/ticket/8442>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system