power consumption depending on screen content

Discussion:

Joerg Reisenweber

2009-10-09 12:23:00 UTC

A completely white screen with ~40% backlight consumes same amount of power as
a completely black screen (-content) with backlight "dim" (=off)

(=> a black screen eats as much power as a backlight set to 40%)

(DUT: FR A7 SHR-U0808)

cheers
jOERG

Timo Juhani Lindfors

2009-10-09 12:56:37 UTC

Permalink

Post by Joerg Reisenweber
A completely white screen with ~40% backlight consumes same amount of power as
a completely black screen (-content) with backlight "dim" (=off)
(=> a black screen eats as much power as a backlight set to 40%)
(DUT: FR A7 SHR-U0808)

I wrote

http://iki.fi/lindi/openmoko/display-current-measurement1.pl

to test this. However, here white screen with 40% brightness consumes

Post by Joerg Reisenweber
dbg running echo 102 > /sys/class/backlight/gta02-bl/brightness
dbg sleeping for 60 seconds for current to stabilize
dbg taking 10 samples with 10-second delays
dbg current now 102750, mean 102750.0000, deviation -1.0000, range 0.0000
dbg current now 99750, mean 101250.0000, deviation 2121.3203, range 3000.0000
dbg current now 99750, mean 100750.0000, deviation 1732.0508, range 3000.0000
dbg current now 91875, mean 98531.2500, deviation 4657.4034, range 10875.0000
dbg current now 91875, mean 97200.0000, deviation 5012.9520, range 10875.0000
dbg current now 91875, mean 96312.5000, deviation 4982.9396, range 10875.0000
dbg current now 90750, mean 95517.8571, deviation 5011.1483, range 12000.0000
dbg current now 90750, mean 94921.8750, deviation 4936.1718, range 12000.0000
dbg current now 101812, mean 95687.4444, deviation 5157.0280, range 12000.0000
dbg current now 101812, mean 96299.9000, deviation 5233.6376, range 12000.0000
dbg running xset dpms force off
dbg sleeping for 60 seconds for current to stabilize
dbg taking 10 samples with 10-second delays
dbg current now 48000, mean 48000.0000, deviation -1.0000, range 0.0000
dbg current now 48000, mean 48000.0000, deviation 0.0000, range 0.0000
dbg current now 52687, mean 49562.3333, deviation 2706.0407, range 4687.0000
dbg current now 52687, mean 50343.5000, deviation 2706.0407, range 4687.0000
dbg current now 52687, mean 50812.2000, deviation 2567.1756, range 4687.0000
dbg current now 62250, mean 52718.5000, deviation 5203.4787, range 14250.0000
dbg current now 62250, mean 54080.1429, deviation 5961.7104, range 14250.0000
dbg current now 62250, mean 55101.3750, deviation 6229.5975, range 14250.0000
dbg current now 54562, mean 55041.4444, deviation 5830.0278, range 14250.0000
dbg current now 54562, mean 54993.5000, deviation 5498.6935, range 14250.0000

Joerg Reisenweber

2009-10-09 21:23:27 UTC

Permalink

Post by Timo Juhani Lindfors

http://iki.fi/lindi/openmoko/display-current-measurement1.pl
to test this. However, here white screen with 40% brightness consumes

[...]

Post by Timo Juhani Lindfors

Post by Joerg Reisenweber
dbg running xset dpms force off
dbg sleeping for 60 seconds for current to stabilize
dbg taking 10 samples with 10-second delays
dbg current now 48000, mean 48000.0000, deviation -1.0000, range

0.0000
[...]

1: I didn't use xset dpms force off, I rather let SHR (or probably
FSO-framework) do it's dim-job, like it always does by default when hooked up
to wallcharger. No idea what is actually done on dim, or if it is
exactly "xset dpms force off"
(btw if dpms force off actually switches off LCM and glamo video clock, I'd be
rather surprised to see any difference from a light or dark video buffer
content)

2: I used current indicator LED on that wallcharger to positively find the
sketch program when dimmed with a virgin white screen uses significantly less
current than when dimmed with a painted black screen (and erasing all the
black brought it back to "power-saving mode", so it's not been any difference
in state of that program - like filled history buffer etc)

3: the probing during suspend was mainly due to the wallwart LED indicator
has its optimum resolution here, whereas it hardly differs for 190mA vs 220mA
e.g
Nevertheless I suspect the power consumption difference will be similar in
normal backlit state, and probably isn't related to dimming at all (except
for the rationale it's rather senseless to keep a black screen during
backlight LED off)

The compare to ~40% backlight power consumption was a rough guess just to
illustrate the magnitude of the effect, based on same LED readout on the
wallwart charger

I can't tell why your results are like that, but obviously your setup differs
somewhat to mine.

cheers
jOERG

Flemming Richter Mikkelsen

2009-10-28 17:39:13 UTC

Permalink

Post by Joerg Reisenweber

Post by Timo Juhani Lindfors

Post by Joerg Reisenweber
A completely white screen with ~40% backlight consumes same amount of

power as

Post by Timo Juhani Lindfors

Post by Joerg Reisenweber
a completely black screen (-content) with backlight "dim" (=off)
(=> a black screen eats as much power as a backlight set to 40%)
(DUT: FR A7 SHR-U0808)

http://iki.fi/lindi/openmoko/display-current-measurement1.pl
to test this. However, here white screen with 40% brightness consumes

[...]

Post by Timo Juhani Lindfors

0.0000
[...]
1: I didn't use xset dpms force off, I rather let SHR (or probably
FSO-framework) do it's dim-job, like it always does by default when hooked up
to wallcharger. No idea what is actually done on dim, or if it is
exactly "xset dpms force off"
(btw if dpms force off actually switches off LCM and glamo video clock, I'd be
rather surprised to see any difference from a light or dark video buffer
content)
2: I used current indicator LED on that wallcharger to positively find the
sketch program when dimmed with a virgin white screen uses significantly less
current than when dimmed with a painted black screen (and erasing all the
black brought it back to "power-saving mode", so it's not been any difference
in state of that program - like filled history buffer etc)
3: the probing during suspend was mainly due to the wallwart LED indicator
has its optimum resolution here, whereas it hardly differs for 190mA vs 220mA
e.g
Nevertheless I suspect the power consumption difference will be similar in
normal backlit state, and probably isn't related to dimming at all (except
for the rationale it's rather senseless to keep a black screen during
backlight LED off)
The compare to ~40% backlight power consumption was a rough guess just to
illustrate the magnitude of the effect, based on same LED readout on the
wallwart charger
I can't tell why your results are like that, but obviously your setup differs
somewhat to mine.

What happens if you do the same test with the screens flex cable
disconnected?

Flemming Richter Mikkelsen

2009-10-28 18:00:45 UTC

Permalink

On 2009-10-28, Flemming Richter Mikkelsen <***@gmail.com> wrote:
--snip--

Post by Flemming Richter Mikkelsen
What happens if you do the same test with the screens flex cable
disconnected?

Never mind... I just read the other thread:)

Joerg Reisenweber

2009-10-28 19:34:10 UTC

Permalink

Post by Flemming Richter Mikkelsen
--snip--

Post by Flemming Richter Mikkelsen
What happens if you do the same test with the screens flex cable
disconnected?

Never mind... I just read the other thread:)

which other thread?

/j

Flemming Richter Mikkelsen

2009-10-28 20:16:19 UTC

Permalink

Post by Joerg Reisenweber

Post by Flemming Richter Mikkelsen
--snip--

Post by Flemming Richter Mikkelsen
What happens if you do the same test with the screens flex cable
disconnected?

Never mind... I just read the other thread:)

which other thread?

The one about backlight. I know it is not the same thing
but it made me think this issue is more complex:)

Anyway, it could be fun to replace the screen with a
resistive load just to check if the power consumptions
still change with the content (black/white). I guess
it might do that.